Local and Aggregate Fiscal Policy Multipliers

Abstract

In this paper, I estimate the effect of defense spending on the U.S. macroeconomy since World War II. First, I construct a new panel dataset of state-level federal defense contracts. Second, I sum observations across states and, using the resulting time series, estimate the aggregate effect of defense spending on national income and employment via instrumental variables. Third, I estimate local multipliers using the state-level data, which measures the relative effect on economic activity due to relative differences in defense spending across states. Comparing the aggregate and local multiplier estimates, I find that the two differ dramatically. I infer that the local multiplier estimates alone do not provide useful information about the aggregate effects of policy. Finally, I use the panel aspect of the data to dramatically increase the precision of estimates of the aggregate multiplier (relative to using the aggregate data alone) by including a spillover term in the panel regressions. My baseline aggregate findings are a long-run multiplier on income equal to 1.6, a moderate long-run effect on employment, and no effect on income or employment effect in the short run. The results suggest that lags in the effects of defense spending are so long that they render countercyclical spending policies ineffective. In addition, I find negative short-run spillovers on employment of spending across state borders.