Irrigation Water Valuation Using Spatial Hedonic Models in GIS Environment

Introduction

Water resources are very important for the humans and the natural environment. The availability of adequate water quantities for all uses is also important for the economic development of every country, because water is a necessary input to production in the most of all economic sectors. Agriculture is the biggest consumer of fresh water among all economic sectors in many regions of the world. The most significant problem in the agriculture sector is the intense use of water, in conjunction with its low efficiency. It is widely accepted in the literature that water must be treated as an economic good and more efficient water prices must be introduced, in order to achieve an efficient water allocation. The establishment of more efficient water prices requires reliable estimation of the economic value of water.

Various valuation techniques have been developed to capture the economic value of environmental resources. These techniques are classified in two major categories, the direct or stated preference methods and the indirect or revealed preference ones. The main methods, which are widely used, and can be used, for the estimation of the economic value of water resources are: hedonic pricing, travel cost, contingent valuation, and choice experiment. The first two of the above belong to the category of indirect methods while the other two to that of direct methods (Birol et al., 2006).

The hedonic pricing method is commonly used in real estate appraisal, real estate economics and consumer price index calculations, but there are also many examples of the use of the method in irrigation water valuation, like the applications of the method presented by Milliman (1959), Hartman and Anderson (1962), Torrel et al. (1990), Faux and Perry (1999). In a hedonic pricing method application a heterogeneous good, usually land property or housing, is treated as a sum of individual goods (characteristics or attributes) that can not be sold separately in the market (Nelson, 1978). Then the individual values of the characteristics of the good under study are estimated (for example the agricultural land value and the individual values of its attributes, such as size, irrigation water etc.) by means of ordinary least squares regression analysis. One of the important assumptions in an OLS regression analysis is that the error must be constant across the sample in order to obtain unbiased parameter estimates, but this is not the case in the most of the hedonic pricing applications. The reason is that the sample, which is used for the estimation of a hedonic pricing model, consists of land property characteristics that are distributed across the study area and, therefore, according to Tobler’s “first law of geography”: “…everything is related to everything, but closer things more so,…” (Tobler, 1979). In other words spatial dependence is here not exception but the rule, and is known as “spatial autocorrelation”.

There are many statistical techniques that have been developed to deal with the problem of spatial autocorrelation. According to Anselin and Berra (1998), spatial autoregressive models are more appropriate for economic data. These models confront the problem of spatial dependence by means of spatial weights and spatial lag operators. The spatial weights define for each observation (property), the “neighborhood sets”, i.e., those locations that interact with it. Therefore, the present research is focused on the estimation of the hedonic pricing function using two well known autoregressive models and, more specifically, the spatial lag and the spatial error model. In a previous work (Mallios et al., 2007b), a specific approach on estimating the value of irrigation water using spatial regression models produced better parameter estimates compared to those obtained by the conventional OLS regression.