A Positive Model of Expenditure Growth
26 Pages Posted: 20 Dec 2007 Last revised: 3 Jan 2014
Date Written: January 1, 2013
A good positive model is merely a good normative model run backwards, Here, we test a positive model of the budget process, where the main goal of jurisdictions, which face a hard budget constraint, is stabilizing expenditure growth, taking existing revenue structures and fiscal assets (savings) as givens, and treating revenue and savings growth as continuous-time, continuous-state stochastic processes. Our empirical tests address to two reduced-form specifications derived from our positive model. The first predicts future spending as a function of a set of current (in the period of budget formulation) state variables, including a jurisdiction’s savings. The second predicts a jurisdiction’s savings as a linear-additive function of past spending, the rate of spending growth, and revenue volatility. We estimated both specifications using OLS and 1995-2006 census data for >14,000 municipalities. The first specification produces results that are consistent with our expectations, but not perfectly so: spending growth is more stable than revenue growth as the model predicts, but less stable than we anticipated. Moreover, spending is less responsive to savings than we predicted. Consequently, our reduced-form specification barely outperforms a simple exponential smoothing model. Perhaps, this is the case because we used a “representative” municipality to calculate this result. Or, because budgetary decision makers are more myopic than we hypothesized. The results for the second specification are perfectly consistent with our expectations. This is arguably a fairly strong outcome: savings levels are uniquely determined in this specification and many observers will find the predicted result counterintuitive.
Keywords: Process, Mechanism, State and local government expenditure, State government savings, Martingale methods, Optimal stopping models
JEL Classification: H71, H72
Suggested Citation: Suggested Citation