Productivity Growth
Trends and cycles

The relative contributions of temporary and permanent components to the total variability of GNP shed light on how economies evolve over time. If permanent changes account for most variance of GNP, technology shocks should dominate demand shocks in explaining economic fluctuations. Interpreting the trend or permanent component as productivity growth is inconsistent with the conventional assumption that it is a random walk, however, which also rules out such well-known features of technological change as learning at the firm level and the coexistence of different capital vintages. In Discussion Paper No. 775, Marco Lippi and Research Fellow Lucrezia Reichlin characterize the permanent component of GNP instead as an S-shaped diffusion curve, which is consistent with findings that innovations are absorbed gradually rather than introduced in one fell swoop.

Lippi and Reichlin then estimate their model on US post-war quarterly data and find that technological innovations are absorbed in about 15 months and the variance of the trend explains most of the total variation of output growth. In contrast, results obtained by modelling the trend as a random walk impose the immediate absorption of permanent shocks throughout the economy, and the cycle explains most output growth. The authors show that their preferred diffusion trend model performs just as well as the standard random walk model and find that there is a `dynamic trade-off': assuming rich dynamics for the cycle leads to a trivial trend while allowing more complicated dynamics for the trend impoverishes the dynamics of the cycle. The data reject all intermediate cases.

Diffusion of Technical Change and the Decomposition of Output into Trend and Cycle
Marco Lippi and Lucrezia Reichlin

Discussion Paper No. 775, March 1993 (IM)