The effect of involuntary maternal job loss on children's behaviour and non-cognitive skills

Peter, Frauke Article — Accepted Manuscript (Postprint) The effect of involuntary maternal job loss on children's behaviour and non-cognitive skills Labour Economics Provided in Cooperation with: German Institute for Economic Research (DIW Berlin) Suggested Citation: Peter, Frauke (2016) : The effect of involuntary maternal job loss on children's behaviour and non-cognitive skills, Labour Economics, ISSN 0927-5371, Elsevier, Amsterdam, Vol. 42, pp. 43-63, https://doi.org/10.1016/j.labeco.2016.06.013 This Version is available at: https://hdl.handle.net/10419/204475 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ The Effect of Involuntary Maternal Job Loss on Children’s Behaviour and Non-cognitive Skills $ June 9, 2016 Frauke Peter DIW Berlin, 10108 Berlin, Germany Abstract This paper uses propensity score methods to analyse the effect of involuntary maternal job loss on children’s non-cognitive skills. My analyses are based on a rich and nationwide random sample, the German Socio-Economic Panel Study (SOEP) that includes information about maternal job loss and child behaviour and non-cognitive skills, in addition to a rich conditioning set. The results show that maternal job loss increases preschool children’s socio-behavioural problems and decreases adolescents’ belief in self-determination. Keywords: Child development, Maternal Job Loss, Non-cognitive Skills, Propensity Score Methods JEL: J13, J63, J65 $I thank the editor and two anonymous referees of this journal for very helpful feedback and suggestions to improve the manuscript. In addition, valuable comments by Nabanita Datta-Gupta, Philipp Oreopoulos and Mari Rege are gratefully acknowledged. Fruitful discussions with participants at the 17th annual meetings of the Society Labor Economists in Chicago and at the 15th IZA European Summer School in Labor Economics in Buch am Ammersee are also appreciated. I thank my colleagues C. Katharina Spiess and Jan Marcus at DIW Berlin and Julia Horstschräer at ZEW for their stimulating impulses as well as Adam Lederer for helpful editorial assistance. Funding from the German Federal Ministry of Education and Research within the framework of the Program for the Promotion of Empirical Educational Research (reference number: 01 JG 0910) is gratefully acknowledged. The usual disclaimer applies. Email address: [email protected] (Frauke Peter) This is the postprint of an article published in Labour Economics 42 (2016), p. 43-6, available online at: https://doi.org/10.1016/j.labeco.2016.06.013 © <2019>. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 1. Introduction Job loss not only leads to a considerable fall in income, it also affects the health and wellbeing of individuals; possibly even leading to divorce (see for example the studies by Charles and Stephens, 2004; Eliason and Storrie, 2009; Marcus, 2014; Rege et al., 2009). Studies also show that it has spillover effects on partners (Marcus, 2013) and children (Huff-Stevens and Schaller, 2011; Kalil and Ziol-Guest, 2005, 2008; Lindo, 2011; Oreopoulos et al., 2008; Rege et al., 2011). This paper contributes to the literature on spillover effects of job loss by analysing the impact of involuntary maternal job loss on children’s behaviour and non-cognitive skills. The paper also contributes to our understanding of factors that impact children’s development of behaviour and non-cognitive skills (Cunha and Heckman, 2007; Cunha et al., 2006, 2010). This paper focuses on non-cognitive skills during preschool ages and during adolescence, as an increasing number of economic studies suggest that non-cognitive skills are important predictors of later educational achievements, health outcomes, and labour market success (Blanden et al., 2007; Carneiro et al., 2007; Cobb-Clark and Schurer, 2013; Currie and Stabile, 2006; Heckman et al., 2013; Prevoo and ter Weel, 2015). Studies show that non-cognitive skills impact cognitive skills, but not vice versa, and that these non-cognitive skills are as important as cognitive skills regarding school performance (Cunha and Heckman, 2007; Heckman et al., 2006). However, less is known about how non-cognitive skills develop if a negative shock occurs to the family environment. Yet, this is particularly relevant, since the family is likely even more important than schools or other institutions for the development of skills (Carneiro and Heckman, 2003). The few existing studies on parental job loss and child outcomes look at children’s academic performance, likelihood of grade repetition, and health, or consider earnings for adult children (Huff-Stevens and Schaller, 2011; Kalil and Ziol-Guest, 2005, 2008; Lindo, 2011; Oreopoulos et al., 2008; Rege et al., 2011). It is plausible that effects of maternal job loss vary with child age and my analyses therefore exploit both a preschool (child age 5/6) and an adolescence/early adulthood (age 17) sample. The preschool sample allows for analyses of child behaviour as measured by the socio-emotional behaviour based on a modified version of the Strength and Difficulties Questionnaire (SDQ) developed by Goodman (1997), which assesses children’s socio-emotional regulation. The adolescence sample, on the other hand, includes information about Locus of Control, which is based on a concept developed by Rotter (1966) and describes to what extent a person believes in self determination or fate. Both ages mark important phases: Age five/six marks the transition to school in Germany while age seventeen is that time at which students make decisions about further education. It is ex ante unclear how maternal job loss affects the well-being of the children of the household. There may be negative as well as positive effects; parental stress caused by the job loss may transmit to the children just as an income loss could lead to deterioration of the family environment. Involuntary maternal job loss may increase the amount of time the mother spends 1 with her children and depending on the quality of that interaction, this may lead to both worse or improved child outcomes. My analyses consider the effects of job loss stemming from plant closures and dismissals. My main analyses combine these two types of job loss into one measure of job loss but robustness analyses acknowledge that their impact on families may vary. Results are based on propensity score matching while drawing on a rich data set informative about the characteristics of families in which women experience job loss and child outcomes. The data used come from the German Socio-Economic Panel Study (SOEP). The SOEP contains a particularly rich set of variables on non-cognitive skills, as well as an extensive set of variables regarding household, parental and child characteristics that is well suited for propensity score methods and its requirement of selection on observables. In addition, with SOEP data, specific mechanisms behind the effect of maternal job loss on children’s behaviour and non-cognitive skills can be examined. The SOEP data comprise information on life satisfaction and household income, as well as on personality traits, which facilitates examining possible mediators of the effect of maternal job loss on children’s skills. This study shows that involuntary maternal job loss negatively affects the non-cognitive skills of children. A mother’s involuntary job loss increases a preschool child’s socio-emotional problems by 51% of a standard deviation and decreases adolescents’ internal locus of control by 26% of a standard deviation. The results also show that job loss decreases the life satisfaction of mothers with preschool children and significantly affects the household income in the adolescence sample. Estimating the effect of maternal job loss on child outcomes including these potential channels decreases the size and the significance of the effect on preschool children’s socioemotional behaviour but not on adolescents’ internal locus of control. The remainder of the paper is structured as follows: Section 2 discusses the related literature and theoretical links of maternal job loss and children’s non-cognitive skills. Section 3 outlines the empirical strategy and in Section 4 the data set is described. In Section 5 the estimation results are discussed. Section 6 presents some robustness tests before Section 7 concludes. 2. Linking maternal job loss and children’s non-cognitive skills As discussed above, the direction of the effects of maternal job loss on child behaviour and non-cognitive skills is not obvious. The skill formation framework proposed by Cunha and Heckman (2007) suggests that children’s skills are produced with parental inputs of time and goods. There are likely dynamic complementarities in skills, i.e. skills acquired in one period depend on those of previous periods, which again are dependent on home and school inputs as well as parental ability (Todd and Wolpin, 2007). Throughout the skill formation process, timing is an important aspect, as inputs impact human capital production differently across childhood stages. In such a framework, a household with a working mother would have less time available to distribute between child and employment than would a household with a non-working mother. 2 Mothers who work full-time or part-time compared to mothers who do not work have different possibilities to divide their time. This production function may, thus, be affected by maternal job loss, as such a disruption in the household could be the source of stress, which may impede child development. The following potential channels of maternal job loss are discussed in this paper: A change in life satisfaction, household income, personality traits, and quality of time. The literature concerned with the relationship between life satisfaction and unemployment (see for example the work by Clark et al. (2010) or Knabe et al. (2010)) finds that job loss (as well as the unemployment level in general) is likely to decrease life satisfaction. This may, in turn, deteriorate the relationship between the mother and her child. An income loss after job loss might be another potential source of instability at home. Tension due to decreased financial resources between parents could spread to their children, leading to an unstable temper of children affecting their relationship with their peers. In addition, maternal job loss also reduces future family income and might impact on children’s development through a reduction in financial resources (e.g. Baum, 2003; Rege et al., 2011). But, because German mothers are often second earners, their job loss might impact the financial situation of the household less than if the father had lost his job. Further, maternal income loss can often be compensated with her partner’s income, assuming dual income; if a single parent loses her job, the effect will be much greater. In addition to their partners’ income, individuals receive generous unemployment benefits in Germany when compared to the US, but not to Scandinavian countries. For example, an unemployed person who has a child receives 67% of their previous income as unemployment insurance (UI).1Furthermore, they are entitled to tax-financed unemployment assistance after UI payments expire or if they fall below a certain threshold. Thus, an income loss as potential mediator of maternal job loss might be less likely. It is argued that mothers who lose their job may be stressed for reasons other than income reduction. Job loss might also lead to direct changes in children’s non-cognitive skills, as it changes maternal characteristics measured by personality traits such as neuroticism (or emotional stability if coded reversely) and internal locus of control. A mother may, for example, regard job loss as something that has happened to her due to others. Thus, adolescents could believe that fate or actions of others influence success in life. Furthermore, job loss may also change maternal time and its allocation towards children’s non-cognitive skills. A change in maternal time might mediate a positive effect of mothers’ displacement, as mothers might spend more time with their children promoting development. For example, a Norwegian study by Rege et al. (2011) finds an insignificant but positive effect of mothers’ displacement due to plant closure on children’s grade point average at sixteen years of age (3% of a standard deviation). Thus, mothers spending increased time with their children 1Length of payment depends on the duration of their own contribution to UI (e.g. Caliendo et al., 2013). 3 seems to have a small but positive effect. Additionally increased maternal time combined with income loss might reduce a child’s time spent in day care. Yet, in Germany day care is not costly (fees are based on income) and therefore it is less likely that children will drop out of day care after maternal job loss.2Thus, children remain in contact with peers from outside the family, which may help in their development.3 3. Empirical strategy The goal of this paper is to identify an impact of involuntary maternal job loss on children’s non-cognitive skills. Equation 1 summarizes the linear relationship of maternal job loss (treatment) on children’s non-cognitive skills in the preschool sample if Si=SEBiand in the adolescence sample if Si=LOCi.4Sicomprises non-cognitive outcome of child i,JOBLiis a variable capturing involuntary job loss, Xiis the vector of conditioning variables and υiis an error term. Si=αi+βiJOBLi+γiXi+υi(1) The treatment variable is defined as a binary measure that equals 1 if a mother experiences an involuntary job loss and 0 if the mother does not. I classify involuntary job loss as plant closure or dismissal by employer. Since the study looks only at involuntary job loss, other types of job loss are not considered and treated mothers are compared to mothers who do not experience job loss during the observation period. Focusing on plant closure and dismissal by employer ensures that job ends of mothers are independent of any change in children’s development before job loss. An effect of maternal job loss on child outcome is identified by using ordinary least squares (OLS) if the “selection on observables” assumption is satisfied (see Heckman, 1979). This means that all variables related to both job loss and children’s non-cognitive skills have to be included in the estimations. Furthermore, for OLS to render consistent estimates the relationship between maternal job loss and children’s non-cognitive skills must be linear, an assumption that cannot be clearly verified. This paper, therefore, uses propensity score methods in addition to OLS to estimate the effect of involuntary maternal job loss on children’s non-cognitive outcomes (βi). Propensity score matching deals with the missing counterfactual problem, as it finds a nearly identical “twin” of each child whose mother experiences a job loss using one or more children whose mothers do not. 2Around 98% of children between ages three and six attend day care (Statistisches Bundesamt, 2014). 3Empirical studies find neutral or positive effects of day care attendance on children’s cognitive and noncognitive skills, especially for children from disadvantaged families (e.g. Apps et al., 2013; Datta Gupta and Simonsen, 2010, 2012; Felfe and Lalive, 2013; Goodman and Sianesi, 2005; Loeb et al., 2007). 4Where SEB is the abbreviation for children’s socio-emotional behaviour and LOC for adolescents’ internal locus of control. 4 The treatment and control group are matched based on the estimated propensity score (P(X)). Before identifying the average treatment effect on the treated (ATT), all observations that do not comply with the common support condition are discarded from the sample.5Hence the sample used for examining involuntary maternal job loss consists only of those treated mothers who have a matched untreated mother based on the same characteristics set X.This paper uses kernel matching to match treatment and control group observations.6Moreover, this paper utilizes a regression-adjusted matching approach as the preferred model specification, which requires controlling for all conditioning variables in the post-matching estimations (see Stuart, 2010). The regression-adjustment method avoids further potential bias if matching is not exact. Equation 2 shows the estimation of the ATT using a regression-adjusted matching approach, where a matching-specific weight Wk,l obtained from kernel matching, is used in the analysis. ATT = k∈T Wk(Y1k−xkˆ β)− l∈C Wk,l(Y0l−xlˆ β)(2) In Equation 2, the symbols T and C stand for treatment group and control group respectively. Wk,l represents a matching-specific weight that is the weight placed on individual lto be comparable to individual k.7The weight Wk,l includes values obtained from kernel matching for the control group of each treated k: Wk,l =G(Pk−Pl bn) h∈CG(Pk−Pl bn)(3) where G(.) is a kernel function, e.g. Gaussian or Epanechnikov, and bnis a bandwidth parameter.8 Since propensity score methods require that selection is based on observable characteristics, unobservables are not addressed. This paper assumes that in the absence of maternal job loss the non-cognitive skills of the treated children and the matched control children would be the same. If this assumption is violated, meaning that treated children differ systematically from children of the control group in terms of unobservable characteristics, the model suffers from endogeneity. For example, children’s non-cognitive skills are likely to be correlated with maternal non-cognitive skills, which, in turn, may be affected by an involuntary job loss. Therefore, this paper first examines whether maternal personality traits are correlated with maternal job loss 5Figure A.1 and A.2 in the appendix show histograms of the propensity score by treatment status in the preschool and adolescence sample respectively and depict the obtained overlap of treatment and control groups. 6Matching is implemented in Stata 11 using the program psmatch2 provided by Leuven and Sianesi (2003). 7Wkequals one in this estimation of the ATT. 8Kernel matching is implemented with an Epanechnikov kernel function and a bandwidth parameter of 0.06 in this paper. 5 and secondly adds these traits as control variables to the preferred model to address “selection on unobservables”. 4. Data Using data from the German Socio-Economic Panel Study (SOEP), the analysis is based on a nationwide random sample and a very rich data set. The SOEP started in 1984 and is an annual household panel9that incorporates a series of mother-child questionnaires as well as a youthspecific questionnaire. Both the child-specific modules and the youth-specific module contain detailed information on children, i.e., non-cognitive skills, birth weight, child care usage, school attendance, and grade repetition among others. In addition, the SOEP has rich information on individual characteristics of the children’s mothers as well as on family characteristics. The SOEP accumulates information on current household compositions as well as on past formations. Based on this vast data set, the probability of involuntary maternal job loss is estimated. 4.1. Sample Since the SOEP includes excellent measures on non-cognitive skill formation for different childhood stages, this paper uses two samples of children - preschoolers and adolescents - to examine the underlying question of this paper. At both stages – either at age five/six or at age 17 – German children face imminent and important transitions: Preschool children start with primary school and adolescents decide upon further education. The preschool sample consists of children aged five/six in the SOEP whose mothers answered the mother-child questionnaire and were 20 years or older when giving birth. I restrict maternal age at birth to twenty years or older, because in Germany a person attains full age at eighteen years of age and apprenticeships as well as the university school track end on average around age twenty, and thus mark a potential labour market entry. Children have to have non-missing information10 on the measured non-cognitive skill and their mothers participated in the survey prior to 2003, with no missing information prior to childbirth. For the implementation of the propensity score methods, a point in time at which mother’s are observed to lose their jobs is determined. Since mothers are entitled to three years of parental leave in Germany, maternal working status is assessed after a child’s third birthday. In period t>3 when children are three years or older, it is observed whether mothers are working and thus may lose their job. A detailed discussion of variables used for modelling the selection decision into treatment is given in Section 4.4.. 9A general overview of the SOEP is given by Wagner et al. (2007), with Schupp et al. (2008) and Siedler et al. (2009) describing the mother-child questionnaires used in this paper. Frick and Lohmann (2010) document the youth questionnaire. 10For preschool children, 1% of the initial sample has missing information on items measuring socio-emotional behaviour. 6 The second sample examined in this study depends on the youth sample of the SOEP, which comprises information of children aged 17 at the time of the survey. This adolescence sample is restricted to children born between 1984 and 1993, living with their parents, and having no missing non-cognitive skill information.11 Moreover adolescents are included in the adolescence sample if their mothers were 20 years or older at childbirth, and have reported their employment status during early childhood. Unlike in the preschool sample, maternal employment patterns prior childbirth cannot be observed for all birth cohorts, since the household panel only started in 1984; and because many households entered the SOEP in 2000. For these children nearly no information prior to 2000 is included in the SOEP. Thus, I choose another cut-off date as in the preschool sample to predict mothers’ propensity of job loss. In the adolescence sample I use age ten as the cut-off point from which mothers are observed to lose their jobs. I do this for three reasons: first, mothers of these birth cohorts were more likely to return to work while children were in secondary school: In 2008, for example, 59% of mothers with children below the age of six were employed compared to 70% of mothers with children age ten or older (Rübennach, 2010). Second, some mothers earlier working information coincides with the German reunification and its transition years of economic and constitutional merger in 1990/91. A third reason for diverting from the cut-off date used for preschool children is related to children’s school careers in Germany. From age ten onward most children move from primary to secondary school.12 If an earlier cut-off date were used, it would result in an even longer time span during which involuntary job loss would occur leading to spurious results from other events. In sum, in period t>10 when children are ten years of age or older, maternal job loss is observed. In order to compare the relationship of maternal job loss and all outcome measures utilized in this paper, I further restrict both samples to observations that have non-missing information for all maternal outcomes.13 4.2. Treatment and control group The treatment and control groups consist of mothers who are working at age three of their child in the preschool school sample and at age ten for the adolescence sample. As described above, this study includes children with valid non-cognitive skill information and whose mothers participated in the survey preand post-treatment. 11Very few observations (2.3% of the initial sample) are dropped due to missing information on items of the locus of control measure. 12In three federal states (Berlin, Brandenburg, and Mecklenburg-Western Pomerania) children transit from primary to secondary school following the completion of grade six, i.e. from age twelve onward. Thus, I also address age twelve as cut-off date in a robustness check in Section 6. The result remains similar to the estimate obtained with the cut-off at age ten. 13The final sample size comprises 229 observations in the preschool sample and 522 observations in the adolescence sample. 7 5. Results In this section the estimates of the impact of involuntary maternal job loss on children’s behavioural and non-cognitive outcomes are presented using OLS and propensity score methods. First, results from stepwise regressions are presented. Then the preferred OLS specification is compared to estimates of regression-adjusted propensity score matching. In the preferred model, estimates are obtained by controlling for all conditioning variables measured pre-treatment. In all tables only the coefficient of the explanatory variable of interest is depicted: involuntary job loss or plant closure and dismissal by employer as separate variables.19 5.1. Preschool sample Table 2 presents the results of the correlation of involuntary job loss with children’s socioemotional behaviour. Involuntary job loss is significantly associated with larger socio-emotional problems (SDQ). A child’s SDQ is estimated to increase by 51% of a standard deviation, if her mother loses a job due to plant closure or dismissal, implying that children are less stable with respect to socio-emotional behaviour (see column 6). Comparing the coefficient of the “full” OLS model (column 6) to the “raw” model (column 1) suggests that estimating the association of involuntary job loss on children’s SDQ is robust to including pre-job loss characteristics. A comparison of the size of the association of SDQ and involuntary job loss as opposed to child’s gender indicates that the effect of job loss is more than twice the size of child’s gender associated with children’s socio-emotional behaviour.20 For girls, the SDQ is 25% of a standard deviation lower than that of boys. Adding two variables to the model estimated in column 6, namely maternal working status at age five/six of the child and regional unemployment rate, renders a similar relationship in terms of size and significance of the estimate (column 7). Controlling for these post-treatment measures approximates mothers’ length of unemployment and its result suggests that the experience of involuntary job loss may last longer than to the point where mothers return to work. As a next step, state and industry fixed effects are added to the model (column 8) to assess whether job loss occurs more frequently in areas with low quality education, which may actually underlie child outcomes. The results are robust and similar in size to whether industry and state fixed effects are included or not, yet, the significance level of maternal job loss decreases, as standard errors increase. Including state and industry fixed effects does not suggest that plant closure or dismissal are more likely to occur in lower quality areas. As I analyse a combined measure of job ends that are plausibly exogenous, I also estimate two models where plant closure and dismissal are included as separate dummy variables (column 9 and 10 of Table 2). The estimates suggest that plant closure may affect socio-emotional 19All models with all covariates are available from the author upon request. For Table 2 and Table 3, I include full model regressions in the appendix (see Table A.5 and A.6). 20The referred relationship is depicted in Table A.5, column 1. 14 behaviour more severely as the size of the coefficient is larger compared to dismissals. But due to a small sample size and few plant closure experiences the correlation is insignificant with large standard errors. The coefficient of dismissal also correlates positively with children’s socioemotional behaviour, implying a negative impact, but is not statistically different from zero. Here standard errors and the size of the correlation are smaller compared to plant closure. Still, both reasons of job ending are positively associated with children’s socio-emotional behaviour suggesting that mothers’ experience of plant closure or dismissal may increase children’s socioemotional problems. In sum, stepwise including important control variables does not affect the “raw” correlation of children’s socio-emotional behaviour and involuntary job loss (column 1). The size of the effect remains mostly stable across all specifications. Yet, as discussed in Section 3, OLS estimates require not only that the selection on observables assumption holds, but also that the underlying relationship of maternal job loss and children’s socio-emotional behaviour is linear. Therefore, column 11 in Table 2 compares the estimate obtained from regression-adjusted propensity score matching to the preferred specification shown in column 6. Using propensity score matching also renders a negative impact of maternal job loss on children’s SDQ, which is larger in size (57% of a standard deviation) than the “full” OLS estimate (51% of a standard deviation, column 6). Looking at the standard errors of both models suggests that applying propensity score matching is more efficient than OLS, as the standard errors are smaller. 15 Table 2: Stepwise estimation of socio-emotional behaviour of preschool children and involuntary maternal job loss Socio-emotional behaviour (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Involuntary job loss 0.544∗0.546∗0.531∗0.543∗0.557∗0.510∗0.550∗0.598 0.565∗∗ (0.3059) (0.3056) (0.3047) (0.3167) (0.3085) (0.3042) (0.3085) (0.3819) (0.2256) Plant closure 0.857 0.935 (0.6582) (0.6516) Dismissal by employer 0.334 0.356 (0.2956) (0.3053) Family/child characteristics No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Maternal characteristics No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Paternal characteristics No No No Yes Yes Yes Yes Yes Yes Yes Yes Regional characteristics No No No No Yes Yes Yes Yes Yes Yes Yes Maternal well-being No No No No No Yes Yes Yes Yes Yes Yes Maternal employment (post) No No No No No No Yes Yes No Yes No State and industry fixed effects No No No No No No No Yes No No No N229 229 229 229 229 229 229 224 229 229 226 R20.018 0.067 0.114 0.124 0.135 0.146 0.160 0.252 0.150 0.164 0.439 Adjusted R20.014 0.033 0.042 0.036 0.033 0.041 0.047 0.030 0.040 0.047 0.369 Note: Each cell depicts the effect of maternal job loss on socio-emotional behaviour of preschool children. Except for model (1) all regressions include a linear time trend. The first column comprises the “raw” correlation between maternal job loss and children’s socio-emotional behaviour. The second column includes family and child characteristics (child’s gender, migration background, household income, partnered, and number of children in household) in the regression; the third column adds maternal socio-economic status variables (maternal employment, maternal education, and age at birth); and the fourth column adds paternal years of education and working status to the set of conditioning variables. The fifth column adds regional characteristics (region of residence (East vs. West Germany), rural-urban dummy) and the sixth column controls for maternal well-being prior to job loss (satisfaction with being a mother). The seventh column includes maternal working status after job loss at age 6 to approximate length of unemployment as well as regional unemployment rate. In column 8 state and industry fixed effects are included. Columns 9 and 10 estimate involuntary job loss with separate dummies for plant closure or dismissal by employer using all variables of the preferred specification in column 6 and employment post-job loss respectively. Model (6) comprises the preferred set of conditioning variables and therefore column 11 depicts regression-adjusted matching using the same specification as the sixth column. Own calculations. Source: SOEP v27 (waves 2008-2010). Robust standard errors in parentheses, significance levels: * p<0.10, ** p<0.05, *** p<0.01. 16 5.2. Adolescence sample Table 3 presents the stepwise estimates of the relationship of adolescents’ internal locus of control and maternal job loss. The estimates in all specifications indicate that maternal job loss decreases adolescents’ belief in self-determination by around 26% of a standard deviation, meaning that adolescents whose mothers experience plant closure or dismissal by employer are less likely to believe that working hard or striving for ones own success will help them achieve their own goals. Comparing the coefficient of the “full” OLS model (column 6) to the one of the “raw” model (column 1) suggests that estimating the association of involuntary job loss with adolescents’ internal locus of control is robust to including pre-treatment characteristics. Analogue to the estimations for the preschool sample, I add all pre-treatment variables stepwise to the estimation of job loss on adolescents’ internal locus of control. Further I also control for maternal employment and regional unemployment rate post-job loss (column 7). The coefficient again remains nearly unchanged in size and significance level suggesting that the experience of job loss impacts longer than up to the point where a mother starts working again. Controlling for state and industry fixed effects does not change the estimate either, suggesting that maternal job loss does not only occur in low quality areas. In order to differentiate involuntary job loss, column 9 and column 10 show the association of job loss separately for plant closure and dismissal by employer. The direction of the effect remains negative for both types of job loss, but the coefficient of plant closure is smaller and not statically different from zero. The results indicate that job ends due to dismissal by employer significantly decrease adolescents’ internal locus of control by around 38% of a standard deviation. However by splitting the incidence of involuntary job loss, the coefficient of plant closure is bound to be insignificant as fewer mothers are exposed to firm closure compared to dismissals. Again the results from the stepwise estimation suggest that the effect of maternal job loss on adolescents’ internal locus of control is robust. The size of the effect remains stable across all specifications. Similar to the preschool sample, column 11 compares the coefficient from regression-adjusted propensity score matching to the preferred model (column 6). The significance of the estimated coefficient remains similar and the size of the effect slightly decreases to 22% of a standard deviation. Overall, the results from Table 3 provide evidence that involuntary maternal job loss is associated with children’s non-cognitive outcomes. The estimates from regression-adjusted propensity score matching are more efficient than those obtained from OLS. When looking at the results from the “full” OLS model without appropriately correcting for the selection effect of maternal job loss, the actual effect is overestimated. Interpreting the results obtained in both samples against the background of Cunha and Heckman (2010) suggests that young children are worse off, as early investments are found to have multiplier effects in later childhood. Thus, although experiencing maternal job loss in later childhood has a negative impact on non-cognitive skills, during early childhood this impact may worsen the development in later childhood. 17 Table 3: Stepwise estimation of internal locus of control of adolescents and involuntary maternal job loss Internal locus of control (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Involuntary job loss -0.268∗-0.244∗-0.247∗-0.248∗-0.248∗-0.258∗-0.265∗-0.261∗-0.216∗ (0.1436) (0.1387) (0.1342) (0.1380) (0.1403) (0.1407) (0.1407) (0.1416) (0.1247) Plant closure -0.066 -0.084 (0.1831) (0.1858) Dismissal by employer -0.386∗∗ -0.384∗∗ (0.1811) (0.1800) Family/child characteristics No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Maternal characteristics No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Paternal characteristics No No No Yes Yes Yes Yes Yes Yes Yes Yes Regional characteristics No No No No Yes Yes Yes Yes Yes Yes Yes Maternal well-being No No No No No Yes Yes Yes Yes Yes Yes Maternal employment (post) No No No No No No Yes Yes No Yes No State and industry fixed effects No No No No No No No Yes No No No N522 522 522 522 522 522 522 522 522 522 521 R20.007 0.104 0.137 0.139 0.150 0.152 0.154 0.205 0.155 0.157 0.283 Adjusted R20.006 0.077 0.088 0.083 0.089 0.089 0.088 0.095 0.090 0.089 0.229 Note: Each cell depicts the effect of maternal job loss on internal locus of control of adolescents. Except for model (1) all regressions include year fixed effects. The first column comprises the “raw” correlation between maternal job loss and adolescents’ internal locus of control. The second column includes family and child characteristics (child’s gender, migration background, household income, partnered, and number of children in household) in the regression; the third column adds maternal socio-economic status variables (maternal employment, maternal education, and age at birth); and the fourth column adds paternal years of education and working status to the set of conditioning variables. The fifth column adds regional characteristics (region of residence (East vs. West Germany), rural-urban dummy) and the sixth column controls for maternal well-being prior to job loss (life satisfaction). The seventh column includes maternal working status after job loss at age 17 to approximate length of unemployment as well as regional unemployment rate. In column 8 state and industry fixed effects are included. Columns 9 and 10 estimate involuntary job loss with separate dummies for plant closure or dismissal by employer using all variables of the preferred specification in column 6 and employment post-job loss respectively. Model (6) comprises the preferred set of conditioning variables and therefore column 11 depicts regression-adjusted matching using the same specification as the sixth column. Own calculations. Source: SOEP v27 (waves 2001-2010). Robust standard errors in parentheses, significance levels: * p<0.10, ** p<0.05, *** p<0.01. 18 5.3. Different child outcomes This section looks at further outcomes measuring child development. First, another noncognitive skills measure, namely personality traits, is examined as an additional outcome. In addition to socio-emotional behaviour and internal locus of control, personality traits are also included in both samples – preschool children and adolescence. However, in the adolescence sample personality traits are only surveyed for a subgroup since 2006. Looking at these particular non-cognitive skills reduces the sample size of the adolescence sample to 224 observations. The second set of outcomes examines the association of maternal job loss on grades (German and mathematics) and grade repetition, which facilitates determining if the evidence found in this paper relates to previous findings in the literature. The measures of personality traits are based on the five-factor model. In the preschool sample personality traits are surveyed with a modified measure based on Asendorpf and van Aken (2003) and in the adolescence sample based on McCrae and Costa (1996). Studies in personality psychology show that parental rating of three-to-five-year-old children’s personality varies in five dimensions: extroversion, neuroticism, conscientiousness, agreeableness, and intellect (openness to experience) (e.g., Kohnstamm et al., 1998). Following Asendorpf et al. (2007), the maternal ratings of their young children for ten items can be transferred to five dimensions of personality traits. For adolescents, the items included in the youth questionnaire are similar to those described in Section 4.4. regarding maternal personality traits. Similar to the locus of control, adolescents rate items on a scale from 1 (completely disagree) to 7 (completely agree). For both samples the surveyed personality characteristics are summed in each dimension and transformed to a standardized score with a mean of zero and a standard deviation of one. Table 4 presents estimates obtained from regressing involuntary maternal job loss on the five personality traits for both samples using the preferred specification.21 For the adolescence sample, no association of maternal job loss on adolescents’ personality traits can be found, which cannot be further interpreted due to the drop in sample size. It might either be the case that job loss does not impact personality traits or that the sample size lacks statistical power to estimate the effect of maternal job loss. For preschool children, on the other hand, a negative effect of maternal job loss on social behaviour is replicated with another outcome measure, as the scale of neuroticism increases for treated children. Neuroticism depicts different notions of social inhibition, insecurity, or impulsiveness (for further information, see Almlund et al., 2011). 21The preferred specification is shown in column 6 of Table 2 for preschool children and in column 6 of Table 3 for adolescents. 19 Table 4: Estimation of other child outcomes and involuntary maternal job loss (OLS) (1) (2) (3) (4) (5) (6) (7) (8) OpenConscientExtraAgreeableNeuroticism German Math Grade ness iousness version ness grade grade repetition Panel A : Sample of preschool children Involuntary job loss 0.142 0.071 0.060 -0.152 0.592∗∗ (0.2052) (0.2739) (0.2849) (0.3587) (0.2571) N228 229 229 229 229 R20.144 0.182 0.103 0.153 0.129 Adjusted R20.038 0.081 -0.008 0.049 0.021 Panel B: Sample of adolescents Involuntary job loss -0.135 -0.143 -0.149 -0.109 -0.185 0.147 0.115 0.109∗ (0.2345) (0.2421) (0.2330) (0.2255) (0.2118) (0.1007) (0.1501) (0.0580) N224 224 224 224 224 518 518 522 R20.186 0.153 0.123 0.150 0.192 0.166 0.098 0.127 Adjusted R20.054 0.017 -0.019 0.012 0.062 0.104 0.031 0.062 Panel C: Sample of adolescents Plant closure 0.129 0.053 -0.262 0.268 -0.056 0.015 0.212 0.151∗ (0.2927) (0.3710) (0.3303) (0.3431) (0.3072) (0.1376) (0.2062) (0.0892) Dismissal by employer -0.312 -0.275 -0.073 -0.363 -0.272 0.235∗0.050 0.081 (0.3108) (0.3050) (0.3042) (0.2994) (0.2568) (0.1309) (0.1952) (0.0729) N224 224 224 224 224 518 518 522 R20.191 0.156 0.124 0.159 0.193 0.168 0.099 0.127 Adjusted R20.055 0.015 -0.023 0.019 0.058 0.104 0.029 0.061 Note: Each cell depicts the effect of maternal job loss on other child outcomes, namely personality traits in the sample of preschool children and of adolescents; and grades and grade repetition only in the sample of adolescents. Panel A contains estimations on personality traits in the preschool sample and Panel B depicts estimates of involuntary job loss on personality traits and “cognitive ability” in adolescence sample. Panel C depicts the same effects as Panel B splitting involuntary job loss in plant closure and dismissal by employer for the sample of adolescents. All models in the sample of preschool children include a linear time trend and the preferred set of conditioning variables (column 6, Table 2), and in the sample of adolescents all models comprise year fixed effects and all controls of the preferred model (column 6, Table 3). Own calculations. Source: SOEP v27 (waves 2001-2010). Robust standard errors in parentheses, significance levels: * p<0.10, ** p<0.05, *** p<0.01. With regard to the previous literature, this paper examines German and mathematics grades as well as grade repetition as additional outcomes, since previous evidence finds spillover effects on these measures of educational attainment (Huff-Stevens and Schaller, 2011; Kalil and ZiolGuest, 2008; Rege et al., 2011, e.g.). Kalil and Ziol-Guest (2008) estimate children’s academic performance as a function of parental employment patterns using U.S. survey data. They find no significant correlation between involuntary maternal job loss and children’s grade repetition. Huff-Stevens and Schaller (2011) analyse job loss and children’s likelihood of grade repetition based on the same data, yet they define involuntary job ends more narrowly, focusing only on dismissals and plant closure.22 Applying child fixed effects, Huff-Stevens and Schaller (2011) show that plausibly exogenous displacements of parents are detrimental for children’s academic performance. Parental job loss increases children’s likelihood of repeating a grade by 0.8 percentage points. Furthermore, Rege et al. (2011) look at the effect of displacements of only plant closures on academic performance measured in terms of grade point average (GPA) at age sixteen. They find that fathers’ exposure to plant closure decreases children’s grade point average (GPA) by 6.3% of a standard deviation, where mothers’ job loss has an insignificant 22Kalil and Ziol-Guest (2008) include quitting, dismissal or illness amongst others in their definition of job loss. 20 but positive effect on children’s GPA. To approximate GPA this paper looks at German and mathematics grades.23 Table 4 shows that mother’s involuntary job loss increases adolescents’ likelihood of grade repetition by 12 percentage points (Panel B, column 8), but has no effect on students’ grades.24 Looking at the effect of involuntary job loss with separate dummies (Panel C) indicates that dismissal by employer marginally increases adolescents’ German grade, rendering a negative effect. The correlation of job loss and grade repetition on the other hand is solely found for mothers experiencing plant closure (Panel C, column 8).25 The results in Table 4 can be cautiously compared to previous evidence found by Huff-Stevens and Schaller (2011), as they also look at dismissals and plant closure. The direction of the effect is very similar to what HuffStevens and Schaller (2011) find, while the size of the effect is not: In their OLS specification they find an increase of 1.7 percentage points due to parental job loss (compared to 12 percentage points in Table 4, Panel B, column 8). Looking at grades the results found in this paper differ to those reported by Rege et al. (2011). Although I also find a negative effect of plant closure on grade repetition, my results using the combined measure of involuntary job loss may still be biased. But, the differences in results may also depend on the quality of family environment after mothers experience a job loss. The quality within families with maternal job loss might be relatively worse in Germany compared to Norway. In general, the analysis regarding other child outcomes shows that results found in previous studies are (to some extent) replicated in this sample. 5.4. Maternal outcomes after job loss At the beginning of this paper potential mediators through which an involuntary job loss might affect children’s outcomes are discussed. This section tests the hypotheses of Section 2 by analysing the relationship of involuntary job loss with these outcomes. Regressing involuntary maternal job loss on life satisfaction, household income and maternal non-cognitive skills may reveal potential channels through which mothers’ experiences may be linked to child outcomes. In the sample of preschool children maternal life satisfaction significantly decreases for mothers who experience a job loss (Table 5, Panel A, column 1). The overall life satisfaction after job loss decreases by 1 scale point.26 All models examining life satisfaction and household income in Table 5 include the pre-treatment level of the potential mechanism variables to 23In Germany grades range from 1 (very good (A)) to 6 (unsatisfactory (F)), therefore, a higher average represents poorer performance in the classroom. 24Table A.7 in the appendix shows the results of Panel A and B using regression-adjusted propensity score matching. Here the negative impact of job loss on neuroticism is also found in the sample of preschool children. Further the results for grade repetition are confirmed in the adolescence sample as well; and in addition, a marginal significant impact of job loss on German grade is identified. 25Due to small sample size in the sample of preschool children, I refrain from looking at effects of plant closure and dismissal separately in this sample. 26In the SOEP, overall life satisfaction is measured with a 11-point Likert-type-scale ranging from 0 (absolutely not satisfied) to 10 (absolutely satisfied). 21 control for individual specific starting points. In the adolescence sample a significant decrease in household income after job loss is observed (Table 5, Panel B, column 2). For mothers in the preschool sample, I find no effect of job loss on household income, as women with small children often only add a small fraction to the household income. Interestingly for mothers of children aged seventeen, overall life satisfaction is not affected by the experience of displacement. The correlation found for mothers of young children could suggest that they are more stressed or frustrated about their job loss than mothers of older children. Maternal personality traits – measured using the dimension neuroticism and the internal locus of control – are not significantly affected by job loss: neither in the preschool nor in the adolescence sample (Table 5, Panel A/B, column 3/4).27 Yet, unfortunately for the analyses of this study, personality traits are only measured in the years 2005 and 2009/2010, which rules out the possibility to include the levels prior to job loss. In order to rule out varying impacts on maternal outcomes by job loss type, i.e. plant closure and dismissal by employer, Panel C in Table 5 splits the analyses of involuntary job loss in the adolescence sample. The significant decrease in household income is found for both types of job loss (Table 5, Panel C, column 2), which may suggest that stress due to financial restrictions is similar for both types of job loss. Since most maternal outcomes are not significantly related with plant closure or dismissal by employer, it remains speculation how the different job loss types impact on these maternal outcomes. Yet, it is worth noticing that plant closure positively correlates with life satisfaction and maternal non-cognitive skills28, whereas negative coefficients are obtained for dismissal by employer. Thus, it may be that the overall negative effect on these outcomes are driven by maternal experience of dismissal, which may lead to a negative change in maternal well-being. 27Table A.8 in the appendix shows the results of Panel A and B using regression-adjusted propensity score matching. Here job loss increases the likelihood of maternal neuroticism in the sample of preschool children. This suggests that the emotional stability of mothers with young children decreases after job loss, although this has to be interpreted with caution, as I cannot control for pre-treatment neuroticism. 28The association of plant closure and internal locus of control is even marginally significant. 22 Table 5: Estimation of maternal outcomes and involuntary maternal job loss (OLS) (1) (2) (3) (4) Life satisfaction Household income Neuroticism Internal locus of control Panel A: Sample of preschool children Involuntary job loss -1.053∗∗∗ -0.001 0.363 (0.3765) (0.0928) (0.2711) N228 229 228 R20.304 0.755 0.120 Adjusted R20.214 0.725 0.011 Panel B: Sample of adolescents Involuntary job loss -0.004 -0.129∗∗ 0.009 0.115 (0.2150) (0.0501) (0.1436) (0.1555) N522 521 522 516 R20.270 0.511 0.050 0.104 Adjusted R20.216 0.475 -0.020 0.037 Panel C: Sample of adolescents Plant closure 0.182 -0.129∗0.043 0.412∗ (0.3228) (0.0771) (0.1654) (0.2370) Dismissal by employer -0.129 -0.129∗∗ -0.013 -0.090 (0.2690) (0.0629) (0.1988) (0.1896) N522 521 522 516 R20.271 0.511 0.050 0.111 Adjusted R20.215 0.474 -0.022 0.042 Note: Each cell depicts the effect of maternal job loss on maternal outcomes. Maternal outcomes include overall life satisfaction, household income, and the personality dimension neuroticism in both samples. In the sample of adolescents internal locus of control is also assessed as maternal outcome post-job loss. Panel A contains estimations on maternal outcomes in the sample of preschool children and Panel B depicts estimates in the sample of adolescents. Panel C depicts the same effects as Panel B splitting involuntary job loss in plant closure and dismissal by employer. All models in the sample of preschool children include a linear time trend and the preferred set of conditioning variables (column 6, Table 2), and in the sample of adolescents all models comprise year fixed effects and all controls of the preferred model (column 6, Table 3). Own calculations. Source: SOEP v27 (waves 2001-2010). Robust standard errors in parentheses, significance levels: * p<0.10, ** p<0.05, *** p<0.01. 5.5. Maternal outcomes as potential mediators of job loss In Table 6 the preand post-treatment levels of the potential mediators – if they are available – are added stepwise to the preferred model. Panel A of Table 6 depicts how the potential channels of maternal job loss affect preschool children’s non-cognitive skills. Controlling for maternal life satisfaction preand post-job loss decreases the effect of maternal job loss on preschoolers’ SDQ in size and significance. Job loss still increases children’s socio-emotional behaviour, but only by 38% of a standard deviation instead of 51%. Further adding household income after job loss or mothers’ level of neuroticism also decreases the significance of job loss on socio-emotional behaviour.29 The estimates suggest that maternal emotional balance might be affected, as maternal life satisfaction is significantly lower after involuntary job loss (see Panel A, 29The adjusted R-squared estimates remain fairly stable across all models in Panel A of Table 6, thus suggesting, that neither household income after job loss nor mothers’ level of neuroticism explain any additional variance of children’s socio-emotional behaviour compared to the level of life satisfaction after job loss. 23 Ermisch, J., 2008. Origins in Social Immobility and Inequality: Parenting and Early Child Development. National Institute Economic Review 205 (1), 62–71. Ermisch, J., Peter, F. H., Spiess, C. K., 2012. Early childhood outcomes and family structure. In: Ermisch, J., Jäntti, M., Smeeding, T. (Eds.), From Parents to Children – The Intergenerational Transmission of Advantage. Russell Sage Foundation, New York, pp. 120– 139. Felfe, C., Lalive, R., 2013. Early child care and child development: For whom it works and why. SOEPpapers on Multidisciplinary Panel Data Research 536, DIW Berlin. Frick, J. R., Lohmann, H., 2010. Biography and life history data in the German Socio-Economic Panel (SOEP, v26, 1984-2009). Data Documentation 52, DIW Berlin. Goodman, A., Sianesi, B., 2005. Early education and children’s outcomes: How long do the impacts last? Fiscal Studies 26 (4), 513–548. Goodman, R., 1997. The strengths and difficulties questionnaire: A research note. Journal of Child Psychology and Psychiatry 38 (5), 581–586. Heckman, J., Pinto, R., Savelyev, P., 2013. Understanding the mechanisms through which an influential early childhood program boosted adult outcomes. American Economic Review 103 (6), 2052–2086. Heckman, J. J., 1979. Sample seleciton bias as specification error. Econometrica 47 (1), 153–161. Heckman, J. J., Moon, S. H., Pinto, R., Savelyev, P. A., Yavitz, A., 2010. The rate of return to the HighScope Perry Preschool Program. Journal of Public Economics 94 (1–2), 114–128. Heckman, J. J., Stixrud, J., Urzua, S., 2006. The effects of cognitive and noncognitive abilities on labor marekt outcomes and social behavior. Journal of Labor Economics 24 (3), 411–482. Heineck, G., Anger, S., 2010. The returns to cognitive abilities and personality traits in Germany. Labour Economics 17 (3), 535–546. Hill, J. L., Waldfogel, J., Brooks-Gun, J., Han, W. J., 2005. Maternal employment and child development: A fresh look using newer methods. Developmental Psychology 41 (6), 833–850. Hirano, K., Imbens, G., 2001. Estimation of causal effects using propensity score weighting: An application to data on right heart catheterization. Health Services and Outcomes Research Methodology 2 (3-4), 259–298. Huff-Stevens, A., Schaller, J., 2011. Short-run effects of parental job loss on children’s academic achievement. Economics of Education Review 30 (2), 289–299. 30 James-Burdumy, S. M., 2005. The effect of maternal labor force participation on child development. Journal of Labor Economics 23 (1), 177–211. Kalil, A., Ziol-Guest, K. M., 2005. Single mother’s employment dynamic and adolescent wellbeing. Child Development 76 (1), 196–211. Kalil, A., Ziol-Guest, K. M., 2008. Parental employment circumstances and children’s academic progress. Social Science Research 37 (2), 500–515. Knabe, A., Raetzel, S., Schoeb, R., Weimann, J., 2010. Dissatisfied with life but having a good day: Time-use and well-being of the unemployed. Economic Journal 120 (547), 867–889. Kohnstamm, G. A., Halversa, C. F., Mervielde, I., Havill, V. I., 1998. Parental descriptions of child personality: Developmental antecedents of the Big Five. Mahwah, NJ: Erlbaum. Leuven, E., Sianesi, B., 2003. PSMATCH2: Stata module to perform full mahalanobis and propensity score matching, common support graphing, and covariate imbalance testing, http://ideas.repec.org/c/boc/bocode/s432001.html. This version 4.0.4 10nov2010. Lindo, J. M., 2011. Parental job loss and infant health. Journal of Health Economics 30 (5), 869–879. Loeb, S., Bridges, M., Bassok, D., Fuller, B., Rumberger, R. W., 2007. How much is too much? The influence of preschool centers on children’s social and cognitive development. Economics of Education Review 26 (1), 52–66. Marcus, J., 2013. The effect of unemployment on the mental health of spouses: Evidence from plant closures in germany. Journal of Health Economics 32 (3), 546–558. Marcus, J., 2014. Does job loss make you smoke and gain weight? Economica 81 (324), 626–648. McCrae, R., Costa, P. J., 1996. Toward a new generation of personality theories: Theoretical contexts for the five-factor model. In: Wiggins, J. (Ed.), The Five Factor Model of Personality: Theoretical Perspectives. Guilford, New York, pp. 51–87. Müller, K. U., Spiess, C. K., Tsiasiot, C., Wrohlich, K., et al., 2013. Evaluationsmodul: Förderung und Wohlergehen von Kindern. Politikberatung kompakt Nr. 73, DIW Berlin. OECD, 2013. Protecting jobs, enhancing flexibility: A new look at employment protection legislation. OECD Employment Outlook, OECD, Paris. Oreopoulos, P., Page, M., Huff Stevens, A., 2008. The intergenerational effect of worker displacement. Journal of Labor Economics 26 (3), 455–483. 31 Prevoo, T., ter Weel, B., 2015. The importance of early conscientiousness for socio-economic outcomes: Evidence from the British Cohort Study. Oxford Economics Papers (First published online March 29, 2015.). Rege, M., Telle, K., Votruba, M., 2011. Parental job loss and children’s school performance. The Review of Economic Studies 78 (3), 1–28. Rege, M., Votruba, M., Kjetill, T., 2009. The effect of plant downsizing on disability pension utilization. Journal of the European Economic Association 7 (4), 754–785. Rosenbaum, P., Rubin, D., 1983. The central role of the propensity score in observational studies for causal effects. Biometrika 70 (1), 41–55. Rosenbaum, P., Rubin, D., 1984. Reducing bias in observational studies: Using subclassifications on the propensity score. Journal of American Statistical Association 79 (387), 516–524. Rotter, J. B., 1966. Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs: General and Applied 80 (1), 1–28. Rübennach, S., March 2010. Alles beim Alten: Mütter stellen Erwerbstätigkeit hinten an. STATmagazin Arbeitsmarkt, Statistisches Bundesamt, Wiesbaden. Ruhm, C. J., 2008. Maternal employment and adolescent development. Labour Economics 15 (5), 958–983. Ruhm, C. J., 2009. Maternal employment and child development. In: Crane, D. R., Hill, E. J. (Eds.), Handbook of Families and Work. University Press of America, pp. 331–408. Schaller, J., Zerpa, M., 2015. Short-run effects of parental job loss on child health. NBER Working Paper No. 21745, NBER. Schupp, J., Spiess, C. K., Wagner, G., 2008. Die verhaltenswissenschaftliche Weiterentwicklung des Erhebungsprogramms des SOEP. Vierteljahreshefte zur Wirtschaftsforschung 77 (3), 63– 76. Siedler, T., Schupp, J., Spiess, C. K., 2009. The German Socio-Economic Panel (SOEP) as Reference Data Set. Schmollers Jahrbuch 129 (2), 367–374. Statistisches Bundesamt, 2014. Kindertagesbetreuung. Betreuungsquote von Kindern unter 6 Jahren mit und ohne Migrationshintergrund in Kindertagesbetreuung. Mitteilung vom 7. Mai 2014, Statistisches Bundesamt, Wiesbaden. Stuart, E., 2010. Matching methods for causal inference: A review and a look forward. Statistical Science 25 (1), 1–21. 32 Todd, P. E., Wolpin, K. I., 2007. The production of cognitive achievement in children: Home, school, and racial test score gaps. Journal of Human Capital 1 (1), 91–136. Wagner, G. G., Frick, J. R., Schupp, J., 2007. The German Socio-Economic Panel Study (SOEP) - Scope, evolution, and enhancements. Schmollers Jahrbuch 127 (1), 139–169. Weinhardt, M., Schupp, J., 2011. Mulit-Itemskalen im SOEP Jugendfragebogen. Data Documentation 60, DIW Berlin. 33 Appendix Tables Table A.1: Summary statistics of adolescents’ outcomes, all conditioning variables, and maternal outcomes post-job loss comparing only treated mothers by type of job loss Job loss types Mean Dismissal by employer Plant closure Mean diff. Child outcomes Internal locus of control −0.42 −0.00 −0.42 Personality traits (Big 5) Openness1−0.33 0.23 −0.56 Conscientiousness1−0.11 −0.09 −0.02 Extroversion 1−0.06 −0.21 0.14 Agreeableness1−0.42 0.40 −0.83∗∗ Neuroticism 1−0.11 0.15 −0.26 German grade 3.06 2.79 0.26 Mathematics grade 3.03 3.17 −0.14 Grade repetition 0.25 0.29 −0.04 Child & family characteristics Gender (female=1) 0.50 0.63 −0.13 Migration background 0.19 0.13 0.07 Log(HH income) (at age 6) 6.69 6.69 −0.00 Partnered (at age 6) 0.92 0.92 0.00 Number of children in HH (at age 6) 2.17 1.96 0.21 Maternal characteristics Migration background 0.14 0.13 0.01 Working full time (at age 6) 0.31 0.50 −0.19 Working part time (at age6) 0.28 0.33 −0.06 Working hours (at age 6) 18.31 28.83 −10.53∗∗ Time at firm (at age 6) 2.08 6.03 −3.95∗∗∗ Size of firm (at age 6) 3.92 6.29 −2.38∗∗ Years schooling (at age 6) 12.90 12.15 0.76 University education (at age 6) 0.33 0.33 0.00 No degree (at age 6) 0.08 0.21 −0.13 Age at childbirth 28.28 27.96 0.32 Age at childbirth (20-25) 0.25 0.29 −0.04 Age at childbirth (30-35) 0.25 0.29 −0.04 Age at childbirth (35+) 0.08 0.04 0.04 Paternal characteristic Working (at age 17) 0.72 0.54 0.18 Years of schooling (at age 17) 11.86 11.57 0.30 Regional characteristics Rural urban (at age 6) 0.22 0.25 −0.03 Region East Germany (at age 6) 0.42 0.54 −0.12 Location in 1989 0.44 0.58 −0.14 Maternal well-being Life satisfaction (at age 6) 6.56 6.58 −0.03 Maternal employment post-job loss Regional unemployment rate 13.46 15.30 −1.83 Mother working (at age 17) 0.86 0.75 0.11 Maternal outcomes post-job loss Overall life satisfaction (at age 17) 6.47 6.71 −0.24 Household income (at age 17) 7.97 7.86 0.11 Neuroticism (2005) 0.01 0.06 −0.05 Internal locus of control (2005) −0.15 0.40 −0.55∗ Note: Summary statistics of all conditioning variables for treated mothers in the sample of adolescents. The first three columns present the means and mean differences of variables for mothers who experienced a dismissal by employer; and columns four to six comprise means and mean differences of mothers who experienced a plant closure. 1Sample size N=224 for the outcome personality traits, as the youth questionnaire only includes these since 2006. Source: SOEP v27 (waves 2001-2010). Own calculations, significance levels: * p<0.10, ** p<0.05, *** p<0.01. 34 Table A.2: Descriptive statistics of the sample of preschool children - before and after matching Means Means Job loss No job loss Standard. Bias (%) Variable unmatched matched unmatched matched Child & family characteristics Age of child (in month) 69.40 69.08 69.40 8.40 −0.10 Gender (female=1) 0.47 0.50 0.47 −5.63 0.30 Migration background 0.27 0.15 0.24 27.25 6.82 Log(HH income) (at birth) 7.81 8.00 7.88 −38.74 −13.69 Partnered (at birth) 0.80 0.93 0.86 −36.08 −15.94 Number children in HH (at birth) 1.53 1.79 1.58 −32.02 −6.31 Maternal characteristics Migration background 0.27 0.13 0.23 35.11 8.26 Working full time (prior birth) 0.47 0.45 0.48 3.56 −2.46 Working part time (prior birth) 0.33 0.22 0.30 24.00 6.20 Working full time (at birth) 0.40 0.35 0.38 11.01 3.99 Working part time (at birth) 0.27 0.23 0.30 8.57 −6.53 Years of schooling (at birth) 12.13 13.18 12.34 −44.38 −9.76 Age at childbirth (20-25) 0.13 0.06 0.11 24.12 8.40 Age at childbirth (25-30) 0.33 0.32 0.35 3.26 −3.22 Age at childbirth (30-35) 0.40 0.36 0.37 9.09 6.15 Age at childbirth (35+) 0.13 0.27 0.18 −33.25 −11.64 Paternal characteristics Working (yes/no) (at age 6) 0.73 0.78 0.76 −9.66 −5.59 Years of schooling (at age 6) 13.11 13.17 13.13 −2.59 −0.81 Missing identifier 0.20 0.11 0.17 25.28 7.94 Missing info on education 0.27 0.20 0.24 15.27 6.95 Regional characteristic Rural-urban (prior birth) 0.33 0.30 0.34 7.23 −0.41 Region East Germany (at birth) 0.40 0.31 0.40 17.87 0.46 Location in 1989 0.47 0.35 0.44 24.32 4.71 Maternal well-being Satisfied w/ being mother (around birth) 0.27 0.15 0.25 27.25 3.49 Time trend Year child outcome is measured 2009.00 2009.04 2009.00 −5.26 −0.25 Note: Summary statistics of all conditioning variables for treated, unmatched and matched control individuals. The first two columns present the means of selected variables before treatment for treatment and control groups. The third column displays the standardised per cent bias before matching. It is the per cent difference of the sample means in the treatment and the matched control sample as a percentage of the square root of the average of the sample variances in both groups. The fourth column shows the standardised per cent bias after matching. Own calculations. Source: SOEP v27 (waves 2008-2010). 35 Table A.3: Descriptive statistics of the sample of adolescents - before and after matching Means Means Job loss No job loss Standard. Bias (%) Variable unmatched matched unmatched matched Child & family characteristics Gender (female=1) 0.55 0.45 0.54 19.96 2.84 Migration background 0.17 0.16 0.17 2.34 −0.22 Log(HH income) (at age 6) 6.69 7.41 6.80 −35.01 −4.73 Partnered (at age 6) 0.92 0.94 0.91 −6.99 2.68 Number of children in HH (at age 6) 2.08 2.09 2.10 −1.18 −1.97 Maternal characteristics Migration background 0.13 0.13 0.13 1.02 1.26 Working full time (at age 6) 0.38 0.30 0.38 17.84 1.25 Working part time (at age 6) 0.30 0.34 0.29 −8.51 1.36 Working hours (at age 6) 22.52 20.47 22.49 11.56 0.14 Time at firm (at age 6) 3.66 4.23 3.68 −11.06 −0.33 Size of firm (at age 6) 4.87 5.37 4.91 −12.66 −1.07 Age at childbirth 28.15 27.60 27.85 11.53 6.19 Age at childbirth (20-25) 0.27 0.26 0.28 0.59 −2.68 Age at childbirth (25-30) 0.38 0.44 0.38 −10.49 1.17 Age at childbirth (30-35) 0.27 0.21 0.26 14.34 1.55 Age at childbirth (35+) 0.07 0.08 0.06 −6.69 0.98 Years of schooling (at age 6) 12.60 12.06 12.55 20.38 1.65 University degree (at age 6) 0.33 0.22 0.32 25.22 3.43 No degree (at age 6) 0.13 0.15 0.13 −5.18 −0.08 Paternal characteristics Working (yes/no) (at age 17) 0.65 0.80 0.66 −33.57 −2.34 Years of schooling (at age 17) 11.75 12.02 11.82 −12.49 −3.60 Missing identifier 0.15 0.11 0.16 11.73 −2.78 Missing info on education 0.17 0.13 0.18 10.32 −3.26 Regional characteristics Rural-urban (at age 6) 0.23 0.32 0.22 −18.52 2.58 Region East Germany (at age 6) 0.47 0.35 0.46 24.57 0.62 Location in 1989 0.50 0.35 0.49 30.41 1.23 Maternal well-being Life satisfaction (at age 6) 6.57 7.04 6.55 −30.11 0.91 Time dummies Year 2001 0.17 0.08 0.14 26.40 6.53 Year 2002 0.13 0.12 0.13 4.28 1.09 Year 2003 0.08 0.12 0.09 −11.15 −2.52 Year 2004 0.07 0.13 0.07 −20.64 −2.13 Year 2005 0.10 0.13 0.10 −8.69 0.61 Year 2006 0.17 0.10 0.19 18.33 −4.84 Year 2007 0.12 0.09 0.12 9.16 0.11 Year 2008 0.07 0.08 0.07 −5.91 −0.47 Year 2009 0.05 0.08 0.05 −12.95 1.30 Year 2010 0.05 0.07 0.05 −8.95 −0.05 Note: Summary statistics of all conditioning variables for treated, unmatched and matched control individuals. The first two columns present the means of selected variables before treatment for treatment and control groups. The third column displays the standardised per cent bias before matching. It is the per cent difference of the sample means in the treatment and the matched control sample as a percentage of the square root of the average of the sample variances in both groups. The fourth column shows the standardised per cent bias after matching. Own calculations. Source: SOEP v27 (waves 2001-2010). 36 Table A.4: Estimation of propensity score in the sample of preschool children and of adolescents Propensity of job loss Sample of preschool children Sample of adolescents Child & family characteristics Age of child (in months) 0.003 (0.0045) Gender (female=1) -0.012 0.039 (0.0356) (0.0288) Migration background -0.027 0.020 (0.0785) (0.0788) Log(HH income) (at birth/age 6) -0.019 -0.023∗ (0.0538) (0.0121) Partnered (at birth/age 6) -0.103 -0.000 (0.0763) (0.0627) Number of children in HH (at birth/age 6) -0.006 -0.014 (0.0274) (0.0190) Maternal characteristics Working full time (prior birth) -0.002 (0.0579) Working part time (prior birth) 0.048 (0.0539) Working full time (at birth/ at age 6) 0.041 0.105 (0.0499) (0.0859) Working part time (at birth/ at age 6) 0.027 0.095 (0.0522) (0.0599) Working hours (at age 6) 0.000 (0.0021) Time at firm (at age 6) -0.004 (0.0032) Size of firm (at age 6) -0.016∗∗ (0.0063) Years schooling (at birth/ at age 6) -0.011 0.014 (0.0087) (0.0104) University degree (at age 6) 0.013 (0.0537) No degree (at age 6) 0.016 (0.0504) Migration background 0.112 0.010 (0.0866) (0.0855) Age at childbirth 0.010 (0.0068) Age at childbirth (20-25) -0.006 0.037 (0.0806) (0.0488) Age at childbirth (25-30 -0.025 (0.0451) Age at childbirth (30-35) 0.027 (0.0484) Age at childbirth (35+) -0.033 -0.053 (0.0464) (0.0840) Paternal characteristics Working (yes/no) (at age 6/ at age 17) 0.048 -0.095∗ (0.1010) (0.0508) Years schooling (at age 6/ at age 17) 0.008 -0.012 (0.0086) (0.0075) Missing identifier 0.078 -0.005 (0.0805) (0.1121) Missing info on education 0.009 -0.055 (0.1113) (0.1083) Regional characteristics Rural urban (prior birth/ at age 6) 0.015 -0.048 (0.0399) (0.0324) Location in 1989 0.014 0.124 (0.0741) (0.0921) Region East Germany (at birth/age 6) 0.025 -0.119 (0.0747) (0.0918) Maternal well-being Satisfied w/ being a mother around birth 0.075 (0.0496) Life satisfaction at 6 -0.016∗ (0.0094) N229 522 R20.074 0.096 Note: This table depicts the linear probability model estimates of the propensity score. The first column presents the results for the sample of preschool children and includes a linear time trend; and column two comprises estimates for the adolescence sample also controlling for year fixed effects. Source: SOEP v27 (waves 2001-2010). Own calculations, significance levels: * p<0.10, ** p<0.05, *** p<0.01. 37 Table A.5: Full regression models of Table 2 (Maternal involuntary maternal job loss and socio-emotional behaviour of preschool children) Socio-emotional behaviour (1) (2) (3) (4) (5) Involuntary job loss 0.510∗0.550∗0.565∗∗ (0.3042) (0.3085) (0.2256) Plant closure 0.857 0.935 (0.6582) (0.6516) Dismissal by employer 0.334 0.356 (0.2956) (0.3053) Child & family characteristics Age of child (in months) -0.012 -0.015 -0.012 -0.015 -0.034 (0.0183) (0.0180) (0.0185) (0.0183) (0.0246) Gender (female=1) -0.254∗-0.238∗-0.245∗-0.228∗-0.712∗∗∗ (0.1323) (0.1329) (0.1355) (0.1363) (0.1946) Migration background 0.139 0.192 0.132 0.186 0.526 (0.3249) (0.3214) (0.3238) (0.3202) (0.3680) Log(HH income) (at birth) 0.190 0.209 0.186 0.206 -0.457∗∗ (0.1969) (0.1986) (0.1948) (0.1961) (0.2131) Partnered (at birth) 0.087 0.113 0.095 0.123 0.662∗∗ (0.2737) (0.2819) (0.2707) (0.2778) (0.3305) Number of children (at birth) -0.203∗∗ -0.187∗∗ -0.206∗∗ -0.190∗∗ -0.177 (0.0925) (0.0925) (0.0938) (0.0938) (0.1681) Maternal characteristics Migration background -0.562∗-0.573∗-0.572∗-0.585∗-1.583∗∗∗ (0.3393) (0.3369) (0.3412) (0.3387) (0.3516) Working full time (prior birth) -0.315 -0.294 -0.325 -0.306 -0.065 (0.2070) (0.2030) (0.2105) (0.2066) (0.3054) Working part time (prior birth) 0.129 0.130 0.134 0.136 0.242 (0.1930) (0.1925) (0.1925) (0.1919) (0.2635) Working full time (at birth) 0.267 0.235 0.248 0.215 -0.100 (0.1869) (0.1812) (0.1869) (0.1811) (0.3646) Working part time (at birth) -0.023 -0.029 -0.025 -0.031 -0.268 (0.1683) (0.1699) (0.1684) (0.1700) (0.2579) Years of education (at birth) -0.051 -0.047 -0.050 -0.046 -0.050 (0.0361) (0.0361) (0.0360) (0.0360) (0.0531) Age at childbirth (20-25) 0.297 0.252 0.320 0.275 0.675∗∗ (0.2927) (0.3129) (0.2941) (0.3137) (0.3132) Age at childbirth (25-30) 0.217 0.235 0.225 0.244 0.677∗∗∗ (0.1690) (0.1669) (0.1719) (0.1697) (0.2347) Age at childbirth (35+) 0.239 0.253 0.226 0.239 1.071∗∗ (0.1915) (0.1899) (0.1878) (0.1860) (0.4820) Paternal characteristics Working (yes/no) (at age 6) 0.084 0.119 0.071 0.103 0.649 (0.3902) (0.3660) (0.3877) (0.3621) (0.5992) Years of schooling (at age 6) 0.003 -0.000 0.005 0.002 0.027 (0.0317) (0.0321) (0.0318) (0.0322) (0.0487) Missing identifier -0.225 -0.268 -0.214 -0.256 0.433 (0.3413) (0.3389) (0.3393) (0.3360) (0.5366) Missing info on education 0.442 0.455 0.440 0.451 0.593 (0.4501) (0.4288) (0.4487) (0.4261) (0.6457) Regional characteristics Rural-urban (prior birth) -0.179 -0.283∗-0.176 -0.281∗-0.511∗∗ (0.1487) (0.1667) (0.1508) (0.1686) (0.2497) Region East Germany (at birth) 0.362 0.127 0.354 0.111 0.205 (0.3124) (0.3190) (0.3114) (0.3179) (0.3814) Location in 1989 -0.411 -0.427 -0.411 -0.426 -0.559 (0.3256) (0.3252) (0.3242) (0.3233) (0.4099) Maternal well-being Satisfied w/ being mother (around birth) 0.315 0.308 0.324 0.319 0.112 (0.2081) (0.2060) (0.2066) (0.2040) (0.2884) Maternal employment post-job loss Regional unemployment rate (at age 6) 0.044∗0.045∗ (0.0264) (0.0263) Working (yes/no) (at age 6) -0.094 -0.103 (0.2448) (0.2456) N229 229 229 229 226 R20.146 0.160 0.150 0.164 0.439 Adjusted R20.041 0.047 0.040 0.047 0.369 Note: This table depicts the same results as Table 2 (column 6, 7, 9, 10, and 11) including the estimates of all conditioning variables. Column 1 is similar to column 6 of Table 2 and includes all conditioning variables summarized in Table 1, i.e. controls for family/child characteristics, maternal, paternal, and regional characteristics as well as maternal well-being. The second column adds maternal working status after job loss at age 6 and regional unemployment rate to the model; and the third and fourth column depict the results obtained analysing plant closure and dismissal by employer separately using the same controls as column 1 and 2 of this table respectively. The last column shows all estimates from regression-adjusted matching using the same set of controls as column 1 of this table (or column 11 of Table 2). All models include a linear time trend. Own calculations. Source: SOEP v27 (waves 2008-2010). Robust standard errors in parentheses, significance levels: * p<0.10, ** p<0.05, *** p<0.01. 38 Table A.6: Full regression models of Table 3 (Maternal involuntary maternal job loss and internal locus of control of adolescents) Internal locus of control (1) (2) (3) (4) (5) Involuntary job loss -0.258∗-0.265∗-0.216∗ (0.1407) (0.1407) (0.1247) Plant closure -0.066 -0.084 (0.1831) (0.1858) Dismissal by employer -0.386∗∗ -0.384∗∗ (0.1811) (0.1800) Child & family characteristics Gender (female=1) 0.123 0.117 0.118 0.113 0.077 (0.0877) (0.0885) (0.0873) (0.0882) (0.1450) Migration background 0.140 0.143 0.154 0.155 0.314 (0.2067) (0.2046) (0.2089) (0.2069) (0.2431) Partnered (at age 6) -0.171 -0.162 -0.175 -0.168 -0.269 (0.2233) (0.2296) (0.2233) (0.2295) (0.3018) Number of children in HH (at age 6) 0.029 0.034 0.030 0.035 0.053 (0.0561) (0.0555) (0.0556) (0.0552) (0.0819) Log(HH income) (at age 6) 0.056 0.056 0.054 0.055 -0.008 (0.0370) (0.0369) (0.0369) (0.0369) (0.0503) Maternal characteristics Migration background -0.072 -0.101 -0.080 -0.107 -0.455 (0.2327) (0.2332) (0.2344) (0.2350) (0.2819) Working full time (at age 6) 0.402 0.408 0.406 0.411 0.696 (0.2806) (0.2828) (0.2792) (0.2816) (0.4600) Working part time (at age 6 ) 0.328∗0.334∗0.334∗0.339∗0.666∗∗ (0.1975) (0.1994) (0.1975) (0.1994) (0.3078) Working hours (at age 6 ) -0.005 -0.005 -0.006 -0.005 -0.011 (0.0068) (0.0068) (0.0067) (0.0067) (0.0119) Time at firm (at age 6 ) 0.020∗∗ 0.021∗∗ 0.019∗∗ 0.020∗∗ 0.035∗∗ (0.0092) (0.0095) (0.0093) (0.0096) (0.0155) Size of firm (at age 6 ) -0.012 -0.011 -0.013 -0.011 -0.026 (0.0218) (0.0217) (0.0217) (0.0216) (0.0294) Years of education (at age 6) -0.031 -0.028 -0.029 -0.027 -0.077 (0.0300) (0.0301) (0.0298) (0.0300) (0.0509) University degree (at age 6) 0.126 0.129 0.122 0.125 0.286 (0.1437) (0.1432) (0.1421) (0.1418) (0.2312) No degree (at age 6 ) -0.158 -0.149 -0.168 -0.158 -0.270 (0.1542) (0.1555) (0.1541) (0.1555) (0.2269) Age at childbirth -0.001 -0.004 -0.002 -0.004 0.005 (0.0171) (0.0179) (0.0172) (0.0180) (0.0209) Age at childbirth (20-25) -0.093 -0.096 -0.094 -0.097 -0.128 (0.1359) (0.1369) (0.1361) (0.1370) (0.1636) Age at childbirth (30-35) 0.001 -0.001 -0.002 -0.004 -0.156 (0.1352) (0.1361) (0.1359) (0.1367) (0.1788) Age at childbirth (35+) -0.187 -0.172 -0.181 -0.167 -0.620∗ (0.2195) (0.2210) (0.2214) (0.2226) (0.3570) Paternal characteristics Working (yes/no) (at age 17) 0.133 0.141 0.139 0.147 0.240 (0.1500) (0.1511) (0.1496) (0.1508) (0.2058) Years of schooling (at age 17) -0.017 -0.017 -0.017 -0.017 -0.016 (0.0231) (0.0233) (0.0230) (0.0232) (0.0407) Missing identifier -0.010 0.030 -0.033 0.006 -0.532 (0.2817) (0.2735) (0.2862) (0.2789) (0.3728) Missing info on education 0.041 -0.002 0.059 0.017 0.680 (0.2605) (0.2536) (0.2641) (0.2581) (0.4275) Regional characteristics Rural-urban (at age 6) 0.133 0.149 0.131 0.148 0.173 (0.0930) (0.0967) (0.0933) (0.0969) (0.1617) Region East Germany (at age 6) -0.177 -0.101 -0.180 -0.104 -0.537 (0.2736) (0.3006) (0.2752) (0.3021) (0.3677) Location in 1989 0.443 0.445 0.441 0.443 0.518 (0.2890) (0.2884) (0.2911) (0.2901) (0.3583) Maternal well-being Life satisfaction (at age 6) -0.025 -0.024 -0.026 -0.024 -0.061 (0.0281) (0.0280) (0.0283) (0.0282) (0.0420) Maternal employment post-job loss Regional unemployment rate -0.008 -0.008 (0.0156) (0.0156) Working (at age 17) -0.167 -0.153 (0.1519) (0.1527) Year fixed effects Yes Yes Yes Yes Yes N522 522 522 522 521 R20.152 0.154 0.155 0.157 0.283 Adjusted R20.089 0.088 0.090 0.089 0.229 Note: This table depicts the same results as Table 3 (column 6, 7, 9, 10, and 11) including the estimates of all conditioning variables. Column 1 is similar to column 6 of Table 3 and includes all conditioning variables summarized in Table 1, i.e. controls for family/child characteristics, maternal, paternal, and regional characteristics as well as maternal well-being. The second column adds maternal working status after job loss at age 6 and regional unemployment rate to the model; and the third and fourth column depict the results obtained analysing plant closure and dismissal by employer separately using the same controls as column 1 and 2 of this table respectively. The last column shows all estimates from regressionadjusted matching using the same set of controls as column 1 of this table (or column 11 of Table 3). All models include year fixed effects. Own calculations. Source: SOEP v27 (waves 2001-2010). Robust standard errors in parentheses, significance levels: * p<0.10, ** p<0.05, *** p<0.01. 39