Multiplicative or compensatory advantage? Multigenerational contribution to grandchildren's educational success in the Soviet and the post-Soviet contexts Open Access

Most research on reproduction of advantages in various contexts has been concerned only with parent–child relationships. Recently, interest in the grandparent effect on grandchildren has increased considerably, partly because of the availability of data for three generations. An emerging literature examines the transmission of educational attainment, social position and income across three or more generations (see, for example, Bol and Kalmijn 2016; Daw et al.2020; Deindl and Tieben 2017; Engzell et al.2020; Erola and Moisio 2007; Erola et al. 2018; Hällstein and Pfeffer 2017; Knigge 2016; Liu 2018; Modin et al.2013; Pfeffer and Killewald 2018; Sheppard and Monden 2018; Song and Mare 2017; Warren and Hauser 1997; Zeng and Xie 2014; Zhang and Li 2019; Ziefle 2016) and national contexts (initially in a range of Western countries and recently covering Chile, Israel, Mexico, Taiwan, Turkey, China and Korea).

The empirical evidence of the direct effect of grandparents, net of the effect of parents, is mixed so far (see review by Anderson et al.2018). Daw et al. (2020) comment that previous mixed results suggest a shift from the question of ‘do grandparents matter?’ to ‘how, when, and why, do grandparents matter?’. Chiang and Park (2015) propose that previous findings about the missing direct effect of grandparents might perhaps hide the heterogeneity of context and subgroups.

Variation across countries and time-periods might be explained by various institutional arrangements that may have an impact on multigenerational transmission of advantages and disadvantages (Alesina and Giuliano 2015; Alesina et al.2020; Mare 2011; Pfeffer 2014). Several recent papers also examine the long-term impact of large macro-level shocks (particularly changes of political regimes and political violence) on multigenerational mobility of specific groups (e.g. Lupu and Peisakhin 2017; Treiman and Walder 2019; Xie and Zhang 2019).

Another suggested reason for the mixed findings and therefore an important area of research on multigenerational stratification is the differentiation of resource transmission between the highest and lowest groups in the socioeconomic hierarchies. Some evidence has been recently provided to support the idea about the concentration of multigenerational association at the very bottom and the very top of the distribution of advantage (Hertel and Groh-Samberg 2014; Lindahl et al.2015; Mare 2011; Park and Kim 2019). The findings on the effects of grandparents among low-status parents seem to indicate that resource compensation takes place (Erola and Kilpi-Jakonen 2017; Jæger 2012; Ziefle 2016). On the other hand, the effect of grandparents among high-status parents might signal multiplication¹ of advantages (Anderson et al.2018; Chiang and Park 2015; DiPrete and Eirich 2006). However, Chan and Boliver (2013) argue that the grandparents’ effect is quite a general social force and operates throughout the class hierarchy and is not restricted to the two extremes. The current study aims to fill the gap in the increasing literature on multigenerational stratification processes by responding to Mare's (2011) requests to examine these processes in post-Soviet societies as well as in various locations in the socioeconomic hierarchy. Central and Eastern Europe countries and post-Soviet societies are particularly informative cases to examine the roles of grandparents in stratification processes, as these grandparents lived primarily in the pre-Soviet era while their grandchildren came of age in post-Soviet era (Mare 2011: 17).

In our previous research (Saar and Helemäe, 2021), we distinguished the short-term (intergenerational) and long-term (multigenerational) impact of Sovietization policies on the reproduction of educational inequalities in Estonia. Our results indicated that these policies (including repressions) were rather ineffective in short term, as the parents of our respondents, being the next generation of Estonian Republic's advantageous social group, had about the same chances to attain higher education as the offspring of Estonian Republic's non-advantageous social groups. In long-term Sovietization, policies produced unintended consequences, facilitating the transmission of advantage in three generational perspectives. In this paper, we consider parental and grandparental advantages as constituents of a long-term joint social background's impact on grandchildren's educational advantage.

Compared to intergenerational mobility, where social background of children is referred only to parental position in the socioeconomic hierarchy, the challenge of multigenerational research is to reflect more complex patterns of ancestors’ socioeconomic locations. While in some families, locations tend to persist across familial generations (e.g. both grandparents and parents top the socioeconomic hierarchy), in other families, parents might be more or even less advantageous as compared to grandparents. Accordingly, in research across three generations, capacity of two kinds of social groups might be juxtaposed: persistently (across both background generations) located in social hierarchy (cumulatively dis/advantageous) versus mobile groups, i.e. parents being the first generation in the respective extremes of the hierarchy. Given that both in research and in practice, grandparental contributions often remain ‘hidden’ compared with the contribution of parents, we call offspring of such first-generation parents, respectively, as being hidden advantageous ((HA) parents at the low end of hierarchy, while grandparents belong to the high end of the hierarchy) and hidden disadvantageous ((HD) parents at the high end of hierarchy, while grandparents belong to a low level of the social hierarchy). In this paper, in addition to traditional focus on compensatory and multiplicative (dis)advantage, we also consider the role of hidden aspects of (dis)advantage in long-term transmission of educational success in Estonia.

Estonia is an informative context because of the changes in Estonian society after WWII (i.e. from the pre-WWII Estonian Republic to the Soviet Estonia and to the re-established post-Soviet independent Estonian Republic), and resulting in (a) the transformation of the social hierarchies and (b) the revaluation of advantages and disadvantages associated with them, e.g. collective social mobility (Kazjulja and Paškov 2011). The forceful Sovietization policies aimed at abruptly halting the political, cultural and social continuity of the Estonian Republic and the intergenerational continuity of its elite (see Saar and Helemäe 2021). Just as in the much more studied China with its radical policies during the Communist Revolution in the 1950s and the Cultural Revolution (Alesina et al. 2020; Deng and Treiman 1997; Xie and Zhang 2019), the Sovietization policies were top-down, i.e. directed at restricting the opportunities of the higher social classes. These policies were different from the bottom-up policies in the Nordic countries, which aimed to equalize the opportunities for those who originate at the bottom of the social pyramid (Esping-Andersen 2015). Accordingly, in countries with experiences of Sovietization policies, peculiarities of multigenerational transmission of advantage are associated with the efficiency of these policies, i.e. with the extent to which the intergenerational linkage between previous advantageous social groups and next generations was destroyed.

More specifically, the advantageous position of the elite in pre-Soviet Estonia, after the regime change might turn into a disadvantage for offspring in Soviet Estonia (especially during the consolidation phase of the regime), and back to advantage for the third generation in post-Soviet (or already in late Soviet) Estonia. In this way, due to the change of regimes, the interplay of parental and grandparental positions in the social hierarchy might have specific consequences in Soviet and post-Soviet contexts as compared to developed democracies and market economies. Accordingly, we specify our main research question. How does regime-specific multigenerational advantage and disadvantage influence the attainment of higher education in the Soviet and post-Soviet Estonian contexts? Which kind of process – compensation or multiplication – prevails? Was the pre-war elite (G1) capable of supporting their grandchildren (G3) to overcome any ‘loss of advantage’ by the in-between (G2) generation? Or did only those grandchildren of the pre-WWII elite take advantage of grandparental (previously high) social status whose parents (G2) managed to secure persistence of G1's advantage?

We make two contributions to the existing literature. First, our analysis with its focus on the joint contribution of grandparents (G1) who reached adulthood in the pre-Soviet era and parents (G2) who reached the age of higher education attainment in Soviet Estonia on grandchildren's (G3) higher education adds evidence about the multigenerational transmission of advantage under changes of political regimes. Changes in mobility patterns were rather complicated: the working class experienced rather intensive group mobility associated with political regimes (first upwards and then downwards) (Kazjulja and Paškov 2011), but there is no such evidence about other social groups. Moreover, in post-Soviet Estonia, even during the period of the greatest changes, there was rather little upward intra-generational upward mobility. This period was characterized primarily by ‘forced social mobility’ (Helemäe 2011). The analysis of intergenerational mobility also indicates the decline of mobility in the 1990s, while the effect of the family background on a person's social position increased considerably (Saar 2010). The exploration of multigenerational mobility under this type of regime change could add an additional dimension to the debate about the processes of social inheritance and their context sensitivity. Second, to contribute to the multiplication versus compensation debate, we focus on interplay between parental and grandparental advantage by exploring interaction effects between the resources of parents and grandparents in Soviet and post-Soviet Estonian contexts. Exploring the importance of multigenerational social background for educational success in context of a wide range of measures applied to hinder intergenerational continuity is a novel contribution to the literature.

To explore the impact of multigenerational social background on the attainment of higher education in Estonia, we use retrospective data from the Estonian Family and Fertility Survey 2004.

A literature review presented by Anderson et al. (2018) indicates 58% of analyses report grandparents’ socioeconomic characteristics are associated with the educational outcomes of grandchildren, independently of the characteristics of parents. Various mechanisms were proposed to account for multigenerational transmission of inequality: direct transfer of material and cultural resources, social networks, preferences (Pfeffer 2014) and genetic effect (Liu 2018). Thus, grandparents might use their material resources and wealth to facilitate the educational attainment of their grandchildren, for example by paying school and university fees, extra-curricular activities, tutoring, etc. (Mare 2011).

Cultural mechanisms operate through shared social networks and family environments by shaping attitudes of grandchildren, their aspirations and academic achievements (Jæger 2012).² Grandparents might read to their grandchildren, help with homework, foster expectations about educational attainment and shape the reference frame for their educational choices (Hertel and Groh-Samberg 2014; Møllegaard and Jæger 2015; Zhang and Li 2019). Family traditions, particularly the social positions of grandparents, might affect educational aspirations (Long and Ferrie 2018) and explain why grandchildren from downwardly mobile families strive to return to the high-status positions of their grandparents (Hertel and Groh-Samberg 2014; Zhang and Li 2019). This multigenerational effect does not have to be based on the direct influence of grandparents (their images in family histories might serve as role models for their grandchildren). Research on intergenerational transmission of educational inequality acknowledges the contingence of the role of cultural mechanisms on socio-contextual factors, especially on the specificity of the educational system (Breinholt and Jæger 2020; Hu and Wu 2021). There are good reasons to suggest it in multigenerational research, as at the societal level, cultural mechanisms are durable. Thus, Wong (1998) argues that a strong intelligentsia tradition in many socialist countries may explain why cultural capital had continued to play a role even under socialism.

Recent research has focused on the heterogeneity of multigenerational effects, suggesting that grandparents influence grandchildren only under certain conditions of parental resources. The interaction between parents’ and grandparents’ education and/or social position provides information on the childhood environment of grandchildren that is not conveyed by either of these two measures considered in isolation (see also Wightman and Danziger 2014). Previous models suggest a stronger multigenerational effect at the extremes of status distribution (see for example Hertel and Groh-Samberg 2014). Some authors have reported stronger effects of grandparents among grandchildren when their parents have low levels of education or socioeconomic status (Braun and Stuhler 2018; Deindl and Tieben 2017; Jæger 2012; Wightman and Danziger 2014), others have found a stronger effect among the parents’ higher level social groups (Chiang and Park 2015; Lindahl et al.2015).

According to Mare (2011), the maintenance of institutions and their enduring impacts on families and individuals may be an important source of multigenerational effects. Institutional contexts, such as the design of the national education system and welfare state arrangements might structure the transmission of social inequality through three generations (Mare 2011; Pfeffer 2014). Elite educational institutions might, for example, consolidate family positions across generations. Also, legal and commercial forces likely determine the degree, to which certain families might maintain socioeconomic advantage across many generations (Pfeffer 2014). But the knowledge about the institutional conditions, under which grandparental resources might be particularly relevant is now rather limited, especially in Soviet and post-Soviet contexts.

Research has started to explore the long-term effects of distinct types of large shocks, such as disruption of social structure, leading to changes in social, economic and political conditions and even to political violence or forced migration (e.g. Becker et al.2020; Lupu and Peisakhin 2017). Legacies of such shocks have been found to be passed down through families from generation to generation, while the affected groups could partially recover from negative shocks. During revolutions and their exports (through annexation) into neighbouring countries, institutions were redesigned to support class-based preferential social policies and to shut down two critical channels of intergenerational transmission: transmission through income and wealth (e.g. inheritance) and transmission through formal human capital accumulation (e.g. schooling). Specifically, studies on multigenerational social mobility in China showed that the impact of the Chinese communist revolution on social stratification has both long- and short-term aspects. The situation of children of people who were from the privileged classes deteriorates immediately in the post-revolution phase, but a reversion towards the grandparents’ generation in social status also emerged (Alesina et al.2020; Xie and Zhang 2019). This paper explores if the Estonian case follows a similar pattern and what were the roles of compensation and multiplication in it.

The first Republic of Estonia was in place between 1918 and 1940. Subsequently, Estonia was part of the Soviet Union for 50 years between 1940 and 1991. In 1991, the independent Estonian state was re-established. These periods were distinct in terms of conditions for intergenerational transmission of resources within the family. A post-WWII social transformation in Eastern and Central European countries might restrict the direct effect of grandparents’ material resources on their grandchildren's educational attainment. It is true especially in Estonia, where wealth accumulation was low before WWII. Moreover, Sovietization policies in the pre- and post-WWII Baltic countries included direct expropriation of private property of ‘social alien elements’. Consequently, the context of state-socialism especially in former Soviet Union countries is hardly supportive of the mobilization of economic resources to secure intergenerational advantage.³

Generally, in neither pre-WWII Estonian Republic nor under Soviet rule, did family cultural resources play a role of signals of distinctive class culture, but rather one of fostering skills that directly enhance educational success. Thus, before the Soviet invasion, there was not any deep cultural differentiation within the population in Estonia. The whole population of the newly independent (1918) Estonian Republic had quite similar (peasants) social background being mainly members of one generational cohort; furthermore, only a half of one per cent of the population over 10-year old had a university education (Sirk 2007: 104). By this time, Estonia's intelligentsia had not yet developed any specific, recognizable cultural distinction. At the same time, those who held white-collar positions, were more educated, more capable of supporting the development of their children's and grandchildren's learning skills as well as the ability to cope with the changing requirements of the educational system.

Intergenerational continuity between the pre-war generation and their children was brutally targeted by Sovietization policies, which increased the perishability of family cultural resources. Especially during the early years of Sovietization, belonging to the intelligentsia might even be a disadvantage (Tomusk 2000). Sovietization policies aimed at the transformation of the social profile of both administrative and professional elites. An important instrument of Sovietization policies was to restrict educational and professional opportunities of offspring of the admittedly disloyal population. These were either both the former urban and rural elites or those whose extended families had been deported (Pilve 2017). One crucial type of restriction operated through an assessment of the political trustworthiness of both the applicants and their relatives. This referred not only to the applicants’ personal qualities but also to their parents’ social involvements and political activities (Veskimägi 2005). Through such regulation of access to higher education, the downward mobility of applicants whose parents were not considered loyal or had the greatest amount of economic and educational resources before 1944 was furthered, and they were excluded from the new social elite.

In the 1960s, the crude measures against the higher levels of the social hierarchy made space for measures of positive discrimination in favour of industrial (and agricultural) workers. The status of students became more important than their family background. Quotas were determined for admission to higher education institutions, which favoured young people who had already been working, as well as those with worker origins. Sovietization policies directly changed the opportunity structures for the parental generation (G2).

Paradoxically, Sovietization policies for most of the Estonian population even increased the symbolic value of education and culture, which were considered part of national awareness and a precondition of ethnic survival. Nevertheless, the same symbolic ethnic value of education might also differentiate the educational outcomes of Estonians, as being even more motivational for the higher positioned pre-war generation of grandparents to support the attainment of higher education by their grandchildren's generation, especially to compensate for the possible educational deficit of generation of their children.⁴ Thus, for the third generation of pre-war elite in late Soviet and post-Soviet Estonia, the main role of familial cultural resources was primarily to support skills formation, promote learning capabilities and to provide motivation. In uncertain times, the knowledge and learning skills of higher-positioned grandparents might directly support the learning process of grandchildren and foster expectations about their educational attainment. These features of cultural resources of pre-WWII white-collar workers might turn into a favour for their grandchildren, when belonging to the intelligentsia no longer had any negative consequences and even turned to have high symbolic and material value in post-Soviet Estonia. Researchers also argue that in the context of relatively economy-related egalitarian settings, the family's cultural environment rather than tangible material support is important for educational outcomes (Sheppard and Monden 2018).

By the late-socialist period, the ‘wrong’ social background had less negative consequences for descendants of pre-war elite (see Saar and Helemäe 2021). Moreover, the post-socialist transformation in the 1990s brought about a change in established social hierarchies: they were overturned and social differences were stretched. Based on these arguments, we could formulate the first hypothesis.

Based on previous literature that suggests a stronger multigenerational effect at the extremes of status distribution, we focus on the interaction between the education of the parental generation (G2) and the social position of respondents’ grandparents (G1). The Estonian context and previous research provide us with arguments in favour of two alternative hypotheses.

Compensation has been previously identified as particularly relevant for ensuring that high-status families do not encounter multigenerational downward mobility (Wightman and Danziger 2014). In the Estonian context, this concerns the capacity of Estonia's pre-war elite (G1) to secure to their grandchildren (G3) high chances to attain higher education. Compared to stable developed Western democracies, the elite in pre-war Estonia had trouble securing higher education for their own children due to the top-down Sovietization policies aimed at restricting the opportunities of those from the parental generation (G2) who had higher social origin. Consequently, G2s with advantageous social background might not be allowed to access higher education for which they were well prepared and motivated. This (in a certain sense downwardly mobile) segment of G2 without higher education (‘hidden’ advantaged) might also be strongly supported by their advantageous parents (G1) to secure higher education for their offspring (G3) (for the same processes in Hungary, see Andorka 1997; in China, see Alesina et al.2020). Among lower educated parents (G2), those with a high-level social background might be better positioned (and more motivated) than those with a low-level one to help their children to achieve higher education. Previous longitudinal research of three generations in Estonia indicates that despite the downward mobility of the first Soviet generation (of parents, i.e. G2), their children tended to rise to the position of their grandparents (G1, who are known as the Generation of the Estonian Republic) (Helemäe et al.2000). These peculiarities of the Estonian context and previous results allow us to predict multigenerational resource compensation.

Some previous studies seem to indicate strong multiplication mechanisms (see especially Chiang and Park 2015), i.e. those for direct grandparental transmission of advantage that is concentrated at the top of the social hierarchy (Anderson et al.2018).

According to the trajectory maintenance theory (Szelenyi 1978; Hanley and McKeever 1997), the advantage of cultural resources might ‘overweigh’ the disadvantage related to the ‘wrong’ origin and secure educational credentials for children of not only the new socialist (i.e. administrative) elites, but also of the former, primarily professional, elites. Wong (1998) argues that a strong intelligentsia tradition in many socialist countries explains why cultural resources and education had continued to play a role even under socialism.⁵ Early (‘consolidation’ with the strong impact of political factors on social selection) and later (‘inclusion’) phases of state-socialist societies⁶ were characterized by the differing roles of familial advantage. At the early stage, previously advantageous family resources were hardly supportive (even negative) in prospering. But later, when the new elites were established, policies were no longer focused on the elimination of advantage of the ‘wrong’ classes.⁷ The role of affirmative action, i.e. the legitimation of the lack of cultural resources, had been decreasing, and institutional compensatory measures to equalize ability were introduced. Accordingly, the roles of parental and grandparental advantage, primarily cultural resources and education in the educational success of offspring had increased, and multiplication tendencies had come into force already under late Soviet and even more in post-Soviet Estonia (Helemäe et al. 2021; Saar and Helemäe 2021).

The types of groups produced by the interplay of grandparental and parental advantage are presented in Table 1. In late Soviet and post-Soviet Estonia, the groups have some distinct features as compared with developed Western market democracies: educational attainment of G3 was conditioned by the experiences of previous generations and educational opportunities for G3, but also by the reaction of all three generations to early Soviet class-based educational policies.

Hypotheses about the comparison of respondents’ (G3) chances to attain higher education are presented in Table 2.

The empirical analyses are based on data from the nationally representative Estonian Family and Fertility Survey 2004. The nationally representative sample of households was randomly selected from the Estonian Population Register on the basis of stratified systematic sampling. We use retrospective data on parents and grandparents of surveyed adult respondents. As a result, the dataset combines data on current (G3) generation outcomes with information on the education of their parents (G2) obtained during Soviet period and grandparents’ (G1) social status in pre-Soviet period. To ensure that data on social status of G1 represent the situation in the Republic of Estonia before WWII, our analyses use data of respondents whose mothers were born between 1922 and 1954 and among them only those whose grandparents (i.e. G1) reached adulthood in pre-war Estonia. For the same reason, we include only those respondents, who had at least one G1 from the maternal and at least one G1 from the paternal lines being born in Estonia. The range of 1922–1954 for the birth year allows us to distinguish those who were old enough to gain admission into higher education institutions during the Soviet period, given that the main restrictions for descendants of pre-war elites were applied at this level of education. Being raised in a family with both parents might be a certain kind of advantage, thus given our interest in the transmission of advantage, we focused on respondents who were raised in families with both a mother and a father. Our final sample comprised 1238 respondents.⁸

We use the attainment of higher education by current (G3) generation respondents as the dependent variable of the analysis. The main independent variables are parental (G2) educational attainment and grandparental (G1) social position, measures that had been proved to be relevant by previous research (see Helemäe et al.2000; Saar and Helemäe 2017). In the questionnaire, the education of the mother and the father of the respondents (G2) were measured by a 7-category scale: never attained school/primary education is not finished; primary education; basic education; secondary education, specialized secondary education, higher education; science degree. These scales formed the basis to create a measure for each of the G2's level of education: low (basic education or lower), medium (secondary) and high (at least higher education). To capture the additive character of (G2) parental resources, a synthetic indicator for parental educational resources was constructed: ‘low level’ was assigned to respondents whose parents both have low education; ‘high level’ was assigned in the case of one of parents having at least higher education; all others were considered as having a ‘medium level’ of parental educational resources. The grandparents’ (G1) pre-Soviet social position was measured on the basis of a question about pre-war social position of respondent's antecedents for maternal and paternal lineages using the scale: agricultural workers and unskilled blue collars; skilled blue collars; farmers; white collars. To take into account, the additive nature of the G1 effective distinguished at least one grandfather had a white-collar job – as the most advantageous social position as contrasted to while all other non-advantageous combinations of G1 social positions of both lineages.

Research in Estonia demonstrates that parental cultural resources, manifest in a large home library, enhance children's attainment of higher education, indicating with parental education, a learning-rich home environment (Saar and Helemäe 2017). To better capture learning richness as a feature of parental (G2) cultural resources, we also included in the analysis an indicator of the size of the parental library.⁹ Based on the question ‘How many books were in your childhood home?’, we aggregated an initial six-category scale (no books at all; less than 50 books; 50–149 books; 150–499 books; 500–999 books; 1000 and more) into a briefer three-category one: less than 150 books; 150–499 books and more than 500 books.

Sheppard and Monden (2018) hypothesize that G1's influence on their G3s depends on the latter's numerical quantity. We include the variable of the number of respondents’ (G3's) siblings into the model because G3 with fewer siblings may benefit more from their G1's social position than G3 who have more siblings. Other control variables include the respondent's gender (0 = male, 1 = female), and birth cohort (1 = 1942–1963, 0 = 1964–1983). Table 3 presents the descriptive statistics for all the variables used in our analysis.

We proceed with the analysis in a series of steps to determine how the advantageous position of G1 is associated with G3's education. First, we descriptively assess the effect of the social position of G1 net of G2 education by comparing the educational attainment of G3 across the G1 social position and G2 education. Second, we estimate a series of logistic regression models. The first three models include the social position of the G1 and the educational level of G2 to test the gross effect of G1 and G2 separately and together. We also estimate separate models where in addition to education parental resources–parental occupation (Model 4) or the size of the library in parental home (Model 5) is added to the variable of G1s’ social position and G2's education. The final one (Model 6) includes, in addition to G1s’ social position and three measures of G2 resources, the number of G3's siblings, G3's birth cohort and G3's gender as demographic controls. In the next step, we estimate two-way interactions between the social position of G1 and the education of G2 to investigate whether multiplication or compensation mechanisms dominate and how combinations of G1's social position and G2's education contribute to the educational success of G3. For all the models, we display average marginal effects (AME). The major advantages of AME for logistic regressions are that it is possible to compare effects across models and groups (Mood 2010) and due to having estimates as predicted probabilities results of interactions can be interpreted more correctly (Mize 2019).

Attainment of education by (G3) respondents varies according to the social position of their G1 and G2 (see Table A1 in Appendix). In line with previous research, this multigenerational effect works largely through G2 education. Thus, at every level of G2's education, the G3's chances to attain higher education are higher among descendants of highly positioned G1 as compared to families where none of the grandparents holds white-collar job in pre-WWII period. Disregarding G1's data tends to underestimate differences in attainment of higher education within G2's educational groups, i.e. multigenerational character of transmission of advantage, especially in case of offspring of highly educated G2. At the same time, there is also room for G1's effect: for each level of G2 education, the chances of G3 respondents to attain higher education are higher for descendants of G1 with higher social position. Thus, among G3 children of G2 with high level of education, higher education was attained by 52% of descendants of highly positioned G1s as compared with 15% among those of lowly positioned G1. For G3 respondents whose parents (G2) did not have higher education, the respective figures are 26% versus 8%.

Results of logistic regressions confirm the importance of multigenerational approach to transmission of educational advantage. Table 4 shows the results of logistic regressions in the format of AMEs on the predicted probabilities.¹⁰ Model 3 estimates the association between the educational level of G2, the social position of G1 and the attainment of higher education by G3. According to AME's, both G1 and G2 generations (statistically significant) contribute to the educational success of G3. A comparison of Model 3 with Models 1 and 2, where associations of G2 and G1 with studied outcomes were estimated separately, indicates an overlap of these associations, i.e. the interplay of G2 and G1 influences. Models 4 and 5 provide estimates of the net G1's social positions and G2's education associations with G3's attainment of higher education controlling for G2's other important resources (G2's occupational status and number of books in respondent's childhood home), demonstrating the persistence of G1's and G2's effects. Model 6 incorporates the full set of G1's and G2's characteristics and control variables (the odds ratios for the whole – basic – model is presented in Appendix A2). Once again, the advantage of both G1 and G2 is clear. This result supports our first hypothesis: In Estonia, G1's social position has a positive effect on G3's educational attainment, net of G2's education.

To investigate the hypothesis of differential effects of the social position of G1 at the top and bottom of the G2's educational hierarchy, we included these specific interaction terms in the final logistic model (Model 6). Following Mize (2019), we rely on AMEs to reveal whether the effect of an advantageous G1's social position on the probabilities for G3 to attain higher education prevails at the lower or upper extreme of the G2's educational distribution (the former has compensatory nature, while latter has a one multiplicative). Table 5 shows that compared to a lower positioned G1, an advantageous G1 provides just slightly higher probabilities to attain higher education for the offspring of parents with low level of education, this (so-called first difference – see note under Table 5) gap between HA and cumulative disadvantage (CD) is not statistically significant. Moreover, this (HA–CD) gap does not significantly differ from analogical (first difference) gap between Medium advantage (MA) and medium disadvantage (MD) for the medium level of parental education, indicating very weak compensatory processes (see respective second difference). This result does not support our hypothesis about compensatory advantage.

Table 5 also shows that an advantageous G1 secures significantly higher probabilities to attain higher education for offspring of parents with high level of education as compared to the contribution of non-advantageous G1. This gap between respondents with cumulative advantage (CA) versus HD social backgrounds is statistically significant and significantly (at the level of p < 0.10) wider than the analogical gap (between MA and MD) for the medium level of parental education, providing support for our multiplicative advantage hypothesis.

Table 6 allows further exploration of the differences in probabilities to attain higher education between respondents with various multigenerational social backgrounds. Respondents with a cumulatively advantageous background have the highest, i.e. statistically significant higher probability to attain higher education as compared to the other five identified types of multigenerational background. The results presented in Table 6 also show that CD respondents have a lower probability to attain higher education compared to all others except for respondents with HA background (G1's advantage does not matter only when G2's education is very low). The picture is more complicated when we turn to the respondents with HD and HA social backgrounds.

Respondents with HD (i.e. newcomers to the high level of social hierarchy) multigenerational background clearly illustrate the importance of advantageous G1's for the educational success of respondents. First, an HD background is associated with lower chances to attain higher education as compared to CA (i.e. persistently advantageous) social background. Second, an HD background is not associated with significant advantage over respondents with lower levels of parental (G2) education if the latter groups have advantageous grandparents (G1). But respondents with a CA background outperform other groups with G1's disadvantageous status.

In summary, the compensation capacity of HA is poor, while the multiplication capacity of CA is strong.

Based on retrospective data from the Estonian Family and Fertility Survey 2004, we investigated the grandparent effect in Estonia. In particular, we explored heterogeneity in the grandparent impact jointly with parental one by including interaction effects between the parent's and grandparent's resources in the models and contributing to the multiplication versus compensation debate. Our analyses yield three main empirical findings. First, it is important to take into account the role of grandparents in the status attainment processes as grandchildren also can benefit from grandparents’ advantageous resources even in a societal context, where at some point, strong policies were enacted to restrict the impact of any such advantages.

We found evidence that the grandparental generation (G1) was able to transfer its advantage associated with a high position in pre-Soviet Estonian society to their grandchildren (G3) in post-Soviet and even in late Soviet Estonia, supporting their attainment of higher education. Our results show that G1 contributes to the formation of G3's educational advantage primarily through processes of multiplication. Thus, the late/post-Soviet G3 generation from families with a high level of parental (Soviet G2 generation) educational resources benefit more from grandparental (pre-Soviet G1 generation) advantageous social position as compared with the Soviet G2 with medium level of education, indicating the condition of multiplicative accumulation. Moreover, those high-positioned pre-Soviet grandparents who succeeded despite the restrictions to secure higher education for their children, i.e. the G2 generation, also strengthen for their grandchildren the benefits of having highly educated (Soviet generation) parents. Respondents (G3) with such (CA) multigenerational social background have the highest chances to attain higher education compared to G3 with all other types of multigenerational background, including offspring of parents who attained higher education (advantageous status) in Soviet Estonia while being themselves offspring of disadvantageous social groups in pre-WWII Estonia.

The process of compensation at the low level of the social hierarchy is rather weak. Indeed, the late/post-Soviet G3 generation from families with low level of parental (Soviet G2 generation) educational resources benefit from grandparental (pre-Soviet G1 generation) advantageous social position to the same extent as the Soviet G2 with medium level of education. Moreover, a few G1 failed to support the attainment of at least secondary education by their children (G2). There is a good reason to consider this type of multigenerational social background (HA) to be the outcome of negative selection. Accordingly, these G1 were unable to compensate for any deficiency in G2's educational skills. Consequently, the pre-Soviet G1 generation contributes to the overall increase in post-Soviet intergenerational inequality.

Second, in the context of regime change and long-term influence of top-down Sovietization policies on the transmission of pre-Soviet advantage, our results overall reflect the limited effectiveness of these policies in the long term. The mere evidence that educational success in the late Soviet and post-Soviet periods is traceable to the G1 generation is the compelling argument in favour of this conclusion. Moreover, the type of multigenerational background that illustrates the effectiveness of restrictive policies (HA, i.e. those G2 who experienced significant loss of advantageous G1's status) is rather small, making up about 8% of respondents with advantageous G1 background. But importantly, having such a rare multigenerational background is associated with a significant loss in G3's educational opportunities only compared with the multigenerational background suggestive of the failure of Sovietization policies (CA, i.e. G2 continues to enjoy G1's advantages despite restrictions brought about by regime change). Sovietization restrictive policies were hardly effective in the sense that they did not result in significant educational success of the least advantaged group in pre-WWII Estonia. Furthermore, according to our results, among descendants of this group, there is a modest percentage (about 13%) of respondents with a multigenerational background that illustrates outcomes of their institutionally supported by Soviet policies access to higher education (HD, i.e. G2 attained advantageous status while being offspring of non-advantaged G1). Respondents with such a social background (both preferable and supported by Soviet authorities) still have a lower probability of attaining higher education as compared to those against whom restrictive policies were designed (CA). With regard to the HD type of social background, the ineffectiveness of Soviet policies against elites is illustrated by our results that descendants of HD have a higher probability of attaining higher education as compared to descendants of other non-advantageous G1 (i.e. MD and CD). However, descendants of HD also have about the same as (but not higher than) offspring of parents with lower level of education but with an advantageous G1 background (i.e. MA, even HA, meaning once again that G1's advantage matters).

To put it in another way: advantageous G1s converted the failure of top-down Sovietization policies towards their children (G2) into an advantage for their late or post-Soviet grandchildren (G3) – a compelling argument in favour of G1's important role in intergenerational transmission of advantage in Soviet context.

How can we reconcile our findings with the previous research? First, most articles have found evidence in favour of the compensation hypothesis (see, for example, Braun and Stuhler 2018; Deindl and Tieben 2017; Jæger 2012; Wightman and Danziger 2014) compared to the multiplication hypothesis (Chiang and Park 2015; Lindahl et al.2015). Our results show that in the late Soviet and post-Soviet context, compensation (at least for strong educational deficiency of parents) is hardly traceable. Secondly, recent literature exploring the short- and long-term impact of societal changes in China has found both of them. Thus, Xie and Zhang (2019) indicate that the effects of the Chinese revolution were most pronounced for the birth cohorts immediately following the revolution, attenuating for the third generation, but the impact of the revolution exerted on the educational attainment by the introduction of class classification remains strong. These results seem to demonstrate the importance of (a) the longevity of the ‘new order’ (in stark contrast with China, the change of orders in Estonia more or less coincided with the change of generations); (b) the dynamics and nature of ‘reversal’ changes in institutions and politics in the instance of the ‘reversal change of order’ (as in Estonia).

In summary, as Mare (2014: 125) indicates ‘ … well-executed descriptive studies of multigenerational associations are a value in their own right’. Our analysis is mainly descriptive, concentrated on pre-Soviet (G1), Soviet (G2), also late Soviet and post-Soviet generations (G3). We added information to the corpus of knowledge about multigenerational inequality under the conditions of massive social transformations from the rather egalitarian society of the pre-WWII Estonian Republic (with a low level of wealth accumulation and without any clear cultural distinctions between levels of the social hierarchy) to Sovietized Estonia and further to the re-independent country that experienced a highly neoliberal transition.

Four limitations could be raised against our study. First, measures of parental education, their occupational position and size of library as part of cultural resources used in this study may not fully capture parental family's backgrounds. As a result, the effect of grandparents’ social position on the educational attainment of grandchildren may be biased by unobserved parental characteristics correlated with both the children's educational attainment and the grandparents’ social position. We are aware, that our inclusion of the size of the parental library is insufficient to overcome this limitation. But bearing in mind low-wealth inequality in the grandparental generation and the Sovietization policies directed against accumulation and inheritability of wealth, the lack of wealth information is not an important limitation. The second limitation of our data is that we only have information on maternal and paternal grandfathers, i.e. we did not have any data for grandmothers. Due to this lack of data, we probably underestimated the effect of grandparents. In the event of having this data, we might be even more confident in our conclusion about the importance of the grandparents’ effect. Third, the retrospective data used in this study suffer from survivor and recall biases (Mare 2011). The grandparent generation is recalled only when they have both children and grandchildren. Unfortunately, we do not have data about the fertility and survival of the grandparental generation. Retrospective data provides only an approximate measure of grandparent's resources. Using retrospective information on G1 information may lead to measurement error and attenuation bias (Anderson et al. 2018). Our possibilities to directly assess the influence of grandparents’ resources are limited until more prospective data for all generations become available. It is a rather empirical question whether the availability of these data will strengthen or weaken the estimated effect of grandparents but it would help to better understand the concrete mechanisms behind the grandparents’ contribution to multigenerational mobility. Fourth, our study is, by design, limited to two parent families because we wanted to rely on all available information on both parents and their grandparents. The role of grandparents could be different in single-parent families (Song 2016).

We should mention methodological issues in the grandparent literature that may well explain the different findings across studies. Anderson et al. (2018) mentioned that the number and detail of parental stratification indicators observed have no relation to the likelihood of detecting an independent grandparent effect (omitted variable bias at the parent level), Engzell et al. (2020) highlighted measurement error at the parent level and Breen (2018) raised concerns about collider bias in the grandparent literature. One of the conclusions of Enzgell et al. (2020) analysis is that ‘compensating’ effects of G3 status at low levels of G2 status emerge because of the greater difficulty of observing status accurately at the lower end of the distribution. However, we suppose that our results somewhat exaggerate grandparents’ importance. We should also have in mind that there are very different direct effects of grandparent social position depending on the parental characteristics. This must be taken into account in further analysis and requires modern causal mediation analysis techniques (see Van der Weele 2015).

We concentrated on the impact of resourceful grandparents and their compensation or multiplication role for their grandchildren. Further analysis would pay more attention to the groups of respondents with disadvantageous social backgrounds using analysis regression models as well as an alternative case-based qualitative comparative analysis (QCA) (see Ragin 2008). The QCA method would enable us to separate an outcome (attainment of higher education) from the analysis of its negation (no attainment). This is important because a causal condition may be consistently connected to one but not the other (Ragin and Fiss 2017).

The strength of the grandparents’ effect may vary by sociodemographic variables (e.g. by gender or race) as the mechanisms and processes through which multigenerational effects are generated may differ (see Pfeffer 2014). Some researchers already approached these topics and found evidence of gender- or race-specificity of grandparents’ effect (Daw et al.2020; Ziefle 2016). In future work, we intend to study gender differences on the strength of multigenerational effects on educational attainment. Further work is also needed to further study both extremely advantaged and disadvantaged social groups in order to explore the multigenerational effects in both the elite and extremely deprived groups. Beyond the compensation versus multiplication debate, an additional mechanism of grandparents’ effects appears in recent literature: the effect of shared lifetimes between grandparents and grandchildren (see Bengtson 2001; Daw et al.2020; Lehti et al. 2018; Zeng and Xie 2014). Due to abrupt social changes in Estonia, this topic would be very interesting for future research.

No potential conflict of interest was reported by the author(s).t

Ellu Saar is Professor of Sociology at Tallinn University. Her main research interests include social stratification and mobility, educational inequalities and lifecourse studies. Her articles have appeared in Journal of Education and Work, European Sociological Review, International Journal of Comparative Sociology, Higher Education, European Societies and East European Politics and Society.

Jelena Helemäe is a Senior Researcher at Institute of International Social Studies, Tallinn University. Important fields of research are social inequalities, ethnic stratification and labour market. Recent articles have appeared in Comparative Sociology, International Journal of Sociology, East European Politics and Societies.