Document Type

Thesis - Open Access

Award Date


Degree Name

Master of Science (MS)

Department / School



One factor which distinguishes the production of non-mobile housing from the production of other consumer goods is the relatively long duration of the production process. A recent study found the median construction period for single family units to be three months, the mean length being 4.3 months. The implication is that, for single family housing, the level of inventory under construction is approximately three to four times the level of monthly starts and completions. If the level of realized sales falls short of the level of sales expected by a builder, it will create undesired increases in the level of inventory, reducing profits and threatening the builder's existence in the industry. In order to reduce inventory to desired levels, the builder must either change his marketing policies in an effort to sell more units or reduce his starts of new units, or both. In this paper the author is concerned only with fluctuations in starts of new single family housing units. The investor, rather than the builder, bears the risk of selling or renting the constructed units. In contrast, between 1963 and 1971, 79 percent of all single family housing units were started without any commitment from buyers. Until these units are sold, the costs and risks of carrying a large unsold inventory remain with the builder. The author of this study developed for statistical analysis three models in an effort to explain fluctuations in residential single unit housing starts: Specifically, the proposition embodied in these models states that residential builders vary their starts of new single family housing units for two reasons. First, builders' expectations of future sales are constantly being revised according to their sales experience. Second, builders attempt to adjust their unsold inventory to desired levels, given their sales expectations. Data recording sales and unsold inventories of new single family housing units have been published monthly since 1963. The data analyzed cover the period from January 1965 to December 1971. The data represent permit and non-permit areas of all 50 states. Since changes in "expectations" and in "desires" are not directly observable, one must resort to models in which changes in expectations or desires are a function of observable phenomena. One class of models which performs this function is referred to as "adaptive expectations" models. The author proposes three alternative adaptive expectation models to explain how builders form sales expectations.

Library of Congress Subject Headings

Housing -- Finance
Housing -- Research



Number of Pages



South Dakota State University