Recent work by Kydland and Prescott to develop a stochastic dynamic general equilibrium version of the `real business cycle' (RBC) model of the US economy has provoked lively debate, but there have been few attempts to test their model with alternative data sets. In Discussion Paper No. 681, Riccardo Fiorito and Research Fellow Tryphon Kollintzas apply their methodology to investigate the stylized facts of business cycles using quarterly data for the G7 countries during 1960-89. They find that current versions of the RBC model can qualitatively (and often quantitatively) account for several important stylized facts in all seven economies, which display considerable regularities in the behaviour of output and most of its expenditure components. Most of these components are procyclical, with consumption generally fluctuating less and investment considerably more than real GNP or GDP; net exports and prices are countercyclical; while money has no clear-cut pattern across countries or definitions of the money stock.

The RBC model's predictions are not supported by their evidence on labour dynamics, however, since variations in employment are too small and those in hours per worker too large; employment lags output everywhere while hours per worker are coincidental or leading, which contradicts model formulations in which adjustments to employment are explicitly or implicitly synchronous with output. Even where the variability of total hours is about right, the variability of its components is not. In some models, hours per worker are fixed, so variability reflects changes in employment only. Models that allow hours per worker to vary grossly underpredict its variability and do not pick up the lagged adjustment of employment.

Fiorito and Kollintzas propose modifying the RBC model to account for these findings by allowing for labour hoarding, so firms that find it more costly to adjust employment than hours per worker use labour intensively in expansions and less so in contractions. This is consistent with survey and time-series data in US manufacturing and elsewhere and also with the observed facts that employment in Europe and Japan fluctuates less than in North America while total hours fluctuate much more than employment. Labour institutions in Europe and Japan present stronger impediments to adjustment and information flows, so labour hoarding may be more important than in North America. The authors conclude by proposing to develop RBC models to account for these empirical findings, which should incorporate the variable utilization of employment and adjustment costs.