On Measuring the Effects of Fiscal Policy in Recessions

This 2011 Journal of Economic Literature forum essay by Jonathan Parker argues that as of early 2009 economists had no good measure of how government spending affects output specifically in a recession, because the two dominant methodologies — vector autoregressions (VARs) and linearized (or close-to-linear) dynamic stochastic general equilibrium (DSGE) models — both estimate a single average multiplier that is, by construction, the same in a boom as in a slump. He traces this to two distinct causes: a methodological one, since even nonlinear DSGE models are usually solved and interpreted in a way that is close to linear, and a data one, since deep recessions are rare, so models that do allow state-dependence are estimated with little statistical power from aggregate time series alone. He also distinguishes the “marginal” multiplier — the effect of a small, unexpected deviation from the typical policy rule, which is what VAR impulse responses and most counterfactual DSGE exercises actually estimate — from the “total” multiplier relevant to an entire countercyclical stimulus program, arguing these can differ substantially if the multiplier declines with the size of the policy. As a path forward, Parker proposes disciplining nonlinear macro models with causal estimates from microeconomic studies — for instance of how households’ spending responds to tax rebates, or how state-level output responds to military spending — on the grounds that such estimates are both independent of the aggregate data the model is trying to explain and directly relevant to the specific causal channels inside it. He surveys several then-ongoing projects building exactly this kind of model and notes, without developing the point, that much of the argument applies equally to measuring the efficacy of monetary policy.

Jonathan A. Parker | Macro Paper Warehouse

On Measuring the Effects of Fiscal Policy in Recessions