Working Paper: CEPR ID: DP17379
Authors: Mario Forni; Luca Gambetti; Marco Lippi; Luca Sala
Abstract: We urge the use of Structural Dynamic Factor Models (DFM) to validate and to guide the construction of Dynamic Stochastic General Equilibrium (DSGE) models. The main reason is that the log-linear solution of a DSGE model has a factor structure which ensures consistency between the representations of the two models. We assess, by means of a few simulations, the validity of SDFM asan empirical tool to complement DSGE analysis. Using a DSGE model as data generating process, the factor model provides very accurate estimates of the true impulse response functions. As an application, we validate a theory of TFP news and surprise shocks.
Keywords: DSGE models; validation; structural VAR; structural factor model; news shocks
JEL Codes: C32; E32
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| log-linear solution of a DSGE model (E13) | consistency between theoretical and empirical models (C52) |
| DFMs (Y10) | accurate estimates of true impulse response functions (C51) |
| DFMs (Y10) | successful estimation of structural shocks and impulse response functions (C51) |
| measurement errors or informational deficiencies in SVAR models (C32) | biases in estimates (C51) |
| theoretical responses of real economic activity variables (E19) | alignment with empirical data (C52) |
| DSGE model (E13) | overestimate effects on consumption (D12) |
| DSGE model (E13) | better performance on GDP, investment, and hours worked (E23) |
| DSGE model (E13) | struggles with accurately predicting dynamics for inflation and interest rates (E47) |
| DFMs (Y10) | conclusions about DSGE model's validity (E13) |