Multigenerational effects on educational outcomes: a systematic - - PDF document

1

Multigenerational effects on educational outcomes: a systematic review

[Multi-generational Social Mobility Workshop version, September 2017]

Lewis Anderson, Paula Sheppard & Christiaan Monden Department of Sociology, University of Oxford lewis.anderson@trinity.ox.ac.uk

Abstract

Is the intergenerational transmission of educational attainment a Markov process? Interest in this question has swelled recently and results from a variety of studies have tended to suggest that grandparents’ socioeconomic characteristics are associated with children’s educational outcomes, independently of the resources of the parental generation – a finding with important implications for the estimation of populations’ social mobility. But how robust is this association? In this paper we critically and systematically review the extant literature on grandparent – or multigenerational – associations with children’s educational outcomes. By comprehensively surveying the field, we explore whether findings vary substantially by study features such as country, data availability, the

• perationalisation of grandparental resources, and controls for parental characteristics. Results

indicate a striking lack of any link between such variables and the likelihood of an analysis finding a significant association. We then discuss what light the literature taken as a whole can shed on recurrent issues such as the validity, interactivity, size, distribution, and mechanisms of grandparent effects. This research is supported by European Research Council consolidator Grant for the FAMSIZEMATTERS project

2

• 1. INTRODUCTION

A long and fruitful tradition in sociology has examined the intergenerational persistence of socioeconomic characteristics. A wide variety of analytical approaches and datasets have been deployed in this enterprise, but the great majority of studies have in common a two-generation approach: for the purposes of determining the influence of family background on an individual’s life chances, the former is effectively equated with the characteristics of that individual’s parents. This approach makes an implicit assumption about the intergenerational transmission of advantage over successive generations, namely that this long-run process can be adequately described as a series of independent associations between adjacent generations. Thus, insofar as grandparents affect the

• utcomes of their grandchildren, this effect is indirect, or Markovian, which is to say that the

grandparental effect is fully mediated through the parental generation. An important recent development in the social sciences has been the increasing availability of data

• n more than the usual two generations (Song & Campbell, 2017). A recurrent finding from these

data is that the Markov process envisaged by the two-generation literature overestimates social

• mobility. That is, taking estimates of the persistence of social status based on two observed

generations and extrapolating over three or more tends to yield an estimate of this persistence over three or more generations which is lower than an estimate based on three or more actually

• bserved generations. Whereas Becker & Tomes concluded that ‘[a]lmost all earnings advantages

and disadvantages of ancestors are wiped out in three generations’ (Becker & Tomes, 1986: S1), more recent empirical findings do not bear this out (e.g. Kroeger & Thompson, 2016; Pfeffer & Killewald, 2016; Lindahl, Palme, Massih, & Sjögren, 2015; Clark, 2014). Socioeconomic status does not appear to decay geometrically across generations. Two types of explanation have been proposed to account for this. Clark (2014) posits an underlying latent factor determining socioeconomic status, which is transmitted between adjacent generations (in a fully Markovian process) at a higher rate than any observed measure, with an intergenerational correlation coefficient around 0.75. According to this argument, intergenerational correlations in

• bserved measures such as income or education are lower because the latent factor does not

translate perfectly into observed measures of socioeconomic status. This is both because individuals are subject to random disturbances (‘market luck’), and because they may choose to trade off dimensions of status, for instance forgoing a high-income occupation for one with more prestige. Clark holds that the intergenerational correlation of this latent factor is constant across a range of contexts, resistant to macroeconomic conditions or policies aiming to increase social mobility, and hence further that it is for the most part genetically transmitted. A major limitation of Clark’s evidence is that it depends on extrapolating from the top (e.g. US physicians, the Swedish nobility, the extremely wealthy) to the whole of the status distribution. Also, Breen observes that genetic relatedness ‘disappears quickly’ over generations, such that an ‘truly remarkable degree of genetic assortative mating’ (2015: 302) would be necessary for the long-term persistence of social status to be driven primarily by genetics. The more common explanation for the non-geometric decay of socioeconomic status across multiple generations is that ancestors more distant than the parental generation directly affect a child’s

• utcomes. That is, their influence is not mediated through intervening generations as in the

traditional two-generation approach or approaches like Clark’s. As figure 1 illustrates, research exploring whether this is the case has proliferated recently, especially since Mare (2011) called for a ‘multigenerational view of inequality’, arguing that the processes generating social stratification are unlikely to be fully described by the usual approach of beginning with individuals and looking at their parents as the determinants of their relative positions. Not only does this usual approach miss the question of how successfully individuals are able to reproduce social advantage – a question which

3 requires attention to demography – it misses the possible channels beyond one’s most immediate ancestors through which family background influences life chances. Figure 1 Studies testing a direct grandparent effect on educational outcomes (N=40) With a focus on educational outcomes, this paper examines efforts to answer Mare’s call, as well as earlier relevant evidence. That is, we critically and systematically review the evidence for a direct effect of grandparental resources or characteristics (Generation 1 or G1) on grandchildren’s educational outcomes (G3). Within this broad survey, we address several specific issues. First, given that the majority of studies report some form of significant G1-G3 association, how robust is this association? It may be, for instance, that apparent direct grandparent effects are due to measurement error in parental socioeconomic status. We examine whether study characteristics such as the level of information available on parental socioeconomic status are correlated with finding a grandparent effect. Second, through what mechanisms might a grandparent effect act, and what is the evidence for these? Third, we examine evidence on the distribution of the grandparent effect, since there are reasons to expect heterogeneity in its operation. We use a mix of quantitative and narrative methods to give a broad overview of the evidence whilst also paying attention to consequential particulars. The paper is organised as follows. Section 2 describes the methods of our systematic literature search and review. Section 3 presents results, both quantitative and narrative. In section 4 we discuss our findings and offer conclusions and directions for future research.

• 2. METHODS

2.1 Literature search

We restrict the scope of this review to educational outcomes as education is arguably the central stratifying characteristic in modern societies. Moreover, this restriction keeps the review

4 manageable and the subject matter cohesive. As well as other standard stratification outcomes such as earnings, occupational status and class (e.g. Chan & Boliver, 2013; Dribe & Helgertz, 2016; Knigge, 2016; Mukherjee, 1954), the literature on multigenerational effects extends to the intergenerational transmission of fertility outcomes (Kolk, 2014), longevity (Piraino, Muller, Cilliers, & Fourie, 2014), and body mass index (Modin & Fritzell, 2009). These fall outside the scope of this review. We also exclude studies which take the ‘horizontal’ approach of examining cousin correlations and thereby capturing the full, observed and unobserved, influence of grandparents. To identify relevant studies, in addition to our initial knowledge of the literature, a database search was conducted, and forward and backward reference searching were also employed.1 We searched Web of Science ('All Databases' option) with the following search terms: (multigeneration* OR "multiple generations" OR grand-parent* OR grandparent* OR great- grandparent* OR grandfather* OR grand-father* OR great-grandfather* OR grandmother* OR grand-mother* OR great-grandmother* OR "three generations" OR "four generations" OR lineage*) AND (socioeconomic OR socio-economic OR "status attainment" OR "social mobility" OR "social class" OR "class mobility" OR education* OR schooling OR occupation* OR income* OR inequalit* OR earning* OR wealth OR stratification OR “human capital”) Studies were included if they estimated any effect of grandparental resources or characteristics (of any kind) on grandchildren’s educational outcomes, whilst also controlling for socioeconomic characteristics in the parental generation.

2.2 Data extraction

Quantitatively summarising a large and heterogenous literature requires simplifying assumptions. Towards the aim of aggregating across studies while remaining sensitive to the loss of nuance this entails, two alternative approaches to aggregation are taken: study-level and analysis-level. We define a study as a single publication, within which is nested one or more analyses. Separate analyses within a study are distinguished only by outcome, population, or both of these. Each study is coded 1 or 0 for finding a grandparent effect according to the following protocol. If the majority (or half) of analyses within a study show a grandparent effect, the study is coded 1. Each analysis within a study is coded 0 or 1 for finding a grandparent effect, i.e. any independent G1-G3 association significant at the 0.05 level (even if others within an analysis are non-significant). A distinct analysis within a study is defined by the presence of any of a) a distinct outcome measure, b) a distinct population, or c) a distinct combination of outcome and population. However, the most aggregated analyses available are used from each study. For instance if results are reported for men, women, and for the full sample, the only analysis taken from the study is the full sample analysis. Populations are regarded as distinct if they differ according to country, cohort, sex of the grandchild,

• r set of three generations (i.e. in a study with data on four generations, a G1-G3 analysis is

regarded as distinct from a G2-G4 analysis). For example, Lindahl et al. (2015) contains three analyses: one tests for a direct G1-G3 association in years of education, a second does the same for G2-G4, and a third analysis tests a direct G2-G4 association for the outcome of whether the

1 We are also very grateful to Guido Neidhoefer, who kindly shared with us the results of his search of the

literature.

5 grandchild has completed an academic high school track. Other changes between samples or analytic approaches are not regarded as constituting distinct analyses: for example examining those with data on maternal and paternal grandparents separately (Chiang & Park, 2015), first reporting preliminary regressions and then structural equation models (Grant, 2005; Warren & Hauser, 1997),

• r excluding those born when their great-grandfathers were still alive (Lindahl, Palme, Sandgren-

Massih, & Sjögren, 2014). Where multiple terms for grandparent resources are included, whether simultaneously (Møllegaard & Jæger, 2015) or in separate models (Bol & Kalmijn, 2016), this also does not constitute a distinct analysis. Where such differences produce different results, this will be reported in the narrative analysis. Where a sequence of models is reported for a given analysis, the model with the greatest number of covariates is used (the fullest model). There are two sorts of exceptions to this.2 First, where interaction terms are introduced, the fullest model prior to the introduction of interactions is used. The descriptive results therefore reflect grandparent associations across whole samples – that is, averaged across the distribution of potentially important moderators. This may of course obscure heterogeneity in G1-G3 associations. Results of analyses including interaction terms are discussed further below. The second exception to the rule of taking the result from the fullest model in an analysis is that where potential mediators of a grandparent effect which are characteristics of the grandchild are included, the model prior to the inclusion of such terms is used. For instance, Yeung & Conley (2008) introduce a term for the child’s self-esteem, which a priori may lie on a causal path from grandparent education to child’s educational outcome. In addition, where applicable, we use the results from the model with the greater number of grandparent terms (e.g. a model with a term each for grandfather’s and grandmother’s education is favoured over one with a single term for the education of the most-educated grandparent). Lastly, where different analytical approaches are taken, the analytical strategy in which the authors place greatest confidence is used. For example, Hällsten & Pfeffer (2017) apply both ordinary least squares and a marginal structural model with inverse probability of treatment weighting, and regard the latter as more robust. Sharkey & Elwert (2011) and Song (2016) also employ marginal structural models, and results from these are reported. Structural equation models are favoured over

• regressions. However instrumental variable analyses are not included (Behrman & Taubman, 1985;

Lindahl et al., 2014; Sauder, 2006) as they are subject to especially strong assumptions. The study-level approach has the advantage that each study is given equal weight. This limits the influence of study-level characteristics such as availability or quality of measures, highly similar populations in multiple separate analyses, particular analytical strategies, and unobserved ‘researcher degrees of freedom’ (Simmons, Nelson, & Simonsohn, 2011). That is, if some feature of a dataset or decision taken by a study author acts to increase the chance of finding a grandparent effect, this is not amplified by variation between studies in the number of analyses reported. If one study reports results from five different populations, and systematically tends towards finding an effect, this does not have five times the weight of a study which reports a single analysis (i.e. one

• utcome for one population). Conversely, the analysis-level approach allows both for a greater N,

and for examining the proportion finding an effect according to factors which vary within studies. Defining distinct analyses only according to outcome and/or population helps restrict multiple- counting of results.

2 In fact there is a third, occurring only once: Stuhler’s (2014: 23) fullest model includes a control for father’s

cognitive ability, which reduces the sample size to 2789, from 47797 in the previous model excluding the

• control. The model excluding this control is used. This does not affect whether a grandparent effect is found.

6

2.3 Descriptive statistics

2.3.1 Proportions finding an effect

For both studies and analyses, we tabulate simple proportions finding a grandparent effect (as defined above), showing how this varies according to a range of study/analysis characteristics which might bias towards or against finding an effect (or in some cases, reflect genuine heterogeneity in the association). These characteristics are: outcome, setting, dataset, sample size, number of grandparents for whom information is available, the specification of grandparent resources, the specification of the equivalent parental resources, and how many – and which sorts of – additional parental socioeconomic controls are included (i.e. in addition to those which are the equivalent of the grandparent term being tested). Some of these measures vary between analyses within studies, and are therefore defined as missing at the study level. Other study/analysis characteristics are a priori interesting, but in fact exhibit little variation – for instance, period. The categorisation of these study/analysis characteristics is mostly self-explanatory. We here describe those which may not be. Outcome is categorised as either level (e.g. years of education, whether any tertiary education, choice of high school track) or ability (e.g. reading ability, GPA, cognitive ability). The two types of outcomes may be subject to different influences since the former involves an element of choice. Sample size refers to number of grandchildren. The ‘other’ category within GP data availability refers to patterns of availability of data which are each rare, e.g. only maternal grandmother (Roksa & Potter, 2011), between one and four grandparents (Ferrie, Massey, & Rothbaum, 2016). For operationalisation of grandparent(s), the category ‘joint/sum’ refers either to the sum of the available (grand)parents’ values of a certain resource (e.g. years of education), or to characteristics which pertain to (or are simply measured at the level of) the (grand)parents as a set, for instance neighbourhood poverty (Sharkey & Elwert, 2011), family income, or social capital. Because the mean of the parents’ values is very rarely used, the categories ‘mean’ and ‘joint/sum’ are grouped together for operationalisation of parent(s). The category ‘other’ for these variables refers to approaches such as including multiple grandparent terms operationalised in different ways, such as each grandparent’s years of education and grandparental family income; another example of an ‘other’ approach is entering the number of post-secondary-educated grandfathers into the model (Sheppard & Monden, 2017). Parental socioeconomic characteristics which are included in addition to the parental values of the grandparent terms being tested are captured in three different approaches (shown in the final three panels of table 1). The first is no. of additional parental SES controls (counting categories), which is a count of how many of the following aspects of parental SES are controlled: income, occupational status / social class, education, wealth / homeownership, social capital, and other (e.g. cognitive skills). Second is no. of additional parental SES controls (counting variables), which counts the number of variables controlled which fall into any of the aspects of SES

• listed. For instance, if mother’s income and father’s income are included as controls, this is counted

as 1 for counting categories and as 2 for counting variables. Separate pieces of information used in constructing these controls are also counted as separate even if just one term is included in the

• model. For instance if the mean of parents’ years of education is included, this is also counted as 2

for counting variables, since it taps both mother’s and father’s education. Third, the additional parental SES controls variables are six dummies indicating whether each aspect of SES is included among the additional parental SES controls.

2.3.2 Comparing strength of association

For the most part, differences between the studies make effect sizes incommensurable, and we are reduced to comparing an imprecise binary. Here we present two approaches which aim at making meaningful comparisons of effect size between studies or analyses.

7 2.3.2.1 Effect size relative to parents Since the analyses reviewed here control for parental resources, it is in many cases possible to calculate the ratio of grandparental to parental effect sizes. This provides a scale-free parameter which can be compared across analyses and itself tabulated, as above, according to study- or analysis-level variation in factors such as which controls are included. [Not yet done] 2.3.2.2 Years of education on years of education Whilst analyses vary widely in the specific measures they employ and hence the parameters they estimate, the grandparental years-of-schooling effect is reported in several studies, and moreover is usually within-generation standardised. Across studies and analyses, these are not estimating precisely the same parameter, as other differences remain, but these estimates nevertheless represent an opportunity to compare absolute effect sizes across studies according to different study characteristics. [Not yet done]

2.4 Narrative review

Analyses testing several hypotheses of interest are either low in number, particularly idiosyncratic,

• r both. For these we give a narrative review of the results and in some cases use summary tables. In

the narrative review we discuss: threats to validity and efficiency, including attempts to address the question of whether apparent grandparent effects might be due to measurement error, and whether results vary according to different specifications within analyses; the possible mechanisms underlying a direct effect of grandparents, which, following Knigge (2016), we divide into contact- based and non-contact based; the distribution of grandparent effects – in particular whether there are nonlinearities and interactions involved; and finally whether there is evidence for particular grandparents being especially important in the transmission of socioeconomic status.

• 3. RESULTS

3.1 Descriptive statistics

Table 1 shows the proportions of studies and analyses finding an independent grandparent effect across a range of study and analysis characteristics.3 Few clear patterns emerge. 65% of studies find evidence of a grandparent effect as defined above. This proportion is higher among those looking at ability rather than level outcomes. The proportions appear relatively stable across settings. Whilst certain suggestive patterns are discernible, there is reason to interpret them with caution. For instance, though analyses twice as often find evidence of a grandparent effect in the US compared with Germany, direct comparative evidence challenges this. Neidhoefer & Stockhausen (2017), included here as a single cross-national analysis, find no evidence for a grandparent effect in the US, but do for Germany, while Hertel & Groh-Samberg report that ‘the pattern of three-generation [social class] mobility is similar in both countries’ (2014: 35). Similarly, the high figure for Sweden

3 Supplementary table 1 lists the 40 included studies and whether each shows a significant GP effect at the

study (overall) level.

8 may owe partly to the extremely large sample sizes available in register data, which may make even substantially very small grandparent effects statistically significant and thus increase the likelihood

• f finding a grandparent effect (as reflected in the increasing proportion across the sample size

categories here). Different analyses using the same dataset often vary in whether or not a grandparent effect is apparent. Finding a main effect may then depend on analytical approach, and/or on the outcome or grandparental characteristics of interest, but it appears that grandparent effects are not obviously either present or absent within a given sample or setting. The similar proportions where data on all four versus one set of grandparents are available suggests that assortative mating in the parental generation leads to a correlation between the characteristics

• f an individual’s two sets of grandparents, such that having information on one set only – most
• ften all that is available – may not lead to drastically different conclusions. Numbers in the other

categories are so low that the proportions are not likely to be informative. The proportions finding an effect do not clearly decline with increasing numbers of parental SES controls, as the measurement error critique of apparent grandparent effects would imply. Half of analyses controlling for four or more additional aspects of parental SES report finding a grandparent effect, while the proportion of analyses with no such additional controls is 0.63. Similarly, proportions are strikingly similar among studies and analyses which do and do not control for particular aspects of parental SES. Table 1 Descriptive statistics and proportions of studies/analyses finding a grandparent effect

No.

• No. finding

GP effect Proportion finding GP effect No.

• No. finding

GP effect Proportion finding GP effect Total 40 26 0.65 70 42 0.60 Outcome Education: level 27 15 0.56 46 25 0.54 Education: ability 10 9 0.90 24 17 0.71 Setting US 17 11 0.65 31 19 0.61 Germany 3 1 0.33 10 3 0.30 Sweden 6 5 0.83 13 10 0.77 Other (Europe) 5 2 0.40 5 2 0.40 Other (non-Europe) 5 4 0.80 7 5 0.71 Cross-national 4 3 0.75 4 3 0.75 Studies Analyses

9

Dataset PSID 8 6 0.75 17 10 0.59 WLS 2 0.00 2 0.00 NLSY79 3 2 0.67 4 3 0.75 NLSY97 2 2 1.00 Swedish register 3 2 0.67 5 4 0.80 Malmo 1 1 1.00 3 3 1.00 Uppsala 2 2 1.00 5 3 0.60 SOEP 2 1 0.50 3 1 0.33 SHARE 3 3 1.00 3 3 1.00 Multiple datasets 4 2 0.50 3 2 0.67 Other 12 7 0.58 23 11 0.48 Sample size <1000 3 1 0.33 6 2 0.33 1000-10000 27 17 0.63 50 28 0.56 >10000 10 8 0.80 14 12 0.86 GP data availability All 11 6 0.55 16 9 0.56 One set 19 13 0.68 35 22 0.63 Grandfather(s) 4 1 0.25 11 4 0.36 Other 5 5 1.00 7 6 0.86 Operationalisation of grandparent(s) Highest 6 5 0.83 10 8 0.80 Only available one 7 4 0.57 21 11 0.52 Mean 3 3 1.00 4 4 1.00 Two separately 7 4 0.57 11 5 0.45 More than two separately 3 1 0.33 4 2 0.50 Joint/sum 9 7 0.78 10 10 1.00 Other 5 2 0.40 2 2 1.00 Operationalisation of parent(s) Highest 5 4 0.80 11 7 0.64 One 9 7 0.78 14 11 0.79 Separately 14 7 0.50 27 12 0.44 Joint/sum/mean 7 6 0.86 13 9 0.69 Other 5 2 0.40 5 3 0.60

10

3.2 Strength of association

3.2.1 Effect size relative to parents

[Not yet done]

• No. of additional

parental SES controls (counting categories) 16 10 0.63 35 22 0.63 1 8 5 0.63 12 5 0.42 2 5 4 0.80 6 5 0.83 3 5 4 0.80 7 5 0.71 4+ 4 2 0.50 10 5 0.50

• No. of additional

parental SES controls (counting variables) 16 10 0.63 35 22 0.63 1 3 3 1.00 6 3 0.50 2-3 10 7 0.70 12 8 0.67 4-5 4 3 0.75 6 4 0.67 6-9 5 2 0.40 9 3 0.33 10+ 1 1 1.00 2 2 1.00 Additional parental SES controls (row below = control absent) Income 15 11 0.73 24 15 0.63 22 14 0.64 44 26 0.59 Occupation / social class 9 6 0.67 16 10 0.63 28 19 0.68 52 31 0.60 Education 7 5 0.71 12 7 0.58 30 20 0.67 56 34 0.61 Wealth / homeowner 7 6 0.86 12 8 0.67 30 19 0.63 56 33 0.59 Social capital 1 0.00 1 0.00 36 25 0.69 67 41 0.61 Other (e.g. cognitive skills) 9 6 0.67 17 10 0.59 28 19 0.68 51 31 0.61

11

3.2.2 Years of education on years of education

[Not yet done]

3.3 Narrative review

In this section we address several recurrent debates in the grandparent effects literature.

3.3.1 Threats to validity and efficiency

Measurement error, omitted variables, and collinearity each merit discussion with regard to assessing the robustness of grandparent effects. First, in certain circumstances, random measurement error is likely to be greater for grandparental characteristics than for parental characteristics, which would act to suppress any true G1-G3

• association. This applies when, as is usually the case, both parental and grandparental characteristics

are reported by G2 – that is, by grandparents’ adult offspring, and in some cases by their sons- or daughters-in-law. Respondents are likely to be able to make a more accurate report of their own and their partner’s socioeconomic resources than of their own parents’ and in-laws’. In addition, information is usually not available on all four grandparents, which may further introduce random measurement error. This depends upon the missingness of grandparent measures being random, which holds in the many cases where either maternal or paternal grandparents are

• bserved, depending on the gender of the G2 survey respondent.

Second, conversely, where information on all four grandparents is used, including terms for each of them may introduce collinearity if the resources of an individual’s grandparents are highly correlated, as may be the case under a high degree of assortative mating. This would decrease efficiency and make a true grandparent effect less likely to be found. Third, a potential source of bias lies in the measurement of G2 resources, both through classical measurement error and through omitted variables. In the standard model regressing G3 education

• n G2 and G1 education, measurement error in the former will attenuate its association with the
• utcome and upwardly bias the G1 coefficient. As mentioned, this issue is more likely to take the
• pposite form, with random measurement error higher among G1 than G2. However, Ferrie et al.

(2016) show that their finding of a significant grandparent effect is consistent with an AR(1) process where the level of measurement error in education is the level which they in fact observe from their multiple data sources: ‘The grandparent effect we observe … either does not exist or is too small for us to distinguish from measurement error’ (2016: 19). Of even greater concern is the omission of G2 characteristics different from the G1 characteristic of interest, but which also mediate the transmission of advantage between generations. For example, if grandparental education exerts indirect effects on grandchild education not only through parental education but also through parental income, failing to control for parental income will upwardly bias the independent G1-G3 association and give the appearance of a direct grandparent effect. Clearly, socioeconomic status in the broad sense extends beyond these two dimensions, and it is not clear that any number of controls in the G2 generation would be sufficient to ensure that an apparent grandparent effect does not in fact reflect an unobserved pathway involving some attribute of the intervening generation. Note also that Clark’s hypothesis is a strong version of this argument: socioeconomic status is not just multidimensional, but not directly observable within one generation.

12 While this argument is most often put in terms of unmeasured parental characteristics, it extends to the characteristics of other family within the parents’ generation. Aunts, uncles, and others embody a pathway for the transmission of advantage from grandparent to grandchild, but their resources are rarely measured in the datasets informing this literature. Loury (2006) reports that aunts, uncles and grandparents each independently affect grandchildren’s educational outcomes, though Jaeger (2012) finds an effect only for aunts’ education among extended family in G2, and no grandparent

• effects. The most striking evidence on this issue is afforded by the unique comprehensiveness of

Swedish register data: Adermon et al. (2016) show that the coefficient for grandparents’ human capital on grandchildren’s years of schooling falls from 0.044 to 0.012 when a greater range of individuals from G2 are taken into account (aunts and uncles, their spouses, the siblings of their spouses, parents’ cousins and their spouses). The attenuation is similar whether schooling, schooling and income, or schooling, income and occupation are used to measure relatives’ human capital. We conclude this section on the robustness of grandparent effects by examining stability and change in results according to within-sample variation in the operationalisation of grandparent resources (table 2). Evidence for grandparent effects will be more compelling if it is not dependent upon a priori minor changes in model specification. This is indeed usually the case, insofar as such checks are reported. It may of course be that authors are reluctant to report instability of findings to specification changes, should this occur. Table 2 Robustness of findings to different operationalisations of grandparental resources

Study Different approaches to including GPs Different results from different models? Sheppard & Monden 2017 Evaluate model fit & GP effects using 14 different operationalisations of GPs (with data on all 4); do this for full sample and 15 countries separately ‘Model fits do not differ much’; ‘The best fitting model differs across countries with no clear patterns as to the best fit, nor with regard to the main findings from the coefficients and p- values’ Neidhoefer & Stockhausen 2017

• - Most educated (of 4) GPs
• - Each GP separately (stratified by sex of

grandchild) Using z-scores: no difference (no GP effect); using unstandardised years of ed.: sig. effect of most-educated GP, but each GP individually NS Hancock et al. 2016

• - All 4 GPs’ (&Ps’) levels of ed.

separately

• - For each set of GPs (&Ps), whether GF,

GM or both have tertiary ed. Broadly similar, results not directly comparable Ferrie et al. 2016

• - Most educated observed GP (&P)
• - Mean education of GPs (&Ps)
• - Most educated observed GP (&P),

with terms for other GPs (&Ps) also included

• - Most educated GP (&P), with mean

education of GPs (&Ps) also included Effect size smallest using mean only & largest using most educated only; ‘Our choice of measure does not change the results we present' (p.14 n.12); findings also ‘robust to inclusion of multiple grandparents and both parents, as well as for matriarchal and patriarchal lines followed and analyzed separately' (p.22) Song 2016

• - Most educated GP (usually either

maternal or paternal GPs observed)

• - Average education of GPs

'I also experimented with using the average education rather than the highest education of parents and

• grandparents. The results are consistent

with those presented in the article' (p.1913) Stuhler 2014

• - Paternal GF only

Paternal GF shows a direct effect, but

13

• - All 4 GPs separately (halves sample

size) all 4 GPs NS when included together; the 4 coefficients sum to well above the P-GF coefficient used alone, but none reaches significance Wightman & Danziger 2014

• - One set of GPs
• - Both sets of GPs

‘conditional on parents’ education, all measures of grandparents’ education are insignificant, this is true whether using one or both sets of grandparents’ information’ (p.67) Zeng & Xie 2014

• - Sum of percentile scores
• - Mean percentile score
• - Maximum percentile score
• - Sum of years of schooling

Highly similar results across all approaches Grant 2005

• - M-GPs’ wealth
• - P-GPs’ wealth

and

• - M-GPs’ income
• - P-GPs’ income

‘no consistently significant differences between the effect of maternal and paternal grandparent wealth’ (p.41 n.1); ‘[no] consistent differences between the effects of maternal and paternal grandparent income’ (p.42 n.2)

3.3.2 Possible mechanisms underlying a direct effect of grandparents

In this section we evaluate the evidence for potential mechanisms for grandparent effects. 3.3.2.1 Contact-based Many of the mechanisms through which a direct grandparent effect is hypothesised to occur depend upon interaction between grandparent and grandchild. Through face-to-face interaction, the established mechanisms through which parents influence children – modelling behaviours, teaching values and norms, reading with the child and so on – are analogously open to grandparents. This implies that, in the presence of true grandparent effects, G1-G3 associations will be stronger in cases where grandparents are able or likely to have been in contact with the grandchild, and weaker when this does not hold – for instance where the grandparent(s) died early in the grandchild’s life or even before their birth. The finding of stronger grandparental effects where the grandparent is closer to the family could, however, also result from lower measurement error in parental reporting of grandparent resources when the grandparent is present, or from unobserved characteristics of grandparents correlating with their availability, so this would not unambiguously confirm the presence of grandparent effects. Several studies test such grandparental availability interaction terms. Table 3 summarises these

• findings. In general, coefficients for grandparental resources do not appear to differ depending on

the likelihood of contact between grandparent and grandchild, though there are exceptions, most notably Zeng & Xie’s (2014) finding that the education of coresident grandparents in rural China is almost as strongly associated with the likelihood of school dropout as mother’s and father’s education. Table 3 Studies testing interactions with grandparental likelihood of contact

Study GP resources interacted with… Result Sheppard & Monden 2017 Lifespan overlap Interaction not significant Neidhoefer & Stockhausen 2017 GP dead when grandchild aged 1 Significant negative interaction only for maternal grandfather, and only in German

14

data Daw & Gaddis 2016 GP still alive when grandchild aged 6 Interaction not significant Braun & Stuhler 2016 Grandfather dead when grandchild born (with WWII deaths as a source of exogenous variation) Interaction not significant Bol & Kalmijn 2016 Tie strength, residential proximity Interactions not significant Kroeger & Thompson 2016 Mother lived with one of the child’s GPs for three months or more after child born Interaction not significant Zeng & Xie 2014 Number and education of coresident, noncoresident, and deceased GPs Effect of education far stronger for coresident than noncoresident or deceased GPs Ferguson & Ready 2011 ‘Co-resident versus caregiver status’ Interaction not significant Loury 2010 GP dead when grandchild born or geographically remote during childhood Effect of grandmother’s education only if not dead or remote (interaction not significant for grandfather)

Results are similar for studies which test for a main, rather than moderating, effect of grandparental

• availability. Wightman & Danziger (2014: 67) find that the child having at least one living

grandparent shows no effect, while Ferguson & Ready report similar null results for variables indicating the number of living grandparents and whether they lived in proximity (2011: 219). Celhay & Gallegos’s main effect for whether any grandparent is still alive is also non-significant (2015: 443). 3.3.2.2 Non-contact-based Direct influence of grandparents on grandchildren is also possible through means which do not require contact. Indeed the most intuitive channel for grandparent effects is direct financial

• transfers. Hällsten & Pfeffer (2017: 331-334) provide a discussion of further mechanisms through

which family wealth may influence educational outcomes. Dead grandparents may also have social capital in the form of contacts which aid their grandchildren, or they may serve as reference points for their descendants’ decision-making in the status attainment process (Hertel & Groh-Samberg, 2014). It is also possible for grandparents to pass on advantageous genetic material which is not expressed in the parental generation. While Braun & Stuhler claim the lack of an interaction between grandfather schooling and early death as evidence for indirect mechanisms of status transmission across three generations (2016: 27), direct evidence on these mechanisms is extremely sparse.

3.3.3 The distribution of grandparent effects

Grandparent effects may be stronger among certain subgroups. Indeed, the loglinear models of Pfeffer (2014, education, USA), Chan & Boliver (2013, social class, UK), and Hertel & Groh-Samberg (2014, social class, Germany & USA) all suggest stronger multigenerational persistence at the ends of the status distribution, as do the transition matrices shown by Lindahl et al. (2015) for both earnings and education. Authors taking a regression approach have tended to frame the nonlinearity issue as one of compensation (stronger grandparent effects where parental resources are low) versus augmentation (stronger grandparent effects where parental resources are high). It is noteworthy that this implicitly

15 rejects the possibility of stronger effects at both ends of the distribution of parental resources than closer to the mean. This framing perhaps results from the difficulty of specifying terms in a regression framework to accommodate testing this possibility. Rather, where nonlinearities are tested, a simple interaction between parent and grandparent resources is included. These tests are reported in table 4. Results are mixed but more often than not, the hypothesis of compensation is

• supported. This is in line with Bengtson’s (2001) argument that grandparents represent an important

source of support which may often not be recognised and harnessed except in conditions of

• difficulty. On the grandparent (supply) side, this accords with ‘the two main cultural norms of

grandparenting [identified in the UK] … ‘being there’ and ‘not interfering’’ (J. Mason, May, & Clarke, 2007: 687). If, as the loglinear evidence mentioned above suggests, grandparent effects are especially strong at both ends of the parental status distribution, the interaction terms tested in this framework may be biased towards zero. Just as there are good reasons to expect grandparents to step up in times of need, there are mechanisms for the direct grandparental transmission of advantage which are especially concentrated at the top of the status distribution, such as college legacy admissions and trust funds. Table 4 Studies testing interactions of grandparent and parent resources

Study Augmentation, compensation

• r linearity?

b1(GP) + b2(P) + b3(GPxP) Sheppard & Monden 2017 Linearity Deindl & Tieben 2016 Compensation 0.49 + 1.20 - 0.27 (high ed.); 0.65 + 0.64 – 0.55 (financial resources) Daw & Gaddis 2016 Augmentation (but not robust to inclusion of cousin fixed effects) e.g. -0.158 + 0.243 + 0.0158 (ed.) Braun & Stuhler 2016 Compensation GP coefficient: 0.334 (father’s ed. = low) v. 0.017 (father’s ed. ≠ low); 0.453 (G2 respondent income = low) v. 0.071 (G2 respondent income’s ≠ low) Chiang & Park 2015 Augmentation e.g. -0.174 + 0.100 + 0.012 (ed.) Wolbers & Ultee 2013 Compensation 0.04 + 0.20 – 0.06 (ed.) Havari & Savegnago 2012 Compensation e.g. 0.516 + 0.906 – 0.357 (ed., proxied by number of books for GPs) Jaeger 2012 Compensation e.g. 0.140 + 0.257 – 0.010 (ed.)

Heterogeneity is also possible according to family structure in the parental generation. Song (2016) focuses on this question, noting that stronger effects are plausible both in single-parent families (analogous to the compensation hypothesis, the necessity of their assistance may ‘activate’ the grandparent effect) and in two-parent families, where more grandparents are likely to be in contact with the grandchild. Song finds a significant G1-G3 association only among married, two-parent families, and not where a child’s parents are divorced or unmarried.

3.3.4 Are certain grandparents more important?

In this section we review evidence on whether certain grandparents or sets of grandparents are more often associated with children’s educational outcomes. In supplementary table 2, we report two types of evidence. Firstly, cases in which terms for grandparents are included separately in a model, such as grandfather’s and grandmother’s education (single sample). Secondly, cases in which the sample is split according to which grandparent(s) are available, and analogous models are fitted

16 to, for instance, the halves of the sample with information on maternal and paternal grandparents. Where samples are split like this, separate models are indicated with a), b), c) etc. Again, there is no clear pattern to the results. It occurs slightly more often that a grandfather term is significant and a grandmother term is not, than vice versa. Even a striking pattern could only be suggestive, since the difference between a significant and a non-significant term may not itself be significant. Sheppard & Monden (2017) note, having fitted 14 different specifications of all four grandparents to data from 15 countries separately, that ‘the only consistent finding across models is that, where there is an association, it tends to be for highly educated grandfathers and there is no evidence for an influence of grandmothers in any country.’

4 CONCLUSIONS & FUTURE DIRECTIONS

The extant evidence gives a remarkably incoherent picture as to whether grandparental resources are associated with the educational outcomes of their grandchildren independently of the characteristics of the parental generation. Around two-thirds of the 70 analyses to model this relationship report a significant association, and it remains unclear what, if anything, might systematically distinguish analyses finding an effect from those which do not. The measurement error critique of grandparent effects is persuasive, but its implication that a grandparent effect should be less likely to be found with more controls for parental characteristics is not borne out here, albeit in a clearly imperfect test. Unobserved heterogeneity is the likely source of this confusion. It is surprising then, that variation in G1-G3 associations by measures of the likelihood of contact is not more evident. If grandparents directly affect their grandchildren, the association is expected to be stronger if the grandparents, for instance, live nearby and overlap substantially in lifespan. This approach is limited, however, since the relation between modes of grandparental influence and measures of availability is not necessarily clear. For instance, geographically distant grandparents may provide fewer economic resources because they have less contact and so take less of an interest in their descendant, or they may provide more to compensate for the relative lack of contact. Similarly, living grandparents may provide more economic resources as they take an active interest in the child’s development, or deceased grandparents may do so because they provide an inheritance and have zero consumption

• themselves. Measures of grandparental availability would usefully be combined with other

information on grandparental characteristics, to enable researchers to determine whether particular mechanisms rely upon some aspect of contact or availability. No research has yet explored the dynamic aspects of multigenerational effects: variation in availability – especially conceived of as health and wellbeing – exists within individual grandparents as well as between them. This raises questions of whether and at what point aging grandparents begin to compete with the youngest generation for resources rather than providing them, and so whether their net effect varies according to the time at which it is measured and depends upon health and longevity. Competing hypotheses are plausible: high SES grandparents have more resources to pass on, yet may live longer and in doing so demand resources from the parental and even child generation. This may have implications for the educational success of grandchildren.

17 Further work is needed to disentangle the nature of grandparent effects. To advance our understanding it will need to attend to measurement error and omitted variables with regard to the parental generation – including aunts and uncles if possible (Prix & Pfeffer, 2017). It will also need to pay close attention to sources of heterogeneity, and in particular the conditions under which, and ways in which, grandparents are drawn out of the ambivalence of their position with respect to their grandchildren’s development (J. Mason et al., 2007) and mobilised as cultivators of their human capital. [To be extended after the Nuffield workshop]

Supplementary material

Supplementary table 1 Included studies (chronological order), whether coded as finding a GP effect (N=40)

Study GP effect (1/0) Sheppard & Monden (2017) 1 Neidhoefer & Stockhausen (2017) Hällsten & Pfeffer (2017) 1 Song (2016) 1 Bol & Kalmijn (2016) Ferrie et al. (2016) 1 Deindl & Tieben (2016) 1 Daw & Gaddis (2016) 1 Ziefle (2016) 1 Adermon et al. (2016) 1 Hancock et al. (2016) 1 Braun & Stuhler (2016) Kroeger & Thompson (2016) 1 Celhay & Gallegos (2015) 1 Chiang & Park (2015) 1 Mollegaard & Jaeger (2015) 1 Lindahl et al. (2015) 1 Stuhler (2014) Wightman & Danziger (2014) Zeng & Xie (2014) 1 Modin et al. (2013) 1 Wolbers & Ultee (2013) Jaeger (2012) Havari & Savegnago (2012) 1 Roksa & Potter (2011) 1 Sharkey & Elwert (2011) 1 Ferguson & Ready (2011) 1 Loury (2010)

18

Modin & Fritzell (2009) 1 Yeung & Conley (2008) Fuchs & Sixt (2007) Mason (2007) 1 Grant (2005) 1 Phillips et al. (1998) 1 Warren & Hauser (1995 / 1997) Hill & O'Neill (1994) 1 Iglesias & Riboud (1988) 1 Heckman & Hotz (1986) Behrman & Taubman (1985) Ridge (1974)

Supplementary table 2 Within-analysis variation in which (sets of) grandparents show an independent effect (see section 3.3.4)

Study (analysis) Separate GPs included GP effect No GP effect Kroeger & Thompson 2016 (sons) a) M-GM b) M-GF a) M-GM b) M-GF Song 2016,4 Online Resource 1 (African Americans) a) Most-educated P-GP b) Most-educated M-GP c) GF d) GM d) GM a) Most-educated P- GP b) Most-educated M-GP c) GF Song 2016, Online Resource 1 (Whites) a) Most-educated P-GP b) Most-educated M-GP c) GF d) GM a) Most- educated P-GP b) Most- educated M-GP c) GF d) GM Hancock et al. 2016 (numeracy outcome) All 4 P-GF M-GM, M-GF, P-GM Hancock et al. 2016 (reading outcome) All 4 M-GM, P-GM M-GF, P-GF Bol & Kalmijn 2016 a) P-GPs’ ed. b) M-GPs’ ed. c) P-GF’s occupational status d) M-GF’s occupational status e) P-GPs’ cultural resources f) M-GPs’ cultural resources a) P-GPs’ ed. b) M-GPs’ ed. c) P-GF’s

• ccupational status

d) M-GF’s

• ccupational status

e) P-GPs’ cultural resources f) M-GPs’ cultural resources Celhay & Gallegos 2015 (sons) a) P-GF & P-GM b) M-GF & M-GM a) P-GF & P-GM b) M-GF b) M-GF Celhay & Gallegos 2015 a) P-GF & P-GM a) P-GM a) P-GF

4 Song states that ‘overall, some (albeit not statistically significant) difference is evident between the average

effects of paternal and maternal grandparents or between grandfathers and grandmothers’ (2016: 1926).

19

(daughters) b) M-GF & M-GM b) M-GM b) M-GF Chiang & Park 2015 (academic HS attendance) a) M-GF & M-GM b) P-GF & P-GM a) M-GF & M-GM b) P-GF & P-GM Chiang & Park 2015 (university attendance) a) M-GF & M-GM b) P-GF & P-GM b) P-GF (negative) a) M-GF & M-GM b) P-GM Stuhler 2014 All 4 All 4 Modin et al. 2013 (maths

• utcome, boys)

a) P-GF b) P-GM c) M-GF d) M-GM a) P-GF c) M-GF b) P-GM d) M-GM Modin et al. 2013 (maths

• utcome, girls)

a) P-GF b) P-GM c) M-GF d) M-GM a) P-GF b) P-GM c) M-GF d) M-GM Modin et al. 2013 (Swedish outcome, boys) a) P-GF b) P-GM c) M-GF d) M-GM a) P-GF b) P-GM c) M-GF d) M-GM Modin et al. 2013 (Swedish outcome, girls) a) P-GF b) P-GM c) M-GF d) M-GM a) P-GF c) M-GF d) M-GM b) P-GM Jaeger 2012 GM & GF GM & GF Loury 2010 M-GM & M-GF M-GM & M-GF Modin & Fritzell 2009 a) P-GF & P-GM b) M-GF & M-GM a) P-GF b) M-GF a) P-GM b) M-GM Yeung & Conley 2008 (Letter-Word outcome, 3-5 year-olds) GM & GF GM & GF Yeung & Conley 2008 (Applied Problems

• utcome, 3-5 year-olds)

GM & GF GM & GF Yeung & Conley 2008 (reading outcome, 6-12 year-olds) GM & GF GF GM Yeung & Conley 2008 (maths outcome, 6-12 year-olds) GM & GF GM & GF Warren & Hauser 1995 (regressions) a) P-GF’s ed., P-GM’s ed., P-GF’s

• ccupational status, P-GPs’ earnings,

M-GF’s ed., M-GF’s occupational status b) P-GF’s ed., P-GF’s occupational status, M-GF’s ed., M-GM’s ed., M- GF’s occupational status, M-GPs’ earnings a) P-GPs’ earnings (negative) b) P-GF’s

• ccupational

status a) P-GF’s ed., P-GM’s ed., P-GF’s

• ccupational status,

M-GF’s ed., M-GF’s

• ccupational status

b) P-GF’s ed., M-GF’s ed., M-GM’s ed., M- GF’s occupational status, M-GPs’ earnings Hill & O’Neill 1994 M-GM & M-GF M-GM M-GF Iglesias & Riboud 1988 P-GF & M-GF P-GF & M-GF Behrman & Taubman 1985 GFs (mean) & GMs (mean) GFs & GMs Ridge 1973 P-GF & M-GF P-GF & M-GF Note: a), b) etc. indicate separate samples (usually divided according to data availability); otherwise the terms noted are tested in the same sample.

20

References

Adermon, A., Lindahl, M., & Palme, M. (2016). Dynastic human capital, inequality and intergenerational mobility (Working Paper 2016:19 No. 16). Uppsala, Sweden: Institute for Evaluation of Labour Market and Education Policy (IFAU). Becker, G. S., & Tomes, N. (1986). Human Capital and the Rise and Fall of Families. Journal of Labor Economics, 4(3), S1–S39. Behrman, J., & Taubman, P. (1985). Intergenerational Earnings Mobility in the United States: Some Estimates and a Test of Becker’s Intergenerational Endowments Model. The Review of Economics and Statistics, 67(1), 144–151. https://doi.org/10.2307/1928446 Bengtson, V. L. (2001). Beyond the Nuclear Family: The Increasing Importance of Multigenerational

• Bonds. Journal of Marriage and Family, 63(1), 1–16. https://doi.org/10.1111/j.1741-

3737.2001.00001.x Bol, T., & Kalmijn, M. (2016). Grandparents’ resources and grandchildren’s schooling: Does grandparental involvement moderate the grandparent effect? Social Science Research, 55, 155–170. https://doi.org/10.1016/j.ssresearch.2015.09.011 Braun, S. T., & Stuhler, J. (2016). The Transmission of Inequality Across Multiple Generations: Testing Recent Theories with Evidence from Germany. The Economic Journal, n/a-n/a. https://doi.org/10.1111/ecoj.12453 Breen, R. (2015). The Son Also Rises: Surnames and the History of Social Mobility. By Gregory Clark. Princeton: Princeton University Press. 2014. Pp. xii, 364. $29.95, hardcover. The Journal of Economic History, 75(1), 301–302. https://doi.org/10.1017/S0022050715000303 Celhay, P., & Gallegos, S. (2015). Persistence in the Transmission of Education: Evidence across Three Generations for Chile. Journal of Human Development and Capabilities, 16(3), 420–451. https://doi.org/10.1080/19452829.2015.1048789 Chan, T. W., & Boliver, V. (2013). The Grandparents Effect in Social Mobility: Evidence from British Birth Cohort Studies. American Sociological Review, 78(4), 662–678. https://doi.org/10.1177/0003122413489130 Chiang, Y. L., & Park, H. (2015). Do grandparents matter? A multigenerational perspective on educational attainment in Taiwan. Social Science Research, 51, 163–173. https://doi.org/10.1016/j.ssresearch.2014.09.013 Clark, G. (2014). The son also rises: surnames and the history of social mobility. Princeton, New Jersey: Princeton University Press. Daw, J., & Gaddis, S. M. (2016). Education in 3G: Heterogeneity of Intergenerational Educational Mobility in the U. S. (SSRN Scholarly Paper No. ID 2710883). Rochester, NY: Social Science Research

• Network. Retrieved from https://papers.ssrn.com/abstract=2710883

Deindl, C., & Tieben, N. (2016). Resources of Grandparents: Educational Outcomes Across Three Generations in Europe and Israel. Journal of Marriage and Family. https://doi.org/10.1111/jomf.12382

21 Dribe, M., & Helgertz, J. (2016). The Lasting Impact of Grandfathers: Class, Occupational Status, and Earnings over Three Generations in Sweden 1815–2011. The Journal of Economic History, 76(4), 969–

• 1000. https://doi.org/10.1017/S0022050716000991

Ferguson, J. L., & Ready, D. D. (2011). Expanding notions of social reproduction: Grandparents’ educational attainment and grandchildren’s cognitive skills. Early Childhood Research Quarterly, 26(2), 216–226. https://doi.org/10.1016/j.ecresq.2010.10.001 Ferrie, J., Massey, C., & Rothbaum, J. (2016). Do Grandparents and Great-Grandparents Matter? Multigenerational Mobility in the US, 1910-2013. National Bureau of Economic Research Working Paper Series, No. 22635. https://doi.org/10.3386/w22635 Fuchs, M., & Sixt, M. (2007). Zur Nachhaltigkeit von Bildungsaufstiegen / On the sustainability of educational improvement. Social heredity transmission of educational success over multiple

• generations. KZfSS Kölner Zeitschrift für Soziologie und Sozialpsychologie, 59(1), 1–29.

https://doi.org/10.1007/s11577-007-0001-6 Grant, E. (2005). Cumulative advantage: A new look at the effect of family social class on children’s educational achievement and the black -white achievement gap (Ph.D.). Stanford University, United States -- California. Retrieved from https://search.proquest.com/docview/305436426/abstract/E08D1FBE267C4B3CPQ/1 Hällsten, M., & Pfeffer, F. T. (2017). Grand Advantage: Family Wealth and Grandchildren’s Educational Achievement in Sweden. American Sociological Review, 82(2), 328–360. https://doi.org/10.1177/0003122417695791 Hancock, K., Mitrou, F., Povey, J., Campbell, A., & Zubrick, S. (2016). Three-Generation Education Patterns among Grandparents, Parents and Grandchildren: Evidence of Grandparent Effects from Australia (Working Paper No. 2016-08). University of Queensland. Havari, E., & Savegnago, M. (2012). Intergenerational transmission of human capital. Evidence from a three generation study. (Working Paper). Retrieved from http://www.webmeets.com/files/papers/res/2013/998/Havari_Savegnago_2012.pdf Heckman, J. J., & Hotz, V. J. (1986). An Investigation of the Labor Market Earnings of Panamanian Males Evaluating the Sources of Inequality. The Journal of Human Resources, 21(4), 507–542. https://doi.org/10.2307/145765 Hertel, F. R., & Groh-Samberg, O. (2014). Class mobility across three generations in the U.S. and

• Germany. Research in Social Stratification and Mobility, 35, 35–52.

https://doi.org/10.1016/j.rssm.2013.09.007 Hill, M. A., & O’Neill, J. (1994). Family Endowments and the Achievement of Young Children with Special Reference to the Underclass. Journal of Human Resources, 29(4), 1064–1100. Iglesias, F. H., & Riboud, M. (1988). Intergenerational Effects on Fertility Behavior and Earnings Mobility in Spain. The Review of Economics and Statistics, 70(2), 253–258. https://doi.org/10.2307/1928309 Jæger, M. M. (2012). The Extended Family and Children’s Educational Success. American Sociological Review, 77(6), 903–922. https://doi.org/10.1177/0003122412464040

22 Knigge, A. (2016). Beyond the Parental Generation: The Influence of Grandfathers and Great- grandfathers on Status Attainment. Demography, 53(4), 1219–1244. https://doi.org/10.1007/s13524-016-0486-6 Kolk, M. (2014). Multigenerational transmission of family size in contemporary Sweden. Population Studies, 68(1), 111–129. https://doi.org/10.1080/00324728.2013.819112 Kroeger, S., & Thompson, O. (2016). Educational mobility across three generations of American

• women. Economics of Education Review, 53, 72–86.

https://doi.org/10.1016/j.econedurev.2016.05.003 Lindahl, M., Palme, M., Massih, S. S., & Sjögren, A. (2015). Long-Term Intergenerational Persistence

• f Human Capital: An Empirical Analysis of Four Generations. Journal of Human Resources, 50(1), 1–
• 33. https://doi.org/10.3368/jhr.50.1.1

Lindahl, M., Palme, M., Sandgren-Massih, S., & Sjögren, A. (2014). A Test of the Becker-Tomes Model

• f Human Capital Transmission Using Microdata on Four Generations. Journal of Human Capital,

8(1), 80–96. https://doi.org/10.1086/674104 Loury, L. D. (2006). All in the Extended Family: Effects of Grandparents, Aunts, and Uncles on Educational Attainment. The American Economic Review, 96(2), 275–278. Loury, L. D. (2010). College Schooling for Grandchildren and Contact with Grandparents (Working Paper). Department of Economics, Tufts University. Retrieved from https://ideas.repec.org/p/tuf/tuftec/0757.html Mare, R. D. (2011). A multigenerational view of inequality. Demography, 48(1), 1–23. https://doi.org/10.1007/s13524-011-0014-7 Mason, J., May, V., & Clarke, L. (2007). Ambivalence and the paradoxes of grandparenting. The Sociological Review, 55(4), 687–706. https://doi.org/10.1111/j.1467-954X.2007.00748.x Mason, P. L. (2007). Intergenerational Mobility and Interracial Inequality: The Return to Family

• Values. Industrial Relations: A Journal of Economy and Society, 46(1), 51–80.

https://doi.org/10.1111/j.1468-232X.2007.00457.x Modin, B., Erikson, R., & V\a agerö, D. (2013). Intergenerational continuity in school performance: Do grandparents matter? European Sociological Review, 29(4), 858–870. https://doi.org/10.1093/esr/jcs064 Modin, B., & Fritzell, J. (2009). The long arm of the family: are parental and grandparental earnings related to young men’s body mass index and cognitive ability? International Journal of Epidemiology, 38(3), 733–744. https://doi.org/10.1093/ije/dyp001 Møllegaard, S., & Jæger, M. M. (2015). The effect of grandparents’ economic, cultural, and social capital on grandchildren’s educational success. Research in Social Stratification and Mobility, 42, 11–

• 19. https://doi.org/10.1016/j.rssm.2015.06.004

Mukherjee, R. (1954). Social Mobility Between Three Generations. In D. V. Glass (Ed.), Social Mobility in Britain (pp. 266–290). London: Routledge & Kegan Paul Limited. Neidhöfer, G., & Stockhausen, M. (2017). Dynastic Inequality Compared: Multigenerational Mobility in the US, the UK, and Germany (Working Paper). Freie Universität Berlin.

23 Pfeffer, F. T. (2014). Multigenerational approaches to social mobility. A multifaceted research

• agenda. Research in Social Stratification and Mobility, 35, 1–12.

https://doi.org/10.1016/j.rssm.2014.01.001 Pfeffer, F. T., & Killewald, A. (2016). Generations of Advantage. Multigenerational Correlations in Wealth from a Life-Course Perspective (Working Paper). Retrieved from http://fabianpfeffer.com/wp-content/uploads/PfefferKillewald2016.pdf Phillips, M., Brooks-Gunn, J., Duncan, G. J., Klebanov, P., & Crane, J. (1998). Family background, parenting practices, and the black-white test score gap. In C. Jencks & M. Phillips (Eds.), The Black- White Test Score Gap (pp. 102–145). Washington, DC: Brookings Institution Press. Piraino, P., Muller, S., Cilliers, J., & Fourie, J. (2014). The transmission of longevity across generations: The case of the settler Cape Colony. Research in Social Stratification and Mobility, 35, 105–119. https://doi.org/10.1016/j.rssm.2013.08.005 Prix, I., & Pfeffer, F. T. (2017). Does Donald need Uncle Scrooge? Extended family wealth and children’s educational attainment in the United States. In J. Erola & E. Kilpi-Jakonen (Eds.), Social Inequality Across the Generations: The Role of Compensation and Multiplication in Resource Accumulation (pp. 112–135). Cheltenham: Edward Elgar Publishing Limited. Ridge, J. M. (1974). Three Generations. In J. M. Ridge (Ed.), Mobility in Britain Reconsidered (pp. 47– 71). Oxford: Clarendon Press. Roksa, J., & Potter, D. (2011). Parenting and Academic Achievement: Intergenerational Transmission

• f Educational Advantage. Sociology of Education, 84(4), 299–321.

https://doi.org/10.1177/0038040711417013 Sauder, U. (2006). Education transmission across three generations - new evidence from NCDS data (Working Paper). University of Warwick. Sharkey, P., & Elwert, F. (2011). The Legacy of Disadvantage: Multigenerational Neighborhood Effects on Cognitive Ability. American Journal of Sociology, 116(6), 1934–81. https://doi.org/10.1086/660009 Sheppard, P., & Monden, C. (2017). The additive advantage of having two educated grandfathers for child education: evidence from a cross-national sample in Europe. Simmons, J. P., Nelson, L. D., & Simonsohn, U. (2011). False-Positive Psychology: Undisclosed Flexibility in Data Collection and Analysis Allows Presenting Anything as Significant. Psychological Science, 22(11), 1359–1366. https://doi.org/10.1177/0956797611417632 Song, X. (2016). Diverging Mobility Trajectories: Grandparent Effects on Educational Attainment in One- and Two-Parent Families in the United States. Demography, 53(6), 1905–1932. https://doi.org/10.1007/s13524-016-0515-5 Song, X., & Campbell, C. D. (2017). Genealogical Microdata and Their Significance for Social Science. Annual Review of Sociology, 43(1), null. https://doi.org/10.1146/annurev-soc-073014-112157 Stuhler, J. (2014). Mobility Across Multiple Generations: The Iterated Regression Fallacy (Working Paper). Warren, J. R., & Hauser, R. M. (1995). Social Stratification Across Three Generations: New Evidence From the Wisconsin Longitudinal Study (Working Paper No. 95-05). Center for Demography and

24 Ecology, University of Wisconsin-Madison. Retrieved from https://www.ssc.wisc.edu/cde/cdewp/95- 05.pdf Warren, J. R., & Hauser, R. M. (1997). Social Stratification across Three Generations: New Evidence from the Wisconsin Longitudinal Study. American Sociological Review, 62(4), 561–572. https://doi.org/10.2307/2657426 Wightman, P., & Danziger, S. (2014). Multi-generational income disadvantage and the educational attainment of young adults. Research in Social Stratification and Mobility, 35, 53–69. https://doi.org/10.1016/j.rssm.2013.09.004 Wolbers, M. H. J., & Ultee, W. C. (2013). Three generations and two parents questions about the level of education of men and women in the Netherlands. Mens & Maatschappij, 88(3), 276–299. Yeung, W. J., & Conley, D. (2008). Black-white achievement gap and family wealth. Child Development, 79(2), 303–324. https://doi.org/10.1111/j.1467-8624.2007.01127.x Zeng, Z., & Xie, Y. (2014). The Effects of Grandparents on Children’s Schooling: Evidence From Rural

• China. Demography, 51(2), 599–617. https://doi.org/10.1007/s13524-013-0275-4

Ziefle, A. (2016). Persistent Educational Advantage Across Three Generations: Empirical Evidence for

• Germany. Sociological Science, 3, 1077–1102. https://doi.org/10.15195/v3.a47