Undoing Gender with Institutions
Men spend more time in paid work, and women more time in housework. In spite of the rise in female participation in the labor market and the feminist struggle for greater symmetry, this gender-wise specialization within couples remains a quasi-universal norm. As a consequence, gender gaps in labor force participation and earnings have not subsided (Bertrand et al., 2015; Blau and Kahn, 2017) and the male breadwinner model remains prevalent.
What is the rationale for the stability of this pattern? Household economists have proposed various explanations based on the notion of comparative advantage (Becker, 1973, 1974), with or without bargaining between spouses (Chiappori, 1988, 1992; Weiss, 1997). These comparative advantages can be seen in turn as partly natural or as being dictated by the nature of economic activity and job characteristics at each stage of a society’s development (Alesina et al., 2013; Autor et al., 2003; Beaudry and Lewis, 2014; Black and Spitz-Oener, 2010). Institutions also certainly play a role in designing the architecture of choices for men and women, and providing incentives for more or less specialization (Esping-Andersen, 2009).
On top of these potential determinants, social scientists have pointed out the superim-position of norms (Akerlof and Kranton, 2000, 2013), postulating that people may attach some value to the roles they endorse per se, such as gender roles. An entire set of stylized facts that cannot be rationalized within standard economic models comes in support of this idea. One of the most striking of these observations is the reaction of households when the male breadwinner norm is violated. While any economic model of decision-making within the family would predict that a spouse should decrease her number of housework hours as her personal contribution to the household income increases, empirical evidence shows that things are not so simple. Beyond spouses’ income equality threshold, when a woman becomes the primary breadwinner, she starts doing gender (West and Zimmerman, 1987) by increasing her number of housework hours, and sometimes withdrawing from the labor force. Such marriages also become more unstable.1 It would be diﬃcult to make sense of such behaviors without recourse to the notion of gender identity. Existing studies have mostly contributed to establishing the existence of this gender-unequal norm, but little is known about its origin and the part played by public policy in constructing or deconstructing it.
This paper brings the first causal evidence that the male breadwinner norm is cultural and can be undone by institutions. Developing a causal test of the role of institutions is a major empirical challenge. Specifically, gender-equalizing institutions are much more likely to emerge in an environment where mentalities have already become more gender friendly. To overcome this empirical hurdle, we focus on Germany and exploit the natural experiment constituted by the 41-year division of the country. Before World War II, prior to the division, gender norms, including female labor force participation, were essentially similar in Eastern and Western regions. During the division, East Germany adopted gender-equalizing policies, in line with the universal « right » (and obligation) to work. Work-family balance programs, kindergarten and other childcare facilities were put in place (Bauernschuster and Rainer, 2012). In the meantime, a traditional family policy prevailed in West Germany. The institutions and policies implemented in the two regions radically diverged and so did gender roles. As a result, in 1989, women’s labor force participation in the German Democratic Republic (GDR) had reached about 89%, one of the highest in the world, against 56% in West Germany (Rosenfeld et al., 2004). After reunification, the government of the former Federal Republic of Germany (FRG) took over East Germany and rapidly dissolved its institutions and structures and absorbed them into those of West Germany, which remained unchanged.
Using data from the German Socio-Economic Panel from 1991 to 2012, we establish that, since reunification, the male breadwinner norm has been prevalent in West Germany but not in the East. First, we show that women who earn more than their husbands « com-pensate » by increasing their number of housework hours in West Germany. But this is not the case in East Germany, where women monotonically keep decreasing the time they spend on housework as their contribution to the household finances rises. Consistently, in West Germany, the risk of divorce increases for couples where the wife switches from earning less to earning more than her husband, whereas this is not the case in East Ger-many. Finally, we show that when a woman’s potential income is higher than that of her husband, she is more likely to withdraw from the labor market, but only in West Germany. Likewise, when a West German woman actually starts earning more than her husband, she is more likely to withdraw from the labor market in the following year. These behaviors are mirrored by self-reported preferences, as East German women attach almost as much value to paid work as men, in contrast with West German women.
To demonstrate the robustness of our results, we run systematic placebo exercises to establish that it is the focal point of equal incomes that triggers these reactions, and not any other alternative cut-oﬀ point. Similarly, we show that it is the former Berlin wall that constitutes the dividing line, and not any other division of Germany. We also provide substantive evidence showing that our results do not stem from pre-existing or current diﬀerences between Eastern and Western regions. First, we show that before the division, Eastern and Western regions had similar industrial, employment and social structures: we show this using first-hand statistical sources pertaining to the year 1933, as well as Prussian data from 1886 and 1849. We also illustrate the diverging trends in terms of female labor market participation. Second, we rule out the suspicion that East-West diﬀerences in household behavior could be due to other historical diﬀerences in unobservables or persisting structural diﬀerences, such as economic conditions. To do so, we focus on areas where it is likely that people face the same structural conditions. We show that among couples who currently live in the West, i.e. in the same environment, those who migrated from the East after 1990 display much more egalitarian behavior in terms of female labor market participation, wage earnings and housework time. In the same spirit, we use another survey (GoGold) to show that among women who currently live in East or West Berlin, those who were born in a former socialist country, or whose mother was born in a former socialist country, are much more likely to be working full time (ceteris paribus) than those who were not. We also focus on couples who live near the former East-West border and document the sharp spatial discontinuity of the female contribution to household income at the border and re-establish our main results on this subsample. Finally, we exploit the heterogeneity within West Germany to rule out the suspicion that our findings may be driven by wage structure diﬀerentials.
These results shed light on the sources of gender inequality. After decades of progress, female labor force participation has recently plateaued (Blau and Kahn, 2017) and re-searchers have started to investigate the determinants of social norms influencing gender equality. These include technological changes (Goldin and Katz, 2002), the influence of ancestors (Fernández et al., 2004; Fernandez and Fogli, 2009), information (Alessandra and Laura, 2011) and past technology (Alesina et al., 2013). We add to this literature by considering the role of institutions in sustaining the male breadwinner norm.
Our results are directly related to the literature on social norms and preferences, beyond gender issues. Experiments in behavioral economics have shown that, in dictator games for instance, the situation of equal earnings is a focal point that powerfully influences decisions (Rabin, 1993; Charness and Rabin, 2002). Cultural economics has also shed light on the evolution of social norms (Bisin and Verdier, 2001, 2011; Alesina and Giuliano, 2015; Fernández, 2013). Here, we illustrate the influence of institutions on social norms and identity.
This paper is in the line of Alesina and Fuchs-Schündeln (2007) who exploited the same episode of socialism in Germany to study the lasting (and progressively withering) eﬀect of socialist institutions on mentalities. Regarding gender attitudes, several studies, such as Breen and Cooke (2005), Bauernschuster and Rainer (2012) or Beblo and Gorges (2018) have illustrated the smaller gender gap in East Germany, as compared with West Germany, in terms of self-reported attachment to work. Campa and Serafinelli (2016) show that this appears to be a hallmark of socialist states. Besides work attitudes, Lippmann and Senik (2018) have also shown that the gender gap in mathematics is smaller in East Germany and in former socialist countries..
The rest of this work is organized as follows. Section 2 recalls the institutional back-ground of East and West Germany. Section 3 describes the data. Section 4 presents the empirical strategy. Section 5 studies the impact of higher female earnings on housework hours, the risk of divorce, and female labor market participation. Section 6 provides ro-bustness checks. Section 7 illustrates the diﬀerences in work attitudes among Western and Eastern couples. Section 8 concludes.
The German Division as a Natural Experiment
Before the Division
The division of Germany was drawn by a postwar agreement between the Allies, on the basis of the zones occupied by the Soviet Union and Western countries. In 1949, five Länder formed the GDR and the remaining eleven constituted the FRG. The line of division was thus arguably unrelated to pre-existing diﬀerences between the two regions.
To provide evidence on this matter, we collected data from German statistical yearbooks before the division. Table 1.1 describes the situation in 1933.2 Columns 1 and 2 provide descriptive statistics for the Eastern and Western regions of Germany. Column 3 computes the diﬀerences between the two regions. In 1933, the employment structure was similar in the two regions. About 45% of East Germans worked in industry against 40% in the West.
As regards gender behavior, the table shows that the female share of employment was 2.8 percentage points higher in the East, and the birth rate (per thousand) 1.95 points higher in the West.
How did the small East/West diﬀerences compare with the average regional diﬀerences? To investigate this question, we performed a permutation test that follows the logic of a Fisher exact test, and compared, for each measure, the East-West diﬀerence with the picture that would emerge from any random partition of Germany (excluding Berlin) into two groups of respectively 15 and 5 regions (column 5). It turns out that in 1933, the structural dissimilarities between the two regions that would later become East Germany and West Germany (excluding Berlin) were not any diﬀerent from what would stem from any random division of the 20 regions into two groups of 5 + 15 regions. This is attested by the p-value (column 5). This result supports the idea that the division of Germany was not influenced by pre-existing regional diﬀerences.
Diverging Trends during the Division
After the division, between 1949 and 1990, the GDR rapidly set up institutions in favor of gender equality. Beyond its constitution ensuring full equality between men and women, the Mother and Child Care and Women’s Rights Acts, adopted in 1950, aimed at ‘[establishing] a range of social services in support of full female employment, including a network of public childcare centers, kindergartens and facilities for free school meals’ (Cooke, 2007, p. 935), as well as paid maternity leave. By 1972, additional policies expanded childcare facilities and extended paid maternity leave to 18 weeks. A final set of reforms implemented between 1972 and 1989 improved childcare facilities, extended parental leave to 20 weeks and allowed fathers as well as grandmothers to take this leave (Cooke, 2007). In summary, these policies were targeted at making participation in the labor force compatible with maternity (see Goldstein and Kreyenfeld, 2011 about fertility trends in both regions).
In the meantime, the FRG’s policies strengthened the traditional family model. Ir-regular school schedules and scarce childcare facilities inhibited female employment. The tax system favored single earner families, as non-employed spouses and children could get public health insurance at no extra cost. Until 1977, the Marriage and Family law stated that: ‘The wife is responsible for running the household. She has the right to be employed as far as this is compatible with her marriage and family duties’ (Civil Code on the Eﬀects of Marriage in General, title Five, Section 1356).3 Subsequent policies then alternated more or less conservative incentives for female participation in the labor market.
As a result of these very diﬀerent policies, the female labor market participation rate started to diverge after the division. To illustrate the impact of the two diﬀerent sets of policies during the division, we collected data from statistical yearbooks from 1959 to 1987. Figure 1.1 displays the diverging trends of women’s share in total employment. In the FRG, the share of employed women, as a percentage of the total female population remains steadily around 30%, whereas in the GDR, it rises from approximately the same level to 50% between 1959 and 1987 (the years for which these statistics are available).4
Persisting Diﬀerences after Reunification
After reunification, the government of the former Federal Republic of Germany (FRG) took over East Germany and rapidly dissolved its institutions and structures, absorbing them into those of West Germany, which remained unchanged. Yet, persisting diﬀerences between the two regions were still observable ten years later in 2000. Labor force partici-pation was still approximately the same across gender in the former GDR (around 80%), whereas the gap remained wide in West Germany, with 65% of women in the labor force against 81% of men (Schenk, 2003). As we will show, these objective diﬀerences are sup-ported by opinions regarding gender roles. In terms of paid work time, in 2000, East German workers generally worked longer hours than West Germans: 35 hours for women and 42 hours for men in the former GDR against respectively 29 and 40 hours in the for-mer FRG. This is probably a legacy of the diﬀerent labor laws that prevailed during the division.5 The status of part-time employment also diﬀered considerably across regions. In West Germany, part-time workers, most of whom were women, often worked less than 20 hours, and were not eligible for the same social benefits as full-time workers (Rosenfeld et al., 2004). In East Germany, part-time workers had longer hours, received identical social benefits and used these contracts primarily as a transition to retirement.
This does not mean that there are no gender diﬀerences at all in East Germany. For instance, Rosenfeld et al. (2004), document the existence of gender wage gaps and occupational segregation. Additionally, within the household, although men participate more in housework in the East than in the West, Eastern women still take on a greater share of housework (Cooke, 2007).
We use the German Socio-Economic Panel, a longitudinal survey run by the German Institute for Economic Research (DIW, Berlin).6 This survey was started in 1984 in West Germany and was extended to East Germany in 1990. In 1998, 2000, 2002, 2006, 2009, 2011 and 2012, additional German households were added to the initial sample. We use 22 waves, from 1991 to 2012.
Main Explanatory Variables
East versus West. Our exercise consists in contrasting the behavior of East versus West Germans. We exploit the biographical information contained in the dataset. The questionnaire asked all individuals: « Where did you live before 1989? ». We define an East dummy variable that takes the value 1 (0) if both spouses lived in GDR (FRG) before 1989, independently of where they live at the time of the survey.7
Relative Income. Our main explanatory variable is a dummy that equals 1 if the wife earns more than her husband and 0 otherwise (hereafter: WifeEarnsMore), where income measures monthly labor earnings.
Housework. The time spent on housework is measured using the following question: « What is a typical weekday like for you? How many hours per normal workday do you spend on housework (washing, cooking, cleaning)? « . The definition of housework, i.e. the list of tasks included in the survey, follows the general usage in the literature. In particular, it does not include the time parents spend with children. The norm concerning childcare has changed since the 1970s: a new norm of intensive parenting has diﬀused, whereby the time spent with children is more and more considered as leisure. This is particularly true of more educated parents and wealthier families (Sullivan, 2010). For this reason, we leave this aspect of couples’ time-use aside from our main housework measure.
Divorce. In Section 5.2, we look at the impact of female relative income on the risk of divorce. We consider the sample of married working couples, aged 18 to 65 years old, and estimate the likelihood of divorce within the coming years, according to their relative income. We use the marital status reported by both spouses at each wave, as well as the biographic data file. As divorce takes time, our main variable of interest is the risk of divorce in a 5-year horizon.
Labor Market Participation. In Section 5.3, we look at the impact of female relative income on her participation in the labor market. We estimate the likelihood of withdrawing from the labor market in a one-year horizon (T+1). Consequently, the variable of interest is a dummy that codes 1 if the individual has no labor earnings in T+1 (out of the labor market) and 0 otherwise.
Attitudinal Variables. We use subjective attitudes elicited in the SOEP survey, namely: How important is success at work for satisfaction? How important is marriage for satisfaction? How important is work for satisfaction? How important is a successful career for satisfaction? How important is family for your satisfaction ? The first two questions were asked in 1992, 1995, 2004, 2008 and 2012. The remaining three questions were asked in 1991, 1994, 1998 and 1999. Given the distribution of preferences (see Figure 1.F1 to Figure 1.F5 in the Appendix), we define dummy variables that equal 1 if the respondent has declared the matter to be very important and 0 otherwise.
Table of contents :
Summary / Résumé
Chapter 1: Undoing Gender with Institutions. Lessons from the German Division and Reunification
Chapter 2: Gender and Lawmaking in Times of Quotas. Evidence from the French Parliament
Chapter 3: Persistence of Incumbents and Female Access to Political Positions