Journal Issue: Children with Disabilities Volume 22 Number 1 Spring 2012
Long-Term Economic Consequences of Childhood Health Problems
Recent studies have examined the role of environmental shocks such as famines and toxins in affecting early childhood and gestational environments and subsequently influencing the path of development and adult health. A large body of work, reviewed by Tessa Roseboom and others,23 examined the effects of growing up during the 1944 Dutch famine on later patterns of heart disease; Douglas Almond and others used a 1 percent sample of China's 2000 census and found a range of adverse economic effects on Chinese exposed to the famine of 1958–61.24
The main advantage of these approaches, as argued by their proponents, is that they provide an opportunity to isolate causal effects of early-life conditions. However, the extent to which the shocks that were examined can be seen as representative of other types of shocks to childhood health remains unknown (see Angus Deaton25). It is important to examine the extent to which shocks generate unintended selection effects, such as differential fertility or mortality, that change the measured average health of surviving babies. However, given the rapid decline of infant mortality in America, and the currently low levels, selection effects due to unhealthy children dying in early childhood are less of an issue.
One common technique scientists use to disentangle cause and effect is to observe "natural experiments," that is, specific changes taking place in nature, and then study the impact of these changes on other aspects of life. Recent natural experiments on the impacts of early-life conditions stem from Robert Barker's hypothesis that foundations for chronic illness in later life are laid in the uterus. Barker argued that stress to the fetus during pregnancy leads to the diversion of resources to protect the brain at the expense of other organs, weakening these organs and predisposing the fetus to later patterns of disease.26 In line with this, several studies found that birth weight (often used as a proxy for the uterine environment) was a predictor of health in later life.27
Caleb Finch and Eileen Crimmins argued in a 2004 article that much of the improvement in adult health over the centuries came about because of reduced exposure to early-life stresses.28 They provided evidence that declining infant mortality in Sweden predicted mortality declines among adults in the same group. They suggested that lower risk of gastroenteritis among infants might lead to lower risk of inflammation in later life. Population studies have also examined whether early physical health adversity affects economic circumstances later on. This research follows individuals from childhood to adulthood or supplements existing studies that do follow individuals over time with data that are missing from those studies. The two mainstays of this research are British studies following individuals from the week of their birth, and long-term American studies.
Using the data from the PSID, Smith found that childhood health bears on a range of adult economic variables including levels and trajectories of family income, household wealth, individual earnings, and volume of work that are robust to controlling for personal attributes that are observed in the data and those that are unobserved.29 Anne Case, Darren Lubotsky, and Christina Paxson found that respondents to the United Kingdom National Child Development Study (NCDS) who had low birth weight and poor childhood health experienced later problems that included lower school and occupational attainment. In another paper, Case and Paxson indicated that childhood health (proxied by height) is associated with many positive life outcomes, only some of which are related to education.
One weakness in examining the effects of childhood illnesses on later health and economic status in America is the lack of data that track people from early life through adulthood. Individual life histories have become a useful tool in examining the effects of early conditions on adult health in panel studies, which follow the same respondents over time.30 Life histories ask respondents to recall important information about their early lives, including general childhood health, health care utilization, and onset and duration of childhood illnesses. Visual and verbal memory cues prompt respondents to remember this information.
Several major studies have employed life-reconstruction data, including the Health and Retirement Study (HRS), the English Longitudinal Study of Ageing, and the Survey of Health, Aging, and Retirement in Europe. (See the data appendix to this volume for additional information about these data sets.) This technique enables researchers to extract relevant information from recent large-scale panel studies that did not interview respondents as children. Using data from HRS and PSID, Smith found that patterns of recalled childhood illnesses closely matched information about illness during respondents' childhoods as measured in the contemporaneously collected American national health surveys, such as NHANES and NHIS.31 The recalled measures of childhood illness act as important predictors of later patterns of illness using these samples.32 On the physical health side, Katayoun Bahadori and others reviewed sixty-eight studies and found evidence pointing to associations between asthma, poorer schooling outcomes, and lower future earnings.33
Lifetime Effects of Childhood Mental Illness
Given the increasing prevalence of mental health problems among young children, the role of childhood mental illness is increasingly important. Janet Currie and others found significant effects of childhood mental health problems.34 They used data based on public health insurance records of 50,000 children born between 1979 and 1987 in Manitoba, Canada. Their design allowed them to compare siblings with noncongenital health problems. They reported that, although childhood physical health problems often lead to future health problems, childhood mental health problems produce significant effects that are not dependent on future physical health problems.
James Smith and Gillian Smith used the retrospective PSID health data to uncover substantial effects of recalled childhood depression on future economic well-being.35 Their estimations showed substantial reductions in income largely caused by a reduction in weeks worked per year. Respondents who reported childhood mental problems also had lower educational attainment, although this effect was small relative to the impact on income. The authors estimated that the family of each affected individual lost about $300,000 over a lifetime, on a discounted net value basis. The corresponding cost to the current American population would be $2.1 trillion. Note that this cost is larger than the annual costs calculated by Mark Stabile and Sara Allin in this volume, in part because it reflects the present discounted value of costs that would be accrued over a lifetime. Currie and Stabile used Canadian data to examine the long-term effects of ADHD, a common form of mental illness among young children. Controlling for confounding factors, they found that the effects of ADHD are much greater than those of physical health problems.36 They reported reductions in future reading and mathematics test scores and increased probability of future grade repetition.
A New Look at the Effects of Childhood Health
Using the retrospective PSID childhood health module, we present new estimates of the impacts of being in excellent or very good health as a child on the economic and health aspects of later life as an adult. Our analyses examine the association of both dimensions of childhood health—physical and mental—with salient aspects of adult life: health, education, work, and income.
In the first column of table 4, we summarize the impacts of having any one of the childhood physical health problems (with the exception of measles, mumps, and chicken pox) and the effect of having any childhood mental health problem on whether a person's self-reported adult health in 2007 was excellent or very good. In the second column of table 4, childhood physical health is separated into its component parts, although some subcomponents are aggregated either because of low prevalence or because of the commonality of the size of the effects of the childhood health problem on adult life. The asterisks in tables 4–6 indicate the likelihood that the effect is statistically different from zero (or no effect), with three asterisks indicating one in a hundred, two asterisks indicating one in twenty, and one asterisk indicating a one in ten chance that there is really no effect.
The two models on the left side of the table are estimated using ordinary least squares (OLS), the most widely used statistical way of showing the relationship of one variable to another, conditional on other variables. All the variation used in the OLS model to estimate an average effect represents variation across people who were originally children in the PSID. In contrast, the two models on the right side of table 4 focus on comparing siblings within the same families (often called sibling models). All of the variation used in sibling models represents differences across siblings in the same family. The sibling models are preferred because they account for all of the common background characteristics shared by siblings (their family, neighborhoods, and schools) whether or not they can be measured in our models. All full age-range models include controls for being in age groups twenty-one through forty and forty-one through sixty; the over-sixty age group is excluded.
The models in table 4 show the association between childhood physical and mental health problems on the probability of being in either excellent or very good health as an adult. These indicate that childhood mental health problems have larger impacts on self-reported adult health than do childhood physical health problems, although the two sets of estimates are close in the sibling models. Using the sibling models, this implies that individuals who had a mental health problem as a child or those who had a physical health problem as a child are 10 percentage points less likely to be in excellent or very good health as an adult.
When the childhood physical health problems are separated into the specific childhood physical health problems, the negative effects on adult health are somewhat larger for hypertension, sight problems, asthma, epilepsy, and diabetes. For some childhood physical health problems, there are strong selection effects whereby children in better-off families are more likely to get a particular disease. A good illustration involves the common infectious diseases, where the across-person estimate is statistically significant and positive. In contrast, the across-sibling estimate is small and statistically insignificant. Table 3 suggests that children in better-off families are more susceptible to these common infectious diseases. This selection effect explains why the estimated effect of having these infectious diseases as a child is positively associated with better health as an adult in the across-person models. This example suggests that some caution is in order when interpreting across-person estimates in various studies in the literature.
Table 5 lists our estimates of the average effects of mental and physical childhood health problems (without the childhood infectious diseases) for adult socioeconomic status, including years of schooling, the number of weeks worked in a year, percentage change in earnings, and percentage change in family income. For the number of years of schooling—the most common adult socioeconomic factor examined in the literature—the across-person estimates suggest that the damage done on adult life is much larger for mental health problems (a loss of 0.8 of a year of schooling) than for physical health problems (where the estimated effect is actually positive). In the preferred across-sibling models, the impact of childhood mental health problems remains statistically significant—a reduction of about a half-year in schooling—but the impact of childhood physical health problems is insignificant. Mental health problems as a child appear to be much more important than physical health problems during childhood on limiting educational opportunities.
For the number of weeks worked in a year, we find the same relationship: childhood mental health problems are much more important than childhood physical health problems. In fact, the impacts of childhood mental illness are about three times greater on the number of weeks worked than those for childhood physical illnesses. The preferred across-sibling model indicates almost seven fewer weeks worked yearly by those who had childhood mental health problems. When we examine the percentage change in adult labor market earnings, estimated impacts again are much larger and more statistically significant for childhood mental health problems than for childhood physical health problems.
Our preferred and most general economic outcome is the percentage change in family income. Using across-sibling models, the estimate for children's mental health problems suggests a 37 percent lower family income— a decline that is three times greater than the estimated impact for a childhood physical health problem.
All of these adult socioeconomic models point to the same conclusion: childhood mental health problems have much larger effects on later adult life than childhood physical health problems.
An important issue not addressed by table 5 concerns how these effects of childhood health problems vary by age. Using the preferred across-sibling models, table 6 provides separate estimates for two age groups, twenty-one through forty and forty-one through sixty.
These estimates across age groups for a single calendar year (2007) could be interpreted either as effects associated with an individual becoming older (aging effects) or as effects associated with individuals being born in different calendar years (birth-cohort effects). There is no way to identify separate birth-cohort or aging effects with data in a single calendar year because an older person must necessarily have been born in an earlier calendar year. But physical improvements in workplace disability accommodations over time may have made physical health problems less limiting over time, especially for younger persons. The extent of accommodation may be much smaller for mental health problems. In fact, the impact of childhood mental health may have increased over time as the U.S. economy increasingly values mental and academic skills over physical skills. Whichever interpretation is preferred, estimates for these socioeconomic outcomes indicate that the effects of childhood mental problems are somewhat smaller in the older age group for all such outcomes, while the childhood physical health outcomes become slightly larger in the older age adult group.37