Economic Evaluations of State Policies

Thesis (Ph.D.)--University of Washington, 2016-08

Alcohol consumption and access to care are important measures of individual and public health. The goal of this dissertation is to examine these outcomes within the context of state-level policy evaluations and via an exploratory methods study. In the first aim of the dissertation, we examine the effect of privatizing retail liquor sales and distribution on household-level alcohol purchases in Washington State by type of alcohol, using data from the Nielsen Consumer Panel Dataset. We use the differences-in-differences approach, with all other U.S. states serving as controls, to estimate the effect of privatization on household alcohol purchases. We find that privatization of liquor is associated with modest increases in liquor, beer, and ethanol purchases in the overall sample. Our subgroup analysis finds that there was variability in responsiveness to the policy. Privatization was associated with increases in liquor and ethanol purchases among low purchasers, an increase in liquor purchases among moderate purchasers, and a decrease in ethanol among high purchasers. In the second aim of the dissertation, we examine whether Massachusetts health reform reduced the overall financial burden of accessing health services, using data from the Behavioral Risk Factor Surveillance System. We use the instrumental variables approach, with the Massachusetts health reform as the instrument for health insurance and, with three other New England states serving as controls, to estimate the effect of gaining health insurance on forgoing care because of cost. We also investigate whether there was a differential response to the policy change by income levels. We find that gaining health insurance is associated with forgoing care because of cost in the overall sample. In the subgroup analysis, we found a differential effect of gaining health insurance, with health insurance benefiting those likely eligible for subsidies and likely not eligible for assistance. There was not a significant effect of gaining health insurance on forgoing care among those likely eligible for Medicaid. In the third aim of the dissertation, we compare two broad instrumental variables approaches, two-stage least squares (2SLS) and two-stage residual inclusion (2SRI), within the setting of a binary endogenous treatment variable. We estimate the local average treatment effect of gaining health insurance on forgoing care, using Massachusetts health reform as an instrument for multiple models. Then we explore whether goodness-of-fit visualizations can assist in model selection. We conclude that the 2SLS model is biased and does not fit the data well. Goodness-of-fit visualizations are useful in distinguishing between choosing 2SLS and 2SRI.