Estimating output gaps

I am grateful to many colleagues at the Reserve Bank for helpful comments. I am particularly indebted to Luci Ellis and Geoffrey Shuetrim for letting me use, and providing me with excellent explanations of, their respective programs for the multivariate HP filter and the state-space model estimated using a Kalman filter. All errors, however, are mine. The views in this paper are my own and do not The output gap, defined as actual less potential output, is an important variable in its own right and as an indicator of incipient changes in inflation. This paper reviews five methods of estimating it for Australian GDP data, including linear time trends, Hodrick-Prescott (HP) filter trends, multivariate HP filter trends, unobservable components models and a production function model. Estimates of the gap vary with the method used and are sensitive to changes in model specification and sample period. While gap estimates at any particular point in time are imprecise, the broad profile of the gap is similar across the range of methods examined. Inflation equations are substantially improved when any measure of the gap is included, and