The World Bank Economic Review, 38(3), 2024, 558–579 https://doi.org10.1093/wber/lhad041 Article Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Job Loss and Household Labor Supply Adjustments in Developing Countries: Evidence from Argentina Matías Ciaschi and Guido Neidhöfer Abstract Using longitudinal data for Argentina, this paper estimates the labor supply reaction of spouses and children, as well as the interactions between them, following the job loss of their husband or father. The findings show that job loss by the household head has a positive and significant impact on the labor supply of other household members. However, it increases the likelihood of spouses to switch to informal and downgraded employment, and of children to drop out from education. While effects are stronger among vulnerable households, cover- age of social security does not provide enough support in coping with unemployment shocks. Mothers’ labor participation, however, may allow their daughters to continue their education. JEL classification: J16, J21, J22, J65 Keywords: job loss, labor supply, female labor participation, educational enrollment, educational drop-out, human capital formation, idiosyncratic shocks 1. Introduction Belonging to a household implies risk-sharing and represents an insurance mechanism to deal with adverse shocks. Aiming to smooth consumption, households can cope with adverse labor-market shocks suffered by the main income earner through changes in the labor-force participation of other household mem- bers. Early theoretical contributions (Humphrey 1940; Woytinsky1940) and empirical studies (Heckman 1983; Lundberg 1985; Maloney 1987, 1991; Spletzer 1997; Stephens 2002) labeled the reaction of wives to their husband’s employment loss as “added worker effect.” More recently, researchers have found sev- eral factors influencing the existence and magnitude of this coping strategy. Among these, the role of social security (Cullen and Gruber 2000; Bentolila and Ichino 2008; Birinci 2019; Wu and Krueger 2021), ag- gregate female labor participation (Bredtmann,Otten, and Rulff 2018; Keldenich and Knabe 2018), labor The authors are grateful to Mariana Marchionni for her continuous support and useful suggestions. They also thank Inés Berniell, Lucila Berniell, Irene Brambilla, Dolores De la Mata, Julia Bredtmann, Andrés Cesar, Guillermo Falcone, Leonardo Gasparini, Carlo Lombardo, Leonardo Peñaloza Pacheco, Friedhelm Pfeiffer, Lucía Ramírez, Leopoldo Tornarolli, two anony- mous referees, and the editor for their valuable comments, as well as participants of seminars and conferences in Buenos Aires, Bahia Blanca, La Plata, and Montevideo. Any errors are our responsibility. A supplementary online appendix is available for this article at the World Bank Economic Review website. Matias Ciaschi is a senior researcher at the Center for Distributive, Labor and Social Studies (CEDLAS-IIE-FCE) at the Na- tional University of La Plata, La Plata, Argentina, and doctoral fellow at the National Scientific and Technical Research Coun- cil (CONICET), Buenos Aires, Argentina; his email address is matiaschiaschi@gmail.com. Guido Neidhoefer (corresponding author) is a senior researcher at the Leibniz Centre for European Economic Research (ZEW), Mannheim, Germany; his email address is guido.neidhoefer@zew.de. C 2024 International Bank for Reconstruction and Development / The World Bank. Published by Oxford University Press The World Bank Economic Review 559 informality (Basu. Genicot, and Stiglitz 2000; Maloney 2004) liquidity restrictions (Ortigueira and Siassi, 2013), and macroeconomic dynamics (Parker and Skoufias 2004; Skoufias and Parker 2006; Mattingly and Smith 2010; Bryan and Longhi 2018; Albanesi 2019; Serrano et al. 2019; Guner, Kulikova, and Vallardares-Esteban 2021). Furthermore, the effects of parental job loss and income shocks on children’s Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 labor and educational decisions were explored separately. It was found that economic shocks may lead to educational drop-out (e.g., Duryea, Laml and Levison 2007; Cardona-Sosa et al. 2018; Cerutti et al. 2019; Di Maio and Nistico 2019; GC Britto et al., 2021) and lower school performance (Rege, Telle, and Votruba 2011), and that parental unemployment can have persistent effects on their offsprings’ human capital investments and future income (Schmidpeter 2020; Kaila, Nix, and Riukula 2021).1 The main contribution of this study is to analyze the joint reaction of spouses and children to sudden unemployment of the breadwinner in the household in the context of a developing country. The joint consideration of labor participation decisions of different household members makes it possible to eval- uate the existence of interactive decisions between them. This paper studies this effect using rich sets of individual panel data for urban Argentina from 1995 to 2015, which makes it possible to obtain estimates abstracting from individual-level heterogeneity. Furthermore, in additional analyses the paper follows the literature and controls for bias deriving from potential sources of endogeneity, such as skill selection and anticipation effects, by focusing on households where the head lost his job in any time period in a panel dataset and by exploiting the variation in the timing of the job loss. The results are consistent when1 abstracting for these potential sources of endogeneity. First, the paper estimates the effect of job loss by the (male) household head on the likelihood of their (female) partner becoming active on the labor mar- ket, or increasing their working hours. Second, it estimates the likelihood of taking up an informal job and of occupational downgrading by the spouse in reaction to her husband’s job loss. Third, the study estimates the impact on older children’s labor-force participation and educational drop-out. This study analyzes heterogeneity in these effects along the income distribution, and for male and female children. Furthermore, it tests how the labor supply reaction of one household member interacts with the reactions of other household members, and whether the effects differ depending on the provision of unemployment benefits. The study also provides evidence regarding women’s job quality when taking on an additional worker role. The role of labor informality and employment downgrading as strategies to cope with unemployment shocks suffered by the partner are still unsolved questions, despite their importance for the evaluation of labor-market prospects, especially of women. While recent contributions found that informal jobs prevent female workers from leaving the labor market upon motherhood (Berniell et al., 2021), studies on the effect of the breadwinner’s unemployment on female labor participation are, so far, limited to examining whether the female partner became active on the labor market and indeed found a job (Bredtmann, Otten, and Rulff 2018). This analysis contributes new evidence on this issue. The study of coping strategies to insure from household shocks is particularly relevant for developing countries. In these contexts, female labor participation and educational enrollment are usually lower than in developed countries, particularly in Latin America. Also, income and unemployment shocks are more frequent since labor markets are likely to show higher instability and social security is more restricted. Cultural factors may also play a role, since traditional gender roles are likely to be more prevalent. Indeed, the literature has found significant differences in female labor supply responses to their husband’s job loss between developed and developing countries: estimates in Latin American countries show that women increase their labor-force participation by between 12 and 20 percentage points (Parker and Skoufias 2004; Fernandes and Felício 2005; Skoufias and Parker 2006; Paz 2009; Cardona-Sosa et al. 2018), while for European countries the estimates show an increase by between 3 and 9 percentage points (Hardoy 1 For a recent review of the evidence and methods regarding the effects of job loss on household outcomes, see Ruiz- Valenzuela (2021). 560 Ciaschi and Neidhöfer and Schøne 2014; Bredtmann Otten, and Rulff 2018; Halla, Schmieder, and Weber 2020; Keldenich and Knabe 2018). Most of the literature studying the factors shaping household coping strategies is focused on developed countries. The findings of this study shed light on the role of insurance mechanisms within the household Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 in a typical context of a developing country, with rather high income volatility, unstable macroeconomic conditions, the prevalence of informal labor arrangements, and relatively low coverage of unemployment benefits. In developed countries these types of social benefits are broader and may crowd out labor par- ticipation adjustments (Bentolila and Ichino 2008; Birinci 2019; Wu and Krueger 2021; Bertheau et al. 2022). The paper tests whether in the context of a developing country such as Argentina these benefits are effective to insure households against unemployment shocks, avoiding informality, downgrading and educational drop-out. The main findings suggest that a substantial number of female spouses who were initially outside of the labor market become active in response to their husband’s job loss. The estimates show an increase in labor supply by 15 percentage points. Those who were already employed increase their labor supply by around two working hours per week. The study also finds significant effects on labor informality and employment downgrading among women exposed to these shocks. Regarding children, after job loss by the household head, the labor-force participation of sons and daughters increases by 7 percentage points. Among sons, the effect is around 6 percentage points, while among daughters it is around 9. Importantly, the results confirm that the labor-supply response of mothers and their adult children acts as a substitute for each other, especially when there are no younger siblings in the household. The estimates also suggest that enrollment in education falls by 14 percentage points due to the job loss of the household head and that this effect is higher among sons than among daughters. Importantly, mothers’ labor participation protects their daughters against educational drop- out. Lastly, the study finds that these effects are stronger among poorer households and that social security plays a rather limited role in helping households to cope with unemployment shocks, given that its coverage only encompasses a relatively low number of households. The remainder of this paper is organized as follows: Section 2 introduces the specific mechanisms under investigation within the context of this case study. Section 3 presents the data. Section 4 explains the estimation strategy, section 5 the results, and section 6 concludes. 2. Institutional Background Argentina represents an interesting case study to examine the mechanisms through which households cope with unemployment shocks. The country is characterized by rather unstable economic performance, which can be illustrated by the evolution of unemployment displayed in fig. 1. Following a deep economic crisis during the 1980s, structural reforms related to trade liberalization and macroeconomic stabilization in the early 1990s helped reduce unemployment. However, the combination of international economic crises (particularly in 1998) with an unsustainable fiscal policy led to an economic crisis in 2001–2002. After this disruptive episode, and helped by windfalls in international commodity prices, the Argentinean economy started another stabilization process. From 2003 to 2008, a stronger labor market, combined with progressive fiscal policies, resulted in a considerable decrease in unemployment rates. Following the 2009 international crisis, the tailwind ended, and government fiscal capacity diminished. Unemployment remained virtually stagnant, and employment rates started to decrease once again. Figure 1 also gives a first aggregate overview of household responses to unemployment shocks. The left panel suggests a positive relationship between male unemployment trends and both female and youth la- bor participation, as documented in previous contributions (Serrano et al. 2019).The right panel suggests a positive correlation between male unemployment and educational drop-out of individuals in households belonging to the three lowest income deciles. This correlation is noticeable, given broadly extended free public education in Argentina (De Hoyos, Ganimian, and Holland 2021). Therefore, in this context, it is The World Bank Economic Review 561 Figure 1. Unemployment, Labor Participation and Educational Attendance. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Source: SEDLAC (CEDLAS and World Bank). plausible that supply-side factors and budget constraints directly associated with educational investments are less likely to contribute to children’s dropping linked to their parents’ job loss. Instead, the associa- tion between job loss and children’s dropping out may be attributed to other possible mechanisms. For instance, labor income shocks could encourage children to drop out of school, as such shocks may dimin- ish the perceived returns or benefits of investing in education (Eckstein and Zilcha 1994). Furthermore, an indirect income effect may be in play, whereby children find themselves compelled to increase their labor participation in response to the household income loss, driven by the necessity to mitigate the financial strain suffered by the family during periods of economic adversity. The empirical set up of this paper tests whether these mechanisms are indeed present at the household level. The described patterns are also silent about the characteristics of the jobs that women and children are taking on. Household labor supply adjustments can, at least partially, compensate income losses; however, this may be at the expense of job quality. For instance, the income loss might lower the opportunity cost and, thus, force household members to accept any job opportunity. This is particularly relevant for women, who, as the recent literature shows, are more likely to be employed in flexible jobs in order to combine employment and motherhood (Berniell et al. 2021). To evaluate the conditions in which women are inserting themselves in the labor market, this paper also studies whether a husband’s job loss induces wives to take on informal work or to take jobs for which they are over skilled (employment downgrading). This topic still represents an important gap in the literature. To understand the effects of household shocks on job quality is particularly important in the context of developing countries, where labor informality is high and the coverage of unemployment insurance rather low. Figure 2 provides a stylized figure of this. The left graph in fig. 2 shows that labor informality affected constantly more than 30 percent of the Argentinean working population from 1995 to 2015, reaching almost 50 percent during the 2001–2002 crisis. As much of the literature documents, labor informality has important consequences in terms of labor-market prospects.2 Importantly, labor benefits may be in- adequate to fully insure against income shocks, making households more likely to trigger other coping strategies, such as individual labor-supply adjustments. With respect to informality, income compensa- tions, such as social protection schemes, tend to have a low coverage2 in developing countries.3 The right graph in fig. 2 shows that in Argentina the coverage of unemployment insurance estimated by the Na- tional Agency of Social Security (ANSES) is, on average, about 6 percent. Previous literature has focused 2 Informality in Latin America is consistently associated with lower earnings and higher job instability (Gasparini and Tornarolli 2009; Tornarolli et al. 2014; Ulyssea 2020) and also to less on-the-job human capital accumulation (Bobba et al. 2021). 562 Ciaschi and Neidhöfer Figure 2. Informality and Unemployment Insurance. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Source: SEDLAC (CEDLAS and World Bank) and ANSES (Administración Nacional de la Seguridad Social), own elaboration. on developed countries, where this type of social assistance is broader and has found that these benefits can crowd out household labor participation adjustments (Bentolila and Ichino 2008; Birinci 2019; Wu and Krueger 2021). The low coverage of these transfers in Argentina raises the question of whether they are effective in reducing the adverse reactions of households to employment and income shocks. This paper will test this hypothesis in the empirical analysis. 3. Data Representative Household Survey This paper uses longitudinal data from the Encuesta Permanente de Hogares (Permanent Household Sur- vey; hereafter, EPH), the primary household survey in Argentina carried out by the INDEC (Instituto Nacional de Estadísticas y Censos). The analysis is restricted to the period between 1995 and 2015 be- cause more recent survey versions do not include the panel structure needed to compute labor transitions. The survey contains information on a large number of socioeconomic variables, including employment and marital status, household structure, individual income, region of residence, and education of each household member. Approximately 19,000 households are surveyed in every round.3 Overall, the survey covers more than 100,000 inhabitants in urban areas representing about 68 percent of the country’s popu- lation. The rural population is not covered. Some questions in the survey are answered by each individual over the age of 10 living in the household. Other questions related to the whole household are answered by the household head. In all cases, it is possible to clearly identify couples, offspring, and their individual characteristics4 During the analyzed period, the EPH had a rotating sample panel structure, which makes it possible to follow households for a maximum of four survey waves over a 1.5 year period. The rotative panel design differs between the periods 1995–2002 and 2003–2015. In the first interval, the survey collected information for two waves in each year, updating 25 percent of the sample in each survey. Consequently, between 2003 and 2015, each household was visited four times, twice in two consecutive trimesters, was 3 Unemployment insurance in Latin America is connected to the labor market functioning, since informality (both in levels and transitions) is high and only formal workers are entitled to receive income compensation benefits (Levy and Schady 2013). Recent advances in noncontributory schemes, which are mainly focused on people in extreme poverty, provide little help against income shocks due to unemployment (Busso et al., 2021). 4 The definition of fathers and mothers is based on whether the individual is the child’s parent or step-parent. The World Bank Economic Review 563 not visited for two trimesters, and then visited two times during the last two trimesters.5 The analysis exploits this rotative panel structure to compute employment and/or educational transitions for each individual, with special attention to periods of unemployment. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Sample The analysis focuses on households with heterosexual couples and male household heads.6 Additionally, the sample is restricted to couples in which both members are between 25 and 60 years old in the initial period to avoid considering the influence of educational and retirement decisions of household heads and their partners on labor-market participation. The study only considers households interviewed four times in order to avoid biases related to panel attrition. Section S1.C1 in the supplementary online appendix shows that nonattriting and attriting house- holds, that is, households observed four times and households observed less than four times in the survey, exhibit no significant differences in their characteristics. While it is important to acknowledge that job loss can be a contributing factor to sample attrition through migration or household dissolution, the presence of attrition in the sample results in fewer observed job losses, potentially leading to an underestimation of the effects. However, as the analysis in section S1.C1 in the supplementary online appendix indicates, the main results remain largely unaffected when both attriting and nonattriting households are included. The analysis uses the second to fourth interviews to compute labor and educational transitions. Since this study is also interested in analyzing changes in labor market participation and educational enrollment of sons and daughters living in the household, and in evaluating possible interactive decisions with their mother’s reactions, labor and educational transition variables are defined at the individual level. The sample is also restricted to households where the male head was employed at the time of the first interview. The main explanatory variable that is used, in the estimations is a binary indicator which equals 1 if the household head became unemployed between two consecutive survey waves. Outcome variables for spouses and children are also defined by exploiting changes between two consecutive survey waves. The analysis considers labor market participation changes at both the extensive margin (labor market entry) and intensive margin (weekly working hours). In the case of older children, a variable indicating educational drop-out is also computed; that is, a negative change in educational enrollment from one period to the next. Each of these main outcome variables is defined within a particular sample. When computing the labor-market participation of female spouses (offspring) at the extensive margin, the analysis only con- siders those spouses (offspring) who were not participating in the labor market in the initial period. When it considers changes in hours worked among female spouses (offspring), the sample is restricted to those spouses (offspring) that were employed across all time periods. Changes in offspring’s educational en- rollment are only defined for those who were enrolled in education in the first period in which they are included in the survey. Hence, after applying the sample restrictions mentioned above, the analysis is per- formed on five different samples according to the outcome variable under analysis. Table 1 illustrates descriptive statistics for the five samples (A to E). While the percentage of household head’s job loss is similar across samples, a relevant difference appears when comparing households with female spouses who were not participating in the labor market in the initial period (sample A), with those households in which spouses were employed in all periods (sample B). The former group has, on average, lower ed- ucation and income, and higher child presence and unemployment exposure than the latter. At the same time, households with at least one son or daughter outside of the labor market (sample C), and those with at least one son or daughter enrolled in education at any level (sample E), show income, education, and unemployment exposure levels between the other samples, as well as a higher number of children. 5 Table S2.1 in the supplementary online appendix illustrates the EPH rotating panel scheme. 6 Households with female household head represent only about 5 percent of the full sample. 564 Ciaschi and Neidhöfer Table 1. Main Variables Descriptive Statistics Sample A mean Sample B mean Sample C mean Sample D mean Sample E mean HH head job loss (%) 12.18 11.08 11.63 12.94 10.64 Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 (32.70) (31.39) (32.06) (33.56) (30.83) Male head’s education (years) 9.70 11.32 10.37 9.45 10.83 (3.70) (4.04) (4.06) (3.86) (4.00) Wives’ education (years) 9.61 11.96 10.40 9.55 10.88 (3.49) (3.97) (3.96) (3.75) (3.84) Male head’s age 44.34 44.03 49.26 50.86 49.17 (10.65) (9.70) (6.55) (6.77) (6.36) Wives’ age 41.66 41.50 46.68 48.35 46.64 (10.90) (9.34) (6.43) (6.74) (6.17) Household income (log) 5.82 6.52 6.02 6.26 6.09 (1.85) (1.66) (1.82) (1.63) (1.84) At least one child (%) 90.99 85.48 100 100 100 (28.64) (35.23) – – – Num. of children 2.30 1.93 3.07 2.64 2.99 (1.52) (1.35) (1.55) (1.40) (1.45) Regional unemployment (%) 14.61 13.92 14.28 14.26 14.43 (4.16) (3.80) (4.02) (3.80) (4.06) Regional male unemployment (%) 13.88 13.15 13.52 13.49 13.68 (4.45) (4.09) (4.33) (4.08) (4.35) Observations 96782 93054 76819 19590 71862 Source: EPH, own estimates. Note: Standard errors in parentheses. Samples: (A) Female spouses not participating in the labor market in the initial period; (B) Female spouses always employed; (C) Households with at least one son or daughter not participating in the labor market in the initial period; (D) Households where at least one son or daughter was employed; (E) Households with at least one son or daughter enrolled in any educational level in the initial period. However, households where at least one son or daughter was employed (sample D) show higher levels of job loss and lower education. Furthermore, the sample of households with 16–25 year old children outside of the labor market (sample C) is composed of older couples.7 4. Empirical Strategy In order to compute changes to labor participation and education, the analysis exploits the time structure of the data in the same spirit as event-study approaches. As mentioned in the previous section, changes at the extensive margin of female partners, sons, or daughters only take into consideration those households in which the female partner, son, or daughter, respectively, was initially not active in the labor market. Regarding changes in labor-force participation at the intensive margin, the analysis measures changes in the hours worked per week of those that were already employed before the event (job-loss of the household head). Changes in educational enrollment are measured for those sons or daughters in the household that were already in education before the event. Educational drop-out is then defined as leaving education between one period and the next without graduating. 7 The sample of offspring is restricted to those that are between 16 and 25 years old. The main rationale for choosing this age bracket is because individuals in this age range are more likely to make decisions on whether to work or to study. Younger children are not included as their labor participation may be under-reported in household surveys since employment of children under the age of 16 is illegal in Argentina. As a result, respondents may choose not to disclose the working habits of underage children. In order to address potential biases due to household composition and coresidency, Table S2.20 in the supplementary online appendix shows that the effects on children aged 16–20 are similar to those aged 16–25. The World Bank Economic Review 565 As previously mentioned, the sample is restricted to the second to fourth interviews and the analysis only considers households where the male head was employed at the time of the first interview. These sample restrictions aim to minimize potential anticipation mechanisms. In fig. S2.1 of the supplementary online appendix, it is possible to observe the correlation between job loss and period 1 heterogeneity for Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 samples A, C, and E as described in table 1. The analysis reveals that households with initially informal male heads and lower incomes are more susceptible to job losses, while this pattern is less pronounced for the sample of initially out-of-labor-market wives. These findings align with the results presented in tables S2.2 to S2.6 of the supplementary online appendix, where the analysis examines the changes in household characteristics under different sampling restrictions. Notably, there are no significant differences between the first and second columns (because almost all male heads are employed), but households experiencing job losses tend to exhibit lower levels of education and initial income, as well as higher exposure to local unemployment. To analyze spouses and offspring’s labor supply responses to job loss of the household head and po- tential interactive decisions, the analysis estimates the following model: Yihrt = α + γ Ehrt + δ Fhrt + ζ Ehrt Fhrt + X ihrt β + ψi + ϕrt + εihrt (1) where Yihrt is the outcome of interest of individual i, living in household h and region r, in time period t. The coefficient γ captures i’s reaction to changes in the employment situation of the household head (Ehrt ). Ehrt is 1 if the male household head was employed in the previous period and is not employed in the current period, and 0 otherwise. Fhrt is the labor supply reaction of other household members (e.g., of the mother when the labor-supply reaction of children is estimated). Hence, ζ identifies the existence of interactive decisions between household members. The analysis estimates five different specifications for Yihrt : (a) female spouses’ labor-force participa- tion at the extensive margin (labor-market entry); (b) female spouses’ labor-force participation at the intensive margin (weekly working hours); (c) children’s labor-force participation at the extensive margin; (d) children’s labor-force participation at the intensive margin; (e) children’s educational enrollment. In specifications (a) and (c), the sample is restricted to spouses or children, respectively, who were not active in the labor market in the first interview. In specifications (b) and (d), Yihrt indicates the weekly hours of work. In specification (e), the analysis considers a sample of children who were enrolled in education in the first interview. In the first specifications of the empirical model, it abstracts from interactive decisions and restricts δ and ζ to be 0. In these models, in the presence of household adjustment mechanisms be- tween spouses, γ , which is the main coefficient of interest, is expected to be positive in each specification of the empirical model. The vector Xihrt is a vector of individual- and household-level control variables including age, pres- ence of children in the household, age of the youngest child in the household, and number of children. In addition, estimations include region-specific time fixed effects (ϕ rt ), and individual fixed effects (ψ i ). The inclusion of individual fixed effects captures unobserved heterogeneity, which represents a potential iden- tification issue that has not been entirely ruled out in most previous contributions (Bredtmann, Otten, and Rulff 2018). They also capture relevant individual characteristics, which are assumed to be fixed during the 1.5 years the household is observed, such as education level, which is likely related to labor partici- pation reaction and job opportunities. Besides, the inclusion of region-specific time fixed effects aims to capture changes in labor-market conditions, particularly from the demand side. ε iht is an idiosyncratic error term. The estimates are performed clustering standard errors at the household level.8 8 Similar results were obtained when considering region-level clustered standard errors in the analysis. 566 Ciaschi and Neidhöfer Two potential caveats of the analysis are related to, first, negative skill selection into unemployment and, second, the household members’ potential anticipation of unemployment loss. If, due to assortative mating and intergenerational transmission of human capital, the labor-force reaction of spouses and chil- dren, as well as children’s educational enrollment decisions, are also negatively correlated with skills, skill Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 selection would bias the estimates upward. Anticipation effects, on the other hand, would bias the esti- mates downward, because spouses and children may have already adjusted their labor-force participation or educational enrollment before their husband or father’s unemployment occurs. Both of these caveats are associated with the panel structure of the data, which does not provide precise information on the timing of job loss. As a result, the estimation faces the limitation of not knowing whether individuals who experience job loss during the unobserved period found a job or remain unemployed. This situation introduces a potential selection bias, as the decision to find a new job or remain unemployed during this period may be influenced by unobserved characteristics.9 “Endogeneity and Robustness” in section ad- dresses these potential endogeneity issues by estimating the effect on a more homogeneous subsample of households where the head lost his job in any of the time periods in the analysis, following Hilger (2016); Halla, Schmieder, and Weber (2020); and Fadlon and Nielsen (2021). Other potential issues are general macroeconomic conditions that could be correlated with both job loss by the household head and employment of other household members. Since this study analyzes labor- force participation (including those active in the labor market but unemployed), rather than employment, the specification does not depend on the actual likelihood of finding a job by household members. How- ever, individuals may become discouraged in the midst of a recession and, thus, stay out of the labor market. The literature on this subject is still inconclusive: some contributions find evidence supporting this hypothesis (Kohara 2010); others show that this effect is particularly strong when other household members expect to find a job in the same industry where the household head was employed (Hardoy and Schøne 2014); recent studies, which were focused on the interaction between household responses and macro conditions, found opposite results (Bredtmann, Otten, and Rulff 2018). In presence of these “discouraged worker effects” the estimates could be considered as a lower bound. Lastly, the analysis acknowledges another potential identification concern related to the estimation strategy, which excludes the simultaneous occurrence of job losses for both fathers and mothers. While the literature does not provide a clear consensus on the effects of maternal job loss on children’s out- comes (Rege, Telle, and Votruba 2011; Ruiz-Valenzuela 2021), this exclusion may introduce bias in the estimators due to the correlation between job losses experienced by both household members. If the im- pact of job losses goes in opposite directions, the strategy could result in downward-biased estimates. Conversely, if the impact of both parents’ job losses are positively correlated, the estimations might be upward biased. In this sense, some recent contributions suggest assessing the robustness of the estimates by considering both job losses (Hupkau et al. 2020; Ruiz-Valenzuela 2020). To address this concern and ensure the robustness of the estimates, table S2.23 of the supplementary online appendix conducts an additional analysis. It examines the effects of the male head’s job loss on children’s outcomes while con- trolling for mother’s job loss, and the interaction term between them. The sample used in this analysis includes households in which both fathers and mothers were employed in the initial period. The results indicate that main effect on children’s outcomes stems from their fathers’ job loss. The interaction Uterm is not statistically significant for both daughters and sons, while the coefficients for maternal job loss are smaller than father’s and not statistically significant, except for the case of daughters’ dropping out. This last result aligns with the main findings, indicating that mothers’ labor-force participation only prevents educational drop-out among daughters. 9 Appendix Table S2.21 shows the main results including fixed effects by household interview. Results are very similar. They suggest that there is no systematic correlation between the timing of the interview and unobservable individual characteristics related to job loss or the likelihood of recovering employment status. The World Bank Economic Review 567 5. Results This section presents the main results of this paper: in “The Effect of Female Labor Force Participation” in section 5 shows the estimates of the effect of breadwinner’s job loss on female labor participation, both at the extensive and the intensive margin, and job quality. It also discusses possible interactive decisions Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 between mother and children’s labor participation. “The Effect on Sons and Daughters” in section 5 focuses on children’s labor-force participation and educational drop-out, also considering mothers’ labor participation. “Heterogeneous Impacts” in section 5 reports heterogeneous effects by initial household income and recipients of social benefits. Finally, “Endogeneity and Robustness “in section 5 discusses potential sources of endogeneity in the estimates and provides robustness checks to test the validity of the findings in terms of causal inference. The Effect on Female Labor-Force Participation Table 2 shows the impact of the household head’s job loss on female labor-force participation at the extensive and intensive margin, and on job quality. Then, this part of the analysis uses samples A and B from table 1. Panel A in table 2 shows the baseline estimates, omitting interactive decisions between mother’s and children’s labor participation, whereas panel B also considers the reaction of other household members. Interestingly, the results do not show significant differences between the two analyses, suggesting that the reaction of mothers is independent of their children’s labor-force participation decisions. The point estimates in column (1) show that female labor participation increases by around 15 percentage points in reaction to their husband’s job loss. Considering that a share of about 47 percent of women are initially out of the labor market, the results point at an increase in female labor participation of 31 percent. This point estimate is similar to previous findings in Latin America that found an effect between 12 and 20 percentage points (Parker and Skoufias 2004; Fernandes and Felício 2005; Skoufias and Parker 2006; Cardona-Sosa et al. 2018; Paz 2009). In comparison to previous contributions, the estimates control for individual-level heterogeneity and estimate the effect over a longer time horizon. Recent contributions for developed countries show lower estimates, between 3 and 9 percentage points (Hardoy and Schøne 2014; Bredtmann, Otten, and Rulff 2018; Keldenich and Knabe 2018; Halla, Schmieder, and Weber 2020). To evaluate whether these women indeed find a job when entering the labor market, column (2) shows estimates where the outcome variable is 1 if the woman, who previously was not active on the labor market, becomes employed in reaction the her husband’s job loss.10 The results suggest that the vast majority of women becoming active on the labor market indeed find a job soon. Column (3) shows the effects at the intensive margin; that is, changes in the amount of weekly working hours. The estimates suggest an increase of about 1.4 working hours. Considering that spouses who were initially employed work, on average, 32 weekly hours, the results suggest an increase of about 4.4 percent. In line with previous contributions for developed countries, which found a positive effect on women’s labor supply responses at the intensive margin (Mattingly and Smith 2010; Bredtmann, Otten, and Rulff 2018; Bryan and Longhi 2018), this paper is the first to find a significant effect in a developing country.11 The findings so far show that women are able to enter the labor market and find employment or increase their working hours in reaction to household income shocks. However, this does not take into account the characteristics of the employment. As mentioned, the income shock could reduce the opportunity cost 10 This aspect was explored by Bredtmann, Otten, and Rulff (2018). 11 Fernandes and Felício (2005) and Martinoty (2015) present results for Brazil and Argentina, respectively, suggesting that the entire women labor supply response materializes at the extensive margin. However, both studies are focused on economic crisis periods, which might be a special case. Generally, comparisons of estimates based on working hours have to be treated cautiously because of potential measurement errors in self-reported measures that could bias labor-supply elasticities (Barrett and Hamermesh 2019). Since these measurement errors have been shown to be positively correlated with the respondent’s wage rate, and job loss is more likely in the lower part of the income distribution, the estimates of this paper are likely to be less affected by this potential problem. 568 Table 2. Estimates on Female Labor Participation Panel A: Baseline estimates. Change into Change into Labor participation Employed Hours Informality informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) HH head job loss 0.148∗∗∗ 0.130∗∗∗ 1.322∗∗∗ 0.025∗∗∗ 0.037∗∗∗ 0.070∗∗∗ 0.071∗∗∗ (0.011) (0.010) (0.513) (0.008) (0.008) (0.013) (0.014) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 96782 96782 93054 70773 55776 60364 49449 Average .47 .47 32.03 .36 .26 .46 .42 Panel B: Considering interactive decisions. Labor participation Employed Hours Informality Change into Downgrading Change into informality downgrading (1) (2) (3) (4) (5) (6) (7) HH head job loss=1 0.149∗∗∗ 0.127∗∗∗ 1.119∗∗ 0.023∗∗∗ 0.035∗∗∗ 0.071∗∗∗ 0.072∗∗∗ (0.011) (0.011) (0.550) (0.008) (0.008) (0.014) (0.015) HH head job loss=1 × Children’s LFP=1 −0.026 (0.026) 0.014 (0.025) 1.452 (1.484) 0.018 (0.020) 0.012 (0.022) −0.020 (0.025) −0.013 (0.028) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 96665 96665 92969 70715 55741 60332 49430 Average .47 .47 32.03 .36 .26 .46 .42 Source: EPH, own estimates. Note: Column (1) measures the effect of male household head’s job loss on the labor-force participation of their wives. Column (2) measures the effect on wives’ employment. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality. Column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses while panel B only includes those with children. Robust standard errors clustering at the household level indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in column (1) and (2) it shows the average probability of being out of labor force. Ciaschi and Neidhöfer Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 The World Bank Economic Review 569 of work and force women to accept any job in order to mitigate the shock. In addition, women may prefer jobs that are more flexible, such as informal jobs, due to childcare and other kind of domestic tasks (Berniell et al. 2021). Columns (4) and (5) show the effect estimates on labor informality, columns (6) and (7) on employment downgrading, as indicators of women’s employment quality. An informal job is Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 hereby defined as such if it does not provide any right to pensions once retired. In column (4), the sample includes all women, while in column (5) the sample includes only women employed in the initial period, and hence measures the effect on changing from formal to informal employment. Both point estimates are positive and statistically significant. The likelihood of working in an informal employment among all women increases by around 3 percentage points and among employed women by around 4. Interestingly, the increase is in the same order of magnitude, and even slightly higher both in absolute and relative terms, among already employed women. This suggests that not only newly employed women are more likely to take informal jobs, but also that women are more likely to switch from a formal to an informal job in reaction to their husband’s sudden unemployment. Table S2.24 of the supplementary online appendix examines the effects of job loss on the working hours of women who transition into labor informality. It also investigates potential variations between women who change their primary job and those who do not. Specifically, the analysis considers job change as the shift in occupation between consecutive interviews, categorized according to the International Standard Classification of Occupations (ISCO). Results reveal that women who transition into informal employment after their husband’s job loss do so by altering their main occupation, thereby increasing their working hours. These findings provide insights into the dynamics of job transitions and their impact on women’s employment patterns in the context of labor informality. The results also show evidence supporting the existence of employment downgrading among women. Columns (6) and (7) only consider skilled women (i.e. those with a completed high school diploma). The INDEC’s Código Nacional de Ocupaciones is used in order to classify occupations and consider low-skilled jobs as those identified as “nonskilled” or “operative,” which correspond to the 1-digit ISCO categories 7, 8, and 9. The results show that the likelihood of highly qualified women to work in low- skilled jobs increases by around 7 percentage points, both for the full sample and among already employed women, in reaction to their husband’s job loss. In sum, the results from the estimations suggest that the labor-force participation of women increases in reaction to their husband’s job loss in Argentina. However, it also induces women to be more likely to work in informal jobs or jobs for which they are overqualified. The Effect on Sons and Daughters Partners are not the only household members who can become “added workers.” Older children may also change their labor-force participation decisions and their enrollment in education in reaction to their father’s job loss, particularly in developing countries. The effect on children can have important implications for their human capital formation and, if the effect is asymmetric across socioeconomic groups, negatively affect intergenerational mobility and increase future inequality. In what follows, this paper estimates the effects on children’s labor supply and educational enrollment. Then, in this part of the analysis samples C, D, and E from table 1 are used.12 Importantly, this section also examines whether mothers’ labor supply reactions can prevent their children’s labor participation and educational drop-out in this context. Table 3 shows the effect of father’s job loss on labor-force participation, employment, hours worked, and educational drop-out of sons and daughters living in the same household. Again, panel A shows the baseline estimates, omitting interactive decisions. Panels B and C consider the joint reaction of children 12 As explained in section 3, the analysis restrict the sample to children aged between 16 and 25. Estimations considering couple’s children aged between 18 and 25 years yield similar results. 570 Table 3. Estimates on Children’s Labor Participation Panel A – Baseline estimates. Daughters Sons Labor participation Employed Hours Drop-out Labor participation Employed Hours Drop-out (1) (2) (3) (4) (5) (6) (7) (8) HH head job loss 0.097∗∗∗ 0.084∗∗∗ −1.859 0.140∗∗∗ 0.062∗∗ 0.059∗∗ −0.340 0.147∗∗∗ (0.029) (0.029) (1.236) (0.020) (0.029) (0.027) (1.343) (0.024) Controls Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Yes Observations 41188 41188 7121 37147 35631 35631 12469 34715 Average .65 .65 9.84 .69 .51 .51 14.02 .58 Panel B: Considering interactive decisions. Daughters Sons Labor employed Hours Drop-out Labor Employed Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) (7) (8) HH head job loss=1 0.103∗∗ 0.112∗∗ −0.873 0.144∗∗∗ 0.078 0.100∗∗ 0.808 0.149∗∗∗ (0.044) (0.044) (1.669) (0.023) (0.056) (0.050) (2.103) (0.020) HH head job loss=1 × Mother −0.121∗ −0.174∗∗∗ 3.797 −0.082∗∗ −0.167∗∗ −0.207∗∗ 0.101 −0.003 LP=1 (0.068) (0.066) (2.904) (0.041) (0.083) (0.081) (3.164) (0.045) Controls Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Yes Ciaschi and Neidhöfer Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Table 3. Continued Panel A – Baseline estimates. Daughters Sons Labor participation Employed Hours Drop-out Labor participation Employed Hours Drop-out (1) (2) (3) (4) (5) (6) (7) (8) Observations 20912 20912 3385 17380 18059 18059 6712 16455 The World Bank Economic Review Average .65 .65 10.23 .65 .49 .49 14.83 .53 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor employed Hours Drop-out labor Employed Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) (7) (8) HH head job loss=1 0.136∗∗∗ 0.127∗∗ −3.693∗∗ 0.121∗∗∗ −0.027 −0.057 −1.151 0.125∗∗∗ (0.052) (0.052) (1.834) (0.024) (0.057) (0.057) (1.937) (0.022) HH head job loss=1 × Mother hrs. −0.004∗ −0.004∗ −0.135 −0.002∗ −0.001 0.001 0.088 −0.001 change (0.002) (0.002) (0.097) (0.001) (0.002) (0.002) (0.106) (0.001) Controls Yes Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Yes Observations 20270 20270 3733 19752 17557 17557 5744 18250 Average .66 .66 9.80 .72 .53 .53 13.42 .63 Source: EPH, own estimates. Note: Columns (1)–(2), and (5)–(6) measure the effect of male household head’s job loss on the labor-force participation of their daughters and sons, respectively. Columns (3) and (7) measure the effect on weekly working hours among employed children. Columns (4) and (8) measure the effect on educational drop-out. Panel A considers all children, panel B only includes children whose mothers were not participating in the labor market in the initial period, and panel C considers children of already working mothers. Robust standard errors clustering at the household level indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (5) it shows the average probability of being out of labor force while in columns (4) and (8) it shows the average educational attendance. 571 Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 572 Ciaschi and Neidhöfer and mothers at the extensive and intensive margin, respectively. The first three columns show the effect estimates for the sample of daughters, the last three columns for the sample of sons. In panel A, for children of both sexes the effects follow the same pattern, and the effect sizes are, on average, in a similar order of magnitude. Job loss by the household head is associated with a significant labor-force participation Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 increase by around 10 (6) percentage points for daughters (sons), and a higher likelihood to drop out from education by 14 (15) percentage points. Estimates for employment are in the same magnitude. However, the analysis does not yield any statistically significant association with weekly working hours for those already in the labor force. However, the results regarding the effects on children’s drop-out are in line with previous estimates for Latin American countries (Duryea, Lam, and Levison 2007; Cardona-Sosa, Flórez and Morales 2018; Cerutti et al. 2019; Britto, Melo, and Sampaio 2021) and other developing countries (Di Maio and Nistico 2019) while previous contributions found negligible effects in developed countries (Hilger 2016). When considering interactive decisions in panel B and C, the results show interesting differences by child gender. While children’s labor-force participation is completely prevented by their mother’s labor- market entry, educational drop-out does not seem fully insured in the case of sons. However, a mother’s labor supply participation counterbalances a daughter’s educational drop-out. This means that even when the mother becomes active on the labor market in reaction to the job loss suffered by the household head, and this substantially reduces the likelihood of children to become active of the labor market themselves, educational drop-out still cannot be fully avoided. Also, when mothers already participating to the la- bor force increase their working hours in reaction to the income shock, this only weakly prevents their children’s reaction to the household shock: given that the average change in mothers’ hours worked is around 11 hours, it only prevents daughter’s labor-force participation reaction by about 32 percent and drop-out by about 18 percent. However, it is important to emphasize that maternal labor participation may not act as a preventive measure against longer-term educational drop-outs. Table S2.26 of the sup- plementary online appendix presents an analysis focusing on the effects of job loss on drop-outs, defined as children who discontinue their education during the second or third interview and remain out of the educational system in subsequent periods. The findings reveal that sons are particularly susceptible to these longer-term drop-outs, experiencing an 8.8 percent decline in educational attainment, while daugh- ters experience a 6.3 percent decrease. Notably, the participation of mothers in the labor force does not mitigate the impact. This indicates that these households may lack sufficient coping strategies to alleviate the adverse consequences of employment shocks. The observed effects on educational drop-outs can likely be attributed to demand-side factors, consid- ering the widespread availability of public free education in Argentina. These effects may stem from the need for compensation due to an income shock or the discouragement of children’s education resulting from decreasing returns to education (Eckstein and Zilcha 1994). This holds significant policy implica- tions, as income compensation policies may not effectively prevent drop-outs if they are primarily driven by low expected returns to education. To explore this further, this study analyzes the heterogeneous ef- fects on children’s outcome by parental background. Findings in Table S2.25 of the supplementary online appendix indicate that the effects on daughters are primarily driven by the income shock resulting from job loss. Specifically, it can be observed that the interaction term between father’s job loss and a binary variable indicating whether the higher-educated parent has completed secondary education or higher is statistically significant only when evaluating the effects on daughter’s drop-outs. This suggests that the effect on sons is less influenced by budget-related mechanisms related to the direct cost of education. In- stead, factors such as changes in returns to education and adverse labor market conditions may play a more dominant role in shaping sons’ decisions regarding labor participation and educational drop-outs. Furthermore, the observed low correlation between job losses among fathers and mothers, as demon- strated in Table S2.23 and fig. S2.2 of the supplementary online appendix, indicates that mothers may be more likely to compensate for the income shock. While there is limited evidence on these mechanisms, a The World Bank Economic Review 573 similar pattern was found by Di Maio and Nistico (2019) in another developing country, Palestine, where job loss had a more pronounced impact on drop-out rates among sons from less educated parents. These findings are also connected to the literature on the income elasticity of educational expenditure in developing countries. It has been shown that households with lower levels of education tend to allo- Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 cate a smaller proportion of their budget to education, which is also more susceptible to income shocks (Qian and Smyth 2011; Huy 2012; Acar, Günalp, Cilason 2016, , among others). Unfortunately, the EPH dataset used in this study does not include information on educational spending to directly test this hy- pothesis. In order to approximate the potential impact on household investment in education, the analysis is extended to examine the influence of parental job loss on public education attendance for children in the sample who were initially enrolled in private education and did not drop out. As shown in fig. S2.3 of the supplementary online appendix, results show that this transition was observed across the entire income distribution despite being higher among daughters from poorer households, which also suggests that the investment in daughters’ education is more sensitive than sons’ to budget constraints. The anal- ysis then investigates whether the effects of parental job loss on educational drop-out varied depending on whether children were enrolled in public or private institutions. To examine this, the estimations repli- cates the baseline analysis and introduces an interaction term between job loss and a binary variable representing attendance in public education. Fig. S2.4 in the supplementary online appendix also demon- strates, results show that educational drop-outs are not statistically more likely to occur among children in public education. Overall, these findings suggest that the transitions from private to public education are not the primary channel through which parental job losses impact children’s education. Instead, the most significant impacts are observed in relation to educational drop-outs. The increase in educational drop-outs is particularly worrisome. In Argentina, educational enrollment is almost universal for children between 6 and 14 years old. However, this rate is substantially lower, around 82 percent, for individuals aged between 15 and 17 (Marchionni, Gasparini, and Edo 2019). Employment shocks might, hence, further decrease children’s human capital formation by leading to edu- cational drop-out and lower overall achievements. The findings of this study, which show a higher effect of parental job loss on educational drop-out than on labor-force participation, suggest that the direct cost of education plays an important role, besides the opportunity cost associated with foregone earnings. Income-support programs or scholarships targeting the direct costs of education might hence be a tool that could reduce disruptions in human capital formation due to household income losses, at least in contexts similar to Argentina. Heterogeneous Impacts Once the study is extended to family members other than the couple, the aim is to analyze some of the mechanisms that explain how households cope with unemployment shocks. Here, the analysis considers asymmetric impacts by initial household characteristics such as income (i.e., reported household income in the first interview) and benefits received by the household (severance payments and unemployment insurance, which depend on the head’s job characteristics in the first interview). To perform these analyses- an interaction term between E in equation (1) and these variables is included. This paper represents the first contribution analyzing the determinants of household coping strategies in this way, particularly in the context of a developing country. Household Income Theoretically, the relationship between labor-supply reactions and household income is, in principle, am- biguous. On the one hand, for high-income households job loss may represent a higher drop in income and, hence, a stronger need to compensate consumption levels by the labor-supply participation of other household members. However, income should also be positively correlated with savings and wealth, which may help to better cope with adverse shocks in the short run. Additionally, higher income can be associ- 574 Ciaschi and Neidhöfer Figure 3. Labor Supply Adjustments and Household Income. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Source: EPH, own estimates. Note: The coefficient associated with “HH head job loss” represents the effect for the first decile of the initial household income distribution. The coefficients for the interaction of initial income deciles and head job loss should be interpreted with respect to it. 90 percent confidence intervals. ated with higher levels of education and skills, which may help to soon find other employment solutions. However, due to assortative mating and intergenerational transmission of human capital, this may also apply for other household members, who, thus, may find it easier to enter the labor market in reaction to income shocks. Consequently, the overall impact of initial household income is, a priori, unclear. Figure 3 shows the estimates of the interaction term between job loss of the household head and initial household income (in the first interview) for three outcomes of interest: women’s labor supply at the extensive margin, children’s labor supply reaction at the extensive margin, and children’s educational drop-out. The results suggest that the magnitude of the female labor supply reaction to their husband’s job loss decreases with initial household income. The probability for women who were initially out of the labor force to become active on the labor market is substantially higher for households with an income level below the median of the distribution. In line with previous contributions for Latin America (such as Serrano et al. 2019), the results suggest that female labor-force participation reactions are stronger among lower-income households. On the other hand, they do not show any clear relationship between household income and children’s labor-supply changes; the effect of a father’s job loss on educational enrollment is suggestively lower among richer households. This last result is in line with a higher exposure to job losses among lower income households, as shown in fig. S2.3 of the supplementary online appendix. Formal Employment and Unemployment Insurance Finally, the role of income compensation mechanisms is studied. The analysis first evaluates differences in the reaction of spouses and children to their household head’s job loss depending on whether the job was in the formal or informal sector. Figure 4 shows the results. While they do not show a statistically significant difference of the effect on children’s labor participation, the labor-force participation response of spouses and educational drop-out of children of workers who lose their employment in the formal sector is about 30–50 percent lower than of workers that were previously employed in the informal sector. Then, the role of social security is analyzed. While previous contributions have demonstrated that social security leads to smaller secondary worker’s responses in developed countries (Cullen and Gruber 2000; Bentolila and Ichino 2008; Birinci 2019; Wu and Krueger 2021), less is known about the role of these transfers in developing countries. This difference is, however, very important since most of the social security schemes (for example, unemployment insurance) are designed for formal employment contexts, while in Argentina about 35 percent of the jobs are informal (SEDLAC 2022). Considering this, the effect for households where the main earner benefited from the receipt of any form of unemployment insurance versus households that did not receive this form of income support is estimated. Figure 4 shows the results. The World Bank Economic Review 575 Figure 4. Labor Supply Adjustments and Income Support. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Source: EPH, own estimates. Note: The coefficients for the interaction of initial head’s formality (upper panel) or unemployment insurance receipt (lower panel) and head job loss should be interpreted with respect to informal workers and households not receiving that benefit, respectively. 90 percent confidence intervals. The results suggest that unemployment insurance may partially or fully insure households against un- employment shocks. While in households that do not receive any unemployment insurance spouses and children are likely to become active on the labor market, and children are likely to drop out of educa- tion, this likelihood is close or indistinguishable from 0 for households that do receive the transfer. This result confirms past findings for developed countries: social security can help in mitigating the conse- quences of employment or income shocks. However, due to the low coverage of unemployment insurance in Argentina, taking into account the full population, this impact is rather limited. Endogeneity and Robustness In previous contributions, such as Bredtmann, Otten, and Rulff (2018), unobserved heterogeneity could not be entirely ruled out due to data restrictions. However, certain unobserved individual characteristics, such as preferences for leisure and opportunity costs, could be correlated with initial household income and employment status and, hence, bias the estimates in an a priori undefined direction. To address this issue, the empirical model includes individual fixed effects. However, it is important to note that while in- dividual fixed effects control for unobservable characteristics of women and children, they do not account for unobservable characteristics of male heads of households. Additionally, the lack of precise information regarding the exact timing of job loss, as well as the different time spans between the second and third interview, and between the third and fourth interview, present a potential limitation in the analysis. This could result in negative skill selection into unemployment, despite the expectation of similar unobservable characteristics among household members.13 To address these potential endogeneity issues, this section reports the results of additional analyses that aim at estimating the employment loss only among households for which it can be considered an idiosyncratic shock. First, following recent contributions this section estimates the effect on a subsample 13 As noted above, section C4 in the supplementary online appendix replicates their main analysis including fixed effects by number of interview and results held. 576 Ciaschi and Neidhöfer of households in which the head lost his job in any of the time periods in the analysis (e.g., Hilger 2016; Halla, Schmieder, and Weber 2020; Fadlon and Nielsen 2021). The objective of this specification is to estimate the effect among a homogeneous sample of households with respect to their unobserved skills and likelihood of unemployment. Hence, this analysis mainly exploits the variation in the timing of the Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 job loss. In each quarter, it compares households affected by unemployment shocks to those currently not affected, but that will lose their job in a later period. By controlling for quarter and household fixed effects, households exposed and non-exposed to job loss should be rather comparable. The results, reported in the supplementary online appendix in tables S2.12 and S2.13, confirm the patterns observed so far and are very close to the baseline results. Finally, the analysis evaluates the results in different time periods. In line with Bredtmann et al. (2018) indicators of macroeconomic performance were considered for the subdivision in periods. Time periods are defined based on the male employment dynamics in Argentina displayed in fig. 1. Section B in the supplementary online appendix shows the results. Generally, the same qualitative pattern of the baseline results is observed in each time period. Interestingly, while the labor supply reaction of spouses seems par- ticularly relevant around the major 2001–2002 crisis that the country suffered, job loss by the household head is consistently associated with the educational drop-out of children in all time periods. Further- more, unlike in other periods, during the crisis mothers’ labor-force participation does not prevent their daughter’s educational drop-out. 6. Conclusions This paper evaluated how households cope with adverse income shocks resulting from employment shocks in the context of a developing country such as Argentina. Accordingly, it took into account the joint re- action of distinct household members, estimated the effect on labor-force participation and enrollment in education, and evaluated the presence of interactive decisions. As argued, the estimates abstract from individual-level heterogeneity, skill selection, and anticipation effects. Therefore, the analysis contributes to a more comprehensive picture of how households cope with unemployment shocks in developing coun- tries. Additionally, the paper studies different heterogeneities, such as by gender of the child and along the distribution of household income. Lastly, it evaluated the role of social protection in shaping coping strategies. The analysis found that both spouses and children substantially increase their labor-force participation after a male breadwinner’s job loss, especially in low-income households. Additionally, the estimates show a substantial impact of the shock on labor informality and employment downgrading among women, and on educational drop-out of children. The paper also found that the labor-force participation of mothers prevents their sons’ and daughters’ labor-force participation but hampers only the educational drop-out of daughters, since it is more influenced by budget-related mechanisms. Finally, the analysis found that income support and social security are indeed able to counterbalance adverse household reactions to income shocks, but that they play a rather minor role due to the low coverage of these social protection schemes. Extending social insurance mechanisms in contexts similar to Argentina, for instance increasing labor formality, could hence be an effective strategy to reduce the vulnerability of households in the middle and at the bottom of the distribution from unexpected life events, such as income shocks. In conclusion, the results of the analysis are relevant from a policy perspective. They show that shocks affect households in an asymmetric way and may damage the process of human capital formation of future generations in developing countries. The findings highlight the need for efficient social security systems and employment policies that target women and young adults in order to match the increased labor supply with its demand and provide support to ensure equality of educational opportunities also in times of crisis. As research has shown, improving equality of opportunities can be a driver of economic development (Neidhöfer et al. 2023). Hence, policies that prevent disruptions to human capital formation, The World Bank Economic Review 577 which may happen through educational drop-out or employment downgrading, should have long run benefits for individual and aggregate economic performance, particularly in developing contexts. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Conflict of interest The authors have no conflicts of interest to declare. Data availability The data underlying this article will be shared on reasonable request to the corresponding author. Raw data for this article is publicly available at INDEC’s website: https://www.indec.gob.ar/indec/web/ Institucional- Indec- bases_de_datos_eph_amp, for the 1995-2002 period; and https://www.indec.gob.ar/ indec/web/Institucional- Indec- BasesDeDatos for the 2003-2015 period. References Acar, E.Ö., B. Günalp, and S.M. Cilasun. 2016. “An Empirical Analysis of Household Education Expenditures in Turkey.” International Journal of Educational Development 51, 23–35. Albanesi, S. 2019. “Changing Business Cycles: The Role of Women’s Employment. Working Paper No 25655. National Bureau of Economic Research. Cambridge, MA, USA. Barrett, G.F., and D.S. Hamermesh. 2019. “Labor Supply Elasticities: Overcoming Nonclassical Measurement Error Using More Accurate Hours Data.” Journal of Human resources 54(1): 255–265. Basu, K., G. Genicot, and J.E. Stiglitz. 2000. “Unemployment and Wage Rigidity When Labor Supply Is A Household Decision.” Economics Working Paper No. 00-01-11. UC Irvine. Irvine, CA, USA. Bentolila, S., and A. Ichino. 2008. “Unemployment and Consumption Near and Far Away from the Mediterranean.” Journal of Population Economics 21(2): 255–280. Berniell, I., L. Berniell, D. De la Mata, M. Edo, and M. Marchionni. 2021. “Gender Gaps in Labor Informality: The Motherhood Effect.” Journal of Development Economics 150: 102599. Bertheau, A., E. M. Acabbi, C. Barcelo, A. Gulyas, S. Lombardi, and R. Saggio. 2022. “The Unequal Cost of Job Loss across Countries.” Working Paper No. 29727. National Bureau of Economic Research. Cambridge, MA, USA. Birinci, S. 2019. “Spousal Labor Supply Response to Job Displacement and Implications for Optimal Transfers.” Working Paper (2019-020). Federal Reserve Bank of St. Louis. St. Louis, MO, USA. Bobba, M., L. Flabbi, S. Levy, and M. Tejada. 2021. “Labor-Market Search, Informality, and On-the-Job Human Capital Accumulation.” Journal of Econometrics 223(2): 433–53. Bredtmann, J., S. Otten, and C. Rulff. 2018. “Husband’s Unemployment and Wife’s Labor Supply: The Added Worker Effect Across Europe.” Ilr Review 71(5). 1201–231. Britto, D.G.C., C. Melo, and B. Sampaio. 2021. “The Kids Aren’t Alright: Parental Job Loss and Children’s Education in Brazil.” SSRN. https://ssrn.com/abstract=4024250 or http://dx.doi.org/10.2139/ssrn.4024250. Bryan, M., and S. Longhi. 2018. “Couples’ Labour Supply Responses to Job Loss: Growth and Recession Compared.” Manchester School 86(3), 333–357. Busso, M., J. Camacho, J. Messina, and G. Montenegro. 2021. “Social Protection and Informality in Latin America During the Covid-19 Pandemic.” PLoS ONE 16(11): e0259050. Cardona-Sosa, L., L.A. Flórez, L.F. Morales, and B. de la República. 2018. “How Does the Household Labour Supply Respond to the Unemployment of the Household Head?” Labour 32(4): 174–212. Cerutti, P., E. Crivellaro, G. Reyes, and L.D. Sousa. 2019. “Hit and Run? Income Shocks and School Dropouts in Latin America.” Labour 33(4): 533–566. Cullen, J.B., and J. Gruber. 2000. “Does Unemployment Insurance Crowd Out Spousal Labor Supply?” Journal of labor Economics 18(3): 546–572. De Hoyos, R., A.J. Ganimian, and P. A. Holland. 2021. “Teaching with the Test: Experimental Evidence on Diagnostic Feedback and Capacity Building for Public Schools in Argentina.” World Bank Economic Review 35(2): 499–520. 578 Ciaschi and Neidhöfer Di Maio, M., and R. Nistico. 2019. “The Effect of Parental Job Loss on Child School Dropout: Evidence from the Occupied Palestinian Territories.” Journal of Development Economics 141, 102375. Duryea, S., D. Lam, and D. Levison. 2007. “Effects of Economic Shocks on Children’s Employment and Schooling in Brazil.” Journal of Development Economics 84(1): 188–214. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Eckstein, Z., and I. Zilcha. 1994. “The Effects of Compulsory Schooling on Growth, Income Distribution and Welfare.” Journal of Public Economics 54(3): 339–359. Fadlon, I., and T.H. Nielsen. 2021. “Family Labor Supply Responses to Severe Health Shocks: Evidence from Danish Administrative Records.” American Economic Journal: Applied Economics 13(3): 1–30. Fernandes, R., and F.D. Felício. 2005. “The Entry of the Wife into the Labor Force in Response to the Husband’s Unemployment: A Study of the Added Worker Effect in Brazilian Metropolitan Areas.” Economic Development and Cultural Change 53(4): 887–911. Gasparini, L., and L. Tornarolli. 2009. “Labor Informality in Latin America and the Caribbean: Patterns and Trends from Household Survey Microdata.” Revista Desarrollo y Sociedad(63): 13–80. Guner, N., Y. Kulikova, and A. Valladares-Esteban. 2021. “Does the Added Worker Effect Matter?” Working Paper No. 2113. Banco de Espana. Madrid, Spain. Halla, M., J. Schmieder, and A. Weber. 2020. “Job Displacement, Family Dynamics, and Spousal Labor Supply.” American Economic Journal: Applied Economics 12(4): 253–87. Hardoy, I., and P. Schøne. 2014. “Displacement and Household Adaptation: Insured by the Spouse or the State?” Journal of Population Economics 27(3): 683–703. Heckman, J. 1983. “A Life-Cycle Model of Family Labour Supply.” In Human Resources, Employment and Develop- ment. Weisbrod, B., Hughes, H. (eds), 213–230. Springer. London. Hilger, N.G. 2016. “Parental Job Loss and Children’s Long-Term Outcomes: Evidence from 7 Million Fathers’ Lay- offs.” American Economic Journal: Applied Economics 8(3): 247–83. Humphrey, D.D. 1940. “‘Alleged Additional Workers’ in the Measurement of Unemployment.” Journal of Political Economy 48(3): 412–19. Hupkau, C., I.E. Isphording, S.J. Machin, and J. Ruiz-Valenzuela. 2020. “Labour Market Shocks during the Covid-19 Pandemic, Inequalities and Child Outcomes.” IZA Discussion Papers, No. 14000, Institute of Labor Economics (IZA), Bonn. Huy, V.Q. 2012. “Determinants of Educational Expenditure in Vietnam.” International Journal of Applied Economics 9(1): 59–72. Kaila, M., E. Nix, and K. Riukula. 2021. “Disparate Impacts of Job Loss by Parental Income and Implications for Intergenerational Mobility.” Working Paper No. 53. Federal Reserve Bank of Minneapolis. Minneapolis, MN, USA. Keldenich, C.,, and A. Knabe. 2018. Labor Market Responses to Partners’ Unemployment and Low-Pay Employment. Kiel, Hamburg, German: ZBW-Leibniz-Informationszentrum Wirtschaft. Kohara, M. 2010. “The Response of Japanese Wives’ Labor Supply to Husbands’ Job Loss.” Journal of Population Economics 23(4): 1133–1149. Levy, S., and N. Schady. 2013. “Latin America’s Social Policy Challenge: Education, Social Insurance, Redistribution.” Journal of Economic Perspectives 27(2): 193–218. Lundberg, S. 1985. “The Added Worker Effect.” Journal of Labor Economics 3(1): 11–37. Maloney, T. 1987. “Employment Constraints and the Labor Supply of Married Women: A Reexamination of the Added Worker Effect. “Journal of Human Resources 22(1): 51–61. Maloney, T. 1991. “Unobserved Variables and the Elusive Added Worker Effect.” Economica 58(230): 173–187. Maloney, W.F. 2004. “Informality Revisited.” World development 32(7): 1159–1178. Marchionni, M., L. Gasparini, and M. Edo 2019. “Brechas de género en américa latina. un estado de situación.” Caracas: CAF. http://scioteca.caf.com/handle/123456789/1401. Martinoty, L. 2015. “Intra-Household Coping Mechanisms in Hard Times: The Added Worker Effect in the 2001 Ar- gentine Economic Crisis.” SSRN. https://ssrn.com/abstract=2581547 or http://dx.doi.org/10.2139/ssrn.2581547. Mattingly, M.J.,, and K.E. Smith. 2010. “Changes in Wives’ Employment When Husbands Stop Working: A Recession- Prosperity Comparison.” Family Relations 59(4): 343–357. Neidhöfer, G., M. Ciaschi, L. Gasparini, and J. Serrano. 2023. “Social Mobility and Economic Development.” Journal of Economic Growth, 1–33. The World Bank Economic Review 579 Ortigueira, S., and N. Siassi. 2013. “How Important Is Intra-Household Risk Sharing for Savings and Labor Supply?” Journal of Monetary Economics 60(6): 650–666. Parker, S.W., and E. Skoufias. 2004. “The Added Worker Effect over the Business Cycle: Evidence from Urban Mexico.” Applied Economics Letters 11(10): 625–630. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Paz, J.A. 2009. “El efecto del trabajador adicional: evidencia para Argentina (2003–2007).” Cuadernos de economía 46(134): 225–41. Qian, J.X., and R. Smyth. 2011. “Educational Expenditure in Urban China: Income Effects, Family Characteristics and the Demand for Domestic and Overseas Education.” Applied Economics 43(24): 3379–94. Rege, M., K. Telle, and M. Votruba. 2011. “Parental Job Loss and Children’s School Performance.” Review of Eco- nomic Studies 78(4): 1462–89. Ruiz-Valenzuela, J. 2020. “Job Loss at Home: Children’s School Performance during the Great Recession.” SERIEs 11(3): 243–286. ———. 2021. “The Effects of Parental Job Loss on Children’s Outcomes.” In Oxford Research Encyclopedia of Economics and Finance, https://doi.org/10.1093/acrefore/9780190625979.013.654. Schmidpeter, B. 2020. “The Long-Term Labor Market Effects of Parental Unemployment.” Ruhr Economic Paper No. 866. Leibniz Institut für Wirtschaftsforschung. Bochum, Germany. SEDLAC 2022. Socio-Economic Database for Latin America and the Caribbean. CEDLAS and The World Bank. La Plata, Argentina. Serrano, J., L. Gasparini, M. Marchionni, and P. Glüzmann. 2019. “Economic Cycle and Deceleration of Female Labor Force Participation in Latin America.” Journal for Labour Market Research 53(1): 1–21. Skoufias, E., and S. W. Parker. 2006. “Job Loss and Family Adjustments in Work and Schooling during the Mexican Peso Crisis.” Journal of Population Economics 19, 163–181. Spletzer, J.R. 1997. “Reexamining the Added Worker Effect.” Economic Inquiry 35(2): 417–427. Stephens, Jr, M. 2002. “Worker Displacement and the Added Worker Effect.” Journal of Labor Economics 20(3): 504–37. Tornarolli, L., D. Battistón, L. Gasparini, and P. Gluzmann 2014. “Exploring Trends in Labor Informality in Latin America, 1990–2010.” Working Paper No. 159. CEDLAS.La Plata, Argentina. Ulyssea, G. 2020. “Informality: Causes and Consequences for Development.” Annual Review of Economics 12, 525– 46. Woytinsky, W.S. 1940. “Additional Workers on the Labor Market in Depressions: A Reply to Mr. Humphrey.” Journal of Political Economy 48(5): 735–39. Wu, C., and D. Krueger. 2021. “Consumption Insurance against Wage Risk: Family Labor Supply and Optimal Pro- gressive Income Taxation.” American Economic Journal: Macroeconomics 13(1): 79–113. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Supplementary Online Appendix Job Loss and Household Labor Supply Adjustments in Developing Countries: Evidence from Argentina Matías Ciaschi and Guido Neidhöfer S1: Descriptives and Results Table S2.1. Encuesta Permanente de Hogares (EPH) Rotative Panel Structure Panel A: Encuesta Permanente de Hogares 1995–2002. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Year 1 Year 2 Year 3 Year 4 May October May October May October May October X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X Panel B : Encuesta Permanente de Hogares 2003–2015. Year 1 Year 2 Year 3 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 X X X X X X X X X X X X X X X X X X X X X X X X X X X X X Source: Own elaboration. Note: Each “X” represents a household observation. Highlighted observations correspond to a panel: every household is surveyed four times. A. Descriptives and Sampling A.1 Job loss and initial incomes and employment characteristics Figure S2.1. Job Loss Probability by Initial Household Income and Head’s Job Formality. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Source: EPH, own estimates. Note: 90 percent confidence intervals. Estimates for regressions of a dummy of job loss on a vector of individual and household-level control variables, region-specific time fixed effects, and individual fixed effects. The upper (lower) panel shows the predicted relationship between job loss and income deciles (male household head’s formality) in the initial period. Standard errors clustered at the household-level. From left to right, each figure were obtained using a sample of: (1) female spouses not participating in the labor market in the initial period; (2) households with at least one son or daughter not participating in the labor market in the initial period; (3) households with at least one son or daughter enrolled in any educational level in the initial period, respectively. A.2 Job losses by gender and male head’s formality Figure S2.2. Job Loss Probability by Gender and Head’s Job Formality. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Source: EPH, own estimates. Note: 95 percent confidence intervals. Job loss estimates were computed using a sample of both men and women employed in the first period. Estimates by male head’s formality consider a sample of initially employed male heads. A.3 Sample A: Female spouses not participating in the labor market in the initial period Table S2.2. Sample A: Main Variables Descriptive Statistics Full sample Employed heads With job loss Without job loss Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Mean Mean Mean Mean Male head’s education (years) 9.59 (3.72) 9.68 (3.70) 8.31 (3.52) 9.87 (3.69) Wives’ education (years) 9.48 (3.51) 9.59 (3.49) 8.34 (3.35) 9.76 (3.47) Male head’s age 45.18 (10.95) 44.32 (10.65) 46.47 (11.77) 44.02 (10.45) Wives’ age 42.47 (11.18) 41.65 (10.89) 43.56 (11.97) 41.38 (10.70) Household income (log) 5.68 (2.06) 5.80 (1.85) 5.34 (2.03) 5.86 (1.81) At least one child (%) 89.91 (30.12) 91.01 (28.60) 88.42 (32.01) 91.38 (28.07) Num. of children 2.26 (1.55) 2.30 (1.53) 2.37 (1.72) 2.30 (1.50) Regional unemployment (%) 14.80 ( 4.23) 14.70 (4.20) 16.01 (4.65) 14.52 (4.10) Regional male Unemployment (%) 14.07 (4.51) 13.97 (4.48) 15.37 (4.94) 13.77 (4.37) Observations 109,984 96,782 12,286 84,496 Source: EPH, own estimates. Note: Standard errors in parentheses. First column represents the entire sample of household with female spouses not participating in the labor market in the initial period. Second column restricts the first column’s sample to households where the head was employed in the initial period. The last two columns divide the second column based on whether the household experienced a job loss or not. Second column coincides with table 1 as it represents the working sample. A.4 Sample B: Female spouses always employed Table S2.3. Sample B: Main Variables Descriptive Statistics Full sample Employed heads With job loss Without job loss Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Mean Mean Mean Mean Male head’s education (years) 11.10 (4.08) 11.31 (4.05) 9.10 (3.80) 11.58 (4.00) Wives’ education (years) 11.74 (4.03) 11.95 (3.97) 9.96 (4.03) 12.20 (3.89) Male head’s age 44.48 (9.89) 44.02 (9.69) 46.14 (10.62) 43.75 (9.54) Wives’ age 41.92 (9.51) 41.49 (9.33) 43.27 (10.01) 41.26 (9.22) Household income (log) 6.40 (1.78) 6.51 (1.67) 5.80 (1.87) 6.60 (1.63) At least one child (%) 85.13 (35.58) 85.51 (35.20) 86.45 (34.23) 85.39 (35.32) Num. of children 1.93 (1.36) 1.94 (1.35) 2.09 (1.56) 1.92 (1.32) Regional unemployment (%) 14.11 (3.91) 13.99 (3.84) 15.46 (4.49) 13.80 (3.71) Regional male Unemployment (%) 13.34 (4.20) 13.21 (4.13) 14.76 (4.81) 13.02 (4.00) Observations 105,342 93,054 10,824 82,230 Source: EPH, own estimates. Note: Standard errors in parentheses. First column represents the entire sample of household with female spouses always employed. Second column restricts the first column’s sample to households where the head was employed in the initial period. The last two columns divide the second column based on whether the household experienced a job loss or not. Second column coincides with table 1 as it represents the working sample. A.5 Sample C: At least one child not participating in the labor market in the initial period Table S2.4. Sample C, main variables descriptive statistics. Full sample Employed heads With job loss Without job loss Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 mean mean mean mean Male head’s education (years) 10.19 (4.06) 10.36 (4.07) 8.72 (3.78) 10.57 (4.05) Wives’ education (years) 10.24 (3.99) 10.38 (3.96) 9.06 (3.87) 10.56 (3.94) Male head’s age 49.62 (6.68) 49.25 (6.54) 50.97 (7.01) 49.02 (6.44) Wives’ age 46.96 (6.53) 46.67 (6.43) 47.88 (7.04) 46.50 (6.33) Household income (log) 5.90 (1.91) 6.00 (1.83) 5.60 (1.93) 6.06 (1.81) At least one child (%) 100 100 100 100 Num. of children −3.08 (1.60) −3.07 (1.55) −3.23 (1.79) −3.05 (1.52) Regional Unemployment (%) 14.47 (4.12) 14.35 (4.06) 15.35 (4.58) 14.22 (3.97) Regional male Unemployment (%) 13.72 (4.42) 13.59 (4.35) 14.66 (4.89) 13.45 (4.26) Observations 87,949 76,819 9088 67,731 Source: EPH, own estimates. Note: Standard errors in parentheses. First column represents the entire sample of households with at least one son or daughter not participating in the labor market in the initial period. Second column restricts the first column’s sample to households where the head was employed in the initial period. The last two columns divide the second column based on whether the household experienced a job loss or not. Second column coincides with table 1 as it represents the working sample. A.6 Sample D: At least one child was employed Table S2.5. Sample D, Main Variables Descriptive Statistics Full sample Employed heads With job loss Without job loss Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Mean Mean Mean Mean Male head’s education (years) 9.29 (3.87) 9.42 (3.87) 8.15 (3.55) 9.61 (3.88) Wives’ education (years) 9.42 (3.76) 9.53 (3.75) 8.54 (3.56) 9.68 (3.76) Male head’s age 51.32 (6.88) 50.84 (6.78) 52.52 (6.69) 50.59 (6.75) Wives’ age 48.71 (6.83) 48.33 (6.75) 49.46 (7.10) 48.16 (6.68) Household income (log) 6.15 (1.72) 6.25 (1.63) 5.81 (1.78) 6.31 (1.60) At least one child (%) 100 100 100 100 Num. of children −2.61 (1.41) −2.64 (1.41) −2.67 (1.63) −2.64 (1.37) Regional unemployment (%) 14.50 (3.92) 14.33 (3.84) 15.44 (4.31) 14.17 (3.74) Regional male unemployment (%) 13.74 (4.20) 13.56 (4.12) 14.79 (4.56) 13.38 (4.01) Observations 23,214 19,590 2619 16,971 Source: EPH, own estimates. Note: Standard errors in parentheses. First column represents the entire sample of households with at least one son or daughter who was employed. Second column restricts the first column’s sample to households where the head was employed in the initial period. The last two columns divide the second column based on whether the household experienced a job loss or not. Second column coincides with table 1 as it represents the working sample. A.7 Sample E: At least one child enrolled in education in the initial period Table S2.6. Sample E, Main Variables Descriptive Statistics Full sample Employed heads With job loss Without job loss Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Mean Mean Mean Mean Male head’s education (years) 10.68 (4.01) 10.81 (4.00) 9.14 (3.83) 11.01 (3.98) Wives’ education (years) 10.76 (3.86) 10.86 (3.84) 9.61 (3.80) 11.01 (3.81) Male head’s age 49.492 (6.48) 49.16 (6.35) 50.61 (6.92) 48.99 (6.26) Wives’ age 46.86 (6.25) 46.63 (6.17) 47.46 (6.68) 46.53 (6.10) Household income (log) 6.00 (1.90) 6.08 (1.85) 5.62 (1.97) 6.13 (1.83) At least one child (%) 100 100 100 100 Num. of children −2.98 (1.48) −2.99 (1.45) −3.13 (1.67) −2.97 (1.42) Regional unemployment (%) 14.61 (4.14) 14.49 (4.09) 15.70 (4.59) 14.35 (4.00) Regional male Unemployment (%) 13.86 (4.42) 13.74 (4.37) 15.03 (4.89) 13.58 (4.28) Observations 80,721 71,862 7835 64,027 Source: EPH, own estimates. Note: Standard errors in parentheses. First column represents the entire sample of household with at least one son or daughter enrolled in any educational level in the initial period. Second column restricts the first column’s sample to households where the head was employed in the initial period. The last two columns divide the second column based on whether the household experienced a job loss or not. Second column coincides with table 1 as it represents the working sample. B. Results by Time Windows B.1 1995–1998 period Table S2.7. Estimates on Female Labor Participation Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Panel A: Baseline estimates. Change into Labor par- Change into downgrad- ticipation Employed Hours Informality informality Downgrading ing (1) (2) (3) (4) (5) (6) (7) HH head job loss 0.166∗∗∗ 0.140∗∗∗ 0.731 (1.423) 0.027 0.049 0.069∗∗ 0.090∗∗ (0.029) (0.026) (0.035) (0.030) (0.029) (0.036) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 16,224 16,224 10,913 8048 5226 7306 5074 Average .56 .56 34.21 .35 .24 .33 .27 Panel B: Considering interactive decisions. Labor par- Employed Hours Informality Change into Change into ticipation informality Downgrading downgrad- ing (1) (2) (3) (4) (5) (6) (7) ∗∗∗ ∗∗∗ HH head job loss = 1 0.150 0.114 0.358 (1.535) 0.034 0.060∗∗ 0.075∗∗ 0.102∗∗∗ (0.032) (0.027) (0.037) (0.030) (0.031) (0.038) HH head job 0.079 0.159∗∗ 3.393 (2.712) −0.065 −0.103 −0.091∗∗ −0.135∗∗∗ loss = 1 × Child LP = 1 (0.062) (0.066) (0.100) (0.103) (0.040) (0.046) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 16,194 16,194 10,899 8038 5221 7301 5072 Average .56 .56 34.21 .35 .24 .33 .27 Source: EPH, own estimates. Note: Column (1) measures the effect of male household head’s job loss on the labor force participation of their wives. Column (2) measures the effect on finding a job. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality. Column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses, while Panel B only includes those with children. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in column (1) and (2) it shows the average probability of being out of labor force. Source: EPH, own estimates. Table S2.8. Estimates on Children Labor Participation Panel A: Baseline estimates. Daughters Sons Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.176∗ (0.094) −1.707 0.250∗∗∗ 0.135∗ (0.074) −0.717 0.116∗ ∗ (2.035) (0.075) (4.434) (0.049) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 5528 1115 5158 4437 1944 4550 Average .59 12.24 .65 .45 16.53 .55 Panel B: Considering interactive decisions. Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.235∗∗ (0.106) −3.771∗∗ 0.212∗∗ 0.300∗∗∗ (0.114) 2.864 (4.070) 0.162∗∗∗ (1.541) (0.090) (0.052) HH head job −0.035 (0.170) 2.679 (2.600) −0.048 0.002 (0.216) −4.306 0.014 (0.114) loss = 1 × Mother LP = 1 (0.107) (7.960) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 3346 598 2982 2635 1216 2536 Average .61 11.84 .64 .43 16.75 .51 Panel C: Considering interactive decisions (in hours worked). Daughters Sons labor Hours Drop-out labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.085 (0.156) 0.008 (3.990) 0.235∗∗ −0.016 (0.134) −7.993 0.068 (0.068) (0.093) (6.242) HH head job 0.002 (0.007) −0.045 −0.005 −0.007∗∗ (0.003) −0.047 −0.003 loss = 1 × Mother hrs. (0.168) (0.004) (0.163) (0.002) change Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 2182 517 2176 1802 728 2014 Average .58 12.13 .65 .47 14.84 .61 Source: EPH, own estimates. Note: Columns (1) and (4) measure the effect of male household head’s job loss on the labor force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Panel A considers all children, Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force while in columns (3) and (6) it shows the average educational attendance. B.2 1999–2003 period Table S2.9. Estimates on Female Labor Participation Panel A: Baseline estimates. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor par- Change into Change into ticipation Employed Hours Informality informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) HH head job loss 0.148∗∗∗ 0.110∗∗∗ 1.456∗ 0.025∗∗ 0.028∗∗ 0.056∗∗∗ 0.059∗∗ (0.018) (0.017) (0.828) (0.011) (0.013) (0.022) (0.025) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 21,988 21,988 16,856 12,712 8290 9778 6815 Average .54 .54 34.29 .35 .22 .38 .32 Panel B: Considering interactive decisions. Labor par- Employed Hours Informality Change into Change into ticipation informality Downgrading downgrad- ing (1) (2) (3) (4) (5) (6) (7) 0.160∗∗∗ 0.113∗∗∗ 1.645∗ 0.021∗∗ 0.022∗∗ 0.060∗∗ 0.064∗∗ HH head job loss = 1 (0.019) (0.018) (0.900) (0.009) (0.010) (0.024) (0.027) HH head job −0.108∗∗∗ −0.024 −1.114 0.043 0.053 −0.044 −0.069∗ loss = 1 × Child LP = 1 (0.038) (0.038) (1.817) (0.051) (0.062) (0.034) (0.038) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 21,963 21,963 16,847 12,708 8289 9775 6814 Average .54 .54 34.29 .35 .22 .38 .32 Source: EPH, own estimates. Note: Column (1) measures the effect of male household head’s job loss on the labor force participation of their wives. Column (2) measures the effect on finding a job. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality, column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses, while Panel B only includes those with children. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in column (1) and (2) it shows the average probability of being out of labor force. Table S2.10. Estimates on Children’s Labor Participation Panel A: Baseline estimates. Daughters Sons Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.115∗∗ (0.046) −5.420 0.147∗∗∗ 0.102∗ (0.059) −0.683 0.147∗∗∗ (3.815) (0.041) (3.735) (0.040) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 8229 1337 7689 6918 2210 6978 Average .65 10.12 .72 .51 12.63 .6 Panel B: Considering interactive decisions. Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 −0.020 (0.066) 0.597 (2.849) 0.141∗∗∗ 0.020 (0.109) 2.676 (4.822) 0.145∗∗∗ (0.040) (0.045) HH head job −0.027 (0.096) 1.010 (5.746) −0.088 −0.149 (0.146) 1.388 (5.629) −0.006 loss = 1 × Mother LP = 1 (0.071) (0.080) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 4621 745 4072 3885 1357 3778 Average .64 10.86 .67 .49 13.81 .55 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.264∗∗∗ (0.084) −5.968∗ 0.097∗∗∗ −0.035 (0.123) 0.379 (4.471) 0.114∗∗∗ (3.245) (0.035) (0.038) HH head job −0.005∗ (0.003) −0.268∗∗ −0.001 0.001 (0.004) 0.194 (0.186) −0.000 loss = 1 × Mother hrs. (0.128) (0.001) (0.001) change Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 3608 592 3617 3033 853 3200 Average .68 9.3 .77 .53 11.74 .64 Source: EPH, own estimates. Note: Columns (1) and (4) measure the effect of male household head’s job loss on the labor-force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Panel A considers all children, Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force, while in columns (3) and (6) it shows the average educational attendance. B.3 2004–2008 period Table S2.11. Estimates on Female Labor Participation Panel A: Baseline estimates. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor Change into Change into participation Employed Hours Informality informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) HH head job loss 0.135∗∗∗ 0.130∗∗∗ 2.032∗ 0.034∗∗∗ 0.044∗∗∗ 0.068∗∗ 0.061∗∗ (0.026) (0.025) (1.177) (0.013) (0.016) (0.028) (0.027) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 21,668 21,668 23,957 18,247 11,645 14,718 9822 Average .44 .44 30.99 .43 .34 .48 .44 Panel B: Considering interactive decisions. Labor Employed Hours Informality Change into change into participation informality Downgrading downgrad- ing (1) (2) (3) (4) (5) (6) (7) 0.130∗∗∗ 0.125∗∗∗ 1.818 0.035∗∗ 0.045∗∗ 0.066∗∗ 0.057∗∗ HH head job loss = 1 (0.026) (0.027) (1.316) (0.014) (0.018) (0.028) (0.026) HH head job 0.022 0.024 1.513 0.000 0.003 0.016 0.027 loss = 1 × Child LP = 1 (0.074) (0.061) (3.166) (0.019) (0.022) (0.055) (0.056) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual Yes Yes Yes Yes Yes Yes Yes FE Observations 21,650 21,650 23,937 18,233 11,639 14,710 9818 Average .44 .44 30.99 .43 .34 .48 .44 Source: EPH, own estimates. Note: Column (1) measures the effect of male household head’s job loss on the labor force participation of their wives. Column (2) measures the effect on finding a job. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality, column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses, while Panel B only includes those with children. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in column (1) and (2) it shows the average probability of being out of labor force. Table S2.12. Estimates on Children Labor Participation Panel A: Baseline estimates. Daughters Sons Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.069 (0.060) −1.520 0.130∗∗∗ −0.025 (0.063) −0.290 0.168∗∗∗ (2.685) (0.043) (2.648) (0.048) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 10,033 2003 9363 8741 3412 8928 Average .63 10.16 .67 .5 14.35 .59 Panel B: Considering interactive decisions. Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.182 (0.137) 2.336 (4.683) 0.096∗∗ −0.013 (0.107) −5.305 0.127∗∗∗ (0.041) (4.655) (0.035) HH head job −0.287∗∗ (0.143) 14.611∗∗ 0.089 (0.102) −0.094 (0.142) −4.281 0.124 (0.088) loss = 1 × Mother LP = 1 (7.044) (9.028) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 4746 846 4114 4233 1680 4079 Average .63 10.28 .64 .5 14.76 .58 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.112 (0.090) 0.529 (4.028) 0.117∗∗∗ −0.057 (0.118) 6.703∗ (3.769) 0.112∗∗∗ (0.044) (0.033) HH head job −0.004∗ (0.002) 0.151 (0.182) 0.003∗ 0.002 (0.003) 0.199 (0.158) −0.000 loss = 1 × Mother hrs. (0.001) (0.001) change Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 5284 1155 5246 4505 1732 4848 Average .62 10.42 .69 .5 14.21 .61 Source: EPH, own estimates. Note: Columns (1) and (4) measure the effect of male household head’s job loss on the labor-force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Panel A considers all children, Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force, while in columns (3) and (6) it shows the average educational attendance. B.4 2009–2015 period Table S2.13. Estimates on Female Labor Participation Panel A: Baseline estimates. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor Change into Change into participation Employed Hours Informality informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) HH head job loss 0.144∗∗∗ 0.145∗∗∗ 1.092 0.011 0.027∗∗ 0.070∗∗∗ 0.063∗∗ (0.019) (0.019) (0.795) (0.014) (0.014) (0.023) (0.026) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 36,902 36,902 41,328 31,766 22,606 28,562 20,846 Average .44 .44 31.68 .31 .22 .49 .44 Panel B: Considering interactive decisions. Labor Employed Hours Informality Change into Change into participation informality Downgrading downgrad- ing (1) (2) (3) (4) (5) (6) (7) 0.147∗∗∗ 0.149∗∗∗ 0.413 0.004 0.019 0.077∗∗∗ 0.068∗∗ HH head job loss = 1 (0.020) (0.020) (0.799) (0.014) (0.012) (0.025) (0.028) HH head job −0.048 −0.045 5.094∗ 0.057 0.065 −0.069∗∗ −0.063∗ loss = 1 × Child LP = 1 (0.046) (0.045) (2.853) (0.040) (0.053) (0.030) (0.037) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 36,858 36,858 41,286 31,736 22,589 28,546 20,838 Average .44 .44 31.68 .31 .22 .49 .44 Source: EPH, own estimates. Note: Column (1) measures the effect of male household head’s job loss on the labor force participation of their wives. Column (2) measures the effect on finding a job. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality. Column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses, while Panel B only includes those with children. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in column (1) and (2) it shows the average probability of being out of labor force. Table S2.14. Estimates on Children’s Labor Participation Panel A: Baseline estimates. Daughters Sons Labor Labor Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 participation Hours Drop-out participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.064 (0.053) −2.231 (2.364) 0.114∗∗∗ 0.051 0.001 0.163∗∗∗ (0.032) (0.048) (1.822) (0.049) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 17,398 2666 14,937 15,535 4903 14,259 Average .68 8.87 .71 .53 13.65 .58 Panel B: Considering interactive decisions. Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) ∗∗∗ 0.090 (0.067) −1.953 (4.044) 0.152 0.094 1.872 0.163∗∗∗ HH head job loss = 1 (0.040) (0.106) (3.123) (0.038) HH head job loss = 1 × Mother −0.114 (0.096) −0.865 (6.924) −0.185∗∗∗ −0.167 0.698 −0.100 LP = 1 (0.067) (0.179) (4.679) (0.089) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 8199 1196 6212 7306 2459 6062 Average .69 9.26 .65 .51 14.61 .51 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) ∗∗∗ 0.052 (0.096) −0.367 (2.672) 0.119 0.007 −3.472 0.136∗∗∗ HH head job loss = 1 (0.041) (0.098) (2.846) (0.043) HH head job loss = 1 × Mother −0.005∗ −0.211 (0.198) −0.003∗∗ −0.002 −0.134 −0.001 hrs. change (0.003) (0.001) (0.005) (0.192) (0.002) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 9196 1469 8713 8217 2431 8188 Average .69 8.95 .74 .55 13.08 .64 Source: EPH, own estimates. Note: Columns (1) and (4) measure the effect of male household head’s job loss on the labor force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Panel A considers all children, Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force, while in columns (3) and (6) it shows the average educational attendance. C. Alternative Sample Results C.1 Attrition: descriptive statistics and main estimates Table S2.15. Attriting and Non-Attriting Households: Descriptive Statistics FLP sample CLP sample CED sample Non-attriting Attriting Non-attriting Attriting Non-attriting Attriting Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Mean Mean Mean HH head job loss (%) 12.29 (32.83) 6.62 (24.86) 11.91 (32.39) 6.67 (24.95) 10.84 (31.09) 6.18 (24.08) Male head’s education (years) 9.70 (3.71) 9.97 (3.69) 10.33 (4.08) 10.47 (4.07) 10.85 (4.03) 10.99 (3.96) Wives’ education (years) 9.59 (3.49) 9.93 (3.50) 10.35 (3.98) 10.58 (3.98) 10.89 (3.86) 11.10 (3.85) Male Head’s age 43.78 (10.87) 41.00 (11.53) 48.48 (6.73) 48.20 (6.90) 48.51 (6.51) 48.31 (6.67) Wives’ age 41.11 (11.11) 38.11 (11.69) 45.89 (6.56) 45.55 (6.60) 45.94 (6.28) 45.70 (6.37) Household income (log) 5.81 (1.86) 5.81 (2.21) 6.00 (1.83) 6.01 (2.25) 6.08 (1.86) 6.14 (2.20) Formal head (%) 75.12 (43.24) 71.56 (45.11) 78.41 (41.15) 77.95 (41.46) 80.44 (39.67) 79.75 (40.19) Regional unemployment (%) 15.05 (4.13) 15.56 (4.67) 14.70 (4.04) 15.32 (4.66) 14.87 (4.03) 15.41 (4.65) Regional male unemployment 14.23 (4.23) 14.79 (4.89) 13.88 (4.18) 14.56 (4.93) 14.06 (4.17) 14.65 (4.91) (%) Households (initial period, %) 47.50 52.50 52.44 47.56 52.34 47.66 Source: EPH, own estimates. Note: Standard errors in parentheses. Estimates were obtained using the same samples as table 2 (FLP Sample) and table 3 (CLP sample and CED sample). C.2 Households affected by job losses Table S2.16. Estimates on Female Labor Participation, Sample with Attrition Panel A: Baseline estimates Labor Change into Change into participation Employed Hours Informality informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 HH head job loss 0.136∗∗∗ 0.117∗∗∗ 1.400∗∗∗ 0.028∗∗∗ 0.037∗∗∗ 0.065∗∗∗ 0.071∗∗∗ (0.009) (0.009) (0.463) (0.007) (0.008) (0.012) (0.014) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 144,409 144,409 143,311 109,858 55,776 96,421 49,449 Average .46 .46 32.11 .35 .26 .47 .42 Panel B: Considering interactive decisions. Labor Employed Hours Informality Change into Change into participation informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) ∗∗∗ ∗∗∗ ∗∗ 0.136 0.113 1.178 0.024∗∗∗ 0.035∗∗∗ 0.064∗∗∗ 0.072∗∗∗ HH head job loss = 1 (0.010) (0.010) (0.495) (0.007) (0.008) (0.013) (0.015) HH head job −0.018 0.017 1.752 0.031 0.012 0.013 −0.013 (0.028) loss = 1 × Children’s (0.023) (0.022) (1.340) (0.022) (0.022) (0.030) LFP = 1 Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 144,239 144,239 143,183 109,765 55,741 96,370 49,430 Average .46 .46 32.11 .35 .26 .47 .42 Source: EPH, own estimates. Note: Estimates including nonattriting and attriting households, i.e., households observed four times and households observed less than four times in the survey. Column (1) measures the effect of male household head’s job loss on the labor force participation of their wives. Column (2) measures the effect on finding a job. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality, column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses, while Panel B only includes those with children. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; column (1) shows the average probability of being out of labor force. C.3 Household composition Table S2.17. Estimates on Children’s Labor Participation, Sample with Attrition Panel A: Baseline estimates Daughters Sons Labor Labor Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 participation Hours Drop-out participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.091∗∗∗ −1.597 0.129∗∗∗ 0.076∗∗∗ −0.183 0.152∗∗∗ (0.026) (1.140) (0.019) (0.027) (1.241) (0.022) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 57,894 10,000 52,573 50,484 17,474 49,378 Average .65 9.84 .69 .51 13.93 .58 Panel B: Considering interactive decisions. Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) ∗∗∗ ∗∗∗ ∗∗∗ 0.123 −0.874 0.137 0.137 0.057 (1.685) 0.149∗∗∗ HH head job loss = 1 (0.034) (1.440) (0.020) (0.044) (0.020) HH head job loss = 1 × Mother’s −0.130∗∗ 4.106 −0.080∗∗ −0.136∗ −3.026 0.019 LFP = 1 (0.060) (2.715) (0.038) (0.073) (2.854) (0.043) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 29,124 4715 24,323 25,319 9264 23,211 Average .66 10.07 .65 .49 14.6 .53 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) ∗∗ ∗∗ ∗∗∗ 0.095 −4.056 0.114 −0.038 −0.390 0.119∗∗∗ HH head job loss = 1 (0.047) (1.770) (0.022) (0.049) (1.946) (0.021) HH head job loss = 1 × Mother’s −0.001 −0.131∗ −0.001 0.002 0.052 (0.096) −0.000 hrs. change (0.002) (0.074) (0.001) (0.002) (0.001) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 28,762 5281 28,230 25,142 8196 26,155 Average .65 9.85 .72 .53 13.42 .63 Source: EPH, own estimates. Note: Estimates including nonattriting and attriting households, i.e., households observed four times and households observed less than four times in the survey. Columns (1) and (4) measure the effect of male household head’s job loss on the labor-force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Panel A considers all children, while Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force while in columns (3) and (6) it shows the average educational attendance. C.4 Fixed Effects by household interview Table S.2.18. Estimates on Female Labor Participation, Job Loss Sample Panel A: Baseline estimates. Labor Change into Change into participation Employed Hours Informality informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 HH head job loss 0.153∗∗∗ 0.132∗∗∗ 1.102∗∗ 0.027∗∗∗ 0.037∗∗∗ 0.073∗∗∗ 0.074∗∗∗ (0.011) (0.010) (0.484) (0.007) (0.007) (0.013) (0.013) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 12,286 12,286 10,824 8239 6010 4730 3625 Average .5 .5 32.57 .5 .39 .55 .51 Panel B: Considering interactive decisions. Labor Employed Hours Informality Change into Change into participation informality Downgrading downgrad- ing (1) (2) (3) (4) (5) (6) (7) 0.159∗∗∗ 0.132∗∗∗ 1.110∗∗ 0.028∗∗∗ 0.041∗∗∗ 0.077∗∗∗ 0.080∗∗∗ HH head job loss = 1 (0.012) (0.011) (0.514) (0.008) (0.008) (0.013) (0.014) HH head job −0.084∗∗ −0.014 −1.571 −0.020 −0.045 −0.073∗∗ −0.096∗∗ loss = 1 × Children’s LFP = 1 (0.033) (0.035) (1.948) (0.031) (0.035) (0.034) (0.038) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 12,266 12,266 10,809 8229 6004 4727 3624 Average .5 .5 32.57 .5 .39 .55 .51 Source: EPH, own estimates. Note: Estimates for a sample of households in which the head lost his job in any of the time periods in the analysis. Column (1) measures the effect of male household head’s job loss on the labor force participation of their wives. Column (2) measures the effect on finding a job. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality, column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses, while Panel B only includes those with children. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in column (1) and (2) it shows the average probability of being out of labor force. C.5 The effects of both father and mother job loss on children outcomes Table S2.19. Estimates on Children’s Labor Participation, Job Loss Sample Panel A:Baseline estimates. Daughters Sons Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.092∗∗∗ (0.026) −2.737∗ 0.164∗∗∗ 0.063∗∗ (0.030) −0.675 0.152∗∗∗ (1.648) (0.021) (1.555) (0.022) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 5149 946 4224 3939 1673 3611 Average .61 10.95 .61 .42 14.56 .47 Panel B: Considering interactive decisions. Daughters Sons Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.171∗∗∗ (0.039) −3.943∗∗ 0.150∗∗∗ 0.115∗ (0.067) 0.764 (3.344) 0.153∗∗∗ (1.804) (0.020) (0.021) HH head job −0.237∗∗∗ (0.087) 8.269∗∗∗ −0.059 −0.211 (0.137) −2.037 0.031 (0.052) loss = 1 × Mother’s (3.048) (0.051) (4.161) LFP = 1 Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 2671 470 1996 2069 933 1781 Average .61 12.07 .55 .41 15.71 .43 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.118∗∗ (0.046) −3.459∗ 0.141∗∗∗ −0.048 (0.097) −5.585∗∗ 0.145∗∗∗ (1.935) (0.025) (2.694) (0.024) HH head job −0.002 (0.002) −0.300∗∗ −0.001 0.005∗ (0.003) 0.147∗ (0.077) −0.001∗ loss = 1 × Mother’s (0.127) (0.001) (0.001) hrs. change Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 2478 476 2228 1870 740 1830 Average .62 10.54 .63 .44 14.07 .49 Source: EPH, own estimates. Note: Estimates for a sample of households in which the head lost his job in any of the time periods in the analysis. Columns (1) and (4) measure the effect of male household head’s job loss on the labor force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Panel A considers all children, while Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force while in columns (3) and (6) it shows the average educational attendance. Table S2.20. Estimates on Children Labor Participation Panel A: Baseline estimates. Daughters Sons Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.101∗∗∗ (0.032) −3.790 (2.681) 0.151∗∗∗ (0.023) 0.090∗∗∗ (0.032) 0.849 (2.227) 0.146∗∗∗ (0.026) Controls Yes Yes Yes Yes Yes Yes Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 26,267 2407 24,777 24,552 5176 24,160 Average .81 6.98 .8 .70 10.26 .72 Panel B: Considering interactive decisions. Daughters Sons Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) 0.118∗∗ −2.978 0.148∗∗∗ 0.121∗∗ −2.244 0.155∗∗∗ (0.046) (4.910) (0.026) (0.060) (4.805) (0.024) HH head job loss = 1 × Mother LP = 1 −0.090 (0.085) 16.306∗∗ (7.217) −0.068 (0.044) −0.188∗∗ (0.093) 5.434 (5.404) 0.002 (0.057) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 12,977 1108 11,614 12,258 2815 11,619 Average .82 7.16 .76 .67 11.03 .67 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) 0.137∗∗ −2.174 0.108∗∗∗ −0.010 −0.785 0.106∗∗∗ (0.057) (3.661) (0.029) (0.064) (3.123) (0.026) HH head job loss = 1 × Mother hrs. −0.003 (0.002) 0.153 (0.160) −0.002∗ (0.001) −0.000 (0.002) −0.188 (0.139) −0.001 (0.001) change Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 13,284 1298 13,151 12,290 2360 12,531 Average .79 7.12 .81 .71 9.64 .75 Source: EPH, own estimates. Note: Sample of children aged 16 to 20. Columns (2) and (5) consider a sample including children employed in the current period. Panel A considers all children, while Panel B only includes children whose mothers were not participating in the labor market in the initial period and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force while in columns (3) C.6 and (6)Working hours, it shows the average informality educational attendance. and occupation change Table S2.21. Estimates on Female Labor Participation Panel A—Baseline estimates. Labor Change into Change into participation Employed Hours Informality informality Downgrading downgrading (1) (2) (3) (4) (5) (6) (7) Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 HH head job loss 0.148∗∗∗ 0.130∗∗∗ 1.321∗∗ 0.025∗∗∗ 0.037∗∗∗ 0.070∗∗∗ 0.071∗∗∗ (0.011) (0.010) (0.513) (0.008) (0.008) (0.013) (0.014) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 96,782 96,782 93,054 70,773 55,776 60,364 49,449 Average .47 .47 32.03 .36 .26 .46 .42 Panel B: Considering interactive decisions Labor Employed Hours Informality Change into Downgrading Change into participation informality downgrading (1) (2) (3) (4) (5) (6) (7) 0.149∗∗∗ 0.127∗∗∗ 1.119∗∗ 0.023∗∗∗ 0.035∗∗∗ 0.071∗∗∗ 0.072∗∗∗ (0.011) (0.011) (0.550) (0.008) (0.008) (0.014) (0.015) HH HH head head job loss = job loss =11 × Child −0.026 0.014 1.450 (1.484) 0.018 (0.020) 0.013 (0.022) −0.021 −0.014 LP = 1 (0.026) (0.025) (0.025) (0.028) Controls Yes Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes No No Quarter × Region FE Yes Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Yes Observations 96,665 96,665 92,969 70,715 55,741 60,332 49,430 Average .47 .47 32.03 .36 .26 .46 .42 Source: EPH, own estimates. Note: Estimates including fixed effects by household interview. Column (1) measures the effect of male household head’s job loss on the labor-force participation of their wives. Column (2) measures the effect on finding a job. Column (3) measures the effect on weekly working hours among employed women. Column (4) measures the effect on informality. Column (5) on switching from a formal to an informal job. Column (6) measures the effect on employment downgrading (i.e., women working in a job for which they are overqualified). Column (7) measures the effect on switching to a “downgrading” employment. Panel A considers all spouses, while Panel B only includes those with children. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in column (1) and (2) it shows the average probability of being out of labor force. C.7 The effect on sons and daughters by educational background Table S2.22. Estimates on Children’s Labor Participation Panel A: Baseline estimates Daughters Sons Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Labor participation Hours Drop-out Labor participation Hours Drop-out (1) (2) (3) (4) (5) (6) HH head job loss 0.096∗∗∗ (0.029) −1.900 0.139∗∗∗ 0.060∗∗ (0.029) −0.368 0.146∗∗∗ (1.245) (0.020) (1.338) (0.024) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 41,188 7121 37,147 35,631 12,469 34,715 Average .65 9.84 .69 .51 14.02 .58 Panel B: Considering interactive decisions. Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.051 (0.044) −0.706 0.143∗∗∗ 0.033 (0.052) 1.067 (2.097) 0.148∗∗∗ (1.682) (0.023) (0.021) HH head job −0.060 (0.068) 3.756 (2.956) −0.082∗∗ −0.127∗ (0.075) −0.380 −0.003 loss = 1 × Mother LP = 1 (0.040) (3.138) (0.045) Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 20,912 3385 17,380 18,059 6712 16,455 Average .65 10.23 .65 .49 14.83 .53 Panel C: Considering interactive decisions (in hours worked). Daughters Sons Labor Hours Drop-out Labor Hours Drop-out participation participation (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.134∗∗ (0.052) −3.590∗ 0.120∗∗∗ −0.028 (0.057) −1.266 0.123∗∗∗ (1.852) (0.024) (1.927) (0.023) HH head job −0.004∗ (0.002) −0.104 −0.002 −0.000 (0.002) 0.094 (0.106) −0.001 loss = 1 × Mother hrs. (0.105) (0.001) (0.001) change Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 20,270 3733 19,752 17,557 5744 18,250 Average .66 9.80 .72 .53 13.42 .63 Source: EPH, own estimates. Note: Estimates including fixed effects by household interview. Columns (1) and (4) measure the effect of male household head’s job loss on the labor force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Panel A considers all children, Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force while in columns (3) and (6) it shows the average educational attendance. Table S2.23. Estimates on Children’s Labor Participation Daughters Sons Labor participation Hours Drop-out Labor participation Hours Drop-out Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 (1) (2) (3) (4) (5) (6) ∗∗ ∗∗∗ ∗∗ HH head job loss 0.102 (0.047) −2.861 0.156 0.115 (0.064) −1.263 0.128∗∗∗ (1.928) (0.030) (2.104) (0.036) Mother job loss 0.015 (0.026) 0.341 (2.433) 0.057∗∗∗ −0.020 (0.026) −0.452 0.012 (0.021) (0.019) (1.832) HH head job −0.039 (0.105) −10.416 −0.106 0.071 (0.090) 1.405 (2.899) 0.016 (0.090) loss × Mother job (6.337) (0.065) loss Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 19,315 3525 18,638 16,707 5375 17,236 Average .65 9.67 .73 .53 13.32 .63 Source: EPH, own estimates. Note: Estimates from a sample of households where both mother and father were employed in the initial period. Columns (2) and (5) consider a sample including children employed in the current period. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force while in columns (3) and (6) it shows the average educational attendance. Table S2.24. Estimates of Working Hours for Women Changing Into Informality 5-digit level 3-digit level 1-digit level (1) (2) (3) HH head job loss = 1 −1.261 (1.066) −1.415 (1.025) −0.847 (0.979) Occup. change = 1 −0.443 (0.679) −0.150 (0.661) −1.324 (0.844) HH head job 5.588∗∗ (2.455) 6.776∗∗∗ (2.586) 9.535∗∗∗ (2.853) loss = 1 × Occup. change = 1 Controls Yes Yes Yes Year FE Yes Yes Yes Quarter FE Yes Yes Yes Quarter × Region FE Yes Yes Yes Individual FE Yes Yes Yes Observations 12,303 12,303 12,303 Average 27.43 27.43 27.43 Source: EPH, own estimates. Note: Estimates from a sample of always employed women who transition into labor informality. “Occup. change” refers to a dummy variable indicating whether the woman changed her occupation between interviews at the 1, 3, and 5-digit level of the ISCO (International Standard Classification of Occupations), in columns (1), (2) and (3), respectively. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row indicates the average of working hours in the first interview, C.8 The effect on sons and daughters’ longer-term drop-out Table S2.25. Estimates on Children’s Labor Participation By Educational Background Daughters Sons Labor participation Hours Drop-out Labor participation Hours Drop-out Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 (1) (2) (3) (4) (5) (6) HH head job loss = 1 0.091∗∗∗ (0.033) −1.767 0.159∗∗∗ 0.055∗ (0.032) 0.368 (1.380) 0.137∗∗∗ (1.475) (0.024) (0.026) Parental 0.005 (0.026) −1.494 0.001 (0.017) 0.001 (0.024) 2.715∗ (1.613) 0.010 (0.019) Secondary = 1 (1.926) HH head job 0.033 (0.058) −0.551 −0.092∗∗∗ 0.032 (0.066) −5.790 0.041 (0.052) loss = 1 × Parental (2.907) (0.033) (3.590) secondary = 1 Controls Yes Yes Yes Yes Yes Yes Year FE Yes Yes Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Yes Yes Observations 41,179 7120 37,138 35,628 12,467 34,710 Average .65 9.84 .69 .51 14.02 .58 Source: EPH, own estimates. Note: “Parental secondary” refers to a dummy variable indicating whether the more highly educated parent has completed secondary education or higher. Columns (1) and (4) measure the effect of male household head’s job loss on the labor force participation of their daughters and sons, respectively. Columns (2) and (5) measure the effect on weekly working hours among employed children. Columns (3) and (6) measure the effect on educational drop-out. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The value in the last row indicates the average of the dependent variable in the first interview; in columns (1) and (4) it shows the average probability of being out of labor force while in columns (3) and (6) it shows the average educational attendance. C.9 Public vs. private education Figure S2.3. Private Education Attendance and Job Loss Probability and by Initial Household Income. Note: Estimates using a sample of children aged 16 to 25 enrolled at any educational level in the initial period (sample E in table 1). Source: EPH, own estimates. Table S2.26. Estimates on Children’s Longer-Term Educational drop-Outs Panel A: Baseline estimates Daughters Sons Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 (1) (2) HH head job loss 0.044∗∗∗ (0.011) 0.051∗∗∗ (0.013) Controls Yes Yes Year FE Yes Yes Quarter FE Yes Yes Quarter × RegionFE Yes Yes Individual FE Yes Yes Observations 37,147 34,715 Average .69 .58 Panel B: Considering interactive decisions Daughters Sons (1) (2) HH head job loss = 1 0.051∗∗∗ (0.011) 0.052∗∗∗ (0.011) HH head job loss = 1 × Mother LP = 1 −0.031 (0.019) 0.032 (0.029) Controls Yes Yes Year FE Yes Yes Quarter FE Yes Yes Quarter × Region FE Yes Yes Individual FE Yes Yes Observations 17,380 16,455 Average .65 .53 Panel C: Considering interactive decisions (in hours worked) Daughters Sons (1) (2) HH head job loss = 1 0.042∗∗ (0.011)∗ 0.052∗∗∗ (0.014) HH head job loss = 1 × Mother hrs. change −0.001∗∗ (0.000) −0.000 (0.000) Controls Yes Yes Year FE Yes Yes Quarter FE Yes Yes Quarter × Region FE Yes Yes Individual FE Yes Yes Observations 19,752 18,250 Average .72 .63 Source: EPH, own estimates. Note: Columns measure the effect on longer-term educational drop-outs, defined as children who discontinue their education during the second or third interview and remain out of the educational system in subsequent periods. Panel A considers all children, Panel B only includes children whose mothers were not participating in the labor market in the initial period, and Panel C considers children of already working mothers. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗ ∗ p < 0.05, ∗∗∗ p < 0.01. The value in the last row shows the average educational attendance. Figure S2.4. Estimates on the Transition from Private to Public Education by Initial Household Income. Downloaded from https://academic.oup.com/wber/article/doi/10.1093/wber/lhad041/7564834 by LEGVP Law Library user on 01 August 2024 Note: Estimates using a sample of children 16 to 25 initially enrolled in private education institutions and not dropping out from education in the following interviews. Income quantiles were computed based on household income in the first interview. The figure shows the coefficients of an interaction term between father’s job loss and initial income quintile; 90 percent confidence intervals. Source: EPH, own estimates. Table S2.27. Estimates on Educational drop-Outs and Private vs. Public Education Daughters Sons High school level High school level (1) All (2) (3) All (4) HH head job loss = 1 0.119∗∗ (0.049) 0.128∗∗∗ (0.049) 0.087 (0.057) 0.157∗∗ (0.069) HH head job loss = 1 × Public 0.015 (0.066) −0.000 (0.056) 0.085 (0.078) 0.005 (0.079) Educ.=1 Controls Yes Yes Yes Yes Year FE Yes Yes Yes Yes Quarter FE Yes Yes Yes Yes Quarter × Region FE Yes Yes Yes Yes Individual FE Yes Yes Yes Yes Observations 10,197 24,553 10,564 23,406 Average .88 .69 .82 .58 Source: EPH, own estimates. Note: Estimates using a sample of children aged 16 to 18 (“high school level”) or 16 to 25 enrolled in any educational level in the initial period (sample E in table 1). “Public Educ.” refers to a dummy variable indicating whether the children was attending public education in the initial period. Robust standard errors clustering at the household level are indicated in parentheses. ∗ p < 0.10, ∗∗ p < 0.05, ∗ ∗ ∗ p < 0.01. The values in the last row show the average educational attendance. C 2024 International Bank for Reconstruction and Development / The World Bank. Published by Oxford University Press