A Discriminant Analysis of the 1980 São Paulo Industrial Location Survey

Abstract

Industrial location surveys typically generate large data sets with information on the characteristics of the companies and plants involved as well as on the relative importance of different locational push and pull factors. Since for the majority of companies locational choice results from the consideration of a range of factors, discriminant analysis can be used to reveal the important groupings of factors which distinguish the choices made by alternative subgroups of firms in a sample or population. This paper uses the results of the 581-firm 1980 Sao Paulo industrial location survey to examine some of the issues and specification choices which arise in using the discriminant model for this purpose. The results presented highlight the plant characteristics and the locational pull factors which distinguish industrial moves from the Sao Paulo metropolitan area out to surrounding cities from moves within the area.

Local economic development: Theory, evidence, and implications for policy in Brazil

“…The magnitude of the productivity advantages gained for cities in the core region of Brazil (the state of São Paulo and its vicinity) has been documented by several studies (Hansen, 1987;Henderson, 1988;Townroe, 1984;Townroe, 1985). Seeking to examine the rate of convergence across different states in Brazil for the period 1939, Azzoni (2001 reports strong signs of increasing divergence within the poorer regions of the country simultaneously with increasing convergence within the richer regions.…”

Assessing the impacts of regional characteristics on the location of manufacturing facilities: A review of recent methods and findings

“…No matter how information is collected, tables that rank faetors as most important, important, or insignificantcan be readily constructed. The factor ratings can also be used in a discriminant model to distinguish between categories of plants or searchers (Townroe 1985). Also, information collected from surveys (such as states considered in the search process) can be used in logit or probit models to test for the significance of factors in the choice of location (Schmenner, Huber, and Cook 11987).…”