Grantee Research Project Results
Final Report: Structural Benefits Transfer: Tying Utility Theory to Ecosystem Valuation
EPA Grant Number: R830923Title: Structural Benefits Transfer: Tying Utility Theory to Ecosystem Valuation
Investigators: Pattanayak, Subhrendu , Smith, V. Kerry , Van Houtven, George L.
Institution: Desert Research Institute , North Carolina State University
EPA Project Officer: Hahn, Intaek
Project Period: July 1, 2003 through December 31, 2007
Project Amount: $260,000
RFA: Decision-Making and Valuation for Environmental Policy (DMVEP) (2002) RFA Text | Recipients Lists
Research Category: Environmental Justice
Objective:
Our objective is to continue developing innovative and integrated approaches to benefit transfer for broad-based environmental policy evaluation. A key part of this process is to test the robustness of the preference calibration method via a set of interrelated conceptual and empirical tasks. By linking benefits transfer to the underlying preference structure in a variety of settings and by exploring its linkages to CGE calibration, we hope to strengthen and broaden the scope of benefit transfer for ecosystem valuation.
Our testing strategy will leverage conceptual frameworks, meta-data sets, and research collaborations we have developed. Specifically, we will conduct several applications of preference calibration for air, forest, and wetland resources; estimate preference parameters using generalized method of moments on a meta-data set of water quality values; and integrate multiple uses and multiple resources to calibrate common preference parameters.
Summary/Accomplishments (Outputs/Outcomes):
Background: A key limitation of conventional approaches to ecosystem valuation is that they propose partial equilibrium solutions to fundamentally general equilibrium problems. A more comprehensive and theoretically consistent approach to ecosystem valuation will require a strategy for integrating information from a wide variety of sources. By drawing on and extending “preference calibration” methods for benefit transfer and general equilibrium modeling approaches, we will develop an approach that we refer to as “structural benefits transfer.” We propose to test and expand the preference calibration logic to include a wider array of environmental resources and services and, in the process, develop guidelines for practical and defensible benefits assessment methods for broad-based environmental initiatives.
Our accomplishments on this project can be organized under the following categories:
- Recreation water quality calibration: Building from our Land Econ paper, we examined whether and how the preference calibration method can be extended other contexts and preference specifications. It uses different (from the Land Econ paper) travel cost and CV estimates for WQ changes to calibrate parameters for five alternative preference specifications and including non-use values in each specification. The results show that structural benefit transfer (SBT) estimates for selected water quality improvements and conditions are sensitive to preference specification; however, they also highlight the strengths and limitations of different specifications, by providing plausibility checks on the range of predicted outcomes. For example, in this application, SBT functions based on a linear trip demand specification produce more plausible benefit predictions than a log-linear demand or a Stone-Geary framework.
- Forest resources – calibration: We identified two published studies that have estimated economic values for changes in forest resources due to fire control using contingent valuation and travel cost methodology. We then developed a common index of forest fire conditions and identified all the other data (income, travel cost, willingness-to-pay) need for a calibration exercise. Unfortunately, the parameters from this calibration were not stable, presumably because one of the original studies generated negative values for fire control (e.g., people were willing to pay for burned over forests, presumably because this improved their viewing opportunities).
- Two-sample generalized method of moments (GMM) estimation: In collaboration with Nick Kuminoff (Virginia Tech), who led the code development in MATLAB, we built a flexible platform for multi-sample generalized method of moment estimation. This code can be critical for structural benefits transfer, and allows us to extend the logic associated with combining revealed and stated preference data in several ways. First, our approach does not require “complete” datasets, each with its own revealed or stated preference response. The GMM logic also does not require equal samples or impose size restrictions. As a result, it is possible in principle to use small samples to search individual preferences in regions that may not be observed with ordinary market choices. The power of the complementary samples logic is that we can work with two or more incomplete data sets at different levels of aggregation (e.g., even focus group findings or expert elicitation) to fill in all the relevant information by following the complementary samples literature on GMM estimation. Furthermore, most of the joint estimation literature has utilized maximum likelihood estimation, which relies on a high level of consistency in the relationship between the error processes associated with each data set. Closed form representations of multivariate probability distributions can be difficult to derive when the methods used to estimate choice parameters involve complex specification (e.g., highly non-linear) of primary choice information. By contrast the GMM logic offers a unifying framework for econometric estimation and a computationally convenient estimation strategy. Our approach also has parallels to recent developments in public economics, empirical industrial organization and development economics.
- GMM Monte Carlo example: In collaboration with Nick Kuminoff (Virginia Tech), we have completed a proof of concept of the GMM logic for structural benefits transfer by developing a Monte Carlo study. First, we construct a micro-sample of 300 observations of recreation demand as a function of travel cost, beach length, and income using a data set collected by George Parsons and Matt Massey. Next, we generate a macro sample of 100 Marshallian average (mean) consumer surplus and Hicksian willingness to pay estimates. We apply the GMM logic illustrated in Arellano and Meghir (1992) and Imbens and Lancaster (1994) to jointly estimate the original recreation demand parameters from these two unequal samples (the code is written in MATLAB).
- Real data for GMM estimation: Under this sub-task, we revisited a meta-data set of water quality values. We short listed approximately 250 water quality values from two types of studies - travel costs and contingent valuation. The GMM code developed under the proof of concept (see previous point) was used to estimate parameters of the preference function. The results from this exercise are preliminary and need a thorough review and evaluation. We have also continued to look for other micro and macro datasets to test the GMM logic. Two of the best candidates for these are: (a) a micro and 2 meta data sets on VSL and related concepts (from a CV, several hedonic wage, and labor supply studies), and (b) two micro and one meta-data set on household willingness to pay for improvements in water quantity and quality (from CV and averting behavior studies). We also collected meta-data on potential ecosystem changes (e.g., watershed protection) that would deliver these watershed services that are typically the focus on economic studies.
- GMM ‘real’ proof of concept. We completed a proof-of-concept for combining the VSL data using GMM and presented the findings at the AEA meetings in 2007. While this initial application provides credible parameter estimates, additional fine tuning is necessary. We also initiated research on specifying a model on joint estimation of CV and averting expenditure data for water supply. The latter was presented at the Southern Economics Association meetings in 2006.
- CGE calibration – lessons learned: Kerry Smith reviewed the literature on calibration of CGE models with Jared Carbone (Williams College, and now University of Calgary) and generated two insights for structural benefits transfers for ecosystem valuation. First, environmental amenities cannot be treated as separable from all market commodities in the utility specification. This has implications for the functional forms we use, the properties of preference functions we apply, and the parameters we estimate or calibrate for structural benefits transfer. Second, the market and the ecosystem interact through a circular flow. This implied recursivity, particularly for large scale policies that trigger services, means that we cannot assume that many critical variables such as prices, incomes and ecosystem services will be exogenous to the model. Based on this review Carbone and Smith developed the first method for general equilibrium calibration in the presence of non-separability. The basic logic of the method recognizes that preference calibration for non-separable resources in a GE framework must be solved with the general equilibrium at an economy level treated as a constraint to the task of calibration. This can be done in the context of a mixed complementarity problem. For large problems with the potential for economy wide impacts this implies that reconciling the parameters of preferences to estimates of WTP or MWTP must be conducted allowing the resulting GE responses in market goods and the generated emissions to be endogenous to the calibration process. The research developed two different approaches for conducting this calibration. To our knowledge this is the first time this problem has been solved at a general level and is reported in three published papers by Carbone and Smith and is part of on-going research for new environmental applications, including the design of non-market national accounts.
- CGE simulations of ecosystem services – simple examples. Subhrendu Pattanayak worked with a global dynamic general equilibrium model (ADAGE), developed by Martin Ross to evaluate US environmental policies, to test the implications for ecosystem service through two examples. First, we used ADAGE to evaluate a program to pay landowners for production of ecosystem services in Costa Rica. Changes in macro-economic indicators (e.g., GDP) deliver one measure of ecosystem service values. Second, we considered the links between environmental conservation and health against a backdrop of climate change, which is likely to increase adverse health outcomes such as cardiovascular, respiratory, and water-borne diseases. Large scale conservation (e.g., expansion of national forests) can mitigate these impacts because it influences (and disrupts) the transmission patterns for vector-borne and water-borne diseases in places such as Brazil. Thus, ADAGE was used to link policy to ecosystem change via adjustments to land and labor markets, and ecosystem values were derived in macro-economic terms.
- Conceptual framework for ecosystem services and valuation. As suggested by the recently published Millennium Ecosystem Assessment, the valuation of ecosystem services remains a major applied research task for the design and evaluation of policies to protect ecosystems. The literature suggest that ecosystem valuation is different from standard non-market valuation for marginal environmental change because it is important to study the long and complex chain from institutions (e.g., policies) to valuation, and isolate the causal impact of policies on valuable ecosystem outcomes. Three sets of functional relationships can be used to estimate values of ecosystem services by linking ecosystem functions and processes directly or indirectly to the well-being of people. The first stage of analysis measures the ecosystem flows (e.g., quantity or rate of stream flow, erosion, or sediment) that result from public polices and private actions for ecosystem conservation (e.g., watershed protection). Since these flows affect the production activities of individuals, the second stage measures how changes in ecosystem flows impact socioeconomic productivity (e.g., health, or income). In the third stage, these productivity changes are expressed in monetary terms, typically relying on derived demand for a related market good (e.g., averting good) or stated preference surveys (e.g., conjoint analysis). It is unlikely (if not impossible) that all of these data will be available from one or a set of related studies with any degree of precision. Nonetheless it is reasonable to expect that each function or set of functions has been estimated and analyzed in a similar context or setting – generating the secondary point estimates and meta-data that are frequently used in benefits transfers. Two distinct ways for integrating data from these multiple sources in a theoretically consistent manner include: (a) the multi-sample unified GMM modeling strategy, and (b) applied dynamic CGE modeling with ecosystem outcomes. These ideas of the links between institutions, evaluation and valuation for ecosystem services were presented at several national and international conferences.
Journal Articles on this Report : 3 Displayed | Download in RIS Format
Other project views: | All 28 publications | 6 publications in selected types | All 3 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Carbone JC, Smith VK. Evaluating policy interventions with general equilibrium externalities. Journal of Public Economics 2008;92(5-6):1254-1274. |
R830923 (Final) |
Exit Exit |
|
Pattanayak SK, Wendland KJ. Nature’s care: diarrhea, watershed protection, and biodiversity conservation in Flores, Indonesia. Biodiversity and Conservation 2007;16(10):2801-2819. |
R830923 (Final) |
Exit Exit Exit |
|
Smith VK, Pattanayak SK, Van Houtven GL. Structural benefit transfer: an example using VSL estimates. Ecological Economics 2006;60(2):361-371. |
R830923 (Final) |
Exit Exit Exit |
Supplemental Keywords:
RFA, Economic, Social, & Behavioral Science Research Program, Scientific Discipline, Air, Ecology, climate change, Air Pollution Effects, decision-making, Atmosphere, Social Science, Economics & Decision Making, benefits transfer, ecosystem valuation, environmental monitoring, decision making, environmental decision making, preference formation, environmental policy, models, water quality value, utility theoryProgress 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.