Title: Tests of policy ineffectiveness
Authors: Ron Smith - Birkbeck University of London (United Kingdom) [presenting]
Abstract: Tests for the effect of a policy intervention usually involve comparing post-intervention realizations of a target variable with a counterfactual: the predicted value in the absence of the policy change. Previous work will be discussed where we consider tests of this form for two cases. The first case uses a dynamic stochastic general equilibrium (DSGE) model. Here tests are unlikely to have much power unless the intervention changes the steady state, e.g. the target rate of inflation. This is because DSGE models use variables that are measured as deviations from the steady state and thus any effects of policy are transitory. The tests will also not be consistent, since as the evaluation horizon expands, the effect of the intervention declines exponentially and the power goes to zero. The second case uses reduced form or final form policy response equations, which explain the target variable by its lags, the policy variable and policy-invariant exogenous variables. Since the counterfactual is a type of forecast and since parsimonious models tend to forecast better, we may obtain more reliable estimates of the counterfactual outcomes and greater power of the test from a parsimonious policy response equation than a DSGE.