Valuing Nature: A Survey of the Non-Market Valuation Literature

The use of geo-spatial differentiation in resource economics has become more widespread in recent years, particularly in the urban and real estate literature. However, previous to use in environmental economics, geo-spatial characteristics were used in international trade and theories of industrial concentration (Krugman, 1998). Geographic Information Systems, or GIS, is a database management system with a spatial reference. This spatial reference is commonly used to " geocode " data; that is, to assign a latitude and longitude based on the street address of the data point. An introduction to GIS as a geo-spatial technique can be found in Star and Estes (1990). used GIS in their analyses of real estate. A more recent paper by Clapp, Rodriguez, and Thrall (1997) both surveys the literature and discusses the advantages, disadvantages, and uses of GIS. Clapp et al. divided recent empirical studies employing GIS into two categories: those using explicit distance as a proxy for transaction costs and those using spatial area characteristics. utilized a straight-line distance measure to the central business district of their studies, as well as other important locations. Later studies that also used explicit distances measures also included spatial autoregressive processes (SAR). The SAR technique tests for and corrects spatial autocorrelation by calculating distances from each transaction to every other transaction. These tests suggest omitted spatial variables and indicate that substantial statistical gains can be obtained by explicitly modeling these processes. Pace and Gilley (1997) employed a standard hedonic model and compared it to a hedonic model with a method of weighted neighborhood geographical areas. They found these geographic variables changed all estimated coefficients by a large amount and increased the R-squared statistic significantly. Earlier, Dubin (1988) found that the consequences of ignoring spatial autocorrelation included: inefficient OLS estimates, biased standard errors, and inaccurate