Recent developments in demography and epidemiology have focused attention on the links between events in early life and mortality later in life. Processes developing over the life course can only be studied with long-run longitudinal data, which is quite difficult to obtain. Historical demography is unusually rich in such data. Indeed, one might say that any longitudinal data set describing complete life histories (e.g. more than fifty years) must be viewed in a historical perspective, because changing epidemiological, economic, and social contexts may affect the impacts of early life conditions on later health. This article is a preliminary attempt to link together demographic events and economic characteristics across the life course from childhood to old age. We are motivated by a particular interest in the effects of early life conditions on old age mortality, but we also want to know whether experiences in childhood can be mediated by experiences in midlife.

A simplified model of interrelations between wealth and health over the life course is illustrated in Figure 1. The arrows in the diagram indicate causal pathways linking outcomes in mid- and later life to earlier conditions. The model describes four outcomes. First, we hypothesize that health in childhood is strongly affected by parents’ wealth (Aïach 1996; Power, Manor and Fox, 1991). We cannot measure health directly, of course, but the heights of men who were examined for military service provide a summary indicator of health and nutrition in childhood. While height is not a measure of health, it is strongly correlated with conditions that lead to good health. Height reflects cumulative experiences of both diet and diseases in childhood. Poor diet has a direct effect on adult stature, and diseases, like diarrhea, also prevent children from absorbing nutrients from the food that they do eat. Haines (1998) found that the heights of nineteenth-century American soldiers were strongly related to epidemiological conditions in their home communities.

Data for this study come from the commune of Sart-lez-Spa, located in the Belgian Ardennes (See Alter, Neven and Oris, 2004a). Nineteenth-century Sart covered a large, sparsely populated area. In spite of its close proximity to the most advanced agricultural and industrial areas on the European continent, Sart was poor and relatively backward. The terrain is steep, soils are poor, and a peat bog covers a large area. At the beginning of the century swidden agriculture was still being practiced in Sart’s extensive, communally-owned forests (Vliebergh and Ulens, 1912). All accounts speak of the area’s poverty, and there are signs of increasing Malthusian pressure as the population grew from 1,791 in 1806 to 2,380 in 1846. After 1850 conditions in Sart improved dramatically. Rapidly expanding factories in the nearby city of Verviers drew migrants from Sart, reducing population pressure. More advanced agricultural practices were introduced, including artificial fertilizers.

After the death of a property owner, the value of the estate and a list of heirs were recorded in the Enregistrement des successions. Accurate valuation of the estate was important, because few people left wills and the Belgian Civil Code, which was based on the Napoleonic Code, divided estates equally among the heirs. Successions were usually registered within three months of the date of death, but large and complicated estates might take longer to inventory. We have collected the estate valuations and the names of heirs for persons who died in Sart between 1818 and 1900, and these records have been linked to the population registers. Successions are available for less than 20 percent of those who died in nineteenth-century Sart (See Appendix). When no succession is available, we assume that the decedent had no wealth.

From 1818 to 1822 the commune of Sart imposed a special tax on all taxpayers. This was a period of extreme distress in much of Europe. The region had been annexed to France in 1794, and records from the French regime show frequent demands for military supplies and conscripts. Napoleon’s defeat in 1815 was followed by harvest failures, high prices, and a major epidemic of typhus. In addition, Sart still had outstanding debts from loans contracted during the eighteenth century. The 1818 and 1822 tax registers show both “personal” taxes and business licenses (patentes), and individuals were assessed approximately 116% of the sum of those amounts. We use these tax assessments as indicators of wealth in 1818 and 1822.

The population register of 1843 includes the amount of land belonging to each household. We use the area (in hectares) listed with each household as an indicator of wealth in 1843.

In the 1870s a map showing landowners and the value of each property was constructed from cadastral records for Sart. This process was conducted throughout Belgium and is known as the Atlas Popp. We use the total value of land and buildings associated with each property-owner as an indicator of wealth in 1870.

From the time that Eastern Belgium was under French control, all young men were required to report for military conscription when they reached age 20 (1794-1814 and 1849-1880) or age 19 (1816 to 1847). The conscription process included a medical examination at age 19 (1816-47) or 20 (1806- 1815 and 1848-1880). Men who were physically unfit or shorter than the minimum height (155 centimeters) could be excused from military service or required to appear for re-examination a year later. Military service could also be canceled or postponed for men who had brothers on active duty. We have linked records from military conscription lists beginning in 1834 to the population registers [7].

Our previous research using military conscription lists from communities in Eastern Belgium suggested a strong link between childhood conditions and height (Alter, Neven and Oris, 2004b). We found several important patterns. First, at the beginning of the nineteenth century men from this region were very short by international standards (Alter, Neven and Oris, 2004b). The average height of men in Sart examined before 1850 was less than 163 centimeters (Table 4). In contrast, Quetelet (1869, 354) computed an average height of 164 centimeters for all Belgian men examined between 1842 and 1865, and even this is below the averages of other contemporary European countries [8]. Second, heights increased substantially, going from 161 for those measured in the 1820s to 165 around 1880 (Alter, Neven and Oris, 2004b, Table 5). The trend in heights in Sart is shown in Table 1. Third, heights were strongly related to conscripts’ occupations. At the extremes, we found an average height of 157 for men working as day laborers compared to 164 and 167 for those who were still in students in rural and urban areas respectively (Alter, Neven and Oris 2004b, Table 8) [9]. Since occupation at age 20 is primarily a reflection of the economic position of the parental household, we interpret this as strong evidence that height reflected conditions in early life. Fourth, socio-economic differences in height narrowed after 1850. While heights among all groups increased, the increase was greatest among those who were shorter in earlier years. This suggests that the improving conditions after 1850 had the greatest impact on the nutrition and health of the poor.

To examine the timing of marriage we use a Cox proportional hazards model (Cox 1972) shown in Table 3. Relative risks estimated in this model show the effect of each covariate on the instantaneous rate of transition from single to married. A relative risk greater (less) than one means that a covariate increased (decreased) the probability of marriage. When the estimated relative risk does not differ from one, the covariate has no association with the timing of marriage. Models have been estimated for unmarried men observed between ages 18 and 50 in each cohort of military examinations.

In Table 4 we ask whether parents’ wealth, height, or marital status affected wealth in mid-life. The dependent variable is our estimate of wealth at age 50 constructed by the method described in the Appendix.

We use the Cox partial likelihood model shown in Table 5 to estimate the effects of early life conditions on old age mortality. Again, we find an important difference between the early and later cohorts. Mid-life wealth and height mattered a great deal in the early cohort and not at all in the later cohort. This suggests that a ten percent increase in wealth would have reduced the risk of dying by more than five percent; while each additional centimeter increase in height was associated with a six percent reduction in the risk of dying. These effects do not appear in the later cohort.

We have emphasized both the long-run effects of conditions early in life and the mediating effects of conditions in mid-life. The interactions between health and wealth over the life course sometimes make it difficult to separate their effects. Wealthier families have more resources to invest in the health of their children. Healthier adults are more likely to marry, and married couples have advantages in accumulating additional wealth. Wealth, marriage, and a healthy childhood all contribute to longer life in old age. The multi-period, longitudinal model used here helps to show these overlapping and reinforcing effects.

This appendix explains how we adjust measurements of wealth of different kinds to a common standard and age. We begin with a simple model of the growth of wealth over the life course: