Causal Effects Of Early-Life Conditions
Empirical Approaches And Empirical Findings
For expositional reasons, this section begins with a subsection on the methodological approaches used in the empirical literature to detect long-run effects of early-life conditions. This includes a discussion of empirical findings that capture the overall causal effect. The overall effect can be a direct causal effect or it can be the result of a causal pathway that involves intermediate events during life. Section Direct and Indirect Long-Run Effects discusses the difference between direct and indirect effects in more detail. Section Indirect Effects: Causal Pathways from Early Childhood by Way of Education to LaterLife Morbidity and Mortality discusses empirical studies of indirect effects that include data information on events occurring along the pathway of interest.
A natural starting point to analyze whether early-life conditions are important is to compare health and mortality outcomes among elderly individuals who faced different living conditions early in life. Empirical studies have shown that adverse socioeconomic conditions early in life are associated with susceptibility to a wide range of health problems later in life. Similarly, medical studies have shown that individuals with a low birth weight (sometimes adjusted for gestation time) are more likely to suffer from health problems later in life.
Observed associations do not necessarily imply the presence of causal effects of early-life conditions. Individual socioeconomic and medical conditions during early childhood and health outcomes later in life may be jointly affected by unobserved heterogeneity. For example, certain genes may simultaneously influence the average level of the parents’ income, the birth weight, and the health outcomes later in life. To be able to detect causal effects, one needs to observe exogenous variation in the early-life conditions, and relate this to outcomes later in life. In all fairness, it should be noted that even if descriptive studies do not capture causal effects, they are still useful from an intervention point of view. Markers for unfavorable future health outcomes can be used as a flag for monitoring or initiating interventions to mitigate such outcomes.
A recent approach has recently become popular to detect causal effects, by using data on indicators Z of individual conditions X early in life with the following property: the only way in which the indicator Z can plausibly affect high-age morbidity or mortality Y is by way of the individual early-life conditions X. (An extreme example is where Z is the outcome of a lottery in which individuals with a baby may win some money. More common examples are given below.) By analogy to the econometrics literature, such indicators Z may be called instrumental variables. Typically, these are not unique characteristics of the newborn individual, his/her family, or household, but rather temporary characteristics of the macroenvironment into which the child is born. In that case they are also called contextual variables. Indicators Z with the above ‘exclusion restriction’ property do not give rise to endogeneity and simultaneity biases, because they are exogenous from the individual’s point of view. Moreover, they do not have direct causal effects on health later in life except through early-life conditions. If one observes an association between such an indicator Z and the health outcome Y later in life, then one can conclude that there is a causal effect of early-life conditions X on that health outcome Y.
In the current context, three types of such ‘instrumental variables’ Z may be distinguished. First is the season of birth. The idea is that the month of birth has no other effect on health outcomes later in life than by way of the early-life conditions of the child. Note that this requires that the composition of newborns is not systematically different across seasons, in terms of unobserved characteristics of the newborns. The literature has typically found significant effects of the season of birth on the mortality rate later in life, with an order of magnitude of a few months of extra lifetime if one is born in the fall, as compared with the late spring. In the southern hemisphere, these effects are mirror-imaged, in the sense that the effect of a month of birth is similar to the effect of the month half a year earlier or later in the other hemisphere. In equatorial areas, seasonal effects are in accordance to what constitutes the rainy (monsoon) and the dry season.
A second type of exogenous variation is provided by epidemics, wars, famines, and other disastrous events. Lumey et al. (2011) provide an excellent overview. For a recent example, see Lindeboom et al. (2010), who examine whether exposure to nutritional shocks early in life affects later-life mortality. They use historical data that include the period of 1845–48, which includes the Dutch potato famine. During this period, potato crops failed due to the Potato Blight disease and bad weather conditions. They found strong evidence for long-run effects of exposure to the Potato famine. The results were stronger for boys than girls and lower social classes appeared to be more affected than higher social classes. Studies based on the Dutch ‘hunger winter’ under German occupation at the end of World War II and on China’s great famine indicated significant long-run effects on adult morbidity, but not on adult mortality. These studies confirmed that malnutrition has a separate effect on adult morbidity (and sometimes) mortality. Experimental animal research has also provided support for the theory that there are long-run effects of malnutrition during pregnancy.
Almond (2002) examines individuals born around the time of the 1918 influenza epidemic. He finds significant effects on the mortality rate later in life, and this finding has been confirmed by subsequent studies using epidemics. Similar to many of these studies, Almond investigates primarily the sign and significance of the mortality-rate differences between birth cohorts, and not the exact size of the effect. This is because the interest ultimately is not in the size of the effect of the indicator Z on the mortality rate, but in the issue of whether there is a causal effect from earlylife conditions X on the mortality rate. Long-run effects may, of course, be nonlinear in terms of early-life conditions. In that case, the relevance of long-run effects of disastrous conditions may be limited, and may not lead to a full understanding of the effects of less spectacular variation in early-life conditions.
A third approach was pioneered by Bengtsson and Lindstro¨ m (2000). They use the transitory component (or deviation) in the price of rye around the time of birth as an indicator of food accessibility early in life – any observed relation between this indicator and the mortality rate later in life signifies the existence of a long-run causal effect of food accessibility on mortality later in life. Similarly, the transitory component in the local infant mortality rate was used as an indicator of exposure to diseases early in life. This study uses data from a relatively small area in Sweden from the eighteenth and nineteenth centuries. The results indicate that individuals born in years with epidemics lived on average a few years less than otherwise, conditional on surviving the epidemic itself. Van den Berg et al. (2006) use the state of the business cycle at early ages as a determinant of individual mortality. Cyclical macroeconomic conditions during the pregnancy of the mother and childhood might affect mortality later in life because they are unanticipated and affect household income. In a recession, the provision of sufficient nutrients and good living conditions for children and pregnant women may be hampered. Van den Berg et al. (2006) find that the average lifetime duration in the Netherlands in the nineteenth century was reduced by approximately 1–3 years if the individual is born in a recession, as compared with having been born in a boom (under otherwise identical conditions during life, and conditional on surviving early childhood). Van den Berg et al. (2011) find analogous effects on cardiovascular mortality, using Danish data.
One important requirement for the analysis of causal longrun effects of early-life conditions is that the individual data cover a sufficiently long time span. After all, the dates of birth and death (or high-age health) must be observed for a substantial number of individuals. An implication of this requirement is that the existing studies have necessarily considered cohorts of individuals who were born a long time ago. In this sense, the most recent evidence comes from studies of individuals born in the Dutch hunger winter (1944–45) and from studies of more recent birth cohorts from developing countries. One way to circumvent this restriction would be to focus on adult health proxies such as adult height (see the upcoming sections).
Direct And Indirect Long-Run Effects
Empirical Approaches and Empirical Findings listed studies that use exogenous variation in the environment to show that there are causal effects from early childhood on later-life morbidity and mortality. The present subsection briefly sets out the main mechanisms underlying these long-term causal effects. Although there are many ways in which early-life conditions may affect outcomes later in life, it can be distinguished roughly between two main views.
First, adverse prenatal and post-neonatal (from birth to 12 months) conditions can have a direct effect on later-life morbidity and mortality. The main idea is that the development of vital organs and the immune system is programmed when the body is exposed prenatally or just after birth to adverse conditions. According to the ‘developmental programming’ or ‘fetal origins’ hypothesis), this may lead to increased vulnerability to chronic diseases in later life. The most commonly mentioned factors mentioned in the literature are malnutrition and exposure to infectious diseases. Other factors are increased stress in the household and lower income to cover housing accommodation costs. Most of the empirical studies mentioned in this section are consistent with a direct effect. As it can be seen, in order to detect long-run effects, it is natural to focus on temporary shocks around the birth date. Any long-run effect found in this way could be a direct effect. Moreover, the estimated size of the mortality effects is usually moderate and in line with the medical evidence. The type of shock is informative regarding whether the effect concerns malnutrition, disease exposure, other adverse conditions, or just bad conditions in general.
Exposure to infectious diseases and malnutrition is likely to be less relevant for the developed world today than it was in the past. However, Bozzoli et al. (2009) recently examined the effect of income and disease exposure on adult height in populations, where height is used as a proxy for lifetime health. They use post-neonatal mortality as a measure for nutrition and disease load in early childhood and examine their effect on height for cohorts born from 1950 to 1980 in the US and 11 European countries. They find a strong negative relationship between adult height and the burden of disease and malnutrition.
According to the second main view, adverse conditions early in life have indirect effects in that they may be the start of a causal chain of events or pathways during life that leads to worse health later in life. For instance, poor early-life conditions may lead to poor health early in life and later in childhood, which may affect educational outcomes and subsequently social status and health in adulthood. Or, more generally, a poor start may affect an individual’s life career, which may ultimately lead to higher mortality rates. The authors discuss this view in more detail below, but before that it is good to note that some studies have stressed that it is the interaction with social factors later in life that determines whether people who are exposed to adverse early-childhood conditions will be more vulnerable to ill health in later life. For example, among individuals born in recessions, the decline in mental fitness after experiencing a negative life event at high ages (such as a stroke, surgery, illness, or death of a family member) is worse. Among women, marriage leads to increased mortality in child-bearing ages, but this increase is smaller if the woman was born under favorable economic conditions around birth, as captured by the business cycle early in life. In a similar vein, the body accommodates to stress, and that it is repeated stress that leads to higher risks of chronic diseases.