Pediatrician Perri Klass wrote an article in the New York Times on May 13, 2013, that focused on the growing interest by pediatricians and other medical professionals in child poverty as a national problem. She argues that evidence suggests that child poverty damages child development and hence the child’s long-run potential as an adult. Poverty may damage child development through producing many stresses on children. Such stresses include frequent housing moves, inadequate nutrition, too little access to health care, and family conflicts.
There is some good research by economists that tries to quantify the relationship between child poverty and future adult outcomes for the child. This research focuses on what child poverty means for the child’s future earnings. For example, there is good recent work by Greg Duncan and his colleagues that looks at effects of childhood poverty on future adult earnings, and on the child’s academic achievement, which will predict future earnings.
Duncan et al.’s work finds that added parental income makes a bigger difference for a child’s future adult earnings for children from low-income families. Duncan et al. also finds that anti-poverty experiments that boost family income appear to boost a child’s academic performance.
Of particular interest to early childhood advocates is the finding that a child’s future adult earnings are more affected by the income of the child’s parents when the child is ages 0-5, than by income in later childhood years. Once one controls for family income in early childhood, the family’s income when the child is ages 6 through 15 has little impact on the child’s future adult earnings.
The special importance of early childhood family income supports the notion that early childhood is a critical developmental period. During early childhood, the economic, social, and educational environment experienced by the child seems especially likely to lead to large permanent changes in life course.
Most of these effects of early childhood family income on the child’s future adult earnings occur due to effects on annual adult work hours. This could be explained if early childhood experiences have more profound effects on future behavior and work habits (“soft skills”) than on more narrowly defined hard skills.
Duncan et al.’s work can be extrapolated to calculate how much the present value of the child’s adult earnings increases for a given increase in parental income. Depending upon what estimates are used and what assumptions are made, the expected present value of the child’s future adult earnings increases somewhere between 1.27 to 2.88 times the boost to the parent’s income. (See below for the gory details on how these calculations are made.)
To boost parental income by $1 may cost the government more than $1 if done through some income transfers (Bos et al., 2007), but could cost less than $1 if done through wage subsidies that encourage work (Eissa and Hoynes, 2011). Therefore, if we were evaluating some government policy to boost children’s future earnings by boosting the income of their parents, the ratio of the child’s future earnings gains to the government’s costs could be less than 1.27 or more than 2.88, depending upon the efficiency of the policy and the assumptions made.
In my work on early childhood programs and local economic development, boosts to future per capita earnings are my definition of “economic development benefits”. (The rationale for this definition is that such boosts to per capita earnings are the main social benefit from the incentive programs we label economic development programs.) Suppose a government income transfer program for parents was perfectly “efficient” – that is, it boosted parental income during early childhood by a dollar per dollar of government costs. Under that assumption, this parental income transfer program would have effects on child development that would yield a ratio of national economic development benefits to government costs of somewhere between 1.27 and 2.88.
Therefore, boosting parental income also deserves policy attention as an early childhood strategy. Investing in helping parents while their children are in early childhood will boost national economic development by boosting the child’s future earnings per capita. Such policy proposals may have net benefits even if evaluated solely on the basis of their benefits for the child’s future earnings. This omits other possible benefits of boosting parental income, most obviously the benefits for the parents.
Such a strategy reinforces the need to expand early childhood programs. Duncan’s work suggests greater child development benefits from boosts to parental income when children are 5 years old or less. How can strategies to help parents target parents with children in that age range? One obvious way to do so is by providing extra subsidies and support for child care, preschool, and parenting support for families with children in this age range, who also have the greatest need for such services.
Early childhood programs that are targeted at the disadvantaged can be seen as a two-generation anti-poverty strategy that boosts early childhood development both by boosting parental real incomes and by providing educational services to children.
(Notes to hard-core policy wonks on how calculations were made. I took Duncan/Ziol-Guest/Kalil (DZK) estimates from Table 2 on effects of early childhood income on adult earnings. They estimate that an extra $10,000 in family income per year from ages 0 to 5 will increase the natural log of future income by 0.52. I calculated the present value of the future earnings of children from poverty backgrounds to be 52% of the average earnings for all persons, based on Appendix A in DZK. I used data from Chapter 12 of my book Investing in Kids to calculate that the present value of future adult earnings for a typical person, in 2007 dollars, using a 3% real discount rate, and evaluated when the person is age 4, is $863,000. This assumes annual secular real wage increases in the national economy of 1.2%. Multiplying this present value by the assumption that persons from poverty backgrounds will earn 52% as much as the average person, yields a future present values of earnings for persons from a poverty background of $432,000. The dollar value of the change in future adult earnings was calculated by adding and subtracting 0.26 from the natural logarithm of future adult earnings, to reflect a 0.52 change in log earnings around the mean. This was then compared with the present value of adding $10,000 in family income for each of the 7 years from the prenatal period to age 5, converted from the 2005 dollars used by DZK to 2007 dollars, and discounted also to age 4. The result is that the ratio of the increased present value of the child’s future earnings to the present value of the shock to parental income is 2.88.
For the calculations using changes in children’s test scores, I used the Duncan/Morris/Rodrigues (DMR) estimates in Table 5 that an extra $1000 in parent’s annual income increased a child’s achievement by 0.06 standard deviations. I used Chetty et al.’s estimates, as explained in Bartik/Gormley/Adelstein, that a 1 percentile boost in student achievement increase will increase future adult earnings by 0.495%. Based on Reardon, I assume that the low-income child’s typical achievement gap from the median child will be 0.50 standard deviations. An improvement from 0.50 standard deviations below the mean to 0.44 standard deviations below the mean boosts achievement in a standard normal distribution by 2.14 percentiles. This percentile boost was multiplied by the 0.495% boost in earnings per percentile and then by the present values of adult earnings of low-income children that were assumed in the previous paragraph. Finally, the experimental data used in DMR seem to typically observe the effects of income boosts at the end of around 3 years. I assumed this period was from ages 2 to 4 for the child, and calculated the present value of this income boost as of age 4. I also adjusted the income boost from the 2001 dollars used by DMR to 2007 dollars, to be comparable with the adult earnings data. The resulting ratio of the boost in the present value of the child’s future adult earnings, to the boost in the present value of the parent’s income, is 1.27.
The latter estimates might be lower for three reasons. First, the DMR estimates include boosts in parental income when the child is older, which is shown in the DZK paper to have lower effects on the child’s future earnings. Second, my extrapolation from the DMR paper relies on effects on educational achievement to predict future earnings, and there might be other channels by which parental income would influence a child’s future earnings. Third, the DMR estimates could be better estimates because they might better control for omitted variable bias than the DZK estimates. The DMR estimates are instrumental variable estimates using experimental assignment as instruments. The DZK estimates control for omitted variable bias using powerful controls, such as later parent income, but perhaps these controls are insufficient. The first two factors argue that the DZK paper may provide better estimates of the effects on future adult earnings of parental income in early childhood. The third factor argues that the DMR estimates may be better estimates. )