Source: Dissertation Abstracts International, Volume: 59-04, Section: A, page: 1274.

Directors: William C. Brainard; Truman Bewley.

This study's main objective is to help explain why seasonal fluctuations in production are so similar in size and timing to seasonal fluctuations in sales. Such non-smoothing of production in anticipation of apparently predictable seasonal fluctuations in sales seems incompatible with the production-smoothing model of production which assumes convex production costs. This study attempts to reconcile seasonal fluctuations in production with the production-smoothing model. In particular, I argue, in contrast to many other authors, that seasonal fluctuations may be particularly uncertain at the relevant level of aggregation, i.e., the product level, while giving the appearance of predictability in aggregate data. When coupled with short life-spans of many seasonal products, cost savings that accrue from production smoothing can be offset by the potential losses of getting stuck with obsolete merchandise. Thus, production and sales can behave quite similarly even under convex production costs. The explanation offered is probably most relevant in final consumer goods industries. In order to investigate the empirical relevance of this hypothesis, I performed case studies of sixteen firms in various industries. Many of these firms face substantial seasonal fluctuations in demand. Some support is found for the hypothesis that product-level uncertainty and high product turnover play an important role in explaining the relatively great seasonal variability in production. In fact, obsolescence is a particularly potent concern for many of the firms facing the largest seasonality in sales. Furthermore, the evidence is supportive of the broad cost assumptions of the production-smoothing model.

This study also considers the possibility that seasonal product turnover and demand uncertainty may be responsible for erroneous rejections of the production-smoothing model in formal statistical tests. To investigate this possibility regressions were run on simulated data. The results of these regressions suggest that even when the production cost function is convex, the nature of the hypothesized seasonality can result in estimates implying negatively sloped marginal costs, non-smoothing of production in expectation of apparently predictable seasonal movements in sales, and downwardly biased estimates of the impact of observed cost variables on production.