Science & Technology Committee: Written evidence ... - Parliament UK

0 downloads 191 Views 3MB Size Report
Dec 14, 2011 - UK census tract data are an invaluable source in facilitating understanding of ...... Joshi, H. E. and Ow
Science & Technology Committee: Written evidence The Census and social science This volume contains the written evidence accepted by the Science & Technology Committee for the Census and social science inquiry. No.

Author

No.

Author

00 01 02 03 04

HM Treasury Julie Selwyn Dr Eldin Fahmy Professor Ceri Peach Emeritus Dr Peter King

05 06

C J Morris ESRC International Centre for Life Course Studies in Society & Health John Stillwell and Oliver Duke-Williams

29 30 31 32 33 33a 34 35

The British Library Royal Statistical Society British Academy The Association of Business Schools Royal Geographical Society (with IBG) Supplementary Institute for Jewish Policy Research Suffolk County Council

36 36a 37 38 39 40 41

Office for National Statistics Supplementary The Salvation Army Joseph Rowntree Foundation Professor Les Mayhew Equality Commission for Northern Ireland Local Government Association

07 08 09 10 11 12 13 14 15 16 17 18 19 20

David Owen TNS-BMRB Professor Heather Joshi Mike Hogan Raj Bhopal David Truswell Dr Julie Fish TWRI Policy and Research Dr James Kirkbride The Institute for Fiscal Studies Dr Stephen Patterson Professor Edward Higgs Centre for Longitudinal Study Information and User Support, London School of Hygiene and Tropical Medicine 21 Dr Nicola Shelton 22 British Society for Population Studies 23 Dr Jennifer Mindell 24 NatCen 25 Tees Valley Unlimited 26 ESRC 26a Supplementary 27 CURDS 28 Welsh Language Board As at 18 January 2012

Written evidence submitted by HM Treasury (Census 00)

The Analytical Community There is an analytical community within government to support those involved in strategy, policy and delivery to develop and use an evidence base. Government analysts contribute to all stages of policy development and implementation though helping to: • • • • • • • • • •

identify issues understand and solve problems present options assess impacts (costs, benefits and risks) interpret existing data specify and gather new evidence test and assess ideas translate research into practical options inform decisions on difficult trade-offs inform decisions on whether to stop, continue or change a policy.

As well as using their own knowledge and skills, government analysts draw on appropriate external expertise, from the academic and broader research community both in the UK and overseas. They can advise on the quality of available external evidence and the confidence that should be given to it in decision-making. Analysts are organised differently in different departments, but generally there are five different groups. Members of each discipline have a unique contribution to make as well as shared skills across the disciplines. At a very high level, the particular contribution can be characterised by: •

Economists (members of the Government Economic Service). Maximising welfare from scarce resources. Microeconomics by providing decision metrics for choosing one option or course of action over another; macroeconomics by fostering prosperity, high employment and stability economy-wide.



Operational Researchers (members of the Government Operational Research Service). Helps people find solutions to complex problems through problem structuring and mathematical and statistical modelling to understand real-world systems, policy options and impact



Social Researchers (members of the Government Social Research Service). Understanding the potential and actual social impacts of policy decisions/practice, including understanding public perceptions and the opportunities for behaviour change. Advising across government on research methodologies and ethics.

1



Statisticians (members of the Government Statistical Service). Ongoing measurement and monitoring of specific and general economic and social trends.



Scientists and engineers (members of the Government Science and Engineering Community). Applying science and engineering knowledge to understand problems and develop policy solutions, including the application of knowledge from individual specialisms

Each of the five analytical disciplines works together to provide the best possible evidence. Each have different, skills and backgrounds – further details provided later – and has their own leadership and management structures but all should be able to look at the problem and indicate which team of analysts would best be able to answer a problem, or signpost to the appropriate analyst for each particular question. Each discipline ensures highly skilled and professional staff through rigorous recruitment procedures; adherence to competency frameworks specific to that discipline; continuous learning and development programmes. Many government analysts are also members of relevant professional bodies and learned societies. As a result, government analysts represent a professional and skilled resource that can add expert analysis and advice to inform choices across government and frontline delivery services. Over the last five years or so the five analytical professions in government have been working increasingly closely together. The most robust and thorough evidence base for a problem will most often come from considering it from a number of perspectives/evidence streams. This can be achieved by ensuring that a multi-disciplinary group of analysts consider the problem. Many departments now ensure analysts work within multi-disciplinary teams to facilitate this approach. Developing or implementing strategy and policy is, of course, possible without the input of the analytical community, but there are risks associated with this course of action. These risks include implementing a policy that does not work; that is more costly than expected or need be; that has unexpected consequences or poor public acceptance. Working collaboratively with government analysts can help mitigate these risks. Signed: Government Economic and Social Research team (GESR) Government Operational Research Service (GORS) Government Statistical Service (GSS) Government Office for Science (GO-Science)/Government Science and Engineering (GSE)

HM Treasury January 2012

2

Written evidence submitted by Julie Selwyn (Census 01)

1.  I am the Director of the Hadley Centre for Adoption and Foster Care Studies,  School for Policy Studies, University of Bristol. I have been a researcher for  over 20 years in the field of alternate care for children unable to live with their  birth parents   2.  My research team has used the Census data to calculate the number of  children growing up in the care of relatives (kinship care).  The data has also  been used to estimate the growth in this form of care and the characteristics of  those carers and children.  3. Our analysis of the 2001 Population Census found that the number of   children in such arrangements had more than doubled since the 1991. In 2001,  one in every 77 children in the UK was living with a relative and many urban  areas had a much higher prevalence (e.g. London Borough of Newham one in  four children). We also found that children’s carers were often elderly, in poor  health/disabled, and living in poverty. The patterns of kinship care also differed  by UK country.   4.  UNICEF has recently described the growth in kinship care as a ‘crisis’, caused  in the UK by parental drug/alcohol misuse.  We expect the 2011 census data to  show another large jump in the numbers of children being cared for by  relatives.    5.  Without a Census it will no longer be possible to monitor these changing   patterns of family care Although the numbers of children in kinship care are  significant, they become invisible in other types of surveys.  They are a hard to  reach group, especially carers and children from minority ethnic groups.   5. Government policy (e.g.  reduction in child poverty, supporting families)   often assumes  that ‘family’ means parental care.  But Census data is showing  that increasingly neither parent is present.   6. Without population data being available, this group of carers are invisible  and their needs are not considered when changes to policy/welfare reforms  are being considered by government.  For example, a grandparent kinship 

3

carer (age 75years)  is extremely unlikely to be able to find work to lift the  family out of poverty.   7. Without this data, Local Authorities will find it hard to plan services at a local  level.     Julie Selwyn  Reader and Director of the Hadley Centre for Adoption and Foster Care   Studies  School for Policy Studies   University of Bristol    10 November 2011      Research referred to  Nandy et al (2011) Spotlight on kinship Care.  Exec summary available at  www.bristol.ac.uk/hadley  

4

Written evidence submitted by Dr Eldin Fahmy (Census 02) 1. UK census tract data are an invaluable source in facilitating understanding of social distinctions and social divisions at a small area level. They are also essential in ensuring that relevant area-based policy interventions are effectively targeted. Using synthetic estimation approaches my colleagues and I have generated, for the first time, methodologically consistent estimates of poverty and wealth at a small area level over the 1971-2001 period (Dorling et al., 2007; Fahmy et al., 2011). This work was well received and widely quoted within the UK print and broadcast media and was quoted verbatim by the then UK Prime Minister, Gordon Brown within the House of Commons. This research demonstrated the growth in spatial inequalities in wealth over this period and is therefore essential in informing policies intended to redress spatial inequalities in socio-economic outcomes. 2. This work extends earlier analyses conducted by these authors into the distribution of deprivation at a small-area level which have informed the development and targeting of area-based regeneration initiatives (Cemlyn et al., 2002; Gordon & Fahmy, 2004), and the targeting of public health interventions (Fahmy & Gordon, 2003). Using a comparable synthetic estimation approach, the authors have developed small-area estimates of fuel poverty in England for the then Department of Trade and Industry (Gordon et al., 2006; Fahmy et al., 2011), and in Wales for the Welsh Assembly Government (Gordon & Fahmy, 2008). Data of this kind are essential in the effective targeting of area-based initiatives to tackle fuel poverty and was recognised as such by the then Parliamentary Under Secretary of State for Energy, Lord Truscott who referred to this work as ‘indispensable in helping understand which areas of England are worst affected by fuel poverty’. 3. However, the decision to axe the decennial population census beyond 2011 will substantially undermine the capacity of UK social science to analyse and understand social processes at a small-area scale. This will have serious consequences for the capacity of policy makers and service providers to effectively target area-based policies and services at those areas with greatest need. A number of alternatives to the traditional decennial population census have been proposed including a rolling census or short-form census. Both of these may be capable of providing robust socio-economic and demographic data at a spatial scale appropriate to the needs of research users interested in understanding societal change and its implications for the effective delivery of social policies and social welfare (e.g. in welfare, health and well being, education, employment, etc). However, it remains unclear to what extent a rolling census would be able to adequately address and adjust for the consequences of change over time in estimates arising from this methodology. Whilst a short-form census allows for the continued collection of ‘key’ socio-demographic data and the subsequent modelling

5

of additional information based on a sample of more detailed questionnaires, the reduction in data quality associated with this approach is likely to be very substantial since the model fit in many areas is likely to be uncertain and (since the census is not designed for this purpose) quite imprecise. 4. However, the implications for the social science community and for research users of other alternatives to a more limited census (such as the use of sample survey and administrative data sources) are likely to be even more serious. Sample surveys may provide valid and robust socio-economic and demographic data for the purposes of estimation at regional level and local authority levels which can inform national policy and provision. However, by their nature (and even allowing for the combining of data from multiple sources) they are incapable of providing the research community, policy makers and service providers with the kind of detailed spatial disaggregation necessary to inform the delivery of services and policies at a small-area level. The development of techniques for combining data from various survey sources in order to create ‘synthetic’ households, whilst undoubtedly an important innovation, will substantially increase measurement error and undermine the validity of inferences drawn on this basis at a small-area level. 5. Clearly, administrative data have important applications in understanding and responding to social, economic and demographic change and will continue to have an important role in informing policy and service delivery. However, it is also vitally important to acknowledge that such data are collected for very different purposes to those associated with the objectives of scientific enquiry and that they reflect the perspectives, priorities, practices and criteria associated with policy development and policy implementation. As such their scientific validity, the extent to which they genuinely measure what we think they are measuring, is in many cases highly questionable with regard to the objectives of scientific enquiry. 6. In conclusion, no other source provides the research community with robust and precise socio-economic and demographic data which can be disaggregated to a high degree of spatial resolution. Whilst it would be highly desirable to supplement existing decennial census data collection with more extensive mid-term estimates than is currently the case (e.g. based upon combining sample survey data sources), the decision to move away altogether from the ‘gold standard’ provided by the UK census of population will have serious consequences for the UK social science communities capacity to understand social distinctions and social divisions and their effects at a small-area level - and for policy makers and service providers to respond to them through the promotion of appropriate area-based initiatives which promote social equity.

6

Dr Eldin Fahmy School for Policy Studies University of Bristol 10 November 2011

REFERENCES (available on request)

Dorling, D., Rigby, J., Wheeler, B., Ballas, D., Thomas, B., Fahmy, E. & Gordon, D. (2007) The Changing Prospects of Place. Bristol: Policy Press/Joseph Rowntree Foundation. (ISBN 958 1861349958). Fahmy E, Gordon D, Patsios D (2011) Predicting Fuel Poverty at a Small-Area Level in England. Energy Policy, 39(7): 4370-4377 (ISSN 0301 4215). Fahmy E, Gordon D, Dorling D, Rigby J, Wheeler B, (2011) Poverty and Place in Britain, 1968–99 Environment and Planning A, 43(3) 594–617 (ISSN 1472 3409). Gordon, D. & Fahmy, E. (2008) A small area fuel poverty indicator for Wales. Cardiff: Welsh Assembly Government. Gordon, D., Fahmy, E. & Patsios, D. (2007) Updating the Fuel Poverty Indicator for England. Report to the Department for Trade and Industry. See also: http://www.fuelpovertyindicator.org.uk/ Gordon, D. & Fahmy, E. (2004). Poverty and Neighbourhood Renewal in West Cornwall: A 2001 Census Update. Final Report to West Cornwall Local Strategic Partnership. Cemlyn, S., Fahmy, E. & Gordon, D. (2002). Poverty and Neighbourhood Renewal in West Cornwall. Final Report to West Cornwall Local Strategic Partnership.

7

Written evidence submitted by Professor Ceri Peach Emeritus (Census 03) Preface (1) The decision to abandon decennial censuses after the publication of the 2011 census is a cultural, academic and statistical disaster. Interestingly, for your committee, it could have a major and negative impact on Members of parliament as their numbers are reduced and their existing constituencies are dismembered. British censuses have been taken every ten years since 1801 . Hitler in 1941 is the only person to have stopped the regular cycle. The census is much more than a count of heads, though that function remains important in the allocation of funds to local authorities, voters to constituencies, assistance to regions facing particular social or economic problems. Race Ethnicity, Religion, migration, segregation and Ghettoisation (2) I am Emeritus professor of Social Geography at Oxford University. I have a vested interest, for the future of my country, that this amazing resource for our national good survives. My research deals with ethnicity, immigration, segregation and ghettoisation of ethno religious groups (Muslims in particular, but Catholic/Protestant relations in N Ireland as well). The British censuses as they have developed over the years since 1951 have delivered data on these topics which are the envy of other European countries. They are also of major importance in the understanding and governance of our country. (3) France for example, which is supposed to have the largest number of Muslims in Western Europe, has no way of verifying whether than is the case. There is no question on race, ethnicity or religion in the French census and the French suffer from that lack. If one traces back the origin of the estimates of 5m Muslims in France, they turn out to be the fiat of a minister of the interior, (M Pasqua or M Sarkozy most recently) but without any demographic data to support them. Michelle Tribalat, a dedicated researcher at INED the National Demographic Institute searched through a huge sample of Census returns to locate likely Muslim names, the best available surrogate for such a calculation and produced an estimate of 3 million Muslims as opposed to the 5 millions cited by the Ministers. She was summoned by Sarkozy, then minister of the interior, and ordered suppress the number as ‘unacceptable’. In Britain in 2001, however, we could look up the census to see 1.6millions of our own. The French however can only speculate, estimate and guess. As

8

their cities banlieus (suburbs) turn into the American Black Inner city, they have no easy way of charting the change. (4) If, in future members of parliament wish to follow the French example and have critical data invented for them by Ministers you will find yourselves in quicksand. (5) The British 1991 census was the first to introduce a question on ethnicity. It greatly clarified the socioeconomic conditions of the minority ethnic populations in this country. It quantified the educational, occupational, domestic housing conditions of the population. Although discrimination against minorities had long been recognised as a general phenomenon, census data made it possible to quantify the extent to which discrimination was acting. (6) This was done through the work of Anthony Heath, Oxford University’s Professor of Sociology Heath calculated the degree of ‘ethnic penalty’ for each ethnic group. An ethnic penalty is the degree to which members of a minority group with educational qualification equal to those of their white peers fail to acquire occupations of the same level. His technique was to use whites as the reference group, take samples from the census of ethnic and white groups with the same educational qualifications and estimate the proportion of those equally educationally qualified minority and white, who achieved jobs in the salariat (jobs that were well paid, had an incremental salary and pension, for example). He not only found the expected systematic ethnic penalty but some unlikely cases as well. For example, Indians as a whole had a socio economic profile as high as or higher than whites. However, although they performed well, they performed less well than they should have done relative to their qualifications. Despite doing well , it is clear that they were nevertheless suffering an ethnic penalty. At the other end of the scale, using avoidance of unemployment as the criterion, Heath found that Bangladeshis with no qualifications performed better than they might have been expected to do. The Longitudinal Study (7) One of the great advantages of the census has been its development of longitudinal studies. This vital research facility is doomed by the threatened demise of the census itself. To explain, the census is a snapshot gives a picture of the population on a particular date each year. We measure change by examining the differences between the populations on ten year gaps. However, changes are not solely the result of the same people ten years later: people are born; people die people emigrate and immigrate. It is theoretically possibly to have identical 9

numbers of people present at two census dates but for the populations to have totally different ethnic compositions. What we need therefore is a way of tracking what changes take place for a sample of individuals whose individual characteristics can be tracked from census to census. ((8) The Longitudinal study selects a largish sample of individuals born on the same date of the year. These individuals’ census forms are retrieved after each census to establish the changes taking places of a constant panel (panel is of course refreshed a little at each census, to compensate for deaths births and migration.) The longitudinal survey produces a different picture from that of the comparison of two national. Censuses as an example, the occupational profile of Caribbean men in London shows rather little change from 1991 to 2001, but longitudinal study shows upward mobility in jobs. The Longitudinal survey also shows a suburbanisation of the Caribbean population. The Sample of Anonymised Records (SARs) (9) A final example of what will be lost with the demise of the Census is the Sample of Anonymised Records. Sars allow one to have very detailed analysis of family structures and relationships without compromising the identity of individuals in such households. It allows detailed examination of society in which the variables that are sought can be specified to a degree that sample surveys by polling agencies cannot be achieved. (10) The bottom line in this argument is that a continual nation census which has been run since 1801 is a national treasure that cannot be replaced by an analysis of Sainsbury loyalty cards. There is a moral her for the preservation of the census from the Climate change debate. The key data on climate change come from ocean sea temperatures. As it happens, techniques of measurements have been changed over time but without overlaying the techniques by which data were captured. Originally a wooden bucket was dropped over the side of a sailing ship pulled onto deck, a thermometer placed in it and the temperature recorded Later, canvas buckets were used and ships’ sides were higher out of the water and both the height and the bucket material affected the temperatures which recorded. At each advance in ships and buckets, the conditions under which the temperature was measures was different but no controls were taken so that the measures using the old system were not overlapped with the new. It thus became impossible to establish to what extent changes in water temperature were due to warming or cooling of the oceans or the process of measurement. The experience of measuring small but critical temperature differences over a long period of time,

10

should be a warning against dismantling the long continuous data collections of our population censuses. Professor Ceri Peach Emeritus, MA D Phil Professor of Social Geography Oxford University and Visiting Professor Institute for Social Change University of Manchester 11 November 2011

11

Written evidence submitted by Dr Peter King (Census 04) I am an economic historian, though mainly concerned with periods before there was a census. I have also been engaged on behalf of an NGO in commenting on planning documents at a local level. I believe that it is important to continue having censuses, because they provide an objective basis for providing future estimates for planning and government expenditure. It is of course true that there are many other sources of information available, but a great many of these are estimates based on samples. It is in the nature of such an estimate that it is only accurate if the sample is a representative one. Since no sample will ever be perfectly representative, any estimate will almost inevitably be imperfect. If a biased sample is used repeatedly, there is liable to be a drift in the estimate away from reality. It is thus important that the “clock” should be returned to zero periodically, by having an objective survey, such as is provided by having a census. Examples can no doubt be multiplied of where public policy has been poor because it has been based on poor estimates: •

Government funding for Derry/Londonderry was based on the census data, but significant parts of the city were not surveyed by the census, because the IRA imposed a boycott on it. This was no doubt for reasons that were sound in their minds, but were counter-productive to the welfare of their supporters.



Slough BC has repeatedly complained in recent years that it was getting inadequate support from government funding. It alleged that it had suffered disproportionately from migration from eastern Europe, but that estimates made by ONS had failed to pick this up.



In my own parish, the Local Education Authority looked at ONS data on the birth-rate in the parish and suggested from this that the Primary School would suffer from falling rolls. In doing so, they ignored that (as an affluent area with good schools) the parish is always under great pressure from in-migration from the adjacent major urban areas. This often consists of parents who have already started a family while living elsewhere. What actually happened was that the Primary School had to be expanded by making provision for a third stream. Even with that the school is always full, and is liable to have to contract the extent of its catchment.

The following will illustrate the tendency for estimates gradually to drift towards error: •

The pre-census estimate for 2001 proved to have a significant error in it, because ONS (or a predecessor) had underestimated emigration, largely by the professional and managerial class. If there had been no census in 2001 to correct this error, the error would have been compounded to an extent that ultimately the estimates would have drifted so far from reality that they would be useless.



Almost everyone is registered with a GP, but o people registering with a new GP without their medical number are assigned a new one. When their medical records are transferred, the necessary cross-

12

reference is made. This will mean that the number of people appearing to be registered will slightly exceed the number of patients. No doubt that is an error for which a correction could be made. o However, it is likely that the medical records of those emigrating will not be transferred. Unless this error can be accurately be estimated, it is likely that the number of medical records will exceed the population and that there will be a continual drift in this. •

Similarly, National Insurance Records are an unsatisfactory source for the size of the population, because it is notorious that one class of benefit cheats defraud the system by creating multiple identities. This is an unquantifiable problem.

As a historian, I observe colleagues obtaining useful historical data from the census. In the period before enumeration records survive, historical data on occupations for example is only scantily available, from directories, parish registers, and other less systematic sources. The result is that we know much less of that period. If the 2021 census is missed, not only will data not be available for government planning, but in 2121, it will not become available to historians. Dr Peter King 12 November 2011

13

Written evidence submitted by C J Morris (Census 05)

The Committee has sought written submissions on the following matters: How do social scientists use Census data? 1. Social scientists (including those employed by government, either directly or on contract) use Census data for two main reasons. 2. Firstly, the Census allows a quick and definitive statement of the numbers and percentages in some defined group of interest eg the number of car-owning households by socio-economic group in Wedmore, Somerset. 3. Although available only once a decade, this figure is at least available. To compile the data from scratch would be virtually impossible by reference to official statistics – government (local or central) most likely have no estimate of how many households there are in Wedmore, much less their socio-economic status, and while the VLA might perhaps, given time, be able to work out the number of cars registered in Wedmore, they have no idea of the socio-economic status of the owners and wouldn’t tell the researcher if they did. 4. Such a figure, once compiled, would, in any case, be a dead-end whereas the Census would allow further cross-reference to other factors such as households containing someone with limiting longterm sickness. The Census is thus is an invaluable starting point for analysing almost any social issue where a relevant question has been asked in the questionnaire, even if the follow-up involves using some other technique or approach. 5. Secondly, it is possible to subject Census data to more detailed analysis. This may be done at aggregate level or (through anonymised data sets such as those made available by the Census and the various UK longitudinal studies – England & Wales, Scotland and Northern Ireland) at individual level. The relationships found are very robust, being based on large numbers, and they allow the analysis to sidestep the problem that not everyone living a poor area is necessarily poor (or as the Duke of Wellington famously put it “Being born in a stable does not make one a horse”). Areal effects can thus be distinguished from personal effects. 6. Variables of particular interest include age, sex, occupation, education, social status, car ownership, religion, ethnic group, health, much as in the Census. This needs to be analysed at both individual and household level eg a household containing both Catholics and

14

Protestants is of particular social interest in a Northern Ireland context (for GB, read “from different ethnic groups”), In fact, there is much greater demand for variables to be included in the Census than there is space to accommodate them, even discounting income which was of considerable interest but proved unacceptable (and probably less than ideally accurate). Not all of these variables are necessarily of particular relevance to those assembling administrative databases. What impact will the ending of the Census have on social science research? 7. The ending of the Census would more or less require social scientists to depend solely on survey data, their own or others. Although at present, they carry out their own surveys to supplement the Census, asking questions or pinpointing groups that the Census does not, the Census provides a benchmark against which to test the results. It is completely impractical, however, for any researcher to carry out a survey even remotely approaching the scale and coverage of the census. 8. Government administrative databases are rarely available to social researchers outside government (and they are not always made available to those within government, either). Computer systems designed solely to facilitate immediate administrative needs (though they do not always achieve even this); reluctance to capture data not directly and demonstrably relevant to the primary function of the system; and of course, the need (real or perceived) for confidentiality all combine to make these a limited asset for research. 9. This is compounded by the fact that many government contacts with individuals at an address (possibly all save those dealing with benefit claims) disregard most if not all other individuals at that address. This loses the key social information on household structure, which would have to be imputed from some other source, probably a survey, or compiled by attempting record matching (and so far as that goes, is the Ms Smith recorded as living at the same address as Mr Jones, a previous occupant, a previous partner of Mr Jones, a current partner of Mr Jones or indeed now the Ms Jones recorded at that address?). 10. The absence of Census data to set a context results in a narrowing of perspective. This is extremely relevant to integrating social science research with policy making. Whilst qualitative research and small surveys may well demonstrate the existence of situations and relationships worthy of policy intervention, it is the scale of these which likely to be the determining factor in policy-making. Small problems attract less help than large problems.

15

What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? 11. The current proposal to abolish the Census seems based on a rather naive view of the accuracy of data sources. Just because a computer record exists does not mean that it is accurate or indeed accessible. Any data source, the Census included, is subject to errors, and it is unrealistic to imagine that there is some panacea approach to data collection and data-matching that will yield error-free results. Different methods will yield different errors and different levels of error, so that it is simply a matter of deciding which is least unacceptable. 12. There are currently no existing reliable alternatives to the Census, and it is open to question whether an acceptable alternative could be constructed. 13. No survey can provide data of comparable quality. Apart from all the problems of non-response and the like: a) Small surveys (ie up to an achieved sample size of say 25,000 or so) can provide very good data on large groups, but for smaller groups (under 10%) of the population, the effective sample is not large overall, and quickly becomes inadequate when disaggregated by eg region or sex or age. b) Large surveys quickly become very expensive and yield ever decreasing benefits for increasing sample size. It is necessary to quadruple the sample, in order to halve the statistical error of the survey c) Very large surveys (eg the 1966 10% Census of Population in Great Britain) carry all the disadvantages of a survey, and are very much like a Census in the set-up costs. It was for that reason that the Northern Ireland government carried out a full census in 1966, calculating that the cost savings from a 10% census were not worth the reduction in data quality. 14. Some of the bigger existing surveys produce estimates expressed in total numbers eg X million in poverty or Y million unemployed. This should not, however be taken as an alternative to the Census. The survey results have been subjected to a complicated grossing process dependent on midyear population estimates that derive ultimately from the most recent Census. They are not, therefore, independent sources. 15. Theoretically, a population register holding a common identity number and basic demographic information could be cross-matched with one or more other data sources where the identity number was held, to produce an ad hoc database which would fulfil the purposes of a census. This, however, would depend on:

16

a) The establishment and maintenance of a universal population register – the Government has abandoned the identity card project for a variety of good reasons, and it would require a U-turn of policy to produce a register b) The modification of existing databases to capture and hold the common identity number c) The accuracy and timeliness of existing databases – many databases are well out of date, because of the infrequency of interaction with the subject and the failure to update them even when changes are notified. Address changes are a crucial issue, since there is a high level of mobility among the poorest members of the community. d) Exclusion of data which although validly held on the subsidiary database, has no place on the compiled database eg lock-up garages are a valid part of the Northern Ireland Valuation List, but form no part of the habitable dwellings which might be occupied by the NI population. It will be noted that there might be justified resistance to adding a common identity number to property details – the owner and the occupier can be different people. e) The implementation of a massive computerised data-matching capability – current evidence eg the NHS experience of integrated databases, suggests that Government lacks this capability and will continue to do so for the foreseeable future 16. Such an ad hoc database would of course have to be set up every time that a research project was required, involving substantial set-up costs each time. The Census database, once set up, is extremely cheap to analyse in response to ad hoc needs. This suggests that the most likely approach would be a kind of “ersatz Census” where once a decade, a database would be set up by data matching and used in a similar way to the Census thereafter. What other existing sources of population and socio-demographic data could be improved upon? 17. All of the existing sources of population and socio-demographic data could be improved, but not perfected, if government were prepared to spend the additional money required. This does not mean that they could be employed as a substitute for the Census. 18. When the 1981 Census of Population in Northern Ireland was the target of a moderately effective campaign of deliberate nonenumeration, it proved possible to arrive at a robust estimate of the level of non-enumeration, using alternative sources of information (as evidenced by the good match between the mid-year estimate and Census in 1991). The alternative sources would have been useless as a substitute for the Census.

17

19. The alternative sources (largely the school Census and past midyear estimates) were partial in coverage, lacking in the range of information and household data required, not aligned to the Census date and sometimes out of date. These problems would be encountered in any alignment of alternative data sources. 20. The Committee may wish to recommend that Office for National Statistics consult with Northern Ireland Statistics and Research Agency, as Northern Ireland has some past experience in dealing with seriously flawed Census returns.

Declaration of interest 21. I am a social researcher and statistician, working as Ulaidh Research Consultancy since my retirement from the NI civil service, where I worked as a statistician for 31 years. C J MORRIS Ulaidh Research Consulting 17 November 2011

18

Written evidence submitted by ESRC International Centre for Life Course Studies in Society and Health (Census 06) I am responding on behalf of ESRC International Centre for Life Course Studies in Society and Health. Our answers to your questions are: 1. How do social scientists use Census data? In our work, Census data are used in many ways. a) The Census serve as the definitive source for describing the health and social circumstances of the contemporary population, particularly for relatively small population groups. For example, the Census is one of the only data sources that allow us to examine the characteristics of specific ethnic minority groups, occupational groups, or residents of small geographic areas such as electoral wards. Census data, including the Samples of Anonymised Records, can serve a source of data for detailed research in their own right or can be used to substantiate or inform findings arising from sample surveys. b) Census allow us to examine and understand social change. With 100 per cent data we can look at change over time at a population level, for example in access to basic amenities, or household structure. Census data linked to other administrative records in the ONS Longitudinal Study open up unrivalled opportunities for looking at change across individuals' lives, inter-generational change in social circumstances, and for understanding socio-demographic differentials in mortality and fertility. For example, the ONS Longitudinal Study is the key source in the UK for understanding differences in mortality by occupational class and provides a rich source for understanding the fertility patterns by ethnic group. c) Census data provide a reference point for ensuring that sample surveys are representative of their intended population, through the generation of population weights or grossing factors. This is a vital function of the Census for the social science community, as it allows us to understand the extent to which research findings from surveys are generalisable to the population at large. 2. What impact will the ending of the Census have on social science research? Given full population coverage of Census, the long run of available Census data for monitoring social and demographic trends, and the relative detail of the UK census, the end of the UK Census is likely to have detrimental impact on the quality of social science research in the UK. The Census is part of the UK social science infrastructure; it is therefore difficult to assess the full ramifications of ending it. We assume that the Census would be replaced without some other method of enumerating the population. The absence of a firm proposal for this alternative, makes it difficult to understand the relative impact of ending the Census. 3. What alternatives to the Census would provide population and sociodemographic data of equivalent or higher quality?

19

In our opinion, it is unlikely that alternatives to the Census would yield data of equivalent quality for social science research. Countries that monitor population trends using register data, have operated in a societal context where this accepted and supported by the general public. To date, there appears to be little public appetite for a central population register in the UK. Any attempt to impose such a register without consensus runs the risk of yielding poor quality information and relatively high rates of data missingness. 4. What other existing sources of population and socio-demographic data could be improved upon Administrative data collected by Government departments could be improved with more attention to the nature of data collection and quality of completion. For example, Her Majesty Revenue and Customs could improve its recording of the occupation and industry of individuals, although this might place burdens on business. However, we are not aware of single data source, or universal point of interaction with state that would provide the opportunity for collecting high quality data about a range population variables. For example, it is not clear where opportunities would present themselves to equally detailed data on issues such as household structure.

David Blane Professor of Medical Sociology Imperial College London. Deputy director ESRC International Centre for Life Course Studies in Society and Health 17 November 2011

20

Written evidence submitted by John Stillwell and Oliver DukeWilliams, School of Geography, University of Leeds (Census 07)

Declaration of Interest  The  authors  are  both  academic  researchers  funded  by  the  Economic  and  Social  Research  (ESRC)  Census Programme to provide users in the UK academic community (academic staff, research staff  and students) with online access to Census and related interaction data via a Web‐based Interface to  Census  Interaction  Data  (WICID)  maintained  by  the  Centre  for  Interaction  Data  Estimation  and  Research  (CIDER).  Both authors have longstanding research interests in migration and commuting  behaviour,  extensive  experience  in  handling  interaction  data  sets  and  publications  in  a  range  of  international  social  science  journals.    The  submission  is  therefore  focused  on  the  outputs  of  the  Census generally referred to as ‘Interaction Data’.    [1] How do social scientists use census data?  1.

  2.

  3.

  4.

Moving  house  is  an  event  that  most  people  experience  at  some  stage  during  their  lifetimes  whereas  going  to  work  or  to  study  is  part  of  daily  life  for  the  majority  of  the  pre‐retirement  population  living  in  the  UK.  Censuses  provide  extremely  valuable  information  about  volumes  and compositional structures of both these types of human behaviour including details of the  origin and destination locations involved.   Data from the 2001 Census show us that migration and commuting are remarkably important  phenomena  in  contemporary  times,  as  they  have  been  throughout  modern  history.  Over  6.2  million people in the UK changed their place of usual residence in the 12 month period before  29  April 2001, representing around 10% of the total population of the UK. In addition, a further  467,000 immigrants arrived from outside the UK in the same period. In comparison, over 28.3  million  people  were  recorded  by  the  2001  Census  as  commuting  to  work  on  a  daily  basis,  around  two  fifths  of  the  total  population,  although  2.1  million  flows  involved  people  whose  home and workplace locations were the same.  Data  on  migration  and  commuting  are  available  from  various  census  products  including  aggregate  statistics  (various  sets  of  tables),  Special  Migration  Statistics  (SMS),  Special  Workplace/Transport  Statistics  (SWS,  STS),  cross‐sectional  micro  data  sets  and  longitudinal  studies.  The  SMS  and  SWS/STS,  known  collectively  as  the  ‘interaction  data‘,  are  particularly  important  because  they  provide  information  about  directional  flows  between  origin  and  destination  areas.  This  means  that,  depending  on  the  geographical  scale,  the  data  sets  may  involve very large yet sparsely populated matrices.  Both migration and commuting have transformed societies and influenced how settlements and  landscapes  have  developed  over  time.  The  role  of  the  Census  is  providing  the  data  to  understand and analyse these transformations in the UK has been fundamental. In the 1970s, 

21

research attention turned from suburbanisation to new processes of population redistribution  such  as  counterurbanisation,  the  movement  of  individuals  and  families  away  from  major  metropolitan centres and down the urban hierarchy to increasingly rural areas, as documented  by a number geographers.     5.

  6.

  7.

  8.

  9.

Whilst  the  process  of  suburbanisation  usually  involves  breadwinners  whose  place  of  work  before  and  after  migration  remains  the  same,  counterurbanisation  includes  those  migrating  away  from  the  cities,  severing  their  commuting  to  work  ties  and  adopting  a  different  style  of  living.    In  some  cases,  these  individuals  are  those  deciding  to  work  locally  or  from  home;  in  other  cases,  they  involve  those  seeking  unconventional  lifestyles;  the  pioneers  of  the  counterurbanisation  movements  were  actually  those  reaching  retirement  age  who,  on  becoming economically inactive, no longer required to maintain ties with a place of work.   Research based on migration data from the last three censuses indicates how the major losses  from the big conurbations have continued in the twenty‐first century, causing policy makers to  fear  implications  that  urban  neighbourhoods  are  being  abandoned  whilst  rural  areas  are  continuing  to  come  under  increasing  house‐building  pressures.    However,  research  on  the  composition  of  migration  streams  has  revealed  important  differences  in  sub‐groups  of  the  population  with  those  in  the  high  mobility  age  ranges,  the  late  teens  and  twenties,  showing  patterns of movement away from rural areas and into cities that run counter to the net flows of  families and the elderly in the opposite direction.   In addition to analysing the spatial variations in the role internal migration vis a vis the natural  change component of demographic restructuring,  migration research has increasingly focused  on  immigration  flows  and  the  ethnic  minority  communities  that  have  established  since  the  initial waves of newcomers arrived from the colonies in the post‐war period. The addition of an  ethnic  question  to  the  Census  in  1991  has  allowed  studies  of  ethnic  populations  in  the  UK  to  flourish and the last decade has seen the emergence of a body of studies on ethnic migration –  suggesting that, contrary to popular opinion that ethnic minorities in British cities are becoming  more  segregated,  there  is  considerable  evidence  to  indicate  processes  of  spatial  deconcentration are taking place, particularly in London, where the majority of ethnic minority  populations are located.    Recent studies of commuting behaviour have demonstrated variations in patterns of mobility by  demographic  characteristics  but  also  by  people  in  different  occupations  or  socio‐economic  categories.  The results of the Census question on the mode of travel to work is particularly well  used by researchers as it is by transport planners seeking to formulate solutions to the critical  problems  of  transport  congestion  that  afflict  almost  every  city  and  small  town  across  the  country at certain peak hours of the day.  Geographers have used commuting flows to construct  self‐contained  labour  market  areas  that  reflect  functional  characteristics  and  the  Census  SWS  have  been  used  to  define  the  official  Travel‐to‐Work  Areas  (TTWAs)  used  by  various  Government Departments.   Research studies based on census interaction data have been carried out by researchers from  various academic disciplines using a range of analysis methodologies and modelling techniques. 

22

Alongside the empirical analysts interested in flow composition and pattern identification, there  are regional scientists or quantitative geographers who seek to model interaction behaviour, in  aggregate of individual form, using statistical or mathematical modelling. This type of research  has  tended  to  focus  either  on  deterministic  approaches  that  seek  to  establish  the  causes  of  interaction  behaviour  or  on  predictive  approaches  that  result  in  very  practical  applications:  a  new road is recommended that will reduce commuting congestion or new housing is created in  response to the pressure of demand from potential in‐migrants.     10. Internal and international migration statistics from the Census play a critically important role in  the production of annual population and household estimates which underpin the provision of  housing and services. They are used in the estimation of disaggregated flows of migrants (e.g.  flows between regions by ethnic group or economic activity based on administrative sources),  they are used by researchers to validate alternative sets of data from administrative or survey  sources, and they are used as input data for projection models that generate better estimates  of future populations.    11. The  following  text  provides  a  number  of  case  studies  of  the  use  of  migration  and  commuting  data, as well as information about the UK Census interaction data sets:  Stillwell,  J.,  Duke‐Williams,  O.  and  Dennett,  A.  (eds.)  (2010)  Technologies  for  Migration  and  Commuting Analysis Spatial Interaction Data Applications, IGI Global, Hershey, pp. 357.    [2] What impact will the ending of the Census have on social science research?  12. The  migration  and  commuting  data  that  are  derived  from  the  Census  have  a  number  of  strengths.    These  include  the  spatial  detail  with  which  results  are  published,  the  comprehensiveness  of  coverage  and  the  level  of  attribute  detail  (the  socio‐demographic  characteristics of the persons or households involved) attached to each flow. In addition, they  share the general authority of the Census as a high quality source of reliable data.    13. The  ending  of  the  Census  will  therefore  have  a  pronounced  impact  on  research  on  migration  and commuting behaviour from a geographical perspective because it is very unlikely that there  would  be  any  alternative  mechanism  or  combination  of  methods  that  would  provide  the  attribute  detail  for  small  areas  such  as  output  areas  or  super  output  areas  across  the  whole  country.    That  said,  it  should  be  acknowledged  that  the  interaction  data  from  the  1991  and  2001  Censuses  were  rendered  considerably  less  useful  by  the  suppression  and  small  cell  adjustment methods (SCAM) that were imposed on the data in both years in order to minimise  the  risk  of  disclosure  and  ensure  the  data  remained  confidential.  The  effects  of  SCAM  effectively destroyed the utility of the flow matrices at output area level for England and Wales  in  2001  and  it  is  heartening  to  note  that  ONS  will  not  be  using  this  method  of  statistical  disclosure control to adjust the data in 2011.     14. Claims have been made by ONS that the 2011 Census has been very successful in terms of the  response and coverage.  The accuracy of the results remains to be seen but if this is the case,  then these results will form the basis of many research projects into the future – since it will be 

23

the last comprehensive count of the nation’s population disaggregated by such a large number  of  socio‐demographic  characteristics.    There  is  no  other  data  source  that  indicates  migration  flows by ethnicity, religion or health status, for example, and there is no equivalent source of  information on commuting flows by demographic attributes, occupation or mode of travel.  We  can  expect  to  see  the  data  from  the  last  Census  being  accessed  and  used  by  social  science  researchers with greater intensity and for a longer period of time than any previous Census.    15. However, the lack of another census will also act as a driver for the search for and exploitation  of  other  existing  administrative  data  sources  as  well  as  the  stimulus  for  new  data  sets  to  be  collected  through  mini‐censuses  or  surveys  and  the  improvement  of  existing  surveys  (such  as  the International Passenger Survey).  It will encourage much greater use of the existing cohort  and longitudinal studies and will drive the effort to link data from different sources together to  provide ‘linked data’. A raft of new data sets are being released by Government Departments  and  other  agencies  that  will  be  of  enormous  interest  to  the  social  science  community  if  agreements can reached for the data suppliers to provide social science researchers with access  to these data sets.  The  ‘Beyond 2011’ Programme is a good example of a context in which new  and  valuable  data  sets  are  being  assembled  and  used  by  ONS.    The  Economic  and  Social  Research  Council  must  take  a  pro‐active  role  in  reaching  agreements  with  data  suppliers  to  disseminate these data sets, taking advantage of the open government licensing arrangements  that are now in place.    [3]  What  alternatives  to  the  Census  would  provide  population  and  socio‐demographic  data  of  equivalent or higher quality?   16. In the case of migration, there are a number of alternative sources which can or could provide  useful data.  These are arguably of ‘higher quality’ in terms of their timeliness and extent, but  they are of less utility than Census data in a number of other respects, including coverage and  level of detail.    17. Data from NHS administrative sources have been made available for the period since the mid‐ 1970s. The structure and nature of these data have changed over the course of time, reflecting  changes in recording and data processing technologies, as well as changes in the administrative  framework  of  the  NHS.  The  current  ‘NHS  derived’  data  series  are  based  on  comparisons  of  patient  registers  at  different  points  in  time,  with  patients  being  identified  as  migrants  if  their  postcode has changed between the start and the end of the observation period.    18. The Patient Register data are currently published on an annual basis, using a mid‐year to mid‐ year reporting period.  A variety of tabular summaries are produced detailing characteristics of  patients involved in various types of move; for the purpose of comparison with the Census flow  data there is one table of interest, which shows the number of persons migrating from specific  origins to specific destinations, using a reporting geography of local authority areas in England  and Wales.  The table shows the total number of persons moving between (or within) locations,  rounded to the nearest 10.   

24

19. These data are of great importance in monitoring and/or estimating migration levels within the  UK  in  the  intercensal  period.    There  are  some  known  shortcomings,  primarily  that  certain  population  groups    –  young  adult  males  in  particular  –  are  less  likely  to  notify  their  GP  of  a  change  of  address,  and  thus  will  not  be  captured  as  migrants;  however  this  weakness  is  understood,  and  is  common  to  many  other  administrative  sources.      However,  the  data  only  show the number of persons or, at best, the number of persons by age and sex.  They do not  include additional socio‐demographic information of the sort common in Census data, such as  individuals’  religion  or  ethnic  groups,  their  marital  or  family  status,  their  economic  status  or  occupation.     20. International  migration  data  are  generated  from  a  number  of  sources,  including  the  International  Passenger  Survey,  and  from  the  National  Insurance  Number  database.  Immigration is a politically sensitive issue, and all data sources including the Census share – to a  greater or lesser extent  – the weakness that  they are avoidable  by those who which to avoid  enumeration.    21. The  International  Passenger  Survey  (IPS)  is  a  sample  survey  which  includes  visitors  (both  long  and  short  term)  to  the  UK,  as  well  as  UK  citizens  leaving  the  UK  (both  long  and  short  term).  Unlike  Census  data,  flows  published  from  the  IPS  are  disaggregated  by  vistors’  reasons  for  entering the UK (holiday, business etc); however the amount of socio‐demographic information  is very limited and the sample size remains relatively small.    22. The Department for Work and Pensions publish annual aggregates (on a fiscal year basis) of the  numbers of new National Insurance Numbers (NINo) allocated to foreign adults entering the UK  to  work.  These  are  disaggregated  by  nationality  (as  opposed  to  origin  country).  These  are  disaggregated  by  local  authority  of  residence.  Information  is  also  published  regarding  the  age  and sex of persons involved, but only at a national level. The NINo data are a valuable source of  immigration information, but are limited in their socio‐demographic  characteristics.    23. Alternative  sources  of  commuting  data  are  very  limited;  this  is  an  area  of  acute  concern  to  those  who  have  used  the  Census  commuting  data  for  research.  The  National  Travel  Survey  (NTS)  is  collected  by  the  Department  for  Transport,  and  has  been  running  on  an  annual  basis  since 1988. It is a sample survey, and in 2009 collected data from over 8,000 households. There  is very limited spatial information in the NTS, with disaggregation by Government Office Region  or  ‘area  type’  only.  Even  at  this  very  coarse  scale,  there  is  no  disaggregation  by  origin,  destination  and  mode  of  travel.  The  essential  reason  for  this  is  the  small  sample  size,  which  does not permit the publication of detailed disaggregated flows.    24. The  Department  for  Transport  also  publish  annual  average  daily  traffic  flow  data,  for  traffic  passing  fixed  points  on  motorways  and  A  roads.    These  provide  important  information  about  traffic levels, but cannot provide information about either the origins or destinations of trips, or  about the individuals making those trips.   

25

[4]  What  other  existing  sources  of  population  and  socio‐demographic  data  could  be  improved  upon?  25. The NHS Patient Register data are captured at the postcode scale, and thus could potentially be  published at more detailed geographic scales than is presently the case.  There is only limited  scope to include additional socio‐demographic information about migrants; however we would  welcome the publication of aggregate observations by age and sex on a geographically detailed  basis.    26. The  lack  of  socio‐demographic  data  in  the  Patient  Register  limits  its  scope  for  use  in  social  science research: disaggregation of internal migration data by characteristics such as ethnicity  or  family  status  often  reveals  significantly  different  patterns  to  those  suggested  by  counts  of  total persons alone.    27. The Schools Census – known as the Pupil Level Annual School Census (PLASC) prior to 2007 –  records  a  variety  of  information  about  pupils  in  state  education  in  England.  In  addition  to  academic  performance  information,  the  data  also  include  student’s  home  addresses.  Comparisons of these over time can be used to generate migration matrices for schoolchildren.   Access to data from the Schools Census is currently limited, especially and necessarily to those  fields  –  such  as  home  address  postcode  –  which  act  as  personally  identifying.    However,  it  would be possible for aggregate observations to be produced showing the migration of school  children.    28. The  Schools  Census  also  has  the  potential  to  generate  a  regular  series  of  origin‐destination  matrices  disaggregated  by  pupils’  modes  of  transport  (walking,  bus,  car  passenger,  etc)  to  school.  Data of this kind were produced as part of the output of the 2001 Census in Scotland  (and  will  again  be  produced  from  the  2011  Census),  but  differences  in  questionnaire  wording  has  not  permitted  the  production  of  these  data  elsewhere  in  the  UK.    These  data  would  thus  deliver information in England and Wales that are not available from the Census.    29. The  Higher  Education  Statistics  Agency  (HESA)  record  information  for  students  at  Higher  Education  Institutes,  including  the  postcode  of  the  student’s  address  (usually,  their  parental  address) prior to entry to an HEI. These could be used to produce migration matrices showing  the movement of students to universities and colleges.    30. There are no obvious alternative sources of data on commuting. The most widely used attribute  of the Census commuting data is the mode of transport to work, and it is this information which  tends to be missing from candidate data.    31. The  Annual  Survey  of  Hours  and  Earnings  (ASHE)  includes  information  about  employees’  workplace  and  home  postcodes,  but  it  does  not  include  any  socio‐demographic  information  about  the  individual,  nor  does  it  include  information  about  the  usual  mode  of  transport  to  work.     

26

32. ASHE is captured through a questionnaire sent to employers (rather than employees) and thus  there is no scope in the current collection model to collect any additional personal data, such as  mode of transport. Even if this were to be the case, the small sample size (1%) means that no  detailed spatial disaggregation could be made.    John Stillwell and Oliver Duke‐Williams  School of Geography  University of Leeds    22 November 2011 

7   

27

Written evidence submitted by David Owen (Census 08) The census and social science 1. How do social scientists use Census data? 1. The Census of Population is the most comprehensive source of data on the demography, economic circumstances and housing of people living in the UK. It achieves a vastly higher response rate than any other social survey and unlike all the administrative sources, permits social, demographic, economic and housing variables to be cross-classified. It is the most detailed source of data on internal migration and commuting. 2. UK social scientists have benefited greatly from the purchase of the whole of the electronic outputs of the Census of Population since 1981 via ESRC/JISC and latterly Treasury funding. The Treasury purchase has permitted a vast increase in the use of Census data in the private and voluntary sectors as well as the public sector and promoted a great improvement in knowledge of the characteristics of the population, more widely spread across society. 3. The disciplines which make most use of the Census are probably Geography and other closely-related subjects with a spatial perspective, such as Town Planning and Regional Science. However, spatial analysis and hence use of the Census is increasing in Economics and Sociology. 4. Geographers use Census data to analyse the characteristics of localities at a range of spatial scales. A wide range of analyses have been undertaken, including exploration of socio-economic differentials, the geography of ethnic composition, geographical patterns of demographic change and migration and the creation of classifications of areas. The latter have been developed by the commercial sector to create powerful geo-demographic classification systems used in credit scoring, insurance and marketing and for predicting individual consumer behaviour. The Census has been an important source of denominators in the creation of indicators of local disadvantage using more regularly up-dated information sources (indeed it is one of the only sources of information on the economic activity of the population at the local level, which is required in calculating small-area unemployment rates). It has also been extensively used as contextual background data in qualitative research on local areas. 5. The analytical tools applied to the Census include regression models, cluster analysis, spatial statistics and the application of Geographical Information Systems to create profiles of areas within distance bands of physical features and within travel time bands. Mapping and graphical display have been important ways of presenting Census data. Census ‘flow’ data (e.g. journey-to-work data) has been used to delineate functional urban regions and city regions, which identify the hierarchy of cities and towns. This is not apparent from any other source of information. Smallarea Census data has also been used to identify the geographical ‘segregation’ of ethnic groups and explore the factors behind this. The detailed geography of the Census is used to link Census of Population data with a wide range of other data sets to enrich the analysis undertaken. 6. The Census of Population is under-used by UK social scientists. The reasons for this are three-fold: (i) the data is collected only once every ten years and hence the potential for time-series data analysis is very limited; (ii) most of the data is in the form of pre-designed cross-tabulations for geographical areas and hence most analysis involves exploring ecological associations; (iii) other data sources exist which are collected more regularly, contain more variables and are available at the level of individuals (i.e. ONS sample social surveys such as the Labour Force Survey).

28

7. On the other hand, the recent UPTAP Initiative funded by ESRC centred around the award of (mostly) small grants for the secondary analysis of quantitative data sets yielded a large amount of research activity around the analysis of the 1991 and 2001 Censuses and yielded a number of new data sets which open new possibilities for analysis, created by linking the Census with other data sources. 8. The availability of microdata from the Census (in the form of the Samples of Anonymised Records) has permitted more in-depth analysis of Census data because the relationships explored are not constrained by the pre-specified tables used for delivering the bulk of the census output. Regression models can identify the separate influence of geographical location and socio-economic characteristics on a range of phenomena ranging from labour market participation to health. 9. The Longitudinal Studies in England and Wales, Scotland and Northern Ireland provide a powerful tool for tracing the evolution of individuals and families over time and have yielded new insights.

2. What impact will the ending of the Census have on social science research? 1. The impact of the ending of the Census of Population on UK social science research is entirely related to what will replace it. 2. The Census is unique in the combination of geographical detail and information about individuals and households which it yields. The 2011 Census includes new questions which allow more detailed analysis of the migrant population, identity, the impact of student populations and the phenomenon of weekly commuting. 3. If the alternative to the Census yields the same range of data, for the same geographical detail and at greater frequency, then the data environment for social scientists will have been improved by abandoning the Census. 4. However, it is more likely that the replacement will be a series of partial products – such as enhanced population estimates, more localised data on housing, economic activity and health. It will be difficult to link these together and the problem of only having ecological associations between variables and not direct causal associations to analyse will be intensified.

3. What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? 1. The UK is hampered (both in terms of research and administration) by the lack of a population register. Without a comprehensive register with a system for updating when individuals move residence, there is no definitive measure of the population and migration is estimated from data sources which were not intended for this purpose. Population registers are not ideal, because they cannot achieve complete coverage and it cannot be ensured that every instance of migration is recorded. Additionally, obtaining socio-economic information for the people on the register still requires large-scale surveys and/or the linkage of official data bases using the individual registration code number (as occurs in Scandinavia). 2. In the UK, the linking of administrative and commercial data bases to yield Censuslike data (which has been suggested by Francis Maude) is difficult because no unique personal identifier is used in the collection of data. 3. The nearest equivalent is the National Insurance number which is allocated to nearly everybody for purposes of collecting tax and payment of benefits and is allocated to new migrants when they seek work. The DWP maintains the “Lifetime Labour Market Database” which can trace individuals over time and across space. This could be used to identify families and their dynamics. However, the NI number has problems – it is not necessarily unique and people leaving the country do not have to notify

29

anybody that they have done so, so there will be numbers which are inactive or used fraudulently. Though the postcode provides location information, there is no method for linking information about the characteristics of the dwelling an individual is living in, unless the Land Registry and Electoral Register also require NI numbers. These sources would also have to be regularly updated with information about the dwelling in which an individual lived in order to yield data on overcrowding and the presence of amenities (e.g. central heating, exclusive use of bathroom). 4. A database built around linking the tax, benefit and other systems does have the potential to yield much more regular information on travel-to-work than the Census. However, in order for systems like this to yield information on ethnicity and other identity variables (such as religion), each person with a NI number would have to be required to report this information at each contact with the government, since these can change over time. 5. Retailers and credit scoring companies hold large amounts of geographically detailed information on the spending patterns and use of financial products of individuals. However, most of the transactions captured would not involve the same individual identifiers used in contact with the public sector and matching would therefore probably depend upon the accurate recording of residential address. There is also the problem of how transactions for multiple credit cards and store cards held by individuals would be combined, especially when personal details recorded differed slightly. 6. While record linkage could yield data on a more frequent basis than the Census of Population, it is not clear that the quality would be equivalent or better. This is because the collection of data is based on administrative processes in which there is less control over the quality and consistency of the information collected. In some instances, information is provided by an individual and in others is provided by an official on behalf of an individual.

4. What other existing sources of population and socio-demographic data could be improved upon? 1. The other sources of population and socio-demographic data in the UK are very limited and should all be improved, whether the Census of Population continues or not. The most reliable element of the UK population data is the registration of births and deaths which are very accurate and form the basis of the annual estimates of population made by the ONS and the Registrars-General of Scotland and Northern Ireland. It would be possible to add more questions about the parents or the deceased. Attempts have been made to add an ethnic group classification to birth registration, but this is difficult because the concept of ethnicity adopted by official statistics is a social construct and cannot be applied to a new-born child. Recording each parent’s ethnicity might be preferable, but this does not necessarily indicate the ethnicity of the child and may not be possible where the father is absent. 2. Most of the information on the socio-economic characteristics of the population between Censuses is derived from ONS social surveys. The main drawback of these sources is the small sample size, which means that data for small geographical areas cannot be generated without potentially risking disclosure of data for identifiable individuals. Much larger sample sizes would improve this situation, but the costs involved in data collection would be substantial. 3. All surveys suffer similar drawbacks to the Census in not asking all the questions that are potentially of interest to data analysts while becoming regarded as intrusive by a section of the population which will not co-operate. Hence all social surveys are experiencing declining participation rates and are biased in (often unknown) ways. The 2011 Census made great improvements, but this involved substantial expenditure to achieve. However, it was undoubtedly cheaper in terms of expenditure

30

per head than each wave of a social survey.

Declaration of interests I have no commercial interest in the Census or connection with suppliers of Census data. Much of my research work is highly dependent upon the availability of Census of Population data.

David Owen University of Warwick 24 November 2011

4

31

Written evidence submitted by TNS-BMRB (Census 09)

BEYOND 2011: AFTER THE CENSUS  SUBMISSION:  1. How do social scientists use Census data?  We work in the field of social surveys and use census data in four main ways: (i) to assess and correct  for non‐response bias in survey samples, (ii) to pre‐classify neighbourhoods and ensure that each  sample of neighbourhoods is as representative as possible, (iii) as additional context data added to  interview data, and (iv) to assist in estimating key survey statistics at the neighbourhood level (direct  estimates from survey samples are unreliable at this very granular level).   2. What impact will the ending of the Census have on social science research?  3. What alternatives to the Census would provide population and socio‐demographic data of  equivalent or higher quality?  The answer to these questions will depend entirely on what it is replaced with.  All survey data  would be weakened if neighbourhood level data was unavailable but the census is not the only  possible source of neighbourhood data.  Administrative data of various types is already available at  local levels but there are problems with it.  Firstly, there is a lack of full linkage between these  sources as well as an inconsistent approach to updates.  Secondly, there are considerable difficulties  in accessing the data in a usable form.  This means that administrative data is considerably more  awkward to use than census data and, for the most part, less reliable (even if more up‐to‐date).   Furthermore, administrative data is inevitably purpose‐specific.  There is no overarching data  strategy to ensure that key evidential gaps are filled.    However, it is not impossible to reduce these problems to a manageable level.  There is no census in  the Netherlands but there is a reliable database of citizens based on a range of linked administrative  data sources.  Neighbourhood level data can be extracted from this and used widely and the  database is also used to draw samples for government and academic‐sponsored surveys.  In Britain,  the uninformative, if comprehensive, postal address file is used as the sample frame for high quality  general population surveys.  A reliable administrative database would represent an improvement,  especially with regard to correcting survey non‐response bias.   Census micro‐data is currently only available on a 5% basis and only for specific purposes.  This is  rarely used in survey research because (a) it is impossible to link specific address‐level data to survey  sample data (for reasons of confidentiality and due to the 1‐in‐20 nature of the sample), and (b)  even a 5% sample (c 1 million households) is too small for micro‐level analysis of specific  neighbourhoods.   Unless a way can be found that allows micro‐level analysis of specific  neighbourhoods the absence of the 5% micro dataset will not greatly affect survey research.  If the census is absent, it would be valuable if the census 2011 figures could be updated in a reliable  way, even if that means replacing actual data with part‐modelled data.  The Annual Population  Survey ‐ as well as many other surveys ‐ provides reliable national and regional level updates of  census 2011 data (albeit with a greater level of uncertainty due to the much higher level of non‐ response than attends the census) but not at sub‐regional level.   

32

However, the APS data could be modelled at the neighbourhood level given a strong enough base of  administrative data.  For our primary purpose – the accurate classification of neighbourhoods – modelled estimates of this type would suffice in place of real data, although there is a risk that these  estimates will become more and more unreliable as time passes without the decennial anchoring  role of the census.    Finally, one possible option is to make the Annual Population Survey mandatory as the census is at  present.  That would go some way to reducing non‐response bias for national and regional  estimates.  Increasing the size of the APS might also allow for reliable updates at local authority level  too (currently estimates are only considered reliable if three‐years of data are aggregated together).   Modelled estimates would then be required only for neighbourhoods rather than for larger  geographies.    Joel Williams  Head of Survey Methods  TNS‐BMRB  24 November 2011 

33

Written evidence submitted by Professor Heather Joshi, Institute of Education (Census 10) 1.

Introduction

My current affiliation is as a Professor Emerita of the Institute of Education, and the President of the Society for Lifecourse and Longitudinal Studies. The latter is an interdisciplinary international society of scholars interested in research on the lifecourse and the development of longitudinal data. Our members come from epidemiology, psychology, geography, sociology, demography, education and economics. I have formerly been president of both the British Society for Population Studies and of the European Society for Population Economics. My illustrations are taken only from my own work for economy of time, but I am sure I have the support of many colleagues in saying that social scientists working in and on the UK find the census a very valuable resource underpinning research as well as vital for informing central and local government.

2.

Statement on Conflict of Interest

I have retired from my post at the Institute of Education, where from 2000 until earlier in 2011 I was director of the Millennium Cohort Study (MCS), making use of some census data as indicated in para 4 below. As also indicated in para 4, I was the head of the team providing academic access to the ONS Longitudinal Study, and I was involved in making the case for it to be linked to the 2001 census. In my retirement I am working part-time on research (education in social mobility, child residential mobility), which will make some use of existing census data. I, personally, would not expect to be a very extensive user of any data which might become available from any future census, but I do need to say that it would be very useful to colleagues in the social science community, as well as those responsible for making and administering policy and the public at large. I would hope you find my interest constructive rather than conflicted.

3.

How do social scientists use Census data?

Turning to your questions, here are some examples, far from exhaustive, of how a social scientist uses the census, which are elaborated in paragraph 4: •

To compare change over time in the demographic and social structure;



To make international comparisons;



To compare locations at various levels of geography, from regions down to very small neighbourhoods;



To track internal and external migration;



To compare workplaces and places of residence, and the generated ‘travel -to-work areas’;



To examine detailed breakdowns of characteristics of individuals;



To compare the combinations of individual characteristics found among members of the same family or household;

34



To chart the living arrangements of individuals at different ages;



To chart the demographic composition of detailed occupations;



To classify areas as rural or urban;



To inform the drawing up of sampling frames for sample surveys;



To inform the assessment of bias in social survey data;



To assist estimates and projections nationally and locally;



To perform micro-level analyses on Samples of Anonymised Records;



To perform - micro-level analyses using the ONS Longitudinal Study.

4.

Here, I briefly amplify these points and give examples of such use from my own (and coauthors’) work. References are appended at the end of this document. I do not pretend to be offering a comprehensive account of the uses made by the social science community of the census. I am offering this eclectic exercise in self citation as the most cost-effective way I have, in the time available, of illustrating a wide range of uses on the basis of a sample of one user. I am not suggesting that the Select Committee needs to consult the publications cited, just to noting a ‘tip-of-the iceberg’ selection of social science research which has used the census.

4.1

Change over time in the demographic and social structure: Censuses offer the building blocks for assessing long-term change over many decades in demographic, social and economic indicators. An example from my own work is the construction of secular time series of female economic participation rates.

4.2

International comparisons: Just as census offers the possibility to gauge continuity and change over time, they also enable comparisons of conditions cross-nationally. An example from my work includes comparing Britain and France on childbearing by women’s occupation. Ongoing work on residential instability as experienced by young children draws on comparison of US and British censuses.

4.3

Comparing locations at various levels of geography: The finer the level of geographical resolution – down to electoral wards, or smaller approximations of neighbourhood such as Output Areas, the more is social science dependent of census indicators on the social profile of inhabitants. Sample surveys cannot cover all areas in any detail, administrative data ( even where reasonably accurate) do not record all relevant indicators, e.g. ethnicity, housing quality, educational attainment of adults, household composition etc. I have made various uses of census-based small area statistics, such as analyses of child outcomes in the Millennium Cohort and National Child Development Study second generation by census (and other) indicatorss of deprivation at ward and Lower Super Output Area level.

35

4.4

Tracking internal and external migration: Though I have not used them myself the special census migration tables seem to me to be invaluable for identifying the characteristics of people moving between localities within the UK. Duration of residence data can provide one half of the demographic profile of international migration. I have used less detailed census migration tables, by age, to inform patterns of non-response to Millennium cohort study.

4.5

Comparing workplaces and places of residence, and the generated ‘travel to work areas’: Have supported other researchers using commuting data in the ONS Longitudinal StudyS (see below).

4.6

Detailed breakdowns of characteristics of individuals, Examples include:. economic activity or carer status, longstanding illness, educational qualifications, country of birth, ethnic group, type of household or housing into gender, details of age by single years and/or detailed geographies. Most valuable is the cross classifications among such characteristics. For instance a comparison of the educational, occupational and health status of ethnic groups controlling for country of birth and duration of residence, and in different regions of the UK. I have not conducted such work myself. Although sample surveys may be adequate to provide details of these cross-classification at a national level, the results become statistically unreliable at more local levels, where they may be needed for planning services.

4.7

Combinations of individual characteristics among co-residents: e.g. ‘workless households’, homogamous or mixed couples with respect to age, ethnic group, occupation or educational attainment. Besides looking at the distribution of paid work across dual, single and zero earner couples, this feature of the census will be used in ongoing work on gender and social mobility, where it will be necessary to compare the education and occupations of spouses.

4.8

Charting living arrangements of individuals at different ages – e.g. one-person household, one parent family, extended family, unrelated adults living together, and the number of dependent children. These can also control for marital status, housing conditions etc. Census evidence on the living arrangements and location of ethnic groups will be of academic as well as policy interest. I have not been involved in work on ethnic minorities or international migrants, but we have used longitudinal census data to look at different groups’ experience of lone parenthood over time.

4.9

Segregation of occupations, industry etc by gender, ethnicity and age to record the extent and evolution of occupational segregation. I have used such results to identify, gender typed occupations in coding occupational data collected in sample surveys in the analysis of gender pay gaps.

36

4.10

Classification of areas as rural or urban: The Rural Evidence Research Centre at Birkbeck College has devised a classification of areas for DEFRA based upon the settlement pattern of the whole population (which could probably not be implemented from survey data). I have applied this Census-based indicator of rurality to the analysis of the Millennium Cohort and the ONS Longitudinal Study.

4.11

Drawing up asampling frames for sample surveys: Although there may be alterative lists of addresses from which sample surveys could be drawn (e.g. postal or electoral) they do not contain much information on which to stratify samples. The over sampling of areas of high ethnic minority settlement for the Millennium Cohort was based upon ward-level census data for 1991, with adjustments for boundary changes provided by Professor Ludi Simpson.

4.12

Assessing bias in social survey data: One example is the study of attrition through mobility in the MCS.

4.13

Assisting estimates and projections nationally and locally: Census data forms the benchmark for estimates and projections of population, households and pension population by marital status.

4.14

Micro-level analyses of Samples of Anonymised Records No personal use of SARS.

4.15

Micro-level analyses of ONS Longitudinal Study: I was head of LS User Support for 7 years, during which time many academics made use of the 1% sample of longitudinally linked individual census records, with their household at each census also linked to vital registration records. I presume there will be fuller evidence about this remarkable resource and rich resource from the CeLsius team currently responsible for access and from the user community. Its classic value, which should remain a raison d’être, is to provide valid denominators for estimating social inequalities in mortality. My own research on it includes a study of the timing of childbearing by the woman’s occupational attainment comparing Britain and France, social profile of movers between urban and rural England, analyses of the geographical and gender dimensions of health inequalities, and the dynamics of children’s living arrangements.

5.

What impact will the ending of the Census have on social science research?

Just at the time when the technology and governance arrangements permit far better academic use of census data through micro data sets, the existence of this increasingly valuable source of evidence is going to be withdrawn. Given the range of uses to which is it currently put, this will be a blow to empirical academic research and the policy uses which flow from it, removing the bedrock on which much else is built.

37

6.

What alternatives to the Census would provide population and sociodemographic data of equivalent or higher quality?

The routine recording of residential moves, including those out of the country, seems to me to be a top priority. This might be achieved through a better use of the NHS central register and better recording of movements across borders. Clearly the current system is inadequate. The social and economic details collected at birth, and marriage registration could be improved. For births I would like to see true birth order, educational attainment and duration of residence at current address included in on the birth certificate. But without supplementary data on the size of the population at risk to give birth at each age and number of previous births, data on birth registration will be hard to interpret. The most important priority for mortality analysis is to have adequate denominators of the population at risk.

7.

What other existing sources of population and socio-demographic data could be improved upon?

The sample size of national surveys should be increased to allow for adequate coverage of small categories and cross classifications. If the census goes, there should be better ways of sample surveys accounting for visitors and transient residents in households, and allowing for people (like children whose care is shared between separated parents, or weekly commuters) being covered. The linkage of vital events from birth and death registration to the existing Longitudinal Studies’ databases could usefully be continued. Consideration might be given to collecting other information from their sample members, if a way could be found to safeguard confidentiality. This would need to be very carefully thought through but there may be an instructive example in France, which has just recruited a large fraction of its national birth cohort (ELFE) from the Echantillon Demographique Permanente (EDP) – the French equivalent of the LS.

References by paragraph 4.1 Joshi, H. E., Layard, P. R. G. and Owen, S. J. (1985) Why are more women working in Britain? Journal of Labor Economics, Vol 3 January 1985, S147-S176. Joshi, H. E. and Owen, S. J. (1987) How Long is a Piece of Elastic? The measurement of female activity rates in British censuses. Cambridge Journal of Economics, Vol 11, 1, 55-74. Joshi, H. E. (1985) ‘Motherhood and Employment: Change and Continuity in Post-war Britain’ in Measuring Socio- Demographic Change, Occasional Paper No 34, London: OPCS, 70-87.

38

4.2 Rendall, M.S., Ekert-Jaffé, O., Joshi, H., Lynch, K. and Mougin, R. (2009) Universal versus economically polarized change in age at first birth: A French-British comparison, Population and Development Review, Vol 35 Number 1, 89 – 115 and work currently in progress

4.3 McCulloch, A. and Joshi, H. (2001) Neighbourhood and family influences on the cognitive ability of children in the British National Child Development Study. Social Science and Medicine, 53, 5, 579-591.

4.4 Plewis, I., Ketende, S.C., Joshi, H. and Hughes, G. (2008) The contribution of residential mobility to sample loss in a birth cohort study: evidence from the first two waves of the UK Millennium Cohort Study, Journal of Official Statistics, Vol.24, No.3, 2008. pp. 365-385.

4.5 and 4.6 nothing to report here at present 4.7 Joshi, H. (1998) The opportunity costs of childbearing: more than mothers’ business. Journal of Population Economics, 11, 161-183. Davies, H., Joshi, H., Peronaci, R. (1998) Dual and Zero-earner Couples in Britain: Longitudinal Evidence on Polarization and Persistence. Discussion Paper in Economics 8/98, Birkbeck College, London.

4.8 Clarke, L. and Joshi, H. (2001) Accessions and Abdications: Men and women as family markers in linked data for England and Wales. In B. Zaba & J. Blacker (eds.) Brass Tacks, Essays in Medical Demography. Athlone Press, London, 207-230. Rendall, M., Joshi H., Oh, J. and Verropoulou, G. (2001) Comparing the Childrearing Lifetimes of Britain’s Divorce-Revolution Men and Women. European Journal of Population, 17, 365-388.

4.9 Sullivan, A., Joshi, H., Leonard, D (2010 in press ) Single-sex schooling and labour market outcomes’ Oxford Review of Education, Vol. 37, Number 3. Paci, P., Joshi, H and Makepeace, G. (1995) Pay gaps facing men and women born in 1958: differences within the labour market in J. Humphries and J. Rubery (eds) The Economics of Equal Opportunity, Manchester: Equal Opportunities Commission, 87-112.

39

4.10 Joshi, H., Dodgeon, B. and Hughes, G. (2008) A profile of population change in Rural England. CLS Working Paper 2008/4, London: Centre for Longitudinal Studies. http://www.cls.ioe.ac.uk/publications.asp?section=0001000100060009

4.11 Plewis (ed) et al (2007) Millennium Cohort Study: Technical Report on Sampling. CLS. IOE .

4.12 see 4.4 4.13 Diamond, I., Testaghiorghis, H. and Joshi, H. E. (1990) The Uses and Users of Population Projections in Australia. Journal of the Australian Population Association, Vol 7, 2, 151-170. Joshi, H.E. and Davies, H.B. (1992) Pensions, Divorce, and Wives’ Double Burden, International Journal of Law and the Family, 6, 289-320. 4.14 No personal use of SARS

4.15 Ekert, O., Joshi, H., Mougin, R. and Rendall, M., (2002) Fécondité, calendrier des naissances et milieu social en France et en Grande Bretagne: des évolutions contrastées. Population 2002 (3) 485-518 English edition: Fertility, Timing of Births and Socio-economic Status in France and Britain: Social Policies and Occupational Polarization 2002(3) 475-507. Wiggins, R.D., Joshi, H., Bartley, M., Gleave, S. and Lynch, K., Cullis A. (2002) Place and personal circumstances in a multilevel account of women’s long-term illness. Social Science and Medicine, 54, 5, 827-838. Wiggins, R.D., Bartley, M., Gleave, H., Joshi, H., Lynch, K. (1998) Limiting long-term illness: a question of where you live or who you are? A multilevel analysis of the 1971-1991 ONS longitudinal study. Risk Decision and Policy, 3,3, 181-198. Gleave, S., Wiggins, R.D., Joshi, H., Lynch, K. (2004) Identifying Area Effects on Health: A Comparison of Single- and Multi-level Models in P. Boyle, S. Curtis, E. Graham, E. Moore, (eds.) The Geography of Health Inequalities in the Developed World, pp. 245-266, Ashgate ISBN 0 7546 1398 4. Joshi, H., Wiggins, R.D., Bartley, M., Mitchell, R., Gleave, S., Lynch, K. (2000) Putting health inequalities on the map; does where you live matter? in H. Graham (ed) Understanding Health Inequalities, Open University Press, 143-155. Sloggett, A. and Joshi, H. (1998) Indicators of Deprivation in Longitudinal Perspective Environment and Planning A, 30, 1055-1076. Sloggett, A and Joshi, H (1998) Indicators of Deprivation as Predictors of Life Events Journal of Epidemiology and Community Health 54, 4, 228-233. Sloggett, A. and Joshi, H. (1994) Higher mortality in deprived areas: community or personal disadvantage? British Medical Journal, December, Vol 309, 1470-1474.

40

Weatherall, R., Joshi, H.E. and Macran, S. (1994) Double Burden or Double Blessing? Employment, Motherhood and Mortality in the Longitudinal Study of England and Wales. Social Science and Medicine, 38, 2, 285-297. Clarke, L. and Joshi, H. (2003) Children’s Changing Families and Family Resources. In Jensen and McKee (eds) Children and the Changing Family: between transformation and negotiation. London: Routledge Falmer, 15-26. ISBN 0 415 27773-6

Professor Heather Joshi Institute of Education November 2011

41

Written evidence submitted by Mike Hogan (Census 11)   Maintain the census    In essence, the census tells us important things about the changing face of our  society.    We will only really know how important and informative it when it is gone.    So lets not axe it.    The demographics of the UK a re changing more rapidly and in complex and  unpredictable ways.    The census is more important than ever.      Mike Hogan  Integrated Commissioning Manager  Adult Mental Health    29 November 2011   

42

Written evidence submitted by Raj Bhopal (Census 12)   Colleagues, without the census, epidemiological research and surveillance will be seriously  dented, and much of it will become untrustworthy for lack of reliable denominators for the  calculation  of  disease  and  death  rates.  Public  health  activity,  both  surveillance  and  evaluation  of  policy,  will  be  undermined.  We  will  also  find  it  very  difficult  to  undertake  reliable  linkage  studies  that  can  inform  us  on  population  health,  with  appropriate  adjustment  for  potential  confounding  factors.  The  only  serious  alternative  is  to  have  a  population register and that is likely to be even more expensive than the Census. Minority  populations will be disadvantaged, as information on them tends to be particularly sparse,  and the census is an essential resource.    At  something  like  a  few  pounds  a  head  the  census  is  very  good  value,  especially  when  compared to other forms of social survey, which costs about £100 per head. I am hopeful  that  your  consultation  will  lead  to  a  reappraisal  of  the  general  view  that  the  census  is  no  longer necessary.      Raj Bhopal  R S Bhopal, Bruce and John Usher Professor of Public Health  Edinburgh Ethnicity and Health Research Group  Centre for Population Health Sciences  University of Edinburgh      29 November 2011   

43

Written evidence submitted by David Truswell (Census 13) Census - Importance of Population Data in Social Science and Health Care While there are well known limitations to the accuracy of Census Data it still remains one of the internationally best regarded studies of detailed population demographics. I have used Census data adjusted by ONS published on annual projected changes in my work in London on mental health related access issues for programmes of work with a national profile. From 2005 to 2008 I led a programme of work on improving mental health seevbvices for people from Black and Minority Ethnic backgrounds. This was the Focussed Implementation Site Programme that was part of the DH Delivering Race equality Action Plan. I worked in a mental health trust that provided services across 5 boroughs in West London. Then from 2009 to 20011 I was the lead for a pan-London programme of work for Commissioning Support for London supporting commissioners in improving services for people living in London in line with the National Dementia Strategy. I have found Census data an invaluable support to cross-reference and either confirm or challenge other information sources in support of improved services, particularly with improving access for Black and Minority Ethnic Groups. One specific areas that the census has been useful for me has been, comparisons between electoral ward level census data, GP referral rates to mental health services of people from Black and minority ethnic groups and mental health ward admission for the same population. I was able to demonstrate that in Brent and Westminster the mental health services were not admitting disproportionately high numbers of these populations. A summary of this work has been published and I have presented it at both national and international conferences. Another area of my work has been the underestimation of the proportion of people with dementia from Black and minority ethnic populations in Central of Census data, and dementia prevalence projections. This work has recently been published and I will be presenting it at a national conference in the next few days

These findings have significant impact on planning of services and investment decisions by commissioners and are an essential contribution to a reliable evidence base for these decisions.

David Truswell Senior Project Manager Central and North West London NHS Foundation Trust 29 November 2011

44

Written evidence submitted by Dr Julie Fish (Census 14)   The  census  provides  valuable  information  that  is  not  available  from  other  sources. It collects baseline data that I draw upon in my research, teaching and  writing.      Dr Julie Fish  Reader in Social Work and Health Inequalities  De Montfort University    29 November 2011   

45

Written evidence submitted by TWRI Policy and Research (Census 15) 1. TWRI Policy and Research (TWRI) appreciates the opportunity to write to the Science and  Technology Committee about the value of the Census to Social Scientists.  We do this in the  context of a number of years 1  experience analysing, manipulating 2  and subsequently  presenting statistics for Tyne & Wear and parts thereof.     TWRI was established in 1986, following the closure of the Tyne & Wear county council, to  support the five Tyne & Wear district councils through the assembly and analysis of statistical  information.   From 1st October 2011, TWRI became a private company which aims to continue  providing and maintaining a range of socio‐economic, demographic and community safety  research and information services.       We would wish to note that, whilst no census is perfect, the absence of a 2021 census will be        extremely detrimental to the understanding of changes in society in the decade 2011 to 2021.    2.  How do Social Scientists use Census data?    In summary, we have used the Census to undertake trend and multivariate analyses.    This is possible because the Census is the one and only source of data collected consistently  across the country at a sub‐district (i.e. at an output area 3 ) level.  In the past this would have  been what was known as an enumeration district, a small area of contiguous addresses  covered by a Census enumerator.     3. Analysis of trends across a sub‐area of a district    Because the Census is collected at a small area level, we can aggregate the data up to any area.   Where appropriate, this can include appropriate apportioning and manipulating.  So if, for  example, Local Authority wards or Parliamentary boundaries change, we can by going back to  the original small area data, manipulate it and aggregate up the old data to new boundaries.     Hence we can establish trends, from the past, in particular indicators, for new local authority  wards and Parliamentary constituencies.    So, for example, we can establish ‘household car ownership’ rates, ‘average household size’ or  ‘percentage aged under 16’, from previous censuses for new wards.  This helps in                                                              

1

 One of us was first involved in understanding and adjusting the 1961 Census to the 1974 Ward boundaries  following on from the 1973 Local Government re‐organisation.  He has worked on all subsequent censuses,  including the 1966 sample census; thus he has had over thirty‐five years working with census material.   Another has over 30 years and another over 20 years experience, again including multivariate analyses.    2  By manipulating the census results we mean a) adding cells to create new variables or new areas, b)  subtracting cells, where this is better than adding, c) creating indicators [e.g. percentage of the population  aged under 16, or over retirement age; percentage of the population/households without access to a car] and  d) apportioning to create measures for boundary changes [e.g. new constituencies and wards], this enabling  local, [e.g. within district], trends to be detected.         3  An output area is a group of full postcode addresses of between 40 and 125 households. 

46

4.

5.

6.

7.

understanding the nature of a particular ward and is useful for politicians and service  providers.      Multivariate analysis of small area Census data     Again, because the Census is collected at a small area level, we can calculate a number of social  and demographic indicators at that level.  For example, we can calculate ‘unemployment rates’  for males and for females, ’percentage long‐term sick’ and ’percentage ethnic minorities’, ‘car  ownership rates’ and ‘percentage owner‐occupiers’.  No other apportioning or manipulating is  required.  We can then input the indicators into a multivariate Factor analysis.    Using Factor analysis, we can determine the importance of each of these indicators in the  underlying Factor, (or Factors if there is more than one) which best describes the data.   Generally speaking, from the Census, we find that the first factor can be described as a  ‘poverty’ measure.     Some of the indicators used in the analysis, may turn out be unimportant.  That would mean  that that indicator was measuring something else. This is important to learn.     We can then calculate Factor scores for each small area.  We can then see which of the small  areas has a low (or high) score on each of the Factors.  This will have policy implications for  intervention to reduce the impact of poverty.    [We can take two (or more) census small area output data sets and undertake a Factor analysis  on both sets together.]    In summary without small area data collected through the Census in a consistent way across  districts this kind of analysis is not possible.    What impact will the ending of the Census have on social science research?     It follows from what was said in the previous paragraphs that unless Districts and/or others  invest in their own data collection, these kinds of analyses will be impossible to undertake,  leading to a lack of understanding of poverty and other measures, at the small area level.      It would also be impossible to get the same kinds of analyses done between different  authorities.      What alternatives to the Census would provide population socio‐demographic data of  equivalent or higher quality?    In theory, it would be possible to use administrative data to undertake these kinds of analyses.   In practice, however, changes in definition make this next to impossible.      For example, the administrative definition of unemployed has changed a number of times  during the last 30 years.  This would mean that whilst we can monitor unemployment at a  ward level, when there are changes in the definition this leads to discontinuities in the  analyses.  Any relative change in unemployment rate could have arisen because of the change  in definition.   

47

Similarly, there have been changes in those allowed to have access to council tax benefit; this  means that this would not be a suitable means of comparing sub‐areas of districts over time.    Additionally there are three problems with administrative data namely confidentiality,  uniformity in collection and non‐comprehensive coverage of the data.  People do not always  register a house move with a GP, so that this information would be inaccurate.  The Census, on  the other hand, counts people at their actual address on Census day.    8. What other existing sources of population and socio‐demographic data could be improved  upon?    It is difficult to say that any source would be adequate.  To enable analyses to be undertaken,  from administrative data, it would be necessary make sure that there is no change in  definition.   This is, as society changes, impossible.    An alternative to the Census would be a population register, as in Sweden.  However, this  would only give Population estimates by age and sex, household size and type.  It would be  unable to give measures of car ownership, tenure, disability, ethnicity and other socio‐ demographic indicators provided by the Census.    9. Summary    It is our view that the Census is irreplaceable.  Any attempt to do the above kinds of analyses,  without it, will fall well short of the desired level of accuracy that can be provided by  undertaking a 100% population census.        TWRI    29 November 2011 

48

Written evidence submitted by Dr James Kirkbride (Census 16)   Re: Census and Beyond 2011    I am writing to share my views of how Census data is currently used in the research that I  conduct. My specialism is epidemiology, and in particular, psychiatric epidemiology. While I will  naturally draw upon my own experiences of how the Census data is used for epidemiological  research into psychiatric disorders, much of what I say will apply to epidemiologists and public  health practitioners interested in a range of other population health issues. To structure my  comments, I have used the question headings you provided for this call. I, as other academic  colleagues I work with, are grateful that you are taking on board the views of those engaged in  social science and public health research in moving forward from the Census.  1. How do social scientists use Census data?  In my field we use Census data to provide an accurate estimate of the population who may be  “at‐risk” of developing a disorder in a given catchment area. This catchment area may be  geographically focussed i.e. on just a single Borough, or may be broad i.e. an estimate of the  population at the national level. This data is primarily used as the “denominator” number which  goes into calculating estimates of the incidence (number of new cases) and prevalence (number  of new and existing cases) of a particular disorder. Estimates of incidence and prevalence are  vital for several reasons related to population health. These include   (i) (ii) (iii) (iv)

accurate health service planning    clues for aetiological (i.e. causal) research  information to develop potential public health campaigns (prevention and health  promotion)  Health economics – to estimate the burden of disorder(s) and potential associated  costs in a given population  

Of these, it is the second (ii) with which I have most direct experience. By calculating incidence  rates of disorders in specific populations (and communities) we can look at variation in those  rates along several factors which might be of interest to us. These include looking at how  incidence rates of disorder vary geographically, or vary by other factors such as age, sex,  ethnicity or occupation. Identifying variation in incidence rates can provide clues about the  aetiology (or cause) of a disorder, which can advance scientific understanding, reduce the  burden of disorder (through prevention strategies which follow) and reduce the direct and  indirect economic costs to health services and society associated with disease. The census  provides essential underlying denominator data, stratified according to such factors, to enable  us to estimate these incidence rates. One of the unique aspects of the census here is that the  data can be stratified according to more than one variable at a range of spatial geographies. By  having this information, we are able to know how many people in a given geographical location  belong to each group we are interested in – for example, we can split the population by age, sex  and ethnicity. By knowing this, we can estimate the number of cases of a given disorder which  occur in particular groups of people, allowing for the estimation of incidence in different groups.  By comparing incidence in different groups, we get valuable information about the burden and  impact of disease in different communities. For example, we have used census data in this way  to show that men have greater rates of psychotic disorder than women at young ages, but that  this pattern reverses in mid‐life (Kirkbride et al., 2006), and we have also shown that black and  49

minority ethnic groups have higher rates of psychotic disorder than their white British  counterparts (Coid et al., 2008, Fearon et al., 2006, Kirkbride et al., 2008b); vital public health  information and important in understanding the causes of such disorder. Furthermore, because  census estimates of the population can be stratified across several factors at once (age, sex,  ethnicity, etc…) this allows us to build this complexity into our models of health and disease,  controlling for the effect of one variable upon another, to eliminate confounding – the chance  that an association between disease and a risk factor is really due to another factor. For  example, in our 2008 paper, we used census data in East London to show that the incidence rate  of schizophrenia remained elevated for ethnic minority groups, after taking into account  possible differences between these populations and the white British group in terms of age, sex  or social class, which could otherwise have explained this association.   Accurate estimates of the population at‐risk – provided for by the census from the national  level to the level of output area – are vital for understanding variation in the incidence and  prevalence of a variety of diseases, which can then be important for aetiology (causal  research), health services planning, health economics and public health. Whatever supersedes  the census will need to provide accurate estimates of population data, stratifiable according to  several factors, down to fine spatial (i.e. geographical) scales.     2. What impact will the ending of the Census have on social science research?  Without an adequate replacement for the census we will not be able to conduct the type of  research outlined above. This will have major impacts on our ability to detect variation in the  rates of disease, geographically and according to different sociodemographic groups, meaning  that models of service provision will be less accurate, it will become harder to investigate  possible aetiological associations between sociodemographic factors and disease, and harder to  forecast (predict) which groups are at greatest risk of developing illness. These consequences  will have downstream effects for health economics.    A particular aspect of my social science research – and that of many others – which will be  affected is our ability to study geographical variation in the incidence of health (well‐being) and  disease. Currently population estimates from the census for different geographical aggregations  (particularly the super output area and the electoral ward) allow us to estimate rates of  disorders in small communities. If these rates vary, we can discover whether any social or  physical factor in our environments varies with them, which has possible implications for causal  understanding. For example, in psychiatric epidemiology we have used census data to show that  the incidence of schizophrenia varies across neighbourhoods (Kirkbride et al., 2007, Kirkbride et  al., 2006), and that this variation is associated with certain characteristics of neighbourhoods  (Kirkbride et al., 2008a), including deprivation and social capital (more urban, more deprived,  less socially cohesive neighbourhoods have more schizophrenia). Interestingly, these findings  have motivated neuroscientists to study differences in the brains of people living in rural and  urban environments, finding that those people in urban environments have a greater reaction to  stressful events. Thus, the epidemiology (facilitated by the census) provided a testable  hypothesis for neuroscientists, with the results suggesting that urban environments may alter  the way we process information in our brain, and very tentatively, govern our risk of developing  certain psychiatric disorders.  

50

This is the kind of translational research which is threatened (and which will be lost to  overseas competitors) if the census ceases to exist, or if suitable equivalent data is not made  available in its place.     3. What alternatives to the Census would provide population and socio‐demographic data of  equivalent or higher quality?  The census provides valuable, accurate estimates of the population. As with any statistics there  is a degree of error in their estimation, but for the kind of social science literature I have  outlined, they represent the gold standard, with currently available UK alternatives lagging some  way behind in scope, scale and precision. The alternatives which would provide equivalent or  better data for population health research of this kind can be separated into (i) research‐led  initiatives or (ii) government‐led initiatives.   (i) Research‐led initiatives would include certain study designs – like birth cohorts – which  follow‐up a group of people born at the same time, with repeated assessments at various  stages over the life course, for the duration of their lives. Examples in the UK exist, such as the  1946 or 1958 Medical Research Council birth cohorts. The advantages of such studies is that  they provide prospective clues for causal research of health and disease (i.e. the risk factor  was measured before the person became ill), providing great power to detect causal  associations. However, such studies are costly, have issues with loss to follow‐up, take a long  time (several decades) and even the largest are small relative to the rate of certain disorders.  For example, the 1946 birth cohort contained only ~5000 singletons in the sample. For  disorders such as schizophrenia, what we know from epidemiology is that this would yield  only roughly 50 cases over the entire life time of the participant members; too small for  meaningful population research. Thus, I believe such options, while powerful in their ability to  detect associations between disease and risk factor, do not provide adequate population data  for the type of research currently facilitated by the census  (ii) Government‐led initiatives – one possibility, although it shamefully does not currently exist in  the UK – is to develop national registers capable of linking up a range of government  databases on the population to provide extremely valuable, prospective, nationally‐collected  linked datasets for the study of health and disease. Such registers are commonplace in  Scandinavian countries where databases are linked through people’s social security number.  Such data is extremely powerful to detect possible associations between disease and risk  factors and provides population level coverage. However, such databases no doubt have  implications for how data is shared between government agencies, and the fiasco over ID  cards illustrates the difficulties in setting up such a registry system, though it would be a  valuable – and in some ways more powerful – method of analysing health data in the  population.   4. What other existing sources of population and socio‐demographic data could be improved  upon?   Unsure. There is nothing currently available in my eyes which comes close to providing the rigour  and precision of the census. Random small sampling with extrapolation to the remainder of the  population is a possibility, but samples will rarely be representative, even if participation if  selected is made compulsory. Another option would be to link up existing government databases  for such research, but this has massive cost, logistical and moral challenges. At best, the decision  51

to scrap the census is very disappointing for social science research, and at worst, this could set  back some areas of social research by several decades.     I am very grateful to the Committee for giving researchers the opportunity to express their views  on this matter. It is of grave concern that this resource should be superseded with something  capable of permitting (and ideally improving) upon the type of population‐based research made  possible by the Census. The estimation of incidence and prevalence rates (of any health  outcome) will only be as good as the denominator data from which those data are ascertained.       Dr James Kirkbride Ph.D. (Cantab), M.Sc, D.LSHTM, B.A. (Hons)    29 November 2011    I have no conflicts of interest to declare  I am a Wellcome Trust research fellow based at the University of Cambridge  References     Coid, J. W., Kirkbride, J. B., Barker, D., Cowden, F., Stamps, R., Yang, M. & Jones, P. B. (2008).  Raised incidence rates of all psychoses among migrant groups: findings from the East London  first episode psychosis study. Arch Gen Psych 65, 1250‐1258.  Fearon, P., Kirkbride, J. B., Morgan, C., Dazzan, P., Morgan, K., Lloyd, T., Hutchinson, G.,  Tarrant, J., Lun Alan Fung, W., Holloway, J., Mallett, R., Harrison, G., Leff, J., Jones, P. B. &  Murray, R. M. (2006). Incidence of schizophrenia and other psychoses in ethnic minority groups:  results from the MRC AESOP Study. Psychological Medicine 36, 1541‐50.  Kirkbride, J., Boydell, J., Ploubidis, G., Morgan, C., Dazzan, P., McKenzie, K., Murray, R. & Jones,  P. (2008a). Testing the association between the incidence of schizophrenia and social capital in  an urban area. Psychological Medicine 38, 1083‐94.  Kirkbride, J. B., Barker, D., Cowden, F., Stamps, R., Yang, M., Jones, P. B. & Coid, J. W. (2008b).  Psychoses, ethnicity and socio‐economic status. British Journal of Psychiatry 193, 18‐24.  Kirkbride, J. B., Fearon, P., Morgan, C., Dazzan, P., Morgan, K., Murray, R. M. & Jones, P. B.  (2007). Neighbourhood variation in the incidence of psychotic disorders in Southeast London.  Social Psychiatry and Psychiatric Epidemiology 42, 438‐445.  Kirkbride, J. B., Fearon, P., Morgan, C., Dazzan, P., Morgan, K., Tarrant, J., Lloyd, T., Holloway,  J., Hutchinson, G., Leff, J. P., Mallett, R. M., Harrison, G. L., Murray, R. M. & Jones, P. B. (2006).  Heterogeneity in Incidence Rates of Schizophrenia and Other Psychotic Syndromes: Findings  From the 3‐Center ÆSOP Study. Archives of General Psychiatry 63, 250‐258.     

52

Written evidence submitted by the Institute for Fiscal Studies (Census 17) Declaration of interest: nil  This is an ESRC funded investment. The views and statements expressed are those of the authors  and do not necessarily reflect the views of the ESRC.  How do social scientists use Census data?   1. There are a number of ways in which researchers at IFS use Census data.   2. The census is the key source of information on population statistics and this information is very  important in their own right. Information on how a population is ageing for example is key to  identifying where policy needs to take into account a changing population structure and to plan  how to spend resources.   3. Population statistics are also important in work that uses data on sub‐samples of the  population. Cross‐sectional and longitudinal samples of the population are inevitably subject to  biases in the type of individual or household that responds. In order to generalise results to the  whole population, one needs to appropriately weight the data to take account of non‐response.  These weights are calculated using information about population totals and the structure of  population taken from the Census.  4. Population total are also used to “gross up” data to work out the number of people or  households affected by a particular policy for example.  5. Information on socio‐economic characteristics is available from the Census at a very  disaggregated geographical level. This information can be used and combined with other data  and will give a very accurate picture of the extent of any deprivation in any local area. For  example we have used information on various indicators of socio‐economic status at the output  area level and merged it into administrative data we have on schools and universities.    What impact will the ending of the Census have on social science           research?   6. The most likely negative impact is that researchers will have to pay for alternative sources of  data.  What alternatives to the Census would provide population and socio‐demographic data of  equivalent or higher quality?   7. While there other sources of information such as ACORN or MOSAIC data which contains  information about socio‐economic status as a very disaggregated geographical level, the census  is the only freely available data.     What other existing sources of population and socio‐demographic  data could be improved upon?  8.   No comment.    Institute for Fiscal Studies    29 November 2011 

53

Written evidence submitted by Dr Stephen Patterson (Census 18) Response to inquiry by Parliamentary Science and Technology Committee into future of Census of England and Wales 1. I have received a copy of the request for submissions to the parliamentary inquiry on impact of ending the Census on social science research. You may be interested in the following comments. 2. There is a long tradition of conducting decennial Censuses in the UK. The first official Census of England and Wales was conducted in 1801. It would be a loss if this tradition came to an end. 3. In the modern day the task of conducting the UK Censuses is complex. Through long experience the Office for National Statistics has developed the expertise to conduct the Census of England and Wales. 4. If the Census is to be replaced, it needs to be determined that alternative sources for all Census data items really are available and accessible. 5. From our point of view in the Public Health Department of NHS Suffolk, the Census is one of the main sources of data for reporting on the health of the local population. Census topics that provide particularly useful data include: • • • • • • • • • • • •

rebased mid-year population estimates by age and sex population estimates by ethnic group limiting long-term illness carers household and family composition lone parents housing and tenure communal establishment populations qualifications occupation and socioeconomic classification religion car ownership

6. These data are widely used in public health, in health needs assessments, Annual Reports of the Director of Public Health, ward profiles and other reports. 7. Census data are available from the level of Output Area to the full range of higher geographies. A similar geographical range of data may not be available from alternative sources. 8. It is also useful to be able to access this whole range of data from a single clearly organised source, e.g. ONS Neighbourhood Statistics,

54

the main ONS website (for 2011 Census data, when it becomes available), SASPAC, InSite or C2001 Census Data Manager. 9. ONS has developed the expertise and infrastructure to conduct the Census. The amount of information requested in the Census is not large and it is not a great burden on the public to complete the forms, as is shown by the high response rates for the 2011 Census. If anything, this foundation should be built upon, with additional questions added to the Census in future years. More questions could be added on lifestyle, including smoking, for example. I hope these comments are of interest.

Dr Stephen Patterson Epidemiologist Department of Public Health NHS Suffolk/Suffolk County Council 29 November 2011

2

55

Written evidence submitted by Professor Edward Higgs (Census 19)

1.  As a British social and economic historian I have been using the raw household returns from the  historic British censuses for the 35 years.  I am also one of the co‐researchers currently working on  the Integrated Census Microdata (I‐CeM) Project, a three year project funded by the Economic and  Social Research Council (ESRC) to create a single integrated dataset of the British census returns,  1851 to 1911, for academic research purposes.  This involves receiving data from a commercial  partner, standardizing it, and coding and classifying it to make it comparable with data from modern  censuses and longitudinal surveys.  We hope to link the people in the censuses across time, and with  modern surveys, so that we can study subjects such a social and physical mobility, the life courses of  the disabled, the intergenerational experience of migrant families, and so on.  2.  As a historian of the nineteenth century the original, individual‐leve,l census returns are a vital source  because they locates everyone spatially, socially (by family and occupational strata), and temporally.   The latter is very important because the census is a de facto rather than a de jure enumeration – it is  a snap shot on one night of the year, and so avoids double counting.  3.  The existence of names and addresses that identify individuals is also of great importance because  they allow historians to link the census returns with other nominal records so enriching our  understanding of individual circumstances in the past.  4.  As a social scientist interested in following families and individuals across time it is also important to  have data that is comparable.  In England and Wales census data has been created by a single body,  the General Register Office (now part of the Office for National Statistics) for the past 170 years, and  this has ensured a degree of continuity in the information collected, and in the definitions used when  collecting that data.  For example, the modern definition of what constitutes a family was developed  in the nineteenth century.  When definitions of terms and variables have changed, the census‐taking  authorities have been scrupulous in explaining those changes.  For example, in some Victorian  censuses women were told not to put down an occupation if they worked at home, except where the  home was a farm or a shop, when they were assumed to be working in the family business.  But in  other censuses this instruction was not given.  We need to be able to reconstruct these instructions  so that we know what the data means.  5.  As a historian and social scientist, I am not particularly wedded to the continuation of the decennial  census, as long as what replaces it can provide comparable information.   That would mean:  a) data collected at one point in time;  b) data which is collected using consistent definitions, which are consistent over time, or if there are  changes in definitions, these we can reconstruct;  c) data which gives names and addresses. 

56

6.  In practice, I am not at all sure that alternative means of data collection to the census could meet  such criteria.  Thus:  a) I cannot see how one could guarantee that data collected from a multitude of differing sources  would not relate to people at different dates, and so be open to double counting.  How do we  know that Edward Higgs at one address is not the same as another Edward Higgs recorded at  another address some time later?  With the abolition of the ID card scheme of the last  government, we do not have a common identifier to tie people down.  b)  In such circumstances I cannot see how one can ensure that data collected in different ways, and  by differing agencies, is strictly comparable.  Do they give the same instructions to respondents  when asking questions and collecting data?  We need to remember the old data processing adage  – GIGO – Garbage in Garbage Out.  c)  Will the data collected contain the names and addresses of identifiable people?  I cannot see how  information from commercial, or non‐census government, databases could be used for census  purposes without breaking the provisions of the Data Protection Acts.  Certainly, I cannot see  businesses being willing to hand over the data they have collected from their customers for  marketing purposes, and I believe that Tesco has already refused to hand over the data collected  via its store loyalty card scheme.  7.  In these circumstances I cannot but think that a shift away from a de facto census every ten years  will seriously undermine the usefulness of the data collected for longitudinal, historical analysis.    Professor Edward Higgs  Department of History  University of Essex    28 November 2011 

57

Written evidence submitted by Centre for Longitudinal Study Information and User Support (Census 20) Responses are sought to the question: ‘How do social scientists use Census data?’ One of the census data resources for social scientists is the subject of this submission.

1.

Declaration of interest 2. This submission is being made by Professor Emily Grundy and staff of the Centre for Longitudinal Study Information and User Support (CeLSIUS) at the London School of Hygiene & Tropical Medicine http://www.celsius.lshtm.ac.uk/. CeLSIUS has been funded by the Economic and Social Research Council since December 2001 to promote and facilitate the research use of the ONS Longitudinal Study by UK academic researchers, both staff and students, and this submission is made on their behalf. Additionally Emily Grundy has made extensive use of data from the LS in her own research. How do social scientists use census data?

3. This response relates to how social (and other) scientists use the census data incorporated in the Office for National Statistics Longitudinal Study (LS). The LS was established in the 1970s largely to address the problem of inadequacies in the data available to analyse mortality differentials in the population of England and Wales and biases resulting from the fact that denominator (population) data came from the Census and numerator (death records) data from vital registration. The initial sample was drawn from the 1971 Census on the basis of birthday, in order to facilitate linkage. All those born on four birthdays per year were included giving a sample amounting to just over 1% of the population of England and Wales. The study has been maintained through the addition of new births and immigrants with the same birth date, and includes individual level data from four censuses (1971, 1981, 1991 and 2001) as well as linked information on births, deaths and cancer registrations. Linkage of data from the 2011 Census is currently underway and will be available to the social science research community in 2013. Access to anonymised data for research purposes is permitted under various strictly adhered-to conditions which include only allowing access to microdata in an ONS secure data laboratory. Advantages of the LS 4. The LS is representative of the whole population of England and Wales, including those in non-private households (such as nursing and residential care homes, prisons etc), includes all age groups and includes census information about other people in sample members’ households at each census which provides additional opportunities for examining intergenerational continuities and changes. The ‘width’ of the sample in terms of size means it is possible to study relatively small groups, such as members of particular ethnic minority groups or older people resident in institutional settings (a group excluded altogether from most surveys). The ‘depth’ of the study over time makes it increasingly valuable for research including a life course or intergenerational perspective, for example, it is possible to examine characteristics of adults aged 31-46 in 2001 by attributes of the parents they lived with as children in 1971. A further strong advantage of the LS

58

is minimal bias due to non-response or attrition, as census coverage is good and rates of linkage high. 5. Since the LS comprises all persons born on four days of the year, the sampling fraction is approximately 1.05% and sampling bias is almost nil. At each census the LS sample numbers over 500,000 and the full database currently includes well over 1 million people. This is by far the largest longitudinal dataset in the UK; it allows analysis of small areas (well below local authority level), particular ethnic groups and specific occupational groups. These are not possible with any other longitudinal dataset currently available. Research based on the LS 6. The current list of publications based on LS research includes over a hundred books, more sections of books, and approximately 350 articles in journals – see list at http://www.celsius.lshtm.ac.uk/publications.html. These publications present results from research on a wide range of topics including, for example, studies of associations between unemployment and mortality; effects of marriage and marital history on health and mortality; ethnic variations in social mobility; trends and differences in fertility patterns; consequences of lone parenthood; social class differences in cancer survival; links between geographic and social mobility; influences on household structures among older people and many others. Current projects, listed at http://www.celsius.lshtm.ac.uk/projects.html, are also wide ranging and include many studies of inter relationships between ethnicity, location and labour market or demographic behaviour; many studies focussed on understanding differences in health and mortality; and many examining effects of regional or local characteristics and change in specific locations (for example, effects of in and out migration on the population of Cornwall). What impact will the ending of the Census have on social science research?

7. Most of the research carried out using the LS could not be undertaken using other existing sources. Indeed, no national, cross-sectional dataset offers the same combination of sufficient numbers for detailed work with repeated measures spanning several decades, and for this reason the LS is used for cross-sectional as well as longitudinal analysis. The older birth cohort studies, for example, provide more information on long periods of study members’ life courses (the LS now spans 40 years) but are not large enough to examine small population sub groups. Even large scale surveys do not provide the potential for looking at sub regional differences or reorganising geographies in the way that the LS permits. What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality?

8. No other source would compensate for the absence of census data in the LS, but some parts of the shortfall might be addressed by more linkage of records from other sources. The Scottish Longitudinal Study (SLS) and the Longitudinal Study of Northern Ireland (NILS) were both created much more recently than the LS and used the design and working of the LS as a template. However, unlike the LS, the two newer studies have steadily expanded the range of existing records which are added to the dataset (or, in some cases, can be linked to it at an individual level for each project that requires these data).

59

9. Records available in the Scottish or Northern Ireland Longitudinal Studies and not in the LS include: -

marriages hospital admissions and discharges (including day hospitals and psychiatric hospitals) in Scotland school census data, including exclusion records, in Scotland migration data within Northern Ireland Valuation and Lands Agency data for Northern Ireland, also value of property records of the prescription of drugs in Northern Ireland records of breast cancer screening in Northern Ireland births and stillbirths to fathers who are sample members – the LS only has mothers; (data for fathers was linked 1971-1974 but was discontinued due to poor linkage rates).

10. There is a legal gateway for similar linkages in England & Wales but to date little progress has been made in using this to enhance opportunities for social science research.

Centre for Longitudinal Study Information and User Support November 2011

60

Written evidence submitted by Dr Nicola Shelton (Census 21)

Declaration of Interests The author Dr Nicola Shelton is a health and population geographer in the Department of Epidemiology and Public Health at UCL. The author has used UK Census data for health geography and health surveillance purposes. Contact [email protected] 1. How do social scientists use Census data? In the context of health geography and health surveillance the Census is used for a variety of purposes. Census data on health have been asked variously since 1991 and have included caring for others since 2001. Health outcome data including cancer registrations and mortality and hospital admissions and discharges are also linked to the Census through the various Longitudinal Studies. The Census is also used to provide sampling frameworks from which sample surveys are drawn and to create reference populations against which sample survey data can be age standardised for comparison between e.g. geographical regions and socio-economic groups. Census data is also used to develop indices that allow small area estimation for health statistics such as smoking and obesity prevalence and for measuring health inequalities. 2. What impact will the ending of the Census have on social science research? The national and regional and socio-economic and demographic variations in health and care will be lost as will the wealth of analysis of health variations from the longitudinal studies. Because the UK does not have a population register there will be no alternative sampling framework other than that of the electoral register which includes data for adults only or the NHS Patient Register data which includes only registered patients. Developing sampling frameworks for sample surveys of socio-economic and geographical and ethnic health inequalities will be very difficult especially for children and immigrants and for those groups less likely to be registered with a GP or to keep that registration up to date e.g. young men. Age standardisation would be problematic as it would be based on increasingly inaccurate population estimates over time. 3. What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? A comprehensive population (identity) register linked to vital events and migrations would deliver the population and socio-demographic data of equivalent or higher quality, but may be a more costly than continuing with the Census in 2021 and beyond if it were to include the full range of variables covered in the Census, and would take some decades to develop. It would not provide continuity for the Longitudinal Studies. 4. What other existing sources of population and socio-demographic data could be improved upon? The NHS Patient Register data could be made available as a sample framework of individuals. This would provide also an opportunity to sample individuals and demographic characteristics and particular health conditions subject to disclosure control.

61

The lack of socio-demographic data in the Patient Register limits its scope for use in social science research: disaggregation of characteristics such as ethnicity often reveal significantly different patterns of health inequality to those from the total population. The electoral register could be extended to include all ages as a basic population register, but again would not provide the socio-economic and demographic detail of the Census.

Dr Nicola Shelton Head of Health and Social Surveys Research Group Epidemiology and Public Health, UCL 30 November 2011

62

Written evidence submitted by the British Society for Population Studies (Census 22) Declaration of interest The British Society for Population Studies (BSPS) comprises persons with a scientific interest in the study of human populations. Its main objectives are to further the scientific study of biological, economic, historical, medical, social and other disciplines connected with human populations and to contribute to the public awareness of them. Our four hundred BSPS members include the public sector (including central and local government and health services), academia and business. They have a wealth of experience in empirical research, with much of it drawing on Censuses in the UK and other countries but also much based on data from other population surveys and from administrative sources not designed primarily for demographic analysis. Summarising points 1.

Good data is the key priority of those who currently rely on Census data for information about the size, distribution, characteristics and welfare of the UK population: good in relation to population coverage, geographical detail, accuracy of information, and range of topics. While many users would prefer greater frequency than once every 10 years, this should not be at the expense of quality on any of these criteria. At the same time, there are some aspects covered by the Census that change relatively slowly, for which 10 years is a satisfactory time interval.

2

The increasing ubiquity of geographically coded data means that the need for small area data will continue to grow inexorably. Any move away from providing such data could not be more detrimentally timed.

3.

Much valuable policy-relevant research relies on the high level of consistency between decennial Censuses to understand change over time and this would be lost by shifting to data collected in different ways.

4.

Most administrative data sources have a very limited number of variables on personal characteristics, so the challenge of replacing the Census is to link the administrative sources together for individuals and households.

5.

Key policy concerns including energy use and social segregation cannot be supported without local analysis of the characteristics of commuting and migration which only the Census provides at present (mode of transport to work; socio-economic and ethnic background of migrants within the UK).

6.

For researchers, the key question is ‘Is there an alternative data source with the range, detail and quality of information that the Census currently provides?’ and not ‘Which is the best alternative, given that Census will be replaced?’

[1] How do social scientists use Census data? 7.

BSPS members use UK Census of Population data for three main purposes: (1) to provide descriptive contextual information on populations of interest; (2) for secondary data analysis, often using advanced quantitative methods, to address questions which improve our understanding of UK society and how it is evolving; and (3) as a means of benchmarking other surveys to ensure that they represent the total population of an area.

8.

The most traditional form of Census outputs, dating back to the first Census in 1801, comprises what has become known as the Area Tables (AT). These provide data for geographical areas ranging in scale from the UK level through the 4-country and regional level down to the very local scale, allowing understanding of neighbourhoods of around 100 households or more. The parts of the AT most heavily used by BSPS members are the ‘multivariate tables’, in which the answers to one Census question are cross-tabulated with the answers to one or more other questions, such as in the table of age by sex by whether living in a household or in some form of institutional accommodation such as a hall of residence or nursing home. The ‘univariate tables’, which present the results for each Census question separately, are less widely used in social science research, because the latter is primarily concerned with the way in which the population’s various individual and household characteristics relate to each other, as for instance between ethnicity and the Census measures of health, deprivation and well-being.

 

1 63

9.

BSPS members are also major users of the Origin-Destination Statistics (ODS) which allow the study of how people move between places, most notably migration and commuting but also (though for Scotland only) travel to place of study. While the Area Tables include some of the data on these topics in summary form, the ODS are highly valued because they show the flows between each geographical area and every other one separately, with each between-place flow broken down by a set of population characteristics and (for data on travel to work and to school) also by mode of transport. Applications of these data include studying whether migration is leading to greater mixing or segregation of people by aspects such as age, ethnicity, employment and occupation, how much occupational groups vary in the distances over which they access work, how reliant each place of employment or residence is on its workers travelling by car as opposed to forms of transport with a lower carbon footprint and – by comparison with the results of previous Censuses – the extent to which these have altered over time. In addition, users are eagerly awaiting new ‘mover’ data sets from the 2011 Census on the whereabouts of people away from home on Census night and on secondary residences.

10.

The suite of datasets known collectively as the Samples of Anonymised Records (SAR) have greatly extended the value of the Census since 1991. Besides allowing users to readily extract alternative crosstabulations, it also permits them to undertake micro-level modelling. The huge benefits of this approach are well rehearsed in the document ‘A business case for microdata files from the 2011 Census’ (www.ccsr.ac.uk/sars/2011/documentrs/businesscase.pdf), which has prompted the Census agencies to repeat this exercise for the third Census in a row.

11.

Even more powerful for social science research is the ability of users to track anonymised individuals through the UK’s three Longitudinal Studies (LS), based on linking their Census records together but also through adding information from other sources. The LS for England and Wales was set up after the 1971 Census with the particular aim of using the Census data on people’s characteristics to help understand differences between them in their life expectancy and proneness to cancer and other notifiable diseases, with data on the latter being linked to their Census records. Analyses have been both retrospective in nature (as in seeing what conditions preceded death or illness) and prospective (seeing what changes in work or housing happened after an event such as birth of a child or widowhood). Following the major insights achieved in this work, similar studies have been set up in Scotland and Northern Ireland.

12.

Finally, a significant number of BSPS members have looked back over longer historical periods. Much of their work has examined long-term trends in places’ size and characteristics by comparing the Area Tables of past Censuses, helped by Censuses including many of the same or similar questions. For Censuses taken more than 100 years ago, it is permitted to access the data for named individuals, providing added richness to these analyses as well as proving extremely popular amongst those wishing to trace their ancestry. It should be noted that, for more recent Censuses, great care has been taken to ensure that no individual persons can be recognised from the data made available to researchers, this being achieved primarily by controlling the detail allowed in microdata releases but also including record swapping and other disclosure control measures in the Area Tables and Origin-Destination Statistics.

[2] What impact will the ending of the Census have on social science research? 13.

Impact is dependent on the nature of any replacement, which means that our answer to this question should be read in conjunction with those to the two following questions. As the Census is unique in the combination of geographical detail and information about individuals and households which it yields, its abandonment without any form of replacement (i.e. with researchers needing to rely on existing alternatives) would have devastating impacts. At the same time, the severity of the impact would vary between the major areas of Census usage outlined above in answer to the first question, as set out below.

14.

Much (but by no means all) of the information provided by the univariate sections of the Area Tables could be proxied by data drawn from existing surveys and administrative sources, though as currently in the Neighbourhood Statistics Service (NeSS) data holdings the population coverage, accuracy and geographical detail would not be as high as in the Census and would vary between items.

15.

The majority of crosstabulations provided by the multivariate sections (i.e. the majority) of the Area Tables could not be generated from existing alternative sources, except at the broadest geographical scales (e.g. whole UK, regions) where sample sizes are large enough, because to get meaningful findings crosstabulations have to be based on the individual persons or households rather than by comparing the univariate area aggregates.

 

2 64

16.

Migration is monitored between Censuses, but the international migration data are based on very small samples with correspondingly high error and lack of geographical detail and the within-UK migration data record only sex and age and do not provide information on moves within local and unitary authority areas, so most applications of the relevant ODS would not be possible. For commuting, there appears to be no viable alternative source.

17.

The value of the SARs over any other survey or administrative dataset lies in its combination of relatively large sample size (allowing a fair degree of detail about small groups and areas), complete coverage of the population (including – unlike many sample surveys – people living in communal establishments) and a wide range of characteristics for each person and household (which can all be crosstabulated against each other at the individual level).

18.

The UK’s three Longitudinal Studies grow ever more valuable as data are added from each successive Census so as to cover an increasing length of people’s lives, and it is extremely unlikely that the records for all these LS members could be updated from non-Census sources on a consistent and comparable basis for more than a very few of the current topics, leading to steadily reducing value over time from a final Census point in 2011.

19.

Longer-term historical research by the social scientists of the future will be rendered less viable if the Census is replaced by data collected in a variety of ways which themselves may change over time.

20.

In this context, it is worth noting that the demands placed on the Census, driven principally by central government but also coming from the data needs of local government, academia and the private sector, have steadily increased over recent decades, and this in spite of the development of new surveys, longitudinal cohort studies and administrative data releases. Most recently, the 2011 Census included several new questions which allow more detailed analysis of the migrant population, self-ascribed identity, the impact of student populations and the phenomenon of weekly commuting.

21.

The impact on social science research and on the intelligence which it provides to policy makers and practitioners in both public and private sectors therefore depends very much on whether new data sources can be developed or existing sources can be accessed and manipulated in novel ways.

[3] What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? 22.

The short answer is that currently there are none. A replacement for the Census would need to provide the combination of small area detail and accuracy that make the Census the one essential data resource.

23.

The nearest alternative is the Annual Population Survey (APS), but much of the specialist Census research for policy makers and others simply could not be done at all with the APS due to it not providing robust data for local authorities and smaller areas. The value of the APS is also diminished by it not attempting full coverage of institutional accommodation. Other official surveys have lower sampling fractions and/or less complete coverage of the population.

24.

Countries which have abandoned the Census now rely on a comprehensive population register with a system for updating when individuals move residence. By themselves population registers are not ideal, because it cannot be ensured that every instance of migration is recorded. Additionally, obtaining socioeconomic information for the people on the register requires linkage to data bases using a unique code number for each person (as occurs in Scandinavia) and/or large-scale surveys.

25.

As regards the linking of data bases, this is difficult in the UK because no single unique personal identifier is used in the original data collection. The National Insurance (NI) number could be used to identify families and their dynamics. However, the NI number is not necessarily unique and people leaving the country do not have to notify anybody that they have done so, so there will be numbers which are inactive or used fraudulently. Also, there is no method for linking information about the characteristics of the dwelling an individual is living in (tenure, type, conditions), unless other sources such as the Land Registry and Electoral Register were also required to record NI numbers and introduced ways of regularly updating their information about the dwellings.

26.

For commuting data, there is no viable alternative to the Census at present. An experiment was undertaken by ONS to see whether APS could provide the necessary data on commuting for updating official Travelto-Work Areas (TTWAs) and it was found that this was very far from true. If the Census is not continued this updating will be impossible, leaving policy-makers and many others without a resource they have had

 

3 65

for many decades (in fact, Eurostat is currently discussing the need for TTWA-type definitions in all EU member countries). The Annual Survey of Hours and Earnings (ASHE) has some potential, but only if its sampling fraction could be raised from 1% to at least 10%. Moreover, ASHE does not collect mode of transport, which is a key policy interest these days in terms of the climate change agenda, nor a sufficient range of personal characteristics to better understand trends in commuting behaviour. 27.

A cautionary note also needs to be sounded about the pitfalls of relying on a variety of surveys and administrative sources. These include the fact that, unlike the Census which is specifically tasked to provide a 100% population count and answer a wide range of questions, most of these other sources have no such remit. For one thing, there is normally less control over the quality and consistency of the information collected. For another, if an alternative was built around one or more key administrative datasets, there would need to be legislation – and no doubt funding – to ensure that the relevant data collection was continued even if the body in charge of the source had no further need of such data or was abolished. Hopefully, lessons have been learnt from the effect of NHS restructuring on migration data availability and the decision to withdraw child benefit from wealthier families next year, as well as the recent failure of Royal Mail to play ball with address registers.

28.

If a population register is selected as a replacement for the Census, then its worth must be properly validated against a full Census enumeration, so the Census should not be withdrawn until the alternative is shown to be of a sufficiently high standard for both policy making and research. It took Sweden some 20 years to bring its register-based data system up to the required standard, and ONS’s Beyond 2011 briefing is suggesting a similar time frame. The latter assumes considerable investment of resources in dataset development over and above the funding needed for continuing the Census to provide this benchmarking.

30.

If some form of population register is adopted as an alternative to the Census or as a supplement to a reduced-length Census, then it is vital that regular downloads of the data are stored in a safe but userfriendly setting in order to enable studies of change over time and to permit historical research by future generations.

[4] What other existing sources of population and socio-demographic data could be improved upon? 31.

The other sources of population and socio-demographic data in the UK are limited and each has its own weaknesses as well as strengths, so improvements should be sought in all of them, whether the Census continues or not.

32.

The most reliable element of the UK population data is the registration of births and deaths which is very accurate and underpins the annual estimates of population made by the ONS and the Registrars-General for Scotland and Northern Ireland. It would be possible to add more questions about the parents or the deceased.

33.

Most of the information on the socio-economic characteristics of the population between Censuses is derived from ONS social surveys. The main drawback of these surveys is their small sample size, which means that reliable and non-disclosive data for smaller geographical areas and sparsely populated rural areas cannot be generated. Much larger sample sizes would improve this situation, but the costs involved in data collection would be substantial.

34.

All surveys suffer similar drawbacks to the Census in not asking all the questions that are potentially of interest to data analysts while becoming regarded as intrusive by a section of the population which will not co-operate. Hence all social surveys are experiencing declining participation rates and are biased in (often unknown) ways. The 2011 Census made great improvements in response, but this involved substantial expenditure to achieve. However, it was undoubtedly cheaper in terms of expenditure per head than each wave of a social survey and it has yet to be shown that a Census is more expensive over a decade than an alternative that provides a similar or better intelligence.

British Society for Population Studies 30 November 2011

 

4 66

Written evidence submitted by Dr Jennifer Mindell (Census 23)

Declaration of interest  Dr Mindell is an academic  public health physician who regularly uses Census data.  She  and her  team at UCL work on the Health Survey for England, the UK‐wide National Diet and Nutrition Survey,  and the Scottish Health Survey, each of which uses Census data for sampling and for weighting the  data to increase the representativeness of the data, and for age‐standardisation.    How do social scientists use Census data?     1. When conducting surveys on a sample of the population, social scientists make extensive  use of Census data, to ensure that their results are representative and hence generalizable  to the population as a whole.  This enables social scientists and others to gross up  percentages to give population numbers, and to use the results for planning policies and  spending.  The census data are used when selecting the sample, assessing the  representativeness of the participants, and when weighting the survey estimates to increase  representativeness, as well as to judge the extent of bias through differential non‐response.     2. When conducting household based surveys, social scientists often use a stratified sample to  better ensure a representative sample.  Using strata ensures that the sample chosen will  closely resemble the population along the stratifiers chosen.  For example, that the  proportion of households sampled in areas of low, medium and high proportions of minority  ethnic groups, match the proportions of households within the target population.  Census  data on areas are often used for this purpose.  Other important stratifiers include, for  example, socio‐economic position (e.g. the proportion of households with a head of  household in a non‐manual occupation).    3. Survey data are often weighted to correct for imbalances in the responding sample, whether  the data are collected by household interview, telephone, internet, or by post.  Such  imbalances may occur as a result of the selection criteria (for example, a survey that places a  limit of 2 adults per household to be included in the survey will therefore under‐represent  adults in 3 or 4 adult households, compared with 1 and 2 adult households).  More  commonly, such imbalances may occur as a result of response bias, whereby certain types of  people are more likely to be contacted and more likely to agree to take part.  To detect, and  correct these imbalances, population estimates provided by ONS on the basis of the Census  are used.  As a result, the survey estimates, and any policy or spending decisions based upon  them, are more accurate.       4. Due to cost constraints, surveys are limited in size, which then limits the size of geographical  area that they can represent accurately.  In large scale household surveys, the nine former  government office regions or 10 strategic health authority areas are commonly the smallest  geographical area presented in the results. Using accurate population figures and  socioeconomic characteristics of smaller areas, available from the census, researchers can  use synthetic estimation to ‘fill in the gaps’ and provide small area estimates, which enable  more detailed planning in smaller local areas.   

67

5. An additional key use of the Census is to provide the definitive population structure by ae  and sex required for age‐standardisation.  This is undertaken to enable comparisons  between groups that are not confounded by age.      6. Surveys face similar limits on the number of categories that their estimates may represent  along other variables, such as income (often analysed by quintiles), deprivation level (often  presented in tertiles or quintiles), education level and so on, and the census data offers the  opportunity of synthetic estimation in the same way.     7. Census data are of course used to provide population level estimates of many variables, and  these too can be used by social scientists conducting other surveys.  For example, health  surveys using the question ‘how is your general health’ can compare their estimates with  those generated in the census.  This may illuminate other biases in who responds to surveys  where participation is voluntary, even after adjustment for demographic and socio‐ economic factors, and also how responses to the same question may differ depending on its  context within a survey labelled ‘health’ and a census gathering general information.    What impact will the ending of the Census have on social science research?     8. Social scientists working on surveys with a sampled population will no longer be able to  sample and weight their surveys to the same degree of accuracy and representativeness.   Less confidence may therefore be attached to the results given, and policy and funding  decisions may have a less robust platform of evidence.     9. Age‐standardisation and other forms of adjustment for age will become increasingly  inaccurate, resulting in misleading findings in social science research.  While mid‐year  population estimates are used, the degree of re‐estimation of these when a subsequent  Census is published demonstrates how crucial the Census is for accurate population  statistics: even with recording of vital events and of inward and outward migration, these  estimates become increasingly inaccurate as time passes sine the previous Census.      What alternatives to the Census would provide population and socio‐demographic data of  equivalent or higher quality?     10. The unique value of the census is its coverage levels of the population, and its bringing  together basic information on several core topics (demographics, population movement,  work, travel, health, education).  Individual surveys cover samples of the population, and  suffer greater levels of non‐response.  In addition, individual surveys focus on different  topics.  Administrative sources of data, such as post‐office address files, health records, tax  records, each cover only certain sections of the population (those registered with a GP,  those with income) or only certain pieces of information (the post‐office address files for  example only enumerate households, not individuals, and are subject to errors such as  empty properties being counted, or properties with more than one household).      In short, we believe that a composite of other sources of data would be unlikely to have the  same accuracy and completeness that the census offers.     What other existing sources of population and socio‐demographic data could be improved upon? 

68

11. It is difficult to think of other sources of data that could replace the Census.  It has been  proposed that the Census is unnecessary because of other sources of routine data and  bespoke surveys.  This completely misses the mark.    a. First, no other data source covers the entire population (excluding, admittedly,  rough sleepers).  All other data sources are limited to certain sub‐groups, such as  school aged children (the Pupil Level School Census); hospital records (people  attending hospital); GP data (those registered with a GP – and individual level data  are available from only a portion of GP practices);   b. Secondly, it is not that the Census is unnecessary because of other data sources but  on the contrary other data sources are useful only because of the information  supplied by the Census that forms the base on which all other population level data  depends, as described above for demographic and socio‐economic data.    c. The Census provides data that are simply not available through other means.  For  example, the Index of Multiple Deprivation (IMD) 2007 comprises some information  that is updated frequently, such as unemployment figures, but others that are based  on Census data, such as housing without central heating, overcrowding, and adult  skills levels.  12. The Census would be less necessary if the UK introduced an individual‐level population  register, as is held in most other European countries.  A population register that contained  the name, sex, date of birth and address of all residents and that was updated annually  would fulfil some, but not all, of the essential functions of the Census.  However, this may  not be considered politically acceptable in the UK.    Dr Jennifer Mindell  Clinical senior lecturer  Research Department of Epidemiology & Public Health  UCL (University College London)    29 November 2011 

69

Written evidence submitted by NatCen (National Centre for Social Research) (Census 24) Introduction 1. This note sets out a response from NatCen (National Centre for Social Research) to a request from the Science and Technology Committee for submissions on the Census and Social Science. 2. NatCen is Britain’s leading independent social research organisation. We are a not-forprofit business, driven only by our mission to ensure that policy making is informed by a rigorous scientific understanding of a wide range of social issues; to lead on research innovation; and to share our knowledge with the wider research and policy communities. 3. The remainder of this note addresses the four questions posed by the Committee. How do social scientists use Census data? 4. Census data are used in a variety of ways by social scientists. Of key interest to our organisation is the use of census data in the design, conduct and analysis of UK population surveys. We would like to draw the Committee’s attention to five key uses: a. Census data are used when designing surveys of particular sub-groups that are recorded by the Census. An example of this is surveys of BME groups (described as ‘ethnic boosts’), for which Census data are used to identify relatively small geographical areas (e.g. postcode sectors, wards, MSOAs and LSOAs) within which higher proportions of residents from BME groups live. This means that fieldwork can be concentrated in areas which are likely to generate a larger number of interviews per issued address, which makes the fieldwork much more cost efficient. b. In addition to being used for over-sampling specific sub-groups for social surveys, Census data are also used for sample stratification; in other words to select a sample in such a way that it is representative of the population for geographical measures such as region. If other measures that are related to the topic of the survey are also used to stratify (order) a sample before selection, this can help to improve the precision of survey estimates and minimise the risk of selecting a sample that is not representative of the population. Most largescale social surveys use Census data when selecting the sample. For example, the National Travel Survey (commissioned by the Department for Transport) is stratified by Census estimates of the proportion of households that own or have a car available for use, and the population density of the area derived from Census population counts. c. Census data are used in analyses to investigate the impact of the area in which someone lives on other measures collected from social surveys. NatCen regularly receives requests for Census variables to be added to our survey datasets to provide context and to enable particular research questions to be explored. A recent example involved adding data on religious composition of local areas to British Social Attitudes data in order to explore the influence of this on individual religious beliefs, values and attitudes. 70

d. Synthetic estimation is an approach that has been used to obtain area-level modelled estimates for a range of measures, for example, to estimate health lifestyle measures such as the prevalence of smoking or binge drinking. These are generated by observing the relationship between individual-level survey estimates (e.g. for smoking status) and accurate external area-level measures, a large number of which are obtained from the Census. This relationship is then used to predict the prevalence of smoking at the area-level. The more accurate and varied the area-level measures available, the more accurate the predicted estimates. e. Finally, census data are used to weight survey data to make them representative of the population of interest. Most large-scale social surveys use mid-year population estimates for calibration weighting to make them representative by age, sex and Government Office Region. This is an essential step that reduces potential sample bias caused by any differential non-response between age/sex groups and across regions. Area-level Census data (such as population density, proportions of ethnic minorities etc) are also used in nonresponse modelling to help reduce non-response bias. And as survey response rates are likely to remain a challenge, trust and confidence in survey estimates will depend heavily on the quality of the population data being used for weighting What impact will the ending of the Census have on social science research? 5. If accurate data on ethnicity were not available, then it would not be possible to carry out ethnic boosts for surveys without them being prohibitively expensive. No other source of population data is available for the relatively small geographic areas with the required level of accuracy. 6. Ending the Census will impact on how samples are stratified and weighted, potentially increasing sampling error for surveys. It will mean that the stratification variables will need to be changed for most of the UK’s large-scale social surveys. It would also be likely to compromise the accuracy of mid-year population estimates used for calibration weighting, as the time lapse since the last accurate population count increases. 7. Census estimates are a key component of synthetic estimation and so ending the Census would impact on the range and accuracy of modelled estimates that could be produced. It would also reduce the range and accuracy of area-level estimates that could be included when modelling the impact of area-level characteristics on individual measures. What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? 8. A rolling large annual survey could be used to replace the Census. The sample size each year would need to be much larger than the current Integrated Household Survey (IHS) and if the IHS model was used it would be necessary to expand it to include the parts of the population that are not covered at the moment (for example, people living in institutions). 9. However, data from such a source would not be as accurate as Census data. All surveys are subject to non-response error. The Census has a response rate of over 95%. The response rate for the IHS, for example, was 65% for the 2009/10 survey. 71

10. The risk is that, in the absence of the Census, population data would become increasingly out of date as the years pass. This will make it increasingly difficult for social surveys to provide accurate estimates and explanations of social phenomena, as the information bedrock on which they are sampled, weighted and analysed becomes weaker. Therefore, in our view, even a very large survey based method would not provide the same quality information as the Census. 11. An alternative method would be a population registration system against which other data sources, such as administrative records and survey data, could be matched. However, substantial improvements would have to be made in the scope, completeness and quality of administrative records in order to create a reliable population registration system. Also issues related to public acceptability, data security and data access would have to be addressed. What other existing sources of population and socio-demographic data could be improved upon? 12. The IHS could become the main source of population and socio-demographic data, but the annual sample size of 420,000 would have be to be increased to be able to obtain estimates that were precise enough for analysis purposes for small areas (e.g. MSOAs), even if several years worth of data were combined. 13. The sample size could be increased by including the core module in other large-scale social surveys so that the data could be combined. This would require collaboration between ONS, and other survey commissioners or survey organisations. However, the limitations of such an alternative, raised in paragraphs 7-10 above, would still be valid. 14. As noted above, the scope, completeness and quality of administrative records and survey data would have to be improved if these data sources, in combination with a population register, are to replace the UK Census. Declaration of interest 15. The key interest to note is that NatCen is a key provider of social survey data to government – conducting studies like the Health Survey for England for the Information Centre for Health and Social Care, the Family Resources Survey for the Department for Work and Pensions and the National Travel Survey for the Department for Transport.

Kevin Pickering Head of Statistics NatCen (National Centre for Social Research) 30 November 2011

72

3

Written evidence submitted by Tees Valley Unlimited (Census 25) 1

Tees Valley Unlimited is the Local Enterprise Partnership covering the boroughs of Darlington, Hartlepool, Middlesbrough, Redcar & Cleveland and Stockton-on-Tees. In addition to our role in understanding and regenerating the local economy, we also work in partnership with our borough councils on the Census, and provide research, expertise and analysis of the Census and other sources of socio-economic information to them.

2

The fundamental requirements of the Census by Local Authorities are: • Accurate local authority population estimates as a key element of financial settlements. • Accurate estimates of particular populations eg children of school age and pre school age, the number of elderly, the number suffering from ill health • Detailed information on local areas to inform needs and priorities, and to find out the individual needs of particular areas. The sorts of information that local authorities find useful at local level include single parent households, old people living alone, housing data (including overcrowding), unemployment, qualifications of residents, health issues, and employment and commuting data.

3

The Censuses have provided an important role in providing this information, and the scope and detail available in the Census has steadily increased to match demand and ability of new technology to process it. Even when the Census process has been flawed, the data produced by the Census was extremely valuable. Until recently, the Census was almost the only source of data for local areas. Whilst some of ONS’ alternative proposals involve some form of a Census but carried out on a different basis, others anticipate the replacement of the Census by the use of alternative, pre existing data sources. Given our reliance on the Census, we therefore view any proposals to cease undertaking the Census with concern – if replacement arrangements provide less data and / or data of inferior quality, this will have an negative impact on Local Authorities. Potential Benefits of Replacing the Census.

4

We are not blind to the possibilities of benefits arising from alternative arrangements. One of the drawbacks of the Census is that data is only published every 10 years, and the extensive processing of the data means that it normally takes around 2 years for each set of data to be published. This means that Census data becomes out of date very rapidly, and the time between Censuses is long, making monitoring of progress over short periods using Census data effectively impossible. The undertaking of more frequent surveys, or use of administrative data on a regular basis could

73

give us more frequent results, with smaller timelags. In addition, use of other data sources means that there is also the potential for publication of data not currently collected by the Census such as data on income and taxation, and second jobs. Challenges for Alternative Data Sources to the Census 5

Administrative datasets are complex entities, compiled for their own specific purpose and not primarily intended to supply statistics. They are liable to changes through time as the rules and procedures of the underlying process evolve. There are likely to be inconsistencies between data sources, in definitions, in coverage and completeness. Some datasets under-record populations (eg the electoral register) whilst others overrecord (eg GP Patient lists). Data on benefits is a useful source of data, but take-up rates will vary and will never be 100%. Ensuring consistency between data sources would be very important for reliable results, but this would be very difficult and time consuming.

6

There may be a greater reliance on sample surveys, though these have many problems, especially of poor response rates. To make matters worse, these vary widely between different socio-economic groups. Furthermore, sample surveys inevitably have a degree of error in their results, which becomes compounded as smaller areas and smaller populations are examined.

7

ONS has experimented with producing modelled outputs for various indicators for small geographies, where the data is not normally collected (eg Average Household income estimates for small areas). In this technique, the data can be collected by a relatively small survey, and the results for small areas estimated by their socio-economic characteristics. This sort of estimate can on occasion be useful, but they are reliant on accurate socio-economic data used in the modelling process. Without accurate base data (eg from the Census), such modelled estimates will be even more problematic. Research Capabilities at Risk

8

ONS are of course yet to make firm proposals as to what arrangements could replace the current Census, so we cannot evaluate specifically the impact of the new proposals on the work of local authorities. However, the following paragraphs list key outputs from the Census (as currently produced), which would appear to be difficult to replicate by other means. The loss of any of these would be a major concern and would limit research that can be performed.

9

Data for very small areas The most unique ability of the Census is to provide detailed information for small areas – not just for standard geographies like Wards or LSOAs, but for even smaller areas (called

74

Output Areas in 2011) which can be combined to create data for neighbourhoods or other custom areas, and used to produce data for revised administrative areas when boundaries change. Furthermore, the Census can provide detailed information for each output area, not just a few counts. The only reason the Census can supply this level of detail is its near 100% sample size. Apart from the Census, we are only aware of a tiny number of data sources currently available for Output Areas, and these only give a few pieces of information for each area. We feel that the level of detailed information from the Census is invaluable, but will be at risk under proposals to replace the Census, since it is hard to see how this level of detail can be replicated by other means. 10

Commuting and Migration Patterns The Census currently provides the only practical and reliable set of data on patterns of commuting and migration, and the only source of data on this at all for small areas. This data is obviously important to study travel patterns and for transport planning, and contains valuable information on mode of travel and type of job as well as the journey itself. ONS did release some alternative data on commuting from the Annual Population Survey, but this was only available at Local Authority level, and the sample size was insufficient to provide reliable estimates for smaller flows. It is not easy to think of a replacement source of local commuting data, and so this valuable resource could be at risk if the Census is replaced.

11

Sample of Anonymised Records and Longitudinal Study data sets are further unique features of the Census, where sets of sample records can be released to researchers allowing more detailed analysis and social, economic and health characteristics to be tracked from one Census to the next. These are specialist areas but are valuable resources, which are likely to be at risk if there are no further Censuses.

12

Data on Ethnicity – there are still no alternative sources of data to the Census on ethnicity below local authority level

13

Household Composition We are not aware of any other datasets that can provide such detailed information on households and their composition. Since families structures are becoming more complex and diverse, retaining a good source of information to understand this issue will be important. Conclusions

14

The Census is a vital source of data for local authorities both for its role in providing accurate population counts and to provide detailed data for small areas. We feel that the current format of the Census should not be abandoned until suitable alternative(s) can be defined that provide as much of the current functionality as possible and are proven to be effective.

75

15

There is a strong case for running a 2021 Census in parallel with the proposed alternative solutions to ensure consistency between the results and to prove the effectiveness of the new procedures.

16

If administrative records are to be used as an alternative source of data, there needs to be significant work undertaken in order to enable them to provide the consistent and reliable data that is needed.

17

There are many important areas of the Census where alternative solutions would appear to be very difficult to achieve, notably the provision of detailed information for very small areas and commuting and migration patterns. The loss of this data would have a significant impact on much research work, especially by local authorities.

18

Whilst the Census has a substantial cost associated with it (approximately £500m for the 2011 Census), ONS are optimistic that the Census will be of significantly better quality than in 2001 and it seems likely that the money spent will produce a very high quality, detailed, reliable set of data. This can be compared with the recent Regional Fire Control Centre scheme where approximately the same amount of money spent produced nothing. The costs, work needed and difficulties involved of producing alternative solutions to the Census should not be underestimated. We feel that a careful analysis of the costs and benefits of any proposed replacement solutions should be undertaken, and that these should compared to those associated with continuing to hold a Census.

Tees Valley Unlimited 30 November 2011

76

Written evidence submitted by the Economic and Social Research Council (ESRC) (Census 26)   Preamble  1. For over 40 years the Economic and Social Research Council (ESRC) has been at the forefront of  developing the UK’s social science data infrastructure, supporting the creation of world leading  data resources and major, pioneering data services which have played a leading and  longstanding role in shaping the international data landscape for social and economic research.   The ESRC works closely in collaboration with a wide range of stakeholders across government  and elsewhere who both produce and fund access to social and economic data resources to  develop the UK’s data infrastructure.  The National Data Strategy, led by the ESRC, provides a  critical framework for identifying future data needs for the academic, policy and practitioner  communities. 2. The ESRC Delivery Plan 2011‐2015 (http://www.esrc.ac.uk/news‐and‐events/news/13751/esrc‐ delivery‐plan‐2011‐2015.aspx) reiterates ESRC commitment to support strategic investment in  core data infrastructure, including the ESRC Census Programme.  This Programme provides  academic access, expert user support and the representation of academic census user  requirements to the census organisations through a network of data support units and a co‐ ordinator. The ESRC has invested extensively in its Census Programme to support data access  services, expert support and training, research and development to underpin social science  research using UK census data.  This investment indicates the widespread and enduring  importance of the census datasets as a key source for understanding British population and  society.  At present, over 20,000 registered users in UK Higher and Further Education take  advantage of these services.    3. In the preparation of this submission, the ESRC has received input from members of its Policy  Committees, key investments, and a from its Census Programme investments via the Census  Coordinator, Professor David Martin.    How do social scientists use Census data?

4. Social scientists use census data for a diverse range of purposes and it is impossible to articulate  one single position in response to this question.  Many social scientists are specifically  concerned with major thematic research questions in fields such as migration, ageing, health  and social circumstances.  These researchers use a range of research data sources, of which the  census is one very important source.     5. The Census serves as the definitive source for describing the health and social circumstances of  the contemporary population, particularly for relatively small population groups. For example,  the Census is the only data source that allows us to examine the characteristics of specific ethnic  minority groups, occupational groups, or residents of small geographic areas smaller than  electoral wards. Census data, including the Samples of Anonymised Records (SARs), can serve a 

77

source of data for detailed research in their own right or can be used to substantiate or inform  findings arising from sample surveys.    6. Social scientists are often concerned with understanding long‐term societal change, which is  only possible by the comparison of census data over multiple decades, and historians in  particular have developed unique insights from individual census responses which are released  after 100 years, allowing analysis of life in the late nineteenth and early twentieth centuries.   With the end of the census such comparative study will no longer be available to future  generations.  Census data allows social scientists to study change over time at a population  level, for example in access to basic amenities or household structures.  A particularly unique  strength of the current census is that its near‐comprehensive population coverage in a single  data collection exercise facilitates a unique range of integrated data outputs and cross‐ tabulations.      Example 1 from Academic A  In order to obtain rates of births, deaths, marriages etc., counts of events (such as births) obtained  through the vital registration system are combined with denominators of the population at risk of  experiencing this event – for example,  women of childbearing age. Of particular importance is the  need for the denominators to be broken down by detailed attributes ( social class, ethnic group,  country of birth or by small geographical area). The census is currently the most reliable source of  such data. For example, the ONS, only publishes age‐specific fertility rates by country of birth for  census years, where they can be more confident of knowing the correct population by sex, age and  country of birth. In my own social research, I rely on these census‐based population estimates: i) to  investigate fertility rates of migrants by country of birth; ii) to map and explain spatial trends in  teenage pregnancy in Welsh local authorities over time; iii) to compare local geographies of family  size in Britain and hence understand and predict demand for school places.   

 

7. Whereas other social surveys may ask for fuller details of housing (for example), no other source  provides the ability to tabulate housing and other data for small geographical areas.  Similarly,  there is currently no other source which permits so many different variables to be cross‐ tabulated and integrated for small areas.  The census currently permits creation of multiple  output data products, essentially aggregated data for small areas, geographical boundaries for  those areas, flow data representing travel to work and migration flows between areas, samples  of anonymised records and a link to the ONS Longitudinal Study (LS).     

78

    Example 2 from Academic B  The census covers the entire population, both those living in private households and those in  institutions. It therefore allows us to look at the size and characteristics of sub‐groups of the  population who are not usually represented sufficiently in sample surveys e.g. those in institutions  and those who are too small in number to be represented in a sample survey. In looking at family  formation among minority ethnic groups I have relied on micro‐data samples from the 1991 and  2001 censuses – the SARS and CAMS. Whilst summary statistics can be gained from large social  surveys such as the Labour Force Survey, it is only by using samples from the census that we have  enough sample size to look, for example, at the factors associated with inter‐ethnic marriages and  cohabitations, and to compare trends for first and second generation migrants.    8. This richness is only possible because of the existence of a single census database derived from  near‐comprehensive coverage of the population across all census questions, geographically  referenced at the address level within the Office for National Statistics.  Currently there is no  other source of such data.       Example 3 from Academic C  Social research is often interested in the role of local area characteristics in affecting individual  behaviour. Currently, census data, alongside other information, are used to construct local area  classifications, e.g. to assign each small area with indices of deprivation. These local area  classifications and indices can then be linked with individual‐level survey data to identify possible  contextual effects. In my own research we have linked local area classifications and information on  local housing markets to individual level data from the British Household Panel Study in order to  assess the impact of local area unemployment rates and housing markets on the ability of young  adults to move out of the parental home and make the transition to residential independence.    9. It should be noted that census data also provide a key input to social science teaching at  undergraduate and masters levels, with the aggregate outputs in particular forming a focus for  many courses, both in data analysis and in substantive fields such as demography, sociology,  public health and population geography.      Example 4 from Academic D  Longitudinal data allow research to move beyond cross‐sectional associations to examine potential  causal mechanisms. We have in the UK some excellent longitudinal data sources for social research,  such as the Birth Cohort Studies and the British Household Panel Study. However, these do not  contain sufficient numbers of particular sub‐groups( ethnic minorities; recent migrants) to be able to  say a great deal about the wellbeing and life course experiences of minority populations. This may  change in the future e.g. if the ethnic boost within Understanding Society is successful and these  minority groups are retained in the study. Linked census data that make up the ONS LS provide a  unique opportunity to study e.g. the longitudinal determinants of health and wellbeing of ethnic  minorities and the dynamics of family formation among recent migrants. Recent research that  colleagues and I have completed uses the ONS LS to examine the childbearing patterns of recent  migrants and hence address part of the question as to extent to which recent increases in the birth  rate in Britain has been caused by recent increased migration.   

79

What impact will the ending of the Census have on social science research?

10. The implications are heavily dependent on the nature of the replacement data sources.  Implicit  in the question is also an assumption that census data can continue to be collected and  published at comparable levels of quality and detail as is presently the case, and there are good  grounds for believing that this may not be the case.  While methodological understanding and  development are very important themes, the majority of social science researchers are  fundamentally concerned with the nature and quality of the data available to answer key  substantive research questions over and above the mechanism which has been employed to  collect and publish these data.  However there are concerns within the social science  community that the end of the UK census without a firm proposal for a viable alternative is likely  to have a negative impact on the quality of social science research in the UK.     11. The most important potential implications of ending census data without an adequate  replacement relate to the loss of explanatory power which comes from the integrated nature of  the census.  Census data allow, for example, the cross‐tabulation of qualifications and  employment status at small area level, facilitating important insights into the relationships  between education and life chances.  Non‐census replacements which may offer independently‐ derived counts of qualifications and classifications of jobs cannot be employed to adequately  answer these questions unless there is integration of data at the individual level.   The ESRC is  proactively working with partners to drive forward use of administrative data through the  establishment of an Administrative Data Taskforce (due to report to Ministers in the summer  2012).  It is hoped that barriers and issues, such as data quality, can be addressed through a  coordinated plan of action.    12. A further concern is that there are some census questions (e.g. ethnicity, self‐reported health)  for which there are no comprehensive alternative sources, hence it would not be possible to  directly continue with some important existing analysis. In some cases it is possible to conceive  of alternative sources for some of these data, but they would take time to develop and are likely  to result in significant changes of definition (e.g. the difference between asking households how  many cars they have available and counting cars by registered address from DVLA records), with  the consequence that many analyses of change would be considerably disrupted.  If existing  large scale surveys are used as an alternative source, there is a question whether they can  continue to provide reliable information, given many social surveys are suffering from declining  overall response rates. Whilst non‐response is an issue for the census as well, non response  rates are far higher on sample surveys were response is voluntary.    13. Another danger is that whereas the definitions and frequency of census data collection are  under the control of the national statistical organizations, many potential alternative sources  are determined by other agencies and policies which are liable to change at short notice  resulting in loss of data, for example, the imminent discontinuation of universal child benefit,  which might have been a potential complementary source of data on counts of children and  families.    What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality?

80

14. Importantly, any alternatives to the Census, whether it is a rolling census, a population register  from administrative sources, a survey based approach, will have to be able to provide similar  information in a timely and accurate manner and provide similar level of coverage and quality.     15. The first results of recent censuses have taken around 18 months to become available and  detailed statistical products of interest to social science researchers take much longer. In some  cases these may not be available until 2‐3 years after the census, hence census data at their  most valuable are often 2‐3 years out of date.  The census is also challenged by complex  household formations, coverage of short‐term visitors to the UK, persons with multiple  residences, and a range of difficult‐to‐enumerate populations, including young men in large  metropolitan areas and students.  In 2011 it appears that many of the response difficulties  experienced in 2001 have been addressed and coverage rates are still high, but at significantly  increased cost and it is currently too soon to examine what biases may have been introduced as  a result of the different enumeration strategies adopted.  The 2001 experience suggests that  once census coverage falls even moderately, it becomes extremely difficult to adequately  correct the population counts by means of statistical adjustment and imputation.      16. In this context, potential alternatives to the census do offer some significant advantages, in  theory.  To provide an adequate population data source for social science research, it would be  essential to develop some form of statistical population register ‐ although this might be created  temporarily in a secure setting by matching existing lists and does not necessitate the creation  of an identity register, against which individual data are continuously matched.  The UK is  undoubtedly hampered by the absence of a population register, or even common personal  identifiers across many of the key datasets.  A population register has played an important role  in other countries (for example Nordic population registers which represent some of the best  sources of data for demographic research, especially when linked to administrative records)  which have moved to administratively‐based alternatives to the census.      17. The matching of individuals and addresses is essential in order to be able to recover the  multivariate cross‐tabulation of data routinely produced by the census and to have adequate  counts for small geographical areas.  ONS is currently conducting a programme of research on  alternative sources, among which the NHS register and DWP/HMRC Customer Information  System appear to be the most promising reference frames.  Also very welcome is the recent  establishment of the GeoPlace national address gazetteer.  A substantial amount of additional  cross‐matching would be necessary in order to overcome some of the current known  weaknesses of these lists in terms of under‐ and over‐counting.      18. Against these lists, a wide range of additional registers could be matched to derive additional  characteristics for example annual school census data, HESA records, benefits data, DVLA  information etc.  No system of matched registers would provide the levels of detailed social  characteristics available from the census but many of these could be included within an  integrated and extended social survey system, which drew on the same address gazetteer as a  sampling frame and was appropriately stratified.  This could in theory address all the current  census themes and more, providing greater responsiveness for the inclusion of additional  questions, but with data inevitably not all available for such detailed geographical areas.  The  quality of many of these sources remains unknown in detail, which is why the current ONS  Beyond 2011 programme of work is to be strongly supported, and is being support by the ESRC.    

81

What other existing sources of population and socio-demographic data could be improved upon?

19. The creation of an integrated non‐census population data system would require deficiencies in  existing data systems to be addressed and a significant expansion in the administrative systems  available to the statistical agencies.  Matching at the individual record level is essential,  potentially based on the NHS register, although there are concerns about selective over‐ and  under‐counting.     20. The current system of monitoring international migration using the International Passenger  Survey would need to be massively augmented by a proper flow of data from the e‐borders  system, which over time would provide the most accurate information on numbers of UK  residents travelling abroad temporarily and long‐term and would also provide the most accurate  figures for overseas citizens within the UK.      21. The Schools Census (Pupil Level Annual School Census pre‐2007) records a variety of information  about pupils in state education in England, including the student’s home addresses. Access to  data from the Schools Census is currently limited, especially fields such as home address  postcode which act as personally identifying. The Schools Census has the potential to generate a  regular series of origin‐destination matrices disaggregated by pupils’ modes of transport  (walking, bus, car passenger, etc) to school.  Differences in questionnaire wording between  Scotland and other parts of the UK have prevented production of these data.     22. The Higher Education Statistics Agency (HESA) record information for students at Higher  Education Institutes, including the postcode of the student’s address (usually, their parental  address) prior to entry to higher education. These have the potential to produce migration  matrices showing the movement of students to universities and colleges, and there are no  obvious alternative sources of such data. The most widely used attribute of the Census  commuting data is the mode of transport to work, and it is this information which tends to be  missing from candidate data.    23. The Annual Survey of Hours and Earnings (ASHE) includes information about employees’  workplace and home postcodes through a questionnaire sent to employers. It does not include  any socio‐demographic information about the individual, nor does it include information about  the usual mode of transport to work.  The current collection model does not allow any  additional personal data, such as mode of transport to be collected, and with the small sample  size (1%) no detailed spatial disaggregation could be made even if this data were collected.    24. The success of the current census is evaluated using an extensive census coverage survey and it  would be essential that a similar survey instrument were used to evaluate the coverage quality  of any alternative system: this would enable users to gain confidence in the counts and also to  understand the patterns of bias and omission which may exist in the new sources.    Economic and Social Research Council (ESRC)  30 November 2011 

6   

82

Supplementary written evidence submitted by the Economic and Social Research Council (ESRC) (Census 26a)   14 December 2011 evidence session You asked if I could send the list of the top ten most heavily exploited datasets used by social scientists, I referred to yesterday in my oral evidence. This is data collected by the Economic and Social Data Service which is the national service ESRC funds to provide access to social and economic data. Those with an asterisk are datasets collected by ONS or other government departments. 1. Census data* 2. Labour Force Survey* 3. Health Survey* 4. British Household Panel Survey 5. British Social Attitudes Survey 6. 1970 British Cohort Study 7. General Household Survey* 8. British Crime Survey* 9. National Child Development Study 10. Workplace Employee Relations Survey   Jeremy Neathey, Deputy Director of Policy  Economic and Social Research Council (ESRC)  December 2011 

83

Written evidence submitted by CURDS (Centre for Urban and Regional Development Studies) (Census 27) 2021 Population Census Declaration of Interest 0

CURDS has undertaken academic and policy-relevant research for 35 years, and the datasets from successive Censuses underpinned much of its geographical analysis. Some of this work was recently part of our contribution to the Spatial Economics Research Centre funded by two government departments plus the national social science research council. In this long-standing contribution to public debate and academic knowledge of CURDS research into spatial dynamics, far less would have been possible without the Census.

[1] How do social scientists use census data? 1

CURDS researchers use UK Census of Population data both directly in secondary data analysis to reveal the spatial patterns and processes at work in the British economy and society, and also indirectly to ‘benchmark’ our own primary research (ie. to validate the coverage by our survey of the total population of an area or group of particular interest). In this respect, research by CURDS is perhaps ‘representative’ of that carried out by many social and spatial scientists. The following paragraphs focus on more CURDS-specific considerations so as not to duplicate the inputs from the learned societies to which CURDS actively contributes.

2

CURDS researchers use the nearest alternative to the Census – datasets from the Annual Population Survey (APS) – but much of the specialist research for policy makers and others simply could not be done at all with this due to it not providing robust data at a sufficiently detailed area level. The irreplaceability of the Census across a wide range of CURDS research stems from its multiple and inter-linked value for • analysing neighbourhood effects (eg. on child poverty), due to its local detail • highlighting the situation of minority groups, due to its ‘sample’ nearing 100% • separating multiple influences on outcomes, due to its wide subject coverage • recognising the role of mobility in social and economic inequalities, due to its unique provision of fine-grain data on migration and commuting.

[2] What impact will the ending of the Census have on social science research? 3

For almost 30 years CURDS has defined Travel-to-Work Areas for the government by analysing each new Census commuting dataset, and now has begun research for Eurostat exploring the possibility of a ‘European standard method’ for defining labour market areas; it would be incredible if the world-leading British research in this field could not be conducted in Britain in future, but this would be the case if the Census was discontinued.

4

CURDS research which relies on Census commuting data (among other datasets) made a prominent contribution to the recent urban and regional policy developments calling for economic and housing policy delivery at the functional economic area (FEA) scale that, for example, is invoked in relation to Local Economic Partnerships. As yet there is no established way to identify these economic ‘places’ in practice,

84

although there is a consensus that the analysis of the Census commuting data is the single most important way forward to such definitions. Most importantly, the patterns observed by such analyses reveal change in the economic self-sufficiency of areas, with the consequence that these analyses need repeating to update understanding: this will not be possible if the key Census datasets no longer exist. 5

With a more academic emphasis, CURDS has been a major user of the Census Origin-Destination Statistics which include data on migration as well as commuting. These datasets are irreplaceable in measuring the flows between each geographical area and every other one in the country, right down to the neighbourhood scale. CURDS research has, for example, shown how migration can alter the social and economic profile of areas through, for example, ‘brain drain’ trends due to the northern cities being unable to retain the graduates from their universities.

6

It should always be remembered that the data collected in 2011 and any subsequent Census will be used for decades or even centuries into the future, just as the data from past Censuses are prized by researchers (along with the ever growing numbers of family historians). If there had been no Census in 2011 it would not be possible now or later to ‘go back and fill that gap’ in knowledge: we have a duty to future generations to collect data on the present, just as we rely on the data collection carried out in the past by our forbears.

[3] What alternatives to the Census would provide population and sociodemographic data of equivalent or higher quality? 7

The idea that there are effective alternatives to the Census is wrong. ¾ It is sensible to test some candidate alternatives, but this must be done without the expectation that at least one can replace the Census: the test should be “can an alternative provide the data the Census currently provides” and not “which is the best alternative, given that Census will be replaced” ¾ None of the proposed alternatives will provide the combination of small area detail and accuracy that make Census data the one essential data resource: the nearest alternative is APS but much specialist research for policy-makers and others simply could not be done at all with this due to it not providing small area data ¾ What makes the Census data absolutely irreplaceable is not the small area detail alone but its combination with a rich range of variables, all of which have been exhaustively evaluated to ensure they are the most important for policy-makers and other users: none of the alternative sources would provide data that can answer such a wide range of key questions about communities (whether these are communities defined in terms of neighbourhoods using small area data, or communities defined by dimensions such as religion, which means for example being unable to answer many questions on the equalities agenda ¾ An experiment was undertaken by ONS to see whether APS could provide the necessary data on commuting for updating official Travel-to-Work Areas (TTWAs) and it was found that this was very far from true: if the Census is not continued this updating will be impossible which will leaving policy-makers and many others without a resource they have had for many decades (in fact Eurostat is currently discussing the need for such definitions in all countries) ¾ The increasing ubiquity of GIS means that the need for small area data will continue to grow inexorably so any move away from providing such data could not be more bizarrely or detrimentally timed

85

¾

8

Much valuable policy-relevant research relies on the high level of consistency between decennial Censuses to understand change over time and this would be lost by shifting to data collected in different ways; a specialist case of this risk is the increasingly valuable Longitudinal Survey whose linkage of Census records provides a unique resource whose value increases with each decade

The only really robust alternative to a Census is provided by a register system such as those in Scandinavian countries where such systems are so familiar and trusted that there is also little public concern at considerable linkage between these systems and many other datasets. However there are considerable public qualms about such systems in the UK (and in fact there is no single unique personal identifier used across official datasets).

[4] What other existing sources of population and socio-demographic data could be improved upon? 9

The key limitation of alternative data sources such as ONS social surveys is their sample size, preventing data being made available for smaller geographical areas. The value of these sources derives from them providing ‘top up’ information between Census years so that aggregate trends can be identified. Reflecting on the way that such surveys were designed to complement the robust comprehensive audit of the national population that the Census provides periodically, it can be seen that the arguments which are made for dropping the Census are wrong. ¾ “Census data ‘out-of-date’ much of the time” – but most analyses for which Census is the only plausible data source are not of rapidly changing trends ¾ “Census is expensive” – averaged over 10 years it is not expensive, and then divided by its myriad uses makes it very low cost: almost certainly the options for replacement will cost at least as much over the decade and will add little by being more frequently available (see above) while being of far less use (see below) so overall of much less cost effective ¾ “Census information is increasingly difficult to collect” – in fact the signals from the 2011 Census are there was less of a problem than in 2001

Prof Mike Coombes on behalf of CURDS (Centre for Urban and Regional Development Studies) Newcastle University 30 November 2011

86

Written evidence submitted by the Welsh Language Board (Census 28) SUMMARY The Census of Population is the only source of information providing full details of the geographic distribution of the Welsh language skills of the population. Sociolinguistic research into Welsh based on samples generally requires Census information concerning the Welsh language for the sample design process and subsequently at the results stage to weight the sample correctly to the population estimates. We do not believe there is currently any other source of information concerning the distribution of Welsh language skills amongst the population which could be utilised for these functions. A version of a rolling Census could provide a viable alternative.

ABOUT THE WELSH LANGUAGE BOARD 1. The Welsh Language Board was established by the Welsh Language Act 1993. The Act was “…to establish a Board having the function of promoting and facilitating the use of the Welsh language, to provide for the preparation by public bodies of schemes giving effect to the principle that in the conduct of public business and the administration of justice in Wales the English and Welsh languages should be treated on a basis of equality, to make further provision relating to the Welsh language…”. 2. The Welsh Language (Wales) Measure 2011 will, with effect from April 2012, abolish the Welsh Language Board and establish the Welsh Language Commissioner.

DECLARATION OF INTERESTS 3. As part of its function to promote and facilitate the use of the Welsh language the Board has undertaken and commissioned research and funded PhD studentships. 4. The Welsh Language Board was represented by one of its officials on the Census Advisory Group for Wales.

HOW DO SOCIAL SCIENTISTS USE CENSUS DATA? 5. Questions about Welsh have been included in the Census since 1891. The distribution of the population able to speak Welsh (and more recently those with other skills in Welsh) has in itself been a recurring theme of research by geographers through the decades.

87

6. Sociolinguistic research (including attitudinal and behavioural research into Welsh) involving the drawing of samples often relies on Census information, frequently both at the sample design stage and subsequently at the results stage. At the sample design stage, Census data permits sample sizes to be determined efficiently to cover the section of interest of the total population. At the results stage, Census data permits the results from the sample to be weighted appropriately so they may be interpreted as reflective of the population. At both stages information about the complete population is required.

WHAT IMPACT WILL THE ENDING OF THE CENSUS HAVE ON SOCIAL SCIENCE RESEARCH? 7. In some respects it could be argued that without the Census information society would have a poorer understanding of the distribution of Welsh within the population than since 1891. Sociolinguistic research could continue but interpretation of the results would be harder as their relationship to the actual population would be unclear.

WHAT ALTERNATIVES TO THE CENSUS WOULD PROVIDE POPULATION AND SOCIO-DEMOGRAPHIC DATA OF EQUIVALENT OR HIGHER QUALITY? 8. Instead of a decennial census covering the whole population at one point of time, a rolling census, covering the whole population in different areas at different points of time might be a viable alternative. As this would ensure that information about Welsh would be updated in some regular pattern this could even be preferable to what some might think would be the more obvious alternative, i.e. a population register, including details of an individual’s skills in Welsh. (The Welsh Language Board is not expressing support for any alternative by noting these two different approaches). The main point is that some means of obtaining complete coverage of the population with some regularity is needed. A sample survey, or a “census” in which participation is voluntary rather than compulsory, could not provide data of equivalent or higher quality.

WHAT OTHER EXISTING SOURCES OF POPULATION AND SOCIODEMOGRAPHIC DATA COULD BE IMPROVED UPON? 9. With the exception of data on school pupils, very few sources provide data concerning an individual’s Welsh language skills. Accordingly many sources could be improved upon.

Welsh Language Board 30 November 2011

88

Written evidence submitted by The British Library (Census 29) I

Introduction

The British Library welcomes the opportunity to contribute our views to the enquiry looking at the impact of ending the Census of Population on social science research. By way of introduction, The British Library is the national library of the United Kingdom and one of the world's greatest research libraries. It supports the UK’s research infrastructure, serving business and industry, researchers, academics and students, world-wide, as well as in the UK. We receive a copy of every publication produced in the UK and Ireland via legal deposit; our collection includes well over 150 million items, in most known languages and grows by approximately three million items a year. The British Library is passionate about providing both physical and digital access to worldclass information where and when people need it. Over 16,000 people use the collections of the British Library each day (on site and online). In the course of a year six million searches are generated by the British Library online catalogue and nearly 400,000 people visit the Reading rooms. We have a specific department for Social Sciences, which works to support and inspire the commissioners, producers and users of social scientific research publications and material, including research monographs, reports, theses, conference proceedings, journals and statistical publications. We also acquire research-level material from around the world, again including statistical information. Our collections include the UK’s Census of Population publications from inception, as well as an extensive collection of official publications, statistical reports covering social and economic trends, incomes data, economic information and material from the major international organisations such as the Economic Community, OECD, World Bank, United Nations and so on. All of this material is used by our readers, who include academic researchers (from undergraduates to postgraduates, post-doctoral fellows and up to professorial levels; government and third sector researchers; and researchers from consultancy companies and other private sector organisations. As well as supporting the social science research community, we are committed to enhancing the public’s understanding of social science. We have an active events programme, including a series entitled Myths & Realities, which looks at major social and economic policy issues. Topics have included crime, migration, welfare and education and many of the speakers have used Census data as supporting evidence. From March to June 2011 we held an exhibition on the Census at the Library: Census and Society: why everyone counts. This included material from the late 18th century to the present day looking at the scope and power of the Census to inform public policy and underpin decision-making. Its content ranged from very early maps of employment distribution, to sermons extolling the virtue of completing the census to detailed information of the distribution of different ethnicities. It was visited by almost 50,000 people. Alongside the exhibition we held a number of conferences and workshops, covering the potential use of the census in research and the value of cohort and longitudinal data (for the research community); on the use and abuse of statistics (for a public audience), and on family history and a historical perspective – broken down by age, sex and religion (also for a public audience). These were all very well-attended. 2

Declaration of interests

No formal interests in the process of running, analysing or producing the Census.

89

3

How do social scientists use Census data?

Social scientists make extensive and multi-faceted use of data provided by the Census. It is the one, single dataset that provides evidence on the characteristics of whole population on a regular basis, across time, and geography. It’s coverage of the whole population provides one major benefit for social scientists: the ability to carry out detailed and, where appropriate, comparative analysis of small areas and neighbourhoods – essential for local service planning, but also providing information on groups often under-represented in even large-scale sample surveys. The Census allows analysis and understanding of the shape of the population on a geographical, regular, and comparative basis; it facilitates understanding of the way household composition, including socio-economic status, age, education, ethnicity, religion, employment patterns, health, travel to work, children and more, all change over time. No other source provides such comprehensive data on the population, and plays such a key role in the development and implementation of both local and central government policy. Even where there are concerns about representation and under-counting, the Census has the ability to provide more reliable evidence than any ad hoc survey could ever do. In fact, the Census is used as a benchmark against which the validity of samples drawn for ad hoc surveys can be measured. While some of the information gathered by the Census can be found via administrative data (which, there is no doubt, should be used to greater effect), there is no substitute to collecting the data consistently and simultaneously from the population as a whole, at the same time. The Census is unique, and provides a unique historical perspective of an evolving population. Some examples of specific uses of the Census follow. 3.1

Local Authorities

Census data are used by local authorities as the basis for local planning decisions about requirements for housing, infrastructure development and services. The Census is also the basis of allocation of central government funding to local authorities. Neil Storer at a meeting of the Royal Statistical Society in October 2009 gave an example of how the Census is used in Camden: “The Census is used for tracking the distribution of the population of the borough, mapping key variations within it (e.g. overcrowding, spread of students across the borough) and profiling areas at a very detailed level. Census information, mashed with local data, is then used to target scarce resources on local areas and communities. (It provides): affordable commissioned tables; national comparisons and rankings; and essential evidence base for creation of the local development framework, the housing strategy, and the economic development strategy. Between 2001 and 2009 Camden received 2199 information requests, of which 37% were answered using Census data. Census information has been used to support: funding applications, area-based initiatives such as Sure Start, policy development, e.g. distribution of green spaces, and service delivery, e.g. defining catchment areas for libraries and post offices.” 3.2

Third Sector Organisations

Census data made available free of charge via the Neighbourhood Statistics service is relied on by voluntary and community organisations, self-help groups and other bodies active at local levels to provide them with a detailed picture of local populations and socioeconomic conditions in small areas. This information is then used to identify need, plan services and support funding bids and campaigns. 3.3

Social historians

Historic census data is used in combination with other sources such as the RegistrarGeneral’s Reports by historical demographers and social historians to:

90



Map changes in population spread over time



Track changes in housing, household composition and employment over time

Census data can be combined with results from other surveys to present demographic change alongside attitudinal change. Examples include family and household composition, the change from manufacturing to service economy, journeys to work, changing household composition, changing housing conditions, and the use of domestic servants (including the more modern, ‘help’). The Online Historical Population Reports website — histpop — is an online resource of almost 200,000 pages of all the published population reports created by the RegistrarsGeneral of (and its predecessors for) England and Wales and Scotland for the period 1801–1920, including all Census Reports for the period 1801–1937, along with ancillary archival material from The National Archives, plus critical essays contextualising much of the material. The whole resource illustrates the changing demographic structure of Britain and Ireland over this period, and for the first time researchers are able to search and browse the entire collection of published pre-World War Two Census and registration material for the British Isles.

4.

What impact will the ending of the Census have on social science research?

As noted earlier, Census data are more complete than other survey data as they capture the whole population at a given point in time rather than based on sampling. One of the Census’s great strengths has been its regularity and coverage. Administrative data cannot substitute for census data as they do not capture the whole population; not everyone in the population uses the NHS or claims social security benefits. Hard-to-find groups (including those who do not want to be found) tend not to use services, nor appear on formal, official registers. They often do not have access to bank accounts, or credit; children may not attend school and so on. Signifiers of age, household composition, even address, may all be recorded differently across different administrative organisations. So, the Census has the potential to provide a greater representation of marginalised groups. Census data thus form a reliable baseline or framework which validates, and is supplemented and enriched by, survey and administrative data. In sum, incomplete administrative and sample-based survey data cannot: • • •

substitute for complete and impartial census data; provide micro-level neighbourhood data required by local authorities and third sector organisations; and, support analysis of long-term trends due to methodological changes.

5. What alternatives to the Census would provide population and sociodemographic data of equivalent or higher quality? User needs would have to be met by a combination of sources including survey data and administrative data collected during the normal operation of central and local government e.g. for education, taxation, payment of benefits, electoral registration, council tax records, local municipal records, births deaths and marriage records, records collected for use in the health service and so on. These are complex datasets, with different coding schemes, units of measurement, geographies and identifiers. We suggest, therefore, that we are some considerable distance from a situation where these sources could replace the unique level of detail, impartiality and completeness of the Census. It is unlikely that administrative data would provide the extensive coverage of the population it obtains as certain groups may not be registered for or access services. It is also a matter of concern that some major

91

surveys are now being discontinued, e.g. the National Citizenship Survey and the General Lifestyle Survey, leading to a significant loss of primary research data for analysis. Conducting surveys in many countries – and certainly in Britain – is proving harder as society changes. Issues that need to be addressed if these are to provide data of equivalent quality to the Census include overcoming the reluctance to fill in surveys, public acceptability as public concern about confidentiality, data security/protection and the ‘surveillance state’ has grown, data processing, storage and data linkage methodologies, estimation methods, and standard concepts/definitions. Examples range from the proposed use of ‘biometric’ data on passports and ID cards, to the vast databases built up from consumer loyalty cards. Most recently, stories of large-scale data loss have commanded news headlines across days at a time. The loss of public confidence in the intentions and competence of those collecting data carries the risk of impacting directly on the amount and quality of data recorded, and the ability of government to make informed decisions based on accurate, reliable statistics. 6.

What other existing sources of population and socio-demographic data could be improved upon?

Moves to improve upon existing sources of population and socio-demographic data would include: • •

• •



Reinstating the major large-scale social surveys as noted above. Instituting additional surveys to cover marginalised groups such as homeless people and asylum seekers. Whatever replaces the Census will need to ensure that “even the most overlooked people in our society are represented so the allocation of public services can be more effectively geared towards the most needy”. Joining up administrative data to ensure that there are no gaps, and supporting efforts to enable government departments to share data more effectively. Ensuring that data is made freely available to researchers outside of higher education through Internet sites such as www.data.gov.uk . It is crucial that access is maintained and links are kept up to date and not broken. Clicking a few other links from data.gov.uk, it appears everything from ONS now sends you to a dead link. Development of Geographic Information Systems (GIS) and other tools for data manipulation

The British Library 30 November 2011

92

Written evidence submitted by the Royal Statistical Society (Census 30)   1. The Royal Statistical Society (RSS) is the UK's only professional and learned society devoted to the  interests of statistics and statisticians. Founded in 1834, it is one of the most influential and  prestigious statistical societies in the world drawing its membership from over 50 countries. It aims  to promote public understanding of statistics and provide professional support to users of statistics  and to statisticians.    2. The RSS Census Study Group provides a cross‐sectoral forum for debate and education among  census users, focusing on the broader methodological issues which should concern all users of  census data rather than on specific sectoral or topic interests.  The group convenes meetings and  contributes to the work of other special interest groups, bringing together census users from across  academia, business, central and local government and the voluntary sector.      3. The Academy of Social Sciences, which exists to be the voice of the social sciences in the UK for the  public benefit and whose membership encompasses 750 individual Academicians and 43 learned  societies (with a total reach of 85,000 social scientists), drawn from right across the social science  disciplines at the academic, policy and practice levels, is glad to add its support for and endorsement  of this submission. The issues that it raises are shared by the wider social science community.    General points    4. The RSS would stress the importance of the Census in providing more than just national figures;  its  value and use depends on a key feature – the ability to provide highly disaggregated  figures (often  for small geographies) that also permit the joint analysis of a range of variables.  The Census is also  resource that is in constant use as a key information source throughout the ten year inter‐censal  period.  Census outputs take two key forms – highly detailed cross‐tabulations for separate locations  of interest and anonymised micro‐data.  Both forms of output must be available, not just to  government, but to a much wider user constituency.  (In addition there is a third form of output in  the longitudinal study.) The current outputs are available to a huge constituency of users outside  government and provide enormous economic benefits.  If the current Census is to be replaced by  something that combines administrative data at the personal level then the issues of access for this  wider constituency need to be addressed as an integral aspect of the evaluation.        5. The RSS supports the “Beyond 2011” project and agrees that it is right to investigate whether an  alternative to the traditional Census would prove advantageous. Nevertheless the traditional census  provides a huge range of information. As well as providing the most vital facts about society at local  and national level, this is a crucial underpinning of large parts of the evidence base for government,  both central and local, for the NHS, and for business and researchers. It is a major foundation block  of the UK’s information system both in respect of the uniquely detailed information it provides and  also as it is the benchmark on which numerous sample surveys are founded and against which they  are calibrated. We believe it provides substantial value for money.     6. Thus it is crucial that alternatives are fully and properly explored. While it would clearly be desirable  to find a lower cost method to replace the census, it would be a false economy to reduce costs if this  resulted in overall much lower information content.  The current ONS "Beyond 2011" programme of  research is to be welcomed as a means of establishing as clear as possible a picture of the genuine  potential for assembling a viable census replacement system.  It is anticipated that any such system 

93

will need to comprise a combination of administrative and survey sources, the latter required to  address important variables which are not collected in any comprehensive administrative systems.    7. The work of the ONS covers England and Wales. Separate studies are under way in respect of  Scotland and Northern Ireland. It would be a major concern if there were substantial differences in  the systems adopted for the different parts of the UK.     Question 1: How do Social Scientists use Census data?    8. Social scientists use census data for an extremely wide range of purposes, reflecting the breadth of  the census as a data collection exercise which is designed to meet the needs of many sectors and  organizations.  It should also be remembered that social scientists are located across all sectors,  with significant census use in central and local government, business and academia in addition to  the large body of social science researchers and students within the academic sector.  Many social  scientists are specifically concerned with major thematic research questions in fields such as  migration or ageing and will use a range of research data sources, of which the census will be one.      9. A particularly unique strength of the census is its near‐comprehensive population coverage which  facilitates a unique range of integrated data outputs and cross‐tabulations which are available down  to a very small area level.  No other single source provides detailed tables for small geographical  areas.   The small area aggregated data provided by the census allow a range of analyses to be  undertaken investigating neighbourhood influences on aspects of life such as educational  achievement, health and crime.  Business users in particular make extensive use of these  multivariate small area data for area classification and to inform decisions about target markets and  site location.  Government users are more likely to be concerned with items such as the assembly of  deprivation indicators and denominator datasets which permit the characterisation of areas to  inform the delivery of services.  These applications demonstrate the enormous importance of the  census as a baseline dataset, with which unique analyses are possible due to its unique combination  of breadth and depth.  The census also provides more specialised research datasets such as flow  data permitting the investigation of travel to work and migration patterns, made possible by asking  place of residence, place of work and place of residence one year before the census and providing  important insights into travel patterns, housing demand and patterns of economic development.     10. A further family of unique analyses is made possible by samples of anonymised records, which  permit users, for example, to create bespoke cross tabulations of census variables and to analyse  household structures.  Although these represent only a small fraction of the census dataset, the  near‐comprehensive census coverage ensures “sample” sizes far greater than even the largest social  surveys.  A further most important research linkage is provided by the ONS Longitudinal Study,  which now contains a sample of linked records from 1971 to 2011 and permits unique insights into  individual life histories, for example understanding the role of past employment and educational  history on health and mortality later in life.    Question 2: What impact will the ending of the census have on social science research?    11. The answer to this question depends entirely on the nature of the replacement data sources.  Public  acceptability will be a major determinant in the success of a potential alternative system although  that issue is not addressed in any detail here.  There is no certainty that census data can continue to  be collected and published while maintaining current quality levels.  In 2001 substantial adjustments 

94

were necessary to the census estimates for some local authorities due to failings in the census  process.  Initial feedback suggests that 2011 enumeration has been successful in addressing these  challenges, but at increased cost and also by using a variety of different data collection methods  (online, post‐back, post‐back following enumerator visits) which may introduce further biases.  It  will not be possible to fully assess the implications of these changes until quality assurance of the  2011 census outputs is complete.     12. Perhaps the greatest risks of ending the census would result from the lack of integrated data which  permit the cross‐classification of different variables, with a resulting loss of explanatory power.   Administrative data bases do not usually provide detailed cross‐tabulations of outcome variables  against socio‐economic and other characteristics. It is also not certain whether alternative sources  will be able to provide all the richer datasets such as flow data, microdata and links to longitudinal  studies.  The extent to which alternative sources will be able to overcome the weaknesses of the  census model, while replacing its combination of geographical and social detail, will be key to  evaluating the impact of ending the census.  There are some census questions for which there are  no alternative sources, or for which alternative sources may take many years to develop for example  questions on health and languages.    13.  While the use of administrative sources may offer the prospect of more timely data, the overall  data infrastructure would become more dependent on the policies of a range of government  departments and organizations whose primary objective is not data collection.  From a social science  perspective this creates a risk that data series may change in unplanned ways and that comparable  datasets through time may be difficult to achieve.  Government departments supplying information  to the statistical system replacing the census should be required to consult with the National  Statistician before implementing major changes to their data collections.    14. Despite this, there is no reason why a non‐census system could not be developed to continue to  create microdata samples and links to longitudinal studies ‐ indeed, the longitudinal studies already  demonstrate much of the power of linked administrative systems. However, the lack of a decennial  census validation would present real barriers to long‐term analyses due to the unknown effect of  such a major break in the methodologies employed.  International experience suggests that it takes  a long time to develop alternative data sources to the point where they can provide replacement  data series of equivalent quality to that which has been achieved by recent UK censuses.      15. Even if an alternative to the Census could be developed which would satisfactorily meet the needs  of social scientists, this will take time.  A main current criticism of the Census is that it is decennial  and that there is a long wait for new data – we will continue to use 2001 data until well into 2013 for  example.  A concern is that if the 2011 census proves to be the last, it may be an unacceptably long  time before a suitable alternative can be developed.    Question 3: What alternatives to the Census would provide population and socio‐demographic data of  equivalent or higher quality?    16. The answer to this question is contingent on the levels of coverage and accuracy which may be  achieved by a notional future census.  Whereas census data typically appear 18 months to 3 years  (in the case of some of the richest research datasets) after the enumeration, the long time period  between each data refresh can be a challenge to many users.  The census continues to be  challenged in terms of overall response rates, difficulties of identifying and gaining access to 

95

properties and a range of hard to reach population groups, including young men, visitors to the UK  and those with multiple residences and complex household structures.  Initial signs are that the  2011 census has achieved a good result in the face of these challenges, but with a considerable  increase in cost.  International experience suggests that the difficulties of achieving near‐ comprehensive coverage will continue to increase. In the 2001 census coverage in some areas fell to  levels which were not amenable to satisfactory adjustment using statistical methods.     17. There are thus attractions to the use of a system of alternative sources which might address these  difficulties more effectively, but it is unlikely that this can be achieved by use of any single approach.  Rather, a combination of methods and alternative sources needs to be evaluated.  It seems likely  that any alternative would need to include some form of statistical population register or database  which would allow the matching of other lists at the individual record level, otherwise it will not be  possible to generate multivariate datasets comparable to those obtained from a census.    18. ONS is currently conducting a welcome programme of research on alternative sources, among which  the NHS register and DWP/HMRC Customer Information System appear to be the most promising  reference frames.  Also very welcome is the recent establishment of the GeoPlace national address  gazetteer.  A substantial amount of additional cross‐matching would be necessary in order to  overcome some of the current known weaknesses of these lists in terms of under‐ and over‐ counting.  Against these lists, a wide range of additional registers could be matched to derive  additional characteristics for example annual school census data, HESA records, benefits data, DVLA  information etc.    19. No system of matched registers would provide the levels of detailed social characteristics available  from the census but many of these could be addressed by an integrated and extended social survey  system, which drew on the same address gazetteer as a sampling frame and was appropriately  stratified.  This could in theory address all the current census themes and more, providing greater  responsiveness for the inclusion of additional questions, but with data inevitably not available for  such detailed geographical areas.  The quality of many of these sources remains unknown in detail,  which is why the current ONS Beyond 2011 programme of work is to be strongly supported.    Question 4: What other existing sources of population and socio‐demographic data could be improved  upon?    20. The creation of an integrated non‐census population data system will require a range of different,  coordinated actions.  Firstly, it will be necessary to identify ways of improving the quality of existing  sources.   For example, reconciliation of addresses from a definitive address register and other  sources such as the NHS register will result in improvements to both.  The NHS register provides the  most promising data source to contribute to population counts and internal migration but it is  subject to recognised weaknesses, including both over‐ and under‐counting of specific population  groups.   Moving to a post‐census statistics system would require issues such as this to be seriously  addressed and supported to an extent which has not previously been done ‐ and also recognition  that complete resolution is most likely unachievable.  Secondly, some existing sources would need  to be augmented by considering the need for additional data collection, for example whether more  systematic ethnicity information could be included in one of the major data series such as the NHS  register.  Developments in other areas may yield specific new variables of interest: for example the  current census question on central heating might in due course be more usefully replaced with  information on energy use consumption from utility records. 

96

  21. Despite recent improvements, the measurement of international migration still remains a major  weakness and the availability of well‐structured data from the e‐borders system would be an  essential element of a workable census replacement system.   A third element would be close  integration of the current social survey system with the census replacement, almost certainly  involving both the integration and expanding of existing surveys so as to provide greater coverage of  socioeconomic detail and the continuation of a substantial coverage survey which would allow both  user and producers to better understand the characteristics of the post‐census population data.   Declining response rates to the major government surveys is already a concern and it should not be  assumed that using larger or more integrated surveys as part of a census replacement will overcome  these challenges without further detailed attention to design and implementation.  Increased use of  internet‐based data collection may offer some advantages in this respect but again its potential role  in a greatly changed data environment requires substantial research effort.      Royal Statistical Society    30 November 2011 

97

Written evidence submitted by the British Academy (Census 31) Introduction 1. The British Academy welcomes the opportunity to submit evidence to the House of Commons Science and Technology Select Committee inquiry on the census and social science. The Census has been an important and highly useful resource for social scientists and historians, enabling them to carry out research, analyse policy development and evaluation social behaviour and mobility. It is vital that, should the Census cease to operate in its current format, its replacement continues to meet the needs of social science research. It may also be worth remembering that the Census has long-run value and can be expected to be used by many future generations of historians and social scientists who wish to understand how, and why, society has changed. 2. Along with other members of the UK Strategic Forum for the Social Sciences 1 , we were concerned when we heard about plans to end the Census in its current form last year. Sir Adam Roberts, KCMG, President of the British Academy wrote to the Right Honourable Francis Maude MP on 30 July 2010 urging that the Census 2011 went ahead with sufficient support to ensure its success. We were clearly very pleased to receive confirmation from Mr Maude in August 2010 that the Census 2011 was to go ahead (as it did in March this year), but we remain worried about how this extremely valuable work will continue in the future. The uses of census data 3. The Census in its current form provides social science researchers with an extremely large sample size, and arguably almost complete coverage of the UK population (because of the legal obligation to respond to the Census). The advantage of this is that it in turn means that social scientists can look at particular sub-groups of the population where there may be specific issues which may require certain policy responses (minority groups such as travellers is one example). This is hugely valuable as it is often not possible to obtain sufficiently large sample sizes from other sources such as the Annual Population Survey (which is derived from the Labour Force Survey, the Welsh Labour Force Survey, the Scottish Labour Force Survey and the English Local Labour Force Survey). 4. Social scientists use Census data for a number of important additional purposes. The data enable them to assess the representativeness of sample survey data that they might use in their own research. Thus, the Census data can be used to provide a yardstick against which to judge the accuracy of the demographic profiles of sample surveys, or to weight sample survey data to ensure they are representative of the presumed national demographic. 1The

Forum brings together a number of funders, commissioners, advocates and users of social sciences to facilitate discussion about areas of shared interest twice-yearly and is hosted by the British Academy. The Forum was established in 2001 to provide an arena to enable the major issues and challenges facing the social sciences to be debated at the highest level, and to ensure the needs of the social sciences are addressed and tackled. 1

98

5. Social scientists can also use the Census in statistical analysis, by using aggregate level (especially geographical) data contained within it, including region, constituency, ward or super output areas (SOA). 2 These data can be used as controls that assist social scientists in properly estimating the relationship between individual-level variables, as predictors of individual-level variables or as predictors of the variations in the relationships between and / or among attitudes and behaviours.   6. Historians also find the Census extremely valuable, as they are able to make excellent use of the long runs of comparable data to study, among other things, patterns of employment and demography over time.   The impact on social science research of ending the Census in its current form 7. Without the data produced by the Census in its current form, social scientists will lose an extremely effective yardstick for judging the representativeness (or otherwise) of their own sample surveys. The overall response rate to the 2001 Census was 98%. 3 We do not yet know the response rates for the 2011 survey but it is reasonable to assume it will remain high, given the legal obligation to complete. This response rate is much higher than that in the best of the other large-scale government surveys, such as the Annual Population Survey and the Labour Force Surveys on which it is based. It is also important to note that response rates for surveys have been falling, and will no doubt continue to do so. For example, the response rate to the Labour Force Survey fell from 80.6% in the third quarter of 1997 to 71.2% in the second quarter of 2006. 4 This suggests that any alternative to the Census that is reliant on surveys may be prone to increasing bias over time. 8. The ending of the Census in its current form would have extremely serious implications. We noted in paragraph 3 the value of the Census in providing data about particular small groups; these data could well be less robust in the future without the Census. Other uses are also important. For example, Census data are used for estimating social composition of constituencies or smaller geographical areas – which is not possible to do from Annual Population Surveys (where data below the level of Unitary Authority is not released). There is a major concern in the social science community about the reluctance of government departments to release data with detailed geographical information because of potential identification issues – the data obtained from the Census enable us to overcome this particular issue (we return to this point in paragraph 10). See the ONS website for more on SOAs http://www.neighbourhood.statistics.gov.uk/dissemination/Info.do;jessionid=C9XqTKsC0p1qCDhhvffJgZP5x C1mQLvbzWMh5jTVz71GgSL06thG!857689500!1321888994656?m=0&s=1321888994656&enc=1&page=aboutneighbourhood/geography/superoutput areas/soa-intro.htm&nsjs=true&nsck=true&nssvg=false&nswid=1280 [accessed 29 November 2011] 3 Data from http://www.ons.gov.uk/ons/guide-method/census/census-2001/data-and-products/quality-ofthe-census-data/response-rates/summary/index.html [accessed 24 November 2011] 4, Labour Force Survey User Guide – Volume 1: Background and Methodology, Office of National Statistics, table 5.3, p33 available from www.ons.gov.uk 2

99

  9. It is true that Census estimates (for example, for constituencies) can rapidly become out of date, hence the need for Office for National Statistics to find ways of estimating mid-Census period adjustments. However, this will only be worthwhile if suitable alternative instruments are put in place to provide information about the changing demographics of the UK public. That is, to replicate (as far as is possible) the functions enabled by Census data discussed in paragraphs 3-6. Alternatives to the Census to provide population and socio-demographic data of equivalent or higher quality 10. We note in paragraphs 2-5 important functions of the Census and some particularly valuable uses for the data provided made by social science researchers and other academics. Any alternative would need to continue to fulfil these functions. 11. To be even more specific, regular targeted surveys (e.g. approximately every five years) could provide a more in-depth study of important population groups (we have already mentioned travellers, but data about young Black men are also less robust than for other groups). Such groups are relatively small, and so sources such as the Annual Population Survey will not provide sufficient numbers of respondents for robust analysis. Even more importantly, such groups tend to exhibit high nonresponse to standard surveys so that survey-based analysis may generate misleading conclusions. Particular arrangements will be needed in order to be confident that government departments, social scientists and other interested parties can be sure that robust data are available. It is very unlikely that administrative data would be able to satisfy this need. 12. The Census is almost certainly less prone to response bias than even the best government and academic surveys. It is true that this is still an issue for the Census (again as examples, coverage of certain disaffected groups like young Black males, or of undocumented residents is less strong) and so any alternative would need a lot of work to identify response bias and to adjust for any such bias. 13. It is not simply a case of alternative surveys improving the way they identify and adjust for response bias 5 (for which we mean both bias in survey response rates and on a question-specific basis within surveys). There would almost certainly need to be supplementary investigations, targeted at particular groups that are believed to be at high risk of non-response, in order to compare their profile with those of respondents to particular surveys (for example, the Labour Force Survey). These supplementary investigations might be expensive, and probably much less definitive than the census in the absence of a legal requirement to respond. There is an important point about adjusting for response bias. While one can in principle use weighting in order to adjust, this assumes that there are no interaction effects with non-response. That is, weighting makes the assumption that young blacks who did respond are similar to the ones who did not. These estimates can then be grossed up. However, they may very well not be similar – for example non respondents may be more likely to be undocumented migrants (e.g. among Black respondents only 5% may be undocumented, but among non-respondents the figure might be 20%, so grossing up using weights could be seriously inaccurate – not that the Census actually tries to ask about legal status of course, but this illustrates the point.)

5

100

14. The Beyond 2011 Programme suggests a number of valuable alternatives to the Census in its current form. It may be possible to achieve high-quality population and socio-demographic data by using a shorter census plus relatively frequent sample surveys to update changing demographic patters between censuses. It may also be possible to achieve high-quality population and socio-demographic data using the administrative data collected by government in a combination with sample surveys. However, it is not clear that either surveys or administrative data on their own or in combination would meet the needs for authoritative data to provide a yardstick for evaluating survey data (given that administrative data may well be subject to unknown biases), or for analysing small groups. 15. We therefore see value in a ‘rolling census’ which, for example, undertook a legally compulsory census on a rolling basis in different geographic areas each year, perhaps revisiting more frequently areas that have faster rates of change/turnover. A census could then be carried out in a different county or borough each year. This might have the advantage of a census plus the benefit of sampling – but in this case, the ONS would be sampling areas for census treatment (stratified sampling). This may also be a less costly option than the Census in its current form. 16. In discussing the use of administrative data, we think it relevant to alert the Committee to numerous current concerns in the academic community about the availability of these data. Social scientists are increasingly finding problems with obtaining access to both administrative and research data held by government departments. The Government Social Research Unit and the departmental Heads of Analysis have promoted good practice, but there still seems to be a lack of clear, consistent guidelines to cover all departments. Instead, there seem to be many obstacles to social scientists obtaining data to support them in their research. The Committee referred to these issues in its report in July this year on peer review 6 as did the British Academy’s response to the Cabinet Office’s consultation on open data in October 7 . Improving on existing sources of population and socio-demographic data 17. As we noted above, data are already available in government departments. An improvement would be to secure, as standard, the open release of all data (suitably anonymised) to archives that grant access to social science researchers and other academics. Protocols for access and use would need to be established, but this would be an improvement on the current situation, and vital if social scientists no longer have the data produced by the Census in its current form. 18. Data on health and education inputs and outcomes that were comparable on a geographical basis (for example by region, constituency, ward, SOA) would be enormously useful for analysing the quality and consequences of policy decisions. House of Commons Science and Technology Committee – Eight Report: Peer Review in Scientific Publications, 2011 7 Available on the British Academy’s website at http://www.britac.ac.uk/policy/policy-centre-reports.cfm [accessed 21 November 2011] 6

101

British Academy submission to the House of Commons Science and Technology Select Committee inquiry into the census and social science – November 2011

19. Regularly collected data on a wide range of social, political and economic attitudes and behaviours would be very useful in informing policymakers about the consequences of their policies. The cancellation of the Citizenship Survey means that there is no longer a vehicle for informing policy-makers about the consequences (if any) of current ‘Big Society’ policies, to take one example. This means, therefore, that the ONS should look at the whole array of government-funded data collection and not treat the Census in isolation. If other relevant surveys are also being cancelled, the argument for maintaining the Census (or establishing something very similar such as a rolling Census) surely becomes stronger. Summary 20. The Census in its current form has enormous value: to government, to other policy organisations, and to historians and social science researchers. Any changes to the Census must ensure that, by and large, the data produced are still able to provide users with valid, reliable and usable information. If changes to the Census result in efficiency savings, then consideration should be given to investing in new targeted research that can investigate some of the issues around non-response and small population groups. Consideration also needs to be given to the issue of bias in nonCensus alternatives such as sample surveys and administrative data. The aim must be to ensure that whatever replaces the Census is at least as effective as the current system.

The British Academy 30 November 2011 The British Academy, established by Royal Charter in 1902, champions and supports the humanities and social sciences across the UK and internationally. It aims to inspire, recognise and support excellence and high achievement across the UK and internationally. As a Fellowship of over 900 UK scholars and social scientists, elected for their distinction in research, the Academy is an independent and self-governing organisation, in receipt of public funding. Views expressed in this submission are not necessarily shared by each individual Fellow.

102

Written evidence submitted by The Association of Business Schools (Census 32)   The Association of Business Schools is a membership organisation representing 125 UK  business schools and related bodies. Our members teach one in seven students in UK Higher  Education and employ over twelve thousand academic staff.    1. How do social scientists use census data?    Within UK business schools, census data is used extensively in the research areas of Health  Economics, Marketing / Consumer behaviour and Economics more generally.    Looking at Health Economics specifically, the use and allocation of healthcare resources is  directly proportioned to age. Mortality rates, incidence and prevalence of common as well  as rare conditions are a function of age, gender and ethnicity distribution of the population.  Those data are only available through census at national level. The link between income and  health is also extensively researched using census data.    2. What impact will the ending of the Census have on social science research?    Research of the kind described above would be very difficult to undertake without census  data. The impact on the evidence base for public policy making would be considerable.    3. What alternatives to the Census would provide population and socio‐demographic  data of equivalent or higher quality?    The lack of census data can only be replaced with large cohort studies, which are lengthy,  expensive and far less significant and broadly applicable than census data.    4. What other existing sources of population and socio‐demographic data could be  improved upon?    See answer to question three.    If you have any questions about our response please get in touch.      Head of Policy development  The Association of Business Schools    30 November 2011   

103

Written evidence submitted by the Royal Geographical Society (with The Institute of British Geographers) (Census 33) 1. The Royal Geographical Society (with The Institute of British Geographers) welcomes this opportunity to comment on the inquiry into the census and social science 2. Given the spatial nature of data collected within the Census, social scientists within geography are amongst those who make greatest use of its data within their research, along with other closelyrelated subjects with a spatial perspective, such as Town Planning and Regional Science. However, spatial analysis and hence use of the Census is increasing in Economics and Sociology. The Society – as the learned Society representing the discipline – focuses on use by the geographical research community. 3. Formed in 1830, the Society’s Royal Charter is for 'the advancement of geographical science'. We are a charity that seeks to develop, promote and support the discipline of geography and its practitioners in the areas of research and higher education, teaching and fieldwork, policy and wider public engagement. The Society has more than 15,000 Fellows and members, of whom a substantial number are academics and other researchers whose work we support through a range of activities. These include holding the largest geographical research conference in Europe, publishing three of the leading international peer-reviewed geography journals in the world (including Transactions of the Institute of British Geographers which is often ranked first), co-ordinating twenty seven specialist research groups, and providing small grants for researchers at all career stages. We work very closely with all Higher Education (HE) geography departments in the UK. 4. In the preparation of this response, the Society consulted with and elicited responses from two of the Society’s research groups (the Quantitative Methods Research Group; and the Population Geography Research Group). In addition we received comments through a number of the Society’s i Fellows . 5.

The key points of the response can be summarised as follows:









Geographers, as social scientists, make wide use of the Census in their research, through aggregation of population data to the areas and regions; flows of population between areas; and examining how these change over time. This includes the understanding of international migration; internal patterns of migration; population estimates and services; public health planning; and the creation of classifications of areas; and transport, including mode of travel to work.   The specific impact of ending the Census is heavily dependent on the exact nature of the replacement population data system, but the main risk would mean deterioration of the information social science research has at its disposal which would compromise Central and Local Government’s ability to make accurate decisions about funding services To achieve an alternative to the Census with data of equivalent or higher quality would need a full investigation of the geographical coverage of the alternatives as it is not clear that any proposed so far provide the same combination of small area detail and accuracy Other existing data sources that could be improved upon include the NHS Patient Register; the Schools Census; the Higher Education Statistics Agency (HESA) record information for students at Higher Education Institutes; the Annual Survey of Hours and Earnings (ASHE); and migration data (based on the NHS register and International Passenger Survey / eborders system data).

How do social scientists use Census data? 6. The census provides a common geographical framework for the integration of data from many sources and its small area counts are essential in allowing: (i) Aggregation of population data to the areas and regions – including mapping, modelling, classification and analysis of small areas such as neighbourhoods. This work ranges from the construction of deprivation indicators and area types to the investigation of 104

geographical patterns of inequality and the relationships (for example) between local social conditions and ill health or life chances, or in providing denominators for studies of disease prevalence or crime rates. This also allows for analysis of sub groups of the population e.g. ethnic or migrant groups; lone parent families; people in institutional accommodation; (ii) Flows of population between areas - for which the census provides data both on travel to work patterns and residential migration. Understanding of such flows is essential to our understanding of daily travel patterns, urbanization, regional development, demand for transportation and housing; (iii) Understanding long-term societal change by comparing census data over multiple decades, and even with the individual responses which are released after 100 years. 7.

Specific examples of research undertaken by geographers using Census data include: (i) Studies of international migration: measuring neighbourhood level population change of ethnic groups. There is considerable evidence to indicate processes of spatial deconcentration are taking place, particularly in London, where the majority of ethnic minority populations are located (ii) Understanding internal patterns of migration: millions of people in Britain change their place of usual residence each year. The Office of National Statistics (ONS) Longitudinal Study enables Census records at the individual level to be linked from one Census to the next, allows these internal movements to be understood. (iii) Population estimates and services: internal and international migration statistics from the Census play a critically important role in the production of annual population and household estimates which underpin the provision of housing and services. (iv) Public health planning: health geographers make extensive use of census data on socioeconomic conditions, demographic characteristics and health related topics and census data are routinely used in public health planning and assessments of service activity and outcomes. (v) Creation of classifications of areas: The Census has been an important source of denominators in the creation of indicators of local disadvantage using more regularly up-dated information sources. These have been developed by the commercial sector to create powerful geo-demographic classification systems used in credit scoring, insurance and marketing and for predicting individual consumer behaviour. (vi) Transport, including mode of travel to work: this is particularly well used by researchers as it is by transport planners, including use to define the official Travel-to-Work Areas (TTWAs) used by various Government Departments. In addition Census ‘flow’ data (e.g. journey-towork) has been used to delineate functional urban regions and city regions, which identify the hierarchy of cities and towns.

What impact will the ending of the Census have on social science research? 8. The Census is an important tool for shaping and evaluating policy in the UK, and is unique in its universal (100%) coverage of the population and in the breadth of subject areas for which information is recorded. It is the only current study which can provide the geographical detail needed to accurately show variations in British society and it is hard to see how attempting to reproduce this coverage by other means would be less resource intensive or expensive. 9. The specific impact of ending the Census is heavily dependent on the exact nature of the replacement population data system. If the alternative provides the same detail of data, but at greater frequency of update, then the data environment for social scientists will have been improved by abandoning the Census. However if, as is more likely, the replacement will be a series of partial products – such as population estimates enhanced by administrative records, combined with local data on housing, economic activity and health based on government surveys, then there is a real risk it will not allow aggregation to a wide enough range of areas needed for research and policy purposes. 10. The design of any census replacement system would need to include a considered strategy for reviewing and updating the geographical basis on which all socioeconomic data are to be 105

published. Further research is required to assess the geographical scales at which data might be produced from an enhanced system of social surveys. 11. The 2011 Census includes new questions which allow more detailed analysis of the migrant population, identity, the impact of student populations and the phenomenon of weekly commuting: further research is also needed on the ability of alternative data systems to capture data such as this – including travel to work and migration flows - for which we currently rely on the Census. 12. The main risk identified is a compromising of Central and Local Government’s ability to make accurate decisions about funding services (including schools and hospitals). For example, unless an alternative data source, based on civil registration of the resident population, is developed it will become much more difficult to make local needs assessments in for health service provision.The census is also important in sparsely populated areas, which are likely to be poorly represented in sample surveys. A very considerable effort to enhance the power of national sample surveys would be necessary to substitute for the census in this respect. 13. There would be further loss in terms of the sources of information compiled in the Longitudinal Study, which is an internationally admired source of data originally based on a sample from the 1971 census. This provides a unique source of information on the currently aging population and other cohort studies have nothing like the same power to analyse local variation in important dynamic ii processes of aging, migration and health outcomes 14. Assuming no satisfactory replacement then this will mean deterioration of the information social science research can use to investigate the well-being of the population, with the real difficulties starting to emerge from around 2022/23. What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? 15. The ONS, NRS, NISRA and WAG iii are all engaged in a project, overseen by the UK Statistics Authority, called Beyond 2011 (2011 to 2014) which aims to develop options for replacing the census iv . Eight options are being investigated: (1) traditional census (e.g. UK Census 2011), (2) short and long form census (e.g. US Census 2000), (3) short form census and continuous survey (e.g. US Census 2010) and (4) rolling census (e.g. France’s current methodology) which uses a very large rolling survey and interpolation and extrapolation methods, (5) address register plus survey, (6) administrative data – aggregate (cf. Index of Multiple Deprivation), (7) administrative data – record level (e.g. Finland) and (8) administrative data - intermediate. 16. For further details of alternatives being considered, the committee is recommended to refer to the National Statistics Beyond 2011 project website v for further details. For example, Option 7 might be based on the reconciliation of existing administrative registers - including the new National Address Gazetteer and population lists such as the NHS and DWP/HMRC lists as the basis for an initial population register. Another person identifier is the National Insurance and the DWP maintains the “Lifetime Labour Market Database” which can trace individuals over time and across space 17. An important consideration in the evaluation of alternatives would be to thoroughly investigate the evenness of their geographical coverage and would be likely to require monitoring via some form of coverage survey, as is used in the 2001 and 2011 censuses, to provide confidence in the coverage of the alternative system across different local authorities and area types. 18. Although the matching of individual records in administrative databases my provide substitutes for some Census variables there are others (eg hours of work or employment status) where it is likely that research will be dependent on existing government surveys (such as the Labour Force Survey/Annual Population Survey) for information. Their existing sample sizes mean that it is not possible to use data from these sources for some smaller local authorities, let alone more detailed local areas. For them to be effective substitutes for the Census their sample sizes would need to be significantly increased – which would involve significant costs. 19. If the UK had a proper population register, as other EU countries do, with a legal requirement on all residents to report their addresses and changes of address and with an universal person identification, then the population spine would be as accurate as it could ever be. This would, as in other countries such as Finland, provide a sensible alternative to the Census. Parliament rejected the 106

project to create a National Identity Card and existing cards have been cancelled vi . Historically, a population register was first proposed (and rejected) in the House of Commons in 1753 vii . The National Health Service Register covers about 98% of the population at present but could be extended to cover all the population by including all individuals not entitled to NHS health care or receiving health care from other bodies. 20. Monitoring of travel and migration could exploit new data sources from traffic and public transport monitoring, together with more determined development of existing systems such as the NHS register for internal migration. Some census questions would inevitably need to be replaced by enhanced social surveys in order to ensure continuity of information on topics not readily captured from any administrative source. 21. Alternative sources of commuting data are very limited and this is an area of acute concern to those who have used the Census commuting data for research. The National Travel Survey (NTS) is collected by the Department for Transport, and has been running on an annual basis since 1988. It is a sample survey, and in 2009 collected data from over 8,000 households. There is very limited spatial information in the NTS, with disaggregation by Government Office Region or ‘area type’ only. Even at this very coarse scale, there is no disaggregation by origin, destination and mode of travel because the small sample size does not permit the publication of detailed disaggregated flows.. 22. None of the proposed alternatives in any of these areas will provide the combination of small area detail and accuracy that make Census data the one essential data resource. It is sensible to test some candidate alternatives, but this must be done without the expectation that at least one can replace the Census: the test should be “can an alternative provide the data the Census currently provides” and not “which is the best alternative, given that Census will be replaced” What other existing sources of population and socio-demographic information could be improved upon? 23. A raft of new data sets are being released by Government Departments and other agencies that will be of enormous interest to the social science community if agreements can reached for the data suppliers to provide social science researchers with access to these data sets. The ‘Beyond 2011’ Programme is a good example of a context in which new and valuable data sets are being assembled and used by ONS. The Economic and Social Research Council must take a pro-active role in reaching agreements with data suppliers to disseminate these data sets, taking advantage of the open government licensing arrangements that are now in place. 24. Researchers highlight that other sources of population and socio-demographic data in the UK are very limited and should all be improved whether the Census of Population continues or not. The most reliable element of the UK population data is the registration of births and deaths which are very accurate and form the basis of the annual estimates of population made by the ONS and the Registrars-General of Scotland and Northern Ireland. It would be possible to add more questions about the parents or the deceased. Attempts have been made to add an ethnic group classification to birth registration, but this is difficult because the concept of ethnicity adopted by official statistics is a social construct and cannot be applied to a new-born child. Recording each parent’s ethnicity might be preferable, but this does not necessarily indicate the ethnicity of the child and may not be possible where the father is absent. 25.

Suggested areas where improvements to existing data sets could be achieved, include: •

Migration data: both internal (based on the NHS register) and international (based on combination of existing International Passenger Survey data with robust counts from the eborders system). Until such time as the latter are available it is difficult to conceive how adequate counts can be produced for international flows. There may be an opportunity to utilise additional data sources from business, particularly relating to energy use, car ownership, communications (including mobile telephony traffic, a potential new alternative for tracking population mobility), financial status and consumption patterns, but these sources are unlikely to be individually comprehensive.



The NHS Patient Register data are captured at the postcode scale, and thus could potentially be published at more detailed geographic scales than is presently the case, but the lack of

107

socio-demographic data in the Patient Register limits its scope for use in social science research. •

The Schools Census – known as the Pupil Level Annual School Census (PLASC) prior to 2007 – records a variety of information about pupils in state education in England. Access to data from the Schools Census is currently limited, especially and necessarily to those fields – such as home address postcode – which act as personally identifying. However, it would be possible for aggregate observations to be produced showing the migration of school children and also has the potential to generate a regular series of origin-destination matrices disaggregated by pupils’ modes of transport (walking, bus, car passenger, etc) to school.



The Higher Education Statistics Agency (HESA) record information for students at Higher Education Institutes, including the postcode of the student’s address (usually, their parental address) prior to entry to an HEI. These could be used to produce migration matrices showing the movement of students to universities and colleges.



The Annual Survey of Hours and Earnings (ASHE) includes information about employees’ workplace and home postcodes, but it does not include any socio-demographic information about the individual, nor does it include information about the usual mode of transport to work. ASHE is captured through a questionnaire sent to employers (rather than employees) and thus there is no scope in the current collection model to collect any additional personal data, such as mode of transport. Even if this were to be the case, the small sample size (1%) means that no detailed spatial disaggregation could be made.

Royal Geographical Society (with The Institute of British Geographers) 30 November 2011

i

David Martin, Professor of Geography, University of Southampton and Coordinator of the ESRC Census Programme; Phil Rees, Emeritus Professor, School of Geography, University of Leeds; Professor John Stilwell; Danny Dorling, Professor of Human Geography, Sheffield University; Dr Martin Frost; Dr Darren Smith, Chair PGRG; Christopher Brunsdon, Chair QMRG and Tony Fielding, Research Professor, Department of Geography and, Sussex Centre for Migration Research, University of Sussex ii

Riva, M. Curtis, S., Norman, P. (2011) Residential mobility within England and urban–rural inequalities in mortality; Social Science and Medicine, doi:10.1016/j.socscimed.2011.09.030. )

iii

ONS (Office for National Statistics), NRS (National Records Scotland), NISRA (Northern Ireland Statistics and Research Agency_) and the WAG (Welsh Assembly Government) iv

ONS (2011a) Beyond 2011. Online at: http://www.ons.gov.uk/ons/about-ons/what-we-do/programmes--projects/beyond-2011/index.html.

v

http://www.ons.gov.uk/ons/about-ons/what-we-do/programmes---projects/beyond-2011/index.html

vi Home Office (2011) Identity cards. The UK National Identity Card and the Identification Card for EEA (European economic area) nationals ceased to be valid legal documents on 21 January 2011.Online at: http://www.homeoffice.gov.uk/agencies-public-bodies/ips/about-us/suppliers/identity-cards/ vii

Coleman, D. (2010) Memorandum from David Coleman, Professor of Geography at Oxford University. Ev 184, House of Commons Treasury Select Committee.

108

Supplementary written evidence submitted by the Royal Geographical Society (with The Institute of British Geographers) (Census 33a) Email to the Committee from Professor Phil Rees, Royal Geographical Society, 15 December 2011

Oral evidence 14 December 2011    I make a couple of observations on a remark made and a question asked by The Right Honourable  Graham Stringer, Labour MP for Blackley and Broughton.    The role of an information infrastructure in building the welfare state  Mr. Stringer made the point that the reforms of the Labour Government of 1945‐51 which  established the foundations of Britain’s welfare state were uniformed by a census because in 1941  the exigencies of the Second World War meant that such an enumeration was not possible.  However, I would point out that crucial to the war effort and the foundation of the National Health  Service was the second National Register, 1939 to 1952. Conscription, labour direction, rationing and  patient lists were founded on the National Register. I would argue that if you do not have a census  after 2011, then such a register is absolutely necessary, where set up, as I argued for, as a full  Population Register or constructed virtually by matching records across existing national  administrative registers, including the NHS Register.    Do poor people report less or more illness than richer people?  Mr. Stringer stated that he was puzzled by higher rates of illness (Limiting Long Term Illness) revealed  in the Census for Buckinghamshire than inner London or Manchester and suggested that the  question on LLTI might achieve different responses from richer than poorer people.  I don’t think this  observation is correct. The maps of LLTI from the 2001 Census (e.g. Daniel Dorling and Bethan  Thomas (2004. People and Places: A 2001 Census Atlas of the UK. The Policy Press, Bristol) show  clearly that Buckinghamshire has lower LLTI rates than inner London or Manchester.  A map from  their atlas showing LLTI for persons age 16+ (adults) by local authority is available online at  http://www.sasi.group.shef.ac.uk/publications/pandpexamples.htm#illness    However,  I would observe that this map underestimates the inequalities because before you can  compare illness rates in different areas, you must standardize for the effects  of age structure. When  you do that the gaps between poorer inner city and affluent suburb are even greater. Studies of the  variation of standardized LLTI across local areas in the UK suggest that about 70% of the variation is  associated with deprivation and a part of the 30% with urban living (linked to pollution, for example).  There is a  concern that the question about self‐reported health or illness may elicit perceptual rather  than real differences. However, a large number of prospective studies that link the answers to a  health or illness question to subsequent presented illness or death confirm that people are telling the  truth. That is why we have a health service that enables people to report their symptoms to their  doctor.   Professor Phil Rees, Royal Geographical Society December 2011

109

Written evidence submitted by the Institute for Jewish Policy Research and the Board of Deputies of British Jews (Census 34)   Assessing the potential impact of the termination of the national census  on the planning and provision capability of Britain’s Jewish community      1. Introduction    The  purpose  of  this  document  is  to  present  before  the  Science  and  Technology  Select  Committee,  the  shared  position  of  the  Institute  for  Jewish  Policy  Research  (JPR)  and  the  Board  of  Deputies  of  British  Jews  on  the  role  of  the  census  in  research  and  policy  planning  carried  out  by  these  organizations on behalf of Britain’s Jewish community.     • JPR is an independent research organization, consultancy and think‐tank.  It aims to advance  the  prospects  of  Jewish  communities  in  Britain  and  across  Europe  by  conducting  research  and developing policy in partnership with those best placed to influence Jewish life.    • The  Board  of  Deputies  of  British  Jews  is  the  representative  body  of  the  British  Jewish  community.    Its  research  unit  carries  responsibility  for  gathering  fundamental  community  statistics that are widely used to inform policy throughout the UK.    Data collection and analytical work conducted by JPR and the Board of Deputies are strongly policy‐ oriented.    The  chief  consumers  of  our  work  are  communal  planning  bodies  at  a  variety  of  levels,  charitable  organizations  and  private  enterprises  operating  within,  and  for,  the  Jewish  community.  Census  data  on  Jews  provide  JPR  and  the  Board  of  Deputies  with  detailed  information  on  the  demographic characteristics and socio‐economic make‐up of Jews in Britain.  No other data source  provides direct information of this kind.  The principal reason for this is that religion is not captured  by any publically available national register in the UK.    We  maintain  that,  as  data  on  religion  are  not  routinely  captured,  all  religious  groups  will  be  disadvantaged  as  a  result  of  potential  discontinuation  of  the  census  and  the  religion  question  contained within it. This is particularly important to religious minority groups which plan and provide  group‐specific services. In the context of Jewish community, the termination of the census, without  an  appropriate  substitute,  would  see  a  return  to  the  pre‐2001  Census  period,  forcing  complex,  expensive and inaccurate methods to be utilised in order to gather data on the basic characteristics  of the Jewish population.    We  would  further  maintain  that  the  gathering  of  data  on  religious  groups  should  be  in  the  government’s  interest.   Given  Britain’s  multicultural  nature  and  some  of  the  challenges  that  exist  within and between religious minorities, it is surely essential to have access to data that provide a  detailed view of the internal dynamics of each sub‐population.      2. How does the Jewish community use census data?   

110

2.1 

Both JPR and the Board of Deputies are engaged in large‐scale collection of data on Britain’s  Jewish population.  The tradition of data collection in the community dates back well over a  century.  The types of data routinely collected and analysed include: 

  2.1.1 

Vital statistics on Jewish births, deaths, marriages and divorces (none of which are  routinely or comprehensively collected elsewhere); 

2.1.2 

Social, economic and political attitudinal data of Jews (collected through community  surveys and again not comprehensively captured elsewhere); 

 

  2.1.3   Data on Jewish faith school education, standards, take‐up and attendance;    2.1.4    2.1.5 

  2.2 

  2.3 

  2.4 

  2.5 

Data on synagogue membership;  Census  data  obtained  from  the  Office  for  National  Statistics  (ONS)  by  means  of  standard and commissioned tables and through the Samples of Anonymized Records  (SAR). 

Data from the 2001 Census has become a principal element of the entire data ‘warehouse’  of the Jewish community.  Census data on Jews are important in their own right by providing  JPR  and  the  Board  of  Deputies  with  detailed  information  on  the  demographic  and  socio‐ economic characteristics of Jews in Britain.    Moreover,  census  data  have  come  to  form  the  backbone  for  the  system  of  demographic  accounting  for  Jews  in  the  UK.    Data  from  the  census  combined  with  data  collected  independently  in  the  Jewish  community  (see  2.1)  constitute  the  base  for  production  of  annual estimates of the size, structure and composition of the Jewish population in the UK,  as well as of summary measures of fertility and mortality pertaining to this population.  The  census  has  dramatically  increased  the  scope  and  accuracy  of  these  types  of  demographic  indicators.  Further,  the  census  provides  vital  benchmark  data  that  are  used  to  calibrate  community  surveys,  making  them  statistically  more  robust  and  valuable.    In  the  absence  of  a  full  sampling  frame  for  Britain’s  Jewish  population  that  would  enable  random  sampling,  sampling processes for Jewish community surveys have to rely on non‐random sampling or  particular indirect methods developed for hidden populations.  These methods are far from  ideal,  as  they  are  costly,  inaccurate  and  potentially  result  in  biased  samples.    In  this  situation,  it  is  especially  important  to  have  a  source  which  provides  a  picture  of  the  basic  distributions  of  demographic,  geographic  and  socio‐economic  characteristics  of  Jews,  allowing for assessment of bias and weighting of community survey data, if necessary.  Below  we  provide  the  most  pertinent  examples  of  the  use  of  census  data  by  JPR  and  the  Board  of  Deputies  for  communal  needs.    Further  examples  will  be  provided  in  the  next  section of this document. 

  2.5.1 

  2.5.2 

Census‐based  age/sex  distributions  of  Jewish  population  are  used  to  estimate  the  current  and  future  care  needs  of  the  Jewish  elderly  population,  both  in  terms  of  provision of residential care facilities and service provision.  Census age/sex data, in combination with measures of fertility, are used to plan for  the future provision of Jewish faith schools and child care facilities. 

111

  2.5.3 

Census‐based  data  on  the  age/sex  structure  and  geographic  distribution  of  the  Jewish  population  are  used  by  community  providers  such  as  community  centres,  kosher shops, charities and synagogues, many of whom have come to rely on census  data for planning their activities and services. 

  2.5.4.  Census  data  on  self‐reported  health,  presence  of  limiting  disability,  access  to  transport,  employment  status,  number  of  persons  per  household,  relationship  status,  mixed‐religious  households,  and  educational  status,  as  well  as  comparisons  of  all  these  metrics  with  other  communities  in  Britain,  has  become  an  important  planning  tool  for  a  multitude  of  organizations  working  inside  the  community  to  provide  relief  and  support  to  the  economically  deprived  sectors  of  Jewish  population,  as  well  as  other  disadvantaged  and  marginal  groups  within  the  Jewish  community.      3. What impact will the termination of the census have on social scientific research in the Jewish  community?     3.1  It is our view that the termination of the census, without  the  provision of alternative data  collection  systems  able  to  identify  Jews  in  a  meaningful  and  comprehensive  way,  will  severely  compromise  the  quality  and  scope  of  our  knowledge  of  the  Jewish  population.   Specifically,  JPR  and  the  Board  of  Deputies  will  no  longer  possess  direct  and  detailed  information  on  the  demographic  and  socio‐economic  composition  of  Jewish  population.   Additionally,  the  continuation  of  demographic  accounting  (e.g.  the  periodic  updating  of  a  given population base with data on births and deaths) will have major systemic drawbacks in  terms of accuracy and cost as compared with the census.  The absence of census data will  take  the  analytical  capacity  of  JPR  and  the  Board  of  Deputies  back  to  a  situation  where  simply establishing the size of Jewish community becomes a major research project in itself,  and one which is ultimately unreliable and uncertain.     3.2  As  a  consequence,  the  planning  and  performance  capacity  of  the  Jewish  communal,  charitable and private sectors will ultimately be curtailed.  British society as a whole, and the  Jewish population within it, faces multiple challenges in relation to the provision of targeted  services (such as education and care support).  Access to reliable and timely data on the size  and characteristics of the Jewish population is a pre‐requisite of informed policy.    3.3  The  termination  of  the  census  will  seriously  negatively  impact  on  the  following  unique  aspects of census data:    3.3.1  The  census  uniquely  provides  the  Jewish  community  with  data  on  ageing,  and  demonstrates  it  exhibits  a  considerably  older  age  profile  than  the  UK  average.   Demographically, this means that the proportion of persons at very advanced ages  (e.g. 95+) among Jews is relatively high compared with other communities.  From a  planning  point  of  view  this  reality  means  there  is  a  high  demand  for  Jewish  care  facilities and caring services, requiring for example special attention to the type and  location of Jewish care homes.    3.3.2  The census uniquely provides the Jewish community with accurate data on fertility  levels  and  the  scope  of  ‘intermarriage’  (Jews  married  to,  or  cohabiting  with  non‐ Jews).  These have a clear bearing on the planning of Jewish educational provision 

112

for  children  and  the  assessment  of  potential  demand  for  educational  and  other  communal services of mixed‐faith families.    3.3.3 

  3.3.4 

  3.3.5 

The  census  uniquely  provides  the  Jewish  community  with  accurate  data  on  its  geographical  distribution  down  to  very  small  and  detailed  areas.    This  means  all  census  metrics  can  be  examined  in  vastly  more  detail  than  was  ever  possible  with  even the most comprehensive of surveys. Specifically, the 2001 Census for the first  time indicated the very wide dispersion of Jewish people across Britain through all  districts and local authorities.  The  census  uniquely  provides  the  Jewish  community  with  data  on  internal  inequalities in terms of educational attainment, access to vehicles, deprivation and  hidden disadvantage and, in particular, of child poverty, within the community.  The census uniquely provides the Jewish community with data on migration in terms  of  both  international  and  national  population  movements.  Again  this  is  very  important from a planning point of view and is unavailable elsewhere. 

  3.3.6 

3.4 

The  census  uniquely  provides  the  Jewish  community  with  baseline  data  which  can  be used to evaluate the representativeness of Jewish population surveys conducted  within the community. 

  These brief examples of the additional insights the census has provided gives an indication of  what is likely to be lost with the discontinuation of the census especially without a suitable  replacement.  Planning and performing activities of a multitude of Jewish organizations that  rely on these unique insights will inevitably be curtailed, and the community will once again  have  to  resort  to  informed  guesswork  and  speculation  about  the  nature  and  future  composition  of  the  community.    Discontinuation  of  the  census,  therefore,  represents  a  threat to good governance in the Jewish part of the third sector. 

    4. What alternatives to the census would provide high quality population and socio‐demographic  data for the Jewish community?     4.1  The UK system of quantitative data collection about the population is one of the oldest and  most  advanced  in  the  world.    However,  in  contrast,  for  example,  to  Canada  and  Australia,  the United Kingdom does not possess a strong tradition of collecting data on religion.  Most  important administrative datasets (such as the National Health Service Central Register, the  Electoral Register, the vital statistics registration) do not include information about religion.   National sample surveys that do ask about religion (such as the Labour Force Survey and the  British  Household  Panel  Survey)  only  locate  a  very  small  number  of  Jews  –  typically  5  per  1,000  respondents,  which,  whilst  a  true  reflection  of  the  community’s  relative  size,  is  insufficient for analysis and communal planning.    4.2  A useful change to the current state of affairs would be the initiation of systematic inclusion  and  publication  of  information  on  religion  in  administrative  data  files:  the  vital  statistics  registration,  National  Health  Service  Central  Register,  and  the  Cancer  Registry  being  chief  examples.    Irrespective  of  this  development,  collection  of  information  on  religion  and  identification of Jews in principal administrative datasets will result in a significant analytical  and practical gain for research and planning agencies in the Jewish sector.   

113

4.3 

  4.4 

  4.5 

A  number  of  countries  of  northern  Europe  possess  population  registers,  incorporating  individual‐level  data  about  each  citizen  and/or  resident  of  a  country,  and  including  characteristics  such  as  date  of  birth  or  immigration,  sex,  ethnicity,  religion,  and  place  of  residence.  Typically, population registers rely on (a) the existence of unique identifiers for  each individual; and (b) operation of subsystems of registration that ‘feed into’ the register  and  keep  it  up  to  date.    Systems  of  vital  registration  and  border  control,  for  example,  are  linked to the register, and any changes in personal status of individuals or their movement  inside and across country borders are adequately reflected in the register.  A  population  register,  with  a  well‐developed  and  functioning  updating  system,  could  potentially  present  an  alternative  to  the  census.    The  experience  of  countries  with  the  register is instructive, however.  Not many countries with the register abandoned the census  completely.    The  chief  reason  for  this  is  that  the  functioning  of  the  registration  systems  updating the register is often imperfect.  This is especially true of registration of cross‐border  migration.  Thus, the census continues to provide a way for benchmarking population size,  structure and composition.  We  maintain  that  the  intention  to  discontinue  the  census  should  be  accompanied  by  thinking  about,  and  taking  practical  steps  towards  adequate  alternatives.    We  identified  a  population register as such an alternative.  It is important to stress that, bearing in mind the  research  and  policy  needs  of  the  Jewish  population,  such  a  register  should  include  information about individuals’ religion.  In parallel, all systems ‘feeding into’ the population  register  should  include  this  information  on  religion  too.    While  by  no  means  impossible  in  theory, currently no such system (most importantly, the system of vital registration) contains  this information. 

    5. Conclusion    5.1  The potential discontinuation of the national census is a cause of major concern to JPR and  the Board of Deputies.  A viable alternative to the census (such as a population register of  the  kind  maintained  in  several  Scandinavian  countries)  must  include  the  collection  of  data  on  religion  if  the  community  is  not  to  be  put  at  a  significant  disadvantage  in  its  data  gathering capacity.      6. Further reading    6.1  The best example of how the British Jewish community has utilized the vast amount of data  the census has provided can be seen in the following report:      Graham  D.J.,  Schmool  M.,  and  Waterman  S.  (2007).  Jews  in  Britain:  a  snapshot  from  the  2001 Census.  London: JPR / Institute for Jewish Policy Research.    6.2   To  understand  what  the  removal  of  the  census  means  in  terms  of  demographic  data  gathering see:    Graham D.J. (2011). ‘Enumerating Britain’s Jewish Population: reassessing the 2001 census in  the context of one hundred years of indirect estimates’. Jewish Journal of Sociology Volume  LIII.   

114

  Signatories to this document      Lord Haskel  President, JPR              Harold Paisner  Chairman, JPR              Jon Benjamin  Chief Executive, Board of Deputies of British Jews    Jonathan Boyd  Executive Director, JPR    Dr Laura Staetsky  Senior Research Fellow, JPR    Dr David Graham  Senior Research Fellow, JPR    Marlena Schmool  Former Director, Community Research Unit, Board of Deputies of British Jews    Daniel Vulkan  Research and Information Officer, Board of Deputies of British Jews      Institute for Jewish Policy Research (JPR) and the Board of Deputies of British Jews    30 November 2011 

115

Written evidence submitted by the Suffolk County Council (Census 35) Summary 1. Local Authorities make extensive use of census counts to assemble the evidence needed for formulating policy, service delivery and monitoring. The census has many advantages over other data sources as it spans the whole Country, is consistently collected, allows meaningful comparisons between areas and provides many complex variables for small geographical areas that can be tessellated to form customised areas. 2. Without the census counts the quality of the evidence underpinning policies would be less robust. 3. Administrative systems, surveys, modelling or linking of data could all be employed to provide counts of straight forward variables, but estimates of complicated variables or data for small areas could not be so readily provided. Consideration would also need to be given to academic researchers testing hypotheses using original anonymised records. 4. Many administrative systems such as Electoral Registers or registration of vital events could be extended to include other variables, but coordination is required to provide sufficient useful data. Legislation will be required to ensure records are comprehensive and fields completed.

1

How do social scientists use Census data?

1.1

Local Authorities make extensive use of Census data despite not specifically being instructed to do so. Councils do not conduct social research using the actual records, preferring to rely instead upon others for this sort of guidance. Instead we use the census counts per geographical area as a key information source.

1.2

There are a number of statutory plans for which we clearly need population characteristics data, such as the Waste Plan, Minerals Plan, Joint Strategic Needs Assessment, Local Transport Plan, Local Development Frameworks and the Survey of the Area (under 2004 Act and proposed to continue in The Town and Country Planning (Local Planning) (England) Regulations [2012]).

1.3

In addition there are several statutory responsibilities for which a background context is desirable, if not essential. Suffolk County Council uses the census results to develop profiles of population subsets living or working in the County, or for specific geographical areas within the County e.g. a particular town. These profiles form the evidence base for anticipating need directly associated with a particular Council service, or by services provided by other organisations in the public sector e.g. public health, or suppliers or contractors over whom the Council has a controlling responsibility e.g. house builders. The greater understanding provided by these profiles facilitates the development of policy, with priorities or targets to address these needs. Finally delivery of these services needs to be planned in a coherent manner. Being able to monitor service effectiveness enables policies to be updated to meet changing circumstances or to adapt them to meet emerging needs and requirements.

116

1.4

We also use the census data indirectly as a denominator for the nationally collated indicators that County Councils are expected to assemble and within indices of deprivation.

1.5

The census results have two main advantages over many other data sources and it is these that the Council exploits fully to prepare its evidence:

1. Census results are consistently available for small geographies 1.6 The Council needs data for small areas within Suffolk that are consistent with data for the County as a whole and with elsewhere in the Country. Such a suite of data makes it possible to both look at the area in question and put it in context; for example by comparing its characteristics with equivalent nearby areas by the simple expedient of ranking using a key variable or mapping values for this key variable across the County to identify ‘hotspots’ or spatial patterns. The County average provides the benchmark comparison. At a County level, comparisons need to be made with elsewhere in the Country using similar techniques. All these comparisons enable a greater understanding of an area so as policies may be tailored accordingly. 1.9

The advantage of having data for small areas is that these constitute building blocks that form other larger more meaningful geographical areas, with data for the smaller areas simply added to give data for the larger area. Data from the 2011 Census will be provided for output areas that can usually be assembled to give the key administrative areas of Parishes and Towns, plus District Council wards.

1.10

The need for data for small areas is not going to diminish; in fact there is a focus on decentralisation and localism for which more data for small areas will be required. The Localism Bill stipulates that the organisations with which County Councils are legally obliged to cooperate need evidence. The Bill also enables town and parish councils or new Neighbourhood Forums to draft neighbourhood Plans for which Local Authorities will be required to provide support and which Local Authorities will be legally required to adopt, the inference being that the Plans need to be properly evidenced. Suffolk County Council is encouraging the private sector or third sector organisations to take more responsibility for running some services but they will only be prepared to do so if a full business case itemising existing and potential client base is available.

2. Census results summarise complex issues 1.11 Categorising the answers to one census question using the answers to another census question amplifies the response to each question by giving extra information. Users are therefore able to explicitly differentiate between separate subgroups to better understand their specific characteristics. The outcome is a suite of data that gives the user a better understanding than if the answers to each question were to be categorised independently. For example, knowing how many pensioners there are in an area or how many people live on their own is valuable, but cross tabulating gives a much richer dataset. The census has the added advantage that all the variables refer to one point in time rather than at different points in time.

2

What Impact will the ending of the Census have on social science research?

2.1

Although much of the use Local Authorities make of the census is not specified by law, many of the documents in which the data is used are legally 117

required. Without the census, these would still have to be prepared but would be less well evidenced, with less clarity. We would therefore have less confidence in what they have to say.

3

What alternatives to the census would provide population and socio-demographic data of equivalent or higher quality?

3.1

It would be too expensive for Local Authorities to conduct commensurate sample surveys themselves. It is widely acknowledged around the world that people are proving less willing to complete such surveys, so the utility of conducting local surveys may be low. There have been experiments in the past involving enhancements to the electoral register, but these have not, by all accounts, yielded comprehensive coverage due to the obligatory nature of the additional questions. Whilst it is true Councils do have sophisticated administrative recording systems these are incomplete (since they refer to just the clients at the moment under current guidelines) so cannot give a full socio demographic picture of the area or a complete evidence base. Without a national effort, we would not be able to put our clients in context or make comparisons with other Authorities.

3.2

It is possible national databases could be used to yield basic counts of the population with particular characteristics, but these will have the same drawback as locally held administrative systems, namely that the records will only refer to clients. Cross tabulations of one variable against another would also not be so readily achievable. In particular, household characteristics such as household size and composition, both exceptionally valuable data, could not be categorised very readily using straight forward administrative systems. Record matching may be required to link data in one database with data from another to produce the more complex variables, but the weakness of this procedure is the quality of the matching and key used to perform the match. Such linked but anonymised records would be of value though to the academic researcher testing hypotheses. Other data, such as travel to work information can really only be reproduced by a survey, which if it is to produce reliable results for small areas, will need to have a high percentage coverage (like a census) or be repeated over a number of years to boost coverage. It may be possible to use modelling techniques to merge data from administrative sources or from surveys, or the two sources combined, to give approximate estimates but this procedure may still not be adequate for small geographic areas.

4

What other existing sources of population and socio-demographic data could be improved upon?

4.1

Some data sources could be easily enhanced to yield additional useful information; for example, vital event registrations could easily include a question about ethnicity information, but without a coherent and comprehensive work plan the results will be uncoordinated and piecemeal and may not be sufficient to yield the basic and necessary evidence required.

4.2

The Electoral Register could be extended to include those who cannot vote (to complete records) and with additional questions that are compulsory; in effect these would be annual censuses.

118

4.3 • • • •

There are a number of issues that could arise if existing sources of population and socio-demographic information were to be used for generating data. Legislation may be required: to avoid changes to any such existing administrative systems; otherwise the data time series will be disrupted to ensure records are fully completed correctly to ensure records are kept up to date to enable individuals to inspect their personal details particularly those derived by matching and linking However if the administrative processes are onerous these will equate to a census form and not reduce the burden on the public.

Response prepared by Mary Moore Senior Business Development Analyst Business Development Suffolk County Council 30 November 2011

119

Written evidence submitted by the Office for National Statistics (Census 36) 1.

The Office for National Statistics (ONS) has recently, at the request of the UK Statistics Authority, established a new Programme - ‘Beyond 2011’ - to look at the most appropriate way of providing population and small area socio-demographic data beyond the 2011 Census.

2.

ONS welcome the inquiry into how social scientists use census data and how changes to the approach might impact upon social science. Our recommendations, which will be made in 2014, will be strongly influenced by an understanding of what matters to users and how statistics are used – and the Committee’s findings will contribute to thinking on the best way forward.

3.

This note provides a high-level summary of the work being undertaken under the Beyond 2011 programme and is provided to the Committee for information. Further details on any aspect of the programme and the options being considered can, of course, be provided on request.

The Beyond 2011 Programme - Background 4.

The system for providing population and socio-demographic statistics for the UK, like that in many other countries, has traditionally been built around a census of the whole population. The traditional censuses in England and Wales, Scotland and Northern Ireland provide a population count and a snapshot of the nation at a single point in time. Supplemented by its coverage survey and quality survey, it currently provides the most reliable national source for a number of sociodemographic topics, and can produce accurate, multi-variate outputs for small areas. Many statistical series are re-based every ten years to take account of census results.

5.

However, the Census has a number of disadvantages not least the fact that its detailed outputs are updated infrequently (every ten years). This can be a significant issue in areas experiencing rapid population change, or when the importance of a particular socio-demographic topic suddenly changes in response to new or emerging Government policies and priorities.

6.

In addition, in recent times it has become increasingly challenging and expensive to conduct censuses and household surveys partly because of the complexities associated with an increasingly mobile population but also because of the reliance on the general public’s willingness to take part. At the same time, users want a greater range of outputs to be available and updated more frequently in order to have a better understanding of the population and how it is changing. Similar pressures are reported in other countries, indeed some countries have recently moved away from a traditional census approach while others are currently considering doing so.

7.

Prior to the decision to hold the 2011 Census, a review clearly established that, in the short term, a traditional census was the best way to meet user requirements in the UK – and all of the signs are that the 2011 Census operation has been highly

120

successful. Nonetheless there are growing concerns, including from the statistical and user communities, that census-taking in its current form may not be the right vehicle looking ahead. While a traditional census can produce high quality statistics for small areas, it only delivers these data every 10 years and there are significant costs and risks associated with the traditional approach to census taking. 8.

In May 2008 the Treasury Select Committee report ‘Counting the Population’ recommended that:“The Statistics Authority set strategic objectives to ensure that the data gathered throughout the UK can be used to produce annual population statistics that are of a quality that will enable the 2011 Census to be the last census in the UK where the population is counted through the collection of census forms”.

9.

Furthermore, in May 2010 Sir Michael Scholar, Chair of the UK Statistics Authority, wrote to the Minister for the Cabinet Office to say that:“As a Board we have been concerned about the increasing costs and difficulties of traditional census-taking. We have therefore already instructed the ONS to work urgently on the alternatives with the intention that the 2011 Census will be the last of its kind”.

10.

The Beyond 2011 Programme was formally established by ONS in April 2011 to look at the alternatives to running a Census in England and Wales in 2021. An iterative programme of research is being taken forward to:•

check and clarify user requirements for population and socio-demographic statistics including the frequency, quality and geographical level required; and



assess the feasibility of alternative statistical options against a range of agreed criteria which will include consideration of issues such as cost, risk and public acceptability.

11.

Close collaboration is planned with the devolved administrations in Scotland and Northern Ireland, who are considering parallel work, to ensure that the obligation to produce UK statistics is met.

12.

Following a period of extensive statistical research, the Beyond 2011 Programme will report its findings and make a recommendation in 2014. The recommendation will be evidence based and supported by a full business case. The outcomes will have implications for all population-based statistics in England and Wales and potentially, in the longer term, for the statistical system as a whole.

121

Options for Meeting Users’ Requirements 13.

A long list of alternative options for meeting users’ needs for small area population and socio-demographic statistics will be tested, evaluated and prototyped before making a recommendation in 2014.

14.

At this early stage we are considering a range of alternatives including: •

census-type solutions;



survey based solutions; and



administrative data solutions.

Annex A provides a brief summary of the main options being considered. (Note that this Annex is already published as part of the current ‘Beyond 2011 Consultation on User Requirements’ – but is attached here for information) Note that although a number of census based options are included, ONS has considerable knowledge and experience of census and survey operations and so the research programme will initially focus more upon assessing options for making better use of existing administrative or commercial datasets. Research will include close consideration of the value that might be added by the re-use of key administrative sources such as the electoral roll or school census data. 15.

All statistical options will be assessed against an agreed set of criteria. The specification of these criteria will take account of users’ requirements and assess each option in terms of quality, cost, risk, public burden and acceptability (amongst others). When considering possible solutions, we recognise there may be tradeoffs to be made between some of the factors important to users - for example, accuracy and frequency - and we are aware of some concerns in this area, for example that a change in model may result in a loss of scope and quality but may bring cost savings. The specification of the criteria to be used will be crucial to ensure that the right balance is struck when considering the options.

Plans for User Engagement 16.

A full understanding of user needs and priorities is essential in order to ensure that maximum benefit is gained from the opportunity to refresh our approach.

17.

Beyond 2011 will be undertaking a comprehensive programme of stakeholder and user engagement. This will build on the work undertaken for the 2011 Census. In particular, we shall be making use of our extensive list of contacts to ensure that key users are aware of the work we are planning and have an opportunity to comment on our proposals.

18.

The first of two public consultation exercises has already started and will run until 20th January 2012. This is designed to give both established users and other interested parties an opportunity to: •

contribute to our understanding of users needs and priorities; and



provide information on methods and data sources which could be used to produce population and other socio-demographic statistics.

122

19.

The consultation also provides an opportunity to properly assess the importance to social scientists and the wider community of research data such as the sample of anonomised records (SARs) and the ONS Longitudinal Study (LS) which are currently entirely dependent upon the census. We are very aware of the central place played by these datasets in much research and more generally of the role of the census as a historical record. Ability to support research work will be one of the considerations that feed into our final recommendations.

20.

The results of this consultation will be published in spring 2012 and will contribute to the development of evaluation criteria for the assessment of the alternative statistical options.

21.

A second public consultation will take place in 2013 to seek comments on the leading options and their relative merits in order to inform the final recommendation. These public consultations will be supplemented by meetings and events with all key stakeholders – including central and local government, commercial users and suppliers, academics, community groups, utilities, privacy groups, genealogists and other interested parties. The views of social scientists will form an important input here – hence our strong interest in the results of this inquiry.

22.

Regular feedback will be provided on emerging research findings, the refinement of the options, evaluation criteria and option scores, in order to ensure that our decision making processes and the rationale behind the final recommendations are understood.

Conclusion 23.

The Beyond 2011 Programme provides an opportunity to refresh our approach to producing population and small area socio-demographic data in England and Wales. In order to ensure we identify the right approach for the nation, the views of all users, including social scientists, are essential. The outcome of the current inquiry will provide a valuable input to help inform the criteria used in selecting the final options - and so the recommendation on the best way forward.

ANNEX A: Statistical Options (Note that this Annex is already published as part of the current ‘Beyond 2011 consultation on User Requirements’ – but is included here for information) A1 This annex provides a high level summary of the main options that will be considered by the Beyond 2011 programme. Each of these options (and sometimes variations upon them) will be tested and assessed against an agreed set of criteria A2 ONS has considerable knowledge and experience of census and survey operations and so the research programme will initially focus more upon the administrative data-based options outlined below. A3 The descriptions here are simplified in the interest of brevity. Equally it is very possible that the solution adopted may use elements of more than one of these options – and we may recommend different solutions for different types of data. Much more information on the lead

123

options and the methodologies that lie behind them, will be released over the course of the Programme, including a more detailed paper in 2012. Traditional Census A4 This option is a continuation of the current approach i.e. conducting a full census of the population at a single point in time (for example, most recently, in March 2011). Everyone completes a ‘long’ form questionnaire and the responses provide the basis for producing a count of the population and details on key characteristics. Historically a census has been taken every ten years in England and Wales, but some countries use a five year cycle instead. In-between censuses the results are supplemented by a series of mid-year population estimates which use the ‘cohort component method’ to add and subtract to take account of births, deaths and net migration. Social surveys on particular topics are used to fill in details, and update estimates as required in the inter-censal period. Short and Long Form Census A5 This option is similar to the current census approach in that information would be collected from the full population at a single point in time. A subset of the population would be asked to complete a traditional ‘long form’ census questionnaire. A ‘short form’ questionnaire collecting basic demographic, household and family information would be sent to the remainder of the population. Estimates would be created from a combination of the two sources. Population estimates would be updated using the existing cohort component method, and sociodemographic statistics by social surveys. This approach has been used historically in Canada. Short Form Census + Continuous Survey A6 In this option, basic demographic, household and family information would be collected from the full population at a point in time through a short form census i.e. a questionnaire covering these basic topics only. Population estimates would use the existing cohort component method in the inter-censal period. Socio-demographic statistics would be collected through a large scale continuous survey. A similar approach has been used in the USA, with the ‘American Community Survey’ replacing the long form census element. Rolling Census A7 In this case, a census is conducted in different areas on a rolling basis. For example, 10% of the country might be selected each year so that over a decade the whole population would be counted. Population estimates would be supplemented by the existing cohort component method, and socio-demographic statistics by social surveys. An approach of this kind has been implemented in France. Research has already been commissioned into the suitability of a rolling census approach in England and Wales. Address Register + Survey A8 Data from an address register is combined with a survey to estimate the population and its characteristics. In its simplest form, the average number of people living at each address included in the survey is multiplied by the number of addresses in an area to give an estimate of the population in that area. Estimates could be updated on an annual basis, but a longer time period may be required to produce estimates for smaller geographic areas. Careful stratification can be used to increase the quality of estimates but this approach is very sensitive to the quality and consistency of the address register – and local variations in household size. Administrative sources or a specific coverage survey would be used for quality assurance, or to supplement survey data to produce small area estimates. Administrative Data – Aggregate A9 In this option, aggregate data from a variety of administrative sources would be used to produce initial population counts. This would be achieved by a weighted average of the various sources, stratified by variables including age, gender and geography. Extra information, such as a coverage check survey would be used to refine weights applied to the

124

stratified initial counts. These weights would be regularly updated, producing annual population estimates. Data to produce socio-demographic statistics would be collected through survey(s), with potential for including information from public and private-sector administrative sources in the future. Administrative Data – Record Level A10 In this option record level administrative sources (including the address register) would be linked together to produce initial population counts. A regular coverage check survey would be used to assess the accuracy of the initial count, measuring both under-coverage and overcoverage. An estimation process (e.g. dual system estimation, ratio estimation) would be used to derive weights to be applied to the initial population counts. Further statistical adjustment, for example through imputation, may also be possible. Data to produce sociodemographic statistics would be collected through survey(s), with potential for including information from administrative sources in the future Administrative Data - Intermediate A11 This option would make use of both aggregate and record level administrative data approaches. The aggregate administrative data approach would be used to produce initial population counts. Data from a coverage check survey, measuring both under-coverage and over-coverage, would be linked to record level administrative data sources and the address register, in the coverage check areas only. An estimation process, e.g. dual system estimation or ratio estimation, would be used to derive weights to be applied to the initial population counts. Data to produce socio-demographic statistics would be collected through survey(s), with potential for including information from administrative sources in the future. A12 Almost all of these options will use an address register either as a source or as a frame for surveys. At the same time, all options are likely to be supported by some form of coverage survey or independent check on quality to help correct for over and under-coverage, in addition to a quality survey to monitor the quality of results. Office for National Statistics 30 November 2011

125

Supplementary written evidence submitted by the Office for National Statistics (Census 36a) As requested at Q76, please find attached a short paper on international comparisons.

International Cost Comparisons In preparing the business case for the 2011 Census in England and Wales, ONS presented equivalent costs that had been supplied by some other countries who undertake similar field enumeration censuses. These are shown below. Note that the costs of the England and Wales Census had been converted to 2005/06 prices for better comparison.  

England and  Wales    2011 

USA        2000 

Canada      2006 

Australia  2006  

New  Zealand  2006 

Cost per head sterling  equivalent at 2005 / 06 prices 

 £7.53

 £17.20

£7.40

£6.10 

£7.50

Cost per head per year 

£0.75

£1.72

£1.48

£1.22 

£1.50

Note also that Canada, Australia and New Zealand carry out a census every five years, England and Wales and the USA, every ten years. Countries with five-yearly censuses are more able to re-use infrastructure (such as processing and output systems) and thus spread the costs between censuses. They also make the census investment twice in the decade, giving significantly higher costs per year than in England and Wales. These figures demonstrate that the England and Wales Census costs are comparable with those in Canada, Australia and New Zealand, who all do five-yearly censuses and have scope for re-using infrastructure, and are significantly lower than the US - where the use of census benchmarks to determine seats in Congress has justified higher levels of spend. Independently, the United Nations Economic Commission for Europe (UN-ECE) issued a report in 2008 (Measuring population and housing*) in which it compared the practices of countries within the ECE region in the 2000/01 round of censuses. In examining the per capita costs it reported that the UK was ranked fourth lowest out of the 15 of those countries that conducted a traditional field enumeration with self-completion questionnaires. * http://www.unece.org/stats/publications/Publication_on_2000_censuses.pdf)

I hope that this has been helpful. Glen Watson, Census Director, Office for National Statistics December 2011

126

Written evidence submitted by The Salvation Army (Census 37) 1. How do social scientists use Census data? Founded in East London by William Booth in 1865, The Salvation Army (The SA) is one of the largest, most diverse providers of social services in the United Kingdom. In the United Kingdom and Ireland, The SA has approximately 50,000 members, 4,000 employees and 1,500 Salvation Army officers (full-time ministers) 1 The SA uses Census demographic data to understand the demographics of a community in which The SA is working, or is considering working. We use the data to assess deprivation, need and development opportunities. The SA produces demographic profiles which report on the population, health, employment, education, housing and deprivation of an area. The Census provides us with a foundation of what a community consists of. Census data directly informs what work The SA will commission and what work we will need to further investigate. We use ONS population projections and estimates regularly to assess the future demographic of an area, e.g. ethnicity projections and age group projections. This helps to future proof any investment we may make and influences what activities we may offer to a particular community. Population projections are an important tool for effective action. We use Census data to show, through longevity, how a community has changed over a period of time. We predominantly use lower super output area and ward based geographies. We will also make data comparisons with local authority geographies. We do not use Government Office Regional data. The ability for the Census to report on small geographic areas robustly and with high statistical significance is invaluable for The Salvation Army. Many of our locations are inner city areas whose intricacies often get dwarfed by the urban area at large. The Census provides us with an intimate picture of small but unique areas. In conjunction with Census data we use The Department of Work and Pensions (DWP) data to report on the health of a community and its employment status. We also report on migrants entering an area through the National Insurance registration. We also report on housing figures and school performances. This data is accessed via various sources, including the DWP tabulation tool, NOMIS, local councils, etc. These other sources of information are beneficial as they use the same geographic categories as the census. This enables us to use fixed data parameters when presenting information. When presenting demographic information on a community, as mentioned above, we use multiple data sources but we make explicit that it is the Census data which forms the foundation of our research.

2. What impact will the ending of the Census have on social science research?

1

The Salvation Army About Us, http://www.salvationarmy.org.uk/uki/www_uki.nsf/vw-dynamicarrays/8EA9C0B3D06D8814802578CC002EA25E?openDocument&charset=utf-8

127

2 We have three main concerns for the removal of the Census: 1. Lack of statistical verification of other data sets: the Census is a comprehensive data set that covers the whole of England, Northern Ireland, Scotland and Wales. We do not believe the same high level of statistical significance and coverage can be achieved by amalgamating other national/regional/local data sets. We believe the same geographic levels that are so consistently reported on now will be lost and, therefore, data will not be used to its fullest potential. Ultimately, the communities on which this data is used to make decisions on, will loose out. 2. Loss of longevity: it will be difficult and time consuming to ensure any historical comparisons will stand up to statistical rigour. 3. Ineffectual integrated working: any amalgamation of data sources will involve in-depth change management across huge organisations (DWP, NHS, DCLG, BIS, Department for Education etc). What guarantees does the charitable sector have that an equivalent integrated data system, like the Census, would be up and running by 2021? Considering the failures of the NHS’ National Programme for IT – is this something the Government is confident that it can deliver on?

3. What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? We do not think there are any alternatives to the Census that would provide equivalent population and socio-demographic data. However, we do believe the data is out there and currently collected by different sources. The difficultly, as mentioned above, would be coordinating the data to provide a consistent national, regional and local output.

4. What other existing sources of population and socio-demographic data could be improved upon? The level of migration data is poor. We are only able to report on migrants that have applied for a National Insurance number. We are not able to report on those that haven’t or track the through flow of migrants within the UK. The categories of ethnicity as presented through current and previous Census collections are not particularly helpful in mapping the ethnicity of respondents. It is assumed that the UKBA collect data on country of origin and ethnicity in greater detail but how much of this data is publicly available to social researchers?

5. Declaration of Interest Fleur Bragaglia, Researcher for The Salvation Army, was a Census Collector for the 2011 Census in the Barnet area. The Salvation Army responded to the Office of National Statistics Work Programme Consultation on 24th December 2010. Research & Development Unit The Salvation Army 22 November 2011

128

Written evidence submitted by the Joseph Rowntree Foundation (Census 38) Declaration of Interests: none  The Joseph Rowntree Foundation is an endowed charity that funds a large, UK-wide research and development programme. We seek to understand the root causes of social problems, to identify ways of overcoming them, and to show how social needs can be met in practice. Our purpose is to influence policy and practice by searching for evidence and demonstrating solutions to improve: • • •

the circumstances of people experiencing poverty and disadvantage; the quality of their homes and communities; the nature of the services and support that foster their well-being and citizenship.

We have no political affiliations and work in partnership with all sectors – private, public and voluntary. We aim to present evidence in a balanced, unbiased way and to stimulate debate on current and emerging issues. In all our work, we look to reflect the diversity of the UK population, learn from others and operate in a sustainable way - socially, environmentally and economically - finding practical and realistic solutions and focusing on the needs of disadvantaged people. Through our housing, community and care services, we aim to achieve the highest professional standards. Across our research programmes we have used information collected in the UK Census to support various projects and these incidences will be highlighted in this submission. It is our intention to answer the four questions posed as requested and to make the case that the Census is an invaluable data source in the field of social science.

129

1. How do social scientists use census data?  1.1. As the census attempts to get information from every person in the UK it is the only survey which provides a detailed picture of the entire population. It is unique because it covers everyone at the same time and asks the same core questions everywhere. This makes it possible to compare different parts of the UK, Local Authorities and even smaller area statistics on many of the range of variables collected. Although this response will focus on the social research value of the Census, it is important to note the key role it plays in providing and developing public services. The information the census provides allows central and local government, health authorities and many other organisations to target their resources more effectively and to plan housing, education, health and transport services for years to come. 1.2. Census data is used by social scientists to form the backbone and in many cases the core of many research projects. Aside from the obvious national representation (despite undercounts for certain types of people such as homeless people and the travelling community) the census data enriches many other datasets by linking with them and therefore really enhancing the analysis possibilities. For example, the ONS Longitudinal Study (LS) is a data set comprising linked census and event records for 1% of the population of England and Wales (about 500,000 people at any one census). It was set up in 1974 to address problems with the adequacy of occupational mortality data, and the lack of longitudinal fertility data, but since then it has been used to address a wide range of other research questions. The LS is invaluable for longitudinal and cross-sectional studies of census and events data as it contains all information from census returns since 1971 (including data on occupation, economic activity, housing, ethnicity, age, sex, marital status and education), administrative events data (on births & deaths, fertility, mortality, migration and cancer registrations) and also geographic data. The linking of the data to the LS allows researchers to look in detail at so many things which would not have been possible otherwise such as; survival rates of cancer in relation to sociodemographic characteristics of the population; and birth outcomes of second generation migrants. 1.3. Secondary data analysis from census data saves time that would otherwise be spent collecting data and, particularly in the case of quantitative data, provides larger and higher-quality databases than any individual researcher or research team could collect on their own. This coverage and robustness would not be possible to carry out or maintain by a small research team. In addition, analysts of social and economic change consider secondary data essential, since it is impossible to conduct a new survey that can adequately capture past change and/or developments.

1.4. Here at JRF we have used information from the Census to support a wide range of projects and programmes, some of which are summarised below:

130



JRF recently commissioned two projects as part of its ‘A Better Life’ programme, to stimulate and inform thinking on alternative approaches to a better life for older people with high support needs. We supported the projects as part of our programme of research and development projects which we hope will be of value to policy-makers, practitioners and service users. Key information from the Census used in this project includes the question asking respondents whether they have ‘any long-term illness, health problem or disability which limits activities or work’. Media coverage from this project includes a round up about how to empower older people on Publicnet.co.uk (October 2010) and an article in The Guardian (October 2010) – “A better life for older people” discussing this research printed on Older People’s Day.



The other project as part of the Better Life Programme resulted in a research paper entitled, ‘Demographic issues, projections and trends: Older people with high support needs in the UK’ (October 2010). The population aged 85 and over is the fastest growing age group in the entire UK population. More older people than ever before live alone and this has implications for residential arrangements and care among this group. This paper provides an overview of demographic issues (from the Census), projections and trends in relation to older people with high support needs in the UK and recommends continued investment in data sources to further understanding of health, disability, economic and social wellbeing in old age.



The Census Programme run by the Place team at JRF produced a number of influential reports based on information collected in the Census. A study on ‘Population turnover and area deprivation ‘(Bailey and Livingston, 2007) draws on detailed analysis of flows of population for neighbourhoods in England and Scotland, based on 2001 Census data. It examines whether deprived neighbourhoods have less stable populations, whether they are poorly connected to the wider housing market through movements of people in and out, and whether they are losing better qualified individuals through net movement out of the area. Using data from the 2001 Census, this study looks at inward and outward flows of population for small areas or neighbourhoods in England and Scotland. The Census provides a unique opportunity to study these population dynamics, as it captures information on all people who moved in the year before the Census. It therefore provides data on inward and outward movement of population and hence net flows of people for every neighbourhood in the country.



Changes in communal provision for adult social care: 1991-2001(Banks, Haynes, Balloch and Hill, 2006) examined shifting trends in care home provision for adults between 1991 and 2001. The report analyses detailed census data from 1991 and 2001 to reveal general trends in adult social care provision. A number of variables are used to understand these trends, including levels of local authority funded places, health and poverty indicators, property prices and provision of other types of social care. This project explored changes in residential and nursing care provision between 1991 and 2001. Using Census data, the

131

researchers were able to look at general trends and also at relative changes in the age, gender and ethnicity of care home residents. •

A particularly influential project that attracted a lot of media attention, both nationally and internationally was on ‘Migration and social mobility: the life chances of Britain’s minority ethnic communities’ (Platt, 2005). This project investigated the impact of class background and ethnicity on class position. Drawing on data from the ONS Longitudinal Study (LS), the report traces patterns of intergenerational social mobility for children from different ethnic groups growing up in England and Wales. For the first time, this report measured their progress and class position compared with those of their parents. It provided a unique insight into the differences between the class positions of parents who had migrated to Britain and their children raised in the country. Taking advantage of the new question on religion in the 2001 Census, the report also asks whether patterns of intergenerational mobility vary by religious affiliation and whether religion can add to our understanding of ethnic group differences.



The relationship between poverty, affluence and area was also analysed using 2001 Census data by Wheeler, Shaw, Richard Mitchell and Dorling. The report used Census data to map the nature and extent of geographical and social inequality in the UK at the beginning of the twenty-first century. Findings included higher numbers of practising, qualified medical practitioners living and working in areas where rates of illness are lower; an ‘inverse education law’ in which areas with the highest proportions of young people with no qualifications tend to have the fewest teachers; well-qualified people living outside London and the South East who accept lower status jobs simply because of where they live; households which might be said to have more cars than they need, while others don’t have any at all. The two groups of households tend to live in very different places.



JRF also publishes annual publications that use information from the Census most notably, Monitoring Poverty and Social Exclusion (JRF’s flagship publication) Monitoring Poverty and Social Exclusion is an essential resource for policy-makers and others wanting to take stock of what is happening and understand the challenges ahead. It highlights a number of issues that need to be addressed through government policy: in-work poverty, the number of children/young adults with few/no qualifications, young adult unemployment, health inequalities and low-income households' lack of access to essential services.

2. What impact will the ending of the census have on social science  research?    

132

2.1. Aside from the rich and robust research that is possible from the census a fundamental concern about ending the census is its impact on overall social science data quality. Data quality in general will be greatly reduced. For many of the most valued and robust social surveys in the UK the census is used to select postcode sectors for the main sample and (in the case of surveys such as Understanding Society) the ethnic boost sample. Census information is therefore integral to the design of these surveys. Virtually all sampling in the UK is based to some extent on census information, as it enables researchers to pick a representative set of areas. Consequently, if the census was cut, it would reduce the quality of all social (and commercial) survey designs. 2.2. As well as the value of linked events data to the Census in the ONS Longitudinal Study (LS), it is also used to carry out invaluable longitudinal research on themes such as mortality, fertility, cancer survival, migration, infant mortality, deprivation, social mobility, ethnicity and ageing. It is the only study that has a complete set of census records for individuals, linked between successive censuses, together with data for various events. Thus, the LS represents a continuous sample of the population of England and Wales, rather than a sample taken at one time point only. It now includes records for over 950,000 study members. Fresh LS members enter the study through birth and immigration and existing members leave through death and emigration. In addition to the census records, the individual LS records contain data for events such as deaths, births to sample mothers, emigrations and cancer registrations. This rich data source based on the census has allowed us to undertake prospective analyses of census and event data. For example: studies of associations between employment status and mortality, between economic status and cancer registrations, and between socio-economic factors and fertility; survival analysis of mortality by area deprivation: prospective analysis of successive census data. For example: analysis of patterns of retirement migration, studies of the effects of divorce and remarriage on housing tenure, comparison of changes in education, employment and migration between the 1970s and 1980s. 2.3. The 2011 Census will be linked into the LS study. The main purpose of this is to link a fifth tranche of census data, laying the foundation for the linkage of a fifth decade of life events data. Information gathered while linking the data will feed into the 2011 Census quality assurance process, identifying areas of possible undercount or overcount. If the 2011 Census is the last of its kind, this clearly has severe implications for the LS as the rich social research that has been carried out using this data source will no longer be possible after 2011. 2.4. Another impact will be the loss of the small area statistics and indeed microdata that enables analyses to be undertaken at the Local Authority level. This level of geography enables mapping, local level tabulations, and even multivariate analyses using local area level variables. These data differ from aggregate outputs such as the Census Area Statistics in that users can look at individual level characteristics within areas and define area level measures of their own choice. Users can define and use subsets, and can create new classifications by grouping existing classifications or combining information from more than one socio-economic characteristic. Microdata from the UK census are a unique source for studying residential mobility, and

133

particularly for providing an insight into ethnic differences in internal migration and the processes of ethnic residential integration. The size of the Sample of Anonymised Records (SARs) and the Controlled Access Microdata Sample (CAMs) enable the migration patterns for relatively small population subgroups, such as minority ethnic groups, to be examined; and to be examined alongside other characteristics of individuals and neighbourhoods.

3. What alternatives to the census would provide population and  socio­demographic data of equivalent or higher quality?  3.1. There are various potential alternatives to the census in providing population and socio-demographic data although each source has its strengths and weaknesses. The new UK Household Longitudinal Study, Understanding Society (USoc), offers the possibility of a much more fine-grained analysis of UK society, with a target sample size of 40,000 households and 100,000 individuals, bigger than any other comparable longitudinal study. The study design includes a significant sample boost for key ethnic minority groups and aims to collect biomedical measures and samples to enable new research on the social determinants and impacts of health in a household context. This will complement data from the English Longitudinal Study of Ageing (ELSA) and from national birth cohort studies. The main problem however with USoc as an alternative to the census, is that it’s longitudinal, so over time because of wave on wave attrition (where the number of respondents decreases over time due to loss to follow up, emigrating or death) it becomes less representative of the population. Although it is large and robust, when you get down to small area statistics which census produces, then USoc is not appropriate. 3.2. There are a number of surveys in the UK that allow detailed analyses at a national level to be carried out due to advanced techniques in weighting the data so that population projections can be made. These surveys include: • The Labour Force Survey (LFS), • The Longitudinal Study of Young People in England (LSYPE) • Households Below Average Income (HBAI) • The Annual Population Survey (APS) • The Family Expenditure Survey (FES) • Health Survey for England (HSE) • Families and Children Study (FCS) • Growing Up in Scotland (GUS) • Millennium Cohort Study (MCS) Although incredibly rich and invaluable in many instances it is not possible to produce small area statistics from these datasets by many indicators due to the size of the samples and confidentiality issues. Another problem for national level analysis is that of missing data where most often the information that is missing is from the key population that we are most interested in. It is of course noted that the detail and information collected in these surveys are invaluable in their own right and that they are all hugely rich data sources.

134

3.3. The LFS Annual Local Area Data Series, for example, and the Unitary Authority/Local Authority (UA/LA series) have been withdrawn on instruction from ONS, due to confidentiality issues. Although LFS databases released for analysis to social science researchers by outside bodies have always been anonymised to ensure that users could not identify any respondent with the information given, advances in technology and software has made it easier to link survey records to either other survey files or other administrative or commercial databases. ‘Although the risk for most respondents is very small, there remains a risk of identification for people with unusual combinations of personal circumstances. Thus the release outside the central government statistical services of social survey databases with small area identifiers, alongside a national database with detailed coding, has now been ceased’ (ONS).

4. What other sources of population and socio­demographic data  could be improved upon?    4.1. A key survey that it may be possible to build on would be The Annual Population Survey (APS). APS represents a major survey which comprises key variables from the Labour Force Survey (LFS), all the LFS boosts and the APS boost. The APS is now able to provide survey data that can produce reliable estimates at local authority level. Key topics in the survey include education, employment, health and ethnicity. There must be potential to build on this survey improving the variety of information captured. The APS combines results from five different sources: the Labour Force Survey (LFS); the English Local Labour Force Survey (LLFS); the Welsh Labour Force Survey (WLFS) and the Scottish Labour Force Survey (SLFS). The APS aims to provide enhanced annual data for England, covering a target sample of at least 510 economically active persons for each Unitary Authority (UA)/Local Authority District (LAD) and at least 450 economically active persons in each Greater London Borough. In combination with local LFS boost samples from Wales (WLFS) and Scotland (SLFS) the survey provides estimates for a range of indicators down to Local Education Authority (LEA) level across the UK. The position of the Joseph Rowntree Foundation is that the Census should certainly stay due to the evidence collected and summarised above. It will always be the gold standard of data collection and social science research would not be as robust without it in terms of data quality, coverage and analysis. Practical and effective policy interventions based on evidence which is not as robust may be less effective. The analysis that is possible using the actual Census datasets and those combined and linked to others opens up realms of additional research opportunities that would no longer be possible if the Census was no longer carried out.  

Joseph Rowntree Foundation  30 November 2011 

135

Written evidence submitted by Professor Leslie Mayhew (Census 39) The Census and social science This memorandum is intended as a contribution to the inquiry into the impact of ending the Census on social science research. I consider the issues to be extremely important to the future of social science and set out in this note my views on certain key aspects, particularly the use of alternatives. I would also be very happy to provide oral evidence if invited to attend the hearing. I am Professor of Statistics at Cass Business School and Director of Mayhew Harper Associates Ltd. I have been working with local administrative data in an academic and private capacity for a number of years both as a user and also as an advisor to local authorities and healthcare providers. As an academic I have published extensively on demography, health and social care, social security, pensions etc. My use of data has spanned both Census and administrative data and so I am able to compare many different sources and also application areas. I am also a current member of the Office for National Statistics (ONS) National Population Projections Expert Advisory Panel. I spent much of my early career in the civil service where I was a regular user of different sources of data (Census, surveys, administrative systems) both as professional analyst and then as a senior civil servant in the DHSS and then DSS. I also spent five years as a director in the Central Statistical Office (then part of HM Treasury) and subsequently in the Office for National Statistics itself. Since leaving the civil service in 1998 to become an academic, I have accumulated 13 years of experience working with administrative data in local authority and PCT environment. My breadth of experience therefore covers national and local administrative data collection systems and surveys of different kinds. I am a past user but also a strong critic of the Census and of the methodology used. I therefore consider myself to be uniquely positioned to express views particularly on questions 3 and 4 in the inquiry: -

What alternatives to the Census would provide population and sociodemographic data of equivalent or higher quality? What other existing sources of population and socio-demographic data could be improved upon?

Two papers published in 2011 by myself and with my colleague Gill Harper are highly relevant to this debate and are recommended reading for the Committee: 1. ‘Applications of Population Counts Based on Administrative Data at Local Level’.

136

G. Harper and L. Mayhew (2011). Jnl. Appl. Spatial Analysis DOI 10.1007/s12061011-9062-z 1 2. ‘Using Administrative Data to Count Local Populations’ .G. Harper and L. Mayhew (2011). Jnl. Appl. Spatial Analysis DOI 10.1007/s12061-011-9063-y 2 A key feature of our approach is that we create comprehensive population data bases by combining different local data sets. These data sets are used by local providers of health, local authority and other services. Applications of the data base are often cross-cutting in policy or research terms and enable useful linkages between sectors of the local economy e.g education, health and social care. The data base produced may be contrasted with information available from the Census -

The results are timelier than the Census which can take several years to process before being released and be more than 12 years out of date before it is refreshed. Based on the 2011 Census basic population aggregates will not be released before July 2012 and more detailed data will not follow in some cases until much later.

-

On Census day this year we obtained and processed snapshots of local administrative data covering the six Olympic Boroughs in London who commissioned the work. Fully geo-referenced databases down to household level by age and sex were completed and handed over to local authorities inside six months.

-

The databases contain much greater granularity than is possible using Census data. Each database is able to produce statistics for any size or shape of geographical area or administrative unit and unlike the Census is not constrained by any pre-determined geographical boundaries.

-

The data can also be flexed by age so that any group of any age width can be considered. This is important for applications in education such as forecasting school places and in health and social welfare for which eligibility or qualifying ages often do not fit easily into standard age groups adopted by the Census.

-

Census data cannot be linked to information on users of different services but our data can. That makes it more valuable to service providers but also to academic users for example to measure inequalities, unmet need, inefficient provision or coverage of services.

1

http://www.cass.city.ac.uk/__data/assets/pdf_file/0009/82692/Applications_of_population_counts_har per_mayhew_full_text_JASAP_0411.pdf 2

http://www.cass.city.ac.uk/__data/assets/pdf_file/0019/82711/Using_administrative_data_to_count_po pulations_harper_mayhew_full_text_JASP_0411.pdf

137

Users believe that our data are more accurate than ONS Mid-Year Estimates (MYEs) especially in areas of London with high population influxes and turnover. For example, in the six borough project, we found that our figures were on average 8.3% higher than the ONS 2010 MYEs but these gaps may be smaller outside London. There are no comprehensive independent comparators but our data is very close to the Child Benefit caseload whereas ONS’s for the six boroughs were not (12.5% under count). The core variables we produce are population by age and sex, household type and size, benefit and tenure status, all by geographical location. Ethnicity data are available for children attending state schools but the same information does cover other age groups in most other administrative sources. To fill this gap, we impute ethnicity of individuals and households using probabilistic methods from our extensive data base of names. By taking different snapshots of administrative data, it is possible to analyse churn and turnover at local levels which is currently a hot topic for research. For specific applications, for example in health, social care, and education, other administrative data sets or variables of interest may be linked together e.g. attendance at hospital, educational attainment. The scope of the database is therefore potentially extensive and much wider than the Census can offer. Projects can span healthcare, public health education, social care, youth crime, housing and so on. On the other hand, variables covered by the Census but not adequately covered by locally available administrative sources include: Information about occupation and place of work, self reported health, caring responsibilities and religion. In our own work, we have been able to fill these gaps to some extent using data from local surveys and linking them to administrative data. A similar approach combining survey information with administrative data could be considered at national level, possibly by using or adapting existing national surveys. Under our method data are processed in a secure environment by a third party (in this case ourselves). No data are shared between data owners and the final data base is anonymised for statistical purposes only i.e. no individuals are identifiable. Our approach is cleared by local Caldicott Guardians and by the Information Commissioner. The cost and timeliness of our approach has much to commend to the Committee. Expanding the system to cover the country would cost roughly one tenth or less of the present cost of the Census, results could be available with six months and the exercise repeated at more regular intervals such as annually. This would benefit both local users and academics whose research would instantly become more accurate, relevant and therefore useful. There is scope to reduce costs further by introducing standard legally binding data sharing protocols and generally reducing bureaucratic impediments in order to make the process more automated. Better addressing and general conformance to national

138

database BS standards etc would also help linking efficiency and reduce the need for any manual processing. I hope this brief analysis is helpful to the Committee. In the time that we have been working on this approach, we have observed steady and substantial improvements to the range and quality of administrative data. In conclusion, I would stress that our approach is practical and there is continual scope for improvement as I hope this memorandum has demonstrated. The main risks to the system would be if any if the local data sets used were scrapped or there were legislative changes that prevented their use. The government should consider therefore either securing their future access or replacing them in due course with an alternative that does the same or better job. Professor Leslie Mayhew Cass Business School, Mayhew Harper Associates Ltd. November 2011

139

Written evidence submitted by the Equality Commission for Northern Ireland (Census 40)

Executive Summary 1. The Equality Commission for Northern Ireland is concerned that the ending of the decennial Census of Population could have considerable implications for the understanding of equality related issues in Northern Ireland. 2. Although we recognise that the Office for National Statistics (ONS) ‘Beyond 2011 Programme’ is centred on assessing alternative options for producing the population and socio-demographic data required in England and Wales, we wish to respond to the inquiry as changes in England and Wales will likely have considerable implications for Northern Ireland 3. The ending of the Census will provide a number of significant challenges to social science research in the UK including the provision of robust statistics on equality-related demographics, small population equality subgroups and small area statistics. 4. In Northern Ireland, the availability of high quality local level data, (which can be reliably disaggregated by equality group) is central to the effective implementation of key pieces of equality legislation such as the Fair Employment and Treatment Order, and Section 75 of the Northern Ireland Act. 5. Surveys conducted using samples of the population, for example the Labour Force Survey, do not currently provide a robust alternative to Census based small area statistics in Northern Ireland, as small sample sizes often do not allow figures to be meaningfully disaggregated to consider equality characteristics at the local area level. 6. Administrative data in Northern Ireland would need to robustly record characteristics across a range of statutory equality (Section 75) grounds and be comprehensively cross-linked to provide any suitable alternative to the Census.

140

7. There are likely to be considerable cost implications of enhancing UK wide sample surveys (i.e. by increasing sample sizes) so as to provide robust small area estimates in Northern Ireland; and in comprehensively linking administrative records to form robust estimates of the population at the local area level. Introduction 8. The Equality Commission for Northern Ireland (‘the Commission’) is an independent public body established under the Northern Ireland Act 1998. 9. The Commission is responsible for implementing the legislation on fair employment, sex discrimination and equal pay, race relations, disability and age. The Commission’s remit also includes overseeing the statutory duties on public authorities to promote equality of opportunity and good relations under Section 75 of the Northern Ireland Act 1998 (Section 75) and the disability duties under the Disability Discrimination Act 1995. 10. The Commission, along with the Northern Ireland Human Rights Commission, has been designated under the United Nations Convention (UNCRPD) as the independent mechanism tasked with promoting, protecting and monitoring implementation of the UNCRPD in Northern Ireland. 11. The Commission welcomes the opportunity to provide evidence to the Science and Technology Select Committee Inquiry on the Census and social science, as we acknowledge that any changes to the Census methodology in England and Wales will have profound implications for the UK as a whole.

141

Questions posed by the Committee Question 1: How do social scientists use Census data? 12. The Census of the Population play a fundamental role in the provision of comprehensive and robust statistics essential for modern policy-making in Northern Ireland and is used extensively throughout the public, private and community and voluntary sectors. The detailed, reliable information provided on, for example, age distribution, spatial distribution, community background, national or ethnic origin etc. is crucially important for developing, targeting, delivering and evaluating policy and programmes and is pivotal in the planning and development of front-line service provision. 13. The Census of Population informs a range of equality relevant work in Northern Ireland – both via research undertaken by the Commission and in facilitating employers and service providers to fulfil their statutory obligations. 14. Northern Ireland has historically been at the forefront of developments in Equality legislation – including in the use of robust data to identify, monitor and challenge inequality. The Fair Employment and Treatment Order, and Section 75 of the Northern Ireland act are two examples of where high quality, local level data provided by the Census is central to their effective implementation. 15. Article 55 of the Fair Employment and Treatment (Northern Ireland) order 1998 1 requires employers to monitor the composition of their workforce and of those applying, appointed, leaving or being promoted (Article 52) and review their workforce composition at least once every three years (Article 55), “for the purposes of determining whether members of the Protestant and Roman Catholic communities are enjoying… fair participation” in their employment. This 1

The Fair Employment and Treatment (Northern Ireland) Order 1998 makes it unlawful to discriminate on the grounds of religious belief and/or political opinion in the fields of employment, the provision of goods, facilities and services, the sale or management of land or property and further and higher education. In addition to providing protection from discrimination, the legislation seeks to promote ’equality of opportunity’ (for those in, or seeking to be in employment or any occupation) and provides for ‘affirmative action’ to secure ‘fair participation’ in employment for members of the Protestant or Roman Catholic communities. http://www.legislation.gov.uk/nisi/1998/3162/article/55/made

142

consideration involves the employer comparing the community composition of their workforce to the composition of the local labour supply. Robust, local level data is central to this comparison. The Commission considers that the Census of Population is the only available data source in Northern Ireland which provides such data at the local area level. 16. Recognising the importance of Equality of Opportunity and Good Relations between different groups in Northern Ireland, Section 75 of the Northern Ireland Act 1998 2 placed a statutory obligation on public authorities in carrying out their various functions relating to Northern Ireland, to have due regard to the need to promote equality of opportunity: o between persons of different religious belief, political opinion, racial group, age, marital status or sexual orientation; o between men and women generally; o between persons with a disability and persons without; and o between persons with dependants and persons without. 17. In addition, without prejudice to this obligation, Public Authorities are also required to have regard to the desirability of promoting good relations between persons of different religious belief, political opinion, and racial group. 18. In delivering these duties, public authorities use data to undertake audits of inequalities to identify key areas within their remit on which they might exert a positive influence. Public authorities also undertake screening and equality impact assessments as part of their policy development processes. Robust local area data is not only required to identify inequalities, but to track the subsequent impacts of any associated policies. The Census of Population is the only data source in Northern Ireland which can provide robust data, by equality ground, where it is required at the local area level. 19. Beyond statutory requirements, social science research using Census data plays a vital role in increasing the understanding of 2

http://www.legislation.gov.uk/ukpga/1998/47/section/75

143

inequality and forms a significant basis for evidence based policymaking in Northern Ireland. 20. Over the years, the Commission (and its predecessor bodies) have used the Census of Population as a data source for key analyses (for example - Equality Commission (2006) Census 2001: Limiting Longterm Illness in Northern Ireland) or as part of the evidence base to inform flagship publications (e.g. Equality Commission (2007) Key Inequalities in Northern Ireland 3 ) 21. The Commissions have also supported academic work, using the Census, to consider a range of key issues over time - for example, with regards to fair employment: • Osborne, R.D.; Shuttleworth, I.G.; (eds) (2004) “Fair Employment in Northern Ireland: A Generation On”, Equality Commission / Blackstaff Press. • Cormack, R.J.; Gallagher, A.M.; Osborne, R.D. (1993); “Fair Enough? Religion and the 1991 Population Census”; Fair Employment Commission • Gallagher, A.M.; Osborne, R.D. Cormack, R.J.; (1994); “Fair Shares? Employment, Unemployment and Economic Activity”; Fair Employment Commission 22.

The Commission also draws on the work of others, for example: • Shuttleworth, I.; Lloyd, C.; (2006) Are Northern Ireland’s Two Communities Dividing?: Evidence from the Census of Population 1971-2001, Shared Space, Northern Ireland Community Relations Council.

23. Often, a key defining characteristic in the decision to use the Census of Population is not only the precision that a census provides, but that sample surveys in Northern Ireland are often insufficient to allow a detailed consideration of patterns at the local (community) level. Census data has allowed the Commission to establish 3

http://www.equalityni.org/archive/pdf/Keyinequalities%28F%291107.pdf

144

quantitative data for small population subgroups (for example, Irish Traveller, BME) and small geographical areas for which sample surveys cannot supply robust statistics 4 . 24. Census data has also played a crucial methodological role in equality-related Commission research. Census data has been used to benchmark and improve the quality of information from other data sources used by the Commission such as sample surveys. For example, census-based population estimates have been used to benchmark population-based quotas in surveys and weight data postsurvey to ensure a representative sample of key equality groupings in Northern Ireland 5 . 25. The harmonisation of the Census across UK regions has allowed the Commission to establish cross-comparisons between equality groupings in Northern Ireland and other parts of the UK. It has also provided a reliable means of tracking change across social and economic indicators across time. 26. In summary, Census based research has been used to provide at least part of the evidence-base to allow the Commission to identify and describe key inequalities in Northern Ireland; to influence policymaking in Northern Ireland, to assist in the monitoring of fair participation in the labour market; and in the evaluation of legislation and policy in Northern Ireland. As noted above, robust local area data, disaggregated by equality grounds, is also central in allowing employers and public authorities to fulfil their legislative obligations. Question 2: What impact will the ending of the Census have on social science research? 27. The ending of the Census will provide a number of significant challenges to social science research in the UK which have the potential to impact on Commission research given the small population in Northern Ireland. These challenges include the 4

See NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. Section 1.13 and 1.17 5 For example in ECNI (2008) Equality Awareness Survey 2008.ECNI: Belfast

145

provision of robust demographics on equality grounds; the provision of robust data on small population equality subgroups and at small geographical levels; disaggregation of data on equality grounds, and benchmarking for sample surveys. 28. In general terms, population data on equality grounds plays a pivotal role in policy-making and is essential for monitoring policy implementation and outcomes. Policy without valid and reliable data is potentially costly and wasteful. The ending of the Census will potentially impact on the ability to obtain data on equality grounds as at present the Census is the most comprehensive means of capturing equality-related statistics 6 . 29. Sample surveys in Northern Ireland are often insufficient to allow a detailed consideration of patterns at the local (community) level. Sampling variability due to small sample size will impact on the ability to provide disaggregated data on equality grounds. In the context of population estimates in Northern Ireland this would mean that surveys would need to be larger and would require costly booster samples to ensure appropriate coverage and reliability. 30. The issue of sample size may have wider compliance implications, for example, Article 31 of the United Nations Convention requires that data on disability should be disaggregated to assess the implementation of the Convention and to identify and address barriers 7 . In addition, the Regulation of the European Parliament and the Council on Population and Housing Censuses requires Member States to provide data at the level of Local Area Unit 2 (LAU2) equivalent to electoral ward level in Northern Ireland8 . Currently this information can only be fully captured at this geographical level by the Census 9 .

6

Except information on sexual orientation, a question on which was still not included in the 2011 Census. 7 United Nations Convention on the Rights of People with Disabilities (2007) Article 31 8 See NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. Section 1.17 9 Ibid

146

31. The small size of the total population in Northern Ireland adds a further layer of complexity. The Labour Force Survey (LFS) User Guide notes that “it is the nature of sampling variability that the smaller the group whose size is being estimated, or from which an estimate is being derived, the less precise that estimate is” 10 . Due to the relatively small size of the Northern Ireland population, population estimates from surveys may be unreliable, in particular for small population subgroups (e.g. ethnic minorities) or small geographical areas as they may be subject to a high degree of sampling variability. 32. Concern with the quality and reliability of survey statistics in the absence of a Census have been raised by the Northern Ireland Statistics and Research Agency (NISRA). NISRA has stated that sample surveys cannot provide robust statistics for small population subgroups, such as ethnic minority populations 11 and at small geographical levels such as electoral ward 12 . In presenting the business case for the 2011 Census, NISRA was of the view that “the Census was still the most authoritative source of information for a wide range of uses. It provides a snapshot of the country, with consistent and comparable information for small areas and subpopulations, and allows multivariate analyses that are not feasible using any other data source” 13 . 33. The ending of the Census may also create challenges in the provision of statistics for persons living in communal establishments (CE) that may receive limited coverage in sample surveys, such as the LFS. In 2009, the ONS stated that “the decennial Population Census is likely to remain the most reliable integrated source of CE population data” 14 .

10

ONS (2009). Labour Force Survey User guide – Volume 1: Background and Methodology. ONS: UK 11 NISRA (2004). The future provision of demographic statistics in Northern Ireland (towards the 2011 Census). Information paper. NISRA: Belfast. 12 NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. Section 1.17 13 NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. Section 1.13 14 ONS (2009). Labour Force Survey User Guide – Volume 1: Background and Methodology. ONS: UK

147

34. Other challenges in relation to the use of administrative and/or survey data include the consistency and stability of concepts used in survey and administrative data. The quality and robustness of equality data for the population is contingent on the consistency of concepts used between surveys. For example, research 15 has identified that the concept of disability used in a survey can impact on population estimates of disability. The challenge would be to ensure stability of concepts between surveys and administrative records used for the purpose of gathering census information and longitudinally across time. 35. In addition, inconsistencies between regional administrative records and surveys due to differences in the regional context have the potential to impact on the harmonisation of UK statistics and will need to be addressed. For example, the Northern Ireland version of the Family Resources Survey (FRS) and Labour Force Survey (LFS) use a different sample design and/or sampling frame than in Great Britain 16 . 36. There is a related danger that the ending of the Census may result in de-harmonisation of statistics between the UK regions and greater challenges in making robust inter-regional and/or whole UK comparisons. It is, therefore, crucial that continued close cooperation and joint working is carried out on Census policy between England and Wales and other UK regions to ensure harmonisation of any alternative arrangements 17 . Question 3: What alternatives to the Census would provide population and socio-demographic data of equivalent or higher quality? 37. We offer a limited response to this question, in seeking to confine our response to areas within our core area of expertise.

15

MSA-Ferndale (2004). Review of disability information project for DFP NISRA. DFP: Belfast Rafferty A (2010). Introduction to Complex Sample Design in UK Government Surveys. ESDS:UK 17 See NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. Section 1.16 16

148

38. We consider that the Census provides the most reliable source of population characteristics in the UK and is used to improve and quality assure information collected from other data sources such as sample surveys. Other sample surveys, for example the LFS, are comparatively less reliable, in that they provide estimates of population characteristics rather than exact measures18. 39. NISRA is of the view that “without the Census, surveys would be less reliable or would need to be larger and more costly” to ensure robustness19. Given the issues with sampling variability noted earlier, this may present challenges in relation to the small population of Northern Ireland. 40. In light of the particular challenges for Northern Ireland the Commission is of the view that should the UK government decide to discontinue the decennial Census the Northern Ireland Executive should be strongly encouraged to introduce a Northern Ireland Census. The Commission recognises that the issue of scale and cost of a Northern Ireland Census would be an important consideration, however, Ireland which has little more than double the population of Northern Ireland has a quinquennial census. Question 4: What other existing sources of population and sociodemographic data could be improved upon? 41. Surveys conducted using samples of the population, for example the Labour Force Survey, do not currently provide a robust alternative to Census based small area statistics in Northern Ireland, as small sample sizes and associated sample error often does not allow these estimated figures to be disaggregated to consider equality characteristics at the local area level. In the context of robust population estimates in Northern Ireland this would mean that surveys would need to be larger and would require costly booster samples to ensure appropriate coverage and reliability.

18

Department of Enterprise, Trade and Investment. (2006). Northern Ireland Labour Force Survey: Performance and Quality Monitoring Report. DETINI: Belfast. 19 See NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. Section 1.8

149

42. Administrative records in Northern Ireland do not always collect information on all equality-grounds covered by Section 75. Where data is collected this may be confined to a limited number of grounds (e.g. age or gender) or data may be incomplete. This is because administrative records are by design focussed on their administrative purpose and thus may not meet wider user needs or may only provide a proxy for the underlying population 20 . 43. To present an alternative to the Census, administrative records, registers would need to be comprehensively linked and crosstabulated 21 and capable of tracking change on a longitudinal basis. This may require a comprehensive population register and system of linkage, such as that used in Denmark, Finland, Norway and Sweden 22 , and such data linkages may require legislative change 23 . Research in Northern Ireland has however indicated that many such records are subject to gaps in the range or quality of data 24 . It is also likely that none of these would at present be able to provide the full range of equality demographics required by the Commission and provided for by the Census. 44. NISRA has indicated that the Census provides the basis for equality monitoring by government, in particular “census information on age, sex, ethnicity, religion and disability help to identify the extent and nature of disadvantage and to measure the success of equal opportunities policies”. 25 45. The Commission is of the view that should an alternative to the Census include administrative records, this must include the recording of equality demographics on Section 75 grounds. This level of information is not only useful to the work of the Commission and 20

NISRA (2004) The future provision of demographic statistics in Northern Ireland (Towards the 2011 Census). DFP: Belfast 21 NISRA (2004) The future provision of demographic statistics in Northern Ireland (Towards the 2011 Census). DFP: Belfast 22 See Ralphs M and Tutton P (2011) Beyond 2011: International models for census taking: current processes and development. Office for National statistics (ONS): UK. 23 See NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. 24 IJpelaar J, Marshall D, Paul S and Moylan K (2011) Quality Report / User Guide – Northern Ireland Population Estimates. NISRA Occasional Paper No. 32: NISRA 25 NISRA (2010). The 2011 Census of Population in Northern Ireland: Proposals. DFP: Belfast. Section 1.8

150

equality-related social science research in general, but would be useful to public authorities in fulfilling their statutory duties under Section 75 of the Northern Ireland Act 1998. 46. The Commission acknowledges that administrative data cannot always be completely accessible due to issues of privacy and confidentiality. If administrative records were used to gather Census data the issue of data access for research whilst maintaining the confidentiality of individual records would need to be addressed. 47. Any decision to move away from a Census of Population must also recognise that there are likely to be considerable cost implications of enhancing UK wide sample surveys (i.e. by increasing sample sizes) so as to provide robust small area estimates in Northern Ireland; and in comprehensively linking administrative records to form robust estimates of the population at the local area level. The Equality Commission for Northern Ireland December 2011

151

 

Written evidence submitted by the Local Government Association (Census 41) 

  1. Introduction

1.1. The Local Government Association is a voluntary membership body with 422 member authorities England and Wales. Together they represent over 50 million people and spend around £113 billion a year on local services. They include county councils, metropolitan district councils, English unitary authorities, London boroughs and shire district councils, along with fire authorities, police authorities, national park authorities and passenger transport authorities. 1.2. The Local Government Association welcomes this opportunity to offer written evidence to the Science and Technology Committee’s inquiry into the Census and Social Science. 2. Key Points 2.1. At a time of public expenditure constraint, the Local Government Association welcomes an examination of alternatives to the Census as timely given both the cost and the weaknesses of a ten yearly approach. However, this examination needs to recognize that reliable, timely and appropriate statistics are key tools that underpin wise, cost effective local spending decisions. 2.2. Accurate local demographic and related statistics are a fundamental enabler for local public services and, to date, local authorities have relied on analysis of the Census as one of the best available sources of such evidence. 2.3. A key priority should be cost effective, better quality and more up to date local statistics that are fit for 21st century approaches to the use of data. Although there are differences of context and requirement, there are lessons to be learned from the way that successful private sector companies do this. 2.4. Abandonment of planning for a traditional census should therefore go hand in hand with the implementation of fit for purpose alternative arrangements. Alternatives will need significant work to offer consistent and reliable data. 2.5. Local government employs a wide range of people under the broad heading of social scientists, many of whom are demographers and researchers providing local evidence to underpin service planning, investment and the engagement of citizens in these processes. It is also, both directly and indirectly, an important customer for wider social research that relies on Census data. 2.6. In welcoming this inquiry and the Office for National Statistics consultation, the Local Government Association believes that they need to

152

trigger technical dialogue to work through options, with local government recognized as one key customer of social science based on data currently supplied through the Census. 3. Detail How do social scientists use Census data? 3.1. Local authorities currently use analysis of Census data to formulate local policy, to plan and deliver local services, to allocate local resources, and to monitor performance. The Census has also provided the most consistent data for both comparison between areas and for informed and customized analysis of the needs of localities. Census currently: •

Underpins local government financial settlements: crudely, some 75% of local authority funding is centrally funded with allocation based to a significant degree on population.



Enables socio economic analysis to identify the distribution and geography of specific groups, for example, for school place planning, meeting the needs of the elderly, understanding the extent of single parent households, or meeting local housing needs.



Provides consistent trend time series data, for example (and despite some weaknesses), it is a key source of travel to work and migration flow data for long term transport and housing strategies.



Informs local authority work required by statute. Population data is important, for example in Waste, Minerals and Transport Planning, Joint Strategic Needs Assessment, Local Development Frameworks and the Survey of the Area (under existing legislation but proposed to continue in The Town and Country Planning (Local Planning) (England) Regulations [2012]).

3.2. Also: •

Census data has been a key component in commercial products that are often purchased by local authorities to provide local insight for policy and planning; and



The wide use of census data by the academic research community is also relevant to local authorities, for example where used in longitudinal studies that inform aspects of health policy and social mobility.



Census results are available for small geographies yet retain consistency nationally and locally, thus enabling examination of the area in question and its context, for example, to identify ‘hotspots’ or spatial patterns. Small area data is a building block that can be configured to illustrate parishes, towns, and district council wards. This

153

is important in supporting localism objectives such as helping neighbourhood forums to draft neighbourhood plans that local authorities will need to adopt.   What impact will the ending of the Census have on social science research?   3.3. This depends on what arrangements are put in place and ONS are yet to make firm proposals. However: •

Census has been important in providing information on which the applied social science, that is particularly important to local public services, relied.



Whilst some Office for National Statistics alternative proposals involve something like a Census but carried out on a different basis, others anticipate the replacement of the Census by the use of alternative, pre existing data sources: a number of our member local authorities are concerned about the potential negative impact if replacement arrangements provide less relevant data and / or data of inferior quality.



As described in the previous section, local authority use of Census is not statutory, but significant plans and activities that use Census data are statutory requirements. Without alternative data of equal or better quality, these requirements would be less well evidenced and less authoritative.



Census data is only published every 10 years with data processing typically taking a further 1 – 3 years so the results are rapidly out of date. More frequent surveys, or improvement and better use of administrative data on a regular basis would provide more up to date results.



In addition, such an approach would have potential to make greater and more effective use of public investment in other data sources that fall outside Census, including data on income, taxation, and second jobs.

3.4. If Census were abolished without alternative arrangements and without ‘read across’ or consistency between census data and the new arrangements, there are a number of questions raised that are illustrated by: •

The potential termination of Samples of Anonymised Records (SARs) and longitudinal studies, both of which enable in depth analysis such as that by Office for National Statistics on the ‘Golden Generation’, health inequalities and trends in family and household formation, but available much more widely to the research

154



How to sustain improvements in population measures: this remains a core requirement for public policy and services. We note that Office for National Statistics report a “successful 2011 census” with good response rates.



Household composition: we are unaware of other datasets that can provide such detailed information on households and their composition. Since family structures are becoming more complex and diverse, retaining a good source of information to understand this issue will be important.



Data for small areas: a key capability of the Census is to provide detailed information for small areas which is consistent nationally and locally, and with close to 100% coverage and therefore malleable, for example, to create data for neighbourhoods or to respond to administrative boundaries change.



Commuting and internal migration: Census currently provides information on commuting and migration, and is the only source of data on this at all for small areas. This data is obviously important to study travel patterns and for transport planning, and contains valuable information on mode of travel and type of job as well as the journey itself. ONS did release some alternative data on commuting from the Annual Population Survey, but this was only available at local authority level, and the sample size was insufficient to provide reliable estimates for smaller flows. It is not easy to think of a replacement source of local commuting data.

What alternatives to the Census would provide population and sociodemographic data of equivalent or higher quality? 3.5. This is an area where further work and engagement with local government (and more widely across the public, private and voluntary sectors) is necessary before detailed conclusions are reached. The answer lies in technical assessment of the feasibility, the affordability of options available and any ‘trade offs’ such as whether greater frequency of statistics militates against the ability to cross tabulate the current range of census characteristics against each other. 3.6. This consideration should address factors such as: •

The role of administrative datasets. These are generally designed to meet specific purposes. As a result, there are understandable inconsistencies between such data sources in terms of definitions, coverage and completeness. Some under-record populations (e.g. the electoral register) and others over-record (e.g. General Practitioner patient lists). This

155

approach is a potential option, but further work is needed to establish and compare the effort and cost needed to make such data fit for wider purposes; however •

Administrative systems such as Electoral Registers may have potential for development to improve records completion provided it is affordable. For example, General Practitioner Lists were first floated as an aid to an Improving Migration and Population and are now incorporated into the mid-year population estimates methodology.



There are a number of experiments and activities that should be evaluated as part of the process. For example, modeling approaches in relation to population projections and small area geographies, or work commissioned by some local authorities that includes Newham, Barking and Dagenham, Luton and Birmingham where they constructed local population estimates by using administrative sources.



The role of sample surveys, which have their place, but also have limitations and challenges in achieving sufficient and representative response rates.



Record matching may be required to link data in one database with data from another to produce the more complex variables, but the weakness of this procedure is the quality of the matching and key used to perform the match. Such linked but anonymised records would be of value though to the academic researcher testing hypotheses.



Other data, such as travel to work information, can only be reproduced by a survey, which if it is to produce reliable results for small areas, will need to have high percentage coverage or be repeated over a number of years to boost coverage. It may be possible to use modeling techniques to merge data from administrative sources or from surveys, or the two sources combined, to give approximate estimates but this procedure may still not be adequate for small geographic areas.

3.7. It will be important to learn from international experience, some of which is well documented: Scandinavian countries use a population register. France has a system of rolling censuses and the US combines a short-form census with surveys. None of these seem entirely successful and some, such as population registers, may raise questions about public acceptability. 3.8. We are also aware of the debate about whether and how private sector sources might offer part of the solution. This is a sensible avenue to explore, although at present, the use of central government data looks more promising (e.g. school census, DWP/HMRC information system, GP patient register). However, these offer a potential means of population count and some basic demographic and employment characteristics rather than the wider range of census variables.

156

What other existing sources of population and socio-demographic data could be improved upon? 3.9. Although not comprehensive, we suggest that the following data could be improved or made more widely accessible: •

Health data on the ageing population and disability would be useful



The availability of data on income and taxation



Second jobs



Broadband take-up



International migrants: sample sizes for international migration from the International Passenger Survey are small so it would be helpful to further explore whether e-borders data might be developed. Also, European economic migrants who stay for less than a year are not counted as residents, but because of ‘churn’ this group is often being replaced by other short-term residents, and so this group is never counted.

Local Government Association December 2011

6

157