Elsevier

Journal of Health Economics

Volume 29, Issue 5, September 2010, Pages 617-629
Journal of Health Economics

Long-run effects on longevity of a nutritional shock early in life: The Dutch Potato famine of 1846–1847

https://doi.org/10.1016/j.jhealeco.2010.06.001 Get rights and content

Abstract

Nutritional conditions in utero and during infancy may causally affect health and mortality during childhood, adulthood, and at old ages. This paper investigates whether exposure to a nutritional shock in early life negatively affects survival at older ages, using individual data. Nutritional conditions are captured by exposure to the Potato famine in the Netherlands in 1846–1847, and by regional and temporal variation in market prices of potato and rye. The data cover the lifetimes of a random sample of Dutch individuals born between 1812 and 1902 and provide individual information on life events and demographic and socioeconomic characteristics. First we non-parametrically compare the total and residual lifetimes of individuals exposed and not exposed to the famine in utero and/or until age 1. Next, we estimate survival models in which we control for individual characteristics and additional (early life) determinants of mortality. We find strong evidence for long-run effects of exposure to the Potato famine. The results are stronger for boys than for girls. Boys and girls lose on average 4, respectively 2.5 years of life after age 50 after exposure at birth to the Potato famine. Lower social classes appear to be more affected by early life exposure to the Potato famine than higher social classes. These results confirm the mechanism linking early life (nutritional) conditions to old-age mortality. Finally, higher food prices at birth appear to reduce later life mortality of children of farmers from higher social classes. We interpret this as an income effect.

Research highlights

▶ Exposure to a nutritional shock in early life negatively may affect survival at older ages. ▶ This paper uses the exogenous variation caused by the severe Potato Dutch famine of 1846–47 to identify causal mechanisms. ▶ We find strong evidence for long-run effects of exposure to the Potato famine. ▶ Boys and girls lose on average 4, respectively 2.5 years of life after age 50 after exposure at birth.

Introduction

Exposure to adverse conditions in early life (such as exposure to infectious diseases, inappropriate nutrition, stress) may causally affect health and mortality at old ages (Forsdahl, 1978, Fridlizius, 1989, Barker, 1998, Kermack et al., 2001, Fogel, 2004). The relationship may be explained by direct effects as suggested in the “Fetal Origin Hypothesis of diseases” of Barker (1998). He postulated that vital organs and the immune system might develop inadequately if the body faces adverse nutritional events in utero and/or the first stages of life, and that this may increase predisposition to chronic diseases at old ages. Alternatively, the relationship may be explained through poorer health and socioeconomic achievements through life (Kuh and Ben-Shlomo, 2004). For instance, a bad start in early life may influence later child health, which in turn may affect future educational attainments, future earnings, and adult health. Testing causal effects in this literature is not trivial because one needs data that cover a sufficiently long time period to measure long-run effects and, in addition, one needs independent variation in early life conditions.

Most evidence is based on studies using birth weight, or other (anthropomorphic) measures at birth or in infancy, as a proxy for conditions early in life (see Doblhammer, 2004, Section 1.3, and Koletzko et al., 2005, for a comprehensive review of the available epidemiological evidence). Most of the studies find associations between early life indicators and increased morbidity rates at older ages (e.g. Barker et al., 1991). However, birth and infant outcomes may depend on unobserved (genetic and socioeconomic) factors that may also explain health at older ages and, as a consequence, the results may be confounded.

Experimental animal research supports the long reach of early nutrition (Barker, 1998, Vickers et al., 2000), but these results may not translate to humans. A recent strand of literature uses techniques based on “natural experiments” to identify causal links. These techniques use “natural” events (such as timing of birth, famine, and epidemic) or macro-events (such as economic changes) as source of external variation to better understand causal mechanisms. These events affect the hypothesized causal factor but are most probably independent of suspected confounders. This is also the approach taken in this paper. More specifically, our focus is on the long-run consequences of malnutrition early in life. We use exposure to the Dutch Potato famine as a main proxy for exposure to severe malnutrition in early life. The Dutch population faced a famine of unprecedented severity in 1846–1847. In this period, all potato harvests in Europe failed due to the potato blight and to bad weather conditions (Mokyr, 1980, Mokyr and Ó Gráda, 2002, Ó Gráda et al., 2007).

Previous work on the long-run effects of famines show mixed results. For instance, no significant effect is found for individuals born in Finland during the famine of 1866–1868 (Kannisto et al., 1997). Studies on the siege of Leningrad in 1941–1944 show some conflicting results: individuals exposed during gestation or in infancy do not experience higher risks for coronary heart diseases in adult life (Stanner et al., 1997) while men exposed at age 6–15 are more vulnerable to cardiovascular diseases and cardiovascular mortality at older ages (Koupil et al., 2007). These results may be explained by “selection problems”, resulting in the fact that the weaker individuals are less likely to survive the crisis. The infant mortality rates in Finland and Leningrad increased in these periods to 47 and 40%, respectively. As a consequence, estimates of impacts of the famine based on data of the survivors may be biased towards zero. On the other hand, studies with a similar research design, such as the studies based on the Dutch “Hunger Winter” under German occupation at the end of World War II (Ravelli et al., 1998, Roseboom et al., 2001, Painter et al., 2005, Lumey et al., 2007) and the studies based on China's great famine (Meng and Qian, 2006, Chen and Zhou, 2007, Almond et al., forthcoming), establish significant long-term effects on adult morbidity. Similarly, in the context of a developing country, Moore et al. (1999) find excess mortality at young adult ages for individuals born during or just after the rainy hungry season in Gambia. Ideally, for the analyses of later life mortality, data covering full lifetimes of individuals are required to take dynamic selection into account. Such data are scarce.

We exploit a historical data set, the Historical Sample of the Netherlands, covering the full lifetimes of about 14,000 individuals born in three large Dutch provinces in the period 1812–1922 (see Mandemakers, 2000, for a description in English). The information is mainly obtained from the individual administrative records of vital events kept by municipalities and provinces. We capture the nutritional conditions in early life by exposure to the Potato famine and by variations in prices of potato and rye. Potato and rye were essential components of the Dutch diet in the 19th century. Market food prices are commonly used indicators for access to food in historical studies (Almond, 2002, Jacobs and Tassenaar, 2004). Our analyses are performed in two steps. First, we non-parametrically compare the total and residual lifetimes of individuals exposed and not exposed in utero and in early infancy (until age 1) to the Potato famine. Second, we perform parametric analyses in which we model the individual mortality rate as a function of exposure to the Potato famine, food prices, other conditions early in life, time-varying covariates measuring current conditions and individual characteristics. It is important to control for other past and current conditions as these may partly explain the link between early life nutritional conditions and survival. If we do not control for this, this may be picked up by the famine parameters. Concerning other conditions early in life, we control for variations in the disease environment and in macro-economic conditions early in life. Indeed, the prevalence of infectious diseases and poverty may considerably increase during famines, due to e.g. social and economic disorders, changes in food composition or debilitated immune systems (Livi-Bacci, 1991, Mokyr and Ó Gráda, 2002, Ó Gráda, 2007). The theory postulates that the pathogen environment and economic conditions early in life may also causally affect health and mortality at old ages (Fridlizius, 1989, Ó Gráda, 2007). Recent empirical studies confirm the long-term effects on mortality at old ages of exposure early in life to infectious diseases (Bengtsson and Lindström, 2000, Bengtsson and Lindström, 2003, Costa, 2003, Almond, 2006, Bozzoli et al., 2009) and to adverse economic conditions (Van den Berg et al., 2006). Following Bengtsson and Lindström, 2000, Bengtsson and Lindström, 2003, we use crude infant and children mortality rates per province to proxy the exposure to infectious diseases. As in Van den Berg et al. (2006), we use the Gross National Product (GNP) to characterize economic conditions in early life. Finally, in our model, we also allow for instantaneous effects of the Potato famine, major epidemics, wars and economic conditions. We perform separate analyses for men and women, for farmers and non-farmers, and per social classes since there is empirical and/or historical evidence that the long-term effects may differ across genders, farmer status, and social classes (Gavrilov and Gavrilova, 2001, Bengtsson and Lindström, 2003, Gavrilova et al., 2003, Van Zanden and van Riel, 2004, Almond, 2006, Van den Berg et al., 2006, Ó Gráda et al., 2007). The non-parametric and parametric results agree and demonstrate the long reach of early life nutritional environment on mortality at older ages.

From a public health policy perspective, understanding the causality linking early life nutritional conditions and survival is important. First, it may provide a better understanding of the increase in life expectancy that most developed countries experienced in the two last centuries. Second, it may help the design of successful strategies to prevent premature death. Say that nutrition in early life plays a substantial role in determining health at older ages. Survival and the quality of life of old individuals may be improved by policies that help young children and pregnant women in periods of shortages of food. All this may currently be particularly relevant since about one-fourth of the children in the developing world are currently underweight or stunted (UNICEF, 2006). Those most at risks may be children of developing countries, but also migrants from these countries and inhabitants of countries undergoing rapid socioeconomic transitions. Research shows indeed that the long-term effects of adverse early nutritional conditions may be reinforced when combined with western diet and lifestyle later in life (Kuh and Ben-Shlomo, 2004). Clearly, nobody will deny the improvement of early life circumstances, but the benefit of these interventions may be even larger than assumed until now.

Section 2 presents a brief overview of the Dutch trends in mortality and the characteristics of the Dutch agricultural sector in the 19th century. Section 3 provides a description of the data. Section 4 presents the non-parametric analyses. Sections 5 Parametric analyses, 6 Sensitivity analysis present the parametric analyses and the sensitivity analysis. Section 7 concludes.

Section snippets

Overview of the Dutch mortality trends and agricultural sector in the 19th century

At the beginning of the 19th century, the Netherlands witnessed relatively high infant mortality compared to the rest of Europe. The rates only began to decrease after the 1870s, with a sharp fall until the Second World War. The conditions of the water and the breastfeeding practices explain to a large extent the high (infant) mortality (Wintle, 2000). We return to that below. The drop in mortality rates after 1870 is mainly related to the increase in the availability of better food, the

Individual information

The HSN data are derived from the registers of birth, marriage, and death certificates. Currently, we have access to a cleansed sample of 13,718 individuals. This is a random sample of individuals born in the provinces of Utrecht, Friesland and Zeeland3 between 1812 and 1922. The three provinces were jointly rather representative of the Netherlands in terms of economic activity and mortality patterns. At the time of the study, the end of the

Non-parametric analyses

We tested equality of means in total and residual life expectancies of individuals exposed (treated) and not exposed (controls) to the Potato famine early in life. We considered three birth cohorts: September 1st 1846–June 1st 1848 (196 individuals born, or in utero during the most tragic famine period); September 1st 1848–September 1st 1855 (945 individuals born after the famine); September 1st 1837–September 1st 1844 (929 individuals born before the famine period).11

Parametric analyses

In order to correct for other relevant mortality determinants, we estimate proportional hazard models with time-varying covariates, in which we interact the age dependence function with current macro-conditions and with the macro-conditions in early life. We will return to this below. We only include individual characteristics at birth since the individual variables later in life may be endogenous (see Section 3.1 for a description of the individual variables). For the age dependence we adopt a

Sensitivity analysis

The results presented in Section 5 are very robust.26 First, with respect to the effect of the Potato famine, the results remain similar when we correct the analyses for the length of exposure to the famine (in days). The coefficients associated with effects at ages older than 50 are smaller (equal to 0.53, T-value equal to 1.6 for males and 0.54, T-value equal to 1.8 for females) when we consider a less catastrophic period (e.g.

Summary and conclusions

Mortality rates at older (50+) ages are higher for those born during the Potato famine. These effects are large and significant. Residual life expectancy at age 50 is reduced by about 3 years. These effects remain after controlling for individual characteristics at birth and early life exposure to infectious diseases (as measured by local infant mortality rates) and economic conditions (measured by the cyclical change in the GNP). With the latter two variables we aimed to capture the pathogen

Acknowledgements

We thank Bertie Lumey, Thomas Lufkin, Kees Mandemakers, Christian Monden, Ryan Mutter, Maurits van Tulder, Cor Zonneveld, and participants of a European workshop on Health Econometrics, the 2008 ASHE conference, and Netspar and Tinbergen Institute conferences, for insightful comments. The “Historical Sample of the Netherlands Dataset release UZF.02” was kindly provided by the International Institute of Social History (IISG) in Amsterdam. We are very grateful to Arthur van Riel for kindly

References (53)

  • D.J.P. Barker

    Mothers, Babies, and Health in Later Life

    (1998)
  • T. Bengtsson et al.

    Childhood misery and disease in later life: the effects on mortality in old age of hazards experienced in early life, southern Sweden 1760–1894

    Population Studies

    (2000)
  • T. Bengtsson et al.

    Airborne infectious diseases during infancy and mortality in later life in southern Sweden, 1766–1894

    International Journal of Epidemiology

    (2003)
  • C. Bozzoli et al.

    Adult height and childhood disease

    Demography

    (2009)
  • Clay, K., Troesken, W., 2006. Deprivation and Disease in Early Twentieth-century America, NBER Working Paper 12, p....
  • Devos, I., 2005. The Evolution of Life Expectancy in Belgium: 18th–20th Century (in Dutch),...
  • G. Doblhammer

    The Late Legacy of Very Early Life

    (2004)
  • T. Dyson

    On the demography of South Asian famines: Part I

    Population Studies

    (1991)
  • R.W. Fogel

    The Escape from Hunger and Premature Death, 1700–2100: Europe, America, and the Third World

    (2004)
  • A. Forsdahl

    Living conditions in childhood and subsequent development of risk factors for arteriosclerotic heart disease

    Journal of Epidemiology and Community Health

    (1978)
  • G. Fridlizius

    The deformation of cohorts: nineteenth century mortality decline in a generational perspective

    Scandinavian Economic History Review

    (1989)
  • L.A. Gavrilov et al.

    The effects of early life events on adult lifespan are sex-specific

    The Gerontologist

    (2001)
  • N.S. Gavrilova et al.

    Early-life predictors of human longevity: analysis of the XIXth century birth cohorts

    Annales de démographie historique

    (2003)
  • V Kannisto et al.

    No increased mortality in later life for cohorts born during famine

    American Journal of Epidemiology

    (1997)
  • W.O. Kermack et al.

    Death-rates in Great-Britain and Sweden. Some general regularities and their significance

    International Journal of Epidemiology

    (2001)
  • B. Koletzko et al.

    Early nutrition and its later consequences: new opportunities. Perinatal Programming of Adult Health

    Advances in Experimental Medicine and Biology

    (2005)
  • Cited by (103)

    • The effects in-utero exposure to influenza on mental health and mortality risk throughout the life-course

      2021, Economics and Human Biology
      Citation Excerpt :

      Our findings of generally stronger health effects of influenza exposure for males are consistent with evidence from other in-utero shocks. Havari and Peracchi (2017), Scholte et al. (2015), van den Berg et al. (2015), Lindeboom et al. (2010) all find that the health effects of in-utero exposure to under-nutrition are stronger for males. Similarly, Arthi (2018) find that the effect of in-utero exposure to the Dust Bowl on physical disability is statistically significant only for males.

    • Local historical context and multigenerational socioeconomic attainment

      2021, Research in Social Stratification and Mobility
    View all citing articles on Scopus
    View full text