The market value of reducing cancer risk: hedonic housing prices with changing information.

• Author: Gayer, Ted

1. Introduction

   Market evidence on the value of a statistical life invariably consists of cross-sectional evidence on risks and prices or wages at a point in time. If information about the level of risk changes over time and people incorporate this information in a rational manner, there will be a corresponding price response. On the basis of these temporal changes, one can estimate the market price-risk tradeoff, eliminating many confounding time-invariant effects that cannot be controlled for using cross-sectional data. This paper examines the market response to the release of government information about the level of risk at hazardous waste sites and provides insight into the rationality of this response.

   Studies based on experimental evidence and survey data often find that individual beliefs may deviate from objective risk levels. (1) People often overestimate highly publicized risks and mortality risks. In the case of hazardous waste risks, bias in risk beliefs often leads to considerable public reaction and pressure for site cleanups, which may be an inefficient outcome. By using market data, one can examine whether this intense public reaction carries over to contexts in which private money is at stake.

   Government agencies frequently use information provision as a regulatory device, particularly since the advent of the right-to-know movement of the 1980s. Examples of information provision efforts include the Food and Drug Administration's requirement that many prescription drugs include information inserts, the Department of Housing and Urban Development's requirement tat sellers of houses built before 1950 inform buyers about the presence of lead-based paints, and the Environmental Protection Agency's (EPA) requirement that manufacturing facilities report their annual releases of chemicals above a threshold amount for a list of over 600 substances. Such regulations imply a belief that citizens can learn from information about risk and rationally adjust their prior beliefs towards the objective risk level in light of the information. This article provides market evidence to bolster the results of some survey studies of information transfer that suggest that individuals revise their risk beliefs in response t o new information. (2)

   The U.S. federal policy dealing with hazardous waste sites is known as the Superfund program. The most hazardous sites are targeted for cleanup and placed on a government priority roster called the National Priorities List (NPL). Extension of the classic theory of compensating differentials to the housing market implies that environmental disamenities (such as hazardous waste sites) will reduce housing prices (Rosen 1974). The negative impact of hazardous waste sites on the housing prices of nearby residences is well documented. For example, in a previous study (Gayer, Hamilton, and Viscusi 2000), we used an analysis that focused on the sale of 16,928 houses from 1988 to 1993 that surround Superfund sites in Greater Grand Rapids, Michigan. We found that before the EPA released its risk report, a reduction in the cancer risk from neighborhood Superfund sites by the mean level of risk would increase the average value of a house by $238 (in 1996 dollars). (3) To estimate this willingness of residents to pay to a void cancer risks before the release of the EPA's risk report, we assumed that residents' prior beliefs were equal to the objectively measured risks suggested by the report. Thus, this analysis was based on the very strong informational assumption that residents could, in effect, predict the results of the EPA's site-specific risk assessments. Before the release of the report, the price-risk tradeoff implied a value of a statistical cancer case of $51 million. Once the EPA released its risk information, the implied value of a statistical cancer case was $4 million. These results suggest that after the release of the EPA's risk report, revealed preferences for avoiding Superfund risks were consistent with surveys of the value of a statistical life in the labor market.

   In this paper, we examine how residents respond to information about Superfund risks by explicitly formulating how risk beliefs may change with the release of the EPA's site-specific information about risk levels. We do not assume that people know the site-specific risks before the release of the EPA's studies providing estimates of the cancer risk levels. Rather, we make the more realistic assumption that people base their estimates of site risks on their general knowledge of Superfund sites. The two reference points we use involve equating residents' prior beliefs about site risks with estimates of the average cancer risk level at Superfund sites nationwide and across the state. Once the EPA's information about the site-specific risk level is released, our model incorporates this objectively estimated risk level into residents' posterior risk beliefs.

   Because our focus is on changes in risk over time, we focus on the subsample of houses sold more than once in the Greater Grand Rapids area from 1988 to 1993. This repeat sales methodology allows us to avoid some econometric problems, such as omitted-variable bias, because all time-invariant effects drop out of the analysis. The estimates consequently isolate whether price effects vary over time as risk beliefs change.

   The average cancer risk for Superfund sites throughout the country and throughout the state is greater than the average risk for the Greater Grand Rapids sites. Consequently, if residents are basing their priors on their general knowledge of Superfund sites, and if they update their beliefs after the site-specific EPA risk data are released, then their risk beliefs will decline after the informational release. We would then expect housing prices to increase. If news stories about sites convey information that also causes a reduction in risk beliefs, then we would expect housing prices to rise with newspaper coverage. We find both of these effects in the analysis of our repeat sales sample. When other factors are controlled for, prices for housing sales after the release of the EPA site-specific risk data are higher. The implied value of a statistical case of cancer is between $4.3 and $8.3 million, depending on whether one assumes prior risk beliefs were based on national average risk levels or statewide aver age risk levels. The similarity between these values and the estimates of the value of a statistical life from other market contexts, such as that of the labor market, suggests that the residents react to risks from hazardous waste sites in ways that closely parallel their reaction to other risks. The finding that newspaper coverage increases housing values further suggests that the circulation of information from private sources about sites can cause people to lower their estimates of risk.

   Some other hedonic studies examine whether the price-distance (rather than the price-risk) gradient changes over time, as information changes. In contrast to our finding, Kiel (1995) found that the price-distance tradeoff did not diminish either after the announcement of the cleanup or after the beginning of the cleanup of two Superfund sites in Woburn, Massachusetts. This is not necessarily inconsistent with our results, since for these sites it may be the case that the new information conveyed by the cleanup announcement and action indicated that the risks from the sites were higher than the baseline perception. Michaels and Smith (1990) found a price-distance effect after the announcement of hazardous waste risks at a site, although their results vary by housing submarkets. (4)

   This paper improves upon our earlier study of hazardous waste risks in at least three ways: In the present study, we (i) explicitly model how risk beliefs may change by positing that prior beliefs are based on general knowledge of hazardous waste sites rather than assuming that risk beliefs do not change, (ii) use a repeat sales methodology that helps avoid omitted-variable biases in the hedonic analysis, and (iii) estimate the value of a statistical cancer case given various assumptions of discount rates and latency periods.

   Discussions of risk policy often assume that individuals cannot accurately perceive risks or that they fail to update their beliefs in the face of new information. This paper provides evidence that individuals can reassess their beliefs of the risks at hazardous waste sites and do spend their own resources to avoid these risks, reflecting tradeoffs that are similar to choices about safety and risk in other markets. In section 2, we describe the repeat sales model specification, which we then link to a risk-learning model in section 3. Our data are described in section 4, and results and conclusions are provided in sections 5 and 6, respectively.

2. The Repeat Sales Estimation Model

   The Hedonic Price Function

   The empirical framework we use to assess the effect of changes in risk levels is the hedonic property model, which postulates that housing prices are a function of structural, neighborhood, and environmental characteristics. (5) The environmental characteristics of a house are the perceived environmental risks associated with living in the house. The price of house i sold at time t is given by

   [Price.sub.it] = f([Structural.sub.i], [Neighborhood.sub.i], [Environment.sub.it], t), (1)

   where [Structural.sub.i] is a vector containing the structural characteristics, [Neighborhood.sub.i] is a vector of the neighborhood characteristics, [Environment.sub.it] is the perceived environmental risk to the household, and t, indicates the year of the sale. Notice that the environmental risk belief variable is subscripted with t, indicating that it varies over time. The model assumes that the structural and neighborhood characteristics are unchanged for each house across sales. (6) Residents have a learning function in which they update their prior...