Abstract
Using data from the German Socio-Economic Panel, the authors examine the impact of leisure, both in terms of the total amount of leisure time and of specific types of leisure activities, on the transition rate to parenthood 2 years later (N = 3,795 women and N = 3,712 men). With regard to the amount of leisure time, it was found that strongly leisure-oriented women exhibited a lower likelihood of first motherhood only if no close family members lived nearby who could provide informal child care support. Analyses concerning specific types of leisure activities revealed clear timing effects: Respondents, particularly women, with a strong preference for out-of-home leisure activities tended to delay their first births. However, at the age of 41 years, they ended up with a level of childlessness comparable to those who engage in out-of-home activities less frequently. No similar delaying effects could be found for other types of activities.
Throughout the past few decades, the menacing vision of an uncontrolled, rapid growth of the world population has given way to concerns about the steep global decline of birth rates, which, since the 1960s, have leveled off well below the replacement level in most industrialized countries (Morgan, 2003). To explain trends of decreasing fertility, economists have pointed out that it may primarily be the costs of parenthood that lead many couples to refrain from engaging in reproductive activity. Specifically, it has been argued that, as a result of the massive educational expansion in Europe that started in the 1960s and that particularly benefited women, the decline in birth rates since that time may be partially explained by the rise in opportunity costs of children (e.g., loss of income and human capital) for mothers (Blossfeld & Huinink, 1991; Blossfeld & Jaenichen, 1992). Issues of combining children and career have become even more pronounced as women’s work aspirations have increased over time (Brewster & Rindfuss, 2000).
However, the work role does not represent the only activity that competes for time with family roles. As early as in the beginning of the 20th century, the economist Lujo Brentano pointed out that the modern dawn of prosperity in the Industrial Era had not paved the way for a baby boom but, in contrast, entailed a variety of new “pleasures” that compete with the love of children:
The increase of new inventions and discoveries, of trade and travel, the general spread of education and culture have widened the circle of interests and the tastes of both men and women, have increased their requirements and made many new pleasures accessible to them. (Brentano, 1910, p. 376)
This point was elaborated on later by Richard Easterlin, who posited, in contrast to traditional economic theory (cf. Stigler & Becker, 1977), that preferences for children are not constant but have to be considered in conjunction with other “objects of expenditure which compete with children—a new home or additions to the present one, recreation (foreign travel, purchase of a summer home or boat), purchase of a business, education of children, leisure, and so on” (Easterlin, 1969, p. 134). A socioeconomic framework of fertility, then, needs to take into account individual preferences for children, relative to other goods that may provide benefit for the individual (Easterlin, 1973). Hence, it is reasonable to hypothesize that, similar to occupational aspirations, a high leisure orientation tends to interfere with both fertility intentions and behavior.
Following this line of thought, the present study aims at exploring the link between leisure and the transition to parenthood. In an extension of earlier work, several conditions are identified under which this link is established or attenuated.
Leisure and the Life Course: Theory and Research
Existing research has primarily focused on the impact of life course transitions on the leisure time budget. Not surprisingly, it has been found that the transition to parenthood imposes considerable constraints on the parents’ leisure time budget, especially for mothers (e.g., Bittman & Wajcman, 2000). Moreover, the birth of a child seems to interfere, at least temporarily, with couples’ engagement in joint leisure activities (Claxton & Perry-Jenkins, 2008; Kalmijn & Bernasco, 2001), even though this leisure-constraining effect of young children appears to have decreased over the past decades (Voorpostel, van der Lippe, & Gershuny, 2008). However, whereas there is substantial empirical evidence concerning the impact of the transition to parenthood on leisure (Henderson, Hodges, & Kivel, 2002), the reverse causal direction has only rarely been studied.
In the present study, it was assumed that individuals are not ignorant of the fact that parenthood reduces the amount of leisure time. Instead, we argue that squeezes in leisure time after the first birth are anticipated and taken into account in fertility decisions. There are surprisingly few studies so far that have explicitly dealt with effects of leisure on fertility. Nomaguchi (2006) reported some evidence for a small negative effect of the unspecific leisure time budget, measured by weekly hours spent on leisure activities 2 years before (t − 2), on the likelihood of first births (at year t). Heaton, Jacobson, and Holland (1999) found that respondents who anticipated that parenthood was going to interfere with their time for leisure and social activities were more likely to remain childless over a 6-year period. The finding that “some of these people, however, eventually decide to have children” (Heaton et al., 1999, p. 536) pointed to a temporary delaying effect of leisure.
Because of the close empirical association between marriage and fertility, some fruitful insights may be drawn from studies that related leisure to marriage propensity. Existing research suggests that the amount of leisure may lower the propensity to marry (Clarkberg, Stolzenberg, & Waite, 1995; Duvander, 1999). Lois (2008) showed that a preference for time-consuming, action-oriented leisure activities decreased the rate of first marriages, whereas activities that were carried out less frequently (i.e., cultural participation) had a positive impact, at least for men. Interestingly, the negative effect of action-oriented leisure activities disappeared once employment and birth of the first child were controlled for, which points to a potential negative association between leisure and fertility.
In summary, there is still a gap in the literature on the nature of the leisure–fertility association that needs to be filled by research based on large, national samples. We hypothesize that the more time an individual spends on leisure in general, the lower the transition rate to parenthood becomes (leisure quantity hypothesis—Hypothesis 1). However, the findings from the Lois (2008) study also suggest that it may be sensible to further distinguish between different types of leisure activities.
Factors That Moderate the Impact of Leisure on Fertility
Apart from determining the impact that leisure has on the decision for parenthood, the present study was aimed at examining conditions under which this association becomes stronger or weaker.
First of all, we suspected that the leisure effect depends on the substantive use of leisure time. That is, fertility decisions should not only be influenced by the overall time budget available for leisure but also by the specific types of activities time is spent on. We assumed that the critical distinction concerns the compatibility of specific activities with child care. Specifically, we expected home-based activities to be the least likely to interfere with parenting chores, partly because they can more easily be suspended and resumed later, according to situational child care demands, and because they require less time (e.g., no travel time). In contrast, leisure activities that are predominantly engaged in outside one’s home, such as going to the cinema, may be considerably harder to keep up after the transition to parenthood (Claxton & Perry-Jenkins, 2008; Montgomery & Robinson, 2010). Therefore, children should be more likely to be perceived as an obstacle to out-of-home activities than to home-based activities; hence, the former should exert a clearer negative effect on the transition rate to first births than the latter (leisure types hypothesis—Hypothesis 2). Apart from testing this hypothesis, we also conducted exploratory analyses on the impact of further activities (see the Method section for details).
From a gendered leisure perspective (Henderson, 1996; Henderson & Hickerson, 2007), differential effects of leisure on fertility decisions were expected for men and women. To begin with, the literature suggests that the role of leisure in shaping life course transitions may be gender specific. Previous research has shown that involvement in action-oriented (e.g., physical) activities is particularly important for women’s identity formation during adolescence because it fulfills otherwise unaddressed needs of independence and autonomy (Shaw, Kleiber, & Caldwell, 1995). At the same time, women often tend to feel that they are not entitled to leisure, which contributes to the contradictory nature of women’s leisure (Henderson, 1996). These issues are especially relevant in the case of Germany. Even though the more recent family policies were designed to encourage the involvement of husbands in child rearing (see Fagnani, 2007), mothers, particularly in West Germany, still devote far more time to parenting chores than fathers do (Eurostat, 2004). Following the predominant “modified male breadwinner model,” German mothers usually attribute the primary responsibility of child care to themselves (DiPrete, Morgan, Engelhardt, & Pacalova, 2003). Compared with the United States, West German mothers were found to be far more likely to exit the labor force after a birth, despite the fact that this frequently resulted in marked loss of income (DiPrete et al., 2003). Other research has shown that, even if mothers retained their jobs after a birth, employment (indicated by personal income) did not seem to release them from parenting obligations (e.g., by leading to a more egalitarian division of labor) but in fact tended to increase the time they reported spending on child care (Gupta, Sayer, & Cohen, 2009). Because parenthood more strongly affects women’s than men’s leisure (Mattingly & Bianchi, 2003), the negative impact of leisure orientation should be stronger for women (gender hypothesis—Hypothesis 3).
Concerning indicators of work orientation and labor force participation, we expected the strongest negative impact of leisure among employed women, because this group should be particularly reluctant to sacrifice the scarce “time of their own” (Nomaguchi, 2006) for parenthood chores (employment hypothesis—Hypothesis 4). Previous research remains inconclusive in that it failed to substantiate the hypothesized moderator effect between leisure and employment status (Nomaguchi, 2006).
Last, we expected that availability of external child care providers may buffer goal conflicts between leisure orientation and fertility. Child care may be a formal or an informal arrangement, the former comprising paid professional child care and the latter referring to unpaid child care services (e.g., by family members). Germany, especially the Western part where our sample was predominantly drawn from, has long been characterized by a marked degree of structural work–family incompatibility (Matysiak & Vignoli, 2006). Traditionally, there has been a considerable undersupply of crèches and day care facilities for toddlers in West Germany (Fagnani, 2007), so German mothers may be particularly dependent on informal child care during the first few years after a birth occurs (Hank, Kreyenfeld, & Spieß, 2004). Hence, the anticipation of child care services provided by close family members should reduce the opportunity costs anticipated by mothers, such as loss of leisure time. Consequently, we hypothesized that availability of informal child care mitigates the negative impact of leisure on the transition to parenthood (child care hypothesis—Hypothesis 5).
A more general research question concerned predictions of level versus timing effects. Not only may involvement in particular leisure activities delay first births, but it may also lead couples to ultimately refrain from having a child at all. We hypothesized that highly leisure-oriented individuals tend to delay the transition to parenthood temporarily during adolescence and young adulthood, when leisure becomes a paramount constituent of identity formation (Shaw et al., 1995); previous life course studies have found peaks in leisure orientation in these developmental stages (Isengard, 2005). However, even though fertility plans are postponed, they may be revitalized once again later on in the life course, after the preference for leisure has faded and motives of generativity have started to prevail (Heckhausen, Wrosch, & Fleeson, 2001). Consequently, we assumed a temporary delaying effect of leisure (timing hypothesis—Hypothesis 6).
The analyses were confined to first births because individuals who have already chosen to become parents might represent a select group of persons who have either adjusted their leisure orientation to the requirements of the parental role or who have successfully reconciled their leisure preferences with their parenting chores in some idiosyncratic way. Hence, it seemed reasonable to assume that opportunity costs are most salient for childless individuals.
Method
Data and Analyzed Sample
The analyses were based on longitudinal data from the German Socio-Economic Panel (GSOEP; see Wagner, Frick, & Schupp, 2007). One advantage of this data set is the annual rhythm of the survey, yielding both short-timed repeated measurements of leisure activities and detailed information on fertility behavior. Although the GSOEP was initiated in 1984, we chose to confine our analyses to Waves I (1992) through Wave Z (2009) because prior to 1992, leisure time use was not assessed routinely. Because our main independent variables, leisure time budget and leisure activities, were considered in lagged form (t − 2), the period of observation for first births spans the years 1994 through 2009; the first two waves (1992 and 1993) only served as a source of covariate information. The unbalanced panel sample for which there was information on first births and leisure comprised a total of N = 3,795 women and N = 3,712 men aged 17 to 41, all of whom were childless, according to their biographic data, at their first time of observation. Because households are sampled in the GSOEP, information on both partners from cohabiting couples can be matched; however, because there was no dyadic data for couples with separate households (see Table 2), we estimated separate models for men and women. To reduce potential selection biases, respondents without a partner were also included in the sample.
To avoid bias resulting from listwise deletion of missing values while taking full advantage of the available data, we used multiple imputation (Little & Rubin, 2002) as implemented in Stata 11 to impute both occasionally missing covariate information and single indicators of leisure (e.g., time spent on leisure on Sundays) for some waves, prior to conducting the analyses. Following Rubin’s (1996) suggestions, we ran five multiple imputations including all covariates from the fully specified model reported in the Results section (see Table 4). In additional analyses, we tested whether the main independent variables, leisure time budget and specific leisure activities, predicted panel attrition. After controlling for respondents’ age and serial number of panel wave, no significant effects of the leisure indicators on the likelihood of dropout were found.
Analytical Approach
To take advantage of the panel data, discrete-time event history analyses were used (Allison, 1982; Singer & Willett, 2003). This analytical technique is appropriate when analyzing events occurring across discrete periods of time (i.e., panel waves); it approximates continuous time survival analyses as measurement intervals become smaller and has been used in previous studies on fertility (Barber, 2001; Nomaguchi, 2006). The analysis is conditioned in that at their first time of observation, all respondents are childless. When analyzing panel data, time-discrete survival analyses represent the best approach because they appropriately deal with this conditioning and the resulting left truncation of the sample (i.e., not all respondents are observed from the beginning of their fertile period) by using a conditional likelihood that yields unbiased estimates (Guo, 1993). Within this framework, the dependent variable is binary and indicates whether the event of interest does or does not occur between two panel observations. The central concept of event history analysis is the transition rate, which, in the discrete-time case, may be interpreted as a conditional probability for a change of state at time t, given that the units of observation belong to the risk set until time t, which means that no event has been observed for them yet.
We used a multispell approach in which subjects were defined to be at risk until the dependent event or right-censoring occurred. The data were arranged in such a way that each calendar year represented a single episode (person-years). The period of observation ended in the case of a first birth or in the case of right-censoring (or if panel attrition occurred).
One strength of discrete-time survival analyses is that timing effects can be explicitly modeled by simple interaction terms between a covariate and the process time indicator (e.g., different age categories; cf. Singer & Willett, 2003, pp. 451-460); for a graphical illustration of level and timing effects, see Figure 1. This way, it was determined whether leisure time affected the transition rate to first birth in specific phases only or equally across different age-groups.

Illustration of timing and level effects.
Measures
The wording of the indicator for leisure time budget on workdays was “How many hours do you spend on hobbies and other leisure activities on a typical workday?” A similar item was used for leisure time on weekends. The workday leisure time budget was assessed at each wave across the whole period of observation (1993 through 2007), whereas the measures for weekends (i.e., Saturday and Sunday) were contained in every other wave. Missing information on single indicators (e.g., leisure time on weekends) as well as obviously implausible assessments (e.g., spending more than 16 hours a day on leisure), coded as missing in the GSOEP, were multiply imputed from all nonmissing model variables (for measurements across waves, see Table 1). Hence, to get an estimate of the average weekly leisure time budget, workday leisure hours were multiplied by 5 (assuming a 5-day work week) and added to the Saturday and Sunday time budgets.
Measures of Leisure Time Budget and Types of Leisure Activities (1992-2007).
Note: x denotes complete assessments with all indicators, (x) refers to partial measurements with a slightly shortened version of the respective scale; short versions were complemented by means of multiple imputation to resemble the long version (see the text for details).
Specific leisure activities were assessed in several waves between 1992 and 2003 (see Table 1 for details), using multiple indicators to measure the frequency of leisure activities on the following 5-point scale: 5 = daily, 4 = at least once per week, 3 = at least once per month, 2 = less frequently, and 1 = never. All corresponding indicators were entered into a principal component analysis, yielding a four-dimensional structure. First, out-of-home activities (5 items covering 10 specific activities) were represented by the frequency of attending cultural events (opera, theater, exhibitions, and concerts), sport events, and entertainment facilities (going to the movies, attending pop and jazz concerts, going to the discotheque), going out to eat and drink, and going on excursions (across panel waves, mean α = .63). Second, home-based activities (three items covering four activities) comprised various activities at home (e.g., doing handicrafts and gardening), doing repairs, and private use of a personal computer (mean α = .49). The third factor referred to mutual visits with friends and relatives (two items, mean α = .45). Fourth, another two indicators measured involvement in social and political nonprofit organizations (mean α = .44). The last two scales were used in exploratory analyses. To reduce potential bias from reverse causation, all leisure time indicators, concerning both time budget and specific activities, were modeled as time lagged (t − 2). This way, it was made sure that the leisure time assessments referred to the period prior to the onset of pregnancy. On some occasions, only a subset of the out-of-home leisure activities were assessed, containing 8 out of the 10 activities (see Table 1). To minimize bias from listwise deletion (i.e., omission of all indicators from waves with partial measurements), single missing indicators were multiply imputed using all available covariate information (Acock, 2005). 1 Supplemental analyses with listwise deletion generally yielded results similar to those reported in Tables 3 and 4. It should be noted that only waves with partial or full measurements of the four types of activities (and, thus, only birth events from the respective t + 2 waves) were included in the analyses; the sample sizes thus differed markedly across types of activities.
Associations between quantity and types of leisure were tested using Pearson correlations across all waves. The associations were slightly negative for home-based activities (r = −.02) and positive for both out-of-home activities (r = .11) and for meeting with relatives and friends (r = .08), whereas the association between total leisure time and social and political involvement was trivial in size (r = .00). As expected, out-of-home leisure activities thus seem to require more free time than home-based activities.
The remaining covariates were coded as follows. Process time was captured by seven age-group dummy indicators (17-19, 20-22, 23-25, 26-28, 29-31, 32–34; omitted, 35-41 years). 2 This approach allows for a nonlinear hazard of the transition to parenthood over time and for explicit tests of timing effects in specific age intervals. To identify period effects, the serial number of the panel wave was controlled. Partnership status was assessed by means of a time-varying categorical variable with one of the following four values: marriage, unmarried couple (nonmarital union with a shared household), “living apart together” (unmarried, separate households), and single (i.e., no partner). Level of education was considered as a time-varying covariate, according to the CASMIN classification, comprising 10 categories (Brauns & Steinmann, 1999); this classification takes school, vocational, and professional education into account and assumes an ascending order of certificates and degrees. To each of these 10 categories, we assigned the years of schooling that are necessary for obtaining the respective educational levels in Germany (range: 8-18 years). Educational status was measured by a time-varying dummy variable (time lagged, t − 2), which was coded as 1 if the subject was currently enrolled in the educational system (secondary education or vocational training and tertiary education, excluding further training). The employment index captured the number of months the respondent was employed in each year, based on retrospective accounts in the subsequent wave (months of part-time employment were weighted by the factor 0.5). Additionally, weighted household income was modeled as a time-varying variable (t − 2). The weight was determined from the number of persons living in the household, according to the OECD scale (principal earner = 1.0, other persons in the household older than 14 years = 0.5, persons in the household younger than 14 years = 0.3). General life satisfaction was assessed on a 10-point response scale (1 = very low satisfaction and 10 = very high satisfaction). Availability of informal child care was considered as a time-varying dummy variable (lagged, t − 2) coded 1 if the respondent’s mother or father lived either in the respondent’s household or in the vicinity. The time-varying dummy indicator for East versus West Germany was based on the current place of residence.
Description of the Sample
Table 2 shows a description of various sample characteristics, pooled over the whole period of observation. Concerning hours of leisure per day, it can be discerned that men reported having more time available for leisure activities, although the gender differences were small: Both on a typical workday and on weekends, men reported to have roughly 20 minutes more leisure time than women. This leisure-related “gender gap” is consistent with other studies (Eurostat, 2004). However, in our sample, there seemed to be more variation within than between groups.
Pooled Means and Standard Deviations for Variables in the Analysis.
Note: Employment index was computed as the weighted number of months in the labor force per year (the weight was 1 for full-time, and .5 for part-time).
Considering specific types of leisure activities, gathering with friends and relatives was the most popular activity. Moreover, the average frequency of home-based activities was substantially higher than that for out-of-home activities, most likely because of the lower time requirements. Concerning gender differences, the male respondents reported engaging in all types of activities, except social activities, slightly more frequently than women.
With respect to sociodemographic characteristics, because of the relatively young sample, a considerable proportion of respondents were still enrolled in the educational system, and many were living in nonmarital arrangements or without a partner. Less than one third of respondents resided in the East German federal states.
Results
The Quantitative Dimension: Leisure Time Budget
The first series of survival models referred to the impact of the quantitative dimension of leisure on the likelihood of first births (see Table 3). Model 3 illustrates the curvilinear, inversely u-shaped hazard rate for women, with the likelihood of first births rising over time until age 29 to 31 and a subsequent decline; the curve for men reached its maximum later (age 32 to 34, see Model 1). Furthermore, there was a negative effect of the panel wave, indicating that the transition rate to parenthood generally tended to decline between 1995 and 2009, for both men and women.
Logistic Regression Estimates for the Transition to First Birth, Predicted by Leisure Time Budget.
p < .05. **p < .01.
However, the main focus of the present study was on the effects of leisure on subsequent first births. Whereas there was an initial negative effect of leisure quantity among both men (Model 1) and women (Model 3), this effect became insignificant once the control variables were introduced (Models 2 and 4). Supplementary analyses showed that partnership type largely accounted for the effect; specifically, the less leisure-oriented respondents were underrepresented in the more committed relationships where the highest transition rates to parenthood were found. This means that, across the whole period of observation, there was no clear impact of weekly hours of leisure on the propensity to start a family. Additional analyses (not shown) intended to detect potential timing effects by means of interaction terms “Leisure time × age-group” revealed that the quantity of leisure time did not lead to delays of first births, either. Hence, the analyses did not support the leisure quantity hypothesis (Hypothesis 1).
The effects of the additional covariates were generally as expected (see Models 2 and 4). First of all, the more committed the type of relationship, the more likely the respondents were to become parents (Heaton et al., 1999). Whereas there was no effect of educational level, being enrolled in the educational system had a negative impact on the transition to parenthood for women (Blossfeld & Huinink, 1991). In the male subsample, employment exerted a positive effect on the likelihood of first births, which is consistent with previous research (e.g., Heaton et al., 1999). Among women, there was an inversely U-shaped effect of the employment index; this might represent a combination of income and opportunity cost effects. As expected, availability of informal child care support increased the rate of first births among women, whereas life satisfaction had a positive impact among men only. Finally, in line with existing research (Arránz Becker, Lois, & Nauck, 2010), East German respondents exhibited an elevated rate of first births, relative to West Germans.
However, despite the failure to find main effects for quantity of leisure, there might still be variations in the strength of the leisure–fertility association across different subgroups. To test moderator effects, interaction terms were added to the model. Whereas there was no evidence for moderator effects among men at all (results not shown), one interaction effect “leisure time budget × informal child care” emerged for women. Because this interaction term was contained in Model 5, the interpretation of the effect of leisure time budget (t − 2) changed to a conditional one (for details, see Frazier, Tix, & Barron, 2004): Given that potential child care support by close family members was unavailable, there was a significant negative impact of leisure time on the risk of first births (b = −.02). Furthermore, as was expected in the child care (moderator) hypothesis (Hypothesis 5), this negative effect actually disappeared if there were informal child care providers available, indicated by the sum of conditional and interaction effects (b = −.02 + .02 = .00). The coefficient for informal child care (b = .20) indicated that for women with an average amount of leisure (the value 0 represents the mean because the variable was standardized), availability of child care support increased the risk of first births. The main result from Model 5 was that only in the subgroup of women who had no family members living nearby, the amount of time spent on leisure adversely affected their propensity to start a family.
To illustrate this moderator effect, Figure 2 displays the predicted hazard rates of four distinct subgroups (cross-tabulation of either low or high scores on leisure time budget and availability versus unavailability of child care support). As can be discerned in the diagram, there was a considerable negative impact of the leisure time budget if informal child care was unavailable (see the lower two lines in the diagram). In contrast, the birth rate was considerably higher if informal child care was available; in that case, the amount of leisure hardly made any difference at all.

Level effects of leisure time on the transition to parenthood, moderated by informal child care.
Further analyses (not shown) yielded no evidence for moderator effects of other characteristics such as relationship type, employment index, and education. Therefore, the remaining moderator hypotheses were not supported.
Specific Types of Leisure Activities
Next, we turn to the effects of different types of leisure activities. As expected in the leisure types hypothesis (Hypothesis 2), whereas home-based activities did not exert any clear effect, neither on the level nor on the timing of first births (results not shown), the analyses yielded some evidence for a negative impact of the out-of-home activities scale (see Table 4). In the initial model without control variables, out-of-home activities were associated with a slightly lower transition rate to first births among women (see Model 3), whereas no effect could be found for men (Model 1). However, this effect was rather small and became insignificant after including controls (especially partnership type; results not shown).
Logistic Regression Estimates for the Transition to First Birth, Predicted by Quantity and Types of Leisure.
*p < .05. **p < .01.
Although these results did not support the notion of a time-invariant causal effect of out-of-home activities, they did not preclude the possibility of timing effects; therefore, we also tested potential delaying effects by introducing interaction terms between the out-of-home activities scale and the age categories (see Models 2 and 4). These additional analyses indeed revealed that the effect of out-of-home leisure activities varied substantially over time. Phase-specific timing effects are indicated by the significant “out-of-home leisure activities × age-group” interaction terms in Table 4. The clearest pattern resulted for female respondents (Model 4): We found strong delaying effects of the frequency of out-of-home activities on women’s propensity to start a family, particularly in the first three age categories (spanning ages 17 through 28, Model 4). The corresponding pattern of effects for men was similar, although the effects were slightly smaller (Model 2); whereas the interaction terms accounted for around 1% of the R2 among women, this proportion was roughly 0.5% for men (results not shown). This finding corroborated the gender hypothesis (Hypothesis 3).
An alternative specification in which the reference category was set to the first age-group (not shown) revealed that, compared with the younger respondents, older women engaging in out-of-home activities exhibited catch-up effects with regard to the transition to parenthood toward the end of their fertile period (all interaction effects except with the age-group 17-22 were positive and significant). For the male subsample, similar but weaker trends emerged (see Model 2). Men who adhered to out-of-home leisure activities tended to delay their transition to parenthood at younger ages (age-groups 17-22 and 23-25) but, in the last two age-groups, appeared to end up with a slightly elevated rate of first births (b = .85). Similar to the results for women, men also exhibited catch-up effects when taking the first age-group (17-22) as the reference category (results not shown); interaction effects with the last two age-groups (32-34 and 35-41) were positive and significant.
Additional exploratory analyses examining effects of the remaining types of leisure activities yielded no further evidence for the impact of specific leisure activities on fertility. A small negative effect of involvement in social and political organizations (p < .10) became insignificant once controls were introduced; no effects resulted for social gatherings with family and friends, neither on the level nor on the timing of first births (results not shown).
Generally, the reported analyses provided some support for the leisure types hypothesis (Hypothesis 2) and timing hypothesis (Hypothesis 6) in that out-of-home leisure activities exerted clear negative (timing) effects on the transition to parenthood, whereas other types of activities, including home-based leisure, generally had no impact. Moreover, findings were largely consistent with Hypothesis 3 in that the negative delaying impact of out-of-home activities was somewhat stronger among women than among men, pointing to gender inequalities in the anticipated impact of parenthood on leisure-related habits.
Timing Versus Level Effects
In the last step, Hypothesis 6 was examined more in detail. Based on the most clear-cut timing effect (i.e., the impact of out-of-home activities on the transition to motherhood), it was analyzed to what degree delaying effects were caught up later on. Figure 3 displays the hazard rates of first births for women with a high versus low preference for out-of-home leisure activities (i.e., those in the upper vs. lowest quartile of the distribution). Across the initial 9-year period, women with a high preference for out-of-home activities seemed to be more hesitant to become mothers than women without such a leisure preference. Subsequently, they started to catch up, though; from the age of 29 onward, their rate of first births equaled or even exceeded that of the other group.

Timing effects of women’s out-of-home leisure activities on the transition to parenthood.
To obtain an estimate of level effects, survival curves for first births were computed for the mentioned two groups (see Figure 4). The figure shows that despite the early delay of parenthood among the women who frequently engaged in out-of-home activities, this group exhibited a precipitously accelerated rate of first births during the middle years. Both groups ended up with a similar level of childlessness toward the end of their fertile period. Hence, with regard to first births, the delaying effects of the preference for out-of-home leisure activities seemed to be fully compensated by later catch-up effects. This finding suggests that the delaying impact of leisure on the transition to parenthood did not ultimately translate into a sustainable level effect.

Predicted survival curves for the transition to first births among women with different preferences for out-of-home leisure activities.
Discussion
The present study was aimed at examining the impact of leisure, both in terms of quantity and specific types of leisure activities, on the process of family formation. Our findings yield a rather complex picture of the leisure–fertility association. Although there was little evidence for an overall inhibiting impact of the quantitative amount of leisure on the rate of first births, net of other potential confounders (disconfirming Hypothesis 1), the analyses clearly indicated that the leisure time budget may inhibit the transition to parenthood if family members who could serve as potential child care providers are unavailable (see Hypothesis 3). Conversely, this implies that family networks (e.g., the geographical proximity of family members) can serve as a means for reducing goal conflicts resulting from structural incompatibilities between family and other life domains on the societal level that are rather pronounced in Germany (Matysiak & Vignoli, 2006). This factor has to be seen in conjunction with the notorious undersupply of institutional child care facilities in Germany, which contributes to young parents being particularly reliant on informal child care arrangements (Hank et al., 2004).
Perhaps, the most striking finding was that the mere quantity of leisure did not seem to be of primary importance for the decision to start a family. Rather, it turned out to be necessary to further qualify the leisure–fertility association by distinguishing between different types of leisure activities (Hypothesis 2). To assess the dynamics of the impact of leisure behavior on family formation over time, it was examined whether involvement in leisure activities primarily delayed first births or whether it ultimately led to a higher level of childlessness. Whereas there was little evidence for level effects on any of the leisure dimensions, a clear timing effect emerged for out-of-home activities. Specifically, the hypothesized negative impact of out-of-home activities was found to be largely temporary and was entirely compensated by catch-up effects (consistent with Hypothesis 6). Perhaps, a preference for out-of-home leisure activities is typical for a group of particularly resilient “sensation seekers” who first intensely enjoy the pleasures of leisure and then, at some point, decide to not forgo the joys (and challenges) of parenthood. However, it would be premature to claim that leisure has no lasting impact on the number of children. It is likely that even though the delay of parenthood may not lead to a deficit in first births, it definitely shortens the time frame for higher parity births, something we could not address with the presented analyses. Besides, it should also be noted that we could not follow all female respondents across their whole fertile period.
Interestingly, home-based as well as other activities did not show any impact at all, neither across the whole period nor in specific phases. The results thus suggest that it is not leisure per se that competes with parenthood for time, but that some leisure activities, particularly those that require leaving one’s home and involve fixed schedules (e.g., a visit at the opera), are anticipated to be less compatible with parenting chores than others. Further research is required to replicate and complement the somewhat exploratory classification of leisure activities used in the present study.
As a further qualification, we also found some evidence for gender differences; women’s leisure generally had a stronger impact than men’s leisure on the transition rate to parenthood. This finding, in conjunction with the literature on the leisure-related gender gap (Henderson & Hickerson, 2007), raises serious concerns about women bearing a disproportionate share of the costs of parenthood. It thus illustrates the pervasiveness of traditional gender roles, despite women’s increasing labor market participation and changes in the gender division of labor (Bianchi, Milkie, Sayer, & Robinson, 2000).
When relating our findings to the general increase in leisure time (Aguiar & Hurst, 2009; Robinson & Martin, 2009) and to some authors’ claims of an increasing importance of leisure for identity formation (Rössel, 2008), it becomes clear that the notion of leisure-related opportunity costs may represent an important yet neglected explanatory factor that may help us better understand the pervasiveness and sustainability of the declining birth rates that many industrialized countries have been witnessing during the late 20th century. However, we think that blaming the “Leisure society” (Seabrook, 1988) for the downward trend in fertility would be premature for several reasons. Although we were able to use panel data starting from the early 1990s, this period of observation obviously does not allow any direct conclusions on the link between positive trends in leisure time budget and parallel negative historical trends in fertility since the 1960s. Hence, the causal association between increasing leisure and decreasing birth rates on the societal level remains highly speculative. Moreover, leisure represents an important element of identity formation during adolescence and early adulthood, particularly for women (Shaw et al., 1995). Similar to the transition to parenthood, such identity formation also represents an important developmental task in the life course that should by no means be undervalued; we think that setting off the benefits of leisure against those of parenthood is rather an individual decision than a scholarly topic. Finally, it should be noted that the subgroup in which there was a negative leisure time budget–fertility relation was rather limited in size. For the majority of the sample, namely, respondents who could resort to relatives for child care support (in the sample studied, about 75% of the respondents), leisure did not seem to be a serious obstacle to the decision for parenthood.
A limitation of the presented work concerns the somewhat heterogeneous measurements of leisure. Whereas it would have been desirable to have more comprehensive information about specific leisure activities assessed by separate indicators, the available measures were somewhat coarse and slightly differed between panel waves. Also, the different response formats (i.e., total time budget vs. frequency assessments) limit the comparability of the results for quantity and types of leisure. Future studies with more sophisticated measurements are required to qualify and extend our findings in this respect. Moreover, we could not address higher parities in this study. Yet to assess the impact of leisure orientation on the final number of children, it would be necessary to extend the analyses to subsequent births; this way, future studies can shed more light on both parity-specific effects of leisure and on contingencies between consecutive births (e.g., to what extent delaying effects translate into shortened time frames for subsequent births).
In summary, the present study extends the economic perspective on parenthood-related opportunity costs by including intangible, nonmonetary losses that couples anticipate when they decide whether to start a family. Previous research has largely focused on work–family incompatibility, particularly for women (Brewster & Rindfuss, 2000), whereas other “competing pleasures” outside the family domain have not been considered equally. Because the outlined impact of leisure appeared to be largely independent from respondents’ educational background and income potential, we believe that our analyses do support the notion of independent leisure-related opportunity costs of parenthood. In our view, this type of costs complements the economic perspective, which focuses on monetary opportunity costs resulting from forgone earnings, in important respects. However, the analyses also showed that the leisure–fertility association is not straightforward but that it is moderated by several factors (e.g., types of leisure activities) as discussed in detail above. Even with these qualifications in mind, the present study adds to a growing literature pointing to the important role that couples’ leisure arrangements play in both shaping and reflecting life course decisions (for the impact of leisure on union stability, see Arránz Becker & Lois, 2010). Hence, we think that couples’ leisure time deserves more attention in family research than it has received so far.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
