The benefits of a semistructural approach
In this paper, Martin Beraja proposes a “semistructural” methodology to analyze the effects of counterfactual policy rule changes without having to fully commit to a structural model. He argues that although structural models are the main approach used currently to study unobserved, potential changes in policy rules, their main disadvantage is that they require making assumptions about model specifications. The semistructural methodology, on the other hand, relies on revealed preference conditions and causal assumptions to partially identify counterfactual distributions. It does not require fully specifying a model. The main advantage of his methodology is that it provides a “middle ground” between reduced-form evidence and structural models for policy analysis. It allows studying a broader set of counterfactuals without relying on potentially incorrect structural assumptions.
The method has three steps. The first is using data from an economy under a benchmark policy rule to estimate a recursive law of motion of the equilibrium of dynamic stochastic models—a reduced-form model. The second is imposing restrictions on these models’ equilibrium equations. They need to be enough to identify all coefficients in such equations—the structural model—given the reduced-form model. The last step is solving for the counterfactual equilibrium under an alternative policy rule, given the identified structural model.
A Semistructural Methodology for Policy Counterfactuals
Author: Martin Beraja
From: MIT
Reduced form approach with contemporaneous and news shocks
In this paper, McKay and Wolf show that in linearized structural macroeconomic models, knowledge of the causal effects of policy shocks is sufficient to construct counterfactual simulations under alternative policy rules. Specifically, if the econometrician can estimate the effects of contemporaneous shocks and news shocks about future changes to the prevailing policy rule, that is enough information to generate counterfactuals for different policy rules. The authors argue that:
- Time-series regression estimates of policy shock effects can be informative for policy analysis by constructing counterfactual distributions and optimal policies without needing to fully specify an underlying structural model.
- With a postulated loss function, the paper’s results also allow recovering the optimal policy rule that minimizes the loss.
- The counterfactual simulations and optimal policy rules derived this way are robust to the Lucas critique under the paper’s assumptions.
- When limited causal evidence is available, the paper discusses strategies such as focusing on a narrower set of counterfactuals consistent with the empirically identified shocks.
What Can Time-Series Regressions Tell Us About Policy Counterfactuals?
Authors: Alisdair McKay, Christian K. Wolf
From: Federal Reserve Bank of Minneapolis, MIT
