NBER Working Papers by Nora Traum

Bayesian prior predictive analysis of five nested DSGE models suggests that model specifications and prior distributions tightly circumscribe the range of possible government spending multipliers. Multipliers are decomposed into wealth and substitution effects, yielding uniform comparisons across models. By constraining the multiplier to tight ranges, model and prior selections bias results, revealing less about fiscal effects in data than about the lenses through which researchers choose to interpret data. When monetary policy actively targets inflation, output multipliers can exceed one, but investment multipliers are likely to be negative. Passive monetary policy produces consistently strong multipliers for output, consumption, and investment.

Dynamic stochastic general equilibrium models that include policy rules for government spending, lump-sum transfers, and distortionary taxation on labor and capital income and on consumption expenditures are fit to U.S. data under a variety of specifications of fiscal policy rules. We obtain several results. First, the best fitting model allows a rich set of fiscal instruments to respond to stabilize debt. Second, responses of aggregate variables to fiscal policy shocks under rich fiscal rules can vary considerably from responses that allow only non-distortionary fiscal instruments to finance debt. Third, based on estimated policy rules, transfers, capital tax rates, and government spending have historically responded strongly to government debt, while labor taxes have responded more weakl...