Elsevier

Journal of Health Economics

Volume 32, Issue 6, December 2013, Pages 1013-1027
Journal of Health Economics

The long-term cognitive consequences of early childhood malnutrition: The case of famine in Ghana

https://doi.org/10.1016/j.jhealeco.2013.08.001Get rights and content

Abstract

We examine the role of early childhood health in human capital accumulation. Using a unique data set from Ghana with comprehensive information on individual, family, community, school quality characteristics and a direct measure of intelligence together with test scores, we examine the long-term cognitive effects of the 1983 famine on survivors. We show that differences in intelligence test scores can be robustly explained by the differential impact of the famine in different parts of the country and the impacts are most severe for children under two years of age during the famine. We also account for model uncertainty by using Bayesian Model Averaging.

Introduction

One of the most notable explanations for the large observed variation in cross-country economic performance has been differences in human capital; see, for example, Mankiw et al. (1992), Barro and Lee (1996), and Kalaitzidakis et al. (2001).1 Studies have also shown that health is an important determinant of human capital outcomes in developing countries; see the comprehensive survey by Bleakley (2010).2 In this paper, we are interested in one component of health that is particularly crucial to human capital formation in developing countries – early childhood nutrition – and how it affects cognitive development.3

Most of the recent work on the determinants of cognitive development has been carried out in developed countries where data has been more readily available; see, Cunha and Heckman (2009) for a comprehensive survey. This paper contributes to the existing literature by employing a unique survey data set from a sub-Saharan African country, Ghana, that includes test scores for a direct measure of intelligence or IQ (along with scores for English comprehension and mathematics) together with comprehensive information on individual, family, community, and school quality characteristics. Using this data, we exploit a natural experiment; i.e., the 1983 famine that swept across much of West Africa, to examine the long-term effects of early childhood malnutrition on the cognitive development of famine survivors who were between the ages of 0 and 8 at the time of the famine.

The health literature has yielded many examples where the natural experiment of famine has been used to suggest that early childhood malnutrition has important negative consequences for adult health. The Dutch Famine of 1944–45 has been shown to have had long-term negative impacts on various adult health (Roseboom et al., 2006), obesity (Lumey et al., 2007), and epigenetic inheritance (Lumey, 1992, Lumey and Stein, 1997) outcomes.4 Chen and Zhou (2007) and Meng and Qian (2009) report that birth cohorts during the most intense period of the Great Famine of China in 1959–1961 were significantly shorter in adulthood and were also likely to work fewer hours and earn less compared to other birth cohorts.5

The development literature has also examined the effects of early childhood malnutrition on various schooling and labor market outcomes; see, Naudeau et al. (2011a). Glewwe and Miguel (2007) provide an extensive review of the literature on the long-term impact of child health and nutrition on schooling outcomes in developing countries. They organize the factors influencing schooling outcomes in terms of the production function for academic skills and also demand-side factors. For example, poor nutrition not only reduces school attendance but also makes learning ineffective. The effects of early shocks on schooling outcomes have been demonstrated in a range of diverse contexts. For example, Neelsen and Stratmann (2011) show that the 1941–42 Greek famine had adverse effects on the cohort of children who were one year olds, infants, and fetuses at the time of famine. This group experienced significantly lower likelihood of being literate, of completing upper secondary education, and also had fewer years of schooling. The negative schooling effects of famine were also found to be stronger for urban versus rural dwellers. In the Chinese context, Meng and Qian (2009) found that children who were in utero at the time of the Great Chinese Famine of 1959–61 achieved significantly lower levels of schooling attainment. Finally, recent work by Maccini and Yang (2009) in the Indonesian context shows that early exposure to positive rainfall shocks resulted in larger positive schooling outcomes for female children.

A particular concern for researchers has been the effects of early childhood nutrition on cognitive outcomes. The timely development of cognitive functions requires sufficient intake of certain proteins and micro-nutrients like zinc and iron that are crucial for brain development (Grantham-McGregor et al., 1997). If a child does not get adequate nutrients brain development could be severely impaired. The literature has been concerned with two key questions. First, in what period of early childhood does the incidence of malnutrition lead to the most severe negative cognitive outcomes in later life? Second, are the effects of childhood malnutrition on cognitive development reversible through remedial efforts in later life?

The consensus in the literature is that cognitive abilities are established relatively early on in life – IQ, for example, is known to stabilize by about 10 years of age – and depend crucially on parental and non-parental resources. In examining the long-term effects of early childhood malnutrition, accurate determination of the critical period is crucial.6 The idea of the critical period is that some needed investments, in this case adequate nutrition and nourishment, should be made in a child's life during this period and failure to do so could result in potentially permanent negative effects. The literature suggests that the critical period for cognitive abilities is up to around 2 years of age. Belli, 1971, Belli, 1975, citing earlier works, highlighted that brain cell development is fastest within the first two years of a child and then slows down sharply afterward. Most of this growth happens within the first six months and if proteins, which are essential to brain development, are in severe shortage during this period, brain development could be sub-optimal and this could impact general intelligence (see Ivanovic et al., 2002, Ivanovic et al., 2004). Glewwe et al. (2001) with panel data from the Philippines show that children who were malnourished in their second year scored lower on IQ tests at age eight. Alderman et al. (2006) also show that the second year of life is most critical for nutritional investments in children for general health outcomes. Malnourished children from 12 to 24 months had lost about 4.6 cm in height-by-age at adolescence.

Nevertheless, there is some evidence that the impact of early childhood malnutrition on health may be partially (though not fully) reversible. For example, Pollitt (1984) suggests that early childhood nutritional shocks that impact cognitive development can be partially reversed over time if the nutritional deficiencies are corrected later in childhood. Alderman et al. (2006) explicitly examine the possibility of regaining some or all of the lost height in the aftermath of the Zimbabwean drought of 1983–84. In birth cohorts aged 12–24 months during the famine, they conclude that only about a third of the 4.6 cm in lost height is recovered through timely nutritional interventions. Importantly, Cunha and Heckman (2009) also suggest that there is a “sensitive period” between ages 6 and 8 where investments can make a large impact for cognitive abilities.

Current works in the literature on the effects of childhood malnutrition on cognitive development suffer from two weaknesses. First, even though several studies have recently examined the long-term impact of childhood malnutrition on health, there have been very few studies that have directly examined the impact on cognitive achievement. The reason for this is largely due to the lack of availability of data where direct measures of cognitive achievement (such as IQ scores) have been collected. Naudeau et al. (2011b) provides a comprehensive review of recent research into the patterns of cognitive development in selected developing countries. Instead, researchers have focused on other outcome measures that are only indirectly related to cognition, such as general measures of health or physical development (e.g., height or height adjusted for age), schooling attainment, performance on tests, or various labor market outcomes (e.g., wages or hours worked). Second, when direct measures of IQ scores have been available (such as in the important work of Glewwe et al., 2001), these scores have been available only for relatively young children. Researchers are therefore not able to definitively answer the question of whether negative impacts on cognitive achievement due to early childhood malnutrition persist into adulthood.

An important exception is Stein et al. (1972). Stein et al. studied the effects of the 1944–45 Dutch Famine on children who were born within 1 year of the famine or who were conceived during the famine (but born after). Their main interest was to evaluate whether there would be significant differences in cognitive outcomes – such as clinically diagnosed measures of severe and mild mental retardation, and also IQ (as measured by scores on a Raven Progressive Matrices test) – between the intrauterine birth cohorts that experienced famine (those who lived in the large cities of Western Holland) through maternal exposure and those that did not (those who lived in cities in the south, east, and north of Holland) by the time the surviving offspring had reached adulthood (age 197). Stein et al. also control for socioeconomic status of the child's family using father's occupation; i.e., whether the father was doing manual or non-manual labor.

Their surprising conclusion was that neither starvation during pregnancy nor early childhood malnutrition appears to have detectable effects on the adult mental performance of surviving male offspring. Stein et al. provide a detailed critique of their methodology and suggest two alternative hypotheses: (1) “selective survival”; that is, only fetuses that were unimpaired by the nutritional deprivations of famine survived, and (2) “compensatory experience”; that is, postnatal education in the period from birth until the time when the individuals were sampled (at military induction) may have (completely, in this case) reversed any early cognitive effects of famine experience. If the compensatory experience hypothesis was true, and the negative cognitive effects of early childhood malnutrition could, in fact, be fully compensated for by subsequent investments, then it would invalidate the “critical period” hypothesis and suggest that more emphasis be placed on establishing the “sensitive period” for childhood investments.

In this paper, we examine the long-term effects of childhood malnutrition that was the consequence of a severe famine in 1983–84 in Ghana on cognitive development in adults 20 years later. In 1982–83 severe droughts and subsequent food shortages plagued most African countries. For example, in 1983, maize, a major food staple, saw a 50% drop in production from the previous year.8 In all, there was a food deficit of 361,000 tons and a request was made to the Food and Agriculture Organization (FAO) for assistance much of which was not delivered until late 1984. According to Derrick (1984), there was a significant drop in daily per capita caloric intake to about 1600 kcal in 1983 from 1900 kcal in 1982. Derrick (1984) reports that drought-prone northern Ghana (collectively composed of the Northern, Upper East, and Upper West regions), which is mostly rural, was the most affected together with the food-growing areas. The lowest calorie intake was experienced in 1984 for most areas in Ghana. Thus it is expected that birth cohorts within this window should be worst affected by the famine in 1983–84. While we do not have data on caloric intake by regions, the variation in under-five mortality rate deviations from trend across regions, our measure of famine intensity, is wide; see Fig. A1 of the Appendix.

Our work differs from the seminal work by Stein et al. in the following ways. First, Stein et al. focused on children between the ages of 0 and 1 years during the famine because they were primarily concerned with investigating the effects of famine on intrauterine birth cohorts. We focus instead on the question of the effects of famine during early childhood malnutrition on adult cognitive outcomes. Consistent with the literature cited above, we define early childhood as children aged 0–2.9 We are therefore naturally interested in the question of whether children who experienced famine when they were younger (in the 0–2 age group) as opposed to when they were older (the 3–8 age groups, in our case) saw differential impacts in the effects of childhood experience with famine. That is, our study focuses on the long-term cognitive outcomes of children within 2 years of age in 1984 compared with older children (up to 8 years old) at the time of famine.

Stein et al. also only focus on famine incidence (i.e., the 7 cities that experienced famine in their treatment group versus the 11 cities that did not in their control group), whereas we consider the variation in famine intensity across the 10 administrative regions of Ghana. Like Stein et al., but unlike most previous studies on this subject, we exploit a unique survey data set from Ghana – the Ghana Education Impact Evaluation Survey (GEIES) in 2003 – that directly measures intelligence or IQ (based on the Raven's Progressive Matrices10) – in addition to scores on tests for English comprehension and mathematics – that was administered on adults who had experienced varying degrees of famine intensity as children in 1983–84 20 years earlier, to examine the impact of early childhood malnutrition on adult cognitive development. Raven is a generally accepted means of measuring general intelligence as it does not depend on crystalized information typically acquired from school (Carpenter et al., 1990).

Further, unlike Stein et al., the data from Ghana makes it possible for us to control for a large number of individual, family, and community characteristics (and not just family socioeconomic status). Importantly, we are able to control for the cumulative effects of childhood investments in health that can confound the direct effects of the famine on adult cognitive development. Specifically, we use height to proxy for accumulated health status. The data also allows us to control for key schooling quality characteristics such as the quality of schooling infrastructure; i.e., the state of classrooms and the availability of textbooks, and the quality of teachers. We are also able to control for the socio-economic status of the family using parental schooling data. Hence, we are able to investigate the effects of early childhood malnutrition (during the critical period of 0–2 years) on long-term cognitive development after controlling for possible subsequent remedial interventions that fall specifically during the sensitive period of a child's development before her IQ stabilizes (at age 10).

Finally, we also make a methodological contribution. In contrast to previous work in this literature, we explicitly address the issue of model uncertainty in investigating the long-term effects of famine. The term model uncertainty was first coined by Brock and Durlauf (2001) in the empirical growth context to refer to the idea that new growth theories are open-ended, which means that any given theory of growth does not logically exclude other theories from also being relevant. In our context, model uncertainty implies that the role of early childhood malnutrition in determining IQ does not automatically preclude any of a large number of other possible determinants related to, for example, either nutritional or schooling investments in the sensitive period from being included in the analysis. However, the estimated partial effect of early childhood malnutrition on IQ may vary dramatically across model specifications depending on which other auxiliary variables are included in the regression. How should one deal with the dependence of inference on model specifications?

To do so, we employ Bayesian model averaging (BMA) methods; see, Leamer (1978), Draper (1995), and Raftery (1995) that have been widely applied in other areas of economics, but are novel to this literature. BMA constructs estimates that do not depend on a particular model specification but rather use information from all candidate models. Specifically, it amounts to forming a weighted average of model specific estimates where the weights are given by the posterior model probabilities. In particular, we implement BMA in both the linear regression context as well as in the structural context. In the latter case, we use data on regional rainfall variations as an instrument for the degree of severity of famine.

Our main findings are as follows. First, we find that, all else equal, famine intensity only affects the cognitive development of children who were in the 0–2 years of age group at the time of famine. The children in the older age group, 3–8, suffered no direct effects from the famine. Second, after controlling for a large set of characteristics including accumulated health, we find that the magnitude of the effect of famine intensity on cognitive development in children who experienced famine between ages 0 and 2 is large. For a standard deviation increase in our famine intensity measure, measured IQ falls on average by about 6% for children in this age group. In terms of performance on Math and English tests, this loss of cognitive ability translates on average to a loss that is consistent with a reduction of about two-fifths to a half of a year of schooling. Overall, our work suggests that early childhood malnutrition has a large and important direct impact on cognitive performance that persists into adulthood. But, the incidence of the malnutrition needs to be early enough for this effect to take hold.

We proceed as follows: Section 2 describes the empirical strategy and data. We then discuss the results in Section 3. Finally, Section 4 concludes.

Section snippets

Empirical strategy and data

Following Behrman and Lavy (1994) and Glewwe et al. (2001), we exploit the differences in famine intensity across Ghana to examine the impact of famine and the resulting malnutrition on survivors. We match data from several sources for the estimation problem at hand. The main data set is the GEIES of 2003 and its precursor the education module of the Ghana Living Standards Survey II of 1988/89. We also use data from the Demographic and Health Survey (DHS) of 1988 and rainfall data from the

Ordinary least squares (LS) and 2SLS results for Raven (IQ) scores

We now turn to a discussion of the results. We first present our findings for IQ (i.e., Raven scores) in Table 2. We start with a discussion of our least squares estimation results. Column (1) of Table 2 presents the LS results for the largest model in the model space (i.e., the “kitchen sink” model in Section 2.1). Our least squares results provide strong evidence for the hypothesis that early childhood malnutrition has an important and significant negative impact on cognitive development. In

Conclusion

In this paper, we investigate the impact of early childhood (children between 0 and 2 years of age) malnutrition resulting from widespread famine in Ghana on cognitive development. A novel feature of our analysis is that we explicitly control for model uncertainty in our estimation. We find a direct, negative, and significant impact of early childhood malnutrition on the cognitive development of famine survivors. These effects persist well into adolescence and adulthood. In turn, this loss of

References (60)

  • H. Alderman et al.

    Long term consequences of early childhood malnutrition

    Oxford Economic Papers

    (2006)
  • M.C. Auld et al.

    Schooling, cognitive ability and health

    Health Economics

    (2005)
  • R.L. Babcock

    Analysis of adult age differences on the Raven's advanced progressive matrices test

    Psychology and Aging

    (1994)
  • R.J. Barro et al.

    International measures of schooling years and schooling quality

    American Economic Review

    (1996)
  • J.R.J. Behrman et al.

    Children's Health and Achievement in School

    (1994)
  • P. Belli

    The economic implications of malnutrition: the dismal science revisited

    Economic Development and Cultural Change

    (1971)
  • P. Belli

    The economic implications of malnutrition: reply

    Economic Development and Cultural Change

    (1975)
  • H. Bleakley

    Health, human capital, and development

    Annual Review of Economics

    (2010)
  • W.A. Brock et al.

    Growth empirics and reality

    World Bank Economic Review

    (2001)
  • P.A. Carpenter et al.

    What one intelligence test measures: a theoretical account of the processing in the Raven progressive matrices test

    Psychological Review

    (1990)
  • E. Cohen-Cole et al.

    Model uncertainty and the deterrent effect of capital punishment

    American Law and Economics Review

    (2009)
  • F. Cunha et al.

    Human capital formation in childhood and adolescence

    CESifo DICE Report, Ifo Institute for Economic Research at the University of Munich

    (2009)
  • F. Cunha et al.

    Formulating, identifying and estimating the technology of cognitive and noncognitive skill formation

    Journal of Human Resources

    (2008)
  • F. Cunha et al.

    Estimating the technology of cognitive and noncognitive skill formation

    Econometrica

    (2010)
  • J. Derrick

    West Africa's worst year of famine

    African Affairs

    (1984)
  • D. Draper

    Assessment and propagation of model uncertainty

    Journal of the Royal Statistical Society, Series B

    (1995)
  • S.N. Durlauf et al.

    Are any growth theories robust?

    Economic Journal

    (2008)
  • S.N. Durlauf et al.

    Is god in the details?. A reexamination of the role of religion in economic growth

    Journal of Applied Econometrics

    (2012)
  • T. Eicher et al.

    Bayesian model averaging and endogeneity under model uncertainty: an application to development determinants

    Econometric Reviews

    (2013)
  • C. Fernandez et al.

    Model uncertainty in cross-country growth regressions

    Journal of Applied Econometrics

    (2001)
  • Cited by (69)

    • Predicting undernutrition among elementary schoolchildren in the Philippines using machine learning algorithms

      2022, Nutrition
      Citation Excerpt :

      Alarmingly, the prevalence of stunted growth has increased in recent years, exacerbated by persistent poverty, natural hazards, and limited access to health care services [5]. Undernutrition increases the risk of morbidity and mortality, lowers short- and long-term immunity [6], stymies cognitive development [7], and is associated with decreased labor-force productivity and lifelong earnings [8], all of which impede a country's development agenda. Early intervention can prevent and mitigate the effects of childhood undernutrition.

    View all citing articles on Scopus

    We would like to thank Sofronis Clerides, Martin Flodén, Jacqueline Geoghegan, Wayne Gray, Yannis Ioannides, Andros Kourtellos, Masayuki Kudamatsu, Anna Larsson-Seim, Mårten Palme, David Seim, Enrico Spolaore, Jonas Vlachos, Yves Zenou, and Xiaobo Zhang for their many valuable comments and suggestions. All errors and omissions are our own.

    View full text