Abstract
This note investigates productivity dynamics of US manufacturing industries during the period 1958–1996. Nonparametric methods are used to calculate total factor productivity relative to a frontier production function and to estimate the transition density. The shape of the transition density leads to the conclusion that persistence relative to the frontier function plays a substantial role in sectoral productivity development, thereby contradicting the existence of strong intertemporal spillover effects between sectors as supposed in recent growth models.
Notes
A Gaussian product kernel is used to estimate the bivariate kernel density. The bandwidth parameters are chosen according to the procedure suggested in Silverman (Citation1986).
The dominating ridge along the diagonal appears also in transition density plots for other values of k. The natural difference to is that plots for k < 5 are less dispersed, whereas plots for k > 5 are more dispersed.