The intergenerational transmission of reproductive behaviour: comparative perspectives

The six papers included in this special issue of The History of the Family have been selected from the proceedings of an International Seminar on The Intergenerational Transmission of Reproductive Behavior: Comparative Perspectives in June 2011 organized by the Scientific Research Community Historical Demography in Leuven, Belgium1. These papers help to fill gaps in our understanding of the nature of intergenerational continuities in reproductive processes and to assess the generalisability of earlier studies by analysing fertility (actual and preferred, and quantum and tempo) and infant mortality. A major focus of the papers is on the key period for investigating this topic, the ‘demographic transition’ – the transformation from a relatively static and moderately high fertility and mortality regime to a more modern low fertility and mortality regime occurring mainly in the nineteenth century.

The existence of intergenerational continuities, which impacts on economic performance and people's life chances, is attracting wider attention in both academic (e.g. Bowles & Gintis, 2001; Lochner, 2008) and in policy contexts (d'Addio, 2007) as ‘Cycles of Disadvantage’ (Rutter & Madge, 1976). Intergenerational transmission of variables – such as behavioural and psychological traits, and educational and socioeconomic status – that affect people's life chances have been extensively analysed, although there have been comparatively fewer studies on continuities in reproductive behaviours (Jennings, Sullivan, & Hacker, 2012, p. 543), and even fewer rigorously investigating mechanisms of transmission.

Intergenerational fertility transmission (IGFT) is arguably the most important continuity of all for a number of reasons. IGFT will influence population size and composition, since people born into large families are themselves more likely to make an above-average contribution to overall numbers in succeeding generations and to have larger kin networks (Murphy & Knudsen, 2002). In particular, the spread of genetic diseases is heavily influenced by such parent–child correlations (Austerlitz & Heyer, 1998). Without children, many forms of transmission cannot happen. The number of children may have an independent effect on transmission of other variables and therefore affect the distribution of such characteristics in the next generation. One of the first major interests in the field was in the interaction between (heritable) IQ and family size based on the Scottish Mental Health Enquiries of 1932 and 1947 (Kevles, 1985). This led to widespread but unfounded fears that average IQ would decline as a result of the reported association. This debate was particularly controversial and may have made many researchers reluctant to engage with the topic.

1. Data sources

Substantial similarities in lifetime experiences of successive generations may be observed when different societies or sub-groups within a society are compared: one sub-population may live longer, have larger family sizes, marry earlier or divorce more frequently than another. The importance of including controls for such heterogeneity was identified by Pearson, Lee, and Bramley-Moore (1899, p. 277) who distinguished between intergenerational correlations that could arise because of the transmission of fertility between parents and children, and ‘spurious’ associations due to differences between subgroups. Macro-level comparisons may provide useful insight into patterns and processes (Parrado, 2011), but such studies provide no information on the extent to which continuities exist at the individual family level. To do so requires high quality micro-data from sufficiently representative populations over extended time periods, including data on socioeconomic status, to control for confounders. Availability of data may constrain or promote particular explanations for observed patterns. Initially the assumption was that inherited physiological factors, essentially fecundability, were the key to intergenerational transmission (Pearson et al., 1899) which was understandable in the context of knowledge at the time. Occasionally biological/genetic effects have subsequently been suggested (e.g. Huestis & Maxwell, 1932) based on speculation rather than evidence. The role of biological factors came to be downplayed for a number of reasons. The introduction of social surveys opened up a range of competing explanations; researchers were no longer forced to rely on the limited information from censuses and vital registration which was unable to add much to analysis of intergenerational issues apart from co-residence. Questions could be included in surveys on topics such as size of family of orientation, which can be analysed together with the fertility of the woman (and sometimes of the man). Only later did the benefits of longitudinal data and historical sources become appreciated, aided by the emergence of methods such as event history analysis. Life course approaches have been much more widely used as the importance of early-life and intergenerational factors came to be recognised. The availability of databases for populations such as in parts of Québec, Spain, Sweden (Umeå and Scania), US (Utah) and Iceland (International Commission of Historical Demography, 2013; Kelly Hall, McCaa, & Thorvaldsen, 2000), together with some of those used in this set of papers, provide a firm basis for investigating long-run patterns.

2. Mechanisms underlying intergenerational fertility continuities

Although Pearson et al. (1899) chose a biological explanation, a view developed that biology was not of major importance and this explanation was rejected by many because of its association with eugenics which came to be vilified because of its association with the racial programs of the Nazi regime and the actions of some ‘progressive’ governments in, for example, sterilisation of those regarded as unfit. The rapid post-War changes in period mortality and later in period fertility appeared to emphasise the importance of the contemporary environment rather than what happened well in the past (child socialisation and education appeared to be exceptions). ‘Socialisation’ came to be widely-cited as the primary mechanism for IGFT in contemporary societies, although the empirical basis for doing so was weak as alternative explanations were rarely investigated.

Since children are usually socialised by their biological parents over a long and formative period, separating out biological and socialisation factors is difficult. Within ‘biology’, explanations may include inheritance of physical and behavioural characteristics. Within ‘socialisation’, it is difficult to establish both pathways of influence and the relative contribution of attitudes, preferences and social learning. Inherited patterns of infant mortality are likely to involve both sets of factors. It is unclear whether a clear-cut distinction between ‘biological’ and ‘social’ processes can be made. Environment affects biological transmission, not only because it may select some individuals for survival and hence reproduction, but it is also becoming apparent that it can affect gene expression over multiple generations (Drake & Walker, 2004; Matthews & Phillips, 2010). Genetic endowment plays a role in filtering people into the environments they experience, which may in turn impact on later outcomes. Nevertheless, the distinction remains useful for presentational purposes.

3. Biosocial approaches

Behavioural genetics and related psychological approaches became a more mainstream area of research, although even here there was resistance to full acceptance, especially since the field became associated with controversies about reasons for observed differences between racial groups. Studies in the 1990s did start to investigate the determinants of inherited fertility from this perspective using the Danish twin register study covering childbearing in the period 1874–1964 to separate out genetic, environmental and shared components (Kohler, Rodgers, & Christensen, 1999; Rodgers, Kohler, Kyvik, & Christensen, 2001). This showed that the genetic contribution to intergenerational fertility correlation increased during the period of demographic transition in that country. At the same time, analysis of a wide range of existing studies suggested that the magnitude of IGFT had increased at about the time of demographic transition from a very low value in pre-transitional populations (Murphy, 1999). The suggestion was that the wider social space created allowed different forms of behaviour to emerge – the relevant organ was the brain rather than the reproductive ones. The paper by Bras, Van Bavel, and Mandemakers (2013) adds considerably to understanding in this area using the large-scale GENLIAS database for sibling pairs born in Zeeland in the period 1812–66. The finding that heritability increased during the demographic transition is confirmed in a new setting, especially among women. In addition, the availability of individual-level data by occupation, type of area (e.g. urban/rural) and area-level data on religion enables the paper to combine mainstream genetic methods with testable hypotheses derived from social science to provide insights into the processes underpinning the long-run and largely irreversible process of fertility reduction, particularly relevant at a time when we are passing the point at which more than half the World is now living in countries with ‘below-replacement’ fertility (UN, 2011). The paper also provides an alternative approach to the disaggregation of variance that separates nature and nurture in addition to the natural experiment of whether a set of same-sex twins is identical or not: the fertility outcomes for children who were or were not wholly brought up by their natural parents.

4. Studies from non-Western countries

Studies from non-Western countries are rare, although recently information has become available from a small number of middle-income countries from the 2001 round of the International Social Survey Programme (ISSP) and for a wide range of developing countries from the Demographic and Health (DHS) programme (although these sources do not include the largest ones such as India and China, Murphy, 2012). Studies of such societies in earlier periods are especially rare. The study by Jennings and Leslie (2013) of the small Caribbean island of Saba (12 km²) is particularly relevant. The existence of black and white (also mixed-race) groups followed across an extended period from 1876 to 2004 permits the role of economic and demographic change on these two separate groups to be assessed. The conclusion of this paper reinforces the finding that the strength of the intergenerational relationship is increasing over time, but emphasises that time is, in part, a proxy for increased control over lives – the context for means, motives and opportunities for IGFT to emerge. The design allows temporal changes to be contrasted with cross-sectional differentials, with reduced options identified among both white women (due to heavy male absence) and black men (who had restricted economic opportunities) being associated with lower IGFT. The underlying mechanism, which is related to the ability to achieve goals, produces a unified explanation for both temporal and cross-sectional patterns. As with all studies in this special edition, the analysis goes beyond presentation of simple bivariate correlation coefficients. They include a range of approaches that can help to identify the contribution of proximate determinants of overall fertility such as age at first birth as well as the role of remote determinants, both in childhood and around the time of childbearing.

5. Migration

A number of studies in this special edition are concerned with the interaction of intergenerational fertility transmission and migration. Early studies often ignored migration since available datasets were often confined to cases where both generations remained in a geographically fixed area from which the records are drawn. Data sources that compare outcomes for two or more successive generations are likely to lose mobile dyads since few sources track individuals over extended geographical areas. Results can be sensitive to the area covered (Gagnon & Heyer, 2001). Migration has a number of advantages when included in studies of IGFT. A key objective is to identify cases where there is sufficient interesting variation to obtain insight into determinants. Migration provides different environments for successive generations and therefore allows the impacts of the household/family and wider environment to be partially separated. It can also test the general finding that the more permissive the environment, the greater the scope for intergenerational continuities to manifest themselves. A general problem with use of migrants to assess the relative importance of influences of areas of origin and destination is that migrants may not be typical of either area. Migrants are selected by being migrants, but the high proportion of migrants to Geneva in the study of Schumacher (in press) suggests that they may be less selected than will often be the case, and therefore they may be closer to the ideal model of random allocation that would allow stronger conclusions to be drawn. He compares the behaviours of different groups of migrants into the urban setting of Geneva in the period 1800–80; that also includes the period of demographic transition, which occurred at different times in the migrants' various original areas. Information on fertility of migrants' area of origin is used to assess possible roles of earlier and contemporary influences, and the nature and extent of socialisation in urban and rural settings. Area of origin retains an influence but a noteworthy fact is that this appears to work only through the husband, although this study does not include individual-level information. The suggestion is that family limitation was dependent on the husband's control over the available methods of celibacy and withdrawal, in which the wife had little role, although Jennings et al. (2012) tested the same hypothesis in the Utah data base and found that the relationship was stronger with mother's than father's sibship size using individual-level data. Other studies have shown that the wife's family of orientation size has an influence at least as great – and often greater – than that of her husband (Reher, Ortega, & Sanz-Gimeno, 2008). The importance of urbanisation for demographic processes is again emphasised – the outcome of positive intergenerational fertility transmission is common, but the mechanisms by which it is achieved may be different since the paper of Bras et al. (2013) emphasises the key role of women.

6. Intergenerational continuities in infant mortality

For analysis of overall population dynamics, effective family size (EFS) rather than completed family size (CFS) is more relevant so mortality needs to be taken into account. Differences in findings for EFS and CFS may help to elucidate the role of alternative mechanisms, including the roles of instrumental assistance from close kin around the time of childbearing as ‘co-operative breeders’ (Kaptijn, Thomese, & van Tilburg, 2010; Kramer, 2010) and of attitude creation and reinforcement (Balbo & Mills, 2011; Lyngstad & Prskawetz, 2010; Matthews & Sear, 2013). The distinction between EFS and CFS is only really important when mortality rates are not low – across a range of 45 developing countries the relationship is slightly stronger with CFS than with EFS (Murphy, 2012). Completed family size and infant mortality continuities might arise because of transmission of biological fecundity or through socialisation if children tend to imitate their parents' behaviours such as short duration of intensive breast-feeding. Historically, such continuities in behaviour are observable in different villages in the same territorial unit (Knodel, 1988), making it difficult to generalise from localised studies. The study by Vandezande and Matthijs (2013) is concerned with the extent to which infant mortality rather than fertility risks are transmitted intergenerationally in Antwerp using the Belgian COR*-database for births in the period 1846–1905. Results indicate a substantial impact of the parents' childhood mortality environment especially in the case of high mortality clusters, which is almost unaffected by inclusion of other controls such as type of area of residence and father's occupation which are weak and generally statistically non-significant (apart from those with unknown father's occupation, which includes unmarried mothers). This paper is also relevant to a wider literature concerned with the role of kin on child survival rather than on fertility per se. Lawson and Mace (2011) reviewed the evidence and concluded that there was generally a positive relationship between fertility and number of sibs, although not in all cases (Tymicki, 2004), although the contribution of siblings to reproductive success is generally small compared to that of surviving parents.

7. Contemporary populations

The paper by Stanfors and Scott (2013) uses the large and rich set of factual information available in Nordic registers to investigate intergenerational continuities in early childbearing, a topic of high policy interest. Findings of a strong effect of the childbearing of the family of origin in Sweden are similar to those from less egalitarian societies where such studies are more common. As expected, all studies also find a strong effect of the socio-economic situation of the younger generation, but this study shows that the effect of the older generation's socio-economic situation has little influence. It may be that this arises because egalitarian societies such as Sweden mitigate the transmission of such factors, but the effect of their parents' age at starting childbearing remains strong and robust to inclusion of controls. As with historical studies, there is no information about the establishment and transmission of preference and attitudes, although the authors conclude that these are plausible explanations.

8. Attitudes

The paper by de Valk (2013) is concerned with the competing roles of influence on childbearing preferences inside the household and outside it among co-resident parent–child dyads in four ethnic groups in comparison with a native Dutch sample. Intergenerational correlations within these groups are about 0.3, higher than values of around 0.2 for actual completed fertility typically found. The strongest variable predicting disagreement between parents and children is the younger generation's number of siblings (i.e. the parents' number of children). When this is included, ethnic group coefficients become statistically insignificant. Dyads from large family sizes are less likely to agree: the odds ratio for the child and parent having different views is 0.2 for each additional sibling, both for the child preferring more children than the parent and vice versa. This finding raises the question of what mechanisms might lead to such a finding.

The second topic considered is preferred age at starting childbearing. In this case, few variables are statistically significant, and, in particular, the family size coefficient is negligible. Given the link between overall fertility and age at starting childbearing, the importance of the family size variable on the overall outcome, but its lack of influence on a key proximate determinant, is surprising. However, selection must be an issue in interpreting these results since the analysis is based on children who were still living with their parents, when many of their contemporaries will have had or are in path to having children.

9. Conclusions

There is a need for research in this area to draw on a range of studies before making generalisations. For example, Pluzhnikov, Nolan, Tan, McPeek, and Ober (2007) analysed genealogies of 161 Hutterite individuals. They found high statistically significant correlation coefficients for the total number of children born to couples and their children, 0.31 for sons and 0.23 for daughters (after controlling for time trend). The Hutterites are a well-documented socio-economically and behaviourally homogeneous natural fertility population that might be expected to exhibit little intergenerational fertility continuity as is the case in most natural fertility populations investigated. It may be that these different findings arise because the strength of the relationship may be sensitive to the precise way the study population is defined and to local circumstances, and therefore vary over time and space so replication is necessary.

Fertility is influenced by direct (‘proximate’) and remote (‘upstream’) causes over extended periods. The only indicator available making for comparisons across a broad range of studies is product moment correlation coefficients of completed or effective fertility in successive generations. However, with such crude measures it is unclear how the relationship changes when even simple socio-economic controls are included and they give no information about the proximate determinants, such as, for example, whether the effect works through continuities in age at first or age at last birth. All of the papers here attempt to address the question of mechanisms using a variety of approaches.

The main finding from these papers is that intergenerational reproductive continuities are highly pervasive in very different contexts, varying from levels of infant mortality of over 200 (in the study of Vandezande & Matthijs, 2013) to under five per 1,000 live births (in the study of Stanfors & Scott, 2013). These results are also robust to the inclusion of socio-economic controls for both generations concerned. In general, the effect of such controls is to reduce the strength of the intergenerational relationship by a relatively small amount. The papers also emphasise that the relationship is not mechanistic, but is context-specific and can vary between different sub-groups and over time, but within an envelope that is tending to show an increasing strength over time. The studies also tend to show that the greater the individual autonomy, the more likely IGFT is to be observed. Such behaviour is usually interpreted as an unachieved preference that can only be met when opportunities arise in the case of the adoption of fertility limitation, which raises the question of whether IGFT is similarly an indicator of more ‘modern’ societies.

It might be thought that a subject that is now in its third century of investigation would have exhausted possibilities, but this is not the case. Clearly trends in the majority of the non-Western World remain unexplored. While the associations with characteristics of mothers and fathers have been explored, few studies have looked at joint parental characteristics, even though the strength of association between partners' family size of origin appears to be as strong as that between parents and children (Murphy, 2006). Another issue is the relationship between continuities in reproductive behaviour and in other areas. How far is intergenerational similarity in behaviour a ‘package’ or one where choices and constraint vary between different domains?

Collectively, these papers emphasise that possibilities now exist to examine intergenerational processes in detail using larger data sets that contain information on more than one generation: in addition to the sources included here, there are increasing numbers of data sources for a range of historical populations, but the existence of clear-cut positive relationships in reproductive behaviours across a number of different historical populations, often associated with the onset of demographic transition, is not in doubt.

Notes

1. The paper by Jennings and Leslie was not presented. The Seminar was organized by the Scientific Research Community Historical Demography.