WEAI/AERE 2009 - Individual Paper Abstract
Title: A Framework for Estimating Willingness-To-Pay To Avoid Endogenous Environmental Risks
Author(s): Yoshifumi KONISHI, South Academic Bldg. 327, Department of Economics, Williams College, Williamstown, MA 01267, yoshifumi.konishi@williams.edu, 413.597.4766; Kenji Adachi, Department of Applied Economics, University of Minnesota
Abstract:
Information about individual averting behaviors is often used to make inferences about willingness-topay (WTP) for health, safety, or environmental quality (e.g. [3], [7], [11], [12]). This approach relies on the earlier theoretical result [1], [2], [6], [16] that the marginal benefit of exogenous pollution reduction equals the cost of self-protection, translated via the marginal rate of technical substitution between pollution reduction and self-protection. Using a continuous-state stochastic model, however, Shogren and Crocker [17] countered this claim, arguing that "unobservable utility terms cannot be eliminated from marginal willingness-to-pay expressions, implying that empirical efforts which identify marginal rates of substitution with willingness-to-pay are misdirected" ([17], p.13).
Konishi and Coggins [10] recently added another dimension to this important discussion, based on the two commonly observed findings from recent empirical studies: (i) consumers are often heterogeneously and imperfectly informed about ambient environmental risks and (ii) WTP estimates are often highly positively correlated with averting behavior (e.g. [8], [15], [18]). One of their important results is that if the marginal return to self-protection increases with the ambient environmental risk, then consumer's WTP should exhibit an inverse relationship to observed self-protection choice. Our goal in this paper is to propose a general empirical strategy to estimate WTP for exogenous environmental-risk reductions, reconciling this gap between the empirical and the theoretical findings.
Our approach consists of a contingent valuation survey design, which incorporates both revealed and stated preference data, and a standard econometric technique, known as Heckman's "dummy endogenous variable" model [9]. The survey must provide (i) a series of questions that elicit the subject's self-protective behavior, (ii) information about the relationship between self-protection activities and health risks, and (iii) a series of contingent-valuation questions. An important step is to include a reminder about a list of private risk-mitigating activities and how these activities, if properly used, reduce personal exposure. The key to the successful survey design is to pay careful attention to framing issues and to properly inform the subjects of all aspects of the contingent commodity that we intend to evaluate--i.e. the welfare value of public risk reduction given self-protection actions.
Econometrically, the key issue in estimation is how to take care of the endogeneity of self-protection in the WTP equation. In the literature (e.g. [4], [5]), a variable describing self-protection activities is sometimes included in the WTP equation, but is often found to insignificantly affect WTP. This result occurs presumably because of the covariates and the errors that simultaneously affect both WTP and selfprotection. To correctly account for correlation between self-protection and WTP, we employ Heckman's dummy endogenous variable model. This empirical model can be augmented to incorporate (i) doublebound dichotomous-choice data, (ii) yea-saying and anchoring effects, and (iii) ordered or multinomial self-protection choice data.
We applied our empirical approach to the data set obtained by Cho, Konishi, and Easter [5] from a contingent valuation survey on arsenic contamination in drinking water. We found that after controlling for this endogeneity, self-protection is significantly negatively associated with WTP. That is, those who take private protective actions are willing to pay less for public risk mitigation efforts. We also evaluate the new arsenic rule of 2000 to demonstrate the significant distributional implications of our approach compared to the conventional approach.
Until recently, literature has been silent as to how valuation analysts might estimate the welfare value of exogenous pollution risk reductions in the presence of heterogeneous consumers taking endogenous selfprotection activities. Neidell [13], however, estimated the effect of ozone pollution on asthma hospitalization controlling for endogenous averting behavior. He found that accounting for averting behavior drastically increased the estimated effect of ozone for children and the elderly, implying that conventional doseresponse estimates were biased downward. Our results are consistent with this finding, but our approach is dissimilar on two important accounts. First, we combine contingent valuation data with (reported) averting behavior and use Heckman's approach to control for endogeneity while Neidell [13] uses reduced-form regression with air quality information, rather than avoidance behavior, as an independent variable. With insufficient data on avoidance behavior, Neidell was unable to recover the structural parameter on averting behavior. Second, our approach allows us to directly estimate consumer's willingness-to-pay for reduction of environmental risk, instead of the health effect of risk.