Heterogeneous risk perceptions : the case of poultry meat purchase intentions in Finland

Int. J. Environ. Res. Public Health 2013, 10, 4925-4943; doi:10.3390/ijerph10104925 International Journal of Environmental Research and Public Health ISSN 1660-4601 www.mdpi.com/journal/ijerph Article Heterogeneous Risk Perceptions: The Case of Poultry Meat Purchase Intentions in Finland Jaakko Heikkilä 1,*, Eija Pouta 1,†, Sari Forsman-Hugg 1,† and Johanna Mäkelä 2,† 1 MTT Economic Research, Latokartanonkaari 9, Helsinki FI-00790, Finland; E-Mails: [email protected] (E.P.); sari.forsm[email protected] (S.F.-H.) 2 Department of Teacher Education, University of Helsinki, P.O. Box 8 (Siltavuorenpenger 10), Helsinki FI-00014, Finland; E-Mail: johanna.m.m[email protected] † These authors contributed equally to this work. * Author to whom correspondence should be addressed; E-Mail: [email protected]; Tel.: +358-29-531-7190; Fax: +358-20-772-040. Received: 21 August 2013; in revised form: 12 September 2013 / Accepted: 29 September 2013 / Published: 11 October 2013 Abstract: This study focused on the heterogeneity of consumer reactions, measured through poultry meat purchase intentions, when facing three cases of risk. The heterogeneity was analysed by latent class logistic regression that included all three risk cases. Approximately 60% of the respondents belonged to the group of production risk avoiders, in which the intention to purchase risk food was significantly lower than in the second group of risk neutrals. In addition to socio-demographic variables, the purchase intentions were statistically associated with several attitude-based variables. We highlighted some policy implications of the heterogeneity. Overall, the study demonstrated that risk matters to consumers, not all risk is equal, and consumer types react somewhat differently to different types of risk. Keywords: poultry; consumer behaviour; risk; food safety; consumer heterogeneity OPEN ACCESS Int. J. Environ. Res. Public Health 2013, 10 4926 1. Introduction and Background Consumer behavioural responses to risks have been studied in relation to consumer food choices, as well as by focusing on purchase intentions under individual food risk types. Typically, these studies have focused on either production-related risks such as hormone treatment [1,2], chemical control substances [3] and genetic modification (GM) [4–7], or on animal disease risks such as Campylobacter [8], Salmonella [9,10] and bovine spongiform encephalopathy (BSE) [11–13]. It has been recognised that the risks are subjectively perceived by consumers, even if objective information on probabilities and impacts is available. Consumers respond to risks based on their subjective perceptions of the probability of the event (i.e., risk perceptions) and on their attitude towards the event (i.e., general positive or negative predisposition) [14]. Additionally, consumer reactions depend on several factors that are risk type-specific, including the extent to which they can affect the risk themselves, how familiar they are with the risk, to what extent the exposure is voluntary, how severe the consequences are (irrespective of their probability), and so forth [15–18]. In this study, we focused on the heterogeneity of consumer reactions to a set of differing risk cases. In previous studies, geographical differences among consumers have been documented as a source of heterogeneity. Consumers in the European Union (EU) have been particularly sceptical regarding genetically modified products (e.g., [15,19]), as well as hormone-treated beef [20], and the risk perception in general has been found to be higher in Europe than in North America or Asia [4]. Country-specific differences have also been documented [21,22]. The analysis of perceptions and attitudes toward risk has often focused on the differences between experts and the public [20,23–25]. However, socio-demographic factors such as age, gender, whether the respondent has children and whether the respondent lives in urban or rural surroundings have also been found to correlate with the risk response (see, e.g., [6,10,26–28]). Beyond individual sources of heterogeneity, segmenting consumers could be useful in risk studies to assess how various types of risk affect heterogeneous consumer segments and their buying behaviour. Identifying the reactions towards specific risks in separate segments helps in targeting information and product development at those segments that are most sensitive [29]. In addition to the design of policies, understanding the behaviour of heterogeneous consumer segments is a necessity for the viable implementation of policies such as risk communication. Segmenting of the population according to information needs and the development of information with high levels of personal relevance to specific segments allows more effective communication of the risks [20]. Additionally, if consumer reactions to risk differ according to the segment, the benefits of precautionary policies will also differ among segments, suggesting that equity considerations are also involved. Therefore, segmentation appears to be a prominent approach in targeting risk management policies, communication and research. However, only a few studies have reported consumer segments based on their food risk perceptions, attitudes or behavioural responses. In a review of consumer reactions to a food crisis, Wansink [30] formed theoretical consumer segments from a combination of risk perceptions and pre-existing attitudes towards risk. Varying the level of risk perception and attitude in a matrix of four fields, he illustrated the behavioural response to risk in these four segments. In an empirical, study Kennedy et al. [31] explored the content of food safety attitudes, which included components of concern, trust, Int. J. Environ. Res. Public Health 2013, 10 4927 the desire for a high level of regulation, acceptance of the number of suffering people, and preference for the right to purchase safe or unsafe food. Based on these attitude components, the authors identified five segments of consumers, with one of them (apprehensive consumers) benefiting the most from safety information. Focusing on Creutzfeldt-Jacob disease, Payne et al. [29] used behavioural, attitude and risk perception information in an experimental setting to segment consumers. They observed that group memberships depend on consumption levels, but demographic variables also provide profile information on the segments. These previous studies have segmented consumers based on their risk perceptions, risk attitudes or behavioural responses towards risk. However, none of these studies has focused on segmentation based on behavioural responses (i.e., purchasing decisions) under various risk types. There have also been a few studies comparing the reactions to various risk types among the general public [32–34], including the case of food risks [35–39]. However, most of these studies have focused on similarities and differences between risk types, not the heterogeneity of people in their reactions. This implies that in food safety issues the general profile of individuals sensitive to various risk types is unknown. In the case of food purchasing decisions, it would also be beneficial to identify those individuals who are particularly sensitive to any of the hazards. In this study, the segments of consumers were formed based on differences in behavioural intentions, but they were identified according to risk-related attitudes and socio-demographic or geographic factors. Moreover, we simultaneously considered several types of risk. By recognizing that beliefs precede attitudes, which further affect purchasing intentions, we followed the ideas of Ajzen [9,40–42] in a simplified form, where the attitudes toward risk were assumed to precede purchasing behaviours. Instead of focusing on subjective norms, which in the food-buying settings are typically inside household norms, or on perceived behavioural control, which in the case of hypothetical risks is challenging to concretize in a survey, we focused on various types of attitudes that can be assumed to affect the buying intentions for poultry under various types of risk. Although there is extensive literature on attitudes related to food risks (for a review, see [43]), the content of attitude constructs applied to model subjective risk evaluations by consumers varies considerably [5,20]. We applied the segmentation concept to the behavioural reactions of Finnish consumers to various risks from poultry meat. The case of poultry is interesting for several reasons. The consumption of poultry meat in Finland has increased significantly in recent years. The increased demand has partly been covered by imports, meaning that international safety risks can rapidly enter the markets due to globalisation of the food chain. At the same time, consumers have become increasingly interested in how meat is produced and how safe it is. If safety is compromised, shortor long-term reductions in demand are to be expected, as demonstrated in Europe, for instance, by avian influenza H5N1 [44,45]. Furthermore, poultry meat can be associated with several risk types and thus fits the aims of the study. We presented the respondents with three different risk cases related to a biological hazard (increased risk of Salmonella in poultry), a chemical hazard (chemical treatment of poultry meat) and a technological hazard (use of genetically modified feed), and analysed how various perceived risks affect purchase intentions, and whether consumer segments exist that react uniformly to all of these risk cases. By analysing the role of attitude constructs such as health orientation, domestic preference, GM negativity and safety orientation, we also aimed to provide information on the characteristics of the segments. Int. J. Environ. Res. Public Health 2013, 10 4928 2. Methods and Data 2.1. The Survey Data Since safety features have no observable variation or direct price in stores, experiments [6] or surveys revealing purchase intentions need to be used to estimate their demand. Here, consumer survey data on poultry meat consumption were used. An online Internet survey conducted in November 2007 provided information on the reaction of consumers to risk types. The data were collected by sampling 2,500 respondents from the Internet panel of a private survey company, Taloustutkimus. The panel comprised 20,000 respondents who had volunteered to participate in the panel [46]. The otherwise random sample was balanced toward population by adding 70 respondents representing the category of older respondents. The consumer data set (N = 1,312), with a response rate of 51%, was a representative sample of Finnish Internet users between the ages of 18 and 79 years. The sample was representative of the general population regarding gender, age, income and geographical location (Table 1), but the educational level in the sample was somewhat higher and the proportion of individuals with children in the family somewhat lower than in the general population. In the data set, 95% of the consumers included poultry in their monthly diet and almost half of them perceived that the proportion of poultry had increased in their diet during the previous five years. Table 1. Descriptive statistics for the respondents in the data set and the 18to 79-year-old population in Finland. In data In population * Proportion of females, % 51 51 Mean age, years 49 47 Proportion of people with a higher educational level (college or university), % 38 26 Proportion of people living in households with a gross income under €40,000, % 42 42 Proportion of people with children (<18 years) in the family, % 29 42 Proportion of people living in northernmost Finland (Lapland), % 4 4 Proportion of consumers having poultry in their monthly diet, % 95 N/A Proportion of consumers who have increased the share of poultry in their diet during the last five years, % 49 N/A * Source: www.stat.fi, 2009. N/A denotes data not available. For dependent variables, the survey measured the conditional purchase intentions of the respondents regarding poultry products under three different risk cases. These three cases are here collectively called “risk products”, but it should be noted that the last two cases are objectively only potentially riskier than standard products. A risk product here thus refers to risk as perceived by the consumer. Case 1 concerned biological risk, the zoonotic risk related to Salmonella, with a quantitative explanation. The levels of morbidity and mortality derived from the Finnish National Salmonella Control Programme were provided for the respondents. The respondents were told that 3,300 persons annually become mildly and 400 seriously ill (requiring attention from a doctor or hospital treatment) after eating poultry meat. They were asked to consider a situation where 19,800 persons become moderately ill, 2,400 seriously ill and additionally four persons die annually due to eating poultry meat. Half of the respondents, chosen randomly, were asked how the risk would affect their consumption Int. J. Environ. Res. Public Health 2013, 10 4929 decisions if the poultry were half of the current price. These respondents were provided a response scale of four alternatives for increased use (25%, 50%, 75%, 100% increase), one alternative of “no effect”, and four alternatives for decreased use (25%, 50%, 75%, no use). Half of the respondents were asked about their reactions if the price remained unchanged. For them, the response scale was “no effect” and four alternatives for decreased use (25%, 50%, 75%, no use). Case 2 concerned chemical safety, and the respondents were presented a scenario where chemical treatment is an alternative approach to maintain product safety: “The safety of poultry meat in Finland is ensured with good production hygiene throughout the production chain. An alternative approach is to treat the meat products before they reach the consumer with chemicals to eliminate potential pathogens. International trade negotiations may lead to the market entry of chemically treated meat in the EU.” The exact name of the possible chemical (chlorine) was not mentioned in the survey. After this information, a four-level scale was presented to the respondents to indicate their willingness to choose the product: (a) if it was cheaper than the conventional product; (b) if it had the same price as the conventional product; (c) even if it was more expensive than the conventional product; or (d) would not choose the product at all. Case 3 related to technological risk, and it was explained to the respondents that genetically modified soya was not at the time of the survey used in poultry production in Finland but was a future alternative (feed containing GM soy entered the Finnish feed markets in 2013). The purchase intention was asked if GM feed was used in production. The same scale was implemented as in Case 2. In all three cases, the respondents who would still buy the product (although possibly at a lower price) were coded in the “yes” category, and the respondents who would stop buying altogether were coded in the “no” category. In other words, in Case 1, a decreased amount of use was also coded in the “yes” category. We tested alternative modelling approaches. First, we used the whole range of scales of the dependent variables, i.e., the original four-level scale for chemical risk and technological risk and the original five-level scale for biological risk. We undertook ordinal regression (ordered logit model). However, the test for parallel lines implied that the slope coefficients in the models were not equal across all response categories. For such data, the appropriate model would be the multinomial logit model instead of ordinal regression. We estimated the multinomial logit models, but in the case of the four-category dependent variable, no stable solution was found because of the low proportion of responses in one category. The model with a three-category dependent variable did not provide us with considerably more information than the binary approach, and because the multicategory approach made the comparisons between the risk types more difficult, we ended up using the binary coded variables, as detailed above. To provide independent variables for modelling, the survey included questions relating to consumer patterns of using poultry meat, several attitude and belief questions relating to poultry production, and socioeconomic background variables. The relative amount of poultry meat in the diet was constructed from the five-class scale measuring the frequency of poultry consumption. The final variable was expressed relative to the consumption frequency of other types of food (vegetarian, fish, beef, pork). The attitude variables included health orientation, domestic preference, GM negativity and safety orientation (Table 2). Structured focus group discussions [47] on poultry meat with different types of consumers were utilized in developing the measures. The salient issues from the focus groups were Int. J. Environ. Res. Public Health 2013, 10 4930 also reflected with measures available from the literature (e.g., [48–51]). The reliability of health orientation, domestic preference and GM negativity measures was analysed with Cronbach’s alpha (Table 2). Health orientation was measured with eight statements from the measures of general health interest in the food setting presented by Roininen et al. [48] using a Likert scale from one (fully disagree) to five (fully agree). One of the items was dropped to obtain an acceptable Cronbach’s alpha and seven items formed the final variable, summarised by the mean of the item responses. The statements were as follows: (1) For me it is important that my daily food is low in fat; (2) For me it is important that my daily food contains plenty of vitamins and micronutrients; (3) The healthiness of snacks is irrelevant to me (reversed); (4) I do not avoid any foods, even if they raise my cholesterol level (reversed); (5) I make sure that all the food I eat is healthy; (6) I eat what I want and do not care much about the healthiness of food (reversed); (7) I always follow a healthy and balanced diet. Table 2. Descriptive statistics for the poultry use variable and the attitude-based variables, including Cronbach’s alphas where applicable. Minimum Maximum Mean Std. deviation Cronbach’s alpha Relative amount of poultry meat in the diet 0.26 2.27 0.81 0.26 Attitude-based variables Health orientation 1.00 5.00 3.63 0.73 0.857 Domestic preference 1.17 5.00 4.17 0.64 0.786 GM negativity 1.00 5.00 3.94 1.03 0.931 Safety orientation 0.26 2.37 1.14 0.18 Based on the focus groups and previous studies on country of origin effects [50,51], the attitude towards Finnish production was measured by six belief statements regarding domestic poultry production: (1) By buying Finnish poultry meat I can affect employment in Finland; (2) I buy Finnish poultry meat, because I value Finnish primary production; (3) I do not pay attention to the origin of the poultry meat (reversed); (4) Finnish poultry meat cannot be distinguished from imported poultry meat (reversed); (5) I buy Finnish poultry meat because I trust Finnish food production; (6) Finnish poultry products should be marked with a label to inform the consumers about the origin. A Likert scale from one (fully disagree) to five (fully agree) was used. The final variable of domestic preference was the mean of the responses to the six statements. Instead of general attitudes towards GM (e.g., [49]), the GM negativity statement was targeted specifically at the use of GM feed in poultry production. The measure was formed as a mean from four statements measuring beliefs related to GM feed in the poultry production chain with a Likert scale from one (fully disagree) to five (fully agree): (1) The impacts of meat grown using GM feed on humans are not known; (2) The impacts of GM feed on production animals are not known; (3) The impacts of GM feed on the natural environment are not known; (4) The genetic modification of agricultural plants is ethically questionable. The safety orientation was constructed from a question measuring the importance of several quality cues [52–54] faced when buying poultry products. In the final variable, the respondents’ rating of safety from one (not at all important) to five (extremely important), was divided by the mean of the Int. J. Environ. Res. Public Health 2013, 10 4931 importance of the other attributes: taste, healthiness, duration of use, price, easiness of use, attractive appearance, suitable package size, wide range of available products, and clearly marked country of origin. In other words, safety orientation measured the importance of safety relative to the importance of the other attributes. The correlations between the attitude variables were relatively low, being under 0.3 between all the variables. Although low, the correlations were significant between domestic preference and all the other attitude variables, indicating a general trust in domestic food, as also reported in many previous studies. The socio-demographic variables tested in the models were gender, age, education, income, families with children, and the residential region. Income and age were analysed in classes, as reported later in Table 5. 2.2. The Statistical Models For each of the risk products, a model of purchase intention was constructed. As the dependent variables were dichotomous, the method used for modelling each purchase intention separately was logistic regression analysis [55]. The dependent variable in the models was whether the respondent would continue buying poultry meat under a change in its safety or stop buying it. The independent variables were selected based on the assumption of attitudes explaining behavioural intentions. We started the model specification with correlation analysis identifying mutually correlating independent variables and variables having a significant correlation with each dependent variable. The full set of alternative independent variables was used as a starting point for the model development. All non-significant variables (p-value over 0.10) were individually removed from the model, starting with those having the highest p-values. In this manner, the intended purchase was explained by the previously described attitude variables, but also by the socio-demographic variables. In addition, the relative amount of poultry in the diet was included in the model. A few methods are available to study heterogeneity, but if there are no strong, theory-based assumptions regarding the sources of heterogeneity, segmentation through latent class models is a good method for the analysis. It does not require a priori assumptions regarding the sources of heterogeneity and deals well with multiple variables. The potential heterogeneity of the respondents regarding the purchase intentions was therefore analysed by latent class logistic regression for binary choices for all three risk products. It was also expected that latent class models would improve the explanatory power of the logit models. The basic assumption in a latent class model for binary choice data is that the parameters of the regression model differ across the estimated classes [56]. In the general case, the model specification is: (1) The dependent variable is a dummy variable for purchasing decisions under risk (yit). The indexes i and t refer to the individual and replication, i.e., three purchasing decisions (one for each risk product) (T = 3). Behind the observed variables exists an unobserved nominal variable x that indicates K separate classes, each having their own distribution of observed variables, y. In the latent class model, ),()(),( 1 1 covcov pred it T t it K x ir pred iii zxyfzxPzzyf i     Int. J. Environ. Res. Public Health 2013, 10 4932 prior assumptions of the reasons for heterogeneity are not needed. Instead, the attitudinal and socio-demographic variables zcov associated with the class membership are empirically examined. The index r indicates each background variable. zpred is a set of Q predictor variables influencing the dependent variable in each of the unobserved classes. In our case, the predictors were the case-specific dummy variables (Cases 1 and 2), while Case 3 (GM feed) was the reference level (Q = 2). In the case of a binary dependent variable, the probability of y = 1 obtains a logistic form: (2) where α and β are model estimates. A detailed description of the methodology is available in Vermut and Magidson [57]. In estimation, the covariates (i.e., attitude and socio-demographic variables) were active, thus affecting the solution, and they were selected in the model if the significance level of the coefficient was below 0.10. The model provides probabilities of class membership for each observation, and the observations were classified into the most probable class. The latent class model for binary choices in Latent Gold software was used to estimate the model. The Bayesian information criterion (BIC) indicated that the optimal model was a two-class model. Although the Akaike information criterion (AIC) continued to decrease, the improvement from the two-class model was minor (Table 3). Table 3. Bayesian (BIC) and Akaike (AIC) information criteria for selecting the number of classes. Number of classes BIC(LL) AIC(LL) L² R² 1 4,264 4,248 3,583 0.24 2 3,571 3,478 2,783 0.56 3 3,604 3,433 2,707 0.66 4 3,668 3,419 2,664 0.57 5 3,715 3,388 2,603 0.58 6 3,786 3,381 2,566 0.58 3. Results The measures of purchase intentions revealed the reactions of the respondents to the three risk cases in poultry production. In Case 1 (biological risk), about half of the respondents stated that they would decrease their consumption of poultry if the risk of morbidity and mortality increased six-fold. In the case of no price impact, only 13% of the respondents would continue the use of poultry at the current level and 57% at a lower level. If the increased disease risk would lead to a price reduction of 50%, 20% of the respondents would continue to use poultry at the current level. Five per cent of the respondents were less sensitive to risk and would consider increasing the use of poultry if the product price decreased by half. However, the proportion of respondents not willing to buy was the same, 30%, regardless of the price effect (Table 4). )exp(1 )exp( ),1( 1 1 pred itq Q q qxx pred itq Q q qxx i z z zxyP          Int. J. Environ. Res. Public Health 2013, 10 4933 Nearly 90% of the respondents were of the opinion that they would not choose chemically treated poultry meat (Case 2). Seven per cent of the respondents, however, were willing to select the chemically treated product if it was cheaper than conventional meat. Table 4. Distribution of buyers based on purchase intentions, proportion of buyers, and correlation between the purchase intentions under different types of risk. Distribution, % of respondents Proportion of buyers % Phi coefficient for correlation... (p-value) Would increase purchases * No effect on purchases Would decrease purchases Would not purchase ...with Case 2 ...with Case 3 Case 1 (Biological risk) 2.3 16.3 51.4 30.0 70.0 0.056 (0.043) 0.148 (0.000) Would purchase if cheaper than conventional Would purchase if the same price as conventional Would purchase even if more expensive than conventional Would not purchase Case 2 (Chemical risk) 7.5 2.9 0.7 88.9 11.1 0.361 (0.000) Case 3 (GM-feed) 23.8 12.0 1.0 63.2 36.8 * This option was available only for the first sub-sample, where it was described that the price would decrease by half. The column “Proportion of buyers” is the sum of the first three data columns. Of the respondents, 63% would not select poultry meat fed with GM feed (Case 3). Approximately 25% would select the GM product if it was cheaper than the conventional product. When asked, over 90% expressed the opinion that GM feed should be marked with a label. Table 4 also reports the correlation coefficients between the purchase intentions to provide a first impression of whether the purchase intentions under various risks are correlated with each other. The correlation coefficients show stronger dependency between the intentions under the two risk types related to production, i.e., chemical risk and GM risk. The purchase intention under biological risk is associated more weakly with the intentions under the other two risk types. Table 5 presents the logistic regression models and the variables that affect the buying intention for each risk product. This provides an opportunity to compare the association of perceived risks with various background variables. For Case 1 (biological risk), the purchase intention probability was significantly affected by the amount of poultry in the respondent’s diet. This probably indicates that when the proportion of poultry in the diet increases, consumers perceive the probability of infection to increase. The effect of the health orientation of the respondent was slightly higher and had a lower p-value in the case of biological risk than in the models for other products. Safety orientation consistently affected the choice in all models, but the effect was slightly lower for biological risk than for the other risks. Women were more likely to reduce their purchase intention as biological risk increased, but the effect of gender was even higher for the other risk products. Younger people reacted Int. J. Environ. Res. Public Health 2013, 10 4940 Overall, the study demonstrated that risk matters to consumers, not all risk is equal, and consumer types react somewhat differently to various types of risk. Taking such diversity into account in risk assessment, risk management and risk communication is a challenge for authorities, as well as for food manufacturers. Conflicts of Interest The authors declare no conflict of interest. References 1. Alfnes, F. Stated preferences for imported and hormone-treated beef: Application of a mixed logit model. Eur. Rev. Agric. Econ. 2004, 31, 19–37. 2. Lusk, J.L.; Roosen, J.; Fox, J.A. Demand for beef from cattle administered growth hormones or fed genetically modified corn: A comparison of consumers in France, Germany, the United Kingdom, and the United States. Am. J. Agric. Econ. 2003, 85, 16–29. 3. Travisi, C.M.; Nijkamp, P. Valuing environmental and health risk in agriculture: A choice experiment approach to pesticides in Italy. Ecol. Econ. 2008, 67, 598–607. 4. Frewer, L.; van der Lans, I.; Fisher, A.; Reinders, M.; Menozzi, D.; Zhang, X.; van den Berg, I.; Zimmermann, K. Public perceptions of agri-food applications of genetic modification— A systematic review and meta-analysis. Trends Food Sci. Technol. 2013, 30, 142–152. 5. Costa-Font, M.; Gil, J.M.; Traill, B. Consumer acceptance, valuation of and attitudes towards genetically modified food: Review and implications for food policy. Food Policy 2008, 33, 99–111. 6. Lusk, J.L.; Coble, K.H. Risk perceptions, risk preference, and acceptance of risky food. Am. J. Agric. Econ. 2005, 87, 393–405. 7. Grunert, K.G.; Lähteenmäki, L.; Nielsen, N.A.; Poulsen, J.B.; Ueland, O.; Åström, A. Consumer perceptions of food products involving genetic modification—Results from a qualitative study in four Nordic countries. Food Qual. Prefer. 2001, 12, 527–542. 8. Christensen, T.; Mørkbak, M.; Denver, S.; Hasler, B. Preferences for Food Safety and Animal Welfare—A Choice Experiment Study Comparing Organic and Conventional Consumers; Joint Organic Congress: Odense, Denmark, 2006. 9. Lobb, A.E.; Mazzocchi, M.; Traill, W.B. Modelling risk perception and trust in food safety information within the theory of planned behaviour. Food Qual. Prefer. 2007, 18, 384–395. 10. Golberg, I.; Roosen, J. Measuring Consumer Willingness to Pay for a Health Risk Reduction of Salmonellosis and Campylobacteriosis. In Proceedings of 11th Congress of the European Association of Agricultural Economists (EAAE), Copenhagen, Denmark, 24–27 August 2005. 11. Kalogeras, N.; Pennings, J.M.E.; van Ittersum, K. Consumer Food Safety Risk Attitudes and Perceptions over Time: The Case of BSE Crisis. In Proceedings of 12th Congress of the European Association of Agricultural Economists (EAAE), Gent, Belgium, 26–29 August 2008. 12. Mazzocchi, M.; Stefani, G.; Henson, S. Consumer welfare and the loss induced by withholding information: The case of BSE in Italy. J. Agric. Econ. 2004, 55, 41–58. 13. Latouche, K.; Rainelli, P.; Vermersch, D. Food safety issues and the BSE scare: Some lessons from the French case. Food Policy 1998, 23, 347–356. Int. J. Environ. Res. Public Health 2013, 10 4941 14. Pennings, J.M.E.; Grossman, D.B. Responding to crises and disasters: The role of risk attitudes and risk perceptions. Disasters 2008, 32, 434–448. 15. Slovic, P. The risk game. J. Hazard. Mater. 2001, 86, 17–24. 16. German Advisory Council on Global Change. World in Transition: Strategies for Managing Global Environmental Risks; Springer-Verlag: Berlin/Heidelberg, Germany, 1998. 17. Fischhoff, B.; Slovic, P.; Lichtenstein, S. How safe is safe enough? A psychometric study of attitudes towards technological risks and benefits. Policy Sci. 1978, 9, 127–152. 18. Yeung, R.M.W.; Morris, J. Food safety risk: Consumer perception and purchase behaviour. Br. Food J. 2001, 103, 170–186. 19. Food Chain Evaluation Consortium. Evaluation of the EU Legislative Framework in the Field of GM Food and Feed; European Commission, DG SANCO: Brussels, Belgium, 2010. 20. Verbeke, W.; Frewer, L.J.; Scholderer, J.; de Brabander, H.F. Why consumers behave as they do with respect to food safety and risk information. Anal. Chim. Acta 2007, 586, 2–8. 21. Mazzocchi, M.; Lobb, A.; Traill, W.B.; Cavicchi, A. Food scares and trust: A European study. J. Agric. Econ. 2008, 59, 2–24. 22. TNS Opinion and Social. Special Eurobarometer 354: Food-Related Risks; TNS Opinion and Social: Brussels, Belgium, 2010. 23. Savadori, L.; Savio, S.; Nicotra, E.; Rumiati, R.; Finucane, M.; Slovic, P. Expert and public perception of risk from biotechnology. Risk Anal. 2004, 24, 1289–1299. 24. Mertz, C.K.; Slovic, P.; Purchase, I.F.H. Judgments of chemical risks: Comparisons among senior managers, toxicologists, and the public. Risk Anal. 1998, 18, 391–404. 25. Hansen, J.; Holm, L.; Frewer, L.; Robinson, P.; Sandøe, P. Beyond the knowledge deficit: Recent research into lay and expert attitudes to food risks. Appetite 2003, 41, 111–121. 26. Adda, J. Behavior towards health risks: An empirical study using the “Mad Cow” crisis as an experiment. J. Risk Uncertain. 2007, 35, 285–305. 27. Grunert, K.G. Current issues in understanding of consumer food choice. Trends Food Sci. Technol. 2002, 13, 275–285. 28. Hammitt, J.K.; Haninger, K. Willingness to pay for food safety: Sensitivity to duration and severity of illness. Am. J. Agric. Econ. 2007, 89, 1170–1175. 29. Payne, C.R.; Messer, K.D.; Kaiser, H.M. Which consumers are most responsive to media-induced food scares? Agric. Resour. Econ. Rev. 2009, 38, 295–310. 30. Wansink, B. Consumer reactions to food safety crises. Adv. Food Nutr. Res. 2004, 48, 103–150. 31. Kennedy, J.; Worosz, M.; Todd, E.C.; Lapinski, M.K. Segmentation of US consumers based on food safety attitudes. Br. Food J. 2008, 110, 691–705. 32. Scotton, C.R.; Taylor, L.O. Valuing risk reductions: Incorporating risk heterogeneity into a revealed preference framework. Resour. Energy Econ. 2011, 33, 381–397. 33. Sunstein, C.R. Bad deaths. J. Risk Uncertain. 1997, 14, 259–282. 34. Slovic, P.; Fischhoff, B.; Lichtenstein, S. Facts and Fears: Understanding Perceived Risk. In Societal Risk Assessment: How Safe is Safe Enough? Schwing, R.C., Albers, W.A., Eds.; Plenum Press: New York, NY, USA, 1980; pp. 181–214. 35. Verbeke, W.; Scholderer, J.; Lähteenmäki, L. Consumer appeal of nutrition and health claims in three existing product concepts. Appetite 2009, 52, 684–692. Int. J. Environ. Res. Public Health 2013, 10 4942 36. Sparks, P.; Shepherd, R. Public perceptions of food-related hazards: Individual and social dimensions. Food Qual. Prefer. 1994, 5, 185–194. 37. Fife-Schaw, C.; Rowe, G. Public perceptions of everyday food hazards: A psychometric study. Risk Anal. 1996, 16, 487–500. 38. Stefani, G.; Cavicchi, A.; Romano, D.; Lobb, A. Determinants of intention to purchase chicken in Italy: The role of consumer risk perception and trust in different information sources. Agribusiness 2008, 24, 523–537. 39. Kirk, S.F.L.; Greenwood, D.; Cade, J.E.; Pearman, A.D. Public perception of a range of potential food risks in the United Kingdom. Appetite 2002, 38, 189–197. 40. Ajzen, I. The theory of planned behavior. Organ. Hum. Decis. Process. 1991, 50, 179–211. 41. Ajzen, I.; Madden, T.J. Prediction of goal-directed behavior: Attitudes, intentions and perceived behavioral control. J. Exp. Soc. Psychol. 1986, 22, 453–474. 42. Cook, A.J.; Kerr, G.N.; Moore, K. Attitudes and intentions towards purchasing GM food. J. Econ. Psychol. 2002, 23, 557–572. 43. Lobb, A. Consumer trust, risk and food safety: A review. Food Econ. Acta Agric. Scand. Sect. C 2005, 2, 3–12. 44. EU Poultry Markets ‘Feeling the Squeeze due to Avian Flu’. AgraEurope, 3 March 2006. 45. EU Poultry Stocks Burgeoning as Demand Falls. AgraEurope, 28 April 2006. 46. Taloustutkimus. Internet Panel. Available online: http://www.taloustutkimus.fi/in-english/ products_services/internet_panel/ (accessed on 28 August 2013). 47. Isoniemi, M.; Paananen, J. Laadullinen tutkimus siipikarjanlihan käytöstä kotitalouksissa ja ruokapalveluissa (Qualitative Study on the Use of Poultry Meat in Households and Food Services); Kuluttajatutkimuskeskus: Helsinki, Finland, 2008; Työselosteita ja esitelmiä (Reports and Presentations) 113, in Finnish. 48. Roininen, K.; Tuorila, H.; Zandstra, E.H.; de Graaf, C.; Vehkalahti, K.; Stubenitsky, K.; Mela, D.J. Differences in health and taste attitudes and reported behaviour among Finnish, Dutch and British consumers: A cross-national validation of the Health and Taste Attitude Scales (HTAS). Appetite 2001, 37, 33–45. 49. Lawson, L.; Larsen, A.; Pedersen, S.M.; Gylling, M. Perceptions of genetically modified crops among Danish farmers. Food Econ. Acta Agric. Scand. Sect. C 2009, 6, 99–118. 50. Verlegh, P.; Steenkamp, J.-B. A review and meta-analysis of country-of-origin research. J. Econ. Psychol. 1999, 20, 521–546. 51. Lobb, A.; Mazzocchi, M. Domestically produced food: Consumer perceptions of origin, safety and the issue of trust. Food Econ. Acta Agric. Scand. Sect. C 2007, 4, 3–12. 52. Glitsch, K. Consumer perceptions of fresh meat quality: Cross-national comparison. Br. Food J. 2000, 102, 177–194. 53. Bernués, A.; Olaizola, A.; Corcoran, K. Extrinsic attributes of red meat as indicators of quality in Europe: An application for market segmentation. Food Qual. Prefer. 2003, 14, 265–276. 54. Krystallis, A.; Arvanitoyannis, I. Investigating the concept of meat quality from the consumers’ perspective: The case of Greece. Meat Sci. 2006, 72, 164–176. 55. Hosmer, D.W.; Lemeshow, S. Applied Logistic Regression, 2nd ed.; John Wiley & Sons, Inc.: New York, NY, USA, 2000. Int. J. Environ. Res. Public Health 2013, 10 4943 56. Wedel, M.; DeSarbo, W.S. A Review of Recent Developments in Latent Class Regression Models. In Advanced Methods of Marketing Research; Bagozzi, R.P., Ed.; Blackwell Publishers: Cambridge, UK, 1994; pp. 352–388. 57. Vermunt, J.K.; Magidson, J. Technical Guide for Latent GOLD 4.0: Basic and Advanced; Statistical Innovations Inc.: Belmont, MA, USA, 2005. 58. Grande, J.; Bjørnstad, E.; Wilson, M.; Hanley, N. Assessment of Consumer Risk Attitudes and Behaviour Related to Countermeasures and Radioactive Contamination of Food; Nord-Trøndelag College: Steinkjer, Norway, 1999. Available online: www.sbes.stir.ac.uk/research/projects/ceser/ Consumer_report.pdf (accessed on 28 August 2013). 59. Slovic, P. Perception of risk. Science 1987, 236, 280–285. 60. Siegrist, M.; Keller, C.; Kiers, H.A.L. Lay people’s perception of food hazards: Comparing aggregated data and individual data. Appetite 2006, 47, 324–332. 61. Kahneman, D.; Tversky, A. Choices, values, and frames. Am. Psychol. 1984, 39, 341–350. 62. Fulham, E.; Mullan, B. Hygienic food handling behaviors: Attempting to bridge the intentionbehavior gap using aspects from temporal self-regulation theory. J. Food Prot. 2011, 74, 925–932 63. Carrington, M.J.; Neville, B.A.; Whitwell, G.J. Lost in translation: Exploring the ethical consumer intention-behavior gap. J. Bus. Res. 2012. Available online: http://dx.doi.org/10.1016/ j.jbusres.2012.09.022 (accessed on 10 October 2013). 64. Klinke, A.; Renn, O. A new approach to risk evaluation and management: Risk-based, precautionbased and discourse-based management. Risk Anal. 2002, 22, 1071–1994. © 2013 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).