Employment Subcenter Identification: A GIS-Based Method Qisheng Pan, Li Ma Abstract This research studied the methods for identifying employment subcenters and examined the effects of subcenters on surrounding density and housing price. Geographic information system (GIS) was used to organize data and model in a convenient way so that the spatial information such as distance, proximity, and adjacency can be utilized to identify employment subcenters. Houston metropolitan area was selected for the empirical analysis. It used the 1990 Census Transportation Planning Package, the 1990 and 2000 Census summary files 3, and the 2000 employment data obtained from the Houston-Galveston Area Council (H-GAC) to explore subcenters in the Houston area and highlight the changes of the subcenters between 1990 and 2000. 1. Introduction Most modern metropolitan areas in the U.S. have decentralized employment that has grouped into one or more subcenters outside of Central Business District (CBD) or dispersed in the entire region. The employment subcenters have been described in multicentric models (White 1976; Wieand 1987; Yinger 1992) and nonmonocentric models (Brueckner 1978; Ogawa 1980; Ogawa and Fujita 1980; Fujita and Ogawa 1982). They have also been identified in many U.S. metropolitan areas in empirical studies of Dunphy (1982); Gordon, Richardson, and Wang (1986); Cervero (1989); McDonald and McMillen (1990); Giuliano and Small (1991); McMillen and McDonald (1997, 1998); Cervero and Wu (1998); Craig and Ng (2001); McMillen (2001, 2003); McMillen and Smith (2003). Two major approaches have been used in subcenter identifications. One is the minimum cutoff point of gross employment density developed by Giuliano and Small (1991) and used by Small and Song (1994), McMillen and McDonald (1998), Cervero and Wu (1997, 1998), and Bogart and Ferry (1999), etc. The other one is the two-stage nonparametric approach proposed by McMillen and McDonald (1997) and revised by McMillen (2001, 2003). Craig and Ng (2001) developed a quantile smoothing splines method, which is also a nonparametric specification in employment density functions but it has not been followed by any other study. Both the minimum cut-off point method and the two-stage nonparametric approach need to analyze the spatial relationship between objects, such as the adjacency of census tracts. However, the spatial relationship has been identified manually and inefficiently because both models do not have appropriate functions to handle it. Similarly, the empirical studies on the effects of subcenters on surrounding density and housing price also have difficulties in conducting spatial analysis.