reports

Empirical studies in the economics of health insurance, health, and fertility

Princeton, NJ, USA: Princeton University (doctoral dissertation)

Princeton University, Princeton, NJ, USA

This thesis is a collection of essays in the economics of health insurance, health, and  fertility. The three essays each explore distinct and unconnected research questions. The  first study asks whether differences in health risk can explain the systematically different  health plan choices made by younger and older US consumers. Finding the answer to  be no, the essay then explores the implications of these differential selection patterns by  age. The second study asks how much of the glaring disparities in life expectancy between blacks and whites in the US can be accounted for by group differences in socioeconomic  factors. The third study asks whether the often-observed negative correlation between  educational attainment and fertility reflects a causal relationship. The first two essays are single-authored, with the second published in the journal Demography. The third essay is co-authored with Damon Clark and Heather Royer.

These three essays share an emphasis on empirical rather than theoretical contributions, and in all cases they benefit from access to unique or restricted datasets. Nonetheless,  the essays differ in the ways they characterize the data used, and the extent to which the quantities of interest are non-parametrically identified. The empirical evidence in the health  insurance essay relies most heavily on economic theory for identification. In part, this first  essay attempts to identify deep parameters associated with expected utility theory. The  fertility essay, in contrast, relies on a natural experiment to identify its parameters of inter-  est, which are the behavioral responses to an educational reform, rather than primitives of a  utility function. Finally, the essay on racial disparities in mortality pushes aside the issue of  identification, and instead provides a novel and careful accounting of partial correlations be-  tween race, socioeconomic status, and mortality. These correlations are important because  they have not been previously reported, despite much interest in the research question.

The first essay, in more detail, examines how health plan choice varies by age and health  risk, and asks whether current pricing restrictions in employer-sponsored health insurance  are optimal. US employers cannot charge different insurance premiums on the basis of  age, gender, or other observable characteristics. Using administrative employer data, I show that younger and older employees in the same firm have very different preferences  over health plans, above and beyond what differences in their health risk would predict.  I build a simple model of plan choice that demonstrates that as long as this is the case, requiring that young and old employees face the same prices can result in socially inefficient self-sorting across the various plans offered by the employer. The phenomenon is distinct from the well-known adverse selection effects. I estimate a structural model of plan choice to quantify the efficiency gains that would result if pricing were allowed to vary with age.  The results indicate that the welfare gains from introducing age-adjusted prices are small, though they are similar in magnitude to recent estimates of the welfare gains from perfectly  correcting adverse selection under uniform prices in employer health plans.

The second essay quantifies the extent to which socioeconomic and demographic characteristics can account for black-white disparities in life expectancy in the United States.  Although many studies have investigated the linkages between race, socioeconomic status, and mortality, this article is the first to measure how much of the life expectancy  gap remains after differences in mortality are purged of the compositional differences in  socioeconomic characteristics between blacks and whites. The decomposition is facilitated  by a reweighting technique that creates counterfactual estimation samples in which the distribution of income, education, employment and occupation, marital status, and other  theoretically relevant variables among blacks is made to match the distribution of these  variables among whites. For males, 80% of the black-white gap in life expectancy at age 1 can be accounted for by differences in characteristics. For females, 70% percent of the gap  can be accounted for.

The third essay examines the effect of school attendance and educational attainment  on fertility. To shed new light on the causal relationship, we exploit a natural experiment generated by a change in UK compulsory schooling laws in 1972, which raised the minimum leaving age from 15 to 16. The reform was recent enough that access to legal abortion and modern contraception at the time was quite similar to today, granting unique insight into  the fertility effects of education in a modern context. We show that the affected girls had significantly lower fertility in their teen years and that the decline was not accompanied by an increase in abortions. We also find that the reform had a negligible impact on completed fertility. Our findings suggest that education-based policies might reduce teen pregnancies  without impacting completed fertility rates.