Printer Friendly

Evaluating the effect of alternative risk communication devices on willingness to pay: results from a dichotomous choice contingent valuation experiment.


Researchers in the field of economics have been improving and refining methods to quantify the benefits associated with the programs or regulations which reduce environmental and health risks. No explicit competitive market price currently exists for the general population to purchase such risk reductions. Therefore estimation of the benefits from reducing this risk requires an alternative or nonmarket method. In addition to wage hedonic models, one method being increasingly used by researchers and governmental agencies is the Contingent Valuation Method (see Mitchell and Carson 1989 for a comprehensive evaluation of the method). Since the Contingent Valuation Method (CVM) provides hypothetical valuations, validation of CVM responses via hedonic property value studies (Brookshire et al. 1982) and cash markets (Bishop and Heberlien 1979; Welsh 1986) have been necessary to demonstrate that credible valuations can be produced by CVM. However, most of these tests are of consumer choices involving certainty. For valuation of resource trade-offs under uncertainty the robustness of CVM to different risk communication devices has yet to be demonstrated.

A review of recent literature on risk communication and public perceptions of risk showed extensive research being conducted by psychologists, sociologists, political scientists, and economists (Fischhoff 1990; Smith and Desvousges 1988). Difficulties facing risk communicators was the main topic at the National Conference on Risk Communication held in 1986. Several factors were cited as large obstacles to effective risk communication: (1) risk information is often highly technical, complex, and uncertain; (2) experts provide widely different risk estimates; (3) regulatory agencies often lack public trust and credibility; (4) there are various ways to define risk; (5) strong beliefs held by the public are resistant to change; and (6) many people have difficulty with probabilistic information (Davies, Covello, and Allen 1987). Most of the past comparisons of the effect of risk communication on behavioral choices have focused on alternative narrative descriptions of risk changes and associated contexts (Tversky and Kahneman 1981). The purpose of this paper is to evaluate CVM responses arising from two commonly used graphical risk communication devices: the risk ladder and risk circles/pie charts. Each of these devices have been used independently to elicit values of risk reduction programs, but the two methods have not been empirically compared for the same magnitude of risk reduction and for the same type of hazard. To allow for progress in value elicitation for risk reduction programs, understanding the implications of using different risk communication devices is very important.


There are three key CVM design elements that must be coordinated in any survey instrument: (1) the good to be valued, (2) the value elicitation procedure, and (3) the payment vehicle. In this study, the commodity or good to be valued is reduction in risk of premature death through state-financed incentives for a California program of hazardous waste minimization by private industry. The public provision of a statewide program and the fact that in California the funding of these types of programs is often decided via a popular referendum made the dichotomous choice referendum elicitation format quite credible. The use of a voter referendum elicitation format (where people vote yes or no) made taxes the most logical or credible payment vehicle. Unfortunately, the payment of higher taxes is not an emotionally neutral subject for many people and such a payment vehicle may increase the number of protest responses. However, it is the realistic possibility of this particular payment mechanism that motivated selection of this method.

In terms of value elicitation procedures, most of the previous surveys on WTP for reductions in risk have used open-ended WTP questions (Smith and Desvousges 1987; Magat, Viscusi, and Huber 1988) or payment cards (Gerking, DeHann, and Schulze 1989). Providing a specific maximum dollar amount for a nonmarket product can sometimes be difficult for respondents and in mail surveys they frequently skip these questions. A dichotomous choice format is easier for the respondent to answer and has been shown to be more incentive compatible for unbiased responses (Hoehn and Randall 1987).

Although the dichotomous choice procedure does not directly provide the maximum WTP for risk reduction by households, there are statistical inference techniques to estimate maximum WTP from data on the probability of a YES or NO response to specific dollar amounts. The probability a respondent will answer "YES" to the WTP question is assumed to be related to the expected gain in well-being obtained from receiving the health risk reduction, over and above the satisfaction lost due to paying higher taxes (Hanemann 1984).

To be more specific, assume a state-dependent utility function (Cook and Graham 1977) such that |U.sub.L~ and |U.sub.D~ are the utility when alive and dead, respectively. Following Smith and Desvousges (1987, 91) this state-dependent utility depends, in part, on income (Y). Further let |P.sub.D~ be the baseline probability of premature death. Baseline expected utility(1) (EU) can be defined as:

EU = |P.sub.D~||U.sub.D~(Y)~ + (1 - |P.sub.D~)||U.sub.L~(Y)~. |1~

The proposed hazardous waste minimization program reduces the probability of premature death from |P.sub.D~ to |P|prime~.sub.D~ but at a proposed cost to the respondent of $X each year. If the reduction in the probability of premature death from |P.sub.D~ to |P|prime~.sub.D~ yields more expected utility than the loss of $X in income, the person will answer YES to the dichotomous choice question. Specifically the expected utility difference (EUD) is given by:

EUD = {|P|prime~.sub.D~||U.sub.D~(Y - $X)~ + (1 - |P|prime~.sub.D~)||U.sub.L~(Y - $X)~} - {|P.sub.D~||U.sub.D~(Y)~ + (1 - |P.sub.D~)||U.sub.L~(Y)~}. |2~

If this expected utility difference is linear in its arguments and the associated additive random error term is distributed logistically, then the probability a respondent will answer YES to a question asking him or her to pay $X for a program that would reduce the risk of premature death from |P.sub.D~ to |P|prime~.sub.D~ would be:

P(YES) = 1 - ||1 + |e.sup.|B.sub.0~-|B.sub.1~($X)~~.sup.-1~. |3~

Maximum likelihood routines can be used to estimate a transformation of equation |3~ in the form of:

Log{P(YES)/|1 - P(YES)~} = |B.sub.0~ - |B.sub.1~($X). |4~

Estimates of the parameters |B.sub.0~ and |B.sub.1~ allow identification of the cumulative distribution function of WTP for the risk reduction program (Hanemann 1984). The mean of the cumulative distribution function is the mean WTP. Since WTP for an unambiguous improvement in expected utility is nonnegative, the mean is given by Hanemann (1989) as:

WTP = (1/|B.sub.1~) *ln(1 + |e.sup.|B.sub.0~~). |5~


The effect of the amount and type of information on consumer's choices has been a source of concern among economists and psychologists for a number of years. Unlike consumer choice under certainty, probabilistic and uncertain events appear to tax the decision-making capability of many consumers. Viscusi and Magat (1987, 7) in summarizing their own research on product labeling and that of others with regard to risk communication state: "The existence of limitations on human cognitive capabilities makes the format and wording of labels particularly important."

There has been much innovation in risk communication devices over the short history of contingent valuation of health-related risk. Jones-Lee, Hammerton, and Phillips (1985, 53) used darkened blocks on graph paper to portray the risk in 100,000 of death from transportation accidents. Risk ladders have been used by Mitchell and Carson (1985) as well as Gerking et al. (1989). In these ladders each rung represented progressively higher and higher risks.

As part of their effort to provide context on risk of death, Smith and Desvousges (1987) used both a risk ladder and pie chart to communicate information on risk. The ladder arrays different probabilities of death from a variety of sources, with the most hazardous at the top. Smith and Desvousges used this risk communication tool primarily to communicate the relative risk from hazardous waste as compared to other risks. Their ladder mixed voluntary risk (e.g., smoking) with involuntary risk (e.g., floods).

To actually communicate the change in probability of death associated with the particular programs they were asking WTP about, Smith and Desvousges used a series of three pie charts. The three pie charts were as follows: the first pie chart illustrated a typical individual's risk of exposure to the hazardous substance. This was done by shading in a pie slice equal to the probability of exposure (i.e., if chances are 33 percent, then one-third of this pie would be shaded in). The second pie chart portrayed the risk of death if exposed to a given dose of the hazardous substance. The third pie chart illustrated the combined (multiplied) results of the first two pie charts. This pie chart was entitled "Combined Risk: Exposure and Death." Generally the size of the darkened slice got smaller as one read left to right. In essence this third chart represented the outcome of a compound lottery. People were asked to pay for a reduction in the risk of exposure, shown as a smaller shaded area in the left most pie. Holding constant the middle pie chart (risk of death if exposed) the program people were asked to pay for reduced the amount of the third darkened pie slice (combined personal risk).

Smith and Desvousges (1987, 96) indicate the separation of risk into three pies was an outcome of several workshop discussions with small groups of the public (called "focus groups") regarding alternative methods for communicating risky situations. The consensus of those focus groups was that having separate pie charts for risk of exposure and risk of death if exposed made it easier for individuals to relate government environmental regulations to changes in risk. While this is certainly an advantage over the ladders, there seems to be several potential drawbacks to relying on the pie charts to elicit WTP for risk reductions as compared to directly using the risk ladder. Perhaps the most important is that for very low risks, it is difficult for people to: (1) relate the small size of the darkened slice to their relative chances of premature death from this hazard compared to more familiar hazards; (2) given the small baseline risk or darkened area it may be difficult to portray noticeable changes in the third pie chart for most reasonable programs. This lower effectiveness to communicate visually perceptible differences in reductions in risk levels would tend to result in respondents giving about the same responses across the risk levels. Since the risk reduction is being displayed without any comparison to familiar risks, it is more difficult for the individual to directly judge how much safety they have bought in terms that are directly related to their life experience (and other risks they face). That is, if one wants the marginal rate of substitution between income and risk, it may be helpful to show how the new market basket of risks compares to the old. The risk ladder does a better job of this. Lastly, Smith and Desvousges presented the three related pie charts in the form of a compound lottery. While this provides additional information, it may be confusing to people not use to thinking in these terms. This may provide unnecessary detail as the risk ladder only presents information on the combined probability.


To test the relative effectiveness of the risk ladder and pie charts for eliciting valuations, two versions of a survey were developed. These versions were identical except for the method used to convey risk information. One method utilized a multi-color risk ladder to show a wide range of involuntary risks and provided a perspective on the size of risk from exposure to heavy metals, relative to the other involuntary risks. This communication technique was utilized by Smith and Desvousges (1987) to display relative risks, but not to display the change in risk for the government program people were asked to value. In contrast, we used this risk ladder to show the reductions in risk level from three alternative programs as movements down the ladder and to elicit WTP responses. Thus the ladder was directly used to provide perspective on how much additional safety was to be purchased relative to other familiar risks. The sides of the ladder were color-coded with high risk shown in red, moderate risks in yellow, and low risk in blue. A black and white copy of the ladder is shown in Figure 1.

An alternative method of communicating risks was patterned after the original pie chart format of Smith and Desvousges. The risk reduction programs were conveyed visually by shading in portions of the three pie charts to depict the level of risk of exposure, risk of death if exposed, and finally the combined personal risk. Figure 2 shows a reduction in personal risk used in the pie chart survey version.(2)


The direct use of the risk ladder for elicitation of WTP may better communicate risks by representing the decrease in risk in terms relative to other experienced risks. In addition, the risk reductions are represented in linear distance rather than an area within a circle. It seems plausible that people can more easily comprehend the magnitude of risk changes when represented as simple lineal distances (which is one dimensional) as compared to area representations of a circle (which involves two dimensions that radiate from small in the center to large at the edge). The main purpose of the research reported here was to compare the effectiveness of risk pies and risk ladders as risk communication devices for eliciting WTP responses for risk reductions. While both techniques have been used to elicit valuations of risk reductions, we are aware of no comparisons of the two techniques for the same level and type of risk.

Our first hypothesis is that the two risk communication devices will yield statistically different logit equations (as in equation |4~). This will be tested in two ways, first is a test of equality of all WTP coefficients. Specifically, the null hypothesis is:

|B.sub.0~ = |A.sub.0~, |B.sub.1~ = |A.sub.1~, |B.sub.2~ = |A.sub.2~, . . . , |B.sub.k~ = |A.sub.k~ |6~

where B's are coefficients from logit equation |4~ estimated from responses obtained using the risk ladder and the A's are coefficients from logit equation |4~ estimated from responses obtained using the pie charts.

The equalities in equation |6~ can be tested using a likelihood ratio (LLR) test. In this case a pooled logit model imposes the restriction in equation |6~ as compared to the unrestricted model which allows the coefficients to be different. The test statistic is computed by comparing the log likelihood function of the pooled sample minus the sum of the two log likelihoods for the unrestricted models. Specifically,

LLR = -2* {LLpooled - (|LL.sub.A~ + |LL.sub.B~)}, |7~

where LL is log likelihood. This statistic is distributed chi-square with k + 1 degrees of freedom.

A more direct, but perhaps more restrictive, test of the two risk communication devices is simply to pool the WTP responses but include a dummy variable indicating device. That is if |B.sub.k+1~ is the dummy then if the two risk communication devices yield equivalent results then |B.sub.k+1~ = 0. We would expect them not to yield equivalent results since the ladder provides more context on relative risk. This test is more restrictive in that it accepts the null hypothesis of coefficient equality in equation |6~, and requires any difference in WTP responses to show up as a parallel shift in the logit WTP function. Nonetheless, this test complements the likelihood ratio test.

Our second hypothesis goes beyond the first to ascertain whether either set of CVM responses vary across risk levels. Specifically, a risk communication device that is effective should produce responses that are consistent with consumer demand theory. In its most general form, this would simply require that WTP be nondecreasing with increases in economic goods. Thus not only is there diminishing marginal valuation, but at some point satiation may result in marginal valuations of zero. However, involuntary risk to one's life and human health, while following the basic tenets of economic theory, are certainly special goods. That is, while diminishing marginal valuation for added years of longevity or improved health is likely, satiation would seem unlikely over the ranges being presented in this survey and in most environmental regulation programs. Therefore we would expect that if a series of substantial reductions in risk is clearly communicated by device X to consumers, we would expect the responses to vary systematically across those risk reductions.

Once again there are two ways to test this hypothesis. The most general is to test whether the coefficients in logit equations estimated for each risk level are statistically different. This forms the alternative hypothesis to the null hypothesis in equation |8~ which states there is no difference in responses across substantial risk levels. Stated in testable form the null is:

|Mathematical Expression Omitted~

where B's are coefficients from logit equation |4~ estimated from responses for three substantially different risk levels (e.g., 25% reduction in risk, 50% reduction, and 75% reduction). The alternative hypothesis is one of statistical difference between coefficients for each risk reduction. As with the first hypothesis, the null hypothesis in equation |8~ can be tested using a likelihood ratio test. This time the restricted likelihood function is computed by pooling observations across risk reduction levels for a given risk communication device. Therefore, the null hypothesis in |8~ will be tested for each of the two risk communication devices.

The second and more direct way to test whether the WTP changes with level of risk reduction is to simply include the absolute level of risk as one of the variables in the logit WTP function. This can be accomplished by pooling responses across risk levels and including a variable for the absolute level of risk reduction. The null hypothesis is the |B.sub.riskreduction~ = 0 if the device does not communicate risk effectively. This null hypothesis is tested twice; once for the risk ladder and once for the pie chart.

Our last hypothesis is even stronger than the first two and states the percent of respondents answering YES to any given risk reduction will be lower when communicated using the pie chart as compared to the risk ladder. Specifically, the null hypothesis is equality of percentage YES's (PRY) for any given risk reduction:

|PRY.sub.A~ = |PRY.sub.B~ |9~

with the alternative hypothesis being

|PRY.sub.A~ |is less than~ |PRY.sub.B~ |10~

for the reasons cited above (i.e., people will more easily perceive the size of the risk reductions and can relate them to other familiar risks more easily with the risk ladder, treatment B). The hypothesis in equation |9~ can be tested using a difference of means test which is distributed with a Student's t-distribution.


A full-sized (8 1/2" x 11") multi-color mail survey instrument of twelve pages in length was divided into three sections. The main section of interest here relates to risk communication and elicitation of WTP. This section was designed to accomplish three important functions within the survey.

This section provided information about pathways of exposure to hazardous wastes from various contamination sources. The information was presented in written form and in a full-page drawing. The respondent received a description of a current hazardous waste minimization program in California and how the risk of exposure to hazardous wastes could be reduced by greater funding to this program.

Besides explaining the pathways of exposure and a mechanism for reducing the risks, the second section contained risk communication devices to convey risk levels. The current risk of premature death from exposure to heavy metals was communicated in narrative and illustrated in either the risk ladder or the pie charts depending on the version of the survey to help provide greater comprehension of the risk magnitudes.

The final task accomplished in the second section was the elicitation of WTP responses. Respondents were asked three WTP questions, one for each size risk reduction program (25%, 50%, 75%). Before answering the WTP questions respondents were told to consider only the value to their household from the reductions in risk of exposure to heavy metals. The value elicitation procedure used in the survey was a close-ended referendum format WTP question to specific dollar amounts.

The risk communication devices and initial bid amounts were pretested with a combined telephone-mail-telephone approach. A sample of 200 households were contacted by phone and told they would receive a survey in the mail. They were to fill out the survey and at a mutually agreed upon time, the interviewer would call them back to obtain their answers and discuss comprehension of the survey elements.


The survey design and administration procedure followed the basic outline of Dillman's "total design method" (Dillman 1978). The survey was in booklet form. It was typeset and was designed by a professional artist. A total of 2,000 booklets (1,000 for each version) were mailed along with a personally addressed cover letter and a postage paid return envelope. A reminder postcard was sent out the week following the initial mailing. Finally a second mailing of the booklet with a new, more emphatic cover letter was sent to people who had not yet returned their survey. The mailing was to a random sample of California households. The response rates of 43 percent and 47 percent were obtained for the risk ladder version and the pie chart version, respectively. This yielded a final sample for each version of 374 and 413 surveys. The undeliverable surveys and deceased persons were omitted for the purpose of response rate calculations. These response rates are somewhat lower than other CVM surveys, although nearly identical to what we have obtained in other CVM mail surveys of general public in California using only the postcard and second survey follow-up (Loomis 1987). This response rate could be increased closer to what others have obtained by using monetary incentives or a third certified mailing.(3) However, the most important concern for methodological comparisons between risk communication device is having a similar response rate, rather than the absolute level of the response rate (which is more important for generalizing the sample responses to the population).

Table 1 presents a comparison of sample characteristics for the risk ladder version and the pie chart version. In general the samples are quite similar. Education and political orientation are nearly identical. The percentage that own their house and mean household income are very close. Because of the similar response rates and characteristics of the two samples, we conclude that any differences in the responses are due to risk communication device, not sample differences.


                                   Risk         Pie
Variable                          Ladder       Chart

EDUCATION                          14.98        15.03
OWN HOUSING                         0.74         0.80
CHILDREN                            0.45         0.36
INCOME                           $48,754      $49,328
POLITICAL ORIENTATION               3.24         3.27
(1 = liberal, 3 = middle of
road, 5 = conservative)


Table 2 provides the logit equations for three risk reduction programs (25% reduction, 50%, and 75%) for each of the two versions of the survey. The same specifications of the logit equation is used for both risk communication devices so as to perform the likelihood ratio test of hypothesis number one. The coefficient on dollar bid is significant in all six logit equations at the .01 level. For nearly all of the regressions across risk levels, the variables have intuitively appealing signs. For example, the more important other community problems are relative to environmental issues (e.g., the Other Problems variable in Table 2), the lower the probability of paying a given dollar amount to reduce hazardous waste. The higher level of education and income a respondent had, the more likely the individual was to pay a given dollar amount. The pseudo |R.sup.2~ for both risk communication devices indicate similar goodness of fit for both approaches.

Table 3 presents statistical results for the two logit equations pooling responses across risk levels for the ladder and pie charts but that include a variable for absolute level of risk change. As before, coefficients have intuitive signs. The absolute level of risk reduction (RISKREDUC) is significant at the .1 level for the risk ladder and .05 level for the risk pies.

Table 4 presents the statistical results when responses to the two risk communication devices are pooled together (and across risk levels). As before, all variables TABULAR DATA OMITTED have the intuitive signs. The risk communication device dummy (RISKCOMDUM) is significant at the .01 level.


Do the Alternative Risk Communication Devices Yield Different Results?

As discussed earlier, a likelihood ratio test is used to assess whether the logit equations associated with the two risk communication devices resulted in equality of the intercept and slope coefficients. Using equation |7~ the likelihood ratio test across the two risk communication devices is computed and reported in Table 5. As is evident in Table 5, the computed likelihood ratio statistics (which are distributed chi-square with k + 1 degrees of freedom) for Hypothesis #1 are statistically different from zero at the 5 percent significance level for the 25% and 75% reduction level. This indicates the two different risk communication devices generally yield statistically different logit equations.

Consistent with this result is the result shown in Table 4: the shifter dummy for risk communication device is significant at TABULAR DATA OMITTED the .01 level. The dummy indicates that the pie charts result in a lower WTP (since RISKCOMDUM = 0 for the pie charts).


Variable                  Coefficient       t-Stat

CONSTANT                  -0.72900          0.868
OTHER PROBLEMS            -0.65118         -4.815
INCOME                     0.0000018        1.754
EDUCATION                  0.11631          5.573
BID                       -0.00298         -9.035
RISKREDUC                  0.00171          2.667
RISKCOMDUM                 0.62374          5.533
Log Likelihood                           -936.08
Pseudo |R.sup.2~                            0.148

Do WTP Responses Change with Risk Levels?

Our second hypothesis is that the responses elicited using the risk ladder would vary across risk reductions and the pie chart responses would be less likely to do so. Results from the Likelihood Ratio Test reported in Table 5 indicates that the coefficients TABULAR DATA OMITTED in the logit WTP functions are not statistically different across the three risk levels using either the ladder or the pie charts.

Since the likelihood ratio tests showed the logit coefficients were equal across the three risk levels, our second test is quite appropriate and no longer restrictive. This second test involves the significance of a variable for the absolute level of risk reduction. Specifically, does the absolute level of risk reduction systematically influence WTP responses, everything else held constant? As shown in Table 3, the answer is yes at the .1 level for the risk ladder and yes at the .05 level for the risk pie charts.

Our third hypothesis test of the equality of the probability of responding YES across survey versions is tested using a difference of means test. The difference of means statistic follows a Student's t-distribution. While the bid distribution on surveys mailed out were identical, the mean bid or offer amounts of the returned surveys differed slightly (by less than 7 percent). For example the difference at the 50 percent risk reduction level is only $10 (i.e., $224 and $234).

The results of this test are also shown in Table 5. The null hypothesis Ho: |PRY.sub.A~ = |PRY.sub.B~ is rejected at the 1 percent significance level for the 25% and 75% reduction programs. The null hypothesis of equality of percentage YES for the 50% reduction programs are different at the 5 percent significance level. As is shown in Table 5, support is given for the alternative hypothesis that |PRY.sub.A~ |is less than~ |PRY.sub.B~ (i.e., greater proportion of YES from the risk ladder).


It is well known in psychology and marketing research that alternative methods of conveying information can have profoundly different impacts in terms of a consumer's interpretation and perceptions of the content. In our survey, version B which moved respondents down the risk ladder, was expected to yield perceptions of larger risk reductions than changes in the small slivers in the pie charts of version A. In addition, the elicitation of WTP within the context of other relative risks respondents face provides potential for greater understanding of just what that change in risk means to them in their everyday lives.

We find that an intercept dummy for risk communication device is statistically significant at the .01 level. As expected, the risk ladder version results in a significantly larger number of people stating YES they would vote in favor of the waste minimization program. Nonetheless, each device was capable of yielding WTP responses that moved in a systematic fashion with the absolute level of risk reduction (as evidenced by the statistically significant coefficient on the absolute risk reduction variable). The confirmation of this for both the risk ladder and the pie charts is encouraging.

Which of these risk communication techniques yields values closer to the "truth" is of course difficult to determine in contingent valuation. Both provide WTP responses that move in a theoretically correct manner with risk reduction. The ladder does a better job providing information on relative risk, i.e., how the risk under study compares with other, often time, more familiar risks. This aids the respondent in forming their marginal rate of substitution between risks and other goods (as represented by income). As developed by Smith and Desvousges (1987) (and used in this survey) the compound lottery depiction of the pie charts appeared to do two things well: (1) communicate how the program would affect the risks facing the respondent (i.e., the program controlled just the risk of exposure and hence death) and (2) the pie charts appeared to do a reasonable job for communicating the absolute level of risk reductions as well. Given the empirical evidence presented in this paper, either device seems adequate for communicating risk reductions in the elicitation portion of a CVM survey. Before this conclusion can be generalized, replication for a wide variety of risk levels is certainly desirable. Of course, psychologists, economists, and other scientists should continue to improve upon these risk communication devices and ideally develop new, even more effective ones.

1 We have purposely chosen a direct and standard formulation of the consumer's choice process under uncertainty. In part this is to maintain consistency with Smith and Desvousges (to whom we are comparing our results) and in part to provide the most direct linkage between the definition of WTP and the utility difference approach underlying our dichotomous choice CVM approach. For alternative and more general representations of consumer decision making in the face of uncertainty the interested reader should see Machina (1987, 132-36) and Smith and Desvousges (1988, 1114-15).

2 A commentor on the earlier version of this paper presented at the W-133 meeting pointed out that perhaps a more comparable test of the risk ladder and pies would have just compared the last of the three risk pies at each risk level. This would have avoided the possibility that any differences between risk communication devices was due, in part, to the compound lottery effect. However, since our goal was to compare the risk ladder and pie charts that were as similar as possible to Smith and Desvousges, we desired to keep the three pie charts.

3 Due to budget limitations it was not possible to perform the third certified mailing. However, in a slightly different version of this survey that focused on risk from solvents we included a one dollar bill with the first mailing. The overall response rate to this survey with just the postcard and second mailing was 64 percent, more in line with what others have obtained elsewhere.


Bishop, Richard, and Thomas Heberlein. 1979. "Measuring Values of Extra-Market Goods: Are Indirect Measures Biased?" American Journal of Agricultural Economics 61 (Dec.):926-30.

Brookshire, David, William Schulze, Mark Thayer, and Ralph d'Arge. 1982. "Valuing Public Goods: A Comparison of Survey and Hedonic Approaches." American Economic Review 72 (Mar.):165-77.

Cook, Philip, and Daniel Graham. 1977. "The Demand for Insurance and Protection: The Case of an Irreplaceable Commodity." Quarterly Journal of Economics 92 (Feb.):143-56.

Davies, J. Clarence, Vincent T. Covello, and Frederick W. Allen, eds. 1987. Risk Communication: Proceedings of the Conference on Risk Communication. Washington, DC: The Conservation Foundation.

Dillman, Don. 1978. Mail and Telephone Surveys. New York: John Wiley and Sons.

Fischhoff, Baruch. 1990. "Understanding Long-Term Environmental Risks." Journal of Risk and Uncertainty 3 (Dec.):315-30.

Gerking, Shelby, Menno DeHann, and William Schulze. 1989. "The Marginal Value of Job Safety: A Contingent Valuation Study." Journal of Risk and Uncertainty 1 (June):185-99.

Hanemann, W. Michael. 1984. "Welfare Evaluations in Contingent Valuation Experiments with Discrete Responses." American Journal of Agricultural Economics 67 (Aug.):332-41.

-----. 1989. "Welfare Evaluations in Contingent Valuation Experiments with Discrete Response Data: Reply." American Journal of Agricultural Economics 71 (Nov.):1057-61.

Hoehn, John, and Alan Randall. 1987. "A Satisfactory Benefit Cost Indicator from Contingent Valuation." Journal of Environmental Economics and Management 14 (Sept.):226-47.

Jones-Lee, M., M. Hammerton, and R. R. Phillips. 1985. "The Value of Safety: Results from a National Survey." Economic Journal 95 (Mar.):49-72.

Loomis, John. 1987. "Expanding Contingent Valuation Sample Estimates to Aggregate Benefit Estimates: Current Practices and Proposed Solutions." Land Economics 63 (Nov.):396-402.

Machina, Mark. 1987. "Choice Under Uncertainty: Problems Solved and Unsolved." Journal of Economic Perspectives 1 (Summer):121-54.

Magat, Wesley, W. Kip Viscusi, and Joel Huber. 1988. "Paired Comparison and Contingent Valuation Approaches to Morbidity Risk Valuation." Journal of Environmental Economics and Management 15 (Dec.):395-411.

Mitchell, Robert, and Richard Carson. 1985. "Valuing Drinking Water Risk Reductions Using the Contingent Valuation Method." Draft Report to U.S. Environmental Protection Agency. Resources for the Future, Washington, DC.

-----. 1989. Using Surveys to Value Public Goods: The Contingent Valuation Method. Washington, DC: Resources for the Future.

Smith, V. Kerry, and William Desvousges. 1987. "An Empirical Analysis of the Economic Value of Risk Changes." Journal of Political Economy 95 (Feb.):89-114.

-----. 1988. "Risk Perception, Learning and Individual Behavior." American Journal of Agricultural Economics 70 (Dec.):1113-17.

Tversky, Amos, and Daniel Kahneman. 1981. "The Framing of Decisions and the Psychology of Choice." Science 211 (Jan.):453-58.

Viscusi, W. Kip, and Wesley A. Magat. 1987. Learning About Risk: Consumer and Worker Responses to Hazard Information. Cambridge: Harvard University Press.

Welsh, Michael. 1986. "Exploring the Accuracy of the Contingent Valuation Method: Comparisons with Simulated Markets." Ph.D. diss., Department of Agricultural Economics, University of Wisconsin, Madison.

@ @
COPYRIGHT 1993 University of Wisconsin Press
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Loomis, John B.; duVair, Pierre H.
Publication:Land Economics
Date:Aug 1, 1993
Previous Article:Joint recreation choices and implied values of time.
Next Article:On the economics of western local water finance: the Central Utah experience.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters