New estimates of the demand for food safety: discussion.

Information on demand for food safety is needed to determine whether the benefit of safety regulations exceed the costs, and it is in this area that the papers in this session make an important contribution. Hammitt and Haninger provide an estimate of peoples' value to avoid an episode of food-borne illness conditional on severity, duration, mortality risk, etc. Roberts, in turn, takes these estimates to calculate the annual cost of food-borne illness in the United States. That the authors find higher demand for food safety than previously obtained using the cost-of-illness approach is entirely to be expected given that the latter estimates ignore factors such as pain and suffering, something most of us would clearly pay to avoid. I do not find higher estimates of demand for food safety disconcerting from a conceptual standpoint; however, the particular survey approach used is likely to have generated larger estimates than what would have been obtained from some other stated- or revealed-preference methods.

Hammitt and Haninger asked people whether they wanted a "standard" product with either a 2 or 4 in 10,000 chance of illness or a "superior safety" product with a 1 in 10,000 chance of illness, assuming the latter costs more than the "standard" product. The elicitation approach represents a mix between contingent valuation and conjoint-choice methods. It is my hypothesis that if people had been given repeated choices between two foods that were explicitly assigned prices, probabilities of illness, etc., a lower valuation estimate would likely have been obtained. A related concern is the hypothetical nature of the task. Previous work suggests that while total willingness-to-pay to obtain a good is typically overstated in hypothetical settings, marginal values are generally not. The framing of Hammitt and Haninger's question places it in a total willingness-to-pay context--the consumer either pays the extra amount or does not.

To illustrate, consider the experimental auction data collected by Hayes et al. (1995). I used these data to calculate the implied value to avoid each food-borne illness reported in table 2 of Hayes et al. (1995). The implied values for a reduction in risk of roughly 87.72, 80.00, 62.89, 4.17, and 0.42 in 10,000 were $98.04, $68.75, $133.56, $1,944.05, and $10,082.42, respectively. Thus, the implied values are much lower than those obtained by Hammitt and Haninger, even when one compares comparable changes in risk (e.g., the largest risk reduction in Hammitt and Haninger is 3 in 10,000, but in Hayes et al. (1995) a reduction of roughly 4.17 in 10,000 is only valued at $1,944). There are many differences in these two studies, such as differences in the samples (students vs. U.S. households), but the comparisons are suggestive.

Hammitt and Haninger also find people's values to be relatively insensitive to changes in severity, length of illness, and mortality risk; however, had a conjoint-type approach been used where each respondent made several repeated choices/rankings, it is likely that greater sensitivity would have been found. It is legitimate to ask whether preferences elicited in within- or between-subject experiments are most appropriate for public policy; however, Hammitt and Haninger's finding that households with children are less sensitive to changes in attribute levels than households without children is probably not due to fundamental differences in preferences, but due to differences in preference elicitation formats (the former group answered only one valuation question related to themselves whereas the latter answered two).

The juxtaposition of Shogren and Stamland's paper with the other two in this session is interesting on at least two fronts. First, Shogren and Stamland's survey results seem to cast doubt that people rationally respond to changes in food safety risk, which if true, would undercut the validity of the Roberts/Hammitt and Haninger valuation estimates. Second, Shogren and Stamland's argument that food safety risk is endogenous is not mentioned in the previous papers, despite some rather important implications for food safety demand.

I am less pessimistic than Shogren and Stamland about their survey results. Their primary concern is the lack of correlation between measured variables. However, it is difficult to know without a formal conceptual framework what relationship to expect among these variables--especially given their argument for endogenous risk. How people interpret the questions posed to them can also have a substantive impact. For example, when people were asked to state their personal risk perceptions, did they interpret this to mean ex ante risk (before self protection such as cooking) or ex post risk (after self protection)? People with higher ex ante beliefs of food safety risk will self-protect more--lowering ex post risk; people with lower ex ante beliefs will do the opposite. The result is that some people self-protect frequently while others do very little, but both types could have roughly the same ex ante belief, a fact that would result in little correlation between self-protection and risk perceptions. Further, the lack of correlation could result because the true underlying relationships are non-linear. For example, it is potentially the interaction between perceived effectiveness of a self-protection measure and its frequency of use that would be expected to relate to ex post food safety risk perceptions. The lack of sensitivity of willingness-to-pay to probability/severity of illness could also result if people substitute self-protection for increases in ex ante food safety risks.

Another seemingly disconcerting finding from Shogren and Stamland's survey is that two seemingly similar questions about the effectiveness of self-protection are virtually uncorrelated. However, close inspection of the questions reveals that one question asks "How effective is--?" whereas the other asks "How effective is doubling/halving--? The lack of correlation between these two questions might relate to diminishing marginal benefit of the self-protection practice. What if separate measures were created from the "doubling" and "halving" questions? Shogren and Stamland are further concerned about the high correlation between people's perception of the effectiveness of a practice for different meat types (beef, pork, and poultry). Even if there are large absolute differences in the "effectiveness" of a practice across meat types, people might have interpreted this word to imply reductions relative to the baseline risk.

Overall, the papers in this session significantly advance our understanding of the demand for food safety and should provide a useful starting point for others working in this field.

Reference

Hayes, D.J., J.F. Shogren, S.Y. Shin, and J.B. Kliebenstein. 1995. "Valuing Food Safety in Experimental Auction Markets." American Journal of Agricultural Economics 77:40-53.

Jayson Lusk is Professor and Willard Sparks Endowed Chair, Department of Agricultural Economics, Oklahoma State University.

This article was presented in a principal paper session at the AAEA annual meeting (Portland, OR, July 2007). The articles in these sessions are not subjected to the journal's standard refereeing process.