Skip to main content

Policy analysis with beliefs about identifying restrictions

Periodic Reporting for period 2 - PABIR (Policy analysis with beliefs about identifying restrictions)

Reporting period: 2018-04-01 to 2019-09-30

This project addresses a key controversy in macroeconomic policy evaluation, that policy conclusions depend on the choice of identifying restrictions of uncertain validity. The objective is to develop new econometric methods for policy analysis that allow researchers to incorporate beliefs about the validity of identifying restrictions. On the empirical front, the tools enable a broad range of new applications and make it possible to search for policy conclusions that are robust to prior assumptions.
In three completed and submitted academic papers, I have developed new econometric methods for conducting multiple-prior Bayesian inference in set-identified models. For example, the new methods can be used to: 1) construct and report a consistent estimator of the impulse-response identified set in Structural Vector Autoregressive models that only impose credible identifying restrictions; 1) make policy conclusions only driven by the information contained in the identifying restrictions, and and not by the choice of arbitrary priors; 3) perform estimation in the presence of ambiguity, for example for the purpose of conducting sensitivity analysis to the choice of a particular prior.
"The new methods can be adopted by policy-makers and empirical researchers, for example to evaluate the effects of monetary policy.

Research that I expect to conduct until the end of the project includes: 1) extending the multiple-prior Bayesian approach to fiscal policy analysis in models identified by external instruments (so-called ""Proxy VARs""), also addressing the interplay between set identification and weak identification that can arise in these models; 2) applying the estimation under ambiguity approach to the empirically relevant context of elasticity estimation; 3) pursuing an alternative approach to model selection for the purpose of policy evaluation (having shown in the completed research that the multiple-prior approach is not fit for this purpose).