This dissertation consists of three chapters that analyze the Affordable Care Act (ACA)’s Health Insurance Exchanges.

The first chapter analyzes plan choice in Covered California, the state-based Exchange of California. Specifically, I analyze three aspects of plan choice: partial year coverage, inefficient plan choice, and plan switching. With respect to partial year coverage, I do not find evidence that supports the common perception that special enrollment enrollees abuse the opportunity to obtain partial year coverage by signing up, getting treated, and then immediately dropping coverage. My analysis of inefficient plan choice shows enrollees to be making mistakes in plan choice by selecting plans that are dominated from the perspective of out-of-pocket spending. As for plan switching, I find the switching rate on Covered California to be close to 15 percent. My estimates from a mixed logit consumer choice model imply that household plan switching costs range from $1,000 to $5,000, depending on the age and income of households.

The second chapter is joint work with Richard Scheffler, Brent Fulton, and Sherry Glied and was published in the May 2016 issue of Health Affairs. The chapter analyzes how health plan, hospital, and medical group market concentration impacted the growth of health insurance premiums between 2014 and 2015 in two ACA state-based Exchanges: Covered California and NY State of Health. The results show that both states exhibited a positive association between hospital concentration and premium growth. Our results for health plan concentration differed between two states: it was positively association with premium growth in New York, but negatively associated with premium growth in California. We suggest the health plan concentration finding in Covered California may be the result of the Exchange selectively contracting with health plans.

The third chapter is joint work with Jon Gabel, Brent Fulton, Sam Stromberg, Matthew Green, Heidi Whitmore, and Richard Scheffler and was published in the January 2017 issue of Health Affairs. The chapter analyzes differences between “offered” premiums (a straight average across plan premiums) and “purchased” premiums (an enrollment-weighted average of plan premiums) on Covered California. We find purchased premiums are 10 to 15 percent lower than offered premiums and premium growth is lower by 2 percentage points when measured using purchased premiums as opposed to offered premiums.

Much work has been done analyzing the determinants of health care expenditures. Much less effort has been devoted to analyzing the determinants of health itself. The focus of the analysis presented here is the production of health, with special attention paid to disaggregating health into pharmaceuticals and other health care. We also analyze the effects that wealth and certain lifestyle factors have on health.

Researchers who have analyzed the determinants of health across geographic units have found certain striking and consistent results. First, basic public health services, in the form of potable water and sanitation services, provide the biggest payoffs in decreased mortality for all age groups. Second, the expansion of health care services does not improve mortality to anywhere near the extent that public health infrastructure development does, if at all. Some researchers have even found positive relationships between some health care inputs and mortality. The results on income and wealth have been more mixed. In studies which have analyzed developing countries, researchers have found that higher incomes are negatively related to mortality. Other researchers have found exactly the opposite result when they have limited their samples to rich countries and/or regions thereof. Many researchers have also found that lifestyle factors such as nutrition, and cigarette and alcohol consumption, are important determinants of health.

Very few studies have estimated the effects of pharmaceutical consumption on mortality rates either directly or indirectly. The studies which have dealt with this directly in an international comparison context have had serious flaws. Some micro studies and many studies of restricted formularies in the United States Medicaid program have provided indirect evidence that pharmaceutical consumption has a positive impact on health.

To investigate whether such an effect could be found in an analysis of international data, we analyze a sample consisting of 21 OECD (Organization for Economic Cooperation and Development) countries as of the early 1990s. We convert pharmaceutical and total health care expenditures to U.S. dollars using purchasing power parity exchange rates for pharmaceuticals and health care, respectively. The purchasing power parities were provided by the OECD. Although other conversions are available for a limited number of countries, the measures of pharmaceutical and other health care consumption used here are the best available for a large number of OECD countries. We measure each country's health crudely, but objectively, using life expectancies at birth, at age 40, and at age 60, along with infant mortality.

The analysis consists of various multivariate regressions in which we estimate production functions for health. A functional form is used that allows for diminishing returns in each of the inputs in the production functions. The explanatory variables in each regression include pharmaceutical consumption, other health care consumption, gross domestic product, alcohol consumption, cigarette consumption, and richness of diet.

In our analysis, we find that pharmaceutical consumption has a positive and significant (both statistically and economically) effect on remaining life expectancy at age 40 and at age 60. It has a small, positive and statistically insignificant effect on life expectancy at birth. The elasticities of pharmaceutical consumption on life expectancy are roughly 0.017 at age 40 and 0.040 at age 60. The estimates are also quite robust to small changes. In a sample comprised of only the 16 European countries for which complete data were available, these elasticities were higher (0.023 for age forty and 0.050 for age sixty) and pharmaceutical consumption even had a small positive significant effect on life expectancy at birth. Pharmaceutical consumption appears to have no significant effect on infant mortality, although it appears that, controlling for lifestyle factors, increased pharmaceutical consumption may even be related to slightly increased infant mortality. Unfortunately, the infant mortality model is not robust to small changes, which does not inspire much confidence.

We also find that gross domestic product has a positive and significant effect on life expectancies at the ages of 40 and 60, although this effect is not present in the European-only sample. The results from the infant mortality regressions are mixed. It also appears that non-pharmaceutical health care consumption has no measurable effect on life expectancy, either at birth, at age 40, or at age 60. However, in one specification, we find that it has a negative effect on infant mortality. Again, where infant mortality is concerned, the results are mixed and not robust.

The lifestyle variable with the biggest effect on health is dietary richness, measured by the consumption of animal fat. Increased richness of diet improves mortality up to a point but the impact becomes negative as a diet becomes very rich. This result is consistent with the idea of the epidemiological transition: the idea that at low nutritional levels, enriching a diet allows one to better fight off infections, but that at high nutritional levels, enriching a diet leads to a greater incidence of degenerative diseases such as cancer and heart disease. This result is slightly surprising. One might have thought that the OECD countries were wealthy enough that nutrition, in this basic sense, would not be an issue.

We believe that this study will add to the debate over how OECD governments should allocate resources both among different health care goods and services and between health care and other goods and services. It improves on much of the existing literature in that it uses better measures of pharmaceutical and other health care consumption and uses a functional form that allows for diminishing returns. The results have been surprising, but they have also been fairly robust in the life expectancy models. The final conclusion is that increased pharmaceutical consumption helps improve mortality outcomes, especially for those at middle age and beyond.