It is well documented by demographers that there is a long and persistent under-coverage of adult black males in the US Census. Despite this, the census undercount has been largely ignored in research using census data. Similar omission patterns also exist in other household based surveys, such as the Current Population Survey and the Survey of Income and Program Participation. I demonstrate that estimates of the undercount that rely on counts from vital statistics data are understated and provide estimates of the undercount of prime age black men in household-based survey data that are robust to under-coverage in vital statistics data. Because the incarcerated are automatically included in the Census, the population at risk of becoming incarcerated provides a unique opportunity for the identification of non-reporter characteristics. I use variation in incarceration by state and year to estimate the impact of increases in incarceration on non-reporting. I find that 90 percent of the incarcerated population would have been non-reporting had they not been incarcerated. Applying reasonable estimates of the size of the population at risk of incarceration, I conclude that non-reporting is almost entirely driven by the population at risk of incarceration. I then use data from the Survey of Inmates on labor market outcomes of inmates prior to incarceration to impute outcomes for the non-reporting population. Accounting for non-reporting meaningfully increases estimated gaps in black-white educational attainment, unemployment rates, and annual earnings.
Are Marginal Jobs Dead End Jobs? Evidence from the Earned Income Tax Credit
Many studies have documented the short-term increase in labor supply for single mothers associated with the Earned Income Tax Credit (“EITC”). But the longer-run impact of encouraging employment in this manner depends on the career development of those induced to enter the labor market. I analyze the career effects of the EITC using data from the Survey of Income and Program Participation matched to earnings data from the Social Security Administration. I compare the career trajectories of a cohort of women who were single mothers in 1984, before the major expansion of the EITC, to a cohort of women who were single mothers in 1998. Women in the 1998 cohort have significantly higher labor market participation rates than women in the 1984 cohort. But the effect is not sustained once the children reach age 18 and the women are no longer EITC eligible. I develop a likelihood weighting technique with which I identify the single mothers who were most likely to have obtained marginal employment. Analyzing this group, I find evidence that only about 10 percent of the women whose labor market participation was induced by the EITC experience wage growth and earn beyond the range of the EITC. I argue that whereas there are career effects for a small fraction of the women in the sample, most have unstable employment. My findings suggest that the additional employment induced by the EITC does little to increase human capital among the majority of single mothers.
The Promise and Pitfalls of Differences-in-Differences: Reflections on "Sixteen and Pregnant" and Other Applications (with Kevin Lang) forthcoming: Journal of Business & Economic Statistics
We use the exchange between Kearney/Levine and Jaeger/Joyce/Kaestner on “16 and Pregnant”
to reexamine the use of DiD as a response to the failure of nature to properly design an
experiment for us. We argue that 1) any DiD paper should address why the original levels of the
experimental and control groups differed, and why this would not impact trends, 2) the parallel
trends argument requires a justification of the chosen functional form and that the use of the
interaction coefficients in probit and logit may be justified in some cases, and 3) parallel trends in
the period prior to treatment is suggestive of counterfactual parallel trends, but parallel pre-trends
is neither necessary nor sufficient for the parallel counterfactual trends condition to hold.
Importantly, the purely statistical approach uses pretesting and thus generates the wrong standard
errors. Moreover, we underline the dangers of implicitly or explicitly accepting the null
hypothesis when failing to reject the absence of a differential pre-trend.