What Women Want: Suffrage, Gender Gaps in ... - Stanford University

0 downloads 95 Views 328KB Size Report
The bottom panel of Table 3 tests whether political gender gaps can be accounted for by a one- dimensional ...... how wo
What Women Want: Suffrage, Gender Gaps in Voter Preferences and Government Expenditures Patricia Funk SITE - Stockholm School of Economics Christina Gathmann Stanford University

This Draft: July 2006

Abstract This paper combines unique individual-level information on ballot votes with state-level data on expenditures to provide new evidence on how women suffrage has affected government spending. Using data from the last country in Europe to adopt suffrage, Switzerland, we demonstrate two main results. First, women suffrage has changed the scope of government much more than its size. Women are more likely to support expenditures for public goods like environment and public transport, but oppose defense spending and subsidies for agriculture. Second, the political gender gap has shifted over time. While women were equally likely than men to support publicly provided health and welfare services shortly after suffrage adoption, their demand for these services has increased over the past two decades. We calculate that ten years after women suffrage, total expenditures are around 7 percent lower than prior to adoption, while welfare expenditures are 10 percent higher. Keywords: Women Suffrage, Fiscal Policy, Voter Preferences, Switzerland

*Correspondence: Patricia Funk, Stockholm Institute for Transition Economics (SITE), Stockholm School of Economics, Email: [email protected]; Christina Gathmann, Department of Economics and Stanford Center for International Development, Stanford University, Email: [email protected]. We thank Renee Adams, Ulf Axelson, Erik Bergloef, Doug Bernheim, Francine Blau, Tore Ellingsen, Daniel Ferreira, Mariassunta Gianetti, Henning Hillmann, Helena Svaleryd, Michele Tertilt, seminar participants of the Stockholm Institute of Transition Economics, Stockholm School of Economics, University of Uppsala, University of St. Gallen, Stanford University and the Midwest Political Science Association for useful comments and suggestions. We are grateful to Werner Seitz, Magdalena Schneider and Elisabeth Willen from the Swiss Bureau of Statistics, Andreas Ladner, Klaus Armingeon, Hans Hirter and Christian Bolliger from the University of Berne and Francois Loretan from SIDOS for invaluable help in collecting the data. Patricia Funk gratefully acknowledges financial support from the Swedish Research Council Vetenskapsradet.

1

Introduction

A dramatic expansion of the electorate occurred when women received the right to vote. Since women suffrage roughly doubled the size of the electorate, we would expect government to take quite a different form, if women have different policy preferences than men.1 Proponents of the anti-suffrage movement even argued that suffrage would destroy families, decline the quality of elected politicians or have a negative influence on the business climate (see Jones, 1991 and Banaszak, 1996). Despite these drastic predictions, surprisingly little is known about the actual policy response to women suffrage. For the United States, Lott and Kenny (1999) report that women suffrage caused a substantial increase in the size of government between 1870 and 1940. In sharp contrast, Aidt et al. (2006) find that female franchise had little effect on expenditures in several European countries in the late 19th and early 20th century. All previous studies rely on aggregate national or state-level data, and cover time periods with limited data availability. This paper makes use of individual voting choices to analyze differences in political preferences between men and women directly. The richness of the data allow us to distinguish between political gender gaps in different policy areas, for example preferences for unemployment insurance from preferences for agricultural subsidies. We combine our individual-level evidence with a detailed aggregate analysis of the fiscal policy responses to women suffrage. This allows us to check the consistency of our results using two very different data sources and assess potential concerns of omitted variables in the aggregate data. Our empirical analysis focuses on Switzerland, which provides a unique empirical setting to analyze voter preferences and the effects of women suffrage.2 While Switzerland was one of the first countries 1 Gender has been found to matter in India: female policy makers invested in projects directly relevant to the needs of their own genders. Depending on the regional area, female leaders allocated resources to drinking water and roads differently than men; see Chattophadhyay and Duflo (2004). 2 Previous studies on Switzerland have also found conflicting results. While Abrams and Settle (1999) document an increase in federal welfare spending in Switzerland after 1971, Stutzer and Kienast (2005) find a negative correlation between women suffrage and cantonal spending.

2

to extend the economic franchise (1848), it was the last country in Europe to extend the franchise to women. Women suffrage was introduced at the federal level in 1971, more than half a century later than in the United States and many European countries. Switzerland also has wide-ranging possibilities for direct democratic participation through voter initiatives and mandatory or optional referendums. Both late adoption and its strong direct democratic tradition provide us with rich data to study political preferences and fiscal policy effects of women suffrage. To analyze gender gaps in voter preferences, we use novel survey data on voting behavior in almost 200 ballots since 1981 covering policy areas like health care, unemployment insurance, defense or subsidies for public transport. The data allow us to study actual voter support for policy projects with predictable and often sizeable financial consequences. Our results suggest that women are more supportive of government intervention in some areas, but oppose them in others. Women consistently support higher expenditures for the environment and public transport, but oppose spending for defense or subsidies for agriculture. We also document a shift in voter preferences over the past three decades: while women are more conservative than men in the 1970s, women turn more left-wing and men more conservative during the 1980s.3 This shift in preferences has lead to an increase in women’s demand for government in the areas of health and welfare. While female political preferences had a strong influence on the types of expenditures supported by the electorate, their effect on the size of government is small: between 1981 and 2003, women were two percent more likely to support more government spending. Focusing on the ballots where women actually changed the results, we calculate that women preferences increased government spending by 3

In the United States, women voted more conservative than men at least until the presidential election of 1964. Berman (1993) also provides evidence that women in Arizona were less supportive of a progressive labor package and the abolishment of the death penalty, but more supportive of the prohibition in 1914 to 1916. A reversal in the political gender gap started to appear in the 1980s when women were much more likely to support Democratic candidates in the United States. A similar shift has been observed in Europe where women became more likely to support left-wing parties in Europe than men since the 1980s (see Box-Steffensmeier et al., 2004; Edlund and Pande, 2002; Seltzer et al., 1997 for the United States. Inglehart and Norris, 2003 and Norris, 2003 for evidence from other countries).

3

just 1.1 percent over that period. We complement our analysis on voter preferences with state-level data set on government spending and canton characteristics from 1950 to 2000. In line with our evidence that women are more conservative upon adoption, we find that spending at the canton level decreased by about 3 percent after women suffrage was introduced. The negative adoption effect is particularly large for welfare, health and education expenditures. We also show that the long-run effect of women suffrage is consistent with our documented change in political preferences. While social expenditures decline upon adoption, they increase with the number of years since women suffrage adoption. Even twenty years after women suffrage was adopted, overall expenditures are lower than prior to adoption. We perform several specification tests to demonstrate the robustness of our results. First, we show that the negative adoption effects are not driven by endogeneity bias. An analysis of the adoption decision at the canton level reveals that support for women suffrage is driven by a combination of timeinvariant cultural differences and an aggregate upward trend in favor of women suffrage. We provide additional evidence that neither differential trends in male preferences nor government spending prior to adoption can explain support for suffrage among the male electorate. In addition, we show that our results are robust to the inclusion of additional control variables and changes in female turnout behavior over time. The paper makes several contributions to the literature. We show that the fiscal policy effects of women suffrage are closely tied to the evolution of aggregate voter preferences. Both data demonstrate that women suffrage affected the scope of government much more than its size. We also for the first time provide evidence that suffrage resulted first in a negative and later a positive effect on health and welfare expenditures.4 We contribute to the literature on the political gender gap in two ways. First, we provide evidence 4

Edlund and Pande (2002) trace women’s shift to the left to increasing divorce rates. Their model implies dynamic suffrage effects that are consistent with our empirical evidence.

4

that women and men systematically differ in which government expenditures they support. Our focus on actual policies with direct financial consequences allows us to link revealed preferences to actual spending patterns at the state level. In contrast, the literature on political gender gaps largely focuses on hypothetical questions from opinion polls, which makes it difficult to assess the actual consequences of changes in electoral rules like women suffrage.5 Second, our results on the reversal of the gender gap in recent decades confirms similar findings for other countries. However, previous studies have focused on the political gender gap measured along a single left-right scale or as the relative support for Democratic presidential candidates. We go beyond this literature in providing evidence that the shift in preferences has mostly affected the demand for health and welfare expenditures with little change in other policy areas like agriculture, transport or environment. The structure of the paper is as follows: Section 2 provides some background information on the introduction of women suffrage in Switzerland. Section 3 analyzes male and female voter preferences for the size and scope of government using data on federal ballots. The results of women suffrage on fiscal policy in the cantons are reported in Section 4. Section 5 shows that the canton-level results are robust to concerns about endogeneity, the inclusion of additional controls and changes in female turnout behavior over time, while Section 6 concludes.

2

The Long Road to Women Suffrage

Switzerland was among the last countries in Europe to introduce voting rights for women. Women suffrage at the federal level was granted in 1971, half a century later than in the United States and many European countries.6 A first attempt to introduce suffrage at the national level was undertaken 5

Studies on the potential sources and determinants of the political gender gap include Alvarez and McCaffery (2003), Conover (1988), Gidengil (1995), Inglehart and Norris (2003), Norrander (1999), Schlesinger and Heldman (2001), Shapiro and Mahajan (1986) and the papers in Mueller (1988). 6 For example, women suffrage was adopted in Finland in 1906, in Denmark in 1915, in 1918 in Germany; Austria, Poland and Russia. Spain and Portugal followed in 1931. France and Italy adopted suffrage in 1945.

5

in 1929, when a petition signed by 250,000 citizens was sent to the federal parliament. However, neither the government nor the parliament reacted to the petition in support of suffrage. Since an extension of the franchise required a change in the Swiss constitution, the male electorate had to approve it in a referendum. The first national referendum on the introduction of women suffrage in Switzerland was held in 1959. It failed as just one out of three men voted for the adoption of women suffrage. The second national referendum in 1971 in contrast passed with almost two-thirds voting in favor of it. Still, in eight out of the twenty-five cantons, less than fifty percent of men voted for its introduction. There was also substantial heterogeneity in the adoption of women suffrage at the canton level. First attempts to introduce women suffrage at the cantonal level failed in Neuchatel, Basle City, Glarus, Zurich, Geneva and St. Gallen in 1920/21. The first column in Table 1 shows that women suffrage was adopted earlier in the French-speaking parts of Switzerland. In contrast, the two Germanspeaking cantons (Appenzell-Ausserrhoden and Appenzell-Innerrhoden) introduced suffrage almost 30 years later. In contrast to the United States, the male electorate in Switzerland also had to approve suffrage in each canton independently of adoption at the federal level. Overall, nine cantons adopted women suffrage in their canton before or when it was introduced at the federal level. Adoption was voluntary in all but the last one (Appenzell-Innerrhoden), where women suffrage was mandated by the Supreme Court in 1990.

— insert Table 1 about here —

As shown in columns (2) and (3) of Table 1, cantons that adopted suffrage earlier at the canton level were also more likely to support voting rights for women in the two federal referendums. This congruence is important since our empirical analysis relies on data both at the federal and cantonal level.

6

A comparison of column (2) and (3) shows that a canton’s ranking in supporting women suffrage was roughly stable over time. The approval of women suffrage in 1971 was therefore a consequence of increasing voter support in each canton, not by changes in a few cantons. As we will demonstrate below, the timing of adoption at the canton level can largely be explained by a combination of timeinvariant cultural differences across cantons and a common upward trend in the support for women suffrage in the whole country. One explanation for the aggregate increase in voter support is that the Swiss Government wanted to sign the European Human Rights Convention, but would do so only with restrictions (”Vorbehalt”) because women did not have the right to vote. This resulted in a lot of media attention and ultimately political pressure in favor of women suffrage.7

3

Political Preferences of Men and Women

3.1

Data on Voting Behavior in Federal Propositions

To study political gender gaps, we exploit the fact that Switzerland has wide-ranging possibilities for direct democratic participation. In Switzerland, citizens may propose initiatives to change the federal constitution or request a referendum about all laws issued by the federal government if 50,000 eligible voters or 8 cantons require it within 100 days after their official publication. In addition, a voter referendum is mandatory for all changes to the constitution and joining international organizations. Our analysis uses surveys of voting behavior in 197 of the 202 federal propositions between 1981 and 2003 for a representative sample of Swiss citizens. The data have a number of advantages over other measures of preferences: first, the votes cover a wide range of political issues, such as health policy, changes in unemployment insurance, new environmental policies, subsidies for agriculture or membership in international organizations. Second, we use information on voting behavior with real 7

See Seitz (2004) and Banaszak (1996) for more details on the political struggle for women suffrage in Switzerland.

7

political and financial consequences instead of relying on opinions polls about political issues. Finally, we identify differential preferences that are representative for the whole population as individuals in all cantons vote on the same proposition.8 The data set is a repeated cross-sectional survey of 500 to 1,000 respondents between 1981 and 2003.9 It contains detailed information on the respondent’s demographics, economic situation, canton of residence, political attitudes as well as voting decisions in the ballots. We dropped all respondents under the age of 21, who were not eligible to vote until March of 1991, and under 18 thereafter. We further restrict our sample to respondents that actually voted in the federal ballot. Table 2 reports summary statistics of the survey data separately for men and women over the sample period. The table reflects the more traditional position of women in Swiss society: women are on average less educated than men and have lower income available to them. The female labor force participation rate is low compared to the United States as is the fraction of divorced people and single parent households. Women are also more likely to live in urban areas and in the French- and Italian-speaking cantons of Switzerland. With respect to political preferences, women position themselves somewhat more on the left than men during the 1980s and 1990s. Women are also on average 7 percent less likely to vote in federal propositions than men.

— insert Table 2 about here —

A first look at the propositions with the largest gender gap shows that women were 18 percent more likely to support an initiative for a reduction in tobacco consumption (see Table A1 in the appendix for a list of those propositions). Not surprisingly, women were also more likely to support votes for the 8

With suffrage, women also gain access to other sources of political influence such as the right to sign or propose voter initiatives. We analyze the responsiveness of different political instruments after the adoption of women in more detail in related work. 9 Even though women received the right to vote at the federal level in 1971, surveys have been conducted only since 1981.

8

equal representation of women in the federal government, equal rights for men and women in general and a reform of marital law. In addition, women were more supportive of anti-discrimination policies in the military, environmental policy and government subsidies for disabled people. While suggestive, the summary statistics show that men and women also differ along many other dimensions like labor force participation or educational attainment. To address this, we now turn to a systematic analysis of political gender gaps conditional on observable characteristics.

3.2

Political Gender Gaps across Policy Areas

We focus our analysis on seven policy areas, where we expect the largest effect on expenditures, and discuss the other votes in footnotes. Three areas cover public goods (environment, transportation and defense), two the public provision of a private good (education, health), and two transfers and redistributive issues (agricultural subsidies, social security provisions). Within each category, the propositions are defined in such a way that it always propagates either more or less of a certain policy.10 The model we estimate is

Yesij = α + βj F emalei + γ 0 Xit + εij

(1)

where Yes ij is a binary variable equal to one if respondent i supported a vote in policy area j and zero otherwise. Xit includes all other characteristics of the respondent such as demographics, economic position or political preferences. Table 3 reports marginal effects from a probit model of the voting decision in each policy area. All specifications include year and canton of residence fixed effects. The first row includes only a female dummy. Women vote more in favor of protecting the environment, support for the elderly and the disabled, the use of public instead of private transportation and a reduction of military spending 10 For instance, all votes on agricultural policy concerned a reduction in agricultural subsidies. See Table A2 in the Appendix for further information on the votes in each policy area.

9

(see also Longchamp and Bieri, 2001). In contrast, they oppose subsidies for agriculture, a nuclear based energy policy, a further construction of roads and a relaxation of speed limits (last three votes not reported). Men, on the other hand, are more likely to support a reform of the unemployment insurance, which included reductions in benefits. Women and men are equally likely to support free secondary education. Row (2) adds controls for demographics, employment status and household income to the specification. Since women are less educated and have lower labor force attachment, gender differences should disappear if education or employment are driving political gender gaps. However, differences in demographic or economic characteristics only affect voting behavior on health insurance, a reduction in unemployment insurance and support for road infrastructure. Most gender gaps cannot be explained by differences in either demographic or economic characteristics.

— insert Table 3 about here —

The bottom panel of Table 3 tests whether political gender gaps can be accounted for by a onedimensional measure of political ideology. Row (4) therefore adds a measure for party ideology to the specification in row (2). Party ideology is measured as the respondent’s self-reported position on a left-right scale from 0 to 10, with ten being conservative and zero being left-wing. Since a respondent’s political ideology is only available after 1987, row (3) reports estimates for the second specification with the sample restricted to the period 1988 to 2003. The results in the third row show that gender gaps are more pronounced in the later sample period. Controlling for the left-right position of the respondent, the gender gap disappears for subsidies for health insurance, two redistributive issues (“reduce unemployment benefits”, “longer maternity leave”) and a reduction in the military. However, a one-dimensional “left-right”indicator cannot capture gender gaps in the area of environmental policy, transportation, agricultural policy or social security.11 11

Using the same specification, we also find the largest gender gap in votes on equal rights for men and women (11

10

Our findings are consistent with previous evidence from public opinion surveys in other European countries. In Sweden for example, women were more in favor of publicly provided medical care, environmental protection and gender equality than men (Eduards, 1982). Similarly, Norris (1988) found that women in several European countries were more supportive of unemployment benefits and less in favor of defense spending. In the US, opinion polls report similar gender gaps in the areas of environmental care, force and violence, and compassion issues (Shapiro and Mahajan (1986)). However, opinion polls include the whole citizenship, and not only the voting population, and can be bad predictors for later voting behavior (Matsusaka and McCarty, 2001).

3.3

Political Gender Gaps and Government Expenditures

While gender gaps in policy issues are informative, they do not tell us whether and how women have changed the size or scope of government. To address this, we select the ballots that would unambiguously increase or decrease government spending. To calculate the fiscal consequences of a ballot, we rely on detailed comments, which are published by the federal government for each initiative, new law and executive orders. These documents include the implied expenditures and taxes in the case of voter approval and are common knowledge before the ballot.12 We are able to identify 71 propositions, where the fiscal consequences were either an unambiguous increase or decrease in government expenditures, taxes or subsidies. Note that the set of propositions we analyze contains both ballots that were approved and therefore actually increased government spending as well as ballots that were not successful. This ensures that we get a representative picture of preferred spending for men and women, which is not affected by the ballot’s actual success. However, we will also analyze women’s role as pivotal voters. percent). Women are also 7 percent more likely to support public spending on culture, 6 percent less likely to support gentechnology and 5 percent more likely to support a more liberal immigration policy. In contrast, no significant gender gaps exist for votes on joining international organizations, regulation of the housing market and on illegal drug policy. 12 All documents by the federal council and parliamentary resolutions can be accessed online at http://www.ads.bar.admin.ch/ADS. The comments including the financial assessment for each proposition are sent by mail to each household prior to the vote. Table A3 in the Appendix contains a detailed list of the 71 votes with predictable financial consequences.

11

Table 4 shows marginal effects from a probit model whether a respondent voted yes on a proposition that would increase (column (1)) or decrease (column (2)) government spending, taxes, subsidies or debt. The first column shows that women are 2.7 percent more likely to support projects that would increase overall government spending. As shown in the lower part of the table, the gender gap changes little with the inclusion of additional control variables.

— insert Table 4 about here —

However, as shown in the last section, political gender gaps vary a lot depending on the policy area involved. Columns (3)-(9) of Table 4 therefore distinguishes propositions with increasing or decreasing expenditures in the seven policy areas. The coefficients show that women are more likely to support higher spending for environmental protection, public transport or welfare, but oppose expenditures for the military or agricultural subsidies. In fact, the gender gap is particularly high for spending on environmental policy (14.3 percent). The gender gaps are somewhat smaller for social expenditures, security, agricultural spending and education.13 The results in Table 3 and 4 suggest that female voters have had a bigger impact on the scope than the size of government. However, the estimates in Table 4 do not allow us to directly measure how women voters changed actual federal spending. The latter is only affected if the proposition is approved by the voters, and women changed the final outcome. Among the fifteen federal votes where men and women had approved different outcomes, women changed the result in their favor in four cases or about two percent of the 202 propositions over that period.14 Based on the information provided by the federal government before the vote, we can get 13

The significant gender gap in the area of education is particularly driven by one vote, where women opposed a reduction of federal subsidies for primary school. 14 Table A4 shows a list of those votes where men and women had accepted different outcomes. The lower support of women for an “ecological and modern agriculture” as well as “easier access to real estate for foreigners” appears to be in

12

a good estimate of the financial consequences for two votes. Women’s opposition to a reduction in unemployment benefits raised federal spending of about 70 million Swiss Francs per year. Women were however also in favor of abolishing subsidies for parking spaces, which saved the federal government about 20 million Swiss Francs per year. Relative to the 46 billion federal expenditures in 1999, the change in voting outcomes by women adds up to a mere 1.1 percent increase in federal spending. The evidence from voter preferences in federal propositions thus suggests that the impact of women on the size of government has been small.

3.4

Changes in Voter Preferences over Time

Our results thus far show that women support government spending in some policy areas, but not in others. Since several studies have documented changes in the political gender gap in recent decades, our findings for 1981 to 2003 might not be representative of political preferences in the years immediately following women suffrage. To check for possible changes in political gender gaps over time, we use data from a survey in 1975 that included questions about four of our seven policy areas: environment, health, education and welfare. The data was part of a cross-national survey (Political Action Study) to study voting behavior and political attitudes for a representative sample of the adult population above 16. For each of the four policy areas, we know how much the respondent thinks that it is an important problem and to which degree he considers it a government responsibility. The response categories in each area ranges from 1 (absolutely unimportant problem / no government responsibility at all) to 4 (very important problem / an essential government responsibility). For welfare, we have two response items: whether the government should redistribute income (column (4)) and care for the elderly (column (5)). contrast to our previous findings. A detailed analysis of these votes however shows that the opponents of the first vote thought the reform did not go far enough, while environmental concerns of new construction played a major role in the second vote. See http://www.polittrends.ch/vox-analysen for a further discussion of these votes.

13

Table 5 shows the results. As in the last section, our main parameter of interest is the coefficient on the gender dummy. The specifications of control variables are the same as in Table 4. Without controlling for any differences in observable characteristics, the first row shows that women consider care for the elderly, providing health care and protecting the environment as more important problems than men. However, the significance decreases as controls are added. As for the government’s responsibility, the lower part of the panel shows that there were no statistically significant gender gaps in any of the four policy areas. Even though women consider protecting the environment, providing or caring for the elderly as important problems, they do not necessarily think that it is a government responsibility. Adding controls for individual demographics, employment status and income, changes the picture somewhat. Women are now less likely to consider the government responsible for the care of the elderly. In contrast, women are now also more likely to consider education a government responsibility. The political gender gap for publicly provided education appears to be driven by differences in household income with more wealthy households supporting publicly provided education.

— insert Table 5 about here —

Comparing the results from Table 4 and 5 shows that women were consistently more likely to support government activity for the environment and education both in 1975 and after 1981. In contrast, women were less likely in 1975 to support government intervention in health care and welfare provisions than after 1981. This suggests an increase in their demand for government in the areas of health and welfare between the 1970s and 1980s. However, the observed changes in preferences could be an artefact of differences in survey questions or some other factor. We therefore turn to an analysis of the political position along a left-right scale, which can be measured consistently over time. Figure 1 shows the fraction of women and men

14

supporting leftist parties in the federal elections since 1971. Two facts are noteworthy. First, women were less supportive of left parties than men throughout the first decade after suffrage was introduced at the federal level. Second, the figure also shows that there was reversal in the political support for left parties over time: women became more supportive of left-wing parties during the 1980s, while men became more conservative. The documented reversal in preferences is similar to observed changes in the political gender gap in other countries (see Edlund and Pande, 2002; Norrander, 1999; Schlesinger and Heldmann, 2001 for the United States; Inglehart and Norris, 2003 for other developed countries).

— insert Figure 1 about here —

Since there are other factors affecting political positions over time, Table 6 analyzes the political position of men and women along a left-right scale for two points in time: 1975 and 1995. The dependent variable ranges from zero (very left-wing) to ten (very right-wing). Columns (1) to (4) show the results for 1975 and columns (5) to (8) for 1995.15 In 1975, women were indeed much more conservative than men in all specifications. The gender gap increases if we control for differences in education and other demographics and decreases once we include employment status and household income. The fourth column adds various interaction effects to allow for different effects of women by employment and marital status. Unemployed women are more likely to be left-wing, but marital status has no effect on the political gender gap.

— insert Table 6 about here —

The picture in 1995 looks completely different: women are now much more left-wing than men and the gender differences become even stronger once we control for demographics, employment status 15 The data for 1975 are again taken from the Political Action Study. The data for 1995 come from the Selects Study, which is a post-electoral survey of more than 5,000 Swiss citizens. The sample is representative of the Swiss population.

15

and income. The final column again adds various interaction effects to see if the results are driven by a subgroup of women. Indeed, the shift to the left was concentrated among women in the labor force and those with children.16 In sum, women were more conservative and less supportive of government intervention immediately after women suffrage. While women are more likely to oppose spending for national defense and agricultural subsidies, they are however more likely to support public goods such as protecting the environment, public transport and the public provision of education. The shift to the left among women in the 1980s and 1990s has increased their demand for government, mostly in the areas of health care and welfare provisions.

4

Women Suffrage and Cantonal Spending Patterns

This section analyzes how women suffrage affected actual spending on the canton level.

4.1

Canton-Level Panel Data

The government sector in Switzerland is with 34 percent of GDP (1996) relatively small, far below the average of the European Union. Cantons play a dominant role in providing public goods and services as the constitution states that all responsibilities retain with the cantons unless they were ceded to the federal government in a referendum.17 In 1998, 33 percent of total government spending was undertaken by the federal government, 40 percent on the cantonal level and 27 percent on the local level. This decentralized structure leaves cantons with a lot of autonomy for redistribution and the provision of public services. For example, cantons decide autonomously on resources for secondary 16

In contrast to previous results from the United States, we could not find evidence that divorced women in general or those with medium income are more left-wing than other women. 17 While the federal government has increased its scope of operation over time, the expansion has been limited by the fact that each redistribution of political responsibilities requires a change in the constitution and is thus subject to mandatory referendum.

16

education, which accounts for 25 percent of all cantonal expenditures, the provision of health care (around 18 percent of expenditures) and social security (16 percent).18 The distribution of revenues between federal and cantonal levels is similarly decentralized. Cantons have the authority to tax labor and capital income which account for roughly 50 percent of canton revenues. This produces substantial variation in the tax burden and tax revenues across cantons. To quantify the effects of women suffrage on fiscal policy, we collected a comprehensive dataset on cantonal expenditures and revenues, demographic and economic characteristics as well as political institutions of the 25 cantons from 1950 to 2000.19 Table 7 shows summary statistics for the canton panel data over the whole period. To analyze the effect of women suffrage on expenditure patterns and compare it to our results on voter preferences, we classified spending into seven policy areas: environment, transport, security, agriculture as well as welfare, education and health. On the revenue side, we analyze overall revenues and public deficits. As can be seen from the last two columns, there is a lot of variation on both the spending and revenue side across cantons and also over time.

— insert Table 7 about here —

In line with our survey data in the last section, the demographic characteristics show that the share of divorcees in the adult population, the fraction of single parents and the female labor force participation rate are low compared to the United States or other European countries. Unemployment rates have also traditionally been below five percent. The education level in each canton is measured as the fraction with a university degree. The 18

Shared responsibilities between canton and federal level exist in agricultural policy, civil and criminal law and taxes. The federal level has the sole responsibility in international relations, defense, customs and currency, atomic energy, media, postal service, telecommunication as well railways and air traffic. In the areas of environmental policy, social security system, roads and industrial and labor regulation, the federal government provides the legal basis while Cantons execute the federal laws. 19 See Appendix A for a more detailed description of the data sources and variables. Our analysis excludes the canton Jura, which was founded in 1977.

17

average is only 9 percent with the share ranging from zero percent to almost one-third of the adult population. The language indicator is a dummy variable equal to one if the dominant language in a canton is French or Italian and zero if it is German. About one-third of the cantons are dominantly French- or Italian-speaking. The share of seats in cantonal parliaments occupied by left parties averages 20 percent. While some cantons have no left parties in their parliaments, other cantons have a majority of their seats held by left parties. Finally, Swiss citizens have access to several direct democratic instruments on the canton level. All cantons in Switzerland allow citizens to put a voter initiative on the ballot. The signature requirement for an initiative varies from 0.01 percent to 17.4 percent of the eligible population. In addition, 64 percent of the cantons have a mandatory budget referendum in place over the period, which requires the electorate to approve projects exceeding a certain threshold. Also, 56 percent allow for a mandatory law referendum, which requires approval of new laws and executive orders.

4.2

Suffrage Adoption Decreases Expenditures in Many Areas

To determine the effect of voting rights for women on expenditures, we estimate the following model

Yst = αs + γt + β · W Sst + δ 0 Xst + ust

(2)

where Yst denotes total (real per capita) annual expenditures, revenues or deficits or expenditures in a certain policy area in canton s and year t. The variable W Sst is a dummy variable equal to one if suffrage was adopted in canton s in year t and zero otherwise. The dependent variable is measured in logs for all categories except cantonal deficit, which can take on negative values and is therefore measured in levels. All specifications include canton and time fixed effects. These pick up aggregate trends in canton expenditures over time as well as all time-invariant heterogeneity in spending across cantons. The effect of women suffrage on government spending is therefore identified from cantons

18

introducing women suffrage in a specific year. Since expenditures and revenues are affected by other time-varying factors, we also include a large number of control variables Xst for each canton. Variables like the age structure of the canton measured by five age categories, the unemployment rate, the fraction of individuals with higher education, the share of commuters in the canton, the percentage of divorced people and single parents as well as the female labor force participation rate control for differences in the demand for government like public transport, social welfare or health care. To control for economies of scale in the provision of public goods, we also add the average population density in the canton. The share in federal revenues and the amount of federal subsidies adjusts for differences in resources available to the cantons. Finally, we also include the percentage of the population that is Catholic as a rough measure of the political ideology in a canton. Table 8 reports the results from regression (2). The estimates show that adopting women suffrage has a negative effect on canton expenditures, revenues and deficit though only the first one is statistically significant. The coefficient suggests that expenditures decrease by 3.3 percent. This implies that suffrage had an economically significant negative effect on the size of government.

— insert Table 8 about here —

As shown in the last section, women are in favor of some expenditures more than others. The righthand side of Table 8 therefore shows the effect of suffrage on the scope of government. The results imply that women suffrage has negative effects of canton expenditures for agriculture, security as well as welfare, education and health. At the same time, suffrage increase canton expenditures for public goods such as public transport and environmental protection though the latter is not statistically significant. Most coefficients on the controls are reasonable. A higher share of federal revenues or subsidies induce canton governments to spend more as does a higher unemployment rate. The fact that higher 19

population density is associated with lower expenditures suggests that there are significant economies of scale in the provision of public goods and services. Also, cantons with a higher fraction of Catholics spend slightly less, which is consistent with the idea that these cantons are more fiscally conservative. The share of single parents has a positive effect on expenditures, while female labor force participation and the percentage of divorced people has a negative effect on spending. The effect of adopting suffrage on spending patterns is broadly consistent with our evidence on preferences in the last section. Women are more supportive of expenditures in the area of public transport and environmental policy, but less in favor of expenditures for agricultural subsidies or security. The negative effect on government size and strong negative effects on expenditures for welfare and health also confirms that women were indeed more fiscally conservative than men when women suffrage was adopted.

4.3

The Dynamic Effect of Women Suffrage on Fiscal Policy

Since women’s preferences shifted over time, the long-run effects of women suffrage might differ from its short-run effect. Given our evidence in the last section, we expect this to affect primarily expenditures on health and welfare. To capture the dynamic effect of women suffrage, we estimate the same model as in (2) where now the variable for women suffrage are the years since adoption. The control variables and expenditure categories are the same as before. Table 9 shows the results. The top panel includes a linear and quadratic term for the years since adoption. The bottom panel allows for a more flexible specification by including dummy variables for each 5 years since adoption. The coefficients suggest that the long-run effect of women suffrage on the size of government is negative. Suffrage not only decrease expenditures, but also canton revenues and deficit. For example, the estimates imply that ten years after the adoption of women suffrage, overall expenditures are 7 percent lower than prior to adoption. The bottom part of the table shows that the negative effect dominates even 20 years after suffrage adoption. 20

Looking at the individual expenditure categories on the right-hand side shows that suffrage decreased subsidies to agriculture also in the long-run. In contrast, there are no long-run effects on spending for the public goods environment, transport and security.

— insert Table 9 about here —

As predicted by our evidence on voter preferences, the largest shift between adoption and over time occurs for welfare and health expenditures: while welfare spending for example declines by 5 percent upon adoption, it is about 8 percent higher 10 years after suffrage relative to the pre-adoption period. Similarly, we find that health expenditures decrease by almost 10 percent upon adoption, but increase by 15 percent in the 10 years after adoption. The bottom part of the table shows that most of the increase in social expenditures occur more than a decade after women suffrage. Finally, expenditures for secondary education decline by a statistically significant 6.3 percent upon suffrage adoption. It is only twenty years after adoption that education expenditures are by 0.5 percent higher than before adoption.

5

Robustness Analysis

5.1

Endogeneity of Adoption

One important concern is that the negative effect of suffrage might be driven by omitted variables affecting both expenditures and the adoption decision. In the absence of a good instrument for suffrage adoption, we present several pieces of evidence showing that omitted variable bias cannot explain our results. Since men in each canton had to approve the extension of the franchise to women in a mandatory referendum, the prime candidate for omitted variables is the preferences of the male electorate prior

21

to adoption. Suppose cantons differ in a one-dimensional preference parameter where higher values indicate a more liberal attitude both towards government spending and extending the franchise. If fiscally liberal cantons are both more likely to adopt women suffrage and have a higher demand for government, our estimate of the adoption effect will be an upper bound and thus an over-estimation of the true effect of women suffrage. Since all our specifications include canton fixed effects, any timeinvariant differences in voter preferences across cantons will be absorbed by canton dummies. The only source of bias can therefore arise from time-varying heterogeneity across cantons not captured by our observable canton characteristics. To check for time-varying trends in male preferences prior to adoption, we again use our data on voting behavior in federal propositions prior to 1971. As in Section 3, we select all propositions that would have implied an increase in federal expenditures, and use the average support for costly projects in each canton and decade as our proxy for male preferences before suffrage adoption.20 The first column in Table 10 reports the result of regressing canton fiscal preferences on an indicator whether a canton introduced suffrage early and decade dummies. Early adopters are cantons that adopted suffrage before it was adopted on a federal level (votes before or at the same day of the national referendum). The results show that cantons that introduce suffrage earlier are indeed fiscally less conservative than late introducers. This difference in preferences will however be absorbed by canton fixed effects. To see whether early introducers also have a different trend in male voter preferences, we add in column (2) a linear time trend and an interaction term between the time trend and the time of adoption. While there is an upward trend in the overall support for higher expenditures among men, there is no differential trend in male fiscal preferences between early and late introducing cantons.

— insert Table 10 about here — 20

As discussed in a related paper, preferences measured from federal voting data are a good proxy for voter preferences at the canton level (see Funk and Gathmann, 2005). Note that we cannot simply add voter preferences into the expenditure regressions as the timing of adoption differed between federal and cantonal level.

22

Since cantons adopt suffrage at different points in time, it could however be the case that canton expenditures prior to adoption influence the adoption decision, for example because they might affect male preferences in that canton. We address this by testing whether early and late introducing cantons had the same growth rates in expenditures prior to adopting women suffrage. The annual growth rate in expenditures is 0.069 (0.055) percent for early introducers and 0.073 (0.067) percent for late introducers in the ten (five) years prior to adoption at the canton level with a T-statistic of 0.38 (0.91). Though early introducers have higher expenditure overall, there is no difference in the growth rate prior to adoption.21 What then determined the decision to adopt women suffrage? The right-hand side of Table 10 provides evidence that support for women suffrage is largely driven by the combination of timeinvariant cultural differences across cantons and a common upward trend, the latter possibly reflecting social change. The dependent variable in the regression is the share of voters supporting women suffrage in the two federal referendums in 1959 and 1971.22 We first test for omitted variable bias by including our proxy for voter preferences. Column (3) shows that cantons favoring higher expenditures at the federal level are also more likely to support women suffrage. The coefficient implies that a ten percent increase in the support for more spending is associated with a nearly six percent increase in the support for women suffrage. This suggests that there is indeed a component of political preferences that drives both demand for government and extension of the franchise. To capture the general upward trend in the support of women suffrage over time, we also include an indicator for the referendum in 1971. Conditional on the other controls, support for suffrage was more than 30 percent higher in the second national referendum. Our control variables are the population density as a proxy for the degree of urbanization, the age structure of the population to capture cohort 21 Results are very similar if we first take out a common linear trend or year dummies and then test for differences in the residual across cantons. There are also no differences in the growth rate of expenditures if we compare cantons in the same time window, for example the 5 years prior to the federal referendums in 1959 or 1971. 22 We analyzed voting behavior in federal referendums in order to use our preference measure constructed from federal ballots. However, we find similar results to the ones reported if we use data on the adoption decision at the canton level instead.

23

effects and the share of university educated people in a canton. We also include whether the canton is located at the border and the number of tourists per capita to capture exposure to outside influences in favor of women suffrage. The results show that only the share of highly educated has a significant, positive effect on adopting women suffrage. Column (4) adds as control whether the dominant language in a canton is German, French or Italian. The results show that language and cultural background is an important determinant of the adoption decision. French- or Italian-speaking cantons (Fribourg, Geneva, Neuchatel, Ticino, Vaud and Valais) are on average 17.1 percent more likely to support women suffrage. The preferences for federal spending are no longer statistically significant once the language variable is included.23 The last two columns uses the same approach to test whether the growth rate of expenditures affects adoption, which would suggest a reverse causality problem. Column (5) uses the same specification as in column (3) where the main independent variable is now the growth rate in canton expenditures in the five years prior to the ballot. The coefficient implies that a higher expansion of canton spending is associated with larger support for suffrage. Column (6) again adds the dominant language of a canton. As in the case with voter preferences, the coefficient on the growth in canton expenditures declines. As in column (4), the language of a canton has a strong and persistent effect on the adoption decision with French- and Italian-speaking being again 17 percent more likely to support women suffrage. Our evidence therefore suggests that neither male preferences nor expenditures exhibit a differential trend prior to adoption. We also find that the adoption decision is strongly influenced by time-invariant cultural factors together with a common upward trend in the support for women suffrage. Since timeinvariant variation is absorbed by canton fixed effects, we conclude that concerns of omitted variable bias are small. Furthermore, any endogeneity bias would lead us to underestimate the effect of suffrage 23

Evidence from individual-level data in 1972 supports this finding at the canton level. The survey asks about the voting behavior of Swiss men in the federal referendum on women suffrage in 1971. Our results showed that language and thus cultural background is the strongest predictor for voting behavior with non-german speaking men being 12 percent more likely to support women suffrage.

24

adoption on canton expenditures and revenues.

5.2

Additional Controls

It could however be the case that other omitted variables bias our expenditure results. For example, the long-run effect on expenditures might be driven by an upward trend in the demand or supply of government independently of suffrage. To address this, we add several variables that control for potential sources of rising demand for government services to the specification in Table 9. Table 11 reports the results for overall expenditures (columns (1)-(3)) as well as welfare expenditures (columns (4)-(6)). The top part of the table shows the result of suffrage adoption, while the bottom part reports the coefficient for years since adoption. The first specification (columns (1) and (4)) adds mean cantonal income to control for rising wealth and its effect on the demand for government. The variable is available since 1965. For total expenditures, the negative adoption effect remains negative but is no longer statistically significant. The dynamic effect in contrast is now even more negative and statistically significant. The adoption effect for welfare expenditures is slightly larger than in the baseline, while the dynamic effect is no longer statistically significant. This suggests that some of the increase in welfare expenditures is driven by income effects. Our second specification adds the share of seats in cantonal parliaments held by left parties to control for other factors that might have increased the demand for government services or redistribution. The negative adoption effect for overall expenditures becomes even stronger, while the dynamic effect is now somewhat weaker. For welfare expenditures, the negative adoption effect is now much weaker and no longer statistically significant, while the dynamic effect is positive and much stronger than in the baseline specification. Even controlling for the strength of left parties, women suffrage has a positive long-run effect on welfare expenditures. We also test whether changes in direct democratic instruments could account for our results. Since the signature requirements for putting an initiative on the ballot was not adjusted in many cantons, 25

women suffrage facilitates the collection of the necessary number of signatures. In addition, several cantons switched the provisions for budget or law referendums. This might have an effect on canton expenditures even in the absence of women suffrage. Columns (3) and (6) therefore add the signature requirement in percent of the eligible population as well as indicator variables whether a canton has a mandatory budget or law referendum in place in a given year. The results imply that this actually makes the negative adoption effect even stronger for both overall and welfare expenditures. The dynamic effect remains unchanged in both cases. We conclude that our results are not very sensitive to the inclusion of controls for other potential changes in the demand for government.

5.3

Political Turnout Behavior

An alternative explanation for the difference between long- and short-term effect of women suffrage on expenditures could be changing participation behavior of women. If women participate more in the years following women suffrage, the long-term effect will be larger than the short-term effect. If in addition the initial non-voters are less fiscally conservative than voters, this could explain why we see an expansion in health or welfare expenditures over time. We address each of these arguments in turn. Our data suggest that the adaptation of turnout between men and women occurred quite rapidly. Figure 2 shows the evolution of the turnout gap calculated as male turnout minus female turnout in federal ballots from 1977 to 2003. While the gender gap in political participation is 8 percent on average, the turnout gap has been steadily declining from 16 percent in 1977 to about 2 percent in 2003.24 To analyze whether the dynamic suffrage effects might be driven by turnout behavior, we need a measure for male and female cantonal turnout. Since data on female turnout in cantonal elections 24 In 1977, political turnout of women in federal ballots was only 40 percent and increased to 55 percent in 2003. Separate turnout data for men and women are unfortunately not available prior to 1977.

26

are not directly available, we estimate them from overall turnout rates in cantonal elections. The basic intuition is that we should see overall turnout rates drop sharply with suffrage if women did not participate in elections. Under the assumption that the male propensity to vote was unaffected by the introduction of suffrage, we can then estimate female participation rates from a regression of overall turnout on the introduction of women suffrage. The first column in Table 12 shows the result of women suffrage on turnout. Overall turnout drops immediately after suffrage by more than ten percent. Over the next ten years, voter turnout is still lower than before suffrage, but no significant effects are found after the first decade. From these estimates, we calculate the evolution of female participation overall (column (2)) and relative to men (column (3)).25 The estimated turnout in canton elections confirm that adjustment of political participation occurred rapidly following a very similar adjustment pattern than turnout in federal ballots.

— insert Table 12 about here —

We then reestimate the effect of women suffrage on cantonal expenditure where our main independent variable is our predicted female political representation to control for turnout behavior. The specification otherwise is the same as in the first column of Table 8. Columns (4)-(7) in Table 12 shows the results for total expenditures as well as expenditures for health, education and welfare. Total expenditures decline for about fifteen years after adoption and then start reversing. Even more than two decades after adoption, total expenditures are not higher than when women were given the right to vote. Health and welfare expenditures show little change initially and then start rising 25

After suffrage, turnout (defined as the number of voters relative to the number of eligible voters) is T urnoutct = (1 − sharect ) ∗ M alect + sharect ∗ F emalect where share denotes the share of females in the electorate, M alect male turnout measured prior to adoption (on average 62.2 percent). Under the assumption that male turnout does not change with women suffrage, this allows us to calculate female turnout in canton c and time t, F emalect . The last step is to sharect ∗F emalect calculate the relative female representation. Define the variable RelativeT urnoutct = (1−share , which is equal ct )∗M alect to zero before suffrage adoption and equal to one if women are equally represented than men. Note that if the overall decline in turnout was in part due to falling male turnout, our estimate of female representation provides a lower bound of the actual political influence of women.

27

about 15 years after adoption. Education expenditures in contrast show a decline for the first 15 years following adoption and then a slight reversal thereafter. We conclude from this that rising female participation rates cannot explain the positive effect of suffrage on social expenditures over time. However, if female non-voters have different political preferences for government than voters, increased participation could still have an effect on expenditures. To address this, we compare the self-reported political position of voters and non-voters shortly after suffrage (in 1975) and two decades later (in 1995). The measure of political position ranges from zero to ten with higher values indicating a more left-wing position. In 1975, voters were not more or less left-wing than nonvoters. While voters had on average a value of 4.96 and nonvoters a value of 4.98, the difference is not statistically significant (T-statistic: 0.08). The pattern is different in 1995: voters now report a value of 4.77 and non-voters 5.06 with a T-statistic of 3.7. This implies female voters have actually become more conservative over time, while female non-voters more left-wing. We would however require the opposite pattern in order to explain the increasing effect of suffrage on expenditures over time.

6

Conclusion

This paper demonstrates how direct evidence on political preferences can be used to study the fiscal effects of women suffrage. Both our analyses of voter preferences and state-level expenditures show that women suffrage affects the scope of government much more than its size. While women are more supportive of government intervention and spending for public goods like the environment, public transport and education, they oppose spending for the military and agricultural subsidies. Our result that women suffrage has little effect on the overall size of government is in line with cross-country correlations of fiscal spending and women suffrage for other European countries. We also show that the shift in the political gender gap in recent decades has lead to an increase

28

in women’s demand for publicly provided health and welfare provisions. Though Switzerland is a more traditional society with low female labor force participation and low divorce rates, shifts in the electoral gender gap similar to the one in Switzerland have been found in other developed countries. This suggests that our results are valid beyond the specific Swiss setting. Since female voter preferences have opposing effects on different expenditure categories, overall expenditures can only provide limited insight in the effects of women’s political participation. Stepping back and asking “What women want” helps us better understand the changes following the inclusion of women in the political process. It would be interesting to extend our analysis as to whether and how women suffrage has affected health and educational outcomes among children. For example, we might expect that infant mortality to decline in countries and states that adopted women suffrage relatively early. We leave this question for future research.

References [1] Abrams, B.A. and R.F. Settle (1999), “Women’s Suffrage and the Growth of the Welfare State,” Public Choice, 100: 289-300. [2] Aidt, T.S., J. Duta and E. Loukoianova (2006), “Democracy comes to Europe: Franchise Extension and Fiscal Outcomes 1830-1938,” European Economic Review, 50, 249-83. [3] Alvarez, R.M. and E.J. McCaffery (2003), “Are there Sex Differences in Fiscal Policy Preferences?” Political Research Quarterly, 56, 5-17. [4] Ballmer-Cao, T.-H. (1988), Le Conservatisme Politique F´eminin en Suisse: Mythe ou R´ealite?, Geneva, Switzerland. [5] Banaszak, L.A. (1996), Why Movements Fail: Opportunity, Culture, and the Struggle for Women Suffrage, Princeton, New Jersey. 29

[6] Berman, D.R. (1993): “Gender and Issue Voting: The Policy Effects of Suffrage Expansion in Arizona,” Social Science Quarterly, 4: 838-850. [7] Box-Steffensmeier, J.M., S. de Boef, and T. Lin (2004), “The Dynamics of the Partisan Gender Gap,” American Political Science Review, 98, 515-28. [8] Chattopadhyay, R. and E. Duflo (2004), “Women as Policy Makers: Evidence from a Randomized Experiment in India,” Econometrica, 72(5): 1409-1443. [9] Conover, P. (1988), “Feminists and the Gender Gap,” Journal of Politics, 50(4): 985-1010. [10] Edlund, L. and R. Pande (2002), “Why Have Women Become More Left-Wing? The Political Gender Gap and the Decline in Marriage,” Quarterly Journal of Economics, 117: 917-61. [11] Eduards, M. (1981), “Sweden,” in Lovenduski J., and J. Hills (eds.), Politics of the Second Electorate, London: Routledge and Kegan Paul. [12] Federal Department of Finance (various years), Oeffentliche Finanzen der Schweiz, Berne, Switzerland. [13] Funk, P. and C. Gathmann (2005), “Estimating the Effect of Voter Preferences on Policy Outcomes: Preferences Matter!,” Stanford Center for International Development Working Paper 248. [14] Gidengil, E. (1995), “Economic Man - Social Women?,” Comparative Political Studies, 28(3): 384-408. [15] Inglehart, R. and P. Norris (2003), Rising Tide: Gender Equality and Cultural Change around the World, Cambridge University Press. [16] Jones, E.B. (1991), “The Economics of Women Suffrage,” Journal of Legal Studies, 20, 423-437. [17] Kueffner, S. (2004), “Stimmen Frauen Anders? Der ’Gender Gap’ in der Schweiz,” Master thesis, University of Berne. 30

[18] Longchamp, C. and U. Bieri (2001), “Frauen und Maenner als Stimmbuerger: Wie haben sie entschieden?,” http://www.gfsbern.ch/gfs/gender.html. [19] Lott, J.R. and L.W. Kenny (1999), “Did Women’s Suffrage Change the Size and Scope of Government?” Journal of Political Economy, 107: 1163-98. [20] Matsusaka, J.G. & N.M. McCarty (2001), ”Political Resource Allocation: Benefits and Costs of Voter Initiatives ”, The Journal of Law, Economcis and Organization, 17: 413-448. [21] Mueller C. (1988), The Politics of the Gender Gap: The Social Construction of Political Influence. London: Sage. [22] Norrander, B. (1999), “The Evolution of the Gender Gap,” Public Opinion Quarterly, 63: 566576. [23] Norris, P. (1988), “The Gender Gap: A Cross National Trend?” in Mueller (ed.), The Politics of the Gender Gap: The Social Construction of Political Influence. London: Sage. [24] Norris, P. (2003), “The Gender Gap: Old Challenges, New Approaches,” in: Women and American Politics: New Questions, New Directions, edited by S. Carroll, New York: Oxford University Press. [25] Schlesinger, M. and C. Heldman (2001), “Gender Gap or Gender Gaps?” Journal of Politics, 63(1), 59-92. [26] Seitz, W. (2004), “Ein Streifzug durch 100 Jahre Kampf um die politische Gleichstellung der Frauen in der Schweiz. Gleichstellungspolitische Strategien und die ambivalenten Wirkungen der direkten Demokratie,” mimeo, Federal Office of Statistics, Switzerland. [27] Seltzer R., Newman J., and M. Leighton C. (1997), Sex as a Political Variable, London: Lynne Rienner Publishers. 31

[28] Shapiro, R. and H. Mahajan (1986), “Gender Differences in Policy Preferences: A Summary of Trends from the 1960’s to the 1980’s,” Public Opinion Quarterly, 50(1), 42-61. [29] Stutzer, A. and L. Kienast (2005), “Demokratische Beteiligung und Staatsausgaben:

Die

Auswirkungen des Frauenstimmrechts,” Swiss Journal of Economics and Statistics, 141(4), 617650. [30] Trechsel, A. and U. Serduelt (1999), Kaleidoskop Volksrechte: Die Institutionen der direkten Demokratie in den Schweizerischen Kantonen 1970-1996, Basle and Munich: Helbing and Lichtenhahn.

A

Canton-Level Panel Data

The data on canton expenditures and revenues are taken from the annual collections on public finances of Switzerland (Federal Department of Finance, various years). Data are available on paper before 1980 and electronically after that. We calculate real per capita expenditures (in 1,000 Swiss Francs and 2000 prices) for the following categories: total cantonal expenditures, security, education, welfare, health, transportation, environment, and agriculture. For the years 1967 and 1968, no separate expenditure data for Cantons and communities are available, but only expenditures for Cantons and their communities together. Values for these years are obtained by linear interpolation. Expenditures for environmental protection are only available since 1970. Total revenue data are available for all years except the years 1968 and 1969. Cantonal shares from federal revenues are unavailable for the years 1968, 1969 and 1990-1993. As for federal subsidies, data are not available for the years 1950-1952, 1968-1977 and 1990-1993. The missing values after 1952 are obtained by linear interpolation. For the canton characteristics, most variables are from the decennial population census with intermediate values interpolated. Data for the population in rural and urban areas is only available 32

since 1970. The education variable is measured as the share of high-school graduates in percentage of the 19 year-old population. Data on average per capita income in the cantons is available since 1965. The unemployment rate is calculated as the number of registered unemployed relative to the active population from the State Secretariat for Economic Affairs after 1975 and as the number of unemployed in percentage of employed persons from the population census before 1975. Population density is measured as the log of the number of people (in 1,000) per square kilometer. Variables of the household structure are again from the population census and include: the percentage of single-parent households, the share of married people in the group older than 20, the group of divorced people in the same age group. Also from the population census are data on the percentage of foreigners in the total population, the share of Catholics and citizens in different age classes (i.e. between 0 and 19, between 20 and 39, between 40 and 64, between 65 and 79, older than 80). Female labor force participation is measured as the share of women older than 15 who work. Information on the existence of a mandatory budget referendum is taken from Trechsel and Serduelt (1999), who systematically collected information for cantons without a town-meeting from 1970 to 1996. For earlier years and the town-meeting cantons, we gather data from the Public Record Offices of all cantons and supplemented any missing information using old canton laws and constitutions.

33

Table 1: Introduction of Women Suffrage in Swiss Cantons

Women Suffrage Adopted in Canton (Year)

Yes Votes in Federal Referendum of 1959 (%)

Yes Votes in Federal Referendum of 1971 (%)

Neuchatel (NE) Vaud (VD) Geneva (GE) Basle-City (BS) Basle-County (BL) Ticino (TI) Zurich (ZH) Valais (VS) Lucerne (LU)

1959 1959 1960 1966 1968 1969 1970 1970 1970

51,3 52,2 60,0 46,8 37,3 37,1 36,2 30,5 21,3

83,9 82,0 91,9 82,2 79,9 75,3 66,8 79,9 62,7

Zug (ZG) Fribourg (FR) Schaffhausen (SH) Aargau (AG) Berne (BE) Glarus (GL) Solothum (SO) Thurgau (TG)

1971 1971 1971 1971 1971 1971 1971 1971

29,8 24,3 31,9 22,8 35,5 19,1 30,0 19,9

71,1 59,9 56,7 50,2 66,5 41,3 64,1 44,1

St. Gallen (SG) Uri (UR) Schwyz (SZ) Graisons (GR) Nidwalden (NW) Obwalden (OW) Jura (JU) Appenzell Aussenrhoden (AR) Appenzell Innerrhoden (AI)

1972 1972 1972 1972 1972 1972 1977 1989 1990

19,3 14,6 14,2 22,4 19,5 14,3 N/A 15,5 4,9

46,5 36,3 42,2 54,8 55,8 46,7 N/A 39,9 28,9

Canton

Notes : The table reports the year women were given the right to vote at the cantonal level and the fraction of voting men that supported the introduction of women suffrage in the two federal referendums 1959 and 1971. Since the canton Jura separated from Bern only in 1977 to become an independent canton, no separate results for the federal referendums are available. Adoption of women suffrage was voluntary in all cantons except one. Appenzell-Innerhoden was forced to adopt women suffrage by the Swiss Supreme Court in 1990.

Table 2: Summary Statistics, VOX-Data

Men Std. Dev

Women Mean Std. Dev

Mean

Demographics Age Protestant Have Kids Single Married Divorced

48,39 0,47 0,41 0,19 0,66 0,05

16,38 0,50 0,49 0,39 0,47 0,21

49,98 0,47 0,33 0,23 0,71 0,03

17,41 0,50 0,47 0,42 0,45 0,17

14,92 -1,08 -13,74 15,57 16,14 -12,02

Education, Work and Income Education: Compulsory Education: Apprentice/Spec Schools Education: University Employed Income House Ownership

0,19 0,74 0,07 0,51 1,76 0,46

0,40 0,44 0,26 0,50 0,82 0,50

0,10 0,77 0,13 0,69 2,21 0,50

0,30 0,42 0,34 0,46 1,06 0,50

-44,52 13,54 31,86 57,72 13,34 9,88

Region of Residence Urban Nongerman

0,66 0,26

0,47 0,44

0,64 0,24

0,48 0,43

-6,77 -4,53

Political Position and Participation Self-Placement Left-Right (0-10) Turnout

4,79 54,35

1,77 49,81

5,09 62,02

1,89 48,53

22,38 32,72

T Statistic Difference

Notes: The summary statistics are based on the sample of voters. Education measures the highest degree, which is either compulsary school, vocational school or university (all dummy variables). Employed is a dummy variable capturing the employment status, and income measures household income in 5 income-classes. House Ownership is a dummy variable whether the person's household owns a house. Married, single and divorced are dummy variables describing the civil status of the respondent. Age is measured in years, protestant, have kids, urban area and non-german speaking area are all dummy-variables. Party-Ideology is measured on a 0-10 left-right scale, where higher number indicate a more right-wing political position. The turnout variable whether the respondent voted in federal parliamentary elections. The last column shows the T-test statistic for differences in means between men and women.

Table 3: Voting Behavior of Men and Women in Federal Propositions

Welfare Reduce Decrease Support for the Unemployment Retirement Benefits Age Disabled

Environment Protecting the Environment

Transport Against further Road Construction

Military Less Military

Agriculture Against Subsidizing Agriculture

Education Free Education

Health Subsidies Health Insurance

(1) Female Dummy

0,086 (0.015)***

0,044 (0.022)**

0,042 (0.020)**

0,098 (0.035)***

0,068 (0,059)

0,065 (0.029)**

-0,053 (0.032)*

0,063 (0.019)***

0,146 (0.045)***

0,062 (0.025)**

(2) Add Income, Job Marriage, Demographics

0,075 (0.016)***

0,023 (0,024)

0,05 (0.025)**

0,118 (0.038)***

0,001 (0,069)

0,029 (0,032)

-0,045 (0,035)

0,05 (0.021)**

0,118 (0.048)**

0,049 (0.029)*

Observations Log-Likelihood

4838 -2975

1969 -1156

2089 -1182

688 -412

251 -137

949 -521

952 -507

2500 -1490

508 -320

1450 -847

(3) Sample 1988-2003

0,078 (0.017)***

0,092 (0.030)***

0,058 (0.027)**

0,192 (0.047)***

0,025 (0,035)

-0,063 (0.037)*

0,052 (0.022)**

0,153 (0.051)***

0,086 (0.049)*

(4) Add Party Ideology

0,049 (0.018)***

0,082 (0.030)***

0,02 (0,028)

0,18 (0.047)***

0,008 (0,035)

-0,046 (0,038)

0,044 (0.023)*

0,116 (0.054)**

0,053 (0,053)

Observations Log-Likelihood

3849 -2222

1487 -808

1854 -956

438 -237

801 -413

848 -423

2189 -1228

461 -261

558 -251

Longer Maternity Leave

Notes: The table reports estimates from a probit model, with marginal coefficients of the female dummy being displayed. The dependent variable is the voting decision, which is equal to one if the respondent supported the proposition and zero otherwise for the propositions shown at the top. The table reports the coefficient on the female dummy. All specifications (rows (1)-(4)) include canton and year fixed effects. The first specification displays the coefficient on the gender dummy without individual controls. The second specification adds the demographics age, education, religion, language and urban area as well as income, civil status and employment measures. Income is proxied by a dummy for houseownership. The employment and civil status variables are binary variables equal to one 1 if the individual is employed or married respectively, and zero otherwise. The fourth specification adds the party ideology, which captures self-placement on a left-right scale, with values from 0-10 and 10 being maximally right. Since party-ideology is only reported since 1988, specification (3) estimates the second specification for the 1988-2003 period. Since the vote on free education was in 1986, no data on the left-right position is available. Robust standard errors in parentheses.

Table 4: Voting in Favor of More Government in Federal Propositions

Size of Government More Less Government Debt

More Environment

More Transport

Scope of Government More More More Defense Agriculture Education

More Health

More Welfare

Female Dummy

0,027 (0.007)***

-0,025 (0,018)

0,143 (0.025)***

0,032 (0.020)*

-0,061 (0.018)***

-0,056 (0.023)**

0,066 (0,051)

0,075 (0.022)***

0,071 (0.015)***

Add Controls

0,024 (0.008)***

-0,028 (0,020)

0,093 (0.028)***

0,03 (0,021)

-0,065 (0.023)***

-0,077 (0.027)***

0,126 (0.060)**

0,046 (0.026)*

0,059 (0.016)***

Year and Canton Dummies Observations Log Likelihood

Yes 19449 -12861,6

Yes 2121 -1080,9

Yes 1529 -921,4

Yes 2429 -1371,7

Yes 2116 -1250,1

Yes 1531 -875,5

Yes 353 -224,2

Yes 1593 -921,3

Yes 4412 -2787,9

Notes : The table resports the marginal effects from a probit model whether the respondent supported a proposition, which would have increased government spending in the respective policy area or opposed it. The classification of the financial consequences of the propositions was done using official documents of the Swiss government. Table A3 shows a list of the federal propositions underlying each column. The table resports the coefficient on the female dummy variable in each row. The specifications (1)-(4) are the same as in Table 3. See notes to Table 3 for further details. Robust standard errors are reported in parentheses.

Table 5: Gender Gaps in 1975

How Important is the Problem?

Protect the Environment

Provide Health Care

Provide Education

Redistribute Income

Care for Elderly

0,0608 (0.0301)**

0,0763 (0.0291)***

-0,0304 (0,0329)

-0,0237 (0,0417)

0,1305 (0.0294)***

0,0353 (0,0395)

0,0195 (0,0385)

-0,0156 (0,0448)

-0,0865 (0,0527)

0,0862 (0.0382)**

3,70 Yes 1226 0,06

3,72 Yes 1230 0,07

3,65 Yes 1231 0,06

3,29 Yes 1226 0,1

3,63 Yes 1233 0,07

Female Dummy

0,0375 (0,0363)

0,0376 (0,0415)

0,054 (0,0394)

0,0487 (0,0471)

-0,0435 (0,0361)

Add Controls

0,03 (0,0463)

-0,0148 (0,0550)

0,1001 (0.0463)**

-0,0263 (0,0591)

-0,0803 (0.0452)*

3,49 Yes 1222 0,05

3,32 Yes 1229 0,1

3,34 Yes 1227 0,08

3,02 Yes 1223 0,09

3,28 Yes 1231 0,08

Female Dummy

Add Controls

Mean Value Canton Dummies Observations R Squared Is it a Government Responsibility?

Mean Value Canton Dummies Observations R Squared

Notes : The dependent variable in the upper part is whether the respondent thinks it is an important problem. The responses range from 1 (absolutely unimportant) to 4 (very important). The dependent variable in the lower part is whether the government is considered responsible for providing the services shown in the respective columns. The response ranges from 1 (no responsibility) to 4 (very essential responsibility). The first rows contains only the female dummy as well as canton dummies. The second rows adds controls for demographics (age, education, language spoken, religion, divorced, single), employment and household income to the specification. Robust standard errors are reported in parentheses. Source : Political Action Study, 1975

Table 6: Women Become more Left-Wing Over Time

(1) Female Dummy

Left-Right Position in 1975 (2) (3)

0,364 (0.127)***

0,418 (0,264) 0,271 (0.133)** -0,232 (0,250) 0,012 (0.005)** -0,141 (0,443) -0,079 (0,202) 0,12 (0,139) -0,216 (0,293) 0,04 (0,250) -0,529 (0,754) -0,762 (0.172)*** -0,688 (0.184)*** -0,562 (0.237)**

-0,177 (0.052)**

-0,238 (0.053)** -0,05 (0,088) -0,118 (0,096) 0,022 (0.002)** -0,131 (0,096) -0,275 (0.067)** 0,138 (0.058)* 0,027 (0,161)

-0,261 (0.054)** -0,014 (0,090) -0,118 (0,099) 0,02 (0.002)** -0,083 (0,098) -0,277 (0.068)** 0,14 (0.058)* 0,022 (0,161) -0,174 (0.063)** -0,207 (0,198) -0,128 (0.065)* 0,027 (0,080) 0,496 (0.110)**

(8)

Mean Left-Right Position

5,84

5,84

5,84

5,84

5,22

5,22

5,22

Canton Dummies Observations R Squared

Yes 1014 0,05

Yes 974 0,08

Yes 974 0,10

Yes 974 0,11

Yes 6595 0,01

Yes 6530 0,06

Yes 6530 0,06

Yes 6530 0,06

Higher Education Age Divorced Single Katholic French-Speaking Employed Unemployed Medium Income High Income Refuse Answer Income

0,298 (0.153)* 0,264 (0.133)** -0,274 (0,248) 0,011 (0.004)** -0,134 (0,390) -0,098 (0,161) 0,115 (0,140) -0,217 (0,294) -0,067 (0,168) -1,209 (0.488)** -0,765 (0.170)*** -0,69 (0.182)*** -0,567 (0.236)**

(5)

0,014 (0,106) 0,006 (0,090) -0,103 (0,099) 0,018 (0.002)** -0,105 (0,149) -0,322 (0.102)** 0,139 (0.058)* 0,027 (0,161) -0,041 (0,094) 0,018 (0,296) -0,112 (0,065) 0,046 (0,080) 0,51 (0.110)** 0,036 (0,085) 0,075 (0,198) 0,006 (0,129) -0,279 (0.118)* -0,428 (0,393) -0,303 (0.113)** 5,22

Medium Education

0,409 (0.127)*** 0,24 (0.131)* -0,362 (0,245) 0,018 (0.004)*** 0,036 (0,382) 0,059 (0,160) 0,114 (0,143) -0,233 (0,302)

Left-Right Position in 1995 (6) (7)

(4)

Have Children Female*Divorced

0,127 (0,838) -0,043 (0,350) -0,152 (0,329) -1,473 (0.823)*

Female*Single Female*Employed Female*Unemployed Female*Have Children

Notes : The table reports coefficients from a regression where the dependent variable is the respondent's left-right position ranging from 0 (very left-wing) to 10 (very right-wing). Column (1)-(4) shows the result for the sample in 1975, while columns (5)-(8) for the sample in 1995. The first specification (columns (1) and (5)) contains only the gender variable and canton dummies. The second specification (columns (2) and (6)) adds individual demographics (age, education, religion, language), while the third specification (columns (3) and (7)) also includes controls for employment status and household income. The missing group of the employment status is retired/housewife. The final specification (columns (4) and (8)) add interaction terms between female and employment status, female and marital status as well as between female and having children [available in 1995 only]. Robust standard errors are reported in parentheses. Source : Political Action Study (1975); Selects Study (1995)

Table 7: Summary Statistics for Canton-Level Data

Mean

Std. Dev

Min

Max

1247 801 1124 1251 1174 1247 1247 1247

5086 224 944 452 478 581 1064 751

3296 169 1112 273 465 562 828 782

597 0 0 0 0 35 40 15

19227 1445 12309 1717 3642 3602 4646 5140

Revenues and Deficit Per Capita (in 2000 SFr) Overall 1246 Federal Subsidies 868 Shares on Federal Revenues 1143 Deficit 1296

4939 748 302 -62

3212 823 244 608

374 50 0 -2983

20392 6129 4457 9473

Control Variables Age 0 to 19 (%) Age 20 to 39 (%) Age 40 to 64 (%) Age 65 to 79 (%) 80 and Older (%) Population Density (per Km2) Unemployment Rate (%) Foreigners (%) Catholics (%) Single Parents (%) Divorced (%) Female Labor Force Participation (%) Education Share Commuters (%) Language: Nongerman

1297 1297 1297 1297 1297 1297 1297 1297 1297 1297 1297 1297 1297 1297 1297

29,3 29,4 28,7 10,0 2,5 431 0,9 13,6 58,5 5,7 3,3 41,4 9,2 4,6 0,3

5,4 2,4 2,5 1,8 1,1 1094 1,4 6,6 25,6 1,2 2,1 6,9 5,8 4,6 0,4

15,7 23,1 23,2 6,2 0,8 19 0,0 2,5 8,1 3,8 0,4 24,3 0,0 0 0,0

40,5 35,3 36,5 15,6 6,3 6350 7,8 38,1 96,2 11,6 9,6 57,8 32,7 23,2 1,0

Additional Controls (Robustness) Rural (%) Mean Annual Income Signature Requirement Initiative Mandatory Budget Referendum Mandatory Law Referendum Share of Left Parties in Parliament

797 896 1265 1297 1297 1195

38,7 7392 3,8 0,64 0,56 21,5

31,2 10282 3,6 0,48 0,50 12,4

0,0 82 0,0 0,0 0,0 0,0

100,0 78174 17,4 1,0 1,0 51,0

Observations Expenditures Per Capita (in 2000 SFr) Overall Environment Transport Security Agriculture Welfare Education Health

Notes: The table reports summary statistics over the whole sample period (1950-2000). The unit is a canton-year observation. All expenditures, revenues and mean cantonal income are deflated to 2000 Swiss Francs. The dummy for the mandatory budget referendum measures whether the Canton has mandatory referendum on (ordinary) expenditures exceeding a certain threshold level. Education is measured as the percentage of the population with a tertiary education. Nongerman denotes the fraction of cantons where the dominant language is any language other than German (Italian, French).

Table 8: Women Suffrage Adoption and Fiscal Policy

Size of Government Expenditures Revenues Deficit Women Suffrage

Environment Transport

Scope of Government Security Agriculture Education

Health

Welfare

-0,033 (0.019)*

0,001 (0,032)

-0,032 (0,059)

0,25 (0,195)

0,087 (0.051)*

-0,037 (0.019)*

-0,055 (0,059)

-0,063 (0.023)***

-0,092 (0.045)**

-0,049 (0.029)*

0,016 (0.007)** -0,462 (0.089)*** -0,004 (0.002)** -0,01 (0.003)*** 0,025 (0.012)** -0,119 (0.023)*** -0,009 (0.003)*** -0,008 (0.002)*** 0,001 (0,016) 0,126 (0.020)***

0,004 (0,010) -0,554 (0.117)*** 0 (0,003) -0,006 (0,004) 0,035 (0.018)* -0,094 (0.030)*** -0,015 (0.004)*** -0,01 (0.002)*** 0,019 (0,029) 0,141 (0.022)***

-0,071 (0.030)** -1,019 (0.557)* 0,004 (0,008) 0,029 (0.015)* 0,015 (0,071) -0,004 (0,089) 0,006 (0,011) -0,008 (0,007) 0,102 (0.055)* 0,03 (0,078)

0,079 (0.027)*** 1,348 (0,996) 0,025 (0.013)** -0,087 (0.023)*** -0,333 (0.126)*** -0,788 (0.125)*** -0,031 (0,030) -0,002 (0,012) 0,027 (0,081) 0,236 (0.075)***

0,073 (0.023)*** 0,012 (0,265) -0,017 (0.006)*** -0,035 (0.008)*** 0,007 (0,040) -0,194 (0.054)*** -0,015 (0.009)* -0,002 (0,005) -0,03 (0,031) 0,346 (0.059)***

-0,012 (0.007)* -0,321 (0.092)*** -0,001 (0,002) -0,016 (0.003)*** 0,011 (0,012) -0,141 (0.015)*** 0,000 (0,003) -0,006 (0.001)*** 0,001 (0,008) -0,029 (0.014)**

0,004 (0,018) -2,122 (0.230)*** -0,018 (0.005)*** 0,004 (0,008) 0,056 (0.034)* -0,19 (0.047)*** 0,020 (0.008)** -0,017 (0.005)*** -0,006 (0,062) 0,104 (0.038)***

-0,033 (0.009)*** 0,213 (0.108)** -0,01 (0.003)*** -0,004 (0,004) -0,038 (0.015)** -0,149 (0.025)*** -0,004 (0,003) -0,003 (0,002) 0,058 (0,036) 0,07 (0.018)***

0 (0,019) 1,176 (0.261)*** 0,006 (0,006) 0,014 (0,009) 0,086 (0.030)*** -0,484 (0.060)*** -0,055 (0.007)*** 0,016 (0.004)*** 0,081 (0.048)* 0,108 (0.037)***

0 (0,011) -0,337 (0.154)** 0,012 (0.003)*** -0,016 (0.005)*** -0,003 (0,019) 0,009 (0,031) 0,001 (0,005) -0,009 (0.003)*** 0,041 (0.013)*** 0,001 (0,022)

Age-Classes Canton-Fixed Effects Time-Fixed Effects

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Yes Yes Yes

Observations R-squared

1200 0,97

1200 0,92

1200 0,55

775 0,7

1200 0,81

1200 0,97

1200 0,9

1200 0,97

1200 0,93

1200 0,97

Unemployment Rate Log Population Density % with Higher Education % Commuters % Single Parents % Divorced Female Labor Force Participation % Katholic Share Federal Revenues Federal Subsidies

Notes: The dependent variable is the log of real per capita expenditures in the respective category. The only exception is the deficit in column (3), which is measured in levels. All expenditures are measured in 1000 Swiss Franks at 2000 prices. Environmental expenditures are only available since 1970. The main independent variable "Women Suffrage" is equal to one if women suffrage has been introduced in the canton and zero otherwise. Population density is measured as the log of the number of people (in 1000) per square kilometer. The unemployment rate, the share of highly educated, the share of divorced people, the share of single parents, the female labor force participation rate and the fraction of katholics are all measured in percentage of the respective population. All specifications include canton and year fixed effects as well as controls for the age structure of the population. Robust standard errors are reported in parentheses.

Table 9: Dynamic Effects of Women Suffrage

Size of Government Expenditures Revenues Deficit Years since Adoption Years Squared R Squared

Years 1 to 5 Years 6 to 10 Years 11 to 15 Years 16 to 20 More than 21 Years

Canton Controls Canton Dummies Year Dummies Observations R Squared F-Test Year Dummies P-value

Environment

Transport

Scope of Government Security Agriculture Health

Education

Welfare

-0,0113 (0.0028)*** 0,0003 (0.0001)*** 0,97

-0,0146 -0,0106 (0.0046)*** -0,0118 0,0001 -0,0006 (0,000) (0.0003)** 0,92 0,56

-0,0262 -0,0233 0,0001 -0,0004 0,7

-0,0124 (0.0074)* 0,0005 (0.0002)*** 0,81

-0,0034 -0,0028 0,0001 -0,0001 0,97

-0,0512 0,0193 -0,0049 0,0082 (0.0079)*** (0.0073)*** -0,0039 (0.0050)* 0,0009 0,0005 0,0003 0,0002 (0.0002)*** (0.0002)** (0.0001)*** (0.0001)** 0,91 0,93 0,97 0,97

-0,013 (0,022) -0,076 (0.024)*** -0,102 (0.030)*** -0,111 (0.036)*** -0,058 (0,046)

-0,005 (0,026) -0,027 (0,057) -0,097 (0.041)** -0,131 (0.057)** -0,094 (0,061)

-0,077 (0,062) -0,054 (0,087) -0,092 (0,113) -0,084 (0,143) -0,204 (0,181)

0,028 (0,143) -0,064 (0,157) 0,01 (0,186) -0,058 (0,212) -0,11 (0,247)

0,102 (0.061)* -0,004 (0,062) -0,077 (0,077) -0,156 (0.095)* 0,019 (0,126)

-0,014 (0,021) -0,038 (0.023)* -0,036 (0,027) -0,027 (0,036) 0 (0,041)

-0,056 -0,019 (0,069) (0,048) -0,335 -0,008 (0.069)*** (0,065) -0,362 0,055 (0.083)*** (0,082) -0,4 0,155 (0.098)*** (0,096) -0,428 0,333 (0.115)*** (0.122)***

Yes Yes Yes 1200 0,97 7,19 (0,000)

Yes Yes Yes 1200 0,92 2,41 (0,035)

Yes Yes Yes 1200 0,55 0,72 (0,612)

Yes Yes Yes 775 0,7 0,61 (0,690)

Yes Yes Yes 1200 0,81 4,39 (0,001)

Yes Yes Yes 1200 0,97 2,02 (0,074)

Yes Yes Yes 1200 0,91 6,4 (0,000)

Yes Yes Yes 1200 0,93 3,76 (0,002)

-0,035 (0,026) -0,062 (0.032)* -0,074 (0.042)* -0,037 (0,047) 0,053 (0,057)

-0,012 (0,031) 0,021 (0,039) 0,037 (0,052) 0,064 (0,058) 0,202 (0.073)***

Yes Yes Yes 1200 0,97 6,27 (0,000)

Yes Yes Yes 1200 0,97 5,05 (0,000)

Notes: The table shows results of a regression of per capita canton expenditures in a given year and policy area on the years since suffrage and years squared (top panel) or 5-years dummy variables to allow for nonlinear effects of women suffrage on expenditures (bottom panel). The other control variables included are the same as in Table 8. All specifications include canton and year fixed effects. Robust standard errors are reported in parentheses. See notes to Table 8 for a description of the variables.

Table 10: Voter Preferences and the Adoption of Women Suffrage

Voter Preferences (1) (2) Early Adopters

0,072 (0.018)***

Linear Trend Trend*Early Adopters

(3)

Suffrage Adoption (4) (5)

0,071 (0.028)** 0,001 (0,023) 0,002 (0,037)

Indicator 1971

0,371 (0.052)*** 0,562 (0.258)**

Voter Preferences

0,339 (0.042)*** 0,014 (0,219)

Growth Rate Expenditures

Border Canton Population Density Foreign Tourists per Capita Fraction Higher Education

No Yes 50 0,25

No Yes 50 0,25

0,289 (0.085)***

0,357 (0.069)***

0,027 (0,067)

0,007 (0,032) -0,015 (0,020) -0,002 (0,004) 0,009 (0.005)**

0,171 (0.031)*** -0,011 (0,026) 0,03 (0.014)** 0,004 (0,003) 0 (0,003)

0,007 (0,034) -0,026 (0,020) -0,002 (0,005) 0,01 (0.005)**

-0,016 (0,052) 0,174 (0.025)*** -0,012 (0,026) 0,031 (0.014)** 0,005 (0,003) 0 (0,003)

Yes No 50 0,89

Yes No 50 0,94

Yes No 50 0,87

Yes No 50 0,94

Language Nongerman

Age Dummies Decade Dummies Observations R-squared

(6)

Notes : Column (1)-(2) report results from a regression, where the dependent variable is support for increased government spending in a decade.

Column (1) only contains a time dummy and whether the canton adopted women suffrage before or in 1971 ("Early Adopters"). Column (2) includes an interaction term between the linear trend and early adoption. Columns (3)-(6) report results from a regression, where the dependent variable is voter support for women suffrage in the federal referendums in 1959 and 1971. All specifications include whether the canton is on the border, number of foreign tourists per capita, population density, fraction with university education and controls for the age structure. Changes over time in the propensity to support suffrage are captured by a dummy for the 1971 ballot ("Indicator 1971"). Column (3) includes average support for higher government spending in federal propositions ("Voter Preferences"), while column (4) adds the dominant language in a canton. Column (5) tests for reverse causality by including the average growth rate in canton expenditures, while column (6) again adds the language variable. Robust standard errors are reported in parentheses.

Table 11: Robustness to Inclusion of Additional Controls

Income (1)

Expenditures Left Party (2)

Direct (3)

Suffrage Adoption

-0,024 (0,019)

-0,061 (0.026)**

-0,07 (0.020)***

-0,0558 (0.0328)*

-0,0255 (0,027)

-0,102 (0.0275)***

Other Controls Year Dummies Canton Dummies Observations R Squared

Yes Yes Yes 875 0,94

Yes Yes Yes 1173 0,97

Yes Yes Yes 1200 0,97

Yes Yes Yes 875 0,96

Yes Yes Yes 1173 0,97

Yes Yes Yes 1243 0,97

Years Since Adoption Years Squared

Other Controls Year Dummies Canton Dummies Observations R Squared

-0,018 -0,005 -0,01 (0.003)*** (0,004) (0.003)*** 0,0002 0,0005 0,0004 (0.0001)*** (0.0001)*** (0.0001)*** Yes Yes Yes 875 0,94

Yes Yes Yes 1173 0,97

Yes Yes Yes 1200 0,97

Welfare Expenditures Income Left Party Direct (4) (5) (6)

0,0084 (0,006) 0,0001 (0,000) Yes Yes Yes 875 0,96

0,0232 -0,0046 (0.0048)*** (0,004) 0,0003 0,0004 (0.0001)*** (0.0001)*** Yes Yes Yes 1173 0,98

Yes Yes Yes 1200 0,97

Notes : The table reports results of a regression where the dependent variable is canton expenditures (Columns (1)-(3)) and welfare expenditures (columns (4)-(6)) respectively. The top part shows the coefficients on suffrage adoption (as in Table 8), while the bottom part shows the coefficient of years since adoption (as in Table 9). Columns (1) and (5) adds mean cantonal income, which is available since 1965. Columns (2) and (6) add the percent of left-party seats in the cantonal parliament. Columns (3) and (7) include three measures of direct democratic instruments: whether the canton has a mandatory budget or law referendum and the signature requirements in percent of eligible voters of the voter initiative. All specifications include canton and year fixed effects as well as the same controls as in Table 8. Robust standard errors are reported in parentheses.

Table 12: Political Turnout Behavior Cannot Account for Dynamic Suffrage Effect

Years 1 to 5 Years 6 to 10 Years 11 to 15 Years 16 to 20 More than 21 Years

Other Controls Observations R Squared

Overall Turnout (1)

Female Turnout (2)

Relative Turnout (3)

-10,507 (1.53)*** -8,061 (2.01)*** -6,275 (2,659) -4,037 (3,099) -3,878 (3,949)

0,517

0,868

0,542

0,914

0,559

0,944

0,582

0,977

0,583

0,995

Yes 920 0,99

Overall Welfare Health Education Expenditures Expenditures Expenditures Expenditures (4) (5) (6) (7) -0,027 (0,023) -0,092 (0.026)*** -0,108 (0.032)*** -0,114 (0.038)*** -0,061 (0,047)

-0,021 (0,036) 0,017 (0,045) 0,027 (0,057) 0,05 (0,062) 0,184 (0.076)***

-0,053 (0,053) -0,034 (0,076) 0,041 (0,092) 0,141 (0,102) 0,322 (0.128)**

-0,044 (0,028) -0,071 (0.036)** -0,082 (0.046)* -0,043 (0,050) 0,047 (0,060)

Yes 1200 0,97

Yes 1200 0,97

Yes 1200 0,93

Yes 1200 0,97

Notes : The table analyzes turnout behavior in cantonal elections and its effect for the evolution of expenditures. Column (1) regresses aggregate turnout in canton elections on 5-year dummies for years since suffrage adoption and the same controls as in Table 8. The next two columns show the evolution of estimated female participation (column (2)) and relative turnout rates (column (3)) based on the estimates from column (1). See Section 6.3. on the details of calculation. Relative turnout rates vary from zero prior to adoption to slightly above one (as the percentage of women in the eligible population is 51.4 percent). The right-hand side of the table reports the effect of estimated relative turnout on overall expenditures (column (4)) as well as welfare, health and education expenditures (columns (5)-(8)). The control variables are the same as in Table 8 and include canton and time fixed effects. Standard errors in (4) to (8) are bootstrapped to account for first-stage estimation of relative female turnout.

Figure 1: Political Gender Gap in Federal Elections, 1971 - 2003

35

Percentage Left Parties

32,5 30 27,5 25 22,5 20 1971

1979

1987

1995

2003

Year of Federal Election Women

Men

Notes : The party reports the percentage of men and women voting for a left party in federal elections. Since the party system in Switzerland is fragmented with many small and regional parties, voting for a left party includes any of the following parties: Social Democratic Party of Switzerland (SPS), Labor Party (PdA), Socialist Party (PSA), Progressive Organisations of Switzerland (POCH), Feminist and Green-Alternative Associations (FGA), Green Party of Switzerland (GPS), Green Alliance of Switzerland (GBS) and the Solidarities. Source : Kueffer (2004)

Figure 2: Turnout Gap in Federal Propositions

Male Turnout Rate - Female Turnout Rate

0,2 0,18 0,16 0,14 0,12 0,1 0,08 0,06 0,04 0,02 0 1977

1981

1985

1989 Year

Source : VOX Surveys, 1977-2003

1993

1997

2001

Table A1: Federal Propositions with the Largest Gender Gap Title of Proposition

Initiative for Reducing Tobacco Consumption Equal Representation of Women in Federal Government Change in Marital Law Against Racial Discrimination in the Military Against Subsidizing Corn Production Initiative for Reducing Alcohol Consumption For Saving the Waters For a Car-Free Sunday per Quarter For Abandoning Nuclear Energy For Equal Rights of the Disabled Initiative for Equal Rights of Men and Women

Vote Number

Year of Vote

Gender Gap (percentage)

404 461 336 414 413 403 381 498 365 500 306

1993 2000 1985 1994 1994 1993 1992 2003 1990 2003 1981

0,177 0,175 0,170 0,168 0,156 0,155 0,153 0,149 0,147 0,146 0,145

Notes : The first column reports the official number of the vote, the second column the year the vote was held. The final column shows the gender gap, the percentage of women approving the proposition minus the percentage of men. Positive numbers imply that women were more supportive of the proposition than men.

Source: VOX Surveys, 1981-2003, Sample of Voters.

Table A2: Titles of Federal Propositions by Policy Area Policy Areas and Title of Proposition

No.

Year Result

Policy Areas and Title of Proposition

Environmental and Nuclear Policy

Social Security

For Environmental Protection

Longer Maternity Leave

No.

Year Result

For Protection of the Moor

349

1987

Yes

For Protecting Motherhood

323

1984

No

For Protecting the Waters

377

1992

Yes

Law on Maternity Insurance

458

1999

No

For Saving the Waters

381

1992

No

Reducing Unemployment Benefits

For Protecting the Alpine Area

408

1994

Yes

Measures for Unemployment Insurance

398

1993

Yes

For Reducing Traffic

463

2000

No

Financing Unemployment Insurance

437

1997

Yes

More Money for Solar Energy

465

2000

No

For Decreasing Retirement Age

Energy-Environment Initiative

466

2000

Yes

For Reform of Public Pension System

444

1998

No

For Taxing Energy instead of Labor

481

2001

No

For a more Flexible Public Pension System

469

2000

No

For a Car-Free Sunday per quarter

498

2003

No

Flexible Retirement Age of 62 for Women and Men

470

2000

No

500

2003

No

Against Nuclear Energy

Supporting the Disabled

For a Future without Nuclear Power Plants

321

1984

No

For Abandoning Nuclear Energy

365

1990

No

Equal Rights for the Disabled

Stop New Nuclear Power Plants

366

1990

Yes

Health

For Electricity without Nuclear Energy

501

2003

No

For an Affordable Health Insurance

373

1992

No

Stop New Nuclear Power Plants

502

2003

No

Initiative for Financially Sustainable Health Insurance

416

1994

No

Health Initiative

499

2003

No

Transport Policy Defense Spending

Against further Road Construction Stop Road Construction

359

1990

No

Less Military

Against A Freeway from Murten to Yverdon

360

1990

No

For a Switzerland without Army

357

1989

No

Against Freeway in the "Knonauer Amt"

361

1990

No

Against Purchasing Fighter Planes

393

1993

No

Pretcting the Landscape near River "Aare"

362

1990

No

No

For Public Transport Promoting Public Transport

370

1991

No

For a Railway though the Alps

382

1992

Yes

For Improving Public Transport

445

1998

Yes

Against subsidizing Parking Space Against Subsidizing Parking at Train Stations

471

2000

482

2001

No

Change Military Law

495

2003

Yes

Abolish Subsidizing Wheat Price

333

1985

Yes

Abolish Reduction of Wheat Price

413

1994

Yes

Agricultural Policy Against Subsidizing Acriculture

429

1996

Yes

Relax Speed Limit Relax Speed Limit to 130/100

Reduce Military Spending For Switzerland without Army

358

1989

No

For Liberalizing Agriculture For an Ecological and Competitive Agriculture

418

1995

No

Education

Resolution on Dairy Farming

419

1995

No

For free Education

Resolution on Agriculture

420

1995

No

For a new Corn Article

446

1998

Yes

Guaranteed Education and Retraining

340

1986

No

Notes: The table reports all federal votes between 1981 and 2003, which fall into one of the following seven broad categories: Environmental Politics and Nuclear Policy, Transportation Politics, Agricultural Policy, Social Security, Education, Health and Military Policy. Results of the federal votes can be found at http://www.admin.ch/ch/d/pore/va/index.html.

Table A3: Propositions with Predictable Financial Consequences

No.

Increase Expenditures

Result

No.

Increase Subsidies/Grants

Result

No.

Decrease Federal Debt

313

Energy Article

No

333

Self-Supply with Corn (A)

Yes

400

For healthy Federal Finances

Result Yes

323

Protection Motherhood (W)

No

335

Risk Guarantee for Small/Medium Enterprises

No

421

For Controlling Expenditures

Yes

339

Culture Initiative

No

341

For Domestic Sugar Production (A)

No

439

For Budget Balancing

Yes

340

Guarantee Vocational Retraining (Edu)

No

425

Revision Language Article

Yes

480

For Controlling Debt

Yes

348

Railway 2000 (T)

Yes

349

Protection Moor (Env)

Yes

350

Change Health Insurance (H)

No

308

Improving Federal Budget

Yes

326

Against Contributions for Primary School (Edu)

Yes

363

Vine Cultivation (A)

No

312

New Regulation Fuel Taxes

Yes

327

Against Federal Contributions for Health (H)

Yes

367

Energy Article

Yes

316

For Taxing Heavy Traffic

Yes

328

Resolutions on Contributions on Education (Edu)

No

370

Promoting Public Transport (T)

No

317

Fees for Road Use

Yes

413

Against Corn Subsidies (A)

Yes

373

For Stable Finances of Health Insurance (H)

No

324

Law on Radio and TV

Yes

428

Against Duty on Schnaps Purchases

Yes

377

Protection of Waters (Env)

Yes

331

Against Cantonal Share in Federal Stamp Duty

Yes

429

Against Federal Contributions for Parking Space

Yes

381

Saving the Waters (Env)

No

332

Distribution of Revenues from Alcohol

Yes

436

Against "Pulverregal"

Yes

382

Construction of Railway through the Alps (T)

Yes

371

For Reorganizing Federal Finances

No

437

Financing Unemployment Insurance (W)

No

386

Salary Parliamentary Members

No

389

For Increasing Fuel Charges

Yes

446

New Corn Article (A)

Yes

387

Improve Infrastructure Members of Parliament

No

399

Resolution on Federal Finances

Yes

410

Promoting Cultural Activities

No

400

For healthy Federal Finances

Yes

416

For a new Health Insurance (H)

No

401

Pro healthy Social Insurance

Yes

384

Change Law on Stamp Duty

423

Securing Invalidity/Age Insurance (W)

No

405

For Fee on Road Use

Yes

430

Counter-Initiative: For Ecological Agriculture

Yes

406

For Fees on Heavy Traffic

Yes

431

Re-Organisation Administration

No

407

Introducing Fees for Heavy Traffic

Yes

346

Vote on Military Expenditures (Def)

444

Revision Age Insurance (W)

No

442

Law on Fees for Heavy Traffic

Yes

393

Against Fighter Planes (Def)

No

445

Infrastructure for Public Transportation (T)

Yes

465

For a Solar Energy Tax

No

421

Resolution on Slowing Expenditures

Yes

458

Law on Motherhood Insurance (W)

No

484

For a Capital Gains Tax

No

422

Change on Pension and Disability Laws (W)

Yes

469

For a flexible Age Insurance (W)

No

427

Cantonal Responsibility for Military Equipment (Def)

No

470

For flexible Retirement Age (W)

No

357

For Switzerland without Army (Def)

No

500

Equal Rights for the Disabled (W)

No

471

Saving at the Military (Def)

No

482

For Switzerland without Army (Def)

No

Increase Taxes

Decrease Subsidies/Grants

Decrease Taxes Yes

Decrease Expenditures No

Notes: The table lists all federal propositions between 1981 and 2003, which either led to an increase in federal expenditures, taxes, subsidies and grants, or a decrease in expenditures, taxes or public debt. The information is taken from publications of the federal government, which publishes the fiscal consequences for each new law and executive order. These are available online at http://www.ads.bar.admin.ch/ADS. Expenditures in individual policy areas are constructed from the respective votes. These are marked after the title using the following abbreviations: (A) for agriculture, (T) for public transport, (Env) for environment, (Def) for Defense, (W) for welfare, (H) for health and (Edu) for education.

Table A4: Propositions where Men and Women had accepted Different Outcomes

Title of Proposition

Year of Vote

Yes Women Yes Men

Decision

Ecological and Modern Agriculture Easier Access to Swiss Real Estate for Non-Residents Abolish Subsidies for Parking Spaces at Train Stations For a Sustainable Unemployment Insurance

1995 1995 1996 1997

0,44 0,43 0,52 0,39

0,50 0,55 0,41 0,52

No No Yes No

New Regulation Fuel Tariffs Introduction of Civil Service Reduce Property Sales, especially to Non-Residents Stop Construction of Nuclear Power Plants Stop Use of Nuclear Energy Reducing Animal Testing For an Ecological Military Against Fighter Planes Flexible Retirement Age 62 for Men and Women For Equal Rights of the Disabled Stop Construction of Nuclear Power Plants

1983 1984 1984 1984 1990 1992 1993 1993 2000 2003 2003

0,48 0,52 0,51 0,54 0,58 0,55 0,51 0,52 0,50 0,55 0,50

0,57 0,45 0,48 0,48 0,43 0,42 0,43 0,43 0,44 0,40 0,44

Yes No No No No No No No No No No

Notes : The second (third) column shows the percentage of women (men) voting in favor of the proposition. The last column shows the official outcome of the federal proposition. The first four rows show the votes where women changed the result. The other rows report the votes, in which men were decisive. Source : VOX Surveys, 1981-2003.