Partitioning Variance in Regression Analyses for Developing Policy Impact Models: The Case of the Federal Medicaid Program

This paper extends the usual regression analysis to include the partition of the variance to examine the possible secondary and higher order impact of policy variables in a policy impact model. The federal Medicaid program is analyzed using the methodology to assess the impact of the policy variable—differential federal matching ratio—on the per capita expenditure from state and local sources. Our result shows that the policy is ineffectual on its own but modestly effective with others when all important factors are considered simultaneously.