Public perceptions of food-related risks: a cross-national investigation of individual and contextual influences

Abstract

Public concerns about food risks have grown in recent decades in response to many food-related scandals. Despite some evidence that risk concerns vary across societies and risk domains, these variations remain understudied. To address this gap, this paper conducts a multi-level analysis of public concerns about biological and chemical/technical food risks in 26 European countries. Findings confirm previous work on individual predictors of risk concern and suggest that several contextual factors contribute to cross-national variations: aggregate perceptions of risks as unnatural, retail concentration in the food sector, and media coverage. The effect of institutional trust on risk concerns varies substantially across nations. Findings also reveal important differences in public concerns about biological versus chemical/technical food risks, supporting the view that food risk perception is multi-dimensional and complex.

Keywords:

• Risk perception
• food safety
• cross-national
• multi-level modeling

Notes

1. Although I would like to model this effect at both levels (individual and country), questions about unnaturalness appear only in an earlier Eurobarometer (73.1), precluding an examination of individual attitudes about unnaturalness.

2. Media coverage is the only predictor with substantial missingness (54.2%). However, this data is mostly missing at random because half of the sample was randomly selected to receive this question. Of the remaining predictors, only two have more than 5% missing: trust in E.U. institutions (6.9%) and institutional performance (5.7%). Missing data are imputed using all individual predictors from the analytical model, after which imputed values of the outcome variables are deleted (Von Hippel 2007). Following Graham, Olchowski, and Gilreath (2007), I perform 40 imputations to improve the efficiency and replicability of estimates. The imputed analyses are replicated with listwise deletion as a robustness check, and the findings are substantively similar. Reported results use the imputed data; the listwise deletion results are available upon request.

3. The remaining six ‘food worry’ items are excluded because they represent conceptually distinct concerns about diet (e.g. weight gain) and ethics (e.g. animal welfare).

4. With this and other descriptive statistics in the paper, the recommended post-stratification weights are applied. For descriptive statistics of individual subsamples, a post-stratification weight is applied to adjust for gender, age, region, and size of locality within each nation. For descriptive statistics of Europe as a whole, another weight is applied to correct for the fact that national samples are similar sizes despite being drawn from populations of vastly different sizes. This weight ensures that nations are represented within Europe according to their population size and includes all of the lower level weighting factors.

5. Only two of the country-level variables include observations for the four subnational samples (Northern Ireland, Great Britain, and East and West Germany): ‘unnaturalness’ and media coverage. For retail concentration and net trade, national data from the U.K. and Germany are applied to the subsamples.