Grantee Research Project Results
2007 Progress Report: Demonstration and Evaluation of Structural Benefit Transfer and Bayesian Benefit Transfer for Valuing Welfare Impacts of Saltwater Beach Quality Changes
EPA Grant Number: R833461Title: Demonstration and Evaluation of Structural Benefit Transfer and Bayesian Benefit Transfer for Valuing Welfare Impacts of Saltwater Beach Quality Changes
Investigators: Poulos, Christine , Phaneuf, Daniel J. , Van Houtven, George L. , Parsons, George , Massey, Matt
Institution: Desert Research Institute , University of Delaware
EPA Project Officer: Hahn, Intaek
Project Period: September 1, 2007 through August 31, 2010
Project Period Covered by this Report: September 1, 2007 through August 31,2010
Project Amount: $199,111
RFA: Methodological Advances in Benefit Transfer Methods (2006) RFA Text | Recipients Lists
Research Category: Environmental Justice
Objective:
To overcome some of the observed shortcomings in existing benefit transfer (BT) methods, economists have developed other methods, including structural benefit transfer (SBT) (e.g., preference calibration and structural meta-analysis) and Bayesian BT methods. Both methods involve data combination, where results from previous studies are systematically combined with each other or with data from the policy site of interest to generate benefit estimates. The methods apply different strategies for combining data; however, mainly because of their novelty and technical requirements, they have been applied infrequently for policy analysis.
This study uses currently available information—prior studies of saltwater beach use, as well as data from the Coastal Module of the National Survey of Recreation and the Environment (NSRE)—to both demonstrate how these two methods can be operationalized and to evaluate their performance.
Progress Summary:
The structural benefit transfer and Bayesian benefit transfer activities are occurring in parallel.
Both activities assume the same policy context, in which some action will have an impact on the availability or characteristics of a recreation site. The project seeks to measure the welfare effect of the change for the relevant population.
For the Bayesian approach, we begin with a standard RUM structure and assume that there are J-1 substitute recreation sites and the utility a person in the population would receive from visiting the site is: (1) where j indexes the available sites and i denotes a person in the population. pij is the travel cost, qj is the policy-relevant quality attribute, is a site-specific fixed effect, and represents unobserved person-specific heterogeneity. The estimation goal is to quantify the parameters (). Absent an original study the objective is to quantify the indirect utility function in (1) using existing studies and existing secondary data and Bayesian benefit transfer.
The Bayesian benefit transfer approach has been conceptualized and a numerical example is being prepared. The approach will identify some of the parameters in (1) primarily based on "prior" information that will be gleaned from other studies, and some of the parameters from the limited amount of secondary data. Due to the fact that parameter estimates from RUM models of recreation site choice are confounded by the scale term, the Bayesian benefit transfer will specify equation (1) in WTP space.
We specify a RUM likelihood function and specify prior distributions for the parameters () using results from previous studies, micro-level data from secondary datasets, and data on the share of aggregate trips. The posterior distribution is proportional to the product of the likelihood function and the priors, and the estimation task is to compute via simulation the posterior means for the unknown parameters. With these posterior means, we are able to characterize the utility function in equation 1.
The next step has been to develop a generated data experiment that provides the opportunity to code the necessary algorithms and provide some Monte Carlo analyses of the properties of the Bayesian estimator. Preliminary results of the generated data experiment make use of travel cost data from the Southern Banks of North Carolina. These experiments were run under different conditions – with and without prior information on the alternative specific constants and marginal WTP. The results of this preliminary exercise demonstrate that while priors on marginal WTP to improve the posterior estimates of and , the WTP estimates are not close to true values when there are no priors on the alternative specific constants. Inclusion of these priors improves the estimates of WTP. The algorithms underlying this generated data experiment are being refined ,and the final exercise will use literature values, micro data (possibly from the NSRE), and aggregate share data for a selected study area to implement the estimator.
To develop a structural benefit transfer (SBT) model for beach recreation, we conducted a detailed literature search for nonmarket valuation studies applied to changes in beach width and/or beach access. We reviewed and abstracted information from over 60 papers and reports measuring values for changes in beach width or access and developed a database summarizing the key elements of these studies and value estimates.
Based on the characteristics of the available studies, we developed an SBT framework that expands on previous SBT applications by defining and parameterizing a spatial, multi-site benefit transfer function. Spatial heterogeneity in quality changes is an important feature of many non-market valuation studies, including random utility models (RUM) of recreation demand. To explicitly incorporate spatial heterogeneity, we adapted the preference framework described by Smith and von Haefen (1997), which generalizes the findings of Anderson, de Palma, and Thisse (1987, 1992) by linking an individual site choice model (with separate quality characteristics for each site) to an aggregate (representative) individual preference function. This framework includes 3-4 preference parameters, depending on the model specification. To estimate values for these parameters, we applied a preference calibration approach using results and summary data from 3 beach width valuation studies: (1) a RUM analysis of Mid-Atlantic beaches, (2) a travel cost study of North Carolina beaches, and (3) a stated preference study of northern New Jersey beaches. To make the analysis tractable, we defined three aggregated beach sites for each study, each comprised for roughly 20-25 miles of beach, we estimated average travel costs for the study sample to the centroid of each site, and we defined the quality variable as the percentage of beach miles in each site with wide beaches. We then inserted the calibrated parameter values into a WTP function, based on the assumed preference structure, and we evaluated the predictive properties of this SBT function by systematically varying travel costs and quality (beach width) for the aggregated sites.
Future Activities:
The Bayesian benefit transfer application will continue in two steps: (1) refinement of the generated data experiment and Monte Carlo simulation and (2) empirical application using the beach width value database being developed. This second step will involve estimating a RUM-type model of beach recreation behavior using a limited amount of primary data, and informative priors on the marginal WTP for beach attributes from the literature. We hope to use the NSRE data to do an application at a location with a primary data application, and compare the findings from transfer and primary data analysis exercises.
The structural benefit transfer application will continue by investigating the multi-site SBT model (developed for the previously described the preference calibration application) in a meta-analytic framework, using generalized method of moments(GMM) estimation . Our previous review of the literature has determined that there are not enough beach width valuation studies to support this type of statistical analysis; however, an alternative approach will be to use studies of beach access/closures, which are more numerous.
Journal Articles on this Report : 1 Displayed | Download in RIS Format
Other project views: | All 3 publications | 1 publications in selected types | All 1 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Van Houtven G, Poulos C. Valuing welfare impacts of beach erosion: an application of the structural benefit transfer method. American Journal of Agricultural Economics 2009;91(5):1343-1350. |
R833461 (2007) |
Exit |
Supplemental Keywords:
beach nourishment, beach erosion, benefit transfer, structural benefit transfer, Bayesian updating, preference calibration, nonmarket valuation, willingness-to-pay,, RFA, Economic, Social, & Behavioral Science Research Program, Scientific Discipline, Water, Ecosystem Protection/Environmental Exposure & Risk, Water & Watershed, Monitoring/Modeling, decision-making, Ecology and Ecosystems, Economics & Decision Making, Watersheds, Social Science, risk assessment, ecosystem modeling, aquatic ecosystem, Bayesian approach, decision analysis, decision making, environmental decision making, water quality, assessment endpoint mechanistic research, ecology assessment models, ecosystem stress, watershed assessment, ecological models, decision support tool, environmental risk assessment, water monitoring, adaptive implementation modelingProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.