Our model of niche width and mortality appears to work reasonably well for semiconductor firms as well as for restaurants. Two more different organizational forms would be hard to imagine. The differences between the restaurant and semiconductor studies are substantial. In the former disbandings are used to define mortality; in the latter, exits were studied. Individual restaurants are on average small organizations, and they are run by sole proprietors for the most part. Semiconductor companies are some- times billion-dollar enterprises, but they are sometimes as small as all but the smallest restaurants. The owners of both kinds of organizations intend them to be profitable, but the technologies used to generate their products stand at the opposite extremes of sophistication. In a comparison of these two studies, “high technology” meets “low technology.”
Restaurant environments were conceptualized as either facing strong seasonal variations or not having them; other kinds of cycles were not studied. In the semiconductor data, we were able to consider a much more complicated pattern of temporal variation in the environment. But in both cases aggregate measures of environments were used for measures of vari- ability. The units of aggregation were quite different, however. Restaurants were presumed to operate in a geographically localized environment, which confronted each of them in the same way regardless of form. So a highly unstable pattern of quarterly sales for a town was assumed to operate as a uniform stimulus, provoking different reactions from restaurants manifesting different forms. Among semiconductor producers, the unit of aggregation was the North American market for a product family. It was assumed that geography was of limited significance, and, indeed, convenience and accuracy of data were the reasons for using North American sales rather than worldwide sales. Research focusing on the industry in more recent years would have to use worldwide data since the industry became a truly global business in the later seventies and eighties.
The restaurant study was prospective, whereas the semiconductor study was retrospective. Consequently, the spans of time covered by the two studies were quite different. Indeed, the short time period and the small number of observations made over time precluded the use of restaurant data for the analyses in Chapters 9, 10, and 11. Although the semiconduc-tor study was not planned primarily as a study of niche width, we did have an advantage in carrying it out, namely the previous experience with the restaurant research. Among other things, this permitted measurement of variables as continua that had been measured as binary variables in the restaurant study (that is, coarseness of grain and generalism). This change and the greater number of uncensored observations undoubtedly helped improve the model’s fit.
Given these great differences between the kinds of organizations under study, and in the designs of the research projects themselves, it is remarkable how consistent the results were.
Source: Hannan Michael T., Freeman John (1993), Organizational Ecology, Harvard University Press; Reprint edition.