Working Paper: NBER ID: w31808
Authors: Jacob Orchard; Valerie A. Ramey; Johannes Wieland
Abstract: Macroeconomics has increasingly adopted tools from the applied micro “credibility revolution” to estimate micro parameters that can inform macro questions. In this paper, we argue that researchers should take advantage of this confluence of micro and macro to take the credibility revolution one step further. We argue that researchers should assess the plausibility of the micro estimates and macro models by constructing macro counterfactuals for historical periods and comparing these counterfactuals with reasonable benchmarks. We illustrate this approach by conducting a case study of the 2001 U.S. tax rebates, as well as briefly summarizing two previous applications of the methodology. In the 2001 rebate case, we calibrate a two-good, two-agent New Keynesian model with the leading estimates of the household marginal propensity to consume (MPC) out of the rebates to construct a counterfactual path for nondurable consumption. The counterfactual path implies that without the tax rebate nondurable consumption spending would have fallen dramatically in the late summer and fall of 2001. Using forecasting regressions and other evidence, we argue that this counterfactual is implausible. When we investigate the source of the discrepancy, we find that the leading MPC estimates are not representative of the response of total consumption.
Keywords: macro counterfactuals; marginal propensity to consume; tax rebates
JEL Codes: E21; E27; E62
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| 2001 tax rebates (H20) | nondurable consumption (E21) |
| 2001 tax rebates not distributed (H23) | nondurable consumption spending would have dramatically fallen (E20) |
| micro estimates (D00) | implausible counterfactuals (D80) |
| MPC estimates (C13) | total consumption responses (E21) |
| macro counterfactuals (E71) | reconcile differences between micro and macro estimates (E10) |