The speed of adjustment and production smoothing: Bayesian estimation

Abstract

The models used for empirical studies of inventory behavior typically include the lagged stock of inventories (as a result of the partial adjustment assumption), and also exhibit serial correlation in the residuals. Some researchers suggested that the very low speed of adjustment of the actual inventories to their desired levels, and also the lack of empirical support for the production smoothing hypothesis may be the results of applying inappropriate estimation methods to the data. Recently, a Bayesian estimator that applies to models containing both lagged-dependent variables and autocorrelated disturbances was developed. In this paper we examine the empirical evidence on the speed of adjustment of inventories to their desired levels and the degree to which firms smooth production by applying this new estimation method to monthly industry data. The results show that the low speeds of adjustment reported in the literature, as well as the low degree, or lack, of production smoothing also reported in the literature, may indeed be artifacts of the estimation procedures used.