Evaluating macro policies

Counterfactual policy analysis

In this paper, McKay and Wolf propose a method that constructs policy counterfactuals using empirical evidence on multiple distinct policy shocks, rather than just a single one.
At the heart of their methodology lies an identification result:

[I]f the econometrician has successfully estimated the effects of contemporaneous shocks to the prevailing rule as well as the effects of news about deviations from the rule at all future horizons, then those estimates contain all the information she needs to construct the counterfactual. Key to the proof is our assumption on how policy rules are allowed to shape private-sector behavior. Since only the expected future path of the policy instrument matters, any given rule—characterized by the instrument path that it implies—can equivalently be synthesized by adding well-chosen shocks to the baseline rule. All that is required is that those policy shocks imply the same expected instrument path from date-0 onwards as the counterfactual rule. Finally we show that, given a loss function, our econometrician can furthermore leverage the same logic to also characterize optimal policy.

What Can Time-Series Regressions Tell Us About Policy Counterfactuals?
Authors: Alisdair McKay, Christian K. Wolf
From: Federal Reserve Board of Minneapolis, MIT

A different way to evaluate a single macro policy

In this paper, Barnichon and Mesters show that it is not necessary to know the full structure of the economy to evaluate a single macro policy decision– in contrast to the above paper that studies systematic changes in the policy rule. They argue that only two statistics are sufficient to detect policies that do not minimize the loss function. The two sufficient statistics are (i) the effects of policy shocks on the policy objectives, and (ii) forecasts for the policy objectives conditional on the policy decision. They introduce the Optimal Policy Perturbation (OPP), the gradient of the loss function, which is entirely determined by the two sufficient statistics, also also has a direct economic interpretation. The OPP is the adjustment to the policy instruments that exactly corrects an optimization failure when the loss function is quadratic and the (unspecified) underlying model is linear.
A Sufficient Statistics Approach for Macro Policy Evaluation
Authors: Régis Barnichon, Geert Mesters
From: Federal Reserve Bank of San Francisco, Universitat Pompeu Fabra