Population Trends No. 111 - Office for National Statistics

Population Trends 111

Spring 2003

How important are intergenerational cycles of teenage motherhood in England and Wales? A comparison with France Michael S. Rendall Population and Demography Division Office for National Statistics

INTRODUCTION

Teenage fertility has fallen substantially in every Western European country except the United Kingdom. This article examines the hypothesis that repetition of teenage motherhood from mother to daughter is a major cause of the UK being the exception. A simple demographic model of fertility across generations is estimated with comparable data from England and Wales and France. The main finding is that mother-daughter repetition can account for only a minor part of the total difference in teenage childbearing between the two countries, especially over the long term. The higher teenage childbearing in England and Wales of those whose mothers began childbearing after their teenage years dominates.

The rates of teenage fertility in Britain stand well above those of other Western European countries.1 The trends, causes and consequences of teenage childbearing in Britain have been the subject of many studies2 and much policy concern.3 A frequently-expressed concern in countries with persistently high rates of teenage childbearing is that the phenomenon is self-perpetuating.4,5 That is, the daughters of teenage mothers become teenage mothers themselves, and the “cycle” thus continues from generation to generation. The present study investigates the quantitative importance of such cycles for Britain’s teenage childbearing. It does so by evaluating the demographic plausibility of the inter-generational cycle hypothesis to explain differences in teenage childbearing between France and Britain, as represented by England and Wales (here treated as one country). Comparison of England and Wales with France is of particular value firstly because of their marked divergence since the 1970s in prevalence of teenage childbearing, having followed parallel upward paths in teenage childbearing in the 1960s and early 1970s. From the late 1970s through the 1990s, sharply falling teenage childbearing in France has contrasted with persistently high teenage childbearing in Britain. Drawing a contrast between Britain and France may also have a wider generality due to the two countries’ representing two international sets of patterns (see Figure 1). France’s path of decline in teenage childbearing since the 1970s has proved to be the typical pattern of Western Europe, whereas the British pattern has been the exception. 6 Britain’s path has been similar, however, to the low- and mediumfertility countries of the English-speaking New World: United States, New Zealand, Australia and Canada.7 These countries, too, have seen levels of teenage childbearing persisting at rates well above those of continental Western Europe. The big question then is, why did England

27

National Statistics


Spring 2003

and Wales follow the Anglo-New-World pattern and not that of the rest of Western Europe? Many explanations have been offered for the high levels of teenage childbearing in the English-speaking countries. These include their higher income inequality, the orientation of their welfare states towards providing for poor and unmarried mothers while adopting a largely laissez-faire approach towards other parents, and the ambivalence in their social attitudes towards contraception, sex education and mass communication about sexuality.8,9,10 The inter-generational cycle hypothesis gives a particular form of demographic impetus to such explanations. It implies that teenage motherhood becomes a selfperpetuating cycle once begun, repeating itself from one generation to the next. Not only are daughters of teenage mothers more likely to themselves start childbearing while teenagers, and so contribute more teenage mothers in the next generation. A combination of earlier and more lifetime childbearing among teenage mothers also implies that children of teenage mothers will be disproportionately more numerous among the next generation of fertility-exposed teenagers. To date, though, no study has investigated whether such intergenerational cycle effects could be powerful enough to explain much of the very large differences in levels of teenage childbearing between the English-speaking countries and those in continental Western Europe. It is as a first step in this direction that the present study investigates the demographic plausibility of the inter-generational cycle hypothesis to explain the diverging paths of France and England and Wales. It does so first by analysing the various demographic components required for an explanation based on the inter-generational cycle hypothesis, and second by simulating teenage and non-teenage childbearing across several generations under different combinations of the French and British values of the demographic components. The study’s outline is as follows. Firstly, the historical time series from 1951 to 2000 of all births and of the number and proportions that are to teenage mothers are compared between France and England and Wales. This comparison shows that the periods of high numbers of births (“baby booms”) in England and Wales and in France in the 1960s provided the demographic conditions for an “echo-boom” of teenage childbearing in the 1980s. The proportions of all births that were to teenage mothers were also high in the 1960s and early 1970s in both countries. The twin phenomena of large numbers of births and large proportions born to teenage mothers thus provide part of the conditions for an inter-generational teenage childbearing “echo” in the 1980s and early 1990s in both countries. Secondly, the teenage childbearing of women born to teenage mothers versus to mothers aged 20 and over is compared between the two countries. A strong correlation between mother’s and daughter’s teenage childbearing is found in both France and England and Wales. This provides a further component needed for an “echo boom” of intergenerational teenage childbearing. The analysis also, however, shows levels of teenage childbearing in England and Wales that are much higher than in France both for daughters of teenage mothers and for daughters of mothers in their 20s and above. Thirdly, a population model of teenage and non-teenage childbearing is constructed. This model is estimated using a plausible background demographic structure that allows for teenage mothers to have approximately three children, versus the two children of other women, and for teenage childbearing to occur on average nine years earlier than non-teenage childbearing (at age 18 compared to age 27). These background parameters are estimated from data on the French 1950s birth cohorts of teenage and non-teenage mothers. The model is run first for France, and then with successive substitutions of teenage

National Statistics

28

Figure 1

Mean number of children born during teenage years in selected countries

Births per 1,000 women 400 1970

1980

1990

2001

300

200

100

0 USA

Austria

England and Wales

France

Denmark

Italy

Sources: Observatoire Démographique Européen, Office for National Statistics. Notes: 1. For USA, France and Italy, data for the year 2000 are used in place of data for 2001. 2. The figures are calculated as the sum of single-year age-specific fertility rates below age 20 (defining age by age at last birthday)

childbearing parameters of England and Wales. These parameters are first the historical birth trends, second the teenage fertility of daughters of teenage mothers only, and third the teenage fertility of daughters of other mothers. The resulting modelled patterns of teenage and nonteenage childbearing are projected over a period stretching from the late 1980s through 2040 to fully capture the nature of inter-generational dynamics. The primary objective of the simulations is to answer two competing questions: (1) Could France’s teenage childbearing have persisted as in England and Wales had France’s daughters of teenage mothers followed the same pattern of teenage childbearing as in England and Wales? (2) Could France’s teenage childbearing have persisted as in England and Wales only if France’s other daughters (those not born to a teenage mother) had also followed the same pattern of teenage childbearing as in England and Wales? The simulation also provides some insight into the effects of greater historical swings in England and Wales in numbers of teenage and total births and in proportions born to teenage mothers. The results of the simulations show all three factors (the teenage childbearing of those born to a teenage mother, the teenage childbearing of those born to a non-teenage mother, and the greater birth swings in England and Wales) have some influence on the divergent paths of teenage childbearing between the two countries. The most important factor, however, is the higher teenage childbearing in England and Wales of those not born themselves to a teenage mother.


Figure 2

Spring 2003

Proportions of births to a teenage mother

Per cent 12

England and Wales, unadjusted

10

8

England and Wales, adjusted France

6

4

2

0

1951

1956

1961

1966

1971

1976 Year

1981

1986

1991

1996

2000

Sources: INSEE, ONS, author’s calculations. Note: The “adjusted” series uses the French definition of age (age attained in the calender year)

TEENAGE AND TOTAL CHILDBEARING IN BRITAIN AND FRANCE OVER THE PAST 50 YEARS This part of the analysis compares national birth registration data for France and England and Wales.11,12 Figure 2 displays the annual proportions of births to teenage mothers in France and in England and Wales over the second half of the 20th century. Because Britain and France use different definitions of age, also plotted is a series that adjusts the England and Wales proportions to match the definition of age used in France. Using a common definition of age has previously been shown to be very important for accurate cross-country comparisons of teenage fertility.13 While British demographers use age at last birthday (‘completed years’), French demographers use age attained in the current calendar year. The French definition of age is used throughout this study. The effect is to reduce the levels of teenage childbearing in Britain compared with the official published series. Under the French definition, births to 19 year olds refer to those women who give birth in the year of their 19th birthday. The effect is to reclassify out of “teenage motherhood” just over half of births to women who have past their 19th birthday but not yet their 20th (that is, 19 year-old mothers according to the British definition). Both countries saw a similar pattern of increase in the proportion of births to teenage mothers from the 1950s through the early 1970s. The increase was greater, however, in England and Wales, peaking at 8.71 per cent of all births in 1972. In France, a peak of 7.18 per cent of all births was observed in 1974. The patterns of decline after the respective peaks differed greatly. In France the proportion fell to a trough as low as 1.77 per cent in 1995, compared to a much higher trough of 5.05 per cent in England and Wales in 1994. By 2000, the French proportion had recovered slightly, such that births to teenage mothers represented 2.02 per cent of all births. In England and Wales, women who were teenage mothers under the French definition represented 6.06 per cent of all births in 2000, exactly three times as high as in France. This percentage translates to 7.58 per cent of all births in England and Wales being to mothers who have yet to reach their 20th birthday (the conventional definition of a teenage mother used on the English side of the Channel).

The historical series of numbers of teenage births in the two countries are shown in Figure 3. These show a somewhat earlier, and also stronger growth in teenage childbearing in England and Wales than in France. The peak level was sustained slightly longer in France, where the number of births to teenage mothers peaked in 1973 at 60,453 births. This was almost exactly coincident with its peak proportion of births to teenage mothers, observed in 1974. In England and Wales, the peak number of births to teenage mothers occurred in 1966, at 69,348 births (again using the French definition). By 1972 in England and Wales, when the proportion of births to teenage mothers reached its peak, the number was already down to 63,225 and about to fall rapidly. By 1977, the number was 43,551, slightly below the 46,145 births to teenage mothers in France that year. From this point onwards, the decline stalled in England and Wales, with the number of births to teenage mothers fluctuating around 45,000 through until the end of the 1980s. Meanwhile, in France, the number of births to teenage mothers continued to decline sharply, so that by the end of the 1980s the number was less than half its 1977 value. The rises and falls in teenage births need to be understood also in the context of overall fertility trends (see Figure 4). Both countries experienced historically high annual numbers of births in the 1960s (“baby booms”). However, the 1960s births in England and Wales stand far above the annual numbers of births either before or after that decade. In France, the 1960s stand out much less. The much greater increase in numbers of birth from the 1950s to the 1960s in England and Wales compared to France simultaneously provided especially large absolute and relative numbers of teenagers at risk of childbearing in the 1980s in England and Wales. There was then a much stronger “baby bust” in England and Wales in the 1970s than in France. This will have resulted in a bigger fall in numbers of teenagers in the late 1980s and 1990s in England and Wales than in France, and so a downward impact on numbers of births to teenage mothers. Further, the 1970s baby bust in England and Wales will have increased the ratio of women in their 20s to those in their teens in the late 1980s and 1990s, pushing down even more the proportion of births to teenage mothers. France’s smoother pattern of rises from the 1950s and falls into the 1970s, on the other hand, will have led to a more even balance between numbers of teenage women and women in their 20s and 30s in the 1980s and 1990s.

29

National Statistics


Figure 3

Spring 2003

Number of births to teenage mothers

Teenage births 100 90 80

Thousands

70 60

England and Wales, unadjusted

50 40 England and Wales, adjusted

30 20

France

10 0

1951

1956

1961

1966

1971

1976 Year

1981

1986

1991

1996

2000

Sources: INSEE, ONS, author’s calculations. Note: The “adjusted” series uses the French definition of age (age attained in the calender year)

Figure 4

Total number of births

Total births 1,000

Thousands

900

800 France

700 England and Wales

600

500 1951

1956

1961

1966

Sources: INSEE, ONS

National Statistics

30

1971

1976 Year

1981

1986

1991

1996

2000


The most important feature of this analysis of historical trends, however, is the coincidence in both countries of historically high numbers and proportions of births to teenage mothers in the 1960s and early 1970s. Both countries could thus be described as having been ripe for an “echoboom” of teenage childbearing to the daughters of the teenage mothers of the 1960s and early-1970s. For this to have occurred, though, the bulge in daughters of teenage mothers reaching their teens in the 1980s and early 1990s would need to have been combined with strong inter-generational correlations of teenage childbearing.

INTER-GENERATIONAL CORRELATIONS OF TEENAGE CHILDBEARING Data and methods used to estimate the teenage motherhood of daughters of teenage and other mothers, in France and Britain, are described in Boxes one and two. Results are presented in Table 1. The likelihood of having a teenage birth in Britain is consistently 2 to 2.5 times higher for a daughter of a teenage mother. This holds whether the definition of a teenage mother based on age at first childbearing (as in the DiSalvo and Rosato studies described in Box two) or at the birth of the current daughter (as in the present study) is used. Significantly, this ratio of 2 to 2.5 times higher for a daughter of a teenage mother holds also for France. The levels of teenage childbearing in Britain are such that 26.1 per cent of daughters born to a teenage mother will themselves become a teenage mother. As many as 10.9 per cent of daughters born to a non-teenage mother will also become a teenage mother, however. This is higher than the percentage of teenage mothers among French daughters of teenage mothers (7.9 per cent). Only 3.7 per cent of daughters of non-teenage mothers in France have a child as a teenager. While the inter-generational teenage childbearing correlation is much higher in England and Wales in France more than three-quarters of daughters born to a teenage mother will not themselves become a teenage mother. In this sense, the inter-generational teenage fertility rates indicate a very weak degree of repetition from mother to daughter. Also noteworthy is that in France, rates for both daughters of teenage and other mothers are only one third as high as those in England and Wales. Further, daughters of non-teenage mothers in England and Wales are more likely to become teenage mothers than are daughters of teenage mothers in France. The present study’s LS results before adjusting for non-linkage are compared to those of the two previous studies cited above that use the LS. The results of the present study are seen to be very close to both those studies, despite the different time periods and definitions of being born to a teenage mother. The 1956–61 cohorts of Rosato’s study saw mean numbers of teenage children born to a teenage mother that are very close to the 1972–78 cohorts of the present study, while those for the 1966–71 cohort are somewhat lower. The 1956–65 cohorts of DiSalvo’s study saw proportions becoming teenage mothers among daughters of teenage and non-teenage mothers that are very similar to the proportions estimated for the present study’s 1972–80 birth cohorts. These results suggest there has been little change in the strength of the inter-generational correlations over the last two decades in England and Wales. The large sample sizes of daughters of teenage mothers in both the LS and EDP are methodologically noteworthy. In the EDP sample, 997 of the 1969 to 1975 French-born female cohorts present in the 1982 and 1990 Censuses were born to a teenage mother (that is, a teenager at the time of the EDP member’s birth). While 352 of the 1969 to 1975 cohorts became teenage mothers (that is, had any births as a teenager), only 56 of those were daughters of the 997 who were born to teenage mothers. Clearly then, a large dataset is required to obtain sufficient observations of such a demographically rare phenomenon as

Spring 2003

Box one THE ENGLAND AND WALES LS AND FRENCH EDP DATA SOURCES

The data source used to estimate the inter-generational teenage childbearing correlations for England and Wales is the ONS Longitudinal Study 14 (LS). In France, it is the French Demographic Panel, or EDP (Echantillon Démographique Permanent). 15 The structure and content of these two datasets are very similar. Both link women’s birth registrations, including of their own birth, and link them also to their Census records. In the files used for the present study, births to EDP members in the years 1968 to 1995 have been linked to their 1968, 1975, 1982, and 1990 Census records, and to their own birth record if after 1968. In the LS, 1971 to 1998 years of births are linked to LS members’ 1971, 1981, and 1991 Census records, and to their own birth record if after 1971. The EDP consistently links births for only half of their EDP sample, resulting in a 1 in 200 sample that may be used in fertility analyses. The LS links births for all LS female sample members, resulting in a 1 in 100 sample. The large number of observations in the LS and EDP greatly facilitates the study of childbearing among small subpopulations such as the daughters of teenage mothers. The long-running, panel nature of the data, moreover, permits the identification of intergenerational fertility. The LS and the EDP are especially useful for comparative research due to the similar length of time for which their data are compiled, and to the similarities of the two countries’ populations and their socioeconomic conditions. Previous studies have demonstrated their value in comparing migration 16, households 17 and fertility. 18

inter-generational teenage motherhood. Of the LS sample, 1,543 were born to a teenage mother. As many as 1,889 of the LS sample themselves became teenage mothers, 332 of them had teenage mothers. These higher numbers in the LS reflect both the persistently high levels of teenage childbearing across British birth cohort and that, procedurally, approximately twice as many LS females as EDP females have their birth records linked to their census records.

A MOTHER-TO-DAUGHTER MODEL The inter-generational teenage childbearing correlations alone do not tell us whether inter-generational teenage fertility is of major importance for overall levels of teenage childbearing. To evaluate this, a model that incorporates the teenage childbearing of both daughters of teenage mothers and of daughters of older mothers is necessary. This is especially important given that, even in a population like England and Wales where teenage fertility is high, far more women are born to nonteenage mothers than are born to teenage mothers. For the model to adequately represent other demographic features of teenage motherhood, it is also necessary to take into account the faster reproduction that occurs when teenage childbearing in one generation is followed by teenage childbearing in the next, and for the higher completed fertility observed for teenage mothers than for women who begin childbearing later (or not at all). With these considerations in

31

National Statistics


Spring 2003

Table 1

Box two LS AND EDP SAMPLES, DEFINITIONS AND ADJUSTMENT FOR NON-LINKAGE To estimate the intergenerational teenage fertility correlations, the present study uses 18,884 LS females born between 1972 and 1978, and compares them to 12,568 EDP members born between 1969 and 1975. Criteria for inclusion in the sample are that they are born in the country, and were present in the 1980s and 1990s Censuses (1981 and 1991 for Britain and 1982 and 1990 for France). Neither the LS nor EDP data, however, can be used accurately without dealing with the problem of incomplete linkage of births. An investigation of the quality of the LS fertility data conducted by Babb and Hattersley,19 following an earlier study by Werner, 20 found a linkage rate of between 83 per cent and 88 per cent in the 1981 to 1988 period. Analyses conducted for this study indicate that this linkage rate has been similar in the 1990s. Analyses of the EDP’s overall success in linking birth records conducted by the author for the present study found similar overall rates of non-linkage to those in the LS. The age pattern of omission is different in the EDP, however, with greater rates of omission at younger ages, and especially the teens. Thus it is important to apply appropriate corrections differentially across the two countries to account for these differences in non-linkage. This is done here assuming no interaction between undercount and mother’s teenage versus non-teenage age at the daughter’s birth. In defining “born to a teenage mother”, the birth record of the LS or EDP member herself is used. This identifies whether the mother was a teenage mother at the LS or EDP member’s own birth. This contrasts with definitions used by previous LS studies, which include also daughters of non-teenage mothers who started childbearing as teenagers. 21,22 DiSalvo examined the teenage fertility of LS females born in the years 1956 to 1965. Rosato examined the teenage fertility of the 1956 to 1961 and 1966 to 1971 LS birth cohorts. To infer whether the mother began her childbearing as a teenager, DiSalvo used 1971 marital fertility histories, while Rosato used the oldest child present in the 1971 and 1981 Censuses. These methods are not easily applied to the EDP data, due to the lack of family and household information available in the EDP from the 1968 Census. Thus the definition based on mother’s age at the LS or EDP member’s own birth is used to achieve comparability between the two countries. The more inclusive, “ever-teenage-mother” definition of DiSalvo and Rosato, however, is used in the simulation model, as is explained in that section.

mind, a simple model of inter-generational teenage and non-teenage fertility that takes into account each of these factors is specified. The model is designed to allow for three paths of influence of intergenerational teenage fertility on overall teenage fertility. The first (‘repetition’) effect is through a higher probability of teenage childbearing among daughters born to ever-teen-mothers than among daughters born to never-teen-mothers. The second (‘timing’) effect is through a faster inter-generational rate of reproduction among teenage childbearers. The third (‘quantity’) effect is through a greater number of children born to ever-teenage mothers.

National Statistics

32

Teenage fertility rates of daughters of teenage and non-teenage mothers in England and Wales and France

A. Mean number of teenage births per woman England and Wales (LS)

France (EDP)

Rosato study’s 1956–61 cohorts

Rosato study’s 1966–71 cohorts

present study’s 1972–80 cohorts

nonlinkageadjusted 1972-80 cohorts


nonlinkageadjusted 1969–76 cohorts

Cohort members with a teenage mother

0.244

0.217

0.252

0.261

0.054

0.088

Cohort members with a non-teenage mother

0.120

0.096

0.101

0.109

0.024

0.039

B. Percentage of women giving birth before age 20 Britain (LS)

Cohort members with a teenage mother (Sample size) Cohort members with a non-teenage mother (Sample size) Sources:

France (EDP) nonlinkageadjusted 1972–80 cohorts


nonlinkageadjusted 1969–76 cohorts

DiSalvo study’s 1956–65 cohorts


20.4 (5,896)

21.5 (1,543)

23.8

5.0 (1,119)

7.9

10.0 (33,554)

9.0 (17,341)

9.9

2.0 (13,146)

3.7

Rosato, M. (1999) and DiSalvo’s, P.M. (1992) (see citations in footnotes 21 and 22); Author’s calculations from the EDP and from LS tabulations provided by the LS Support Group.

Notes: 1 Age is defined as age attained in the year in the present study, while it is defined in completed years in Rosato’s and DiSalvo’s studies. Whether mother was a teen mother is defined according to mother’s age at LS or EDP daughter’s birth in the present study, but according to mother’s age at presumed first childbearing in Rosato’s and DiSalvo’s studies. 2. Mean number of teenage births was calculated by multiplying Rosato’s annual teen birth rates by five (the number of years of exposure between ages 15 and 19). For the present study, it was calculated from real cohort experience.

The modelling of the repetition effect takes advantage of the parameters estimated from the EDP and LS data as described in the previous section. The model, however, employs the less strict definition of intergenerational teenage motherhood to allow daughters of ever-teenagemothers to become teenage mothers at higher rates than never-teenage mothers, even for daughters born while the mother was no longer in her teens. These daughters of ever-teenage-mothers are modelled to become teenage mothers at the rates of daughters born to mother who was a teenager at the time of this daughter’s birth. This allows for the estimation of an upper bound of the influence of repetition of teenage childbearing on overall teenage childbearing. The timing and quantity effects of teenage childbearing in the model, on the other hand, are given more of a background role. That role is simply to allow for the effects of the large difference Britain’s and France’s between ‘repetition’ rates to play out in overall teenage and non-teenage childbearing. To incorporate timing differences into the model, the average ages of teenage and non-teenage childbearing are set to 18 and 27 respectively. These are approximately the French averages over the 1978 to 1995 period. From a modelling perspective, their being multiples of nine conveniently allows for a nine-year projection interval. Births in nine-year period t are projected by


applying the teenage fertility rates to births in period t-2, and the nonteenage fertility rates to births in period t-3. The non-teenage childbearing of teenage childbearers in period t is projected by applying to the teenage mothers of period t-1 the constant mean number of children in a teenage mother’s non-teenage years. The assumptions for this component are clearly approximate only. The mean number estimated from French data is used equally for France and for England and Wales. Further, the average age of non-teenage fertility does not vary in the model according to whether the woman began childbearing as a teenage mother, whereas it is likely that ever-teenage mothers would also have lower average ages of non-teenage fertility. Thus the timing effect of inter-generational teenage fertility will be somewhat understated in the model. The largest reproductive age difference, however, will clearly be that between age at the teenage childbearing of ever-teenage mothers and age of (non-teenage) childbearing of never-teenage mothers. This difference is captured. With respect to ‘quantity’ effects of inter-generational teenage childbearing, the model incorporates the mean numbers of teenage births of a teenage mother respectively in Britain and France, and the mean numbers of non-teenage births respectively of ever-teenagemothers and of never-teenage-mothers (including those who remain childless) in France. These are estimated from the mean numbers of children born between ages 20 and 39 (the presumed oldest age of childbearing) among the EDP sample of women who were under 20 in 1969 to 1975. Thus these women are from the earlier, 1950s birth cohorts. Results are again adjusted for estimated non-linkage of births in the EDP. Among women who respectively were and were not born to a teenage mother, 1.866 and 2.054 non-teenage births are thereby estimated. Since the average number of teenage births to an ever-teenmother in France is estimated at 1.097, an ever-teen-mother will have a mean lifetime total of 2.963 births, or almost one lifetime birth in excess of that of a never-teen-mother (2.054). These lifetime mean numbers of children may be weighted by the proportions born respectively to a teenage mother and to an older mother among the 1950s cohorts in the two countries. This produces an aggregated 2.074 children per woman in France and 2.088 children per woman in England and Wales, the small difference being generated by the slightly larger percentage of teenage births in the 1950s in England and Wales. These are reasonably comparable to the estimates of completed fertility for the 1950s cohorts in the two countries. These were 2.13 children per woman in France and 2.02 in England and Wales.23 Further, the mean ages of childbearing of the 1950s cohorts in the two countries were close to the age 27 that the projection model assumes for the more than 90 per cent of children born in each country to a non-teenage mother. Subsequent cohorts in both countries, and especially in England and Wales, have seen trends towards later and fewer children. Because of these differences between the fertility parameters assumed for the projection model and the true fertility parameters, and because of the very simplified structure of the simulation model itself, none of the results of the projections can be interpreted as forecasts. Instead their value is in their interpretation as simulations that capture the main dynamics of teenage and non-teenage fertility from generation to generation, and of how those dynamics are influenced by differences in the teenage fertility of women respectively born to teenage mothers and non-teenage mothers. Inter-generational childbearing is simulated over a period of 54 years, from 1987 to 2040. The period of 1951 to 1986, divided into four nineyear subperiods, is used to establish the simulation model’s initial conditions. These are mothers’ teenage and non-teenage childbearing statuses in France or in England and Wales. Non-teenage births to ever-teenage-mothers are estimated by applying to the teenage births of

Spring 2003

each subperiod the ratio of non-teenage births to teenage births estimated from the French 1950s cohorts as described above. Projection over subsequent generations of women is then performed from the 1960–68, and 1969–77 and 1978–86 base populations for six nine-year subperiods from 1987–95 through to 2032–40. The model is ‘single-sex’ in that no effect of bearing sons as a teenager is allowed for. Only daughters are considered. This has no direct distorting effect on the teenage proportion of all fertility, however, as only daughters are considered also among non-teenage births. Finally, the model takes no account of socioeconomic correlates of teenage and non-teenage fertility. While these will clearly play a strong role in influencing the teenage and non-teenage fertility parameters used in the model, just how they do so is beyond the scope of the present study.

Simulation results to 1987–1995 and to 1996–2004 The first set of simulation results are presented in Figure 5 simulations I and II. Here the purpose is to evaluate the role of the differences in historical birth trends between the two countries. These are primarily the historically-higher proportions of teenage mothers in England and Wales than in France, and the larger swings in birth numbers around the peak 1960s decade of childbearing in England and Wales. The simulation model is first seen to reproduce reasonably well the observed proportions of births to teenage mothers in France (simulation I). The 2.18 per cent and 2.16 per cent projected proportions born to a teenage mother are approximately comparable to the 2.32 per cent and 2.03 per cent observed teenage births of Figure 2 (using the mid-period year 2000 to approximate the 1996–2004 period). The majority of teenage mothers are the daughters of women who were not themselves teenage mothers: 1.53 per cent both in the 2.18 per cent of 1987–95 and 2.16 per cent of 1996–2004. Those teenage mothers who were born to a mother in her teens make up only 0.27 per cent and 0.16 per cent of all births in the respective periods. Adding those teenage mothers born to a mother who began childbearing as a teenager gives an extra 0.39 per cent and 0.47 per cent of daughters of teenage mothers who may be counted among all teenage mothers in 1987–95 and 1996–2004. Substituting the French base population with that of England and Wales (simulation II) results in the daughters of teenage mothers accounting for a higher proportion of all teenage mothers. This is due to the higher proportions of teenage mothers among all mothers in the 1960s, 1970s and 1980s base populations in England and Wales. Daughters of teenage mothers in 1987–95 and 1996–2004, however, still only account for 0.28 per cent and 0.27 per cent of all births in those respective periods on the strict definition (of having been born to a mother in her teens), and add only an extra 0.49 per cent and 0.58 per cent of daughters of teenage mothers when those teenage mothers born to a mother who began childbearing as a teenager are also included. The effect on the overall percentage of teenage mothers of substituting in the England and Wales base populations, however, is small and of different magnitude from one period to the next. For 1987–95, it results in a smaller overall percentage of teenage mothers (only 1.99 per cent of all births) than that produced when using the French base population (2.18 per cent of all births). This is due to the sharper decline in total births in the 1970s in England and Wales (seen in Figure 4), which led to a lower ratio of teenagers to women in their 20s in 1987–95 than in France in the same period. The England and Wales teenage childbearing rate of daughters of teenage mothers is next substituted into the model. Results are presented in simulation III. This involves the replacement of the 0.086 children on average born to daughters of teenage mothers in France

33

National Statistics


Figure 5 Per cent 8

Spring 2003

Simulated proportions of births to teenage mothers, by own mother’s age at childbearing, 1987 to 2004

Simulation II France with England and Wales’ 1960s, 70s and 80s births

Simulation I France

7

Simulation III France, with teenage fertility rate of England and Wales’ daughters of teenage mothers

Simulation IV France, with teenage fertility rate of England and Wales’ daughters of teenage and non-teenage mothers

6 5 4 3 2 1 0 1987–95

1996–2004

1987–95

1996–2004

1987–95

1996–2004

1987–95

1996–2004

Year

Own mother was never a teenage mother

Own mother was a teenager at this daughter’s birth

Own mother began childbearing in teenage years, but was no longer a teenager at this daughter’s birth ("ever–teenage mother")

Sources: Author’s simulations from INSEE and ONS data sources (see text)

with the 0.261 children on average born to daughters of teenage mothers in England and Wales. The substitution is done using the French historical base population. The result is a tripling of the percentages of births to daughters of teenage mothers in both 1987– 1995 and 1996–2004, following directly from the England and Wales rate being triple that of France’s. The numbers involved are still relatively small, these inter-generational teenage births amounting to only 1.98 per cent of all births in 1987–1995 and 1.93 per cent of all births in 1996–2004. The total percentage of births to teenage mothers in France (including those of daughters of never-teenage-mothers) would then have been 3.51 per cent in 1987–1995 and 3.46 per cent in 1996–2004. These are considerably above the 2.32 per cent observed in France for 1987–1995 and the 2.16 per cent for 1996–2004, but still far below the England and Wales percentages of 5.94 per cent in 1987– 1995 and 6.06 per cent in 1996–2004. In the final simulation IV, the England and Wales teenage childbearing of daughters of never-teenage mothers (that is, those born to a mother whose childbearing began in her 20s or later) is additionally substituted in the model. This involves the replacement of the 0.040 children on average born to daughters of such mothers in France with the 0.109 children on average born to daughters of never-teenage mothers in England and Wales. This substitution, again done using French historical population parameters, makes a bigger difference to overall proportions of teenage births. Just as the previous substitution tripled the percentage of inter-generational teenage births, this substitution triples the percentage of other teenage births. The latter now amount to 4.01 per cent of all births in both 1987–1995 and 1996–2004. The total percentage of births to teenage mothers in France (including also intergenerational teenage births) would then have been 5.93 per cent in 1987–1995 and 5.87 per cent in 1996–2004, very similar to the observed England and Wales percentages of 5.94 per cent in 1987–1995 and 6.06 per cent in 2000. Thus substituting in the England and Wales teenage childbearing rates, of both daughters born to teenage mothers and daughters born to non-teenage mothers is sufficient, when applied to the French 1960s, 1970s and 1980s base population, to bring

National Statistics

34

France’s proportion of births to teenage mothers from the mid-1980s to mid-2000s up to the levels observed in England and Wales. This result follows from France and England and Wales having had similarly favourable conditions for an echo boom of teenage childbearing in the 1980s and 1990s. When the Franco-British differences that are due to the rates of teenage childbearing respectively of daughters of teenage and nonteenage mothers are apportioned out, both are seen to contribute significantly to the overall difference in late 1980s to early 2000s teenage childbearing. In approximate terms, the difference between the 2 per cent of births to teenage mothers in France and the 6 per cent in England and Wales is bridged by about 1.5 per cent added when the daughters of ever-teenage mothers in France are given the England and Wales teenage fertility rates, and by about 2.5 per cent that is added when the daughters of never-teenage mothers in France are given the England and Wales teenage fertility rates. Thus England and Wales’ teenage childbearing that is not a repetition from one generation to the next accounts for the majority of the difference, but inter-generational repetition was nevertheless a substantial contributor to the overall difference.

Simulation results from 2005–2013 to 2032–2040 Figure 6 presents the teenage childbearing simulations that are projected out from 2005–2013 through to 2032–2040. There are three different simulations, all using the French base populations. The simulations are numbered consistently with Figure 5. In simulation I, the model is run with French parameters (base populations and teenage childbearing rates) throughout. The England and Wales teenage childbearing rate of daughters of teenage mothers is substituted in the model whose results are presented in simulation III. The England and Wales teenage childbearing of daughters of never-teenage mothers is additionally substituted in the model whose results are presented in simulation IV.


The main finding from these analyses is of the greater role of the teenage childbearing of daughters of non-teenage mothers in explaining the long-term trend of lower teenage childbearing in a population with France’s levels of teenage fertility as contrasted with those in a population with the levels of teenage fertility of England and Wales. France’s teenage-mother proportions of all births over 1987–2004 are sustained into the future by the continued application of the same teenage and non-teenage fertility rates into the future. The periods 2014–22 and 2023–31 each see a slightly lower 1.96 per cent of all births, while 2.30 per cent and 2.28 per cent are projected for 2005–13 and 2032–40. This three-period cycle of higher teenage fertility arises from the echo 27 years later of the birth swing that saw large numbers of French teenagers in the 1978–86 period. This swing is partly an artifact of the model structure, whereby all non-teenage births are forced to occur at age 27. The meaningful part of the swing, though, is that it shows that fluctuations in teenage fertility in a French-type population are not related to inter-generational correlations of teenage childbearing, but instead to the fluctuations in numbers of teenagers in a given period, irrespective of whether they were born to a teenage or (much more commonly) a non-teenage mother. When only the daughters of ever-teenage mothers in France are given the British teenage fertility rates (simulation III), the percentage of teenage mothers falls from that modelled for the 1987–1995 and 1996– 2004 periods (as seen in Figure 5). The periods 2014–22 and 2023–31 see 2.67 per cent and 2.68 per cent of all births to teenage mothers, while 3.08 per cent and 2.94 per cent are projected for 2005–13 and 2032–40. These falls from the 3.51 per cent in 1987–1995 and 3.46 per cent in 1996–2004 in this same simulation are entirely due to a fall in inter-generational teenage childbearing. The 1.98 per cent of all births in 1987–1995 and 1.93 per cent of all births in 1996–2004 that are teenage births to a daughter of an ever-teenage-mother fall steadily from 1.09 per cent in 2005–13 down to 0.92 per cent in 2032–40. That is, even with British rates of mother-daughter repetition of teenage childbearing, only 1 per cent of all births each year would be to

Figure 6 Per cent 8

Spring 2003

daughters of a teenage mother were France’s rates of teenage motherhood among the daughters of non-teenage mothers to apply. When the daughters of never-teenage mothers in France are additionally given the British teenage fertility rates (Simulation IV), the overall percentage of teenage mothers is then sustained at approximately the England and Wales levels observed in the 1987–1995 and 1996–2004 periods. The periods 2014–22 and 2023–31 each see approximately comparable levels of 5.91 per cent and 6.02 per cent of all births, while 7.06 per cent and 7.02 per cent of 2005–13 and 2032–40 are produced by the continuation of the 27-year-apart echo of higher total numbers of teenagers at risk. Approximately two-thirds of these mothers of teenage children are born to never-teenage mothers, and one-third born to everteenage mothers. As an important caveat, the majority of this latter quantity are births to daughters whose mother began childbearing as a teenager but who was no longer a teenager at this daughter’s birth. The mean numbers of such births per woman who began childbearing as a teenager were estimated from French teenage mothers only. An earlier study24 in England and Wales estimated 26.5 per cent of all teenage births being to a daughter of an teenage mother under somewhat less favourable conditions for an inter-generational echo of intergenerational childbearing. Thus the proportion of the present study appears to be reasonable, though perhaps a little high. Taking as typical the periods 2014–22 and 2023–31, in approximate terms the difference between the 2 per cent of births to teenage mothers in France and the 6 per cent in England and Wales is bridged by about 0.7 per cent added when the daughters of ever-teenage mothers in France are given the British teenage fertility rates, and by about 3.3 per cent that is added when the daughters of never-teenage mothers in France are given the England and Wales teenage fertility rates. Thus England and Wales’ teenage childbearing that is not a repetition from one generation to the next accounts for about four-fifths of the difference, and inter-generational repetition only one-fifth of the difference between the two countries.

Simulated proportions of births to teenage mothers, by own mother’s age at childbearing, 2005 to 2040

Simulation I France

Simulation III France, with the teenage fertility rates of England and Wales’ daughters of teenage mothers

7

Simulation IV France, with teenage fertility rate of England and Wales’ daughters of teenage and non-teenage mothers

6 5 4 3 2 1 0 2005–13

2014–22

2023–31

2032–40

2005–13

Own mother was never a teenage mother

2014–22 2023–31 Year

2032–40

Own mother was a teenager at this daughter’s birth

2005–13

2014–22

2023–31

Own mother began childbearing in teenage years, but no longer a teenager at this daughter’s birth ("ever–teenage mother")

Sources: Author’s simulations from INSEE and ONS data sources (see text)

35

National Statistics

2032–40


Spring 2003

While the decline in the total teenage childbearing contribution of everteenage-mothers would be slowed somewhat by inter-generational teenage childbearing at British levels, eventually only the high teenage childbearing also of daughters of non-teenage mothers is able to prevent the decline in importance of inter-generational teenage childbearing. This latter result demonstrates that while England and Wales’ intergenerational fertility dynamics make inter-generational teenage motherhood a significant contributor to the high overall levels of teenage childbearing, its significance derives to a large extent from interaction with England and Wales’ high levels of teenage fertility among daughters of non-teenage mothers.

SUMMARY AND CONCLUSION This study set out to investigate the quantitative importance of intergenerational cycles of teenage childbearing (where daughter’s teenage childbearing follows mother’s) for Britain’s high overall teenage childbearing. It did so by evaluating the demographic plausibility of the inter-generational cycle hypothesis to explain differences between England and Wales and France. The French definition of age was used throughout this study, and this needs to be borne in mind when interpreting and comparing this study’s numerical results. The effect of the French definition is to reduce the levels of teenage childbearing as compared to those of published tabulations. Under the French definition, births to 19 year olds refer to those women who give birth in the year of their 19th birthday. This means, for example, that the 6.06 per cent of all births to teenage mothers in England and Wales as presented here translates to 7.58 per cent under the usual British definition of the teenage years. A review of the historical numbers of teenage births in the context of all births illustrated similar patterns in the two countries. Each saw “baby booms” in the 1960s that coincided with high proportions of teenage mothers among all births. These high teenage-mother proportions continued into the early 1970s in both countries. The twin phenomena of large numbers of births and large proportions born to teenage mothers thus provided part of the conditions for an inter-generational teenage childbearing “echo” in the 1980s and early 1990s in both countries. By 2000, however, the two countries had taken very different paths, such that the proportion of teenage mothers among all births were three times higher (at 6 per cent) in England and Wales than in France (at 2 per cent).

In the long run, only about 1 per cent of all births in England and Wales would be inter-generational teenage births – that is, births to a teenage mother who was herself born to a teenage mother – if it wasn’t also for the higher teenage fertility in England and Wales among daughters of non-teenage mothers. With the higher teenage fertility in England and Wales among daughters of non-teenage mothers, about 2 per cent of all births in England and Wales, or up to one third of all teenage births, would be inter-generational teenage births. Only with the ongoing replenishment of the pool of teenage mothers from own mothers who began childbearing after their teenage years, however, can the 2 per cent level be sustained instead of dropping to a 1 per cent level. Without the ongoing replenishment of teenage mothers from families where the mother started childbearing after her teenage years, the overall proportion of teenage births in England and Wales would likely fall to a long-term level just under 3 per cent of all births (compared with around 2 per cent in France). This suggests that efforts to reduce the numbers of inter-generational “repeaters”, even if they were to be effective for this subgroup, would not do very much to reduce the overall level of teenage motherhood in England and Wales from its current 6 per cent of all births. To be effective on a substantial scale, efforts must reduce the teenage childbearing of a wider risk group that includes the daughters of nonteenage mothers. While inter-generational teenage childbearers are a significant group, they do not by themselves sustain more than a minor part of all teenage childbearing.

Key findings ●

●

To analyse how the Franco-British divergence occurred, simulation of teenage and non-teenage childbearing across generations was performed. Its results were described in two parts. First, the childbearing of the mid-1980s to mid-2000s was projected alternately from the French and British base populations (births to teenage and non-teenage mothers) of the 1960s, 1970s and 1980s, and alternately substituting in the British teenage childbearing rates of daughters of teenage and non-teenage mothers for the French rates. Second, the simulations were projected out a further 36 years into the 21st century to fully capture the dynamic features of inter-generational teenage childbearing and other teenage and non-teenage childbearing. Results showed that in the short run, inter-generational teenage childbearing (the daughters of teenage mothers becoming teenage mothers themselves) can account for about one-third of the 4 per cent difference in proportion of teenage mothers in all births (2 per cent versus 6 per cent) between France and England and Wales. In the long run, however, inter-generational childbearing could account for only about one-fifth of the difference. The larger contribution in the shortrun can be interpreted as being due to the interaction of England and Wales’ higher teenage childbearing of daughters of non-teenage mothers with inter-generational teenage childbearing. This interaction is explained as follows. National Statistics

36

●

●

●

The 1960s and early 1970s saw peaks in numbers and proportions of births to teenage mothers in both England and Wales and France. These provided conditions favourable to an “echo” occurring 20 years later, in the form of greater numbers of births to teenage mothers and to the teenage daughters of teenage mothers (“inter-generational teenage births”) in both countries. In both England and Wales and France, the daughters of teenage mothers are more than twice as likely to become teenage mothers themselves than are the daughters of mothers in their 20s or older (“nonteenage mothers”). The daughters of both teenage and non-teenage mothers in England and Wales were more than three times more likely to become teenage mothers than were the daughters respectively of teenage and nonteenage mothers in France. The above combination of conditions resulted in a teenage childbearing “echo” occurring in England and Wales, but not in France. Up to one third of all teenage childbearing in England and Wales in the late 1980s through 1990s may have been to the daughters of teenage mothers. These still constitute, however, only about two per cent of all births. In the long-term, inter-generational teenage births would fall to only one per cent of all births, and overall teenage childbearing would halve, if it were not for the higher levels of teenage childbearing of the daughters of non-teenage mothers in England and Wales than in France.


ACKNOWLEDGEMENTS The author is grateful to Myer Glickman, Heather Joshi, David Pearce and Steve Smallwood for helpful comments on an earlier draft. Longitudinal Study (LS) analyses presented here use data provided by Kevin Lynch then of the LS Support Group at the Institute of Education, from data collections and files produced by the Office for National Statistics (ONS). The equivalent French analyses presented here use French Demographic Panel (EDP) data collected and compiled by the Institut National de la Statistique et des Etudes Economiques (INSEE), who generously provided the author with the on-site resources for access to those data. Historical numbers of teenage and all births in France were provided by Fabienne Daguet and Isabelle Robert-Bobée at INSEE. The teenage fertility rates for use in adjusting the LS estimates to the national teenage fertility rates were supplied by Alec Ross of the Centre for LS Information and User Support at the London School of Hygiene and Tropical Medicine. The data for the cross-national comparisons of teenage fertility in Figure 1 were provided by Jean-Paul Sardon at the Observatoire Démographique Européen. Detailed tabulations of teenage fertility in England and Wales were provided by Denis Till. Funding was provided by the Nuffield Foundation and the Institut National d’Etudes Demographiques (INED). None of these agencies or persons bear any responsibility for the analyses and interpretations presented here.

REFERENCES

Spring 2003

16. Cribier F and Kych A (1993). A comparison of retirement migration from Paris and London. Special LS issue of Environment and Planning A 25(10), pp. 1399–1420. 17. Hall R, Ogden PE and Hill C (1999). Living alone: Evidence from England and Wales and France for the last two decades, in McRae S (ed.) Changing Britain: Families and Households in the 1990s. Oxford University Press: Oxford. 18. Ekert-Jaffé O, Joshi H, Lynch K, Mougin R and Rendall M S (2002). Fertility, Timing of Births and Socioeconomic Status in France and Great Britain: Social Policies and Occupational Polarisation. Population-E 57(3), pp. 475–508 19. Babb P and Hattersley L (1992). An examination of the quality of LS data for fertility analysis. LS User Guide no. 10. Centre for Longitudinal Studies : London. 20. Werner (1984). Fertility and family background: from the OPCS Longitudinal Study. Population Trends 35, pp. 5–10. 21. DiSalvo P M (1992). Intergenerational patterns of teenage fertility in England and Wales. MSc thesis. London School of Hygiene and Tropical Medicine. 22. Rosato M (1999). Teenage fertility in England and Wales: trends in socioeconomic circumstances between the 1971 and 1981 Censuses. LS Working Paper 78. Office for National Statistics: London. 23. Pearce D, Cantisani G and Laihonen A (1999). Changes in fertility and family sizes in Europe. Population Trends 95, pp. 33–40. 24. See note 21.

1. Office for National Statistics (2003). Social Trends no.33, 2003 edition. TSO: London, p.49. 2. For example, Babb P (1993). Teenage conceptions and fertility in England and Wales, 1971–91. Population Trends 74, pp.12–17, and Kiernan K E (1997) Becoming a young parent: A longitudinal study of associated factors. British Journal of Sociology 3, pp. 406–428. 3. For example, Social Exclusion Unit (1999). Teenage Pregnancy. TSO: London. 4. Kahn J R and Anderson K E (1992). Intergenerational patterns of teenage fertility. Demography 29(1), pp. 39–57. 5. Maynard R, ed. (1997). Kids having Kids: Economic Costs and Social Consequences of Teenage Pregnancy. Urban Institute Press: Washington D.C. 6. Chandola T, Coleman D A and Hiorns R W (1999). Recent European fertility patterns: Fitting curves to ‘distorted’ distributions. Population Studies 53, pp. 317–329. 7. Chandola T, Coleman D A and Hiorns R W (2002). Distinctive features of age-specific fertility profiles in the English-speaking world: Common patterns in Australia, New Zealand and the United States, 1970–98. Population Studies 56, pp. 181–200. 8. Jones E F et al (1986). Teenage Pregnancy in Industrialized Countries. Yale University Press: New Haven. 9. Gauthier A H (1996). The State and the Family: A Comparative Analysis of Family Policies in Industrialized Countries. Oxford University Press: Oxford. 10. Wellings K. and Wadsworth J (1999). Family influences on teenage fertility in McRae S (ed.) Changing Britain: Families and Households in the 1990s. Oxford University Press: Oxford. 11. Office for National Statistics (various years). Birth Statistics, series FM1. The Stationery Office: London. 12. INSEE (various years). Mouvement de la Population. INSEE: Paris. 13. Teitler J. The relationship between adult fertility and adolescent fertility in developed countries. Paper presented at the 1999 Annual Meeting of the Population Association of America. 14. Hattersley L and Creeser R (1995). Longitudinal Study 1971–1991: History, organisation and quality of data. Office for National Statistics, series LS No.7. HMSO: London. 15. INSEE (1995). Echantillon Démographique Permanent: Manuel de l’Utilisation. INSEE Departement de la Démographie: Paris.

37

National Statistics