Working time preferences, hours mismatch and well-being of couples: Are there spillovers?

Labour Economics 24 (2013) 244–252

Contents lists available at ScienceDirect

Labour Economics journal homepage: www.elsevier.com/locate/labeco

Working time preferences, hours mismatch and well-being of couples: Are there spillovers? Christoph Wunder a,⁎, Guido Heineck b a b

University of Erlangen-Nuremberg, Germany University of Bamberg, IZA Bonn, Germany

H I G H L I G H T S • • • • •

We analyze how well-being is related to working time preferences and hours mismatch. Particularly underemployment is detrimental for well-being. We provide first evidence on spillovers from the partner's mismatch. Findings suggest that the spillover is due to caring preferences. Redistribution of work hours could improve well-being of workers and their families.

a r t i c l e

i n f o

Article history: Received 12 November 2012 Received in revised form 21 August 2013 Accepted 1 September 2013 Available online 11 September 2013 JEL Classification: I31 J21 J22

a b s t r a c t Working time arrangements determine, to a large extent, the successful balancing of work and family life. This study investigates the role of working time preferences and hours mismatch for well-being among couples. The empirical evidence indicates that well-being is generally lower among those with working time mismatch. Particularly underemployment is detrimental for well-being. We further provide first evidence on spillovers from the partner's working time mismatch that are, however, no longer significant once we control for the partner's well-being. This suggests that well-being is contagious, and that the spillover is due to caring preferences. © 2013 Elsevier B.V. All rights reserved.

Keywords: Subjective well-being Life satisfaction Working time preferences Working time mismatch Spillovers Caring preferences

1. Introduction Recent decades have been marked by profound changes of the family roles of men and women. The male breadwinner family is on the decline and females are increasingly participating in the labor force while they continue to raise children and care for elderly relatives (Folbre and Nelson, 2000; Goldin, 2006; Aguiar and Hurst, 2007). To support families in effectively reconciling work and family life, many governments and firms have implemented family time related policies and flexible ⁎ Corresponding author at: University of Erlangen-Nuremberg, Department of Economics, Lange Gasse 20, 90403 Nuremberg, Germany. Tel.: +49 911 5302 260; fax: + 49 911 5302 178. E-mail address: [email protected] (C. Wunder). 0927-5371/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.labeco.2013.09.002

working time arrangements (for an international overview, see Lee et al., 2007).1 Such policies are based, however, more or less on coarse indicators of the situation of families, such as fertility rates and female labor force participation rates. In consequence, little is known about how the well-being of families depends on their working time

1 Issues of reconciling work and family life are high on the political agenda and family time related policies receive much attention in Germany. See, e.g., the 2012 Federal Government's Family Report with its focus on family time policies (BMFSFJ, 2012). Furthermore, a recent reform of the maternity leave regulations in 2007 substituted an earnings-replacing benefit system for means-tested lump-sum payments to give highly educated mothers the opportunity to return to work within one year of the birth of their child and to enable fathers to enhance their childcare involvement (for details, see, e.g., Kluve and Tamm, 2013).

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

arrangements. Do families manage to reconcile work and family life? Are working couples happy with their working time arrangements? However, answers to these questions may help to take appropriate measures to promote the work–life balance. This study uses a direct approach to obtain insights into the lives of families by analyzing the relationship between well-being and working time arrangements among couples. First, we investigate the consequences of mismatches between actual and preferred hours of work for well-being. Second, we study the spillover of one partner's working time mismatch onto the other partner's well-being. Our focus is on couples with children where both partners are employed because they are expected to respond more sensibly to working time mismatches than individuals without caring obligations due to the parental time required for child care. We use measures of life satisfaction to approximate true well-being (or utility). This approach is by now well established in the economic literature (for overviews, see, e.g., Frey and Stutzer, 2002; Layard, 2005; Van Praag and Ferrer-i-Carbonell, 2008). Although subjective measures of well-being were extensively used to analyze various labor market issues—including unemployment (Winkelmann and Winkelmann, 1998; Kassenboehmer and Haisken-DeNew, 2009), unemployment duration (Clark, 2006), income comparisons (Clark and Senik, 2010), or co-worker wages (Clark et al., 2009) — , researchers have so far paid only little attention to an empirical investigation of the wellbeing of mismatched individuals. Our study makes several contributions to the literature. First, we provide new and comprehensive results on how well-being is related to working time mismatches using data on couples from the German Socio-Economic Panel Study (SOEP). Except for Wooden et al. (2009) and Booth and Van Ours (2008, 2009), who provide evidence for Australia and Great Britain, the role of working time mismatch for well-being has remained largely unexplored. As a result, we do not know whether well-being effects of working time mismatch are a common problem or a phenomenon limited to specific labor markets. As a second contribution, we provide, to the best of our knowledge, first evidence on whether and to what extent well-being is related to the partner's working time mismatch. Prior research has not investigated spillovers of a mismatch within couples so far. On the one hand, existing research on working time mismatch did not consider the role of the partner's mismatch (e.g., Wooden et al., 2009). On the other hand, research on family well-being has hardly investigated partner's employment situation. Notable exceptions are Winkelmann and Winkelmann (1995), Clark (2003), and Booth and Van Ours (2008, 2009). However, these researchers focused on the partner's unemployment and actual hours of work while a working time mismatch of the partner was not taken into account. The present study goes some way towards enhancing our understanding of the work–life balance by combining research on working time mismatches and family well-being: taking spillovers of working time mismatches into account for the first time, we are able to draw a more comprehensive picture of the role of working time arrangements for well-being than previous studies. Our third contribution is to disentangle the transmission channels through which one partner's working time mismatch impinges on the other partner's well-being. Here, we distinguish between (i) spillovers due to the shared household environment and the joint consumption of household public goods and (ii) spillovers due to caring preferences. This distinction is of interest not only because it provides deeper insights into the well-being generating processes within families but also because caring preferences have important implications for the redistribution of resources within the family (e.g., Lundberg and Pollak, 1996; Ermisch, 2003) or for non-market insurance arrangements (Chami and Fischer, 1996). The empirical results of this study indicate that working time mismatch is negatively correlated with well-being. Underemployment provokes a stronger response in well-being than overemployment, particular among males. Our study provides first evidence on the

245

existence of a significant spillover from the partner's working time mismatch: both males and females suffer from their partners' underemployment. However, the partner's working time mismatch is no longer significant once we control for the partner's well-being, suggesting that well-being is contagious, and that the spillover is due to caring preferences between partners. 2. Literature review So far, only a small literature discussed questions on the well-being effects of working time mismatch, including spillover effects. First, the general pattern from cross-sectional evidence on working time mismatch is that deviations from preferred working hours are negatively associated with well-being and other outcomes, including health, job, and life satisfaction. Such evidence was provided, for example, by Wilkins (2007), who uses data from the Household, Income and Labour Dynamics in Australia (HILDA) survey, or Grözinger et al. (2008), whose analysis is based on the 2004 wave of the German Socio-Economic Panel (SOEP) Study. Only very few studies employ panel data. The results of Friedland and Price (2003), who use two waves from the Americans' Changing Life (ACL) study, are somewhat at odds with the above mentioned general pattern inasmuch as the authors find a positive relation between underemployment and job satisfaction.2 Closer to our analysis is the study by Wooden et al. (2009) who use panel data from the HILDA survey from 2001 to 2005. They apply pooled and fixed effects estimators to examine the association between working time mismatch and subjective well-being. Their results suggest that a mismatch between actual and desired working hours matters for both life satisfaction and job satisfaction among both males and females, with the (negative) effects being larger for overemployment than for underemployment. As for cross-partner effects of working time itself, Booth and Van Ours (2008) use data from the British Household Panel Survey (BHPS) and report that satisfaction with hours of work and job satisfaction is highest in couples where the male is full-time working and the female is part-time working. This finding was reinforced using 2001–2004 data from the HILDA Survey (Booth and Van Ours, 2009). In their 2009 study, the authors furthermore show that female life satisfaction increases when the male partner is working full-time while male life satisfaction is unaffected by their partner's market hours. Yet, they do not examine whether the individual's and partner's working time preferences or working time mismatch are related to well-being, neither do they look at possible spillover effects of the partner's well-being on the other partner's well-being. Next, we present some previous research on the spillovers of workrelated factors within families. The literature reports negative associations between well-being and, for example, the work stress of the partner (Bolger et al., 1989), the partner's or the parent's unemployment (Clark, 2003; Kind and Haisken-DeNew, 2012), shared components such as household income or housing (Schimmack and Lucas, 2010), or unobserved intrahousehold characteristics (Shields and Price, 2005). These studies do, however, not explicitly address the question of whether these spillovers are driven by the shared context itself or by caring preferences (or altruism). Finally, recent research provides first evidence of cross-individual effects of well-being, i.e., of caring preferences and altruistic links. The existing research in this field has, however, mainly looked at the relation between parents and children. Irrespective of that, this literature consistently reports positive correlations of the well-being of family members (e.g., Winkelmann, 2005; Powdthavee and Vignoles, 2008; Bruhin and Winkelmann, 2009; Schwarze and Winkelmann, 2011). Investigating caring preferences among partners, Powdthavee (2009) applies GMM2 The authors do not have a convincing explanation for this unexpected finding, and it is unclear whether it is driven by their empirical approach as the panel structure is only exploited by including lagged indicators of health and well-being to account for reverse causation.

246

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

system estimators in order to take into account correlated effects of partner's life satisfaction and measurement error bias. He finds a statistically significant spillover effect of life satisfaction within couples. Except for the individual's own employment status, he, however, does not take working time arrangements into account. In sum, only few studies have empirically examined the relationship between well-being and working time mismatch so far, and no research has been found that explicitly looked at spillovers from the partner's mismatch onto the other partner's well-being. In addition, we differ from most of the prior research by employing longitudinal data.

mismatched worker, who in this case is overemployed, deviates from his or her utility maximizing labor supply, i.e. U′(w′,H′) b U1(w′,S(w′)). A similar reasoning holds for a mismatched worker who accepts a job package with short working hours H″ though he or she prefers the higher level of working hours S(w′) given the wage rate w′. In the latter case, the worker is underemployed. The loss in utility, ΔU, resulting from the mismatch is equal to the difference between U1(w′, S(w′)) and U′(w′,H′). Hence, the worker's current utility (i.e. from the actual hours of work) can be expressed by two components: the utility from desired hours of work and the loss in utility resulting from the mismatch:

3. Methodology

 

′ U ′ w′ ; H ¼ U 1 w′ ; S w′ −ΔU ðΔHÞ;

ð1Þ

3.1. Conceptual framework Working time mismatches emerge when workers receive job offers consisting of fixed wage–hours combinations and they are not able to realize their preferred hours of work given the wage offer (Altonji and Paxson, 1988). The wage-hours combinations may, however, vary between employers because of firm-specific requirements, production technology, recruiting costs, legal regulations, or collective bargaining agreements, for instance. Workers adjust their working time by selecting wage-hours combinations across employers offering different job packages, provided that the utility gain exceeds search and mobility costs (Altonji and Paxson, 1992). Thus, we may observe persistent mismatches in the labor market if utility gains are not sufficiently large or the worker has imperfect information about job opportunities, for instance. A working time mismatch leads to lower utility compared to a working time match between preferred and actual hours of work. Fig. 1 illustrates the loss in well-being that results from the working time mismatch. The labor supply curve, S, shows the preferred working hours level at different wage rates. At wage w⁎, the worker maximizes his or her utility, U, at the preferred working hours level S(w⁎). However, we assume that the job package (w∗,S(w∗)) is not in the worker's choice set but that a firm offers the alternative hours-wage combination (w′,H′) that requires longer hours and provides compensation in the form of higher wages (w′ N w∗). The worker may be indifferent between these combinations (i.e., U0(w∗,S(w∗)) = U′(w′,H′)). Yet, the utility level U′ is lower compared to the utility in a situation in which the worker is able to realize preferred working hours given the wage rate w′. The w

where ΔU denotes the loss in utility and ΔH = S(w′) − H′. Considering the work–life balance of families, a working time mismatch may not only affect the worker's own well-being but it may also transmit to the well-being of the partner (i) through the shared household environment and the joint consumption of household public goods or (ii) due to caring preferences between partners (Becker, 1974; Tomes, 1986). In the latter case, the partner's mismatch affects his own well-being and the other partner is affected because he or she cares for his or her partner. The partners are then said to be altruistically linked (Altonji et al., 1992). Knowledge about the relevance of caring preferences with respect to working time mismatches is of interest not only because it provides deeper insights into the well-being generating processes within families but also because of their further economic significance. Theoretical reasoning predicts that caring preferences may serve as a partial insurance against work-related risk (Chami and Fischer, 1996). Since altruisticallylinked partners tend to internalize their partner's loss in utility, caring preferences can be interpreted as an implicit non-market agreement. Therefore, one partner may work more overtime hours if the other partner becomes underemployed, for instance. Furthermore, economic models of the family suggest that caring preferences have important implications for the redistribution of family resources (e.g., Ermisch, 2003). If one partner is an effective altruist (Becker, 1981), then a redistribution of working time, for example, from the overemployed male to the underemployed female will have no impact on household consumption. If, in contrast, neither partner is an effective altruist, then redistribution may affect household consumption because it has an effect on the partners' bargaining power (e.g., Konrad and Lommerud, 1995; Lundberg and Pollak, 1996). 3.2. Estimation controlling for endogeneity

S

We use the following regression equation to explain the variation in subjective well-being, SWB, which is an indicator of current utility of individual i at time t:

U1

SWBit ¼ β1 ′xit þ β2 ′xjt þ δSWBjt þ α i þ it :

U0 =U

w

ð2Þ

We assume that well-being arises from individual characteristics, represented by a vector xit that includes age, health, education, variables identifying the occupational group and the sector of industry, an indicator variable for foreign citizenship, the (logarithm of the) wage rate and the time used for household work. Furthermore, the model takes into account the actual working time arrangements. Following the considerations expressed in Eq. (1), we include the desired hours of work and the hours mismatch as explanatory variables on the right hand side.3

w*

H S(w*) S(w )

H Δ Hunder

H Δ Hover

Fig. 1. Labor supply and fixed wage-hours combinations.

3 The model can, in general, be transformed into an equivalent specification where actual working time is substituted for working time preferences. It can be shown that the coefficient of actual working time is equal to the coefficient of working time preference in the reparameterized model. The coefficient of the working time preference in Eq. (2) therefore represents also the effect of working time on well-being, after controlling for the working time mismatch.

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

The vector xjt denotes variables of the partner that are relevant for the shared household environment, including his or her working time preferences, working time mismatch, hourly wage rate, and time use for housework. β1 and β2 represent vectors of coefficients describing the relationship between response and explanatory variables. As a further component, the model includes the well-being of partner j at time t, SWBjt, to allow for caring preferences. Finally, the error consists of two components: αi denotes an individual-specific error term component, and it is the idiosyncratic component. Including the partner's well-being as an explanatory variable raises concerns about potential problems of endogeneity that may arise from simultaneity or unobserved variables. For example, endogeneity may arise from unobserved variables—such as unobserved household consumption, norms related to the family, or personality traits—that have an impact on both, SWBit and SWBjt. We apply two estimation methods to account for these concerns: As a first approach, we use a fixed effects estimator that allows for unmeasured time-invariant factors that influence both SWBit and SWBjt. This implies that the unobservables represented in the error term it are assumed to be uncorrelated with the partner's well-being. However, a bias in the fixed effects estimates may result if the partner's well-being correlates with the unobservables in the error it. Such a case occurs, for instance, if unobserved well-being generating factors change over time. To address this problem, we employ an instrumental variable estimator as a second approach that estimates an additional equation for the partner's well-being: SWBjt ¼ θ1 ′zit þ θ2 ′zjt þ θ3 ′wjt þ α j þ ν jt ;

ð3Þ

where zit and zjt denote overlapping explanatory variables that also have an impact on SWBit. wjt represents a vector of non-overlapping variables that only affect SWBjt, but not SWBit. αj models time-invariant unobserved heterogeneity and νjt is the remaining idiosyncratic disturbance. We estimate Eq. (3) simultaneously with Eq. (2) applying Limited Information Maximum Likelihood (LIML) (Baum et al., 2007). The selection of our instrumental variables in the vector wjt is based on theoretical reasoning as well as on statistical testing using diagnostic statistics, particularly the Cragg-Donald test for weak instruments (Stock and Yogo, 2002).4 Following Schwarze and Winkelmann (2011),5 we use characteristics of the partner j and various variables describing the interview context of the partner, particularly indicator variables for time of the interview (weekends or during the week) and the interview mode. The following interview modes are used: oral, written questionnaire with interviewer assistance, self-completion without interviewer, written and oral, written by mail, and CAPI. Also, variables describing the interview context of individual i are included in the covariate vector xit in Eq. (2).6 We argue that the interview variables are suitable instruments: First, the literature on survey methodology provides evidence that the interview context is relevant for self-reported well-being. Akay and Martinsson (2009) and Helliwell and Wang (2011), for example, report day-of-the-week effects on subjective well-being. Furthermore, Conti and Pudney (2008) find more positive well-being reports in oral interviews done by an interviewer. The interview context may therefore induce exogenous variation in the partner's well-being. Second, it is plausible to assume that the context of the partner's interview is 4 The online appendix reports diagnostic statistics in detail: using the Cragg-Donald Wald F statistic, we find no evidence for weak instruments that can produce biased IV results; the Sargan-test of overidentifying restrictions of all instruments indicates that the orthogonality assumptions cannot be rejected at any reasonable significance level; finally, our underidentification LM test shows that the excluded instruments are relevant. 5 Their study uses the child's characteristics to instrument the child's well-being in a regression of the parent's well-being on the child's well-being. 6 Variation in the interview context within couples exists for almost 15% of our sample. In these cases, the partners were interviewed with different interview modi and/or on different week days.

247

uncorrelated with unobserved factors affecting the partner's wellbeing, such as unobserved household consumption.

4. Data The analysis is based on data from the German Socio-Economic Panel Study (SOEP), covering the period 1985–2011 (Wagner et al., 2007).7 We focus on families with children because their working time constraints are quite likely more severe, as the children's needs have to be taken into account on top. Since the working time preferences of young couples that are in an early stage of their professional lives and of family formation might contain some measurement error, we select a sample of couples where both partners are employed and of main working age (between 30 and 60). We exclude those individuals where at least one partner is in training, non-working, unemployed, or on maternity leave. As the conditions on the labor markets and the supply of child care facilities are quite different between East and West Germany, we restrict the analysis to West Germany. Table 1 presents descriptive statistics separately for females and males. Males tend to be older than females; the proportion of males in the young age groups is smaller than that of females, while the pattern is reversed in the old age groups. The raw gender wage gap is substantial and amounts to 0.405 log points. Females use considerable more time for household duties (household work, child care, garden work, etc.) than males: while females contribute approximately 8 h per day, males supply less than 3 h. On average, 1.7 children live in each household. The vast majority of working couples has children of school age (i.e. six and older); only 6% have children younger than three and 19% have children between three and six. Table 2 provides a description of the joint distribution of actual and preferred hours of work. The working time preferences were surveyed with an open hour per week response option: If you could choose your own number of working hours, taking into account that your income would change according to the number of hours: How many hours would you want to work? The observed pattern shows a clear difference in labor supply between females and males. While female labor supply ranges predominantly in the intervals between 12 and 28 actual working hours, male labor supply is clustered in the upper categories of actual working hours (≥36 weekly hours). Thus, labor supply follows a gender-specific pattern in which females are part-time employed and males are full-time employed. The main diagonal is dominant in Table 2, indicating a strong positive correlation between actual and preferred hours of work. Nevertheless, we also observe substantial differences between actual and preferred working hours. For example, the majority of men with more than 44 actual hours want to reduce labor supply. In total, 2/3 of women and 3/4 of men report preferences for working hours other than their actual working time. 63% of men and 39% of women are overemployed while 11% of men and 26% of women are underemployed. We use self-reported measures of life satisfaction as an empirical approximation of true well-being (or utility). Life satisfaction is surveyed by the following question: How satisfied are you with your life, all things considered?. Responses are collected on an 11-point scale ranging from 0 (completely dissatisfied) to 10 (completely satisfied). A first impression of the spillover of one partner's working time mismatch on the other partner's well-being can be obtained from Table 3 that shows average life satisfaction by mismatch type of the couple. On average, life satisfaction is lower when the partner is either underemployed or overemployed compared to a state in which both partners realize their working time preference. The spillover is, however, stronger when the partner is underemployed than when he 7 Due to missing data on the respondents' disability status, we cannot use the survey years 1986, 1990, and 1993. Information about working time preferences was not collected in 1984 and 1996. In total, we are able to use 23 survey years.

248

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

for well-being in Subsection 5.1. After that, we look at the spillover in Subsection 5.2.8

Table 1 Descriptive statistics. Females Variable

Mean

Life satisfaction 7.352 Preferred working time (hours per week) 23.224 Actual working time (hours per week) 23.991 Overemployed 0.389 Underemployed 0.267 Overemployment (in hours) 2.958 Underemployment (in hours) 2.191 Time use for housework (hours per week) 7.986 Log of hourly wage 2.333 Age 30–35 0.237 Age 36–40 0.327 Age 41–45 0.280 Age 46–50 0.125 Age 51–55 0.029 Age 56–60 0.002 Education in years 12.123 Foreign citizenship 0.134 Disabled 0.024 Number of children in HH 1.652 Young children LT3 0.059 Young children GE3 LT6 0.190 Self employed 0.083 Occ.: Missing 0.037 Occ.: Managers 0.025 Occ.: Professionals 0.137 Occ.: Technicians 0.250 Occ.: Clerical support workers 0.173 Occ.: Service and sales workers 0.183 Occ.: Skilled agricultural workers 0.007 Occ.: Craft workers 0.033 Occ.: Operators and assemblers 0.044 Occ.: Elementary 0.110 NACE: Missing 0.055 NACE: Other 0.009 NACE: Agriculture and mining 0.031 NACE: Manufacturing 0.086 NACE: Electricity and gas supply 0.043 NACE: Water supply/construction 0.015 NACE: Trade, retail 0.183 NACE: Information/finance 0.065 NACE: Administration activities 0.153 NACE: Education 0.284 NACE: Arts, entertainment 0.075 Interview: oral 0.269 Interview: written without interviewer 0.274 Interview: written with interviewer 0.049 Interview: oral and written 0.042 Interview: mail 0.069 Interview: CAPI 0.221 Interview on weekend 0.184 Number of persons 2,583 Number of person-year observations 12,464

Males Std. dev. 1.513 9.842 12.247 0.488 0.443 5.663 4.873 5.012 0.567 0.425 0.469 0.449 0.331 0.169 0.043 2.700 0.340 0.154 0.721 0.236 0.392 0.275 0.189 0.156 0.344 0.433 0.379 0.386 0.085 0.179 0.206 0.312 0.227 0.094 0.175 0.280 0.204 0.123 0.386 0.247 0.360 0.451 0.263 0.443 0.446 0.217 0.200 0.253 0.415 0.388

Mean

7.268 39.432 44.422 0.631 0.105 5.671 0.681 2.736 2.738 0.128 0.273 0.304 0.193 0.079 0.024 12.362 0.137 0.048 1.652 0.059 0.190 0.113 0.015 0.081 0.199 0.165 0.064 0.039 0.009 0.254 0.121 0.052 0.025 0.014 0.032 0.215 0.119 0.114 0.093 0.101 0.153 0.083 0.049 0.266 0.293 0.045 0.036 0.069 0.210 0.191 2,587 12,464

Std. dev. 1.489 7.032 9.202 0.483 0.306 7.326 3.139 2.312 0.477 0.334 0.446 0.460 0.395 0.269 0.154 2.874 0.344 0.214 0.721 0.236 0.392 0.316 0.122 0.273 0.400 0.371 0.245 0.195 0.097 0.435 0.326 0.221 0.157 0.117 0.176 0.411 0.324 0.318 0.290 0.301 0.360 0.276 0.215 0.442 0.455 0.208 0.187 0.253 0.407 0.393

5.1. Well-being and working time mismatch

5. Results

The results indicate that well-being is considerably lower among workers who are not able to realize their working time preference. In general, the (negative) effect of underemployment is stronger than that of overemployment. The hypothesis that the coefficients of overemployment and underemployment are equal can, however, not be rejected for females. For men, the highly statistically significant coefficient of the underemployment variable shows a reduction in well-being by approximately 0.02 per hour mismatch. Since underemployed men, on average, want to work 6.5 additional hours per week, the corresponding reduction in well-being is 0.12 points. This is equivalent to a decrease in wages of approximately 50%.9 In contrast, our results do not indicate a significant loss in well-being from overemployment for males. The finding that underemployment appears to be a more serious problem for the well-being of individuals than overemployment is in contrast to the analysis of Wooden et al. (2009) who provide evidence that overemployment is more detrimental than underemployment. Yet, our results may reflect the development on the German labor market in recent decades, where policies aiming at reconciling family and work life have become more important. Measures to achieve this goal focus mostly on a reduction of working time (EIRO, 2006). The increased prevalence of part-time work, job-sharing schemes, etc. made it easier in the first place to reduce working time. Unions also promoted a reduction in working hours, especially in the 1980s and 1990s. The collectively agreed normal weekly working hours in Germany are relatively low in European comparison (EIRO, 2008). As a consequence of these developments, it may be easier to reduce working time whereas a worker who wants to increase his or her labor supply is more likely to encounter institutional obstacles. The finding that overemployment is associated with only small losses in well-being implies that the potential utility gain from a reduction of overemployment is also small. The worker will reduce his working time only if the utility gain from the adjustment is greater than the associated costs. Such costs may be of a considerable magnitude because adjustment is often achieved through changing jobs (see our discussion in Section 3.1). As a consequence, an overemployed worker's push for shorter hours is expected to be rather weak. This may explain the relatively large fraction of overemployed workers in the data. Next, we inspect the relationship between well-being and desired hours of work. The null hypothesis of an F-test on the joint significance of all coefficients on the desired hour categories cannot be rejected at the 5% level for both genders (females: p = 0.10, males: p = 0.34). As explained in Section 3.2, the coefficients on working time preferences are equal to those of actual hours of work in a reparametrized model. Thus, we interpret the result as evidence that working time appears to have no extra effect on well-being, once the mismatch is taken into account. This replicates prior findings in the literature, where studies show either no or a significantly positive relationship between actual working time and well-being (job or life satisfaction) once the working time mismatch is controlled for (e.g., Friedland and Price, 2003; Green and Tsitsianis, 2005). In contrast, studies that do not explicitly control for a mismatch often find that longer working hours are detrimental for well-being (e.g. Clark et al., 1996).10 This suggests that not working

This section presents the empirical evidence. Estimation results for females and males are shown in Tables 4 and 5, respectively. We use two versions of the regression model in Eq. (2): the restricted specification in column 1 omits the partner's well-being and provides evidence on the net effect of the partner's working time arrangements on wellbeing. The full specifications in columns 2 and 3 allow for caring preferences and give further insights into the mechanism underlying the spillover. We begin with the results on the role played by own mismatches

8 All econometric models assume cardinality of the response variable. Ferrer-iCarbonell and Frijters (2004) show that assuming ordinality or cardinality of satisfaction scores makes little difference to the results of regression analyses (for a detailed discussion, see Kristoffersen, 2010). 9 The estimated coefficient of log wages is 0.27. Compensating for the loss in well-being therefore requires an increase in wages by 0.44 log points. 10 Also, the epidemiological literature on the relationship between (long) working hours and individuals' mental health (Spurgeon et al., 1997; Sparks et al., 1997; Amagasa and Nakayama, 2012) or physical health (Virtanen et al., 2012) provides evidence that long working hours might be harmful.

Source: SOEP v28.

or she is overemployed. In general, life satisfaction is lowest when both partners are underemployed.

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

249

Table 2 Actual and preferred hours of work (couples with children). Preferred hours Actual hours

N0 … b4

≥4 … b12

≥12 … b20

≥20 … b28

≥28 … b36

≥36 … b44

≥44 … b52

≥52

Females N0 … b4 N4 … b12 ≥12 … b20 ≥20 … b28 ≥28 … b36 ≥36 … b44 ≥44 … b52 ≥52 Total

21 22 16 26 2 6 1 0 94

58 993 146 92 25 15 0 1 1330

20 354 1303 516 45 19 2 2 2261

17 467 747 2590 553 289 47 10 4720

3 49 128 417 879 741 165 32 2414

1 48 55 104 111 899 188 90 1496

0 2 1 5 2 14 50 30 104

0 1 0 0 3 3 6 32 45

120 1936 2396 3750 1620 1986 459 197 12,464

Males N0 … b4 ≥4 … b12 ≥12 … b20 ≥20 … b28 ≥28 … b36 ≥36 … b44 ≥44 … b52 ≥52 Total

0 0 0 2 1 9 6 3 21

0 13 2 4 2 15 7 6 49

0 2 16 3 1 4 4 3 33

1 9 6 69 36 98 35 15 269

0 13 16 28 313 1353 572 115 2410

1 27 13 30 118 4395 2252 815 7651

1 3 6 1 5 286 756 529 1587

0 1 2 1 1 16 45 378 444

3 68 61 138 477 6176 3677 1864 12,464

Total

Note: Table shows number of observations in each category. Source: SOEP v28.

time itself is a primary source of disutility, but the mismatch between preferred and actual working hours. Finally, we turn to the results for some of the socio-economic control variables. The estimation yields a positive coefficient on the hourly wages, suggesting a positive correlation between income and wellbeing, as shown, e.g., by Kahneman and Deaton (2010). For age, we find a monotonic decline for males, but a less clear-cut and imprecisely estimated gradient for females. Since our sample only includes couples of working age, this finding mirrors the first half of the U-shape in wellbeing reported in the literature (Blanchflower and Oswald, 2008; Wunder et al., 2013). The presence and number of children are significantly positive in the regression for females. In particular, the presence of children younger than three contributes to their well-being. Similar evidence for Germany was reported by Schwarze and Härpfer (2007), though the international literature on the effects of having children for well-being reports mixed results (Dolan et al., 2008). For both genders, disability is detrimental for well-being, though the estimate is only half as large for males as for females. Although the literature points to some adaptation over time, disability is generally associated with large drops in happiness (Lucas, 2007).

5.2. Spillovers This subsection presents the empirical evidence on the role of the partner's working time mismatch for the well-being of the other Table 3 Average life satisfaction of females by own and partner's mismatch status. Partner j Individual i

Matched

Underemployed

Overemployed

Total

Females Matched Underemployed Overemployed Total

7.50 7.22 7.39 7.40

7.30 7.00 7.12 7.15

7.50 7.19 7.27 7.32

7.46 7.16 7.27 7.31

Males Matched Underemployed Overemployed Total

7.55 7.24 7.45 7.44

7.36 7.06 7.12 7.17

7.53 7.17 7.31 7.35

7.52 7.17 7.32 7.35

partner. Tables 4 and 5 present the estimation results for females and males, respectively. We begin with the results for a restricted model that excludes the partner's well-being (column 1). After that, we compare the results to the unrestricted model that allows for caring preferences (columns 2 and 3). The partner's underemployment is negatively associated with wellbeing. For both females and males, we observe a loss in well-being of 0.01 life satisfaction points for each weekly hour their partner wants to work longer hours, but is not able to do so (Tables 4 and 5, column 1). In contrast, the coefficient of overemployment of the (female or male) partner is statistically insignificant. This suggests that the partner's overemployment is not important for individuals' well-being. One important implication emerging from these findings is that a situation in which both partners are underemployed is associated to a severe loss in well-being. For females, the loss from a double mismatch, in which both partners are underemployed, is equivalent to approximately 40% of the loss from disability. For males, the loss from a double mismatch is fully equivalent to the utility loss from disability.11 Turning to the results for the full model that includes the partner's life satisfaction as an additional explanatory variable allows deeper insights into the transmission channels that connect the partner's working time arrangements and the other partner's well-being. As outlined above, the transmission may act through contextual factors of the shared environment and/or caring preferences. The estimation results of a fixed effects approach are given in Tables 4 and 5, column 2, for females and males, respectively. We find statistically significant coefficients on the partner's well-being, amounting to approximately 0.3 for both men and women. At the same time, the absolute value of the coefficient of the partner's working time mismatch variables is considerably smaller than in the regressions omitting the partner's well-being. For females, the coefficient on the partner's underemployment decreases by approximately two thirds of its initial value and is no longer statistically significant. The parameter of the partner's overemployment does not change and remains statistically insignificant near zero. Similarly for males, the coefficients of their partner's working time mismatch variables also decrease considerably 11 The calculations used these values: underemployed women (men), on average, want to work 8.2 (6.5) additional hours per week. Using the regression results in column 1 of Tables 4 and 5, the total loss is calculated to be 0.16 and 0.19 points on the 11-point well-being scale for females and males, respectively.

250

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

Table 4 Estimation results: females.

Table 5 Estimation results: males.

Variable

(1)

(2)

(3)

Variable

(1)

(2)

(3)

Desired hours GE 1 LE 15

0.045 (0.039) 0.013 (0.044) 0.265** (0.120) −0.012*** (0.003) −0.006** (0.003) −0.006 (0.004) −0.020 (0.060) 0.021 (0.040) 0.008 (0.041) −0.066 (0.087) −0.189 (0.163) −0.009** (0.004) 0.002 (0.002) 0.004 (0.007) 0.008 (0.052) –

0.061 (0.037) −0.013 (0.042) 0.282** (0.115) −0.009*** (0.003) −0.006** (0.003) −0.006 (0.004) 0.010 (0.057) 0.031 (0.039) −0.007 (0.040) −0.107 (0.083) −0.184 (0.156) −0.003 (0.004) 0.002 (0.002) 0.002 (0.007) −0.074 (0.050) 0.301*** (0.010) 0.055* (0.032) 0.020 (0.042) −0.088 (0.063) −0.201** (0.089) −0.216 (0.132) −0.524 (0.352) 0.046* (0.027) 0.121** (0.055) 0.057 (0.035) 0.092*** (0.031) −0.022 (0.167) −0.357*** (0.117)

0.064 (0.039) −0.019 (0.046) 0.294** (0.115) −0.008** (0.003) −0.005* (0.003) −0.004 (0.003) 0.013 (0.061) 0.035 (0.039) −0.005 (0.041) −0.113 (0.088) −0.186 (0.156) −0.002 (0.006) 0.002 (0.002) 0.002 (0.007) −0.086 (0.077) 0.374* (0.208) 0.055 (0.034) 0.033 (0.043) −0.056 (0.066) −0.l62* (0.092) −0.174 (0.134) −0.473 (0.366) 0.041 (0.029) 0.117* (0.061) 0.051 (0.035) 0.101*** (0.034) −0.029 (0.177) −0.326*** (0.117)

Desired hours GE 1 LE 30

−0.083 (0.058) −0.030 (0.039) 0.050 (0.040) 0.137 (0.085) −0.009 (0.158) −0.018*** (0.004) −0.001 (0.002) 0.007 (0.007) −0.058 (0.038) 0.098** (0.043) −0.047 (0.117) −0.009*** (0.003) −0.003 (0.003) −0.002 (0.004) 0.049 (0.033) –

−0.079 (0.056) −0.035 (0.037) 0.046 (0.038) 0.154* (0.081) 0.045 (0.152) −0.016*** (0.004) −0.001 (0.002) 0.006 (0.006) −0.069* (0.036) 0.094** (0.041) −0.120 (0.112) −0.006** (0.003) −0.001 (0.002) 0.000 (0.003) 0.030 (0.031) 0.282*** (0.009) 0.266*** (0.049) −0.097** (0.047) −0.216*** (0.065) −0.376*** (0.086) −0.547*** (0.114) −0.605*** (0.160) 0.021 (0.026) 0.070 (0.054) −0.015 (0.034) 0.020 (0.031) 0.111 (0.219) −0.156* (0.088)

−0.077 (0.055) −0.033 (0.037) 0.045 (0.038) 0.156* (0.081) 0.057 (0.153) −0.015*** (0.004) −0.001 (0.002) 0.006 (0.006) −0.076** (0.036) 0.093** (0.041) −0.134 (0.117) −0.004 (0.003) 0.000 (0.003) 0.002 (0.003) 0.029 (0.033) 0.353** (0.138) 0.271*** (0.049) −0.071 (0.051) −0.178** (0.070) −0.337*** (0.088) −0.493*** (0.122) −0.547*** (0.164) 0.015 (0.027) 0.060 (0.058) −0.023 (0.035) 0.017 (0.032) 0.150 (0.229) −0.138 (0.089)

Desired hours GT 30 LE 40 DesiredhoursGT40 Underemployment (in hours) Overemployment (in hours) Time use for housework Partner: Desired hours GE 1 LE 30 Partner: Desired hours GT 30 LE 35 Partner: Desired hours GT 40 LE 50 Partner: Desired hours GT 50 LE 60 Partner: Desired hours GT 60 Partner: Underemployment (in hours) Partner: Overemployment (in hours) Partner: Time use for housework Partner: Log of hourly wage Partner: life satisfaction Log of hourly wage Age 36–40 Age 41–45 Age 46–50 Age 51–55 Age 56–60 Number of children in HH Young children (b3 years) Young children (≥3 and b6 years) Education in years Foreign citizenship Disabled

0.070** (0.034) 0.007 (0.044) −0.l26* (0.065) −0.248*** (0.093) −0.26 l* (0.138) −0.679* (0.369) 0.060** (0.028) 0.156*** (0.058) 0.058 (0.036) 0.071** (0.033) 0.063 (0.174) −0.366*** (0.122)

Desired hours GT 30 LE 35 Desired hours GT 40 LE 50 Desired hours GT 50 LE 60 Desired hours GT 60 Underemployment (in hours) Overemployment (in hours) Time use for housework Partner: Desired hours GE 1 LE 15 Partner: Desired hours GT 30 LE 40 Partner: Desired hours GT 40 Partner: Underemployment (in hours) Partner: Overemployment (in hours) Partner: Time use for housework Partner: Log of hourly wage Partner: life satisfaction Log of hourly wage Age 36–40 Age 41–45 Age 46–50 Age 51–55 Age 56–60 Number of children in HH Young children (b3 years) Young children (≥3 and b6 years) Education in years Foreign citizenship Disabled

0.269*** (0.051) −0.144*** (0.050) −0.276*** (0.068) −0.427*** (0.089) −0.643*** (0.119) −0.692*** (0.167) 0.037 (0.028) 0.115** (0.056) 0.004 (0.035) 0.005 (0.032) −0.034 (0.229) −0.197** (0.091)

Note: Life satisfaction is the dependent variable. Col. (1) and (2) are fixed effects estimations. Col. (3) is a fixed effects instrumental variable model. Instruments are variables of the interview context and characteristics of the partner. First stage estimation results and diagnostic statistics are in the online appendix. All estimations include further controls for occupation, branch of industry, self employment and the partner’s age. Standard errors in parentheses. Sample sizes: n = 2, 583. nT = 12, 464. Significance level: * b 0.1, ** b 0.05, *** b 0.01. Source: SOEP v28.

Note: Life satisfaction is the dependent variable. Col. (1) and (2) are fixed effects estimations. Col. (3) is a fixed effects instrumental variable model. Instruments are variables of the interview context and characteristics of the partner. First stage estimation results and diagnostic statistics are in the online appendix. All estimations include further controls for occupation, branch of industry, self employment and the partner's age. Standard errors are in parentheses. Sample sizes: n = 2587. nT = 12,464. Significance level: *b0.1, **b0.05, ***b0.01. Source: SOEP v28.

though the coefficient of the female partner's underemployment is still significantly different from zero. The evidence suggests that adding the partner's well-being eliminates the correlation previously captured by the partner's working time mismatch. Hence, the partner's mismatch may not directly impinge on well-being. Instead we suppose that caring preferences are relevant and contextual factors (i.e. the working time arrangements themselves) are not the primary source of disutility.

In a further analysis step, we apply an instrumental variables approach to take into account that the positive coefficient of the partner's well-being could be driven by factors affecting both partners' wellbeing simultaneously. The results from panel IV estimations are given in Tables 4 and 5, column 3, for females and males, respectively.12 12 The first step estimation results including diagnostic statistics are in the online appendix.

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

251

Here, the coefficients of the partner's mismatch variables are generally small and statistically insignificant, which is in line with the results from the fixed effects estimation above. The results further point to a statistically significant link between the partner's well-being and the individual's own well-being while controlling for unobserved individual-specific effects and a large number of context variables. Therefore, the interdependence of well-being within couples cannot be attributed to the large number of observed and unobserved factors controlled for in the regressions. We conclude that the partner's working time mismatch lost much of its “effect” for one's own well-being once the partner's well-being is taken into account. Well-being appears to be contagious so that individuals are less satisfied with their lives if their partner is dissatisfied with his or her life. This evidence is consistent with the hypothesis that persons in partnerships are altruistically linked and that the partner's hours mismatch produces a considerable spillover within couples because of caring preferences.

consumption of household public goods or for the stability of marriages in the long-run. Finally, we conclude that employees' working time preferences should be taken into account more seriously. Evidently, job packages offer too little flexibility in the choice of working hours. 2/3 of women and 3/4 of men report preferences for working hours other than their actual working time. Since a mismatch is connected with significant losses in well-being, employees and their partners have to bear extra costs resulting from inflexible working time arrangements. Not least because of the serious spillovers from partner mismatches and its consequences for the partner's well-being, we believe that introducing or enhancing greater flexibility in working hours—which may include measures such as bilateral agreements over working time, working time accounts or a working time honor system13—creates an enormous potential to increase individuals' and families' well-being.

6. Conclusion

We thank two anonymous referees, David Kiss, Tom Mroz, Rainer Winkelmann, Michael Zibrowius, participants of the Scottish Economic Society 2012 Conference, of the Austrian Economic Association Meeting 2012, and of the International German Socio-Economic Panel User Conference 2012 for helpful comments.

Issues of work–life balance are the topic of much public debate. This study analyzed how working time mismatches between actual and preferred working time affect the well-being of couples. As a contribution, we merged research on working time mismatch and research on family well-being to examine spillovers within couples. Using longitudinal German SOEP data, our evidence contributes to the so far small literature on working time mismatch, which is mainly based on cross-sectional data. The empirical evidence yields the following main findings: First, there exists a clear relationship between working time mismatch and well-being, suggesting that both females and males experience significant losses in well-being if they are constrained in their hours of work choices. In general, losses from underemployment are larger than losses from overemployment. We regard underemployment as a severe problem because individuals are deprived from utility gains resulting from monetary and non-monetary job aspects, such as the potential of developing skills and the social interaction with, for example, colleagues or customers. Moreover, underemployment reduces aggregate welfare because the work force potential of the economy is underused (e.g., Eichhorst et al., 2011). Second, considering spillovers within couples, the results show a significant relationship between one's own well-being and the partner's working time mismatch: both genders experience lower well-being when their partner is underemployed. The analysis gives rise to the conjecture that the spillover occurs through caring preferences between partners rather than through the shared context only, supporting the hypothesis of altruistic links between partners. Since altruisticallylinked partners may provide partial insurance against working time mismatches (Chami and Fischer, 1996), they may adjust their labor supply in response to the partner's mismatch to lower the repercussions for family life. Since recent evidence provided by Benjamin et al. (2012) shows that respondents' SWB predictions are a powerful predictor of their choices, we regard it as highly plausible that families adjust their labor supply due to spillovers through caring preferences. However, our evidence represents only a first indication of motives for such behavior and more research on this topic needs to be undertaken in the future. Despite of potential compensation in the form of nonmarket insurance within couples, we see the urgent need for a redistribution of working time to support families to balance work and family life. Since underemployment is more detrimental to well-being than overemployment of equal size, a redistribution of working hours from the overemployed to the underemployed might considerably increase the well-being of employees and their families (for a similar argument, see Golden, 1998). However, since the redistribution of working time may potentially also affect bargaining between partners, future research should investigate possible consequences of such redistribution for the

Acknowledgments

Appendix A. Additional results Additional results to this article can be found online at http://dx.doi. org/10.1016/j.labeco.2013.09.002. References Aguiar, M., Hurst, E., 2007. Measuring trends in leisure: the allocation of time over five decades. Q. J. Econ. 122 (3), 969–1006. Akay, A., Martinsson, P., 2009. Sundays are blue: aren't they? The day-of-the-week effect on subjective well-being and socio-economic status. IZA Discussion Papers 4563. Institute for the Study of Labor (IZA). Altonji, J.G., Paxson, C.H., 1988. Labor supply preferences, hours constraints, and hourswage trade-offs. J. Labor Econ. 6 (2), 254–276. Altonji, J.G., Paxson, C.H., 1992. Labor supply, hours constraints, and job mobility. J. Hum. Resour. 27 (2), 256–278. Altonji, J.G., Hayashi, F., Kotlikoff, L.J., 1992. Is the extended family altruistically linked? Direct tests using micro data. Am. Econ. Rev. 82 (5), 1177–1198. Amagasa, T., Nakayama, T., 2012. Relationship between long working hours and depression in two working populations: a structural equation model approach. J. Occup. Environ. Med. 54 (7), 868–874. Baum, C.F., Schaffer, M.E., Stillman, S., 2007. Enhanced routines for instrumental variables/ generalized method of moments estimation and testing. Stata J. 7 (4), 465–506. Becker, G.S., 1974. A theory of social interactions. J. Polit. Econ. 82 (6), 1063–1093. Becker, G.S., 1981. Altruism in the family and selfishness in the market place. Economica 48 (189), 1–15. Benjamin, D.J., Kimball, M.S., Heffetz, O., Rees-Jones, A., 2012. What do you think would make you happier? What do you think you would choose? Am. Econ. Rev. 102 (5), 2083. Blanchflower, D.G., Oswald, A.J., 2008. Is well-being U-shaped over the life cycle? Soc. Sci. Med. 66 (8), 1733–1749. BMFSFJ, 2012. Zeit für Familie — Familienzeitpolitik als Chance einer nachhaltigen Familienpolitik. Technical report. Bundesministerium für Familie, Senioren, Frauen und Jugend (http://www.bmfsfj.de/BMFSFJ/Service/Publikationen/publikationsliste, did=186954.html). Bolger, N., DeLongis, A., Kessler, R.C., Wethington, E., 1989. The contagion of stress across multiple roles. J. Marriage Fam. 51 (1), 175–183. Booth, A.L., Van Ours, J.C., 2008. Job satisfaction and family happiness: the part-time work puzzle. Econ. J. 118 (526), F77–F99. Booth, A.L., Van Ours, J.C., 2009. Hours of work and gender identity: does part-time work make the family happier? Economica 76 (301), 176–196. Bruhin, A., Winkelmann, R., 2009. Happiness functions with preference interdependence and heterogeneity: the case of altruism within the family. J. Popul. Econ. 22 (4), 1063–1080. Chami, R., Fischer, J.H., 1996. Altruism, matching, and nonmarket insurance. Econ. Inq. 34 (4), 630–647. Clark, A.E., 2003. Unemployment as a social norm: psychological evidence from panel data. J. Labor Econ. 21 (2), 289–322. Clark, A.E., 2006. A note on unhappiness and unemployment duration. Appl. Econ. Q. 52 (4), 291–308. 13

For further policy recommendations, see e.g., Eichhorst et al. (2011).

252

C. Wunder, G. Heineck / Labour Economics 24 (2013) 244–252

Clark, A.E., Senik, C., 2010. Who compares to whom? The anatomy of income comparisons in Europe. Econ. J. 120 (544), 573–594. Clark, A., Oswald, A., Warr, P., 1996. Is job satisfaction U-shaped in age? J. Occup. Organ. Psychol. 96 (1), 57–81. Clark, A., Kristensen, N., Westergå̊rd-Nielsen, N., 2009. Job satisfaction and co-worker wages: status or signal? Econ. J. 119 (536), 430–447. Conti, G., Pudney, S., 2008. If you're happy and you know it, clap your hands! Survey design and the analysis of satisfaction. ISER Working Paper Series 2008-39. Institute for Social and Economic Research. Dolan, P., Peasgood, T., White, M., 2008. Do we really know what makes us happy? A review of the economic literature on the factors associated with subjective well-being. J. Econ. Psychol. 29 (1), 94–122. Eichhorst, W., Marx, P., Tobsch, V., 2011. Familienfreundliche flexible Arbeitszeiten — ein Baustein zur Bewältigung des Fachkräftemangels. IZA Research Reports, 33. Institute for the Study of Labor (IZA). Ermisch, J.F., 2003. An Economic Analysis of the Family. Princeton University Press, Princeton, New Jersey. European Industrial Relations Observatory (EIRO), 2006. Reconciliation of Work and Family Life and Collective Bargaining in the European Union. An Analysis of EIRO Articles. http://www.eurofound.europa.eu/eiro/other_reports/work_family_life.pdf. European Industrial Relations Observatory (EIRO), 2008. Working Time in the EU and Other Global Economie — Industrial Relations in the EU and Other Global Economies 2006–2007. http://www.eurofound.europa.eu/docs/eiro/tn0804058s/tn0804058s.pdf. Ferrer-i-Carbonell, A., Frijters, P., 2004. How important is methodology for the estimates of the determinants of happiness? Econ. J. 114 (497), 641–659. Folbre, N., Nelson, J.A., 2000. For love or money–or both? J. Econ. Perspect. 14 (4), 123–140. Frey, B.S., Stutzer, A., 2002. What can economists learn from happiness research? J. Econ. Lit. 40 (2), 402–435. Friedland, D., Price, R., 2003. Underemployment: consequences for the health and wellbeing of workers. Am. J. Community Psychol. 32, 33–45. Golden, L., 1998. Working time and the impact of policy institutions: reforming the overtime hours law and regulation. Rev. Soc. Econ. 4, 522–541. Goldin, C., 2006. The quiet revolution that transformed women's employment, education, and family. Am. Econ. Rev. 96 (2), 1–21. Green, F., Tsitsianis, N., 2005. An investigation of national trends in job satisfaction in Britain and Germany. Br. J. Ind. Relat. 3, 401–429. Grözinger, G., Matiaske, W., Tobsch, V., 2008. Arbeitszeitwünsche, Arbeitslosigkeit und Arbeitszeitpolitik. SOEP papers 103. DIW Berlin, The German Socio-Economic Panel (SOEP). Helliwell, J.F., Wang, S., 2011. Weekends and subjective well-being. NBER Working Papers 17180. National Bureau of Economic Research, Inc. Kahneman, D., Deaton, A., 2010. High income improves evaluation of life but not emotional well-being. Proc. Natl. Acad. Sci. 38, 16489–16493. Kassenboehmer, S.C., Haisken-DeNew, J.P., 2009. You're fired! The causal negative effect of entry unemployment on life satisfaction. Econ. J. 119 (536), 448–462. Kind, M., Haisken-DeNew, J.P., 2012. Unexpected victims: how parents' unemployment affects their children's life satisfaction. Working Paper 2/12. Melbourne Institute. Kluve, J., Tamm, M., 2013. Parental leave regulations, mothers' labor force attachment and fathers' childcare involvement: evidence from a natural experiment. J. Popul. Econ. 26 (3), 983–1005. Konrad, K.A., Lommerud, K.E., 1995. Family policy with non-cooperative families. Scand. J. Econ. 581–601. Kristoffersen, I., 2010. The metrics of subjective wellbeing: cardinality, neutrality and additivity*. Econ. Rec. 86 (272), 98–123.

Layard, R., 2005. Happiness: Lessons from a New Science. Allen Lane, London. Lee, S., McCann, D., Messenger, J.C., 2007. Working Time Around the World: Trends in Working Hours, Laws, and Policies in a Global Comparative Perspective. Routledge, London. Lucas, R.E., 2007. Long-term disability is associated with lasting changes in subjective well-being: evidence from two nationally representative longitudinal studies. J. Personal. Soc. Psychol. 92 (4), 717–730. Lundberg, S., Pollak, R.A., 1996. Bargaining and distribution in marriage. J. Econ. Perspect. 10 (4), 139–158. Powdthavee, N., 2009. I can't smile without you: spousal correlation in life satisfaction. J. Econ. Psychol. 30 (4), 675–689. Powdthavee, N., Vignoles, A., 2008. Mental health of parents and life satisfaction of children: a within-family analysis of intergenerational transmission of well-being. Soc. Indic. Res. 88, 397–422. Schimmack, U., Lucas, R., 2010. Environmental influences on well-being: a dyadic latent panel analysis of spousal similarity. Soc. Indic. Res. 1–21. Schwarze, J., Härpfer, M., 2007. Are people inequality averse, and do they prefer redistribution by the state? Evidence from German longitudinal data on life satisfaction. J. Socio-Econ. 36 (2), 233–249. Schwarze, J., Winkelmann, R., 2011. Happiness and altruism within the extended family. J. Popul. Econ. 24 (3), 1033–1051. Shields, M.A., Price, S.W., 2005. Exploring the economic and social determinants of psychological well-being and perceived social support in England. J. R. Stat. Soc. Ser. A (Stat. Soc.) 3, 513–537. Sparks, K., Cooper, C., Fried, Y., Shirom, A., 1997. The effects of hours of work on health: a meta-analytic review. J. Occup. Organ. Psychol. 70 (4), 391–408. Spurgeon, A., Harrington, J.M., Cooper, C.L., 1997. Health and safety problems associated with long working hours: a review of the current position. Occup. Environ. Med. 6, 367–375. Stock, J.H., Yogo, M., 2002. Testing for weak instruments in linear IV regression. NBER Technical Working Papers 0284. National Bureau of Economic Research, Inc., Cambridge, Massachusetts. Tomes, N., 1986. Income distribution, happiness and satisfaction: a direct test of the interdependent preferences model. J. Econ. Psychol. 4, 425–446. Van Praag, B.M.S., Ferrer-i-Carbonell, A., 2008. Happiness Quantified. A Satisfaction Calculus Approach. Oxford University Press, Oxford. Virtanen, M., Heikkilä, K., Jokela, M., Ferrie, J.E., Batty, G.D., Vahtera, J., Kivimäki, M., 2012. Long working hours and coronary heart disease: a systematic review and metaanalysis. Am. J. Epidemiol. 7, 586–596. Wagner, G.G., Frick, J.R., Schupp, J., 2007. The German Socio-Economic Panel Study (SOEP) — scope, evolution and enhancements. Schmollers Jahrbuch J. Appl. Soc. Sci. Stud. 127 (1), 139–169. Wilkins, R., 2007. The consequences of underemployment for the underemployed. J. Ind. Relat. 49 (2), 247–275. Winkelmann, R., 2005. Subjective well-being and the family: results from an ordered probit model with multiple random effects. Empir. Econ. 3, 749–761. Winkelmann, L., Winkelmann, R., 1995. Happiness and unemployment: a panel data analysis for Germany. Appl. Econ. Q. (4), 293–307. Winkelmann, L., Winkelmann, R., 1998. Why are the unemployed so unhappy? Evidence from panel data. Economica 65 (257), 1–15. Wooden, M., Warren, D., Drago, R., 2009. Working time mismatch and subjective wellbeing. Br. J. Ind. Relat. 47 (1), 147–179. Wunder, C., Wiencierz, A., Schwarze, J., Küchenhoff, H., 2013. Well-being over the life span: semiparametric evidence from British and German longitudinal data. Rev. Econ. Stat. 95 (1), 154–167.