Maternity Leave in Turbulent Times: Effects on Labor Market Transitions and Fertility in Russia, 1985-2000

Sociologists and demographers recognize an inherent conflict between women's employment and fertility: primary responsibility for childcare falls on women, so mothers face difficulties combining work outside the home and early childrearing (Brewster and Rindfuss 2000; Rindfuss et al. 2003). To participate in the labor force, women must curtail childrearing (and hence childbearing) or find alternative means of childcare. Maternity leave potentially reduces this conflict, at least temporarily. Governments have enacted maternity leave policies to maintain women's participation in the labor force after childbirth, promote gender equality and encourage couples to have more children. However, the evidence that parental leave accomplishes these objectives is mixed (Ruhm 1998; Gauthier 2007), and it mainly consists of macro-level analyses that cannot control for factors that may jointly influence maternity leave and other outcomes (Neyer and Andersson 2008).

We evaluate the effects of maternity leave on employment outcomes and ­childbearing behavior in Russia using individual-level data from the Survey on Stratification and Migration Dynamics in Russia. The SSMDR includes retrospective employment histories with information on maternity leaves taken from 1985 through 2000, which we use to estimate discrete-time hazard models that show the relationships between maternity leave and women's labor market transitions (entry or exit into the labor market and job mobility), as well as second birth rates. Russia holds particular interest for studying maternity leave due to its history of socialist family policies, the rapid institutional transformation associated with the Soviet collapse, the ensuing economic and political turbulence that may have undermined maternity leave, and a cultural tradition of low father involvement in childrearing. We find that maternity leaves up to 36 months consistently helped Russian mothers of young ­children maintain a foothold in the labor market, even during the considerable economic and political turbulence that followed the collapse of the Soviet Union (1992-2000). Moreover, women who took leaves of 6-36 months with their first birth conceived second children at higher rates after returning to the labor force. Thus, our results provide evidence that maternity leave maintains women's participation in the labor force and contributes to higher fertility rates.

Our analyses address, to the extent our data permit, a major methodological problem in studying the effects of maternity leave: the potential endogeneity of the decision to take maternity leave. Mothers predisposed to having additional children may be more likely to take longer maternity leaves with their first birth, upwardly biasing estimates of the effects of maternity leave on second birth rates. We correct for this type of endogeneity by estimating the propensity to take long maternity leave and including the propensity as a control in our models for second births. We also perform sensitivity analyses to assess the robustness of our findings. Endogeneity may also bias our estimates of the effects of maternity leave on employment outcomes: for example, women with greater job security may take longer leaves and have lower rates of job loss. Women may choose “family-friendly occupations” that are more favorable to staying on extended maternity leave. To address these possibilities we include extensive covariate controls for occupational, firm and regional labor market characteristics.

Maternity Leave and Employment

All industrialized countries have some form of maternity leave policy to help alleviate the conflict between childbearing and employment (Gornick et al. 1998), yet the impact of these policies on the work outcomes of mothers is uncertain. Some studies link the availability of maternity leave to positive outcomes, such as a greater probability of job retention after childbirth (Waldfogel, Higuchi and Abe 1999), a greater likelihood of returning to a previous employer, or a wage premium that offsets the wage penalty of motherhood (Waldfogel 1998), and (at the country level) less chance of reduced employment among mothers of infants and preschoolers (Gornick et al. 1998)

However, maternity leave policies may have a negative effect on women's labor force participation and gender equality. Although women may be guaranteed employment in the same firm when they return from leave, time out of the labor force may result in lower job retention and poorer wages in the long term (Gangl and Ziefle 2009). While short periods of paid leave probably have little impact on wages, longer periods could reduce wages (Ruhm 1998).

In Soviet Russia, the state encouraged women to work and sought to keep them in the labor force by expanding access to quality daycare and allowing mothers to work shorter days with more flexible hours. Paid maternity leave, first introduced in 1981, was another measure designed to help women balance work and childrearing. All women had the right to take maternity leave regardless of occupational sector or type of job. Women on maternity leave received full pay up to 112 days (16 weeks), partial-pay up to 12 months and unpaid leave from 12 to 18 months. Women were not required to return to work in order to receive maternity leave after a second birth. In 1990 the system was changed: the guaranteed period of full pay was extended to 126 days (i.e., 18 weeks), partial pay to 18 months and unpaid leave to 36 months. Also, the amount of pay during the “partial pay period” was linked to the minimum wage. In 2009 the “full pay” period was extended to 140 days, but this is well after the period covered by our data.

We distinguish conceptually and empirically three durations of leave. Short leave refers to the period during which women receive full pay while on leave. This period (and the others) changed in 1990, but for simplicity we adopt the post-1990 periodization and define leaves of 0 to 5 months as short.¹ For working women, short leaves should be fairly universal because they involve no immediate cost in lost pay. Women who take an extended leave (6-36 months) receive reduced pay or no pay during this period but legally should face no other penalty (like job loss). Women had no legal entitlement to leave beyond 36 months: such excess leaves probably reflect unusual circumstances (such as the need to care for a sick child), specific arrangements with employers or perhaps measurement errors. Excess leaves should be unstable, because employers are not legally bound to honor them. Therefore, while we expect short and extended leaves to protect women's attachment to work, women on excess leave probably have higher rates of all exits from leave, including returning to work (at the previous job or a new job) and exiting work (by quitting or being dismissed).

Despite very high female labor force participation rates in Soviet Russia (Gerber and Mayorova 2006), gender equality in the labor market or home never materialized, as reflected in the exclusion of women from certain occupations, gender wage gaps comparable to those in other countries, and high levels of occupational sex segregation (Ogloblin 1999; Brainerd 2000). Women were expected to work outside the home and also do the majority of childrearing and domestic chores. In short, conservative gender roles and norms persisted.

The crisis-like conditions of the Russian economy for much of the 1990s may have undermined the role of maternity leave in maintaining women's attachment to the labor force for several reasons. First, the collapse of the state socialist economy and the introduction of market reforms at the outset of 1992 disrupted the labor market, yielding rampant inflation, sharp declines in real wages, structural dislocations, growing unemployment and widespread wage arrears (Gerber and Hout 1998; Gerber 2002, 2006). These developments placed pressure on household budgets, making it more costly for women to forego the difference between their wages and their maternity benefits, and women may have lacked the financial resources to take leave longer than four months. Second, some have argued that discrimination against women increased during this period (CEDAW 1999). State authorities no longer explicitly encouraged women to participate in the labor force, and some policy makers called for women to return to the home to perform their “natural” duties (Teplova 2007). Market forces eroded Soviet institutions and policies that suppressed gender discrimination, potentially giving free rein to latent patriarchal attitudes among male employers (Linz 1996; Ogloblin 1999; Khotkina 2001). As social institutions and government enforcement agencies faltered, employers may have been able to ignore mandates to protect gender equality. Whether driven by new competitive pressures or by latent sexism, employers in the post-Soviet context may have targeted new mothers for layoff and denied maternity leave, even in violation of Russian law. Technically, although employers pay maternity leave benefits directly, they are fully compensated for such payments by the state in both the Soviet and post-Soviet systems: in the former case via an adjustment to the enterprise's wage fund, in the latter via an offsetting payroll tax credit. However, post-Soviet employers, particularly private firms, have more incentive to conceal wage payments to evade payroll taxes, and doing so limits their government compensation for maternity leave payments. Even though maternity leave should have been available to everyone, access to maternity leave may have varied by type of employing firm, occupational characteristics or region, as these variables are related to both employer and employee considerations that, in turn, affect the costs of taking or denying a leave.

Nevertheless, maternity leave could play its designated role of protecting women's positions in the labor market despite the turbulent economic upheavals. The erosion of the strong social protections of the Soviet state after the collapse may be exaggerated, and Russian women probably resemble Ukrainian women (Perelli-Harris 2006) in that they still expect the government to provide support for families and childbearing, especially paid maternity leave. Thus, although the real value of maternity leave payments declined in post-Soviet Russia, we hypothesize that it still helped women keep a foothold in the labor market.

Maternity Leave and Fertility

Maternity leave provides time to care for an infant when the infant is establishing secure attachments and developing basic physiological and behavioral regulation (Clark et al. 1997). It allows mothers to avoid, at least temporarily, the role incompatibility and associated stress that come with attempting to raise young children and work simultaneously (Rindfuss et al. 2003). By reducing the opportunity costs of having a child, maternity leave can make it easier for women to achieve their desired number of children. Accordingly, some governments have implemented parental leave policies in order to encourage working women to have more children.

However, evidence on the relationship between maternity leave and fertility is mixed (Gauthier 2007). Sweden's “speed premium,” which expanded the length of time parents could retain parental leave benefits to two years, increased second birth rates (Hoem 1993). In 1990 Austrians hastened the tempo of third births after the government extended the parental leave period, helping reverse a decades-long decline in third birth rates; however, the changes in parental leave did not affect total fertility (Hoem, Prskawetz and Neyer 2001). Father's uptake of parental leave in Sweden is related to higher second- and third-birth propensities, but not when periods of leave are extended (Duvander and Andersson 2006).

Soviet officials hoped their 1981 maternity leave policies would reverse falling birth rates. Russia's total fertility rate grew from 1.89 in 1980 to 2.22 in 1987, but only due to faster progression to second births, not increased cohort fertility (Zakharov 2006). The TFR then fell from 1.89 in 1991 to 1.17 in 1999, mainly because of postponement or elimination of second births (Zakharov 2008). Thus, we analyze the effects of maternity leave on second births, the key parity progression for Russia's fertility decline.

We hypothesize that working first-time mothers who take extended leave from their jobs (6-36 months) have higher second birth rates when they return to work because on average they have a better experience following their first births. They have more time to bond with and look after their infants without losing their jobs. They do not need to leave their children in state institutions, which often have high child-to-caregiver ratios and are of poor quality, or with grandparents, who may be struggling with their own issues; they also need not take time off from work when their children are sick, which occurs frequently among children in childcare centers. The opportunity to care for their small children may reduce feelings of guilt at leaving the child with others and alleviate the daily stress of working while maintaining a household. In contrast, women who return to work quickly experience, on average, more difficulties and stress, making the prospect of combining work with a second child unattractive. In sum, extended maternity leave with a first birth gives many women a more positive experience than a fast return to work, increasing the appeal of a second birth.

Data and Methods

We analyze the relationships of maternity leave with employment and fertility outcomes from 1985-2000 using the Survey on Stratification and Migration Dynamics in Russia. The SSMDR was given to a multistage, stratified probability sample of 7176 Russian adults in three waves from September 2001-January 2002. For information about pre-testing, sampling, fieldwork and quality control see Gerber (2006b). The instrument included special batteries of questions that elicited the respondent's fertility, marital, work and residential histories from December 1984 through the month of the survey. These histories allow us to estimate hazard models, relating births and labor market transitions to maternity leave, with time-varying covariates.

To test whether maternity leave protects the workforce attachment and job mobility of women who have children, we estimate discrete time hazard models for four outcomes: entry to work (for those currently unemployed, out of the labor force or on maternity leave), and job loss, voluntary employment exit (quits) and employer change (for those currently working or on maternity leave). Throughout, we only consider maternity leave from a job, because we want to know whether and how maternity leave from paid employment helps women combine work and childbearing. Respondents are at risk for all four outcomes during months when they have the appropriate employment status, including maternity leave from a job. In contrast to the risk set for our fertility models, which is limited to mothers with one child because we wish to model second conceptions, the risk sets for our models of the four labor market events include all women who are at risk for the particular outcome (regardless of how many children they have). However, we also estimate the same models on more limited risk sets consisting of women with at least one child under 3 in the home (again, regardless of how many children overall the woman has). We estimate the effects of being on leave relative to being out of the labor force (for job entry) or to hired employment (for job loss, voluntary quit and employer change). In the employment models, we distinguish between being on leave for 0-36 months and more than 36 months, reflecting our expectations that legally protected maternity leave protects attachment to work, while excess leave is untenable in the long term and is thus associated with both re-entry to work and exit from work altogether.

We treat either a return to the same job or the start of a new job while on maternity leave as employment entry, while we also treat the latter event as an example of employer change. It may seem contradictory to treat women on maternity leave from a job as simultaneously at risk for employment entry (implying they are not currently employed) and for job/employment exit (implying they are). However, maternity leave shares some characteristics with both non-employment (women on leave from a job are not actively working) and employment (they formally hold a job, have an employer, etc.) Our flexible treatment is consistent with the ambiguous character of maternity leave itself as an employment status, and in all our analyses we make reasonable comparisons of maternity leave to other statuses. The observation window for these analyses is 1988-2000, because the survey did not identify the number of children under 3 at home for 1985-1987 or pregnancies that began after December 2000.

We next analyze how maternity leave influences subsequent fertility using discrete-time hazard models of second conception rates for women who remain in the workforce after first births. The dependent variable is the log-odds in a particular month of a conception resulting in a second birth. The data only identify births, not conceptions, so we assign conceptions to eight months before the birth, when the decision to keep a pregnancy is typically made. We cannot observe conceptions resulting in a miscarriage or abortion. Because we cannot tell if respondents are pregnant at the time of the survey (September 2001-January 2002), we truncate our observation window in December 2000 (to ensure that we measure all conceptions leading to live births in our window). Respondents enter the risk set the month following their first birth, as long as it occurs after the start of the observation window (January 1985), and are censored when they conceive their second child, when they turn 45 or in December 2000.² Because we are interested in whether maternity leave is associated with higher rates of second conceptions among working women, we also censor person-months when respondents are out of the labor force (neither working, nor unemployed and looking for work, nor on maternity leave from work). Women who leave the labor force (e.g., to study or keep house fulltime) re-enter the risk set if and when they re-enter the labor force. To test whether restricting the risk set to women in the workforce introduces selection bias, we re-estimate our model after restoring person-months when first-time mothers are not in the labor force. We try additional specifications in a series of sensitivity analyses.

Key Variables

Main Activity

Respondents indicated the month and year when they started a “leave due to pregnancy or to look after a child,” and also when they returned from maternity leave to their prior job or to a new job (the latter two alternatives were distinguished, allowing us to identify job changes by women while on maternity leave). They also provided the month and year of any job losses, voluntary quits, employer changes, new jobs and other changes of workforce status (starting to look for work, entering full-time school, leaving the work force due to retirement or disability, etc.) We use this information to identify the respondent's main activity each month during the observation period. We code respondents who are not working as “not in the labor force” when they are not looking for work and as “unemployed” when they are actively looking. Other options include: working for hire, self-employed, studying in school, serving in the military, retired/disabled or other activity.

Maternity Leave

We measure the duration of leave based on the months elapsed since the respondent began her current leave: when a woman has a child while on leave from a previous birth, her leave duration is re-set to zero. In the employment models, we use two dummy variables to indicate being on maternity leave in a particular month: one for months 1 through 36 of leave, the other for all months beyond 36.³ As noted, the effects of “excess” leave (more than 36 months) are unpredictable because they are not guaranteed by law and are inherently unstable. In the fertility models we distinguish the effects of current and prior maternity leave by incorporating separate time-varying variables for each, and we distinguish between women on short, extended and excess leaves. While a woman is on extended maternity leave (6-36 months), the reduced income might discourage her from having an additional child; alternatively, the ability to retain her job may encourage her to become pregnant again. After they return to work from an extended leave, we expect women to have higher second birth rates because their extended leave allowed them the chance to devote time to early child-rearing free from anxiety about losing their jobs. Thus, in these models the key distinction should be between leaves of extended and short duration.⁴ However, we also enter separate dummy variables for excess leave in light of our concern that their effects differ due to their unusual character.

Structural Variables and Unobserved Heterogeneity

We attempt to control for unobserved factors that may jointly affect the duration of maternity leave and the employment outcomes of interest by incorporating a range of variables characterizing the woman's job and locality at the time of a first birth. These variables could represent the availability of maternity leave (if employers were illegally discouraging women from taking extended leave), or women's decisions to enter into “family friendly” occupations. In the fertility models, we are especially concerned that the length of prior maternity leave may be endogenous to intentions to have a second child. To control as best we can for this form of unobserved heterogeneity, we estimate each respondent's propensity (using probit regression) to take a leave of at least six months in connection with a first birth based on both individual and structural characteristics at the time of the first birth. Then we enter this estimated propensity as a covariate in the model for second conceptions. We cannot use the propensity to obtain the standard “matching” estimates of the effect of prior maternity leave due to the person-month structure of the data, but incorporating the estimated propensity as a person-level, time-invariant covariate is a viable alternative (Winship and Morgan 1999).

Three measured aspects of the employing firm may be relevant: sector, branch and size. We expect private employers to discourage employees from taking extended leave because they face tighter budget constraints and are focused on profit motives that make them less willing to hire and train replacements. Firms in different branches of the economy may differ by work process flexibility, the costs of hiring and training replacements, and the degree of competitiveness. Based on preliminary analysis we collapsed the 13 branches into three categories: (1. health and social services, government and public services and communications; (2. finance and insurance; and (3. other (including manufacturing and extraction, agriculture and forestry, construction, transportation, wholesale/retail trade and catering, housing and communal services, science and education, arts and entertainment, and other). We expect the effects of firm size to be non-linear: small firms find it harder to replace workers on leave, but the largest firms are most affected by the difficult economic conditions and more prone to lay off workers. We thus included dummy variables denoting firms with more than 1,000 and less than 51 employees.

We employ three measures of occupational characteristics. The occupation associated with each job held by the respondent was initially coded using the 1988 ISCO classification. We used scales developed in other studies (see Gerber and Mayorova 2006) to code occupations by their percentage of female incumbents and their average logged earnings. Following human capital reasoning about gender differences in occupational choice (Polachek 1981), we expect more feminized occupations to be associated with higher rates of extended maternity leave and better-paying occupations to be associated with lower rates of extended maternity leave. We also incorporate a dummy variable denoting managerial and proprietor occupations, as incumbents in these jobs should experience more pressure to return to work and resume their duties.

We test for systematic variation by regional characteristics using the logged population size of the respondent's locality at the time she gives birth, a dummy variable for residence in Moscow, and five variables describing the province (“oblast”) of residence: average logged wages (adjusted for inflation), the unemployment rate, the percentages of the workforce in small businesses and in the private sector, and the percentage of ethnic Russians according to the 1989 census. We expect more extended maternity leaves in regions that are performing better economically, as workers in such regions might be less fearful of dismissal in retaliation for taking an extended leave. At the same time, the average wage penalty for taking an extended leave would be higher in higher-wage regions, which might create a disincentive for extended leave. Small business and private sector employment capture the degree of penetration of the regional economy by market forces. Following Gerber and Berman (2010), we treat the percentage of ethnic Russians as a proxy for relatively stronger cultural orientation toward Western norms regarding work and family, because many non-Russians are Muslims. As a Western feminist institution that assumes women will return to work, extended maternity leave might be more culturally accepted in regions with fewer Muslims.

Period Measures

The sweeping changes in Russian society during our study period require controls for period effects. In brief, initial signs of economic trouble and political instability emerged in the late 1980s, when Russia's fertility decline began. Full blown crisis conditions erupted with the Soviet collapse at the end of 1991 and continued until a period of sustained growth started in 1999. Our models specify period using dummy variables for 1985-1988, 1992-4, 1995-8 and 1999-2000, representing the early pre-transition and early, middle, and later transition periods, which are compared to the late pre-transition baseline of 1989-1991. To test whether the effects of maternity leave diminished after the Soviet collapse for the reasons outlined, we interacted the maternity leave measures with a dummy variable for the post-Soviet period (January 1992 onward).⁵

Other Controls

The employment outcome models include controls for current pregnancy and number of children under 3 and 3+. All models include controls for marital status, age (specified using the optimal polynomial specification determined empirically), and education (dummy variables for university, specialized secondary, and less than secondary, contrasted with secondary education, including vocational secondary). The fertility models also incorporate months elapsed since first birth, optimally specified as a third-order polynomial (months, months squared and months cubed).

Results

Descriptive Statistics

Table 1 shows the duration of maternity leaves taken by female SSMDR respondents in connection with first births reported during the observation window. The modal duration is an extended leave, taken about half the time, followed by short leaves, taken 37 percent of the time. About 14 percent took an “excess” leave of more than three years. Given the difficult economic conditions during this period, the women who took only a short leave may have been unable to forego any earnings. Also, they may have had access to alternative forms of childcare, such as subsidized daycare and help from grandmothers.

The SSMDR data reveal no clear trend in the duration of maternity leave taken with first births in Russia, suggesting that the collapse of the Soviet system had little impact on such decisions.⁶ In addition, leave duration is not significantly associated with standard demographic variables such as age at birth, education and marital status. As we expected women working for private sector firms are less likely to take a leave longer than five months.

For descriptive statistics on the control variables in our models, see Appendix Table A1. The trends in these variables over time reflect known trends in Russia during the period in question, such as rising unemployment and private sector employment, stable rates of female educational attainment and declining marriage and fertility.

Effects of Maternity Leave on Labor Market Transitions

We examine four labor market outcomes: job loss, job entry, voluntary quit and job mobility (employer change). To illustrate our modeling procedure, we show three detailed model specifications for job loss (Table 2). Model 1 controls for individual demographics and household composition and treats the effects of maternity leave durations as constant across periods. Model 2 enters interaction terms testing for changing effects after the Soviet collapse. These models show that leaves up to 36 months lower the hazard of job loss (relative to the baseline category of active employment), but excess leave (more than 36 months) elevates the hazard. Neither effect changes significantly after the Soviet collapse. Model 3 adds controls for the structural variables that may affect both the odds of taking an extended leave and also the risk of job loss, in order to correct for potential bias in the estimate of the effect of maternity leave on the outcome. The estimates of the effects of maternity leave barely change and remain statistically significant. Generally, the controls behave as in prior research (Gerber 2002). University education reduces the hazard of job loss, as does an employer in health, government or communications (this aggregation of branches is associated with lower rates of taking leaves longer than five months). Private sector employment and higher regional unemployment increase the hazard.

Table 3 shows significant effects of the two maternity leave durations on all four labor market outcomes. We report separate sets of estimates for the entire risk sets of 15- to 44-year-old women and for restricted risk sets of women with children under 3 in the household (no matter how many children they have overall). The former offer more statistical power, the latter clearer interpretation, because they pertain only to women likely to be on maternity leave (since they have children under 3). For each outcome, maternity leave improves women's labor force attachment. The same pattern of results holds whether we analyze the sample of person-months for all women ages 15-44 or women with children under 3. Moreover, the effects of maternity leave either remain unchanged during the post-Soviet era or only emerge at all in that period.

Compared to women who are not in the labor force – i.e., those with no job who are not looking for work – women on extended maternity leave with otherwise identical characteristics were significantly more likely to enter work after the Soviet collapse. The effect is especially strong for those with a child under 3 at home: the hazard ratio is 1.78. Apparently, maternity leave had little impact on women's return to employment during the Soviet era, when jobs were generally plentiful, but a positive effect in the post-Soviet era, when jobs are scarce.

Our results show that maternity leaves up to three years also offer protection against being laid off, particularly for mothers of young children. The persistence of the pattern in the post-Soviet era implies that despite concerns about waning enforcement, the protective function of legally-guaranteed maternity leave remained in force. After the Soviet collapse, when layoffs became far more common due to the sharp economic contraction accompanying market reforms, extended maternity leave dramatically reduced exposure to layoffs: while on maternity leave, women's hazard of layoff was only 17 percent that of otherwise similar women who were actively working (11% among women with children under 3 at home).⁷

Our results also indicate that women on leave up to 36 months were about half as likely to quit their jobs as women who were actively working and had the same characteristics on all the control variables, both during and after the Soviet period, further evidence that maternity leave has helped Russian women maintain their place in the workforce. Finally, women on leave up to 36 months had elevated rates of job mobility (employer change) compared to women currently working with similar characteristics. This finding is somewhat counterintuitive, because women who are on maternity leave would not seem attractive hires for prospective employers. Women on maternity leave might experience more pressure to change jobs from employers who want them to leave, or perhaps they are replaced at their original firms, offered less attractive jobs when they return, and decide to seek a new job instead. It is also plausible that some women on maternity leave – particularly those with other caretakers in the household – have more time to search for new job opportunities compared to women with young children who are actively working. If a woman is unhappy with the conditions of her current job, maternity leave might afford her the chance to seek a better situation elsewhere.

Excess leave (over 36 months) is associated with substantially higher rates of all four outcomes: job entry, loss, and quit and employer change. Because these outcomes are inconsistent in terms of what they imply for labor force attachment, the most reasonable conclusion is that excess leave is an inherently unstable state: women on leave from their jobs for more than the legally guaranteed period of three years are likely to exit that state, either to active work or to non-employment, and so they have higher rates of all four types of transitions than the corresponding baseline categories. It would, therefore, be unwise to attach much substantive importance to the strong “effects” of being on excess leave. Also, it is well worth distinguishing the effects of leaves on labor market outcomes by their durations to avoid either inflating or deflating the estimates of the effects of extended leave by combining it with excess leave.

Our evidence that maternity leave up to 36 months helps Russian women stay in the labor force is clear and consistent. Contrary to the pessimistic scenarios raised by concerned observers who feared that maternity leave policies had become ineffective after the Soviet collapse, the labor market benefits of leave remained stable or (in the case of job entry) increased during the post-Soviet era. The strong “effects” of excess leave on both positive and negative employment outcomes probably reflect its inherently unstable nature. Employers are under no obligation to provide it, and it brings no income to the household, so women exit excess leave at high rates, whether by returning to work at the same or a different job, quitting or being dismissed.

Employers may punish women for maternity leaves after the fact, when they have greater legal leeway to do so. To test this possibility, we incorporated a dummy variable indicating prior maternity leave in our preferred models for job loss and voluntary quit, in this case estimated for currently working women with one child. In neither case did prior leave have a significant effect (results not shown). Thus, our findings are consistent that maternity leave helps Russian mothers retain a foothold in the labor force.

Effects of Maternity Leave on Second Births

To get an initial sense of whether the SSMDR data confirm our hypothesis that prior extended maternity leave increases second birth rates, we plotted smoothed hazards of second conceptions for those who had and had not taken a prior extended leave using standard life-table methods (Figure 1). This plot pertains to all women who were working at the time of their first birth, and the x-axis is simply the months elapsed since first birth. The smoothing procedure weights the local estimates in a moving window around each month using an Epanechnikov kernel function, which is why the series start at 12 months even though some conceptions occurred earlier.⁸ Consistent with our hypothesis, after extended leave women have much higher hazards of second conception during the first two years after a first birth, and this gap generally persists for at least a decade, aside from a brief period at around seven years. The pattern is also evidence in Figure 2, which shows the cumulative hazards of second conception for women who did and did not take extended leave. These plots indicate that the elevated conception rates of women who took extended leave are not just artifacts of their “catching up” with women who did not take extended leave, as they persist for years. They also confirm that Russian women have relatively long birth intervals, with an average spacing of about five years in the mid-1990s (Zakharov 2008).

Our hazard models confirm that prior extended maternity leave elevates second conception rates (Table 4).⁹ Model 1, estimated on person-months when women had returned to the workforce after a first birth or remain on maternity leave, shows no statistically significant effect of current maternity leave (relative to active work). But after returning to work women who took extended maternity leave have second conception rates 1.58 times (e^.457) higher than those who did not take extended leave. The effects of control variables conform to expectations, aside from the non-significant contrast between unemployed women and working women. Women with university educations have lower odds of a second conception than women with a general secondary degree or less. Married and cohabiting women have higher rates of second conceptions than single mothers. The effect of age is curvilinear, and the third order specification of time since first birth fits the data. The odds of a second conception initially increase, but at a markedly slower rate than observed in countries where the modal spacing between first and second births is approximately two years. Second conception rates are lower in big cities, but net of city size they are higher in Moscow. The period effects capture the dramatic decline in second birth rates during these years.

To correct for the possible effects of unobserved factors that jointly influence maternity leave and fertility decisions, we estimate person-level propensities for taking a leave of greater than six months, then include this propensity as a covariate.¹⁰ We calculate each respondent's propensity based on a regression where the dependent variable is the probit transformation of the probability of taking a leave of six months or more (Table 5). Our probit specification satisfies the balancing property necessary for matching on the basis of propensities.

Age, education and marital status are not significantly related to the probability of taking extended leave, but a number of structural variables have significant effects. As predicted, women working for private sector firms are less likely to take long leaves. This effect could stem from greater fears among young mothers working at private firms about losing their jobs, better compliance with labor and tax laws and, on average, fewer competitive pressures to reduce labor costs on the part of state firms, or higher private sector wages and associated opportunity costs. In Russia, opportunity costs may not reduce first births, which are nearly universal, but instead affect second births by limiting time spent on maternity leave. We found no significant effect of firm size, but we did identify branches where, net of the other effects, extended maternity leaves are less common: health and social services, communications, public safety and administration, and finance and insurance. The latter branch is known for its especially high wage rates, again raising the opportunity costs of long maternity leave. We cannot explain why women who work in the other branches (which we aggregate because their coefficients are statistically indistinguishable) would be less likely to take extended leaves.

Women in the least feminized occupations are less likely to take a leave of more than five months in connection with a first birth. This is consistent with straightforward human capital reasoning: women more oriented toward childrearing than career achievement may avoid occupations that do not offer flexible maternity leave arrangements. Thus, the most masculine occupations serve as a proxy for an unobservable orientation toward career over child-rearing rather than a causal factor encouraging maternity leave. In contrast, better paying occupations are also associated with elevated odds of taking longer leave, which contradicts human capital logic. Perhaps women in higher-paying occupations are better off economically, both due to their relatively high-paying jobs and to assortative mating, and can more easily sacrifice some of their wages while on extended leave. Or, the maternity leave rights of salaried employees could be better protected than those of wage employees. At the same time, net of these other effects women who work as managers or proprietors are more likely to return to work in five months or less.

Net of the other variables in the model, short maternity leave is significantly less common in Moscow than in other locations. We speculate that this reflects more reliable enforcement of maternity leave compensation policies there. We find no evidence for effects of regional unemployment, average wage or percent of the labor force in the private sector. Small business employment (a proxy for regional penetration of market forces) and the percentage of ethnic Russians in the region do have the effects we anticipated. Finally, net of the structural controls, longer leaves increased in 1989-1991 (when the full pay period was extended to five months) and then spiked in 1999-2001 (when Russia's economy began growing again).

The effect of prior extended leave remains statistically significant when we enter the estimated propensity of taking a leave longer than five months in the model for second conceptions (Table 4, model 2). This strengthens our confidence that extended leave is at least partly responsible for elevating the hazard of a second conception after first-time mothers return to work: to the extent that we can control for the unobserved preference for taking an extended leave, our key result is robust. Of course, this claim rests on the assumption that net of propensity to take longer leave, selection into extended leave is random.

Our results could also be biased by the limitation of the risk set to mothers who had returned to work. Accordingly, we estimated our preferred model after including person-months drawn from women with one child when they were not in the labor force or in school fulltime. This expansion of the sample did not affect our key finding regarding the elevated rate of second conception hazards (Table 4, model 3). Our key finding also proved robust in additional sensitivity analyses (results available upon request.) To ensure that our results were not driven by the tendency of self-employed women to avoid both extended leave and second conceptions, we omitted person-months when respondents were self-employed. We also omitted unemployed women and women currently on maternity leave, alone and in combination with self-employment spells, in case their distinctive patterns drive the effect of prior extended leave. Because conception rates are typically low in the months immediately after a birth, which are too soon for a woman to be on extended leave, we tried omitting spells within five months of first birth. We tried additional approaches to control for unobserved heterogeneity. Rather than estimated propensities for longer leaves, we incorporated the structural variables from our probit model in Table 5 directly. Even if the covariates of interest are uncorrelated with unobserved factors that influence the outcome at the outset of the observation period, they may become correlated over time due to differential censoring of respondents, producing dynamic selection bias. To check for this form of bias, we specified the preferred model in a continuous time framework with a person-level random effect. In all these analyses, we obtained similar, statistically significant point estimates for prior extended leave. Although there may still be unobserved characteristics that jointly influence maternity leave and second birth rates or other forms of bias that are responsible for our findings, our sensitivity analyses increase our confidence that extended maternity leave increases the rate of second conceptions among working women.

Discussion

In Russia maternity leave helps alleviate the conflict between work and childbearing: it both promotes women's attachment to the labor force and it increases fertility. Women on maternity leave for the legally-protected period of up to 36 months have more positive work outcomes – higher rates of returning to work, lower rates of quitting or losing their jobs, and higher rates of changing employers – than comparable women not on maternity leave. These findings hold net of extensive controls for individual and structural characteristics that also shape employment outcomes. Maternity leave protected Russian mothers' access to employment as much or more during the post-Soviet economic turbulence as it did in the Soviet period. Excess leave, on the other hand, has no consistent effect, most likely because leave beyond 36 months is not legally protected. We find no evidence that the introduction of capitalism allowed for increased discrimination against mothers or lower compliance with maternity leave regulations. Instead, our findings indicate that maternity leave policies continued to help women negotiate the early years of childrearing while staying attached to the labor market.

Women who took extended leave with a first birth had significantly higher odds of a second conception after returning to work than women who had previously taken short or excess leave. This result holds up in a thorough set of sensitivity analyses. The impact of prior extended maternity leave on second birth rates is particularly important given Russia's recent fertility decline and the long interval between first and second births. Prior extended maternity leave, taken after half of first births, shortened the interval in both Soviet and post-Soviet periods. The effect of maternity leave was only significant after a woman returned to work, perhaps because, consistent with economic theories of childbearing decisions (Kohler, Billari and Ortega 2002), mothers had to work to afford a second child.

This study leaves some questions unanswered. We cannot measure the impact of changes in maternity leave policies or access to benefits. Because our retrospective employment histories only ascertain the duration of leaves women took, we do not know whether employer pressures, women's own preferences, spouse's preferences or other factors were the key influences on how much leave they took. We incorporated all available individual and structural characteristics as proxies for women's and employer's characteristics that could be related to leave duration. We lack data altogether on husband's or partner's characteristics. We also do not know what kinds of child care arrangements are available to women until their children start formal school at age 6. Without direct information about women's options, motives and calculations regarding how much leave to take – which are very difficult to obtain for current leaves, and impossible in a retrospective study like ours – we can only analyze the effects of short vs. extended maternity leave, correcting as best we can for possible endogeneity. We lack the space here to analyze in detail the many possible ways that women might “game” the maternity leave system, for example, timing conception to maximize maternity leave benefits (Hoem 1993). Nonetheless, we find no evidence that women have higher rates of second conception while on leave or quit their jobs when leave payments expire.

Although we find that maternity leave helps maintain Russian women's position in the labor force for up to three years, it could be the case that women suffer earnings penalties or have higher rates of job loss after they return to work from an extended leave. We did conduct preliminary analyses to see if prior extended leaves are linked to elevated risk of job loss after return to work: the coefficient was in the expected direction (indicating higher rates of job loss after the return from an extended leave), but it was not statistically significant. More careful analysis of this issue is warranted.

Our estimated effects of maternity leave on both employment and fertility may still reflect unobserved heterogeneity: women who are more oriented to childrearing than career may be more likely to choose certain types of jobs, take maternity leave and have second births. Our fertility models controlled for this type of unobserved heterogeneity using estimated propensities to take a leave of more than five months. The robustness of our findings increases our confidence that the effect is valid and reflects the mechanism we propose: extended leave allows women to stay at home with an infant until they are ready to return to work without risking job loss, thereby encouraging women to combine work and motherhood again. But we cannot rule out the possibility that our estimated propensities do not fully account for unobserved factors that jointly influence uptake of extended maternity leave and second conception hazards. Our approach to dealing with unobserved heterogeneity may be useful for future studies: the richer the set of variables to model antecedent propensities for taking longer maternity leave, the greater one's confidence that the estimated effects are causal in nature.

Limitations notwithstanding, our study breaks new ground by directly examining individual-level associations of maternity leave with employment transitions and fertility. Previous research has posited that women limit their fertility to accommodate their labor force activity or adjust their labor force behavior to their fertility, depending on institutional support (Brewster and Rindfuss 2000). Our findings suggest that maternity leave provides a mechanism for balancing work and family, an undertaking which has become increasingly more difficult in industrialized countries. These effects are particularly noteworthy in light of the rapidly changing economic, political and social environment in Russia during the post-Soviet period. Given that uncertainties in the labor market and low fertility continue to plague Russia, this research suggests that maternity leave policies should be maintained.

Notes

References

Appendix