Manageable risks and the demand for food products: the case of oyster

Abstract

The risk of norovirus food poisoning from oyster consumption can be minimized by consumers, as thorough cooking can destroy norovirus. This paper uses data from 23 prefectures in Japan during 1970–2010 to examine how the demand for oysters responds to this manageable risk. We find that the incidence rate of norovirus has no effect on the price of oysters in prefectures that specialize in oysters for cooking, whereas it has significant negative impacts in prefectures that produce oysters to be eaten raw. These results indicate that consumers respond differently to manageable and non-manageable risks.

Keywords:

• Demand
• Japan
• norovirus
• oyster
• price
• risk

Notes

Acknowledgments

We are grateful to Hikaru Oikawa for his excellent research assistance. This research was supported by Japan Fisheries Research and Education Agency.

Notes

1 Oysters are also known for its risk of vibrio vulnificus infection, but it is very rare in Japan. Although the formal statistics is not available, Ooishi Ura, Mitsumizo and Nakajima (2006) report that there were only 185 cases during 1975–2005. Furthermore, unlike in the United States, vibrio vulnificus incidences in Japan originate from various seafood products, and raw oysters are not the primal cause.

2 We do have access to a confidential advertising catalog dataset compiled by Japan Fisheries Information Service Center. It shows the average price of oysters to be eaten raw (2610 yen/kg) is higher than that of oysters for cooking (2569 yen/kg), during January, February, October, November and December in 2010. These prices substantially vary by month; oysters to be eaten raw are more expensive in January, February and December, whereas oysters for cooking are more expensive in October and November. This dataset is useful in obtaining qualitative information, but not suitable for statistical analyses due to its small sample of member shops.

3 We add 1 to all the observations of the number of norovirus cases before ln transformation.

4 Shamkhali Chenar and Deng (2017) report that water temperature, solar radiation and gage height are among the most important factors to explain oyster norovirus outbreaks.

5 This does not violate the law, as long as they put the correct label on their product stating whether it is to be eaten raw or for cooking.

6 We use data on oysters with shell for our base estimations, but using data on oysters without shell gives very similar results.

7 Demura (2012) reports 90% of oysters from Miyagi Prefecture are sold as oysters to be eaten raw, while “most of” the oysters from Hiroshima Prefectures are sold as oysters for cooking. Similarly, Miyata et al. (2014) report the oysters from Miyagi and Iwate Prefectures are “mostly” those to be eaten raw. Miyata (2005) also reports that Hokkaido, Iwate and Mie Prefectures produce oysters with shells, and Tokyo Metropolitan Central Wholesale Market (2018) explain that oysters with shells are “mostly” sold as those to be eaten raw.

8 Assuming 0% instead of 10% and 100% instead of 90% gives similar results.

9 For this reason, we excluded five prefectures from our analysis whose production data are available before 1998 but is missing during 1998–2010.

10 Let y and x denote the perceived risk of infection and the incidence of norovirus, respectively. Suppose y is a linear function of x and the parameter is $b_1$ before 1998 and $b_2$ after 1998: $\mathrm{y}=a+b_1*D*x+b_2*(1-D)*x$, where D takes 1 before 1998 and 0 after. If the effect of the incidence is zero before 1998 (i.e., $b_1=0$), then $y=\mathrm{a}+b_2*\left(1-D\right)*x=a+b_2*x\prime$, where the new variable $x\prime$ takes zero before 1998 and equals x after 1998. Thus, even if x is not actually zero before 1998, we can estimate $b_2$ by assuming $x=0$ before 1998, as long as the effect of x on y is zero before 1998.

11 It is also possible that consumers consciously or subconsciously used other sources of information to infer the risk from norovirus before 1998. A primal candidate is the rate of food poisonings from related food products or pathogens. As such, we also included in our regression the incidence rate of food poisonings from shellfish, Vibrio and Escherichia. However, none of them became statistically significant, and the point estimates for the norovirus incidence did not change much. Also note that the statistics on the number of infectious gastroenteritis cases (that includes norovirus infection) is only available since 1999.

12 The calculation is based on Column 1 in Table 4. As we add 1 before ln transformation, the predicted price before 1998 from our model is slightly different from the actual price. For illustrative purpose, we simply assume the predicted price before 1998 is exactly the same as the actual price.