Independent Variables: Measures of Environmental Variability. We measured HMO penetration as the number of persons enrolled in HMO plans (allocated to metropolitan statistical area [MSA] counties within an HMO’s service area) divided by the total population of each MSA (Thorpe et al., 2001). We used a Herfindahl Hirschman Index (HHI) calculated on the basis of hospital admissions to measure competition at the county level. Ideally, it would be best to measure uninsured demand directly by using the number of uninsured individuals in the county. However, there are no publicly available data that allow us to measure this variable over time. Hence, we used total population, per capita income, and unemployment rate at the county level as proxy variables to capture uninsured demand. The demographic data were from the Area Resource File. In Virginia, there are five distinct administrative geographic regions, with one major safety net hospital (SNH) serving the central region and another serving the northwest region. As such, we define a binary variable called SNH in region to indicate whether a hospital is located within a region having one of these major safety net hospitals. Medicare share is defined as Medicare inpatient days divided by total inpatient days, and Medicaid share is defined as Medicaid inpatient days divided by total inpatient days. The inpatient data were from the AHA annual survey data.