Identifying cyclical and trend components of labour productivity has important implications for macroeconomic analysis, since distinguishing transitory from permanent productivity changes is useful for judgements of structural reform programmes and the sustainability of activity levels. Estimating trend productivity permits measuring `gaps', interpreted as deviations from potential; this information may help to guide macroeconomic policy. In Discussion Paper No. 808, Giuseppe Nicoletti and Research Fellow Lucrezia Reichlin construct trend and cyclical components of business sector labour productivity in the major OECD countries using quarterly data for 1960-92, defining the `trend' as the long-run forecast of the productivity level adjusted for mean growth. At each point in time, this `normal' level is the sum of the current value of and all expected future changes in this series, while the `cycle' is the difference between this series and its normal level.

Nicoletti and Reichlin adopt a multivariate generalization of the BeveridgeNelson approach which improves on traditional univariate procedures in two ways. First, their estimation of trends is based on the simultaneous observation of productivity and other series that account in part for its cyclical movements. Second, it uses long-run empirical regularities linking labour productivity to other series to enlarge the information set on which trend computations are based, so the resulting decompositions are more reliable. They consider a wide range of variables displaying a cyclical and/or long-run relationship with labour productivity and include all those with explanatory power in their estimated models. The resulting estimated cycles outperform those generated by the HodrickPrescott filter in terms of stability properties and predictive performance.

Trends and Cycles in Labour Productivity in the Major OECD Countries
Giuseppe Nicoletti and Lucrezia Reichlin

Discussion Paper No. 808, July 1993 (IM)