Why Is It So Hard to Measure the Current Output Gap?

In order for time series estimates of the output gap to be useful to policy makers, this paper argues that two factors will be critical. First, they must be able to produce an estimate of the current output gap based only on past information. Put another way, to evaluate the performance of such estimators, we should focus on their properties as _filters_ rather than _smoothers_. Second, the decomposition of actual output into potential output and the output gap must not be based on arbitrary assumptions about the time-series behaviour of these variables. In general, this leads to a multiplicity of possible solutions that may differ greatly in their policy implications. The arbitrary selection of one of these creates arbitrary policy advice. A trivial but flexible estimator of the output gap (dubbed TOFU) can be constructed which satisfies both of the above criteria. It estimates the output gap as part of a larger economic relationship. A discussion of its properties shows that except under very strict circumstances, such time series estimates will produce better estimates of the output gap ex post than they will ex ante. Furthermore, under certain conditions, contemporary and historical data will be of no use in estimating the output gap whatsoever. This in turn places limits on the extent to which time series methods can hope to improve upon structural estimates of the output gap.