Is a Good Normative Model of Public Spending Run Backward a Good Positive Model?

We test a positive model of government spending and savings, where jurisdictions seek to stabilize spending growth and where revenue growth and savings are assumed to be continuous-time stochastic processes, the Dothan–Thompson optimal budget model. Our empirical tests address two reduced-form specifications derived from this model, each applied to two different subsets of municipalities. For each of the two specifications, we first estimated models using data for all available municipalities (more than 20,000) covered by the U.S. Census government finance data, and then tested them again on a much smaller balanced subset of the full panel. We find that municipalities do not, on average, optimally stabilize spending in the face of revenue volatility, nor do they consistently use savings to do so. There is, however, at least modest statistical evidence that savings reflect revenue volatility at typical levels of volatility. For all model specifications tested, the single greatest determinant of savings at the municipal level is the state in which the municipality is located. The next challenge is to identify and measure the impact of the underlying factors that explain the dramatic state-to-state differences observed in municipal savings patterns.