Comrades in the family? Soviet communism and demand for family insurance

Funding information Polish National Science Centre, Grant/Award Number: 2019/33/B/HS4/00387 Abstract We study how exposure to (Soviet) communism (EC), a political-economic regime based on collectivist state planning, affected the preferences for family support, which we refer to as informal family insurance. Against the backdrop that ‘communism gave rise to the abolition of the family’, we document that it actually strengthened the preference (the demand) for informal family insurance without depressing individuals' preferences for social insurance. We exploit cross-country and cohort variation in EC on more than 314,000 individuals living in 33 Central and Eastern European countries, among which 14 had been subject to communist regimes. We estimate that EC gave rise to 9.6 percentage point (pp) increase in the preference for family care for older parent and 4.3 pp increase in the support (both financial and nonfinancial) for children. These effects are explained by the strengthening of social and family networks that resulted from the erosion of generalized, interpersonal and institutional trust, rather than by ‘indoctrination effects’ during Soviet communism times.

. Although throughout history the family has served as a source of informal insurance, one of the central aims of Soviet communism 1 was to abolish the traditional family, as it was perceived to be perpetuating capitalism (Marx & Engels, 2013). Hence, it is an empirical question whether entry to and exit from Soviet communism influenced the demand (or preference) for family insurance. So far, previous studies document that exposure to (Soviet) communism (EC) leads to a preference for social insurance drawing on evidence from Eastern Germany (Alesina & Fuchs-Schündeln, 2007). However, we know very little about the effects of EC on the organization of the family and specifically on the preferences for the family as a source of informal financial and nonfinancial support.
The main purpose of this paper is to fill this gap.
Although family structures are argued to reflect the ideologies that have shaped the history of the 20th century (Todd, 1985) and to reinforce inequality (Marx & Engels, 2013), we contend that preferences for family support (or informal family insurance) are endogenous to political regimes. By abolishing formal wealth accumulation, communism might have created parallel informal incentives to develop family networks further, being a source of informal insurance, which we define as the 'informality hypothesis'.
Other studies suggest that conversely, informal family support can be crowded out after the extension of market insurance (Becker, 1981). However, we contend that such effects depend on wider institutional incentives specific of a political-economic regime (Bowles & Gintis, 1975). In this paper, we show that in the absence of free markets, and in a setting where privilege does not reflect in wealth accumulation, 'internal family networks become a salient way to have access to privilege, including education and elite positions' (Filtzer, 2014).
Nonetheless, the effect of EC on the family is far from trivial given the strong presence of the state in Soviet regimes, which included extensive public benefits potentially crowding family support out. Furthermore, Soviet regimes would be followed by propaganda instilling Marxism-Leninism, which openly aspired to the abolition of the traditional family. This latter effect is referred in the literature as the indoctrination hypothesis. This paper will test the latter hypothesis, which was the core of the seminal analysis by Alesina and Fuchs-Schündeln (2007). This paper is the first to document that along with limited precautionary saving opportunities, exposure to Soviet communism brought two additional effects. First, we document that it shifted the demand for all types of formal and informal insurance (Di Tella & MacCulloch, 2002), thereby strengthening the networks of family and informal support, while at the same time demanding more publicly funded social programmes (Alesina et al., 2001;Alesina & Fuchs-Schündeln, 2007). Second, we show that the effect of communism exposure was heterogeneous across countries, which has not been analysed before.
The current paper extends the analysis of the EC effects to a larger set of countries under the Soviet influence besides Germany, which, to the best of our knowledge, has been the sole focus of previous literature. Furthermore, we conduct a battery of robustness checks referring to various definitions of regime exposure, and different samples of countries, including the effect of historical legacies (Simpser et al., 2018;Wittenberg, 2015). We document robust evidence of a significant increase in the preference for family support (informal family insurance) resulting from EC. The mechanisms driving the effect include the erosion of both generalized trust as well as of the trust in the regime-specific institutions, driving the reliance on family networks which in turn helps to understand the increase of the demand for either insurance type.
The next section describes how the paper adds to the wider and specific literature. Section 3 describes the data. Next, Section 4 contains the empirical strategy and Section 5 reports the main results and heterogeneity analysis, followed by mechanisms, threads to the identification, and robustness checks. Section 7 concludes. 1 A political-economic regime based on the collectivist planning of human needs was introduced to the Russian empire with the 1917 Revolution and imposed on the number of neighbouring countries after World Word II.

| RELATED LITERATURE
Soviet communism was characterized by the absence of free markets, a strong state level control, and the provision of extensive public benefits and services. These features coexisted with prevalent corruption in public institutions (Karklins, 2002) and a culture of fear, persecution, and repression (Bohacek & Myck, 2011;Rozenas & Zhukov, 2019).
However, as we argue in this paper, by abolishing wealth accumulation, communism created parallel incentives to develop informal family networks. Although under certain circumstances, market economies can potentially erode family ties (Alesina & Giuliano, 2010), when wealth accumulation is not an option, or an economy is 'weakly monetized', privileged groups turn to the family to cultivate their groups connections that provide access to elite positions and other forms of priviledge (Filtzer, 2014). That said, such strategies differ across gender and urban-rural areas.
Some studies find that strong reliance on the family can erode generalized trust and inhibit support for social insurance (Alesina & Giuliano, 2010), or in its extreme form, give rise to phenomenon known as 'amoral familism' (Banfield, 1967). Consistently, Fukuyama (1996) argued that low-trust societies are characterized by large families and strong internal ties. Given the well-documented detrimental effects of Soviet communism on trust (Letki, 2018;Traps, 2009), one might hence hypothesize that Soviet communism might encompass strong family ties. However, in both communist and noncommunist countries, similar changes were brought by modernization when conservative gender norms (and in turn family values) were replaced by values oriented towards gender equality (Naumann, 2005). According to Unified Growth Theory (Galor, 2011), such a phenomenon was possible, because in modern growth regimes, the importance of family weakens. Hence, one might expect that the advancement of gender equality was enhanced by state-controlled employment in Soviet communism. In contrast, free labour markets are more prone to gender discrimination. Hence, the effects of Soviet communism on the preferences for family insurance are far from obvious.
Previous studies on the impact of EC have placed the focus on social rather than family preferences. Corneo and Grüner (2002) documented significant differences in social preferences between Eastern and Western European countries, resulting from 'indoctrination effects', namely, that exposure to Soviet communism increased people's egalitarian preferences (Alesina et al., 2001;Alesina & Fuchs-Schündeln, 2007) in line with the socialist thought. Consistently, EC brought a reduced individual self-reliance (Bauernschuster & Rainer, 2012). More recent evidence suggests that after 20 years of communism, Eastern Germans exhibit weaker prosocial behaviour (Brosig-Koch et al., 2011), an effect driven by changes in the cohorts socialized prior to the demise of Soviet communism (Huber & Mikula, 2019).
Previous evidence is, on first sight, inconsistent with other studies (Brosig-Koch et al., 2011;Ockenfels & Weimann, 1999;Shiller et al., 1990Shiller et al., , 1992. Indeed, some literature has attempted to reconcile the lack of empirical consensus by examining the differences in social values between East and West Germany. Van Hoorn and Maseland (2010) identify differences between Eastern and Western Germans using happiness data and conclude that, contrary to expectations, Easterners appeared to entertain values more conducive to economic growth, which questions the myth of pro-entrepreneurial values in the West. Campa and Serafinelli (2019) compare attitudes towards work in East versus West Germany. They show that women were more likely to work in Eastern Europe as state-socialist governments promoted women's economic independence. Similarly, they show that US migrants educated under the state-socialist regimes become less traditionalistic compared with Western European countries. This paper attempts to contribute to this literature by examining the effect of EC on the preferences for family support, which helps to reconcile some of the apparent inconsistencies of the previous literature.

| Data sources
Our analysis draws on three primary data sources: the Generations and Gender Survey (GGS), the World Values Survey (WVS) and the European Social Survey (ESS). We supplement the main data sources with the 2006 wave of the Life in Transition Survey (LITS) sampling all countries exposed to Soviet communism. We use a quasi-experimental design to examine the exposure to EC by using postcommunist countries varying with respect to EC at different stages of political regime maturity, along with other European countries as controls, and different cohorts of individuals that exhibit a differential exposure over time. Migrants are excluded, because their spatial mobility is likely to alter the EC effects. 2 The surveys provide data on social attitudes and behaviours of adults observed from 1981 to 2017. We employ post-1989 observations on Europeans living in 33 countries, including citizens born in 14 postcommunist countries, yielding a total of about 221,000; 67,000; and 314,000 observations in GGS, WVS, and ESS, respectively. 3

| Preference for family insurance
We operationalize the preference for family insurance with questions on the preferred role of one's own family in the provision of informal support. In this paper, we focus on measures of preferences for family support concerning intergenerational (flowing upward or downward) transfers of care and money exchanged between family members.
Respondents in GGS answered 'to what extent do you agree or disagree with the following statements': 'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so'; 'children ought to provide financial help for their parents when their parents are having financial difficulties'; 'parents ought to provide financial help for their adult children when the children are having financial difficulties'; and 'children should take responsibility for caring for their parents when parents are in need' using the following scale: 1-strongly agree, 2-agree, 3-neither agree nor disagree, 4-disagree, 5-strongly disagree. Figure A1 in Appendix A shows distributions of responses to these questions in Eastern and Western European countries. Descriptive statistics indicate that familiaristic attitudes are more common in the East than in the West, with the case of care for the elderly being the most pronounced example.

| Preference for social insurance
Another group of measures capturing beliefs related to social insurance refers to the role of the state to ensure social equality. In order to operationalize the preference for social insurance, we refer to the measures for the following preferences: income equality, income redistribution, and equality of opportunities. Using 6-point scale ESS respondents answered to what extent they agree with a statement that 'government should reduce differences in income levels' and on 7-point scale how much they resemble individuals believing that it is 'important that people are treated equally and have equal opportunities'. WVS respondents were asked 'Which of these two statements comes closest to your own opinion? A. Incomes should be made more equal. B. We need larger income differences as incentives for individual effort'. Figure A2 illustrates distributions of these measures. The descriptive statistics are suggestive of stronger support for income redistribution as observed in ESS (though not in the WVS) in the East than in the West of Europe. 2 We loosen this assumption in the robustness analysis. After the promulgation of the Soviet Constitution in 1936 migration was very rare if not existent with the exception of family reunification and some forced deportations (Dowty, 1989;Marshall, 2000). However, forced migrations between Soviet republics were frequent in the 1940s (Stola, 1992). Polian (2003) estimates that six million Soviets were resettled before Stalin's death. It is estimated that after the Second World War, twelve million ethnic Germans were deported out of Germany to other Eastern European countries, though until the early 1950s, the lines between the East and the West in some of Eastern European borders were easily crossed. Furthermore, Boenisch and Schneider (2013) document related evidence suggesting that exposure to communism affects the probability of spatial mobility. 3 See Table A1 in Appendix A for descriptive statistics of the research samples. Tables with detailed information on the research samples' structure by country and wave as well as the composition of the research samples by country of residence and birth cohorts are available upon request.

| Exposure to communism
Our study concerns post-war Soviet communism. Living in an Eastern or Western European country provides a crude measure of external margin of EC, but it fails to inform precisely on the extent of the exposure. Thus, we measure the number of years an individual lived under a communist regime to capture EC. Because more accurate measures of actual instalment of communist regimes are unavailable, we use the year when the socialist constitution of the state was announced as an indication of the maturity of communist institutions, which ranges from 1936 to 1952 (see Table 1 for details). 4 The end of communism is also country-specific ranging from 1989 to 1995 and operationalized as the year of the first democratic parliamentary elections, with the exception of Romania and Russia where dates of the death sentence for the Romanian Communist Party general secretary and legislative election were used, respectively. In some countries (e.g., Romania) the dates of nationalization of private property and the introduction of communist constitution coincide.
Most of previous studies measure the EC as the total number of years of exposure, ignoring possible differences between exposures across various stages of life. However, one can identify different periods of Soviet communism (Stalinism, Post-Stalinism, Reformism) and its propaganda that would result in heterogeneity of EC in different stages of life between cohorts.

| Other measures
To examine parallel effects of EC shedding more light on the results obtained in the main analysis, we supplement our measures of the preferences with generalized trust, as in Butler et al. (2016), as well as measures of trust in selected public institutions available in our data sources. 5 We identify deep differences with respect to institutional trust in Europe, in line with the empirical studies discussed before (cf . Table A2). Furthermore, we refer to a number of relevant measures, including political and civic participation, certain dimensions of religiosity as well as traditionalism in gender roles. Table A2 provides more details on these T A B L E 1 The dates of the beginning and the end of the exposure to Soviet communism (EC) in analysed postcommunist counties.
measures, pointing to significantly more pronounced civic participation associated with stronger secularization and gender equality in the West than in the East of Europe. These patterns are observed in all three data sources.

| EMPIRICAL STRATEGY
Shocks in the institutional environment, such as political-economic regime changes, are extremely uncommon (Schelling, 2006). Though, there are a few exceptions (Nee & Swedberg, 2005). The institution of Soviet communism after Second World War was imposed on a number of countries, some of which gained political independence only in 1918. The collapse of communism, even if anticipated, occurred gradually over time beginning in 1989 in Poland and Romania, and extending through Soviet Union dissolution and the first free parliamentary election in Russia in 1991 and 1995, respectively. Thus, the exposure to Soviet communism can be treated as a natural experiment allowing for an examination of the effects of political-economic regime. Our main estimates exploit cross-section and longitudinal data as well as cohort specific variation in the exposure to communism. More specifically, we estimate Equation (1): We examine the effect of the extensive margin of the exposure to communism (EC it -ever exposed to communism) to explain preferences for social and family insurance Y it , where i refers to individuals, t to survey year, g to the country, and c to the birth cohort group.
We conduct pooled OLS regressions of social preferences and behaviours that were influenced by the exposure to communism controlling for demographic, socio-economic, and other individual characteristics. That is, with the set of such control variables X it , we adjust the estimates for age (in a quadratic form) and gender. However, in some specifications, in order to proxy variation in socio-economic status that occurred only after communism collapse, we control for current income and education, even though such variables might potentially be 'bad controls'.
Our parameter of interest testing the informality and indoctrination hypotheses is γ 1 , as its positive value in the estimations explaining preferences for informal family insurance supports the informality hypothesis. The indoctrination hypothesis implies negative relation between EC and the preference for family insurance. As for the preferences for social insurance, positive effects of EC are in line with the indoctrination to Soviet values.
In the next step, we examine in detail the effects of the total length of exposure to communism according to Equation (2): where EC k it stands for k years of total exposure to communism experienced by an individual i, k ¼ 0, …, K g , and K g stands for the maximum length of the exposure to communism in country g. In order to avoid mutlicolinearity between EC and age, rather than as a continuous variable, we treat EC as a categorical one, estimating the effect for each integral of total years of EC separately.
The identification of EC relies on the assumption that selection into communism was random, leading to potential threats that we address in the empirical analysis below. First, we examine alternative operationalizations of the EC (see Tables A28-A31 in Appendix D.3) and cohort measures. Second, we pay particular attention to the choice of the control group. Third, we conduct robustness checks by controlling for other relevant factors, including those related to the effects of transformation from communism (such as occupation type, age at communism collapse, and exposure to recession). 6 In addition to linear estimates using multiple-wave data with time trends (linear, cf. Table A26, quadratic, and country-specific), we confirm that estimates obtained on a single wave in each sample remain stable over time after communism collapsed. Moreover, we estimate random and fixed effects specifications for the panel subsample of GGS (cf . Tables A32 and A33) along with models for binary variables, namely, logit and   probit for dichotomized measures of the preferences related to family and social insurance (cf. Tables A34 and A35).
Finally, it is worth mentioning that the decision to exploit the EC in a number of Central-European countries (rather than relying on the East-West Germany division) is motivated mainly by the fact that EC in Eastern Germany might have been different than in other postcommunist countries, due to its stronger economic, religious, and cultural ties to Western Europe. East and West of Germany were far from being randomly selected into specific treatment and control groups (Becker et al., 2020). 7 Furthermore, migration flows from East to West Germany were more pronounced throughout the duration of communism, a mass arrival to East Germany of individuals expelled from the German lands that become part of Poland took place after 1945 (Gatz, 1989) and forced migrations between Soviet republics were frequent in the 1940s (Stola, 1992).

| Main results
We find that individuals exposed to communism more often report that members of own family should support each other, especially when personal care to older parents and financial support to own adult children is needed (by 9.6 percentual points (pp) and 4.3 pp in the total sample, respectively; cf. Table 2), than individuals unexposed to communism. Our results are consistent with the informality hypothesis. The effects of EC are positive for intergenerational transfers of care both to older and younger generation. In the case of the preference for family financial transfers towards older individuals, we find insignificant yet a negative EC effect, which might be explained by the extensive and generous retirement schemes in formerly communist countries.
These results shown in Table 2 help reconciling previous estimates from Alesina and Fuchs-Schündeln (2007), who find that exposure to communism did increase preferences for redistribution. We document a significant increase in the preferences for income equality and income redistribution due to EC in all estimation specifications.
In the left panel we use all countries (i.e., the control group consists of individuals living in the countries never exposed to communism and individuals in formerly communist countries from the cohorts that were never exposed to communism), whereas in the right panel the sub-sample of countries ever exposed to communism. The main specification yields results robust to alternative control groups.
The effects discussed above are likely to be even larger due to intergenerational transmission of preferences (Bisin & Verdier, 2000). That is, the effect sizes using total sample are predominantly larger than those retrived using the sub-sample of formerly communist countries alone. Indeed, in the total sample we assume that individuals born after the communism in formerly communist countries do not differ from their European counterparts with respect to the preference for family and social insurance. However, it is likely that the preferences shaped under communism in older generations are adapted too by younger individuals never exposed directly to communist institutions. 6 In particular, we control for household size, individual religiosity, experience of war, especially during one's impressionable years, country democratic index, and others, in addition to the three main model specifications. See Appendix D.3 Tables A19-A25 and A27 for results of these robustenss checks.
Moreover, we exclude groups of countries to check if the results remain intact (cf . Tables A14-A17). 7 A selective fifth of the population fled from East to West Germany before the building of the Wall in 1961, which influences differences in working-class, employment in manufacturing, and self-employment prevalence. Reichstag election in 1924 already showed twice as large a communist vote share.
Moreover, share of Protestants was higher in the East and exhibited lower church attendance. Finally, in 1925, female labour-force participation was higher by 6 percentage points in Eastern than in Western Germany.
T A B L E 2 Effects of the exposure to communism (EC) on the preference for family and social insurance (extensive margin).  Table 3. We examine generalized trust measured on a 10-point scale in ESS, as well as its dichotomous version from GGS and WVS and we identify major differences between individuals exposed to communism and those unexposed. The EC coefficient equal to À0.07 (and À0.10) in GGS (and in WVS) suggests that the probability of agreeing with the statement that most 'people can be trusted' would be reduced by 11 (19) per cent due to EC for an average individual living in the West of Europe.
Similarly, we document that individuals exposed to communism declare substantially lower levels of trust in family than those unexposed. That might be partially explained by differences in the perceptions of the family, because in the Eastern European countries the term family is more likely to refer to extended family, while in Western Europe it may only refer to nuclear family. Nevertheless, taking into account this effect as well as the reduced interpersonal trust along with the negative EC effects on confidence in several public institutions, including the legal system, police, and political parties (cf. Figure 1), we are inclined to interpret the overall results on trust as an indication of a strong and long-lasting sense of uncertainty and insecurity being a result of living under communism.

| Civic capital and gender norms
Consistently with evidence documenting a negative correlation between government regulation and trust (Aghion et al., 2010), we find that communism reduced not only trust (cf. Figure A5) but also civic participation and the importance of democratic values (cf . Table A11). If public institutions are perceived as corrupted and people withdraw from expressing their voice in public due to little reliance on democratic institutions, then the preference for placing the responsibility for individuals in need of support on family networks rather than the state becomes a rational strategy.
Furthermore, we find that tradition (cf. Table A12) rather than family is more important to individuals exposed to EC, and older cohorts exposed to communism used to have more children as compared with all those unexposed.
One explanation of this result is that societies exposed to communism were predominantly peasant before entry to communism, and this was even more so after the wartime. Therefore, we find no suggestive evidence to credit the more pronounced preference for family insurance in postcommunist countries to the demise of family importance in Old Europe. Instead, this might result from a stronger adherence to tradition in postcommunist countries.
Against the backdrop that motherhood is less often reported as relevant for female fulfilment and the diminished role of religion during communism, we find strong evidence of a substantial increase in the support for traditional gender roles with EC (cf. Figure A6). This finding reflects the heterogeneity in child-care policy under Soviet communism (Szelewa & Polakowski, 2008). Perhaps more importantly, attitudes towards gender equality in Central and Eastern Europe seem to be taken for granted, which is not surprising taking into account rights to vote and paid maternity leave dating back to 1910s (Wikander et al., 1995), co-education of girls and boys in schools and forced female employment under communism. In such circumstances, pro-choice behaviour might entail freedom from nontraditionalistic commodification of economic activity.

| Effects of the length of EC exposure
We estimate the effects of the length of the EC exposure to examine possible nonlinearity in the effects. Figure 2 shows that EC has ambiguous impact on the preference for family insurance depending on the extent of the communism exposure.
Note: FE-fixed effects. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Demographic controls: age (quadratic), gender. Education controls: highest education level attained. Generalized trust-'generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?' with the answer 'most people can be trusted' opposed to 'need to be very careful'. People being helpful-'people mostly try to be helpful' or 'people mostly look out for themselves'. Trust in family-how much respondents trust their own family measured on five point scale ('trust them completely' … 'do not trust them at all'). Trust (ESS)-how much respondents trust in legal system, political parties, politicians, and police measured on 11-point scale ('no trust at all' … 'complete trust'). Trust (LITS)-how much respondents trust in president, parliament, government, political parties, and armed forces measured on five point scale ('complete distrust' … 'complete trust'). Confidence (WVS)-how much confidence respondents have in the press, political parties, police, army, labour unions, and justice system measured on five point scale. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 2-6 (release 2015_04_18), ESS waves 1-8, and LITS wave 1.
F I G U R E 1 Effects of the exposure to communism (EC) on the generalized trust and confidence in public institutions. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 2-6 (release 2015_04_18), ESS waves 1-8. Note: Point estimates with 95% confidence intervals, controlling for ability to make ends meet or scale of incomes, age (quadratic), gender, education, country, as well as time and cohort fixed effects. Dashed line shows the average effect of EC. Trust-'generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?' Confidence-how much confidence respondents have in the justice system press, police measured on five point scale. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05. *p < .1. [Colour figure can be viewed at wileyonlinelibrary.com] The positive effects of EC on the preference for social insurance are stronger for longer periods of exposure, particularly with respect to equality of opportunities (cf. Figure 3).
The difference in the EC effects on the role of informal family insurance with respect to children and elderly parents is striking and presumably related to the security provided by formal social insurance. The public pension support was relatively generous under communism, which might explain why shorter periods of EC are in line with the indoctrination hypothesis. However, we find that the EC for longer periods increased the preference for family support towards older family members, in line with the informality hypothesis. Exposure to communism for short periods, and more specifically less than 33 years, significantly strengthened the preference for family insurance with respect to the support to younger generations, in line with the informality hypothesis. Consistent with our main results, EC effects on family support towards older adults differ from support to the younger counterparts, and so do the nonlinearities in the EC effects when we focus on its length.
The effects of EC length on other outcomes (cf. Figure A6) suggest thatthe EC effects are nonlinear. A deep change in EC effects depends on its length in the case for support for inequality in gender roles within the labour market and in care-giving. The same is true, for the preference for income taxation, and for the actual number of children. Thus, it seems that the changes in formal and informal institutions occurred in parallel, creating a dynamic environment affecting the preferences for social and family insurance.

| Heterogeneity
Next, we conduct an analysis of the heterogeneous impacts of EC at the intensive margins depending on demographic (gender, birth cohort), regional (rural versus urban and various country groups), and cultural (historical and religious heritage) characteristics. 8 Consistent with previous literature, we find a different effect of EC on social preferences for men and women (cf. Figure A3). That is, our results for the strengthening of the preference for social insurance are driven by men primarily. In contrast, our results suggest strong evidence showing that EC strengthened the preference for family insurance among women, while it reduced the preference for family insurance among men. Taking into account the fact that traditionally women bore responsibility for taking care over the dependent family members (both among older and younger generations), our evidence supports the informality hypothesis. Women's preferences are indeed suggestive of a stronger demand for family insurance, while men's preferences appear to be more in line with the Marxism-Leninist ideology, consistent with the indoctrination hypothesis.
F I G U R E 3 Effects of the exposure to communism (EC) on preference for social insurance by the length of the exposure. Source: Authors' own estimations based on WVS waves 1-6 (release 2015_04_18) and ESS waves 1-8. Note: Point estimates with 95% confidence intervals, controlling for scale of incomes, age (quadratic), gender, education, country, as well as time and cohort fixed effects. Dashed line shows the average effect of EC. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities', equality-more important that 'nobody is underprivileged and that social class differences are not so strong'. Robust standard errors clustered by year of birth and country. ***p < .01, **p < .05. *p < .1. [Colour figure can be viewed at wileyonlinelibrary.com] We observe significant heterogeneity with respect to cohort-specific deviations from general patterns (cf. Tables A5, A6 and A13). The results show that EC exerted the strongest effect on the preference for social insurance in cohorts born before 1950. In contrast, cohorts of individuals born after 1960 reveal only weak evidence of EC strengthening a preference for income equality, and even negative effects on the preference for income redistribution. As for the preference for family insurance, we find that the effect is driven by changes among cohorts born between 1950 and 1969, entering young adulthood in the last (Reformist) stage of communism. These individuals were disillusioned with the performance of communist institutions and thus were more critical with the state propaganda aimed at Soviet indoctrination. Therefore, we observe the more pronounced effects in line with the informality hypothesis for these cohorts.
We find that our estimates on EC effects on the preference for social insurance are driven by urban areas, while the increase in the preference for family insurance is mainly explained by the variation rural areas (cf. Table A4). The latter reflects that the quality and presence of public services and institutions were less resourced in countryside than in more urbanized areas in Central and Eastern Europe, which in turn might explain the regional differences in the effect of EC on family support. Indeed, access to courts and willingness to voice complaints and rights protection was hampered by the individual skills needed for successful completion of such actions (Jegorow, 2003). These differences alongside the higher accessibility of key social services (e.g., nurseries and schools) in urban areas explain the regional diversity of positive EC effects on preferences for examined types of insurance.
Next, the analysis of EC effects in various country groups (cf. Tables A9 and A10) reveals substantial heterogeneity between countries exposed to Soviet communism, 9 which is not surprising taking into account various levels of support for income redistribution across all European countries (Olivera, 2015). Indeed, the general patterns observed for the whole sample of countries are similar to the results for Eastern compared with Western Germany, but the effect sizes differ depending on the choice of country sets. This indicates that infering the effects of Soviet communism only from one country might provide results with limited external validity.
Our analysis suggests that the institutional environment pre-existing on the Habsburg and Prussian lands before regime change to Soviet communism was important when evaluating the preference for social insurance, while EC made hardly any impact in areas already under Russian Empire in the 18th and 19th century. Our analysis (see Table A7) reveals that strengthening of the preference for social insurance due to EC is driven almost entirely by individuals living in the lands of former Prussia, that is, current Germany, Western Poland (Western Prussia), Western Lithuania, and lands formerly belonging to Eastern Prussia (North-Eastern Poland). Moreover, the positive effects of EC on the preference for social insurance are driven mainly by the predominantly protestant countries. Hence, the indoctrination effects of communism are predominantly present in protestant societies living on the lands of former Prussian empire.
In contrast to previous results, when we examine the effect of EC on the preference for family insurance, we find that the informality hypothesis is less evident in the former countries of the Habsburg empire. Indeed, EC effects are stronger in the former Russian empire, and often operate in the opposite direction than those in the former Prussian empire. That is, we find an increase in the preference for family insurance due to EC in predominantly orthodox or Greek Catholic countries, while we observe a decrease in Roman Catholic countries.
In sum, the significant differences between countries exposed to communism and the vast heterogeneity in historic and religious heritage allows us to place the results obtained previously for Eastern and Western Germany in a wider context. 9 Additional results for alternative country groupings are available upon request.
6 | THREATS TO IDENTIFICATION AND ROBUSTNESS CHECKS

| Threats to identification
The dates of the beginning and the end of the communism period are critical for the identification of EC effects. 10 Therefore, we examined alternative operationalizations 11 of communism timelines and we document that the results remain intact, also after setting the end of the communism to a fixed year (either 1989 or 1991). We observe almost the same results for both preference for family and for social insurance regardless of the EC measure, which is reassuring with respect to the credibility of our main results.
Another pivotal source of heterogeneity allowing the identification of EC effects lies in the differential exposure to communism by cohort. Therefore, we examine cohort groups alternative to the year of birth. Specifically, we employ as a robustness check a 10-and 5-year cohort group-reflecting the 5-year span of economic central planning-and find negligible changes in EC effects between alternative cohort groupings. Moreover, we test alternative approaches to clustering over only the year of birth or only the country of birth, and we find negligible differences from the main approach relying on the variation over birth year and country.
In the main analysis, we removed migrants from the research sample in order to mitigate selective migration.
However, the inclusion of migrants in the sample does not alter overall EC effects on the preference for family insurance and the preference for income redistribution.
Next, we examine whether the selection of countries unexposed to communism affects our estimates excluding the West of Germany. The effects including all other postcommunist countries juxtaposed with Western Germany are similar to the main results with entire sample of Western Europe, despite minor differences in their magnitude.
However, removing Western Germany from the sample significantly alters the results (cf. Table 4). Individuals exposed to communism do not differ fromindividuals unexposed to communism living in non-German postcommunist European countries with respect to the preference for income equality and income redistribution, but they are significantly more inclined to report a preference for family insurance with respect to care. This result strongly supports our findings on positive EC effects on the preference for family insurance with respect to care, in line with the informality hypothesis.
An additional analysis presented in Table A8 suggests that the reason for the crucial role of Western Germany in control group might be partially credited to the protestant confession, work ethics and attitudes towards wealth and money (Weber & Kalberg, 2013). Communism changed individuals in orthodox and catholic societies to a greater extent with respect to financial family support and income redistribution, presumably because the welfare state institutions of financial public benefits had been developed in the industrialized areas of Germany, unlike in the rest of Europe (Hennock, 2007;Mommsen, 2018). In Germany, both Christian and social democrats have traditionally supported welfare state, whereas in Russia prior to Soviet revolution, both monarchy and church were against it. 12 These might explain why Western Germany is crucial for our control group including noncommunist countries to find EC effects on the preference for income equality (redistribution) and for financial informal (family) insurance.
It is worth pointing out that communism collapse resulted in deep economic change, which brought about structural unemployment and recession. We conduct estimations controlling for the experience of recession, especially in formative years, as well as the occupation at the time of communism collapse (and, in addition, occupation interacted 10 Detailed results of the analyses discussed in this section can be found in Appendix A.4. 11 Baseline EC measure uses the dates of the socialist constitution and first free democratic elections as the dates of entry to and exit from Soviet communism, respectively. Other EC measures use either fixed dates of entry (1945) or of exit (1990), or both the dates fixed. Another EC measure refers to the dates of the erection (1961) and the fall (1989) of the Berlin Wall as the entry and exit dates, respectively. 12 The Orthodox Church in 19th century Russia supported monarchy and was under the authority of state officials (Jowitt, 1992).
with graduation year), as well as the age at communism collapse, to proxy both the risk of unemployment during the transitional period as well as the chances of having relatively high socio-economic status under communism. Controlling for these factors yields results consistent with the main analysis. Moreover, the significance of the extensive margin of the exposure to Soviet communism in our results points to the key role of the communism rather than transition to free market economies.
The fact that we document substantial EC effects using various start and end dates is reassuring for the communism impact rather than the impact of transition to market economics after communism collapsed.
Finally, in an attempt to examine whether the exposure to Soviet communism can be reduced to the exposure of any type of authoritarian regime that had been present in Europe, we use as the control group the countries with highest score in the current democracy index and the set of countries with authoritarian past. We find that exposure to nondemocratic regimes cannot explain our main results found for EC, that is, the exposure to Soviet communism.
T A B L E 4 Coefficients on the exposure to communism (EC) on the preference for family and social insurance (intensive margin) excluding Western Germany from the research sample.
(1) (2) Note: FE-fixed effects. Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.

| Robustness checks
The data on which we base our empirical analysis come from years preceded by the collapse of communism; therefore, the effects observed using numerous waves should also be present for a single wave, if EC poses a long-lasting impact. Thus, we use the first wave of GGS (composed of larger country set than the second) to find EC effects on the preference for family insurance similar to the main results. Most recent waves of WVS and ESS reveal still present positive EC effects on the preference for social insurance, especially on their intensive margins, indicating that the EC effects are indeed long lasting.
We examine the robustness of our results to alternative estimation techniques. In particular, probit and logit models with dichotomized dependent variables on the preference for family and for social insurance, which continue to support our main results. Furthermore, we employ random effects models in the estimations of the preference for family insurance for the panel subsample of GGS and find that all significant effects remain, and their magnitudes remain at a similar level.
Another relevant consideration is that omitted variables may bias main estimates; thus, we examine the bias resulting from the selection on unobservables referring to Oster (2019) coefficient stability test. We find that relative degree of bias resulting from the unobservables' impact on preference for family insurance in all specifications is negligible, with values of delta below 0.001 (c.f. Table A18). Although the values of delta parameters are larger in the case of estimations for the preference for social insurance, they are close to zero and not greater than 0.06.
In order to use more detailed data on household socio-economic status instead of crude measures of the ability to make ends meet or scale of incomes, we draw on multiple imputation techniques. We use household average monthly income per capita with imputations (in logs) for the GGS sample. The sign and significance of EC effects in this specification remain similar to main results. In additional checks, we control for the experience of war. Our results are robust to specifications controlling for these experiences, particularly during impressionable years, which are found to be relevant for the preferences in later life. In addition, we control for rural area, country democracy index, European marriage patterns, individual religiosity, and household size when such information is available, to find results reassuring our main findings.
Finally, it can be argued that if there have been ongoing changes in individual preferences for family insurance, the inclusion of a time trend might help to remove the bias in main estimations insensitive to such changes. We allow for separate time trends in the East and West of Europe in additional analyses. The EC effects, remain significant and positive regardless of the analysed time trends, which is not the case for other examined preferences. By allowing for separate time trends (either linear or quadratic) in postcommunist and other countries, we identify negligible EC effects on the preference for income equality and income redistribution. As expected, time trends specific to country of birth show that the main results for particular familiarism remain robust.

| CONCLUSION
Against the backdrop of communism weakening the family, we document robust evidence that EC increased the demand for informal family insurance. Although political regimes are argued to result from family structures (Todd, 1985), in this paper we document that exposure to a political regime: Soviet communism increased the strength of family supports (insurance), alongside other forms of insurance. More specifically we show evidence of an average increase in the preference for family support to care for older parents by 10 percentag ponts (pp) and by 4 pp with regards to care and support for pre-school and adult children. Such effects are robust to a series of identification threats, sample compositions and other robustness checks.
The results presented in this paper provide a new perspective to the previous evidence (Alesina & Fuchs-Schündeln, 2007) of EC effects on the demand for insurance and the role of indoctrination to Marxist-Leninist values in formerly communist countries. Against the view of the previous litertaure, we find that the 'indoctrination hypothesis' explanation suggesting an increase in the demand for social insurance is weaker than the 'informality hypothesis', predicting the increase in the preference for different forms of informal family insurance. Our findings document that EC increased informality or network effects within the family, and consistently with other studies, we find evidence that it eroded social, institutional and interpersonal trust (Aghion et al., 2010;Booth et al., 2018;Rainer & Siedler, 2009).
Taken together, this study shows that in an environment that outlaws significant wealth accumulation, informal family networks become an additional form of insurance to protect against care needs and financial adversity, consistent with seminal economic theory (Becker, 1981). We find that during Soviet communism times, family networks became instrumentally valuable as a source of connection and status, despite the state propaganda aimed at the abo-  ' with the answer 'most people can be trusted' opposed to 'need to be very careful'; people being helpful-'people mostly try to be helpful' or 'people mostly look out for themselves'; confidence-how much confidence respondents have in press, political parties, police, labour unions, and justice system measured on five point scale; trust-how much respondents trust in parliament, politicians, police, and legal system, measured on 11-point scale ('no trust at all' … 'complete trust'). Civic participation: interest in politics-'how interested in politics'; discussing politics-'discussing political matters with friends'; political actions-dummy generated using responses to questions on undertaking various political actions; active (passive) organization membership-dummy using data on membership in up to ten civil society organizations, excluding religious ones; current (ever) trade union membership-'membership of trade union or similar organization' currently (currently or previously). Religiosity: church answers social (family) problems-whether 'your church is giving, in your country, adequate answers' to the 'social problems facing our country today' ('the problems of family life'); frequency of praying-praying 'outside of religious services' in ESS ranging from 'never' to 'every day' and in WVS from 'never or practically never' to 'several times a day'; importance of religion (God; family)-'how important in life is religion (God; family)' on four point scale ('not at all important … 'very important'); religiosity-'attending religious services apart from special occasions' in GGS recoded to the scale 0 = never, 1 = less than once every 3 months, 2 = one to three times every 3 months, 3 = one to three times every month, 4 = at least once a week, and in ESS ranging from every day to never at a 7-point scale. We use questions to what extent individuals disagree with statements presenting traditional gender roles: financial independence-women should be financially independent from their husbands; working women-men have more right to job than women, if jobs are scarce; working mother-pre-school children suffer when their mothers work; marriage-women should try to marry and have a child; motherhood-a women without children is fulfilled; parenting after divorce-children should stay with mother rather than father after divorce; importance of tradition-'important to follow traditions and customs'. Family trust based on question 'how much do you trust your family' ('do not trust them at all' … 'trust them completely'). All comparable variables statistically significantly different between the East and the West with the t test significant at 1% level. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-6 (release 2015_04_18), and ESS waves 1-8. T A B L E A 6 Effects of EC in different cohorts on preference for social insurance.
T A B L E A 7 Effects of EC in the lands of former Russian, Prussian, and Habsburg empires. (1) T A B L E A 9 Effects of EC in selected country groups on preference for family insurance.
(1) T A B L E A 1 0 Effects of EC in selected country groups on preference for social insurance.
(1)  (1956). Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on WVS waves 1-5 (release 2015_04_18) and ESS waves 1-8.

APP E NDIX C : MECHANISMS
F I G U R E A 5 Effects of the exposure to communism (EC) on the beliefs about people being helpful and fair Source: Authors' own estimations based on ESS waves 1-8. Note: Point estimates with 95% confidence intervals, controlling for scale of incomes, age (quadratic), gender, education, country, as well as time and cohort fixed effects. Dashed line shows the average effect of EC. People are helpful-'people mostly try to be helpful' or 'people mostly look out for themselves'. People are fair-'most people would try to be fair' or 'most people would try to take advantage of me'. Robust standard errors clustered by year of birth and country. ***p < .01, **p < .05. *p < .1. [Colour figure can be viewed at wileyonlinelibrary.com] T A B L E A 1 1 Average effects of the exposure to communism (EC) on civic participation and democratic values.
(1) (2) discussing politics-'discussing political matters with friends'; political actions-dummy generated using responses to questions on undertaking various political actions; active (passive) membership-dummy using data on membership in up to ten civil society organizations, excluding religious ones; current (ever) trade union membership-'membership of trade union or similar organization' currently (currently or previously). Democratic values: democracy-importance of democracy, taxation in democracy-taxes on the rich subsidizing the poor essential to democracy; incomes in democracy-state making incomes equal essential to democracy, rules in democracy-obeying the rules essential to democracy; women rights in democracy-women having the same rights as men essential to democracy. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 1 2 Average effects of the exposure to communism (EC) on religiosity and traditional family and gender roles.
(1) (2) Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Cohorts: year of birth. Religiosity: church answers social (family) problems-dummies whether 'your church is giving, in your country, adequate answers' to the 'social problems facing our country today' ('the problems of family life'); frequency of praying-praying 'outside of religious services' in ESS on seven point scale ranging from 'never' to 'every day' and in WVS on eight point scale ranging from 'never or practically never' to 'several times a day'; importance of religion (God)-'how important in life is religion (God)' on four point scale ('not at all important … 'very important'); religiosity-'attending religious services apart from special occasions' in GGS recoded to the scale 0 = never, 1 = less than once every 3 months, 2 = one to three times every 3 months, 3 = one to three times every month, 4 = at least once a week, and in ESS ranging from every day to never at seven point scale. Traditionalism: financial independence-women should be financially independent from their husbands on five point scale; working women-men have more right to job than women, if jobs are scarce on five point scale; working mothers-pre-school children suffer when their mothers work on five point scale; marriage-women should try to marry and have a child on five point scale; motherhood-women without children is fulfilled; parenting after divorce-children should stay with mother rather than and 2 (release 1.3) and ESS waves 1-8. Note: Point estimates with 95% confidence intervals, controlling for scale of incomes, age (quadratic), gender, education, country, as well as time and cohort fixed effects. Dashed line shows the average effect of EC. Insignificant coefficients not reported. Motherhood-women without children is fulfilled. Financial independence-women should be financially independent from their husbands. Working mothers-pre-school children suffer when their mothers work. Working women-men have more right to job than women, if jobs are scarce. Parenting after divorce-children should stay with mother rather than father after divorce. Equal earnings-it is not good for relationship of woman earns more than partner. All measured on five point scale where 1 denotes the least and 5 the most pronounced gender inequality. Robust standard errors clustered by year of birth and country. ***p < .01, **p < .05. *p < . Note: Young are the individuals born after 1989 in formerly communist countries. Old are the individuals exposed to communism for at least 14 years. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Demographic controls: age (quadratic), gender. Education controls: highest education level attained. Cohorts: year of birth. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 2-6 (release 2015_04_18), and ESS waves 1-8. Note: Young are the individuals born after 1989 in formerly communist countries. Old are the individuals exposed to communism for at least 14 years. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Demographic controls: age (quadratic), gender. Education controls: highest education level attained. Cohorts: year of birth. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 2-6 (release 2015_04_18), and ESS waves 1-8.

T A B L E A 1 3 (Continued)
T A B L E A 1 4 Coefficients on the exposure to communism (EC) in the research samples excluding Russia. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Family preferred over social insurance: upwards care-'care for older persons in need of care at their home', downwards care-'care for pre-school children', upwards money-'financial support for older people who live below subsistence level', downwards money-'financial support for younger people with children who live below subsistence level is mainly a task for society or mainly a task for family'. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Family preferred over social insurance: upwards care-'care for older persons in need of care at their home', downwards care-'care for pre-school children', upwards money-'financial support for older people who live below subsistence level', downwards money-'financial support for younger people with children who live below subsistence level is mainly a task for society or mainly a task for family'. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.

T A B L E A 1 4 (Continued)
T A B L E A 1 5 Coefficients on the exposure to communism (EC) in the research samples excluding Germany. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Uprisings in Czech Republic, Germany, Slovakia and Hungary. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.

D.2 | OMITTED VARIABLES BIAS
T A B L E A 1 8 Results of the coefficient stability test for the omission bias in the effects of the exposure to communism (EC) on the preference for family and social insurance (extensive margin). Note: Parameters delta were obtained conditional on the assumption that beta coefficient is equal to its estimate obtained in the examined model. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Demographic controls: age (quadratic), gender. Education controls: highest education level attained. Cohorts: year of birth. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 2-6 (release 2015_04_18), and ESS waves 1-8. Note: Parameters delta were obtained conditional on the assumption that beta coefficient is equal to its estimate obtained in the examined model. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Demographic controls: age (quadratic), gender. Education controls: highest education level attained. Cohorts: year of birth. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 2-6 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 1 9 Coefficients on the exposure to communism (EC) controlling for the experience of war. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 1 9 (Continued) Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 2 0 Coefficients on the exposure to communism (EC) controlling for the experience of economic recession during impressionable years. Note: Economic recession based on the World Bank data on unemployment rate. Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained.Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 2 0 (Continued) Note: Economic recession based on the World Bank data on unemployment rate. Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained.Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 2 1 Coefficients on the exposure to communism (EC) controlling for the occupation. T A B L E A 2 2 Coefficients on the exposure to communism controlling for democracy index. T A B L E A 2 9 Coefficients on the exposure to communism (EC) using fixed exit date. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 3 0 Coefficients on the exposure to communism (EC) using fixed entry date. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.
T A B L E A 3 1 Coefficients on the exposure to communism (EC) using dates of the Berlin Wall erection and fall. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own estimations based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.

D.4 | ALTERNATIVE ESTIMATION STRATEGIES
T A B L E A 3 2 Coefficients on the exposure to communism (EC) in the fixed effects panel estimation. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3).
T A B L E A 3 3 Coefficients on the exposure to communism (EC) in the random effects panel estimation. Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3). Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3).

T A B L E A 3 3 (Continued)
T A B L E A 3 4 Coefficients on the exposure to communism (EC) in the probit model.
( Note: Demographic controls: age (quadratic), gender. Income controls: ability to make ends meet (GGS) or scale of incomes (WVS, ESS). Education controls: highest education level attained. Preference for family insurance: upwards care-'children should take responsibility for caring for their parents when parents are in need', downwards care-'grandparents should look after their grandchildren if the parents of these grandchildren are unable to do so', upwards (downwards) money-'children (parents) ought to provide financial help for their parents (adult children) when their parents (the children) are having financial difficulties'. Preference for social insurance: income equality-'incomes should be made more equal', income redistribution-'government should reduce differences in income levels', equality of opportunities-'important that people are treated equally and have equal opportunities'. Robust standard errors clustered by year of birth and country. Statistical significance: ***p < .01, **p < .05, and *p < .1. Source: Authors' own tabulation based on GGS wave 1 (release 4.2) and 2 (release 1.3), WVS waves 1-5 (release 2015_04_18), and ESS waves 1-8.