A reliable output gap that reflects policymakers' beliefs
When imposing a low signal-to-noise ratio on output growth, the Beveridge-Nelson decomposition is able to produce an output gap that is consistent with policymakers' beliefs, but does not suffer the reliability problems that plague other methods that assume a similarly smooth trend. In particular, we show that the output gap based on the Beveridge-Nelson decomposition using Bayesian estimation of an autoregressive model of US quarterly output growth with a dogmatic prior on the sum of the autoregressive coefficients is persistent and large in amplitude, moves closely with the NBER-dated expansions and recessions, and leads to better out-of-sample forecasts of output growth and inflation than output gaps based on a quadratic trend, the Hodrick-Prescott filter, or a bandpass filter.