Intergenerational Transmission of Welfare Benefit Receipt: Evidence from Germany

Riphahn, Regina T.; Feichtmayer, Jennifer Article — Published Version Intergenerational Transmission of Welfare Benefit Receipt: Evidence from Germany Review of Income and Wealth Provided in Cooperation with: John Wiley & Sons Suggested Citation: Riphahn, Regina T.; Feichtmayer, Jennifer (2024) : Intergenerational Transmission of Welfare Benefit Receipt: Evidence from Germany, Review of Income and Wealth, ISSN 1475-4991, Wiley Periodicals, Inc., Hoboken, NJ, Vol. 70, Iss. 4, pp. 1226-1251, https://doi.org/10.1111/roiw.12680 This Version is available at: https://hdl.handle.net/10419/313731 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ bs_bs_banner Review of Income and Wealth Series 70, Number 4, December 2024 DOI: 10.1111/roiw.12680 INTERGENERATIONAL TRANSMISSION OF WELFARE BENEFIT RECEIPT: EVIDENCE FROM GERMANY BY REGINA T. RIPHAHN*and JENNIFER FEICHTMAYER Friedrich-Alexander-University Erlangen-Nürnberg We study the intergenerational transmission of welfare benefit receipt in Germany. We first describe the correlation between welfare receipt experienced in the parental household and subsequent own welfare receipt of young adults. In a second step, we investigate whether the observed correlations reflect causal effects of past welfare experience. We use family fixed effects estimations and Gottschalk’s (1996) approach and take advantage of the long-running German Socio-Economic Panel Survey to contribute to a sparse literature. We find strong positive correlations between parental and own welfare receipt. These patterns do, however, not persist after controlling for unobserved heterogeneities. Therefore, our results suggest that the strong intergenerational correlation of welfare benefit receipt is determined by family background rather than by the experience of parental welfare benefit receipt. JEL Codes: I32, I38, J62, C36 Keywords: causal effect, family fixed effects, Gottschalk estimator, intergenerational mobility, social assistance, welfare 1. INTRODUCTION It is well known that parent well-being affects child well-being. Intergenerational transmission patterns are studied intensely as the transmission of disadvantage from parents to their children indicates inequality of opportunities.1 This paper investigates the intergenerational transmission of participation in means-tested minimum-income protection programs. The purpose of such welfare programs is to lift households from the most pressing economic troubles and to protect the next generation. If, however, welfare programs cause welfare receipt to be passed from generation to generation then these programs do not work properly for the young. Instead, they impose negative externalities and harm the next generation. Here, welfare reforms that reduce parental participation can be beneficial and pay off for the next generation, as well. Various mechanisms may determine the intergenerational transmission of welfare benefit receipt: after experiencing parental welfare receipt youths may be better informed about application procedures and institutional features; they may We thank Libertad González, Kristiina Huttunen, Kundu Anustup, Che-Yuan Liang, and participants of the 33rd Annual (virtual) Conference of the European Association of Labour Economists (EALE) 2021, the 77th Annual Congress of the International Institute of Public Finance 2021, and the Annual Meeting of the German Economic Association 2021 for helpful comments and suggestions. Open Access funding enabled and organized by Projekt DEAL. *Correspondence to: Regina T. Riphahn r[email protected] 1For an early survey see, for example, Black and Devereux (2011). Later contributions on the transmission of earnings, education, and place-based effects are, for example, Adermon et al. (2021), Blanden (2013), and Chetty and Hendren (2018a,2018b). © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. 1226 Review of Income and Wealth, Series 70, Number 4, December 2024 be affected by parental role models and be less subject to stigma concerns; they may know less about the labor market, and receive less parental support with respect to human capital investments or labor market networks, compared to peers who grow up without welfare. If the experience of welfare receipt in the parental household increases the next generation’s welfare receipt by any such mechanism the welfare program has negative externalities. Internationally, most studies confirm positive intergenerational correlations of welfare benefit receipt, but the evidence on causal effects is mixed. Most of the literature on the intergenerational transmission of welfare covers either the United States or Scandinavian countries such as Sweden and Norway. We are the first to offer evidence on the recent intergenerational transmission of welfare receipt for Germany which provides an interesting laboratory to study the causal transmission of welfare benefit receipt and the existence of what the literature termed a “welfare trap”. Germany is a relevant intermediate case as it ranges between the Scandinavian countries and the U.S. with respect to income inequality, the prevalence of poverty, and the generosity of minimum income protection and social spending (OECD, 2019). Germany has much lower poverty rates after taxes and transfers than the U.S. but higher ones than Sweden (Immervoll et al., 2022). The institutions of the welfare state affect the effectiveness of minimum income protection including its intergenerational effects; therefore, a comparison of intergenerational transmission effects across the different national welfare systems (e.g., following Esping-Andersen, 1990, liberal vs. conservative vs. social-democratic systems) can be informative. While the German welfare system protects the poor comparatively well, we do not know yet whether this relatively generous system also succeeds in an intergenerational perspective, that is, by protecting the next generation from inherited dependence. This is our research question. We contribute to the international literature on intergenerational welfare transmission by offering evidence from more than three decades of survey data. Long-running longitudinal data on parents and children are required to examine the transmission of welfare across generations. For most countries, such data is not available. The German Socio-Economic Panel Survey allows us to study the transmission of youth welfare experience for individuals born 1969–1991. We can consider parental welfare receipt when the youth is 10–18 years old and investigate its association with the young person’s own welfare receipt at ages 25–29. Thereby our analysis uses wider observation windows than much of the prior literature.2 In a first step of our analysis, we describe the correlation between parent and child welfare receipt. We study the correlation patterns before and after a major welfare reform which is useful to assess the sensitivity of correlation patterns to institutional change. In contrast to much of the literature which focuses on mother-daughter pairs, we compare outcomes for young men and women and separately evaluate the transmission from fathers and mothers. Our data allow 2Appendix Table A.1 characterizes the number of years of observations used in prior studies (see columns entitled “Exposure (t0)” and “Own welfare (t1)”). For parental welfare receipt, Beaulieu et al. (2005) can use 10years of observations. However, several contributions have fewer years: Antel (1992) and Levine and Zimmerman (1996) observe parental welfare receipt only for 1 year, Edmark and Hanspers (2015) use only 3 years, and Boschman et al. (2019) 2years of child welfare outcomes in adult age. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1227 Review of Income and Wealth, Series 70, Number 4, December 2024 us to describe the relevance of the age at which youths are exposed to parental welfare receipt and thus to determine the most impressionable years (Krosnick & Alwin, 1989). We look into the potential mediation effect of child educational outcomes. In a second step, we address the potential impact of unobserved heterogeneities that render parental welfare receipt endogenous to the next generation’s outcomes. In particular, parental characteristics, such as human capital, attitudes towards work and family, health, addictions, and emotional well-being may affect both generations’ welfare receipt and thus can generate spurious intergenerational welfare correlations. To account for this, we consider the empirical strategy developed by Gottschalk (1996) and apply family fixed effects estimation. The two approaches identify different effects and apply different methods to control for the potential endogeneity of parental welfare receipt. If the identifying assumptions hold, applying both methods allows us to get closer to answering the question of whether parental welfare receipt causally affects the welfare receipt of the next generation in Germany. We find a strong intergenerational correlation in welfare outcomes for our three welfare indicators. The correlations are larger for females than for males. We do not find important differences in welfare transmission from fathers versus mothers. Exposure to parental welfare receipt at the ages of 10–12 and 16–18 yields stronger correlation patterns than in the 13–15 age window. Comparing the correlation patterns before and after a major welfare reform we obtain inconclusive results and cannot confirm that intergenerational transmission declined post-reform. Both, the family fixed effects and the Gottschalk (1996) method identify causal effects under certain, yet different assumptions. In our case, both strategies fail to find evidence of a causal impact of parental welfare receipt on child welfare outcomes. Thus, in the German institutional framework, it does not appear to be the experience of parental welfare receipt that drives subsequent child welfare receipt but the correlation of individual characteristics and circumstances in the child and parent household. Levine and Zimmerman (1996) call this situation a “poverty trap” as opposed to a “welfare trap”. These results have clear policy implications as they show that it is not the character of welfare institutions themselves that leaves the offspring of welfare recipients at an elevated risk of welfare receipt. Therefore, any initiative to reduce intergenerational correlation in welfare receipt must not focus on the institutions of the welfare system but address characteristics at the individual and household level and, for example, improve human capital, health, and labor market involvement. In the next Section, we summarize the state of the literature. Section 3then provides institutional background. We outline our empirical approach in Section 4 and describe our data in Section 5. Next, we present the results of our descriptive analyses of intergenerational transmission patterns and of our causal estimates in Section 6. Finally, we draw conclusions in Section 7. 2. PRIOR LITERATURE While a broad international literature describes intergenerational correlation in welfare receipt, fewer studies identify causal effects of minimum income programs; Appendix Table A.1 offers a brief characterization of prior contributions © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1228 Review of Income and Wealth, Series 70, Number 4, December 2024 and their results. Early contributions applied structural estimation approaches and U.S. survey data (Antel, 1992; Levine & Zimmerman, 1996) with opposite results. While Antel (1992) concludes that maternal welfare use causally affects daughters’ receipt, Levine and Zimmerman (1996) find only a correlation in incomes. Gottschalk (1996) studies U.S. welfare transmission by applying event study methods and confirms a causal relationship between mothers’ and daughters’ welfare receipt. Pepper (2000) compares alternative empirical approaches and confirms a causal relationship. Hartley et al. (2022) use instrumental variables and difference-in-differences strategies based on regional heterogeneities and find that mothers’ welfare receipt increases the probability of their daughters’ welfare participation. However, welfare reforms did attenuate the transmission. Finally, Mitnik (2010) studies the intensive margin of welfare receipt; applying matching and family fixed effects estimators he does not find causal effects. There are only a few studies covering countries outside the U.S. Beaulieu et al. (2005) exploit administrative data on social assistance receipt in Quebec, Canada, and confirm causal intergenerational effects. Edmark and Hanspers (2015) apply family fixed effects estimation to Swedish register data and find no causal effects. In their study using administrative data from Norway, De Haan and Schreiner (2018) apply bounds analyses with instrumental variables and confirm significant positive causal transmission effects. Boschman et al. (2019)and Cobb-Clark et al. (2022) apply the Gottschalk (1996) approach to Dutch and Australian administrative data, respectively, and find no significant causal effects for social assistance benefits. Overall, the evidence on the causal intergenerational transmission of welfare receipt is mixed.3 So far, little research has addressed intergenerational welfare transmission in Germany. While there are a number of studies on income and unemployment transmission, research on welfare receipt is limited. Closest to our analysis is Siedler (2004): using early data from the German Socioeconomic Panel (1984–2002) he investigates intergenerational correlation in social assistance receipt. He focuses on young adults’ benefit receipt at age 22 or above, that is, at an age when almost 40 percent of the sample still live in the parental household. He applies regional characteristics as instruments as well as bounds analyses and concludes that parental benefit receipt is exogenous. Therefore, the correlation patterns are interpreted as causal effects. 3. INSTITUTIONAL BACKGROUND The German constitution guarantees each resident the right to a “dignified life”: if an individual or household cannot muster the financial means for a “dignified life”, the person or household can demand the support of the state. Different 3The literatures on the transmission of disability and unemployment benefits apply similar methods with mixed results: Dahl et al. (2014), Dahl and Gielen (2021), and Grübl et al. (2020) confirm causal intergenerational transmission while Ekhaugen (2009) and Maeder et al. (2015) reject it. Bratberg et al. (2015) and Mueller et al. (2017) find causal transmission patterns for some family relationships but not for others. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1229 Review of Income and Wealth, Series 70, Number 4, December 2024 Figure 1. Unemployment and welfare institutions for the working-age population in Germany before and after the 2005 reform. Source: Own illustration . programs provide assistance for groups such as the unemployed, the elderly, the disabled, and the poor.4In our analysis, we jointly consider those branches of the welfare state that provide means-tested minimum income support to individuals below retirement age (for a similar strategy see Boschman et al., 2019). As the welfare state underwent a major reform in 2005, we distinguish between preand post-reform institutions (see Figure 1). We consider the receipt of social assistance (Sozialhilfe) and unemployment assistance (Arbeitslosenhilfe) before the reform and social assistance, unemployment benefit II (UB II), and social money after the reform to capture means-tested minimum income support. We label the combined institutions “welfare” throughout. Before the reform, individuals could claim means-tested social assistance (Sozialhilfe) if their household income, that is, the combination of earnings or other income, unemployment benefits, or unemployment assistance, was too low to cover the formally defined financial need of the household. Social assistance provided general income support to the employed, the unemployed, and those out of the labor force. In addition, those who had exhausted their insurance-based unemployment benefits and those who were not (yet) entitled to unemployment benefits were eligible for a second, tax-financed and means-tested unemployment assistance (Arbeitslosenhilfe). Unemployment assistance replaced up to 57 percent of previous net labor earnings and in most cases was provided without a time limit, that is, at most until retirement. On December 24, 2003, the reform law (Viertes Gesetz für moderne Dienstleistungen am Arbeitsmarkt called “Hartz IV”) was passed which came into effect on January 1, 2005. Its objective was to reduce transfer dependence and shorten the 4Poverty is established in a means test: first, the financial need of a given household is formally determined. It consists of administratively fixed amounts for all household members plus housing expenditures (rent and heating). If household income and wealth are too low to cover the thus calculated financial need the household can claim government support. While institutional regulations are gender neutral, females are more affected by poverty than males. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1230 Review of Income and Wealth, Series 70, Number 4, December 2024 transfer receipt period. Except for shortened payout periods, the unemployment insurance benefit was not affected by the reform; Riphahn and Schrader (2020) study the effect of the reduced payout period. Figure 1summarizes the institutional changes caused by the reform: the former unemployment assistance and social assistance programs were combined in the new UB II program, a means-tested and tax-financed benefit for those able to work. Since the reform, individuals who exhaust their unemployment insurance benefit (i.e., UB I) or whose UB I claim is insufficient to cover the household’s financial need may be eligible for UB II (possibly in addition to UB I). The UB II benefit covers the legally defined minimum income (household financial need). Generally, all individuals—including those who are employed or out of the labor force—can claim UB II if their household passes the means test and if they are physically able to work at least 15 h per week. Their children or other household members who are not able to work can claim a similar benefit called social money (Sozialgeld). Independent individual claims against the UB II system are possible starting at age 25; then, financial means are compared to their needs for the young person or the person’s own core family. Since the reform, the previous social assistance (Sozialhilfe) program is available only for those who are not able to work, for example, due to sickness, disability, or care responsibilities, and who do not have an employable household member. The main change induced by the reform was the abolition of the unemployment assistance program. Individuals with high prior labor earnings who previously received unemployment assistance faced cuts: their benefit claims declined and in addition, they had to pass more stringent means tests than before. Those who received social assistance before the reform continued to be eligible for UB II as long as they were able to work; for details on the German welfare system see BMAS (2019,2020). Figure 2describes the utilization of the welfare programs over time. The absolute number of social assistance recipients (dashed line) increased since 1980 from below 1million to almost 3million individuals in 2004.5Similarly, the number of unemployment assistance recipients (dotted line) increased substantially over time—since 1991 covering East Germany, as well. The thin dotted grey line presents the sum of social assistance and unemployment assistance beneficiaries; as some individuals may have benefitted from both programs the addition generates an overcount. The unemployment assistance program disappeared in 2005. Immediately after the reform, the number of unemployment benefit II (UB II) recipients (bold black line) surpassed 5million basically continuing where the sum of the two prior benefits left off. The number declined in subsequent years. The number of social money recipients was constant at about 0.8 million and reflects individuals in the household of UB II recipients who cannot work, that is, mostly children. After the reform, the social assistance benefit was used only by individuals unable to work at least 3 h per day and dropped. The figure suggests that the joint consideration of the two means-tested programs of social and unemployment assistance before 5This covers welfare recipients who live independently (Hilfe zum Lebensunterhalt außerhalb von Einrichtungen). The group of disabled individuals was supported by a different social assistance program (Hilfe in besonderen Lebenslagen) and is not reflected in Figure 2. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1231 Review of Income and Wealth, Series 70, Number 4, December 2024 0 10,00,000 20,00,000 30,00,000 40,00,000 50,00,000 60,00,000 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 Soc. Assist. Un. Assist. UB II Social Money SA + UA Figure 2. Utilization of welfare programs over time. Source: Own depiction based on information from different sources. Until 1990 only West Germany, starting 1991 East and West Germany. Social Assistance (recipients as of 31.12. annually) from https://www.destatis.de/DE/Themen/ Gesellschaft-Umwelt/Soziales/Sozialhilfe/Tabellen/ liste-hilfe-lebensunterhalt-empfaenger-zr.html [last accessed July 28, 2021]. Unemployment assistance (annual average number of recipients), BA (2020), Arbeitslosengeld und Arbeitslosenhilfe von 1991–2004 (Zeitreihen Monatsund Jahreszahlen); for earlier years: annual publications of Amtliche Nachrichten der Bundesagentur für Arbeit (ANBA). Unemployment benefit II (UB II) and Social Money recipients as of December each year: BA (2021), Strukturen der Grundsicherung SGB II – Deutschland, West/Ost, Länder und Kreise (Zeitreihe Monatsund Jahreszahlen ab 2005), Table 1 (erwerbsfähige und nichterwerbsfähige Leistungsberechtigte). and the UB II program after the reform generates a plausible reflection of welfare receipt.6 4. EMPIRICAL MODEL AND METHODS 4.1. The Model We are interested in whether the welfare receipt of young adults, that is, the child generation, is associated with and potentially caused by experiencing the welfare receipt of their parents. We follow the previous literature and model child i’s 6There is substantial non-take-up in the German welfare system of more than 40 percent of the eligible population. However, this is moderate by European comparison (Eurofound, 2015,Table1). Also, other social policy programs in Germany feature even higher non-take-up rates (Bruckmeier & Wiemers, 2018,Table3). These authors point to a complicated benefit structure and argue the expected utility of the entitlements as well as information costs and stigmatization explain take-up behavior. For recent evidence see, for example, Bruckmeier et al. (2021), Bruckmeier and Wiemers (2017), and for the pre-reform welfare system Riphahn (2001). © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1232 Review of Income and Wealth, Series 70, Number 4, December 2024 welfare receipt (WiC) in observation period t1as a function of parental welfare receipt (WiP) in an earlier observation period t0: (1) WiC=WiP𝛽0+ε 0iC. The estimate of coefficient 𝛽0reflects the unconditional intergenerational correlation in welfare receipt. As this correlation may be affected by various factors, we consider an extended specification that controls for a set of individual and household level covariates (X) such as age, gender, migration background, and region of residence: (2) WiC=WiP𝛽1+Xiγ+ε 1iC, where 𝛽and 𝛾are coefficients to be estimated. The estimate of 𝛽1reflects the conditional correlation of welfare receipt across generations. While it may not provide the causal transmission effect, it quantifies the overall association between parent and child outcomes. It is interesting to compare this association for different subgroups and for different types of exposure. Estimates of 𝛽1can be interpreted as a causal effect only if parental welfare receipt is exogenous, that is, uncorrelated with the error term ε1iC. However, this is unlikely if parent and child welfare participation are affected by unobserved heterogeneities (e.g., tastes, preferences, biological factors, abilities, or unobserved regional characteristics). Let parental welfare receipt be modeled by (3) WiP=XiPδ+ε iP. Then, the error terms for child and parent welfare receipt may follow (4) εiC=𝛼iC+𝜇iC (5) εiP=𝛼iP+𝜇iP, where 𝜇iCand 𝜇iPare uncorrelated random error components. If there are unobserved family characteristics we expect corr(𝛼iC,𝛼iP)≠0. This correlation causes a bias in the OLS estimate of 𝛽in Equations (1) and (2): the coefficient estimate mixes the causal effect of experiencing parental welfare receipt in period t0and the effects of shared family unobservables. In the first step of our analysis, we estimate the intergenerational correlation of welfare receipt using two model specifications. In a basic specification, we do not consider a detailed set of control variables. In an extended specification, we account for heterogeneity along individual and parental background dimensions. In particular, we control for characteristics of the individual (year of birth, gender, immigration background, and parity, i.e., the rank position in the family birth order), characteristics of parents (year of birth, parental education) and household characteristics at age 17 of the individual (household size, number of children in parental household, federal state of residence). Holding these dimensions constant © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1233 Review of Income and Wealth, Series 70, Number 4, December 2024 In panel B of Table 2, we present the estimates of the extended specification. The controls account for some of the intergenerational correlation in welfare receipt: the coefficients decline in magnitude but remain highly statistically significant. Conditional on individual, parent, and household characteristics young individuals are about 14 percentage points more likely to receive welfare when their parents received welfare during their teen years, a substantial difference; Table A.4 in the Appendix presents the full set of estimation results. The estimate in column 2 shows an increasein the number of ownyears of welfare experiencebyabout 0.16 for each year of parental welfare receipt. Column 3 suggests that the share of observed years on welfare as an adult is associated with a significant increase of 21 percentage points when parents were on welfare for the full observation period. The reduced intergenerational correlations in Panel B compared to Panel A suggest that the control variables are indeed correlated with the propensity to receive welfare benefits: the intergenerational correlation of welfare receipt is smaller within demographic groups than on average. Below we inspect these heterogeneities in greater detail. Next, we investigate whether the association between child welfare receipt and the duration of parental receipt is indeed linear. We separately regress the extensive margin of child welfare receipt (i.e., ever welfare in period t1) on having experienced at least xnumber of years of parental receipt, where x runs from 1 to 9. Figure A.1 in the appendix shows the results for both specifications: the propensity to ever receive welfare increases with the number of years of parental welfare receipt experienced, however, confidence intervals are wide. In Figure A.2 we describe the development of correlation patterns as estimated by the basic regression specification separately for subsequent birth cohorts. We use rolling regressions on three neighboring birth cohorts. The patterns are similar for all three outcomes with a peak in correlations in the early 1970s and a significant positive trend for more recent birth cohorts. 6.2. Heterogeneity by Child and Parent Gender Next, we follow the literature and investigate whether intergenerational welfare correlation differs for young men and women; descriptive statistics yield higher welfare receipt among females than males.9We apply different strategies to describe the gender-specific patterns in our data. First, we re-estimated the extended specification described in Table 2and additionally interacted parental welfare receipt with child gender. Panel A of Table 3shows that the correlation between parent and child welfare receipt is substantially but mostly insignificantly higher for females. Panels B and C of Table 3show separate estimations of the basic and extended specifications by gender and confirm higher intergenerational correlations for females than males across all welfare indicators. This agrees with the literature (e.g., Dahl & Gielen, 2021; Hoynes et al., 2016). One mechanism may be that the single parenthood risk is larger for females and can be transmitted across generations (Musick & Mare, 2004). Also, role model expectations and social norms may contribute to gender differences in economic independence. 9In our sample, 14 and 11 percent of females and males ever receive welfare, respectively. For females, we observe on average 0.62 and for males 0.48years of welfare receipt. The differences in the parent generation are small and insignificant (see Table A.5 for descriptive statistics by gender). © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1240 Review of Income and Wealth, Series 70, Number 4, December 2024 TABLE 3 GENDER-SPECIFIC EFFECTS Dependent variables: Welfare receipt t1 (1) (2) (3) Ever (0/1) Number (years) Share (%) Panel A: Extended specification with gender interaction (N=2403) Parent welfare, t00.100*** 0.083** 0.151*** (0.036) (0.038) (0.050) Female ×Parent welfare, t00.074 0.145** 0.117 (0.050) (0.059) (0.039) Panel B: Male sample (N=1205) Parent welfare, t0, basic specification 0.161*** 0.134*** 0.201*** (0.036) (0.038) (0.051) Parent welfare, t0, extended specification 0.109*** 0.089** 0.154*** (0.037) (0.039) (0.049) Panel C: Female sample (N=1198) Parent welfare, t0, basic specification 0.213*** 0.269*** 0.306*** (0.037) (0.047) (0.058) Parent welfare, t0, extended specification 0.166*** 0.214*** 0.254*** (0.035) (0.047) (0.056) Panel D: Maternal welfare receipt—basic specification (N=2375) Maternal welfare, t00.195*** 0.208*** 0.243*** (0.028) (0.035) (0.041) Panel E: Paternal welfare receipt—basic specification (N=2284) Paternal welfare, t00.197*** 0.233*** 0.296*** (0.032) (0.048) (0.055) Panel F: Maternal welfare receipt—ext. specification with gender interaction (N=2375) Maternal welfare, t00.158*** 0.126*** 0.178*** (0.039) (0.042) (0.054) Female ×Maternal welfare, t00.071 0.154** 0.122 (0.056) (0.066) (0.081) Panel G: Paternal welfare receipt—ext. specification with gender interaction (N=2284) Paternal welfare, t00.163*** 0.165*** 0.264*** (0.045) (0.060) (0.084) Female ×Paternal welfare, t00.061 0.126 0.052 (0.063) (0.093) (0.110) Notes: Each cell entry represents a separate regression where the parental welfare measure matches the dependent variable as listed in the column headers (see Table 2). Panels D–G additionally control for an indicator reflecting whether an individual ever lived with a single parent (i.e., in a non-couple household). Robust standard errors are reported in parentheses. For details on the basic and extended specification see notes of Table 2.∗∗∗p<0.01; ∗∗p<0.05. Source: SOEP (1984–2017), own calculations. In Panels D and E of Table 3, we present separate estimates based on whether maternal or paternal welfare receipt was observed during childhood.10 Since any 10As welfare is provided at the household as opposed to the individual level the welfare outcome in our data was identical for 88 percent of parent couples. The gender-specific effects are identified from separated couples or single parents where the children live with only one of the two parents. In very few cases (28 for mothers and 119 for fathers) we have no information on the person-specific welfare history. There are no major differences by parent gender in extended specification. Results are available upon request. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1241 Review of Income and Wealth, Series 70, Number 4, December 2024 differences might be due to living with only one parent rather than to the parental gender in particular, we additionally control for single parenthood in these specifications. Our results yield only minor differences in parent-specific correlation patterns for the basic specification; in separate estimations (not presented to save space) we observe larger intergenerational correlation coefficients if maternal welfare receipt was experienced in a single-parent household. In separate estimations, we considered child gender interaction terms in the estimations for fathers’ and mothers’ welfare outcomes (see panels F and G of Table 3). These results indicate positive but again mostly insignificant coefficient estimates confirming the stronger correlations for female children but no major differences by parent gender. Overall, the findings confirm patterns found in other studies: using data for Germany, Mueller et al. (2017) obtained stronger intergenerational unemployment correlations for daughters than for sons. Using Dutch data, Boschman et al. (2019) also found the correlation patterns for maternal and paternal social assistance receipt to be similar.11 6.3. Heterogeneity by Age of Exposure Numerous contributions discuss the relevance of a child’s age at exposure to intergenerational transmission effects. Bratberg et al. (2015) and Dahl and Gielen (2021) study the relevance of age at exposure with respect to the transmission of parental disability.12 Carneiro et al. (2021) studied the connection between the timing of parental income shocks and the next generation’s human capital outcomes. Conditional on household permanent income they find that children benefit most from positive income shocks during age 0–5 and 12–17. Edmark and Hanspers (2015) and Hartley et al. (2022) compare the relevance of parental welfare receipt across child exposure ages. The former find the strongest intergenerational correlation if the young generation was exposed at age 17–19 and argue that this may reflect role-model or network-related effects that are strongest in the formative years of the late teens. Hartley et al. (2022) find larger correlations for older ages at exposure, that is, at ages 10–14 through 13–17. The authors similarly suggest that learning effects increase when children experience welfare receipt at older ages. In our analysis, we consider exposure to parental welfare receipt at ages 10–12, 13–15, and 16–18. As our survey does not allow us to go back in time for all individuals, we start out with age-group-specific estimations which vary in sample size. Panel A of Table 4shows the results. Across all welfare indicators, we find stronger correlations for the youngest and oldest age groups and the smallest correlations for the middle age group of 13–15-year-olds. In order to compare the age-specific 11The studies on intergenerational transmission of disability benefits in Norway disagree on this issue: while Dahl and Gielen (2021) find larger transmissions from mothers, Bratberg et al. (2015) observe larger effects for fathers. While Dahl and Gielen (2021) find no heterogeneity by child gender, Bratberg et al. (2015) observe larger effects for daughters than sons. 12Bratberg et al. (2015) compare child age categories from below 15 to up to 40 and do not find clear heterogeneities for exposure at younger ages. Dahl and Gielen (2021) compare effects for children up to age 14, up to age 18 or at age 19 plus. They find larger intergenerational spillover effects if the younger generation is young at the time of parental treatment. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1242 Review of Income and Wealth, Series 70, Number 4, December 2024 TABLE 4 HETEROGENEITY OF WELFARE CORRELATION BY AGE OF EXPOSURE Dependent variables: Welfare receipt t1 (1) (2) (3) Ever (0/1) Number (years) Share (%) Panel A—Separate estimations by age group—Basic specification Age group 10–12, t0(N=1242) 0.225*** 0.559** 0.282*** (0.046) (0.117) (0.060) Age group 13–15, t0(N=1835) 0.189*** 0.400*** 0.196*** (0.037) (0.081) (0.040) Age group 16–18, t0(N=2403) 0.201*** 0.427*** 0.203*** (0.030) (0.065) (0.032) Panel B—Joint Estimations for all Age Groups (N=1216)—Basic Specification Age group 10–12, t00.135** 0.303*0.147** (0.055) (0.156) (0.071) Age group 13–15, t00.051 0.129 0.013 (0.052) (0.150) (0.068) Age group 16–18, t00.123** 0.325*** 0.156*** (0.048) (0.121) (0.056) Notes: Each cell entry represents a separate regression where the parental welfare measure matches the dependent variable as listed in the column headers (see Table 2). Robust standard errors are reported in parentheses. For details on the basic specification see notes of Table 2.∗∗∗p<0.01; ∗∗p<0.05; ∗p<0.10. Source: SOEP (1984–2017), own calculations. correlations for a given yet smaller sample we pooled the three age-group-specific measures in panel B of Table 4and estimated the correlation patterns in one joint model. We continue to find the weakest correlation for the middle age group and larger impacts for the youngest and the oldest group. Table 4shows results for the basic specification only. The results are similar when the extended specification is estimated (available upon request). The finding of larger coefficients for the oldest group agrees with the literature. The strong correlation for 10–12-year-olds is somewhat surprising. Possibly it is related to the German secondary schooling system where at around age 10 important tracking decisions are taken. If these decisions are negatively affected by financial problems in the parental household the effects may reduce average human capital with long-run effects. Boschman et al. (2019) consider the heterogeneity of correlation patterns by recency of parental welfare receipt as a potential indicator of the relevance of information transmission. That we find correlations of similar magnitude for 10–12 and 16–18-year-olds does not support the idea of recent information as an important mediator. 6.4. Pre Versus Post Reform Patterns During our observation period, the German welfare program underwent an important reform in 2005 that is intensely debated to this day. The reform aimed to activate welfare recipients who are able to work (see Section 3). It increased job search monitoring and it reduced benefits for some long-term unemployed. We describe intergenerational correlation patterns before and after the reform; for analyses of preand post-reform state dependence in welfare receipt at the individual level over time see, for example, Riphahn and Wunder (2013,2016). We consider © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1243 Review of Income and Wealth, Series 70, Number 4, December 2024 individuals who reached age 29 before 2005 (birth cohorts 1969–1975) to be subject to the pre-reform welfare regime and those who turned 25 in 2005 and after (birth cohorts 1980–1991) to be affected by the reform. Appendix Table A.6 describes the two groups’ welfare outcomes and intergenerational correlation patterns. While the welfare outcomes for the two subsamples in t1 are similar, surprisingly, we find much higher parental welfare receipt for the post reform group (see Panel A). Panel B additionally shows higher intergenerational correlations for the post-reform of, for example, 0.24 versus 0.15 for the “ever welfare” outcome. This may reflect aggregate trends to higher welfare use over time (see Figure 2). Table 5shows the estimation results for both subsamples with the basic and extended specifications in Panels A and B, respectively. Panels C and D offer estimation results on the pooled samples with an interaction term. In Panel A, estimation TABLE 5 PRE-VERSUS POST-REFORM OUTCOMES Dependent variables: Welfare receipt t1 (1) (2) (3) Ever (0/1) Number (years) Share (%) Panel A—Basic specification Pre-Reform (N=778) Parent welfare, t00.202*** 0.216** 0.163** (0.0624) (0.0895) (0.0746) Post-Reform (N=1210) Parent welfare, t00.196*** 0.218*** 0.308*** (0.0317) (0.0342) (0.0475) Panel B—Extended specification Pre-Reform (N=778) Parent welfare, t00.184*** 0.205** 0.146** (0.0620) (0.0904) (0.0742) Post-Reform (N=1210) Parent welfare,t00.143*** 0.176*** 0.264*** (0.0313) (0.0348) (0.0459) Panel C—Basic specification with interaction terms Full period (N=1988) Parent welfare, t00.191*** 0.207** 0.159** (0.0619) (0.0880) (0.0725) Post ×Paternal welfare, t00.006 0.013 0.147* (0.0695) (0.0945) (0.0865) Panel D—Extended specification with interaction terms Full period (N=1988) Parent welfare, t00.170*** 0.199** 0.142** (0.0603) (0.0877) (0.0724) Post ×Paternal welfare, t0−0.031 −0.029 0.111 (0.0679) (0.0942) (0.0851) Notes: In Panels A and B each cell entry represents a separate regression where the parental welfare measure matches the dependent variable as listed in the column headers (see Table 2). Robust standard errors are reported in parentheses. For details on the basic and extended specification see notes of Table 2. In Panels C and D the preand post-reform observations were pooled and an interaction term of the parental welfare indicator with the post-reform indicator was added to the specification. ∗∗∗p<0.01; ∗∗p<0.05; ∗p<0.10. Source: SOEP (1984–2017), own calculations. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1244 Review of Income and Wealth, Series 70, Number 4, December 2024 results for the first two welfare outcomes yield that the intergenerational correlation did not change substantively after the reform. This pattern is not supported by the outcomes reported in Panel B, where correlations declined in the post-reform period; however, the interaction term estimates in Panels C and D are imprecise. In contrast, correlation patterns for the third outcome increased substantially after the reform (see Panels A and B). This is confirmed by the statistically significant estimate of the interaction term coefficient in Panel C. Overall, these results are inconclusive: we find neither strong evidence of increasing nor of decreasing correlation patterns. 6.5. Relevance of Mediator Variable: Child Education It is possible that child education acts as a mediator of the parent–child connection in welfare receipt. If parental welfare receipt negatively affects child educational attainment (e.g., via role-model effects, stigmatization in school, low parental self-esteem, or residential instability) then low child human capital, that is, cognitive and possibly non-cognitive skills, may limit labor market opportunities and eventually economic independence. We can test whether child educational attainment is a mediator by adding child educational outcomes as a control variable in the estimations shown in Table 2where they had been omitted so far to avoid endogeneity issues. If the intergenerational correlation declines once we condition on child education then mediation effects are likely which may point to useful policy strategies. We consider four indicators of the child’s highest educational degree obtained. Table 6shows the estimated correlation patterns that result after adding the child education controls to the set of covariates in the basic and extended specifications. All coefficient estimates continue to be positive and highly statistically significant. However, in comparison to the results in Table 2they are smaller in magnitude by about 20 percent. Thus, a considerable part of the intergenerational correlation may operate via attenuated educational attainment of children in welfare-receiving households. This agrees well with the literature (see Boschman et al., 2019 or Bubonya & Cobb-Clark, 2021).13 6.6. Causal Estimation Approaches—Family Fixed Effects The correlations investigated so far cannot generally uncover causal effects. To get closer to causal effect estimation we take advantage of siblings from the same family in the family fixed effects model. This allows us to account for time-constant family unobservables. If these are the only biasing factors then the family fixed effects models provide causal effects (see the discussion in Section 4). Our family fixed effects sample offers information on 1161 siblings from 514 different families. Panel A of Table 7presents the baseline correlation estimates for the basic and extended specifications for this particular subsample. The results 13Child education differs significantly for the groups with and without parental welfare receipt. Those with parental welfare receipt are more than twice as likely to be in the lowest (shares of 25 vs. 12 percent) and less than half as likely to be in the highest category (12 vs. 28 percent). © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1245 Review of Income and Wealth, Series 70, Number 4, December 2024 TABLE 6 CONTROLLING FOR CHILD EDUCATION AS A POTENTIAL MEDIATOR Dependent variables: Welfare receipt t1 (1) (2) (3) Ever (0/1) Number (yrs) Share (%) Parent welfare, t0(basic specification) 0.148*** 0.161*** 0.212*** (0.025) (0.031) (0.039) Parent welfare, t0(extended specification) 0.112*** 0.129*** 0.180*** (0.025) (0.031) (0.038) Notes: Estimations used 2403 observations. Each cell entry represents a separate regression where the parental welfare measure matches the dependent variable as listed inthe columnheaders (seeTable 2). Robust standard errors are reported in parentheses. For details on the basic and extended specification see notes of Table 2; all estimations additionally control for three indicators of child educational attainment. ∗∗∗p<0.01. Source: SOEP (1984–2017), own calculations. TABLE 7 FAMILY FIXED EFFECTS ESTIMATION Dependent variables: Welfare receipt t1 (1) (2) (3) Ever (0/1) Number (years) Share (%) Panel A: OLS results for the FE Sample (N=1161) Parent welfare (basic) 0.199*** 0.233*** 0.290*** (0.037) (0.044) (0.058) Parent welfare (extended) 0.132*** 0.162*** 0.211*** (0.036) (0.043) (0.054) Panel B: FE Regressions (N=1161) Parent welfare (basic) −0.007 −0.028 −0.150 (0.076) (0.087) (0.104) Parent welfare (extended) 0.009 −0.031 −0.163 (0.78) (0.089) (0.115) Panel C: FE Regressions w/o welfare experience of oldest child (N=1150) Parent welfare (basic) 0.003 0.002 −0.155 (0.081) (0.086) (0.112) Parent welfare (extended) 0.009 −0.031 −0.163 (0.078) (0.089) (0.115) Notes: Each cell entry represents a separate regression where the parental welfare measure matches the dependent variable as listed in the column headers (see Table 2). Robust standard errors are reported in parentheses. For details on the basic and extended specification see notes of Table 2.∗∗∗p<0.01. Source: SOEP (1984–2017), own calculations. are rather similar to those of the full sample in Table 2.PanelBofTable7shows the coefficient estimates on parental welfare when we apply the family fixed effects estimator to both the basic and extended specifications: the positive significant correlation coefficients do not hold up to fixed effects controls. This result does not support the existence of causal intergenerational treatment effects. As a robustness test, we show the fixed effects results when those families are omitted from the sibling sample where only the older sibling experienced parental welfare receipt. In these cases, the mechanisms that generate intergenerational transmission such as reduced stigma, availability of institutional information © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1246 Review of Income and Wealth, Series 70, Number 4, December 2024 might persist in the family even though a welfare receipt is not observed for the younger sibling. Panel C in Table 7yields that the results obtained so far, that is, no significant positive effects, are robust to this additional test.14 The finding of no causal effects agrees with the family fixed effects estimations forunemploymentbenefit transmissionin Ekhaugen(2009) andMuelleret al.(2017), and for the transmission of maternal (not paternal) benefit transmission in Bratberg et al. (2015). In their fixed effects analyses, Edmark and Hanspers (2015) even obtained negative intergenerational transmission results for welfare receipt in Sweden. The authors argue that either children of welfare recipients are particularly eager to avoid welfare or the coefficients on parental welfare receipt capture other differences between siblings that correlate with the welfare experience. 6.7. Causal Estimation Approaches—Gottschalk Estimation As our second approach to approximate causal welfare transmission effects, we apply the procedure developed by Gottschalk (1996) as characterized in Section 4.3 above. Here, we account for family-specific unobservables that might otherwise bias causal effect estimation by controlling for parental welfare receipt in the period after observing the second generation’s welfare receipt. As we do not observe these outcomes for all families the estimation with an additional parental welfare control can only be performed on a subsample. Appendix Table A.7 shows descriptive statistics for the full sample and the Gottschalk subsample for whom information on late parental welfare receipt is available. Not surprisingly, individuals in the Gottschalk subsample and their parents are on average 2 years older than the main sample. The child generation is insignificantly more likely to use welfare (incidence in t1of 14.3 vs. 12.8 percent) than the main sample whereas the parents are less likely to use welfare in t0. Overall, the subsample characteristics do not appear to differ in important ways (significant age differences are by construction). Next, we investigate whether the correlation patterns in the Gottschalk subsample reflect our results from Table 2.PanelAofTable8shows the basic and extended specification estimates for the Gottschalk subsample. The coefficient estimates do not differ in important ways from prior results. Panels B and C of Table 8show the estimation results of the actual Gottschalk estimation approach for the basic and extended specifications. Each individual parental welfare receipt indicator yields positive and significant coefficient estimates. The row labeled “Gottschalk effect” in each panel presents the difference between the two parental effects as estimated based on Equation (6). In no case do we obtain significantly positive differences which would be indicative of causal 14We pursued two strategies for the robustness test after we determined the set of families where parental “ever-welfare” outcomes varied across siblings; out of 1161 children in 514 families, only 103 children in 38 families had varying parental outcomes across siblings. Our first strategy omitted 49 observations from families where already the first-born child experienced parental welfare receipt (estimating with N=1112 observations) because the one-time experience may affect the family characteristics permanently. In our second strategy, we omitted observations from families where only the first-born child experienced welfare receipt (see Panel C of Table 7). The results hardly differed between the strategies. Nevertheless, we cannot exclude the possibility that families with time-varying benefit receipt differ in unobservable characteristics from other families. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1247 Review of Income and Wealth, Series 70, Number 4, December 2024 TABLE 8 GOTTSCHALK ESTIMATION Dependent variables: Welfare receipt t1 (1) (2) (3) Ever (0/1) Number (years) Share (%) Panel A: OLS results for Gottschalk estimation sample (N=1221) Parent welfare, t0, basic specification 0.178*** 0.249*** 0.260*** (0.039) (0.058) (0.062) Parent welfare, t0, extended specification 0.131*** 0.192*** 0.217*** (0.038) (0.057) (0.060) Panel B: Gottschalk approach estimation results, basic specification (N=1221) Parent welfare, t00.130*** 0.191*** 0.201*** (0.039) (0.060) (0.063) Parent welfare, t20.318*** 0.350*** 0.188*** (0.072) (0.076) (0.046) Gottschalk effect −0.188** −0.160 0.014 (0.089) (0.104) (0.084) Panel C: Gottschalk approach estimation results, extended specification (N=1221) Parent welfare, t00.097*** 0.149*** 0.172*** (0.039) (0.059) (0.061) Parent welfare, t20.254*** 0.271*** 0.156*** (0.073) (0.080) (0.046) Gottschalk effect −0.157** −0.122 0.015 (0.089) (0.108) (0.083) Notes: In Panel A, each cell entry represents a separate regression where the parental welfare measure matches the dependent variable as listed in the column headers (see Table 2). In Panels B and C the parent indicators of periods t0and t2are controlled jointly in the same regression model. The rows labelled “Gottschalk effect” present the difference between the two period-specific coefficient estimates. Robust standard errors are reported in parentheses. For details on the basic and extended specification see notes of Table 2.∗∗∗p<0.01; ∗∗p<0.05. Source: SOEP (1984–2017), own calculations. intergenerational transmission effects. Therefore, the finding of a lack of causal transmission from the fixed effects estimations is confirmed with the Gottschalk approach. In fact, the overall effects on the incidence of welfare receipt even turn out significantly negative. While the negative total effect is surprising, it reflects the findings of other authors who studied possibly heterogeneous policy programs in different countries: Ekhaugen (2009) and Mueller et al. (2017) similarly obtained negative estimates for the transmission of unemployment benefits. Also, Boschman et al. (2019) find negative effects for social assistance and disability programs. These authors argue that it is not the experience of the welfare program itself that causes the next generation’s participation. Instead, family-specific characteristics such as norms and attitudes that arenot attachedto actuallyreceivingthe benefit may drivetheintergenerational correlations. The same patterns appear to hold for our data. 7. CONCLUSIONS The international literature discusses whether experiencing parental welfare receipt in childhood or adolescence is correlated with and causally determines own © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1248 Review of Income and Wealth, Series 70, Number 4, December 2024 welfare receipt later in life. This is an important policy question because intergenerational transmission of welfare receipt indicates a failure of welfare programs: government support does not succeed in lifting families out of poverty and may even impose negative externalities on the next generation. We take advantage of a long-running household panel (SOEP) survey to study the intergenerational transmission of welfare receipt for the case of Germany. Comparative research suggests that Germany offers a relatively generous welfare system; however, its intergenerational characteristics have not been investigated before. The richness of our data allows us to add informative analyses of intergenerational correlation patterns to the literature. We consider three welfare indicators and find strong intergenerational correlation patterns. The correlations are larger for recent than for older birth cohorts and for females than for males. We do not find important differences in the transmission of welfare from fathers vs. mothers. Exposure to parental welfare receipt at the ages of 10–12 and 16–18 yields stronger correlation patterns than exposure in the 13–15 age window. Child educational attainment appears to be a mediator between parent and child welfare receipt which may offer an opportunity for policy interventions. We use family fixed effects and the Gottschalk (1996) method to go beyond correlation analyses and to identify causal effects of parental welfare receipt. Both strategies identify causal effects under specific, yet different assumptions and therefore complement each other. Interestingly, both strategies fail to find evidence of a causal impact of parental welfare receipt on child welfare outcomes. Thus, we do not find evidence that it is the experience of parental welfare receipt itself and a “welfare culture” (Dahl et al., 2014) that drives subsequent child welfare receipt. Instead, the correlation of individual characteristics and circumstances in the child and parent household seems to determine transmission patterns. This suggests that it is not the character of welfare institutions themselves that leaves the offspring of welfare recipients at an elevated risk of welfare receipt. This is a highly policy-relevant finding. It clarifies that any initiative to reduce intergenerational correlation in welfare receipt must not focus on the institutions of the welfare system but more plausibly on characteristics at the individual and household level and, for example, improve human capital, health, and labor market engagement. Our conclusions are subject to strong identifying assumptions and should be reinvestigated when larger samples are available. It seems worthwhile to direct future research to study the determinants and relevance of youth educational attainment, which might be malleable by public policy. Also, it is important to re-analyze any changes in intergenerational correlation after the reform of the welfare system. Finally, we agree with Hartley et al. (2022), who point out that in a situation of low benefit take-up, intergenerational spillovers and correlations can be a good thing if they reduce non-take-up. REFERENCES Adermon, A., Lindahl, M., & Palme, M. (2021). Dynastic human capital, inequality, and intergenerational mobility. American Economic Review,111, 1523–48. Angelini, V., Bertoni, M., & Corazzini, L. (2018). Does paternal unemployment affect young adult offspring’s personality? Journal of Human Capital,12, 542–67. © 2024 The Author(s). Review of Income and Wealth published by John Wiley & Sons Ltd on behalf of International Association for Research in Income and Wealth. 1249