Transferring multivariate benefit functions using geographical information systems

Recreation value studies have to date generally failed to adequately transfer functions predicting visitor numbers and benefit values from 'source' to 'target' recreational sites. One reason behind such failure may lie in the difficulty of successfully modelling the inter-site variation in predictors. The present study employs a geographical information system (GIS) to tackle this issue by integrating data from a number of sources (including a recreational site survey, the national census, regional road networks, etc.) permitting improved and standardised measurement and analysis of predictors at both source and target sites. The paper applies Poisson regression techniques to a set of GIS derived explanatory variables predicting arrivals and consumer surplus estimates at a single target woodland recreation site. GIS techniques are then used to define a corresponding set of predictors at 30 other target woodland recreation sites across England for which actual arrival numbers were known. Comparison of actual versus predicted arrivals provided a test of transferral validity which proved satisfactory and consumer surplus estimates were derived as before.