ABSTRACT
The production smoothing model incorporating stocks of finished goods and backlogs of unfilled orders has long been the paradigm within which empirical research on stocks and unfilled orders have been conducted. The basic hypothesis embedded in the model is that stocks of finished goods primarily serve to smooth production levels when the firm experiences fluctuating demand and convex cost functions. This is commonly known as the linear-quadratic stock model that was introduced by Holt et al. (1960). Blinder (1981, 1982) introduced the linear-quadratic functions (proposed by Holt et al.) for demand, cost of production and production change, as well as cost of holding stocks into his model which assumes that a firm will optimise net revenue. The conditions under which the optimisation results are obtained from this model are actually considered to be rather weak (see König and Seitz (1991)). Yet, empirical research has since presented evidence that is in sharp contrast to the theoretical model. The production smoothing hypothesis emerging from this model is contradicted in three respects:
Variance of production exceeds the variance of sales in most 2-digit industries,
Changes in sales and stocks are positively correlated, and
Estimated adjustment speeds of partial stock adjustment models turn out to be extremely low.
