Accounting for heterogeneity in behavioural responses to health-risk information treatments

ABSTRACT

Traditional revealed and stated preference models consider a typical individual's behavioural responses to various policy-based information treatments. For some cost–benefit applications in which resource managers are concerned with responses from a representative individual, this is sufficient. However, as behavioural responses to information treatments can vary across respondents, we develop a latent class analysis with covariates to examine unobserved heterogeneity responses to health-risk information treatments. Results from a probabilistic model indicate that classes of consumers respond differently to the health-risk information treatments. Principally, we find that the media form of the information treatment is important, with raw consumer groups typically more responsive to a brochure information treatment, while cooked oyster consumers are more responsive to the same information in a video format. We also find that a proposed US Food and Drug Administration policy on processing all raw oysters before market has a greater effect on reducing demand for consumers of cooked oysters. However, with an associated price premium, all consumer classes reduce demand. Overall, the results suggest that future policy-based research could benefit from examining potential heterogeneity in individuals’ responses to risk information treatments in order to fully understand the efficacy of treatments on behaviour.

KEYWORDS:

• Revealed and stated preference
• health-risk information treatments
• oyster consumers
• latent class analysis

JEL codes:

• Q18
• Q22
• Q51

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. For example, Earnhart (2001) examines the benefits to homeowners from environmental amenities and coastal restoration; Whitehead, Haab, and Huang (2000) estimate the recreation benefits of environmental quality improvements; Morgan, Massey, and Huth (2009) consider the welfare effects of creating an artificial reef diving site in the Gulf of Mexico; and Huang et al. (2011) consider the impact of erosion and erosion control programs on beach recreation.

2. A full discussion of the potential health risks to oyster eaters and the Vibrio vulnificus (V. vulnificus) bacteria, the current regulatory stance of the FDA, and a full description of the survey design developed for this research can be found in Morgan et al. (2013).

3. A ‘case state’ refers to a state in which an individual has died from consuming a raw, Gulf of Mexico oyster. These are Florida, Alabama, Mississippi, Louisiana, Texas, and California,

4. Each respondent receives a price increase of $1, $3, $5, or $7, or a price decrease of either $1, $2, $3, $4, while being informed that the price of all other food products remains the same.

5. The brochure can be viewed at < http://www.issc.org/client_resources/Education/English_Vv_Risk.pdf>

6. The full video is available for viewing at http://vimeo.com/7035554

7. Mortality risk is difficult to quantify precisely. One estimate used by Bruner et al. (2014) was calculated by pairing the results of two related studies. Hlady and Klontz (1996) report on the incidence of Vibrio vulnificus infection among oyster consumers in the state of Florida. They estimate an average of 10.3 adults contracting a V. vulnificus infection for every 1,000,000 raw oyster-consuming adults. Their estimate uses reported cases of V. vulnificus to the Department of Health and Rehabilitative Services between 1981 and 1993. Of course, many cases of food-borne illness, including V. vulnificus, do not get reported because infected persons do not always seek medical help. To account for this, Bruner et al. rely on the adjustment factor used by Mead et al. (1999) and estimate the frequency of V. vulnificus infection at roughly twice the reported rate. Therefore, Bruner et al. estimate an average annual incidence of V. vulnificus at 20 cases for every 1,000,000 (1 in 50,000) oyster-consuming adults in Florida.

8. Price premiums are varied randomly across respondents as $1, $3, $5, or $7.

9. Respondents were prompted to choose from a five-level Likert scale of ‘Not very likely at all; Somewhat unlikely; I don't know; Somewhat likely; Very likely.’

10. All observations were collected before the BP Deepwater Horizon Gulf oil spill.

11. Results from the standard model (Table 3) are not directly comparable to the LCM results (Table 5) as the standard model accounts for the panel nature of the dataset via random effects while the LCM analysis does not.

Additional information

Funding

The research was supported by Gulf Oyster Industry Program [grant number R/LR-Q-32].