Estimating the nonmarket value of green technologies using partial data enrichment techniques: Brett R. Gelso
Introduction Many cities throughout the United States currently face important water management issues. Among the most prominent of these issues is the control of storm water runoff from urban and nearby agricultural areas, which potentially contaminates water bodies and contributes to the risk of property damage from flooding. As an alternative to managing water flow with traditional technologies, civil engineers and urban planners are becoming increasingly interested in “green technologies.” Examples of these technologies, those that will both prevent flooding and mediate contamination before runoff enters water bodies, are planting trees in strategic locations and constructing wetlands. Green technologies may better be viewed as restoring natural water control methods, since urban growth replaces vegetation with unnatural impervious surfaces such as buildings and pavement. Recent literature has suggested that there needs to be a greater empirical basis to estimate the benefits from scenery, wildlife, and the social values associated with urban trees. Gelso (2002) developed a cost-minimization model that chose the optimal combination of traditional facilities, urban trees, and restored wetlands, where the non-water benefits of green technologies are taken explicitly into account. Based on likely parameter values for the city of Topeka, numerical simulations of this model suggested that green technologies may be a cost-effective way to manage urban runoff. Yet, the results were sensitive to unquantified external benefits. Although revealed preference hedonic methods may be used to estimate how much a resident will pay for scenery provided from residential trees, some research, such as Earnhart (2001), has suggested that a combination of revealed and stated preference methods will more accurately account for the amenity benefits from urban trees in residential neighborhoods. As noted repeatedly in this book, combining revealed preference (RP) and stated preference (SP) methods may ameliorate respective weaknesses of each data source, while taking advantage of respective strengths. Such “combined” models often impose the restriction that all model parameters are independent of the data source-i.e. that RP and SP data are generated from the same set of preferences. However, Swait et al. (1994) suggested that the hypothesis of parameter equality is not supported in many applications, perhaps due to the
known sources of bias inherent in these types of data. In such cases, Louviere et al. (2000) investigated “partial data enrichment,” when combining RP and SP data, where only certain model parameters are restricted to be equal across data sources and others are data-specific. Combined models offer many attractive economic and econometric properties. By adding stated preference data to observed data, variables may be orthogonal by design, non-use values are included, statistical efficiency may be improved, and stated preference data is at least partially based on observed data. In the current chapter, we develop a combined choice experiment (stated preference), recreational site choice (revealed preference) model to estimate the benefit of green technologies. In the process, we test and correct for scale differences between RP and SP data. Further, we test for structural differences in preference functions between RP and SP data sources and use partial data enrichment techniques to allow for flexibility in the modeling of behavioral functions derived from different data sources.