Smoking status and public responses to ambiguous scientific risk evidence.

• Author: Viscusi, W. Kip

1. Introduction

   Most risks that we face are not known with precision. The risks posed by climate change, the cancer risks created by breast implants, the potential for adverse reactions to a new drug, and the threat of mad cow disease are all highly uncertain. For example, one prominent British scientist offered the rather imprecise risk judgment that the human form of mad cow disease, or Creutzfeldt-Jakob disease, would kill from 500 to 500,000 British consumers.(1) Such uncertainty is the norm rather than the exception. Substantial ambiguity exists regarding the extent of the hazard even for risks for which there is substantial agreement, such as the hazards of cigarettes. The extent of the health risks posed by environmental tobacco smoke remain hotly debated, as some studies indicate substantial risks and others fail to indicate any significant risk.

   How do people respond to situations in which there is a diversity of opinion regarding the level of the risk? Ideally, they should make risk inferences in a rational manner; where we will take Bayesian learning as the rationality reference point. When there are competing scientific risk judgments, how people process the information will depend on their prior beliefs, the precision of their prior beliefs, the risks indicated by the scientific studies, and the weight they place on these studies. There are many ways in which people might use such scientific studies, such as by averaging their implications or placing a higher weight on the more credible study. Some of these changes in risk beliefs in response to new information will be consistent with rational learning, whereas others may not. In this paper, we present original survey evidence in which respondents consider a variety of information scenarios, making it possible to assess the consistency of the treatment of different scientific risk studies.

   To the extent that behavioral predictions are possible based on evidence in the literature, it would suggest that people may exhibit anomalous behavior when there is ambiguity regarding the magnitude of the risk. Consider, for example, the implications of the classic Ellsberg (1961) paradox. In that experiment, subjects considered two urns, one of which offered a known probability of winning a prize, whereas the other offered an uncertain chance of winning a prize. Respondents were averse to ambiguous chances of winning the prize, compared to a lottery ticket with precisely known probabilities and the same mean probability value. Researchers have identified similar phenomena with respect to small probabilities of a loss, because there is often a tendency toward ambiguity aversion that is not consistent with rational expected utility theory.

   The issues addressed in this paper represent a variant on this ambiguous risk structure. Moreover, we have coupled the presence of risk ambiguity with different risk information sources, whereas in earlier studies, such as Viscusi and Magat (1987), there is no difference in the source's identity. Suppose that there are two different parties providing risk information, and their views are different. How do people process the divergent risk judgments? Consider the situation of an industrial polluter, which might be expected to have a vested economic interest in downplaying the risk or even misrepresenting it. To what extent does it matter whether the risk evidence is from industry or government? When is individual differentiation between those risk sources entirely rational, and is there any reason to believe that people fail to process diverse risk information rationally?

   The substantive focus of the study will be on how people respond to different assessments of the risk of cancer due to air pollution. Cancer is a chief source of individual risk that is often ambiguous. The use of an air pollution context enables us to consider risk estimates in the policy arena, including both business and government. In each case, respondents will consider two different sources of risk information about the potential cancer risk. These informational sources could be both government or industry sources, or possibly a mixture of industry and government sources. Because the polluting firm will have less of an incentive to reveal accurate risk information about the hazards it generates than might the government, an interesting economic issue is how people will process these divergent risk judgments depending on their source and the risk levels associated with them. Unequal weighting of such scientific evidence may be entirely rational. In the case of our study, however, respondents will consider a variety of different informational combinations. Ideally, people should be consistent in how they weight risk evidence across different risk scenarios.

   The key issue we explore here is how risk judgments and relative weighting of diverse scientific evidence varies with personal characteristics. Demographic factors may be consequential. Women, for example, have shown themselves to be more averse to facing health hazards than are men.(2) A personal characteristic of particular saliency is individual smoking status. Do people who have self-selected themselves into this extremely dangerous consumption activity process risk information differently than do nonsmokers with similar demographic profiles? The analysis of the distinctive behavior of cigarette smokers with respect to this informational context is of interest since cigarette smoking involves a risk for which there are continuing battles between the cigarette industry and government officials focused on the hazards of environmental tobacco smoke.

   The organization of this paper is as follows. Section 2 outlines the theoretical structure for the study and the empirical formulation of the model that will be tested. Section 3 presents the empirical estimates of the effect of the different risk information sources and the different risk levels on individual risk judgments. Section 4 focuses on the role of smoking status and its effect on the processing of risk information. As indicated in the concluding section 5, individuals do not treat the different kinds of information based solely on the credibility of the source. Rather, they tend to place disproportionate weight on the high risk assessment, particularly when there are conflicting risk judgments. Cigarette smokers are less likely to exhibit such alarmist behavior.

2. Theoretical Structure and Empirical Framework

   Theoretical Structure

   The basis for the study is a survey of adults in which they considered moving to one of two areas, area 1 or area 2. Each area posed a risk of cancer, but they differed in terms of how the risk was characterized. Area 1 posed an uncertain cancer risk level S for which there were two divergent risk judgments involving industry, government, or a mix of sources. Area 2 has cancer risk level R that is estimated with precision and will thus serve as the reference point for the level of the precisely understood risk that is equivalent to uncertain risks. The survey proceeded iteratively using an interactive computer program until respondents identified the known risk R in area 2 that they believed was equivalent to the ambiguous risk in area 1.

   The advantage of considering a move to an undefined new area is that respondents will not bring to bear theft prior beliefs regarding theft current risk situation. Each area was, in effect, an abstraction for which there was only survey information regarding the area's chemical hazards based on two different studies. The study could consequently be a pure informational experiment in which all that transpired was an effort by respondents to find a known risk value that was equivalent to an area in which there were two studies that differed in terms of the implied risk level.

   The nature of the informational context is noteworthy as well. For personal risk-taking activities, such as smoking, some researchers have hypothesized that people may discount risk evidence for a risky activity they have chosen. Air pollution, which is the case study for this analysis, is a hazard outside of the individual's control. This risk is involuntary and inflicted by a polluting industry in which the respondent has no economic interest. Moreover, air pollution is a public hazard that poses similar exposures to all, although the actual risks may differ if, for example, one is asthmatic. The survey design consequently avoids any influence of risk perceptions being influenced by a desire to justify relatively great personal risk taking.

   The two information sources were scientists from the polluting chemical industry (denoted by i) and scientists from the government (denoted by g). Respondents received no information about the nature of the scientific studies or theft timeliness, so there was no reason otherwise to differentiate between them. The Appendix presents sample survey questions.

   If we let U denote utility in the healthy state and V denote utility with cancer, the task is for respondents to find the value of the known risk R in area 2 that gives them the same expected utility in both locations, or

   (1 - S) U + SV = (1 - R)U + RV. (1)

   From the standpoint of the survey structure, respondents are picking the known risk R that satisfies

   R = S. (2)

   Thus, the structure of the utility functions and the character of the rewards structure are not consequential because the U and V terms drop out. For this binary lottery, the task is simply to find the precisely understood probability R that is equivalent to the subjective risk assessment S in terms of the attractiveness of the chance of getting cancer in that area.

   The nature of the study makes the value of the equivalent precise risk R observable, because that precise risk value is simply the cancer in area 2, where the risk is known. By construction, the survey determines the value of R that is equal to the respondent's subjective risk...