«Dalhousie University, IZA and HICN University of Mannheim, IFAU-Uppsala, VU University Amsterdam and IZA University of Mannheim Very Preliminary - ...»
Exploiting spatial and temporal variation in nutrition at
birth, using data on infant mortality and famines:
a study of long-run health eﬀects
Mevlude Akbulut-Yuksel1, Gerard J. van den Berg2, and Katharina Walliczek3
Dalhousie University, IZA and HICN
University of Mannheim, IFAU-Uppsala, VU University Amsterdam and IZA
University of Mannheim
Very Preliminary - Not For Citation
We study long-run health eﬀects of nutritional conditions early in life exploiting spatial and temporal variation in those conditions. For this purpose we consider individuals from a large set of municipalities in Germany, born in the years 1933-1950. Time series of local infant mortality rates are used to distinguish between aﬄuent municipalities and municipalities with adverse early-life conditions, and these series are also used to assess the local impact of the post-World War II famine in Germany. This methodology is novel and useful if the impact of a famine is strongly heterogeneous across regions and if local variation in the impact of the famine is not observed in the data. Individuals from municipalities where the famine was severe display an average realized adult height loss of about 2.5 cm. This predicts adverse health outcomes late in their life.
JEL classiﬁcation: I10, I12, J13, N34 Keywords: nutrition, famine, early-life conditions, developmental origins, infant mortality, height, adult health 1 Introduction The long-term eﬀects of health conditions during early childhood are subject of a growing literature. The present analysis adds to this literature by exploiting regional variation in infant mortality rates during the second world war (WWII). WWII was characterized by massive civilian-targeted warfare, which in Germany was followed by a huge famine in the aftermath of the war. This had catastrophic consequences for the well-being of children. During warfare, children are negatively aﬀected in their development in many ways. The well-being as infant can determine health and productivity for a whole lifetime. We quantify the long-term eﬀect of adverse health conditions in the region of birth on health much later in life. We apply diﬀerence-in-diﬀerences estimation to a unique data set of German municipalities before and after WWII, linked to individual-level health variables from the Socio-Economic Panel (SOEP).
Warfare is just one cause for bad economic and health situations during infancy. Economic crises and poverty can have similar eﬀects. However, the destruction and the hunger period caused by the war build an exogenous shock which allows to isolate their eﬀect. This is more diﬃcult to do for an economic crisis or poverty. In principle, the results in this study should be generalizable to the broader context of adverse medical or economic conditions during infancy.
In this study, we consider WWII and the subsequent famine as negative shock to infants’ living conditions. This shock did not equally aﬀect all regions in Germany. We can distinguish the level of exposure between the municipalities by comparing the infant mortality rate (IMR) between districts and over time. To measure the eﬀect on health later in life, we build up a quasi-natural experiment with individuals born between 1933 and 1950, considering their health quality measured by adult height as long-term outcome of the shock in early life. Since adult height seems to react systematically differently among men and women, see Barker, Osmond and Golding (1990), we restrict our analysis to men. The novelty of this study is the low-scale regional level of health conditions in infancy, which can much better capture variation in the exposure to food shortages than higher-level measures. We ﬁnd a negative and signiﬁcant eﬀect of high infant mortality in the region of living during the year of birth on adult height.
In the next section, we give a short overview on the relevant literature, followed by a description of the children’s situation in WWII and potential eﬀects on health later in life in Section 3. Data and methods are presented in Sections 4 and 5, respectively, and in Sections 6 and 7 we give the descriptive and multivariate results. Section 8 concludes.
2.1 The importance of early-life conditions The analysis of the long-term eﬀect of adverse conditions in utero and early childhood is important for estimating the impact of e.g. economic crises on the individual. Until now, this question is not suﬃciently answered: Since Barker formulated his popular “fetal programming hypothesis”, many medical, but also economic studies address the consequences of adverse conditions at the time of birth on the birth cohort, see e. g.
Lumey, Stein and Susser (2011), Almond and Currie (2011) Lawlor (2008) or Pollitt et al. (2005) for an overview of research on early-life conditions.
Barker (1990) summarizes evidence for the “fetal origins hypothesis” and uses the maternal nutrition, birth weight and infant mortality rates to proxy the quality of the infants’ living conditions. He states that the intrauterine environment and conditions very early in life can explain regional diﬀerences in cardiovascular mortality later in life.
Kuzawa and Quinn (2009) point out that during critical periods in infant development, the body reacts to speciﬁc biological signals and changes its settings, which also aﬀects disease outcomes later in life.
Also Catalano and Bruckner (2006) test the “diminished entelechy hypothesis”, which states that adverse environment early in life, measured as high infant mortality, goes together with a decreased longevity. They test this with time series data from Sweden, Denmark, England and Wales and ﬁnd support for their hypothesis. The observed eﬀect is more signiﬁcant for males than for women.
Crimmins and Finch (2006) concentrate on infant mortality caused by infections to test the “cohort morbidity hypothesis”. Their results concerning long-term eﬀects indicate that lower mortality in infancy is positively related to low mortality throughout adulthood. Additionally, adult height can also be explained by low rates of chronic infections, since infections eat up substantial energy, which otherwise would have been used for growth. They use historical data of cohorts born up to 1900, when the infection level was high and antibiotics and childhood immunizations were not common yet. Crimmins and Finch claim that the mechanism they modelled cannot play a role in explaining later life outcomes when the mortality caused by infection in infancy is low already.
To identify a causal eﬀect of conditions at early life on later life health, one has to introduce exogenous variation to the childhood situation. Otherwise, early and late life could be jointly determined. In contrast to the majority of medical studies in the area of fetal programming, economic research often tries to tackle this problem of endogeneity.
E.g. van den Berg, Lindeboom and Portrait (2006) address the eﬀect of socio-economic conditions in early childhood, proxied by annual GDP, on individual mortality in the Netherlands and ﬁnd that being born in a recession reduces the life span compared to being born in a boom by a few years. Also van den Berg, Doblhammer-Reiter and Christensen (2008) analyse the business cycle’s eﬀect on health and focus on the share of cardiovascular mortality later in life. When surviving until age 40, otherwise identical individuals live 11 months shorter when born in a recession rather than in a boom only because of a higher risk of cardiovascular mortality.
2.2 Infant mortality as measure for health conditions
Infant mortality is popular for measuring health conditions of a society, see for example Reidpath and Allotey (2003). The IMR is deﬁned as the share of live-born infants that die before turning one year old. The reasoning behind this measure is that infant mortality mirrors health quality and rates of illness, economic development and the social and environmental situation of the whole population, since both infants and the rest of the population are determined by the same underlying factors. IMR is an important measure also in the present analysis. It is very useful, since IMR is very well documented and thus it is available even in times of few resources.
Mackenbach, Bouvier-Colle and Jougla (1990) review studies using regional mortality rates for “avoidable” mortality to relate them to health service resources, both between and within countries, between the 1950s to the 1980s. “Avoidable” here means that mortality is caused by adverse conditions avoidable by medical intervention. In the reported studies, however, the geographical variation in “avoidable” mortality is not consistently explained by diﬀerences in health care services. In contrast to Mackenbach, Bouvier-Colle and Jougla (1990), we expect that during the 1940s avoidable infant mortality still played a high role due to the broad destruction of health care facilities and the lack of trained personnel and other resources. Since penicillin was not discovered until the second world war and not introduced to the German market before 1946, infectious diseases probably were much more important in the years covered here than for the timespan observed by Mackenbach, Bouvier-Colle and Jougla (1990).
Myrskylä (2010) studies the eﬀect of shocks in mortality before age 5 on later life mortality until age 30, using data from six European countries. He ﬁnds that an increased infant mortality decreases life expectancy later in life. However, this relation holds only for England and Wales, but not for the other countries examined. Furthermore, the eﬀect is signiﬁcant only for life expectancy at ages 5-30, but vanishes for older ages.
Thus, the bad health shock early in life seems to have a transitory eﬀect only.
2.3 The relation of health and height
As aggregate proxy for adult health in general, adult height is a widely used measure.
Elo and Preston (1992) examine the relation between childhood mortality and adult mortality in a summary study and ﬁnd adult height to be the best single indicator of health conditions as child, since it reﬂects the early life conditions with respect to nutrition, diseases or stress on the very individual level. Additionally, height is very closely negatively correlated with adult mortality, suggesting that there are underlying factors in childhood determining both adult height and illnesses later in life.
In a review study, Silventoinen (2003) ﬁnds that environmental conditions determine about 20 % of adult height variation in modern Western societies. In poorer societies, this fraction is even bigger. Most important among the environmental factors are nutrition and disease load. This is conﬁrmed by Deaton (2007), who examines the environmental driving forces of height and ﬁnds on the sample basis of 43 countries that income or disease history are able to well predict the mean height in a country.
Only in African countries, where infant mortality is extremely high and positively related with adult height, the selection eﬀect of infant mortality seems to predominate the scarring eﬀect. Also Steckel (1995) ﬁnds that the environment accounts for most of the height diﬀerences across countries and that the height potential is more or less uniformly distributed among ethnic groups.
The adult height of immigrants is the focus of van den Berg et al. (2009b). Observing brothers moving to Sweden at diﬀerent ages, they ﬁnd that the timing of migration to Sweden plays an important role in determining the adult height, especially for immigrants from poorer countries that oﬀer less healthy conditions.
According to Steckel (2009), Case and Paxson (2008) and Case and Paxson (2006), also cognitive ability is possibly associated with adult height. Case and Paxson run tests during early childhood and ﬁnd that people who are tall as adults did already well as children. These ﬁndings suggest that early-life health and nutritional quality, measured by adult height, have important implications for well-being and ability in later life.
2.4 WWII and famines as health shocks Both medical and economic papers also use war as exogenous variation to estimate its eﬀect on the involved individuals. However, to our knowledge no one yet has estimated the causal relation of an adverse health shock in terms of high regional infant mortality using such a dense grid and the long-term consequences on health.
The “Dutch Hungerwinter” in 1944/1945, caused by an embargo by Germany, has been often studied as an example for adverse living conditions, see e. g. Banning (1946), Lumey, Stein and Susser (2011), and Painter, Roseboom, Bleker et al. (2005), where cardiovascular diseases, lower self-rate health, obesity and diabetes are among the consequences of malnutrition. Susser and Stein (1994) also ﬁnd a decrease in adult stature, if having been exposed to the famine shortly after birth.
Droese and Rominger (1950) assess the midterm health consequences of malnutrition for children in Schleswig-Holstein. They build upon mortality data as well as information on the children’s weight and height development and nutritional changes. Droese and Rominger (1950) ﬁnd descriptive evidence for qualitative and quantitative malnutrition in the aftermath of the war. Observed associations are decreased body growth and increased infection and infant mortality rates. Directly after the war, children from refugees were as well oﬀ as children from residents. However, in the following years their relative situation became worse. Diﬀerent from young children, older cohorts could not catch up in terms of body height and weight until 1949. The mechanisms behind the poor development for children are not identiﬁed by Droese and Rominger (1950) - intake of refugees, malnutrition or the lack of health care.
Jürges (2012) analyzes cohort eﬀects on educational attainment, occupational status and unemployment rates to assess the impact of quantitative and qualitative malnutrition, stress and other adverse conditions in the months around the end of the war.