This paper uses a multivariate generalisation of the Beveridge and Nelson methodology to model trends and cycles of business-sector labour productivity in the major OECD countries. The method implies that the trend is the long-term forecast of productivity, given all available information; the cycle is thus interpreted as the total excess growth that one would forecast beyond "normal" rates of productivity (see Evans and Reichlin, 1992). Multivariate trends in productivity were estimated including series that Granger-cause and, possibly, are cointegrated with productivity. The corresponding cycles were compared with those generated by the Hodrick-Prescott filter and with the business-cycle dating of the OECD. The stability and predictive properties of the Beveridge-Nelson and Hodrick-Prescott trends were compared. Finally, the estimated productivity gaps were used as proxies for capacity utilisation in econometric models of price formation in order to assess their empirical ...