Working Paper: NBER ID: w16169
Authors: Ryan Chahrour; Stephanie Schmitt-Grohe; Martin Uribe
Abstract: The SVAR and narrative approaches to estimating tax multipliers deliver significantly different results. The former yields multipliers of about 1 percent, whereas the latter produces much larger multipliers of about 3 percent. The SVAR and narrative approaches differ along two important dimensions: the identification scheme and the reduced-form transmission mechanism. This paper uses a DSGE-model approach to evaluate the hypothesis that the different tax multipliers stemming from the SVAR and narrative approaches are due to differences in the assumed reduced-form transmission mechanisms. The main finding of the paper is that in the context of the DSGE model employed this hypothesis is rejected. Instead, the observed differences in estimated multipliers are due either to both models failing to identify the same tax shock, or to small-sample uncertainty.
Keywords: No keywords provided
JEL Codes: E32; E62; H20
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| small sample uncertainty (C83) | discrepancies in empirical estimates (C51) |
| correct identification of tax shocks (H22) | similar average tax multipliers (H29) |
| Blanchard-Perotti model (E19) | fails to identify the same tax shocks as Romer-Romer model (E19) |
| Romer-Romer model (O41) | fails to identify the same tax shocks as Blanchard-Perotti model (H31) |
| tax shocks (H26) | output (C67) |
| SVAR approach (C32) | smaller tax multipliers (H29) |
| narrative approach (B53) | larger tax multipliers (H29) |