This article by Monica Costa Dias also appears in the MethodsNews Spring 2014 issue.
Joint research by PEPA and Pathways is investigating some of the mechanisms that link life experiences and later life health. In particular, the focus is on the interconnections between fertility histories and wealth, and how these impact on later life health.
Earlier research has established the importance of the associations between these variables. But the interpretation of these associations is notoriously difficult, confounded by the classic simultaneity problem. The difficulty lies in the fact that wealth, fertility and health are closely intertwined life-long processes. Hence, the many distinct mechanisms that may contribute to some observed correlation cannot be easily disentangled.
The links between fertility histories, wealth and later life health
Take, for instance, the strong positive correlation between health status and wealth that has been repeatedly found for many countries and time periods. One can think of three competing explanations for such association. First, it is conceivable that differences in SES have lasting effects on health. Second, it is equally plausible that the reverse is true, with poor health having cumulative effects on SES if, for instance, it impairs investments in skills, working and earnings capacity. And finally, underlying but (partly) unobserved factors including ability, parental background or other early life experiences may determine both health and economic wellbeing.
The relationship between fertility histories and health is potentially equally complex. Again, three alternative explanations can be considered. Women, in particular, may experience physiological consequences of pregnancy and childbirth, but health itself can be one of the determinants of fertility. A third explanation relates to the many costs and benefits of parenting, in the form of emotional fulfilment and/or strain, social interactions and support, economic costs or time use, to name only a few. While health outcomes may be affected by these factors, potentially in different directions, the relative importance of these costs and benefits may well depend on health in the first place.
The effects of fertility and wealth on later life health may partly result from the interaction of these factors over time. For instance, the time and economic resources needed for child-rearing may reduce other expenditures and put a downward pressure on wealth accumulation, with negative impacts on health. On the other hand, parenthood may affect many economic decisions such as whether or not to work or in which assets to invest, with unknown effects on wealth accumulation. It is also possible that offspring provide protection against negative economic shocks, both when young by supporting eligibility to family policies or promoting/facilitating stronger social networks, and as adults by actively assisting their parents.
Quantifying some of these links
It is difficult to visualise how one could disentangle all these interactions without resorting to a model of the life-course that explicitly models the income, wealth and health processes together with fertility decisions and their consequences. However, appropriate data may help shed light on some of these relationships in a flexible way.
This is the approach pursued by this project. It relies on the especially rich information available in the English Longitudinal Study of Aging (ELSA). ELSA contains detailed longitudinal data on health, employment, income, wealth and demographic variables for a large sample of individuals aged 50 and above and their partners. It also records retrospective data on a number of SES variables and health of respondent as a child, and on marital, fertility, education and employment histories. Finally, ELSA data can be linked to NI records for detailed earnings histories.
These data are used to investigate the joint effects of fertility histories and wealth shocks (or innovations) on later life health. This requires two crucial steps. The first is to ensure that individuals being compared are indeed similar. A particular worry is that early differences in health or socio-economic conditions may persist and later affect fertility choices, earnings capacity and health trajectories simultaneously. The common factors could lead to spurious correlations between these variables. Hence, the first step is to carefully match on background health and socio-economic information.
The second step is to find some variables associated with differences in fertility histories or wealth shocks but plausibly otherwise unrelated to health. Suppose one could find one such variable or instrument for fertility histories. The idea behind the use of an instrument is that individuals who have different levels or values of the instrument, but are otherwise identical, experience different fertility patterns for reasons unrelated to health. Variation across individuals singled out by the instrument can then be used to separate the impact of fertility on later health. This is the Instrumental Variables approach.
One instrument that has been used in the past to provide some random variation in fertility is the gender of the older two children among parents of two or more. The rationale is that parents of two daughters or two sons will be more willing to have a third child. In ELSA, this instrument is a strong predictor of one additional fertility episode among parents of at least two children. It is thus used to identify the impact of fertility on later life health.
A more challenging task is to determine a good instrument for wealth. The alternative currently being explored in this project is to focus on the impact of unexpected wealth shocks on health later in life, and the role of (earlier) fertility in protecting against those shocks. The most recent recession may well provide the variation needed to separate unexpected changes in wealth.