| This dissertation is comprises three essays. The first attempts to answer the following question. Is fiscal policy more effective, as measured by the government spending multiplier, when the economy is "weak" relative to when it is "strong?" Results in the empirical literature have been mixed on this question. I use local projection techniques to estimate the impulse response functions of real output and real government spending to a shock to military spending. In addition, I attempt to endogenously estimate the level of the unemployment rate that distinguishes between states of the economy. I find that fiscal multipliers are near two at horizons of two to four years when unemployment is relatively high, compared to below 1 when unemployment is low. The second paper seeks to understand why disagreement in the empirical literature is so pervasive and if there are certain modeling choices that systematically lead to particular findings on the state dependence of the government spending multiplier. I identify eight dimensions along which many of the studies in the literature vary and determine if choices along these dimensions have a systematic impact on the results. I conclude that estimation of a state-dependent multiplier is, in general, not robust to various plausible specification assumptions. Finally, I estimate the effect of government spending at the county level using a previously little studied spending program, the Vinson-Trammell Act of 1934. Stimulated by fears about Japanese military expansion, this act aimed to build up the United States Navy to treaty allowances. I am able to identify local areas in the United States that hosted shipyards in 1934, and I estimate the effects of government spending on these areas. I find that manufacturing output, employment, and earnings all rise faster over the course of the 1930s in counties hosting shipyards at the time of the bill's passage. Also, I see significantly faster growth in county level retail sales and a positive effect on household consumption. Attempting to scale these results to an aggregate government spending multiplier, however, leads to a wide range of estimates for the effect on overall output. |
