UC Berkeley

In the United States, health insurance is often necessary for access to regular, affordable health care. With only eight of every hundred Americans buying private insurance plans on the individual market, the main sources for health insurance traditionally have been employers and the government. As new laws are being debated and introduced to reform an expensive health care industry in which nearly one-sixth of the population is uninsured, research is needed in order to evaluate the costs and benefits of these policy changes and to predict their success. To this end, in addition to understanding how likely individuals are to adopt new health insurance policies, we also should be interested in knowing how the demand for health insurance and changes in its accessibility will affect non-medical decisions. Specifically, labor market choices have been theorized to be directly related to decisions involving insurance coverage. If the availability of health insurance distorts a workers' job-related decisions, then the changing the landscape for how to access insurance may reverberate in employment outcomes.

My dissertation focuses on understanding the factors that influence the demand for health insurance and the role that health insurance plays in an individual's decision to work, where to work, and how much to work. Specifically, I focus on the following three related questions: how does the demand for insurance affect labor market decisions such as when to exit unemployment? what drives insurance demand, and in particular, what motivators work best to increase demand for health coverage among the uninsured? and lastly, what are the supply-side employment responses to the provision of free or reduced-cost public health insurance?

My first chapter explores how the demand for health insurance can change re-employment decisions among the unemployed, as well as the speed at which individuals return to work. Past research on this issue focuses on job-to-job switches and "job lock" but has yet to focus on individuals looking for work. This chapter uses data on laid-off individuals from the Medical Expenditure Panel Survey to compare the job search behavior and outcomes of individuals who differ in their demand for health insurance. I use three proxies for demand, based on spousal health and past insurance offer take-up decisions. Although each is potentially confounded by unobserved determinants of job search, I use a difference-in-differences and propensity score designs to isolate plausibly causal effects. I find consistent patterns across all three proxies (despite different potential omitted variables biases). Overall unemployment durations do not vary with demand for insurance, but this masks variation in the types of jobs taken. Individuals with higher demand for insurance have higher hazards for exiting unemployment into a job with insurance, but lower hazards for exiting to a job without insurance. This points to effects of insurance demand on both search effort and reservation wages, and to potentially important distorting effects of employer-linked health insurance.

Whereas the first chapter takes variation in demand for insurance as a given, my second chapter digs deeper into the basis for this variation and whether it can be affected. In this chapter, I investigate the reasons the uninsured choose to forego insurance coverage and the impact of different messages on their insurance demand. Working with Enroll America, a large non-profit dedicated to decreasing the number of uninsured Americans, I conducted a stratified experiment to determine the best communication strategies to encourage participation in the healthcare exchanges. We test a combination of the following behavioral and information treatments: a risk treatment that emphasizes the average financial risk for someone without health insurance; a norms treatment that alerts our participants that staying uninsured will be against the law; a savings treatment that highlights the average savings available at the exchanges; a wording treatment where we refer to the Affordable Care Act (ACA) as "Obamacare"; and lastly, a cost-calculator treatment that allows individuals to explore the likely cost of insurance based on their own characteristics. Among the uninsured, we find that the cost-calculator treatment, the risk treatment, and the mandate are most effective in increasing intention to purchase insurance. The cost-calculator and the risk treatment increase informedness among this population, but the cost-calculator (when paired with the savings treatment) is the only treatment that increases willingness to pay for insurance. We use the information on willingness to pay to construct sub-group price elasticities of demand to compare to previous work interested in the demand for health insurance. Overall, the results of this chapter highlight the importance of informational campaigns to increase awareness of the costs and benefits of health coverage, particularly after large changes such as those implemented by the ACA.

My third chapter continues by looking at the changes that have been introduced as a result of the ACA. Specifically, it explores whether expanding access to government-provided insurance affects individuals' decisions regarding employment and overall hours of work. Recent findings have suggested that increasing access to health insurance outside of employment has a sizable, negative impact on labor force participation. Along these lines, the Congressional Budget Office predicted that the expansion of Medicaid and private health insurance will cause a 1.5 to 2% reduction in hours worked in the first ten years. Comparing states by whether they chose to expand Medicaid under reforms introduced by the ACA, I look at changes in the probability a childless adult receives Medicaid, as well as changes in this group's employment likelihood and hours of work. Using household survey data from the CPS monthly survey and ASEC Supplement, I confirm a marked increase in the percent of childless adults insured by Medicaid but find no statistically significant changes in employment outcomes. I compare these results to other estimates of "employment lock" in recent literature. These results, though imprecise, align with the findings in Chapter 1 which suggest that overall employment is not drastically affected by insurance demand.