The Location and Employment Choices of New Firms: An Econometric Model with Discrete and Continuous Endogenous Variables
构建了一个同时建模新分支工厂选址和就业选择的计量模型，利用离散和连续内生变量，并基于对偶理论揭示两者关联，为区域经济分析和政策制定提供参考。
B USINESS location is a subject of great interest. Businessmen obviously want to know where to locate their plants. State planners want to know the best way to attract new employment to their state. Regional economists use business location to get an advance reading on the health of an economy. Newly locating plants are responding to current incentives in making their locational choices and are therefore a better barometer of a region's future than employment at existing plants whose decisions are obviously influenced by their prior locational decision. Despite all the interest, economists know very little about the factors influencing new business location. (See Carlton (1979) for a survey.) Part of the reason is undoubtedly the lack of data on new business formation. As far as I know, Dun and Bradstreet is the only systematic data source available for studying new business formation. However, there have been only a handful of studies using this data set. And only two, Schmenner (1975) and Carlton (1979), have econometrically attempted to model location by economic variables. Schmenner's study looked at location within an SMSA and failed to find much explanatory power for his economic variables. Carlton (1979) focused on interregional location of new single establishment (one plant) firms and was able to uncover some significant economic effects. I also examined the location of branch plants in that paper, but failed to model the size of the branch plant. Obviously, what is of interest is not only where new location will occur but how much employment will be generated. No one to my knowledge has linked the two. This paper simultaneously models both the location and employment choice of new branch plants across SMSAs. The important methodological contributions are showing that the two decisions are linked (via duality theory) and exploiting this link in the model estimation. A special bonus of the methodology is that it allows for direct testing of the independence of irrelevant alternatives assumption in a logit model. The study takes special care to use information on individual plants in narrowly defined industries (4 digit SIC code) and narrowly defined geographic regions (SMSAs). The use of such disaggregate data is a distinguishing feature of the work. Some of the specific findings of the study are the following: