Essays on Generational Economic Links Between Childhood and Adulthood

Abstract

My dissertation examines the economic links between people's experiences in early and later-life. It offers new empirical evidence on the effect of income in infancy on later-life outcomes, and investigates the performance and econometric properties of the linking tools often used to create data for these long-term empirical investigations.

In my first chapter, I estimate a relationship between family income in infancy and later-life outcomes for children. Eligibility for child-related tax benefits depends on the calendar year in which a child is born. Families with children born in December are eligible for tax benefits a year earlier than families with children born a few days later in January. These differences create a discontinuity in after-tax income in infancy worth on average approximately $2,000 for families in tax year 2016. I use regression discontinuity techniques to calculate the effect of this change in after-tax income on outcomes for children and young adults in Census data. Evidence show that a $1,000 increase in after-tax income in infancy results in a 1.2 percentage point increase in the probability of a student being grade-for-age by high school, a basic indicator of academic achievement and social maturity. Effects of this income shock are larger for children from families that are more likely disadvantaged at a child's birth, including Black families, and families with low education attainment. After high school, small differences in labor force attachment, earnings and education attainment persist for the adults who experienced the income increase as children. These effects are again pronounced for Black adults and adults born in counties with low average education attainment.

In my second and third chapters, I investigate methodological problems that arise when linking data. Linking is often necessary to investigate generational economic links between childhood and adulthood.

In the second chapter, my coauthors Martha Bailey, Catherine Massey, Morgan Henderson and I review the literature in historical record linkage in the U.S., and examine the performance of widely-used automated record linking algorithms. Focusing on algorithms in current practice, our findings highlight the important effects of linking methods on data quality. We then extend our analysis to look at the consequences of these differences in data quality on inference by computing intergenerational income elasticities between fathers and sons. Many of the methods produce estimated elasticities that are statistically distinguishable from the estimated intergenerational elasticity with hand-linked data, suggesting that the linking algorithms themselves may bias inference. However, eliminating false matches renders elasticity estimates similar to each other, and statistically indistinguishable from the elasticity estimated with the hand-linked data.

In the third chapter, my coauthors Martha Bailey, Catherine Massey and I investigate two complementary strategies to address the issues we highlight in my second chapter. We investigate the use of validation variables to identify higher quality links and a regression-based weighting procedure to increase the representativeness of custom research samples. We demonstrate the potential value of these strategies using the 1850-1930 Integrated Public Use Microdata Series Linked Representative Samples (IPUMS-LRS). We show that, while incorrect linking rates appear low in the IPUMS-LRS, researchers can reduce error rates further using validation variables. We also show researchers can reweight linked samples to balance observed characteristics in the linked sample with those in a reference population using a simple regression-based procedure.