An assessment of meta-analysis benefit transfer methods for valuing wetland ecosystem services in US National Wildlife Refuges
Patton, Douglas Arthur
MetadataShow full item record
In this dissertation, I conduct an assessment of meta-analysis benefit transfer methods by examining both existing meta-analysis models and by constructing a novel meta-analysis database and estimator. The analysis of both a newly developed and published meta-analysis studies focuses on ecosystem services provided by wetlands in U.S. National Wildlife Refuges (NWRs). An application of the meta-analysis models to four case study NWRs provides empirical examples of how meta-analysis benefit transfer can be used to value water quality enhancements and flood control/storm protection ecosystem services in specific wetland landscapes. In Chapter 3, A Monte Carlo simulation and forecast combination of the dependent variable from three published meta-analysis models indicates that the models can be useful for rank ordering sites for low-stakes decision making or for preliminary assessments of ecosystem service values. However, the simulation also indicates that an assessment of the accuracy of the existing models cannot be determined without strong assumptions about the structure of each model’s error term covariance matrix. A novel meta-analysis database in Chapter 4 with a similar focus to those assessed in Chapter 3 is developed in order to better quantify benefit transfer accuracy. Regression analysis of the meta-analysis database indicates similar rank orderings of predicted values relative to the results in the previous chapter. Advancements include more flexible modeling of local substitute wetlands and local populations as well as the inclusion of a user population variable omitted from earlier studies. Chapter 5 develops a novel Parametric Locally Weighted Least Squares (PLWLS) estimator that empirically models the theoretical notion of correspondence that is often mentioned but has not yet been quantified in the ecosystem service valuation literature. The non-linear PLWLS estimator improves on previous meta-analysis models by establishing a systematic approach to resampling while improving the efficiency of benefit transfers. Employing a jackknife simulation to approximate resampling, I verify that the PLWLS method generates forecasts that are more accurate than the conventional modeling approach. The results also indicate that the PLWLS estimator in conjunction with the novel data set generates forecasts that are more precise than the forecasts from the three earlier meta-analysis models.