The land use conversion modeling component being undertaken by the economists involved in this project constitutes a separate analysis, one of human decisions and how they affect the pattern of land use. Predictions from this modeling effort, under different scenarios about future policies and growth, can be passed to the PLM model, allowing us to move beyond exogenously imposing scenarios regarding development and to incorporate the results of more sophisticated spatial and dynamic models of human decisions. Because spatially explicit data is able to characterize landscape patterns and features for which humans have preferences, decisions with regard to land use change can be modeled in a spatially disaggregated way. The result is an economic model that can predict probabilities of land conversion from forest or agriculture to different densities of residential use within each 100 m2 cell in the seven-county area of the Patuxent basin.
The first step in the economists' modeling process is to estimate, statistically, models that explain the value of land parcels in different uses. Prior work has been used to approximate values in agriculture and forest, but current work involves modeling the value of land in residential use have been developed (Bockstael and Bell 1997, Bockstael 1996, Bockstael et al. 1995, Geoghegan et al. 1997).
This was made possible by an extensive GIS data base that includes geo-coded records of all parcels in the tax assessment databases of the seven counties. The tax assessment data base includes historical data on actual transactions (selling prices) together with characteristics and location of parcels. In addition, an extensive spatial data base of land use, zoning and other natural and human-imposed characteristics that might influence values in residential and alternative uses, as well as conversion costs, has been assembled. The value of land in residential use is explained by an extensive array of features of the location: distance to employment centers, access to public infrastructure (roads, recreational facilities, shopping centers, sewer and water services), and proximity to desirable (e.g. waterfront) and undesirable (e.g. waste dumps) land uses to name a few. Also included are some less obvious explanatory variables that describe the nature of the land uses around a parcel. The estimation techniques used are maximum likelihood and generalized method of moments, the latter being an approach that allows for treatment of the obvious spatial autocorrelation in the model (Bell and Bockstael, 1997).
The ability to spatially locate transactions and account for locational characteristics, explicity, has provided an improved technique to test assumptions about what affects residential land values. The model demonstrates the importance of scale (e.g. in lot size or development density) considerations and the non-linearities associated with distance effects. For instance, proximity to some features of the landscape, such as major highways, are positive amenities up to a point, but become disamenities when too close. This first stage modeling exercise also provides a means of creating predicted spatial maps of value of undeveloped land were it to be put in residential use, given the existing set of zoning ordinances, public utilities provision, highway network, etc. These predictions can then be used in the second modeling stage.
The second stage of the economics modeling process involves estimating qualitative- dependent variable models (i.e. discrete choice models) of historical land use conversion decisions. In this stage, historical decisions as to whether or not to convert a parcel in an open space use agricultural or forest) to residential use are modeled as functions of the value in original use, predicted value in residential use (derived from the first stage model), and proxies for the relative costs of conversion. The purpose of this model is to determine what factors affect land use conversion and to estimate parameters of those conversion functions.
Once the parameters of the two stages of the model are estimated for any given submarket, the model is used to generate the relative likelihoods of conversion of different parcels in the landscape. Thus a spatial pattern of relative development pressure is obtained as a function of characteristics of the parcels and their locations. Since the explanatory variables used to predict the values in residential and alternative uses and the costs of conversion are all functions of ecological features, human infrastructure, and government policies, the effects of changes in any of these variables can be simulated.
Contact: Professor Nancy Bockstael
Dept. of Agricultural and Resource Economics
University of Maryland
2200 Symons Hall
College Park, MD 20742