European consumers’ willingness to pay for red meat attributes

Food consumption in Europe is changing. Red meat consumption has been steadily decreasing in the past decades. The rising interest of consumers for healthier and more sustainable meat products provide red meat producers with the opportunity to differentiate their offers by ecolabels, origin and health claims. This international study analyses the European consumer preferences for red meat (beef, lamb and goat) in seven countries: Finland, France, Greece, Italy, Spain, Turkey and the United Kingdom. Through a choice experiment, 2.900 responses were collected. Mixed multinomial logit models were estimated to identify heterogeneous preferences among consumers at the country level. Results indicate substantial differences between the most relevant attributes for the average consumers, as well as their willingness to pay for them in each country. Nevertheless, national origin and organic labels were highly valued in most countries.


Introduction
Food production and consumption tendencies in Europe show significant changes in recent years, especially regarding meat consumption. While beef and lamb meat production has decreased since the Moreover, although the average consumer is concerned about animal welfare [28, 56,57], previous studies showed heterogeneous preferences [32,37]. According to de Jonge and van Triip [21], this variety of results is a product of the complexity of the animal welfare concept, which is highly dependent on consumer perception, their inferences on the diverse animal welfare levels, as well as how the message is communicated. In this study, given the difficulty of communicating animal welfare claims beyond the EU-wide animal welfare standards, animal welfare was not included in the attributes. Instead, the Halal certification label was included. Though Islam does not tolerate animal abuse, most halal certification does not incorporate specific animal welfare requirements, while pre-slaughter stunning is still a contentious issue [58]. Islamic rules require the animals to be alive, healthy, fed and watered before the slaughtering. The animal must be killed incurring the least pain possible, and the carcass must be drained of blood [59]. Some European Union countries (i.e., Belgium and Denmark) require that all animals be previously stunned before Halal slaughtering; while a recent EU Court ruling has established that all organic meat need to be stunned, including those bearing the halal label [60,61]. Research on reversible pre-slaughter stunning is still ongoing to improve animal welfare during slaughter yet still meet the spiritual requirements for halal [58]. Previous research also identified that many consumers, including non-Muslims, perceive the Halal label as a credence quality attribute [62,63], and this is a growing trend [64]. There is not much evidence on consumer perception of Halal certification in Europe, though one study showed that some non-Muslims question animal welfare issues in relation to the Halal certification in the UK [65]. However, a recent EU-funded study concluded that since "there is little expressed use of animal welfare as a purchase criterion, little understanding of the slaughter process and an inability to distinguish between different methods of stun, providing information on the different methods used would not appear to aid a consumer decision" [66].
There are also a wide variety of nutritional and health labels used for diverse proposes. As red meat remains as the most important dietary source of protein [67], a "high protein content" claim was selected for the present study. There is little previous literature reporting studies on consumers preferences of protein content in meat products [36,68], showing significant differences in cross-cultural settings. Moreover, there is a growing trend to use ready to cook/eat foods [69] and to look for healthier products [56] with less fat. However, previous literature does not offer clear and homogeneous results in both cases, as consumers present opposite views regarding these attributes' desirability, according to their culture and depending from the meat cut [38].
Given the variability of results from prior studies, all the previously mentioned attributes were included in this research. Besides, to account for cross-cultural differences, the willingness to pay (WTP) for each attribute was estimated at country level. Previous studies have established that product differences can be product differentiators only if they create valued benefits that consumers are willing to pay for [70]. Understanding the value that consumers give to each meat attribute and their WTP will provide essential information to producers, allowing them to identify which labels and quality claims consumers are willing to pay for.
The present paper is organised as follows. In the next section, there is a description of the methods applied, followed by the presentation of the results for each country. Then, results are discussed; conclusions are presented in the final section.

Materials and methods
The discrete choice model where the researcher can only observe Vij, the deterministic part of the utility. The remaining part of the utility (εij) is unobservable for the researcher and treated as random. The systematic component (Vij) can be approximated by a linear function of observed attributes in the vector Xij and the utility parameters of each attribute collected in the vector β: Assuming heterogeneous individual preferences across respondents [74], the discrete choice model may be specified as a mixed multinomial logit (MMNL) model that allows the unobserved, random part of the utility to follow any distribution [75].
Thus, the utility of individual i from alternative j is specified as [73]: Uij= β′i Xij + εij where β'i, is a vector of unobserved coefficients that varies between individuals but not over alternatives; Xij is a vector of observed variables that relate to each alternative j and respondents i and εij is a random term distributed i.i.d. extreme value over individuals and alternatives. For the present research, the linear specification of the utility for an individual i for the alternative j is: Uij = αij + βprice Price + βhalal Halal + βnational origin National origin + βeu origin EU origin + βPGI PGI + βorganic Organic + βcarbon Carbon + βLow fat Low fat + βprotein Protein + βformat Format + εij The random parameters β' (except the price) were assumed to be normally distributed to allow both positive and negative preferences for each attribute. A negative lognormal distribution was assumed for the price parameter to attain better behavioural fit to microeconomic theory. The utility was specified directly in the WTP space [76] to obtain better results and reduce the range of behavioural implausibility [77]: Uij = αij -λi pij + (λi γi)' xij + εij where λi = (βi price/μi), βi price and μi being respectively, an individual-specific coefficient for price and an individual-specific scale parameter; while γi = (ci/λi), where ci = (βi/μi).

Product and attributes selection
Based on the literature review results and a qualitative study [78], nine credence attributes were selected (  Table 1). Given the high number of attributes, the number of levels for each attribute was kept low to avoid excessive participants' cognitive burden [79]. The product name, origin and price were introduced through text in each alternative, while the other attributes were presented in a graphic way ( Figure 1). The base price was calculated as the average price for each meat type and cut in each country. Then, it was pivoted in three levels, with variations of ± 30%. The price was expressed in local currencies. The origin attribute was set to three levels: national, EU and New Zealand, as most lamb meat imports to the EU, come from this country [80]. New Zealand was set as the reference category for differences in utility in the choice model. As a consequence, the estimated willingness to pay (WTP) for both national and EU origin needs to be interpreted as a premium price above the New Zealand origin.
The PGI and PDO labels were presented through their official labels in the primary language of each country. The choice to choose either PGI or PDO was based on the presence/absence or predominance of each certification scheme for lamb meat in each country. In Finland, there exist only one PDO for red meat. In Turkey, being outside the EU, PGI and PDO regulations don't apply. Therefore, in both countries, the PGI and PDO labels for national meat were omitted as there are no certifications available in these countries [81].
The organic label was also adapted by country. Turkish respondents were presented with their local logo, while, for the European countries, the mandatory EU organic logo was used. The Halal certification, the carbon footprint and the "high protein content" labels were also adapted to each country. With respect to the carbon footprint attribute, the Carbon Trust label was used, which has gained popularity in the UK in recent years. The label was presented in the primary language of each country. In the case of the "high protein content" claim, a simple label of red letters on a white background was created for the present research, as there was no homogeneous label used across the countries under study.
The attributes fat content and format (convenience) were introduced to the respondents by modifying the alternative's image to keep it closer to reality, as people in the supermarket usually see the meat cut when making their choice ( Figure 2). To avoid biases, each alternative's image was modified using image-editing software, by adding/removing the presence of fat (as a proxy to healthiness and meat quality) [23, 82,83], as well as the presence of additional ingredients (e.g. rosemary, pepper, etc.) to feature a ready-to-cook product.

Figure 2 Examples of fat content presented in red meat cuts
In the case of the type of meat and cuts, these were based on the most consumed cuts of lamb and goat meat in the countries investigated, as they often present less variety than beef. As the most consumed lamb and goat cuts were chops and legs (for lamb), a T-bone steak was considered as the most comparable beef cut, to avoid bias for more convenient cuts or with no presence of bones. As a result, a labelled DCE was developed including four labelled alternatives (lamb leg, lamb chops, goat chops, T-bone steak) and a no-choice option.

Data collection and analysis
The data were simultaneously collected through an online survey from March to May 2019 in seven countries: Finland (FI), France (FR), Greece (GR), Italy (IT), Spain (ES), Turkey (TR) and the United Kingdom (UK). The respondents were selected by a third party (Qualtrics) using a quota sampling approach to achieve between-country comparability [84,85]. The quotas were established according to age and occupation by gender [86]. Only red meat consumers (at least once a year) were sampled. Respondents were between 18 and 64 years old, fully or partially responsible for the grocery shopping in their household and did not work or had a close relative working in the meat or catering industry. Two thousand nine hundred responses were collected. After removing incomplete or low-quality surveys (e.g., speeders, line-responses), 2866 usable responses were left (around 400 per country -see Table 2). The questionnaire was developed in English. Then it was translated and back-translated to the primary language of each country [87,88]. Active collaboration and feedback from international researchers allowed cross-country conceptual, functional, and category equivalence [89,90]. The questionnaire included consumer sociodemographic and geographic data, as well as a hypothetical discrete choice experiment (DCE). It was pilot tested in each country before launch.

DCE design and estimation
A fractional D-efficient design with priors consisting of twenty-four choice sets in two blocks was generated using the Ngene software (D-error = 0.256, A-error = 0.754). Each respondent was presented with twelve choice sets of five alternatives. All choice sets and alternatives were presented randomly to avoid ordereffect [91]. A "cheap talk" was introduced to respondents before answering the DCE, aiming to reduce the hypothetical bias [92,93]. Moreover, participants were introduced with all labels, their meanings and definitions [94], as well as an example of a choice set. Respondents had unlimited time to answer each choice set [95]. A total of 34,392 choices were collected.
The data was analysed using the APOLLO package in R [96]. For each country, an MMNL model in WTPspace was estimated based on the preference-space MMNL model's priors. The log-likelihood was attained by Monte-Carlo simulation-based integration using Halton draws with 1000 replications. A scaling factor was used to ease convergence in the WTP-space.

Results
The respondents' sociodemographic characteristics and meat consumption frequency are presented in Table 3. In all countries, most consumers declare to consume beef meat at least once a month. However, the percentage of consumers that consume lamb or goat meat with the same frequency varies in each nation. Turkey is the country with higher lamb and goat meat consumption, followed by the United Kingdom and Spain. Finland is the country with lower consumption frequency for both lamb and goat meat. The consumers' choices regarding the meat cut also varied between countries (Table 4). Beef T-bone was the preferred choice in Finland. Beef T-bone was the preferred cut in Greece too, though it was a close race with lamb chops; on the contrary, in Italy, the lamb chops were the preferred cut closely followed by beef Tbone. In Spain, Turkey and the UK, there was a clear preference for the lamb chops over other meat types. Only in France, the lamb leg was the preferred cut. In all countries, most consumers preferred to choose a meat alternative than selecting none.  FI  FR  GR  IT  ES  TR  UK  Beef T-bone  31  20  24  27  24  24  15  Goat chops  11  8  16  12  11  18  10  Lamb chops  18  21  23  29  29  29  30  Lamb leg  17  29  19  18  20  18  26  None  23  22  18  14  16  11  19  Total respondents  413  414  400  417  417  391  414 Willingness to pay for each country Results of the MMNL in WTP-space are presented in Table 5. National origin and the organic label are the only statistically significant and positive attributes for all countries. The attributes EU origin, PGI/PDO, Carbon footprint and low fat had positive coefficients when significant, while the attributes ready to cook and the high protein content presented negative coefficients when significant. The Halal attribute was statistically significant and positively valued only in Turkey, while in the other countries where it was significant, it was perceived negatively. However, it is worth mentioning that the statistical significance of the standard deviation of the Halal attribute is relatively high in several countries, implying a high heterogeneity of taste for this attribute within those countries.   Consumers from all countries presented the highest WTP for the meat of national origin if compared with New Zealand or EU origin. French consumers demonstrated the highest WTP (€) for national or EU origin, while the UK had the lowest WTP (€ 0.47) for national meat. In general, meat consumers were willing to pay more for the geographical origin of the meat than for certification labels linked to the origin of the meat (PGO/PGI). The PGO/PGI certification was significant and positive only in Greece, Italy, Turkey and the UK.
In most countries, the second preferred attribute was the organic label (besides in the UK, where the WTP for organic meat was higher than for the national origin). France presented the highest WTP for the organic label, followed by Greece, Finland and Italy. Another certification label included in the meat attributes was the Carbon footprint, which was significant and positive in France, Greece, Spain and Turkey. However, besides Spain, most countries presented a higher WTP for other attributes (e.g., organic) than for the Carbon footprint.
While the low-fat attribute was significant and positively valued only by French and Italian consumers, the high protein content was significant and negatively valued by Finnish, Greek and Turkish consumers. Moreover, the ready to cook format was also perceived negatively by France, Greece and Spain. However, the statistical significance of the standard deviations for this attribute might suggest heterogeneous preferences within the countries.

Discussion
Altogether, the results showed heterogeneity in consumer preferences and their WTP, both within and between countries. However, the national origin was always among the most valued attributes in all countries, in line with previous studies showing a higher preference for local and national products [41,52,97]. The EU origin was also valued positively and preferred over New Zealand imports, except for the UK. There was no significant difference in their WTP for meat coming from the EU and New Zealand for British consumers. New Zealand had a well-established position in the UK as a red meat supplier, at least before Brexit [98], which results in a quite accepted alternative to local lamb and, partly, beef.
Results were quite different when geographical quality assurance schemes were taken into consideration instead of geographical origin. The implementation of PDO and PGI labels had significant and positive influence only in Greece, Italy, Turkey and the UK. However, in all four countries, the PDO/PGI certification was always among the top three preferred attributes. Nevertheless, it always presented a lower WTP than the national origin. Red meat consumers were more inclined to pay more for the meat's geographical origin than for the quality cue linked to the production process being explicitly developed in a particular geographical area.
Organic products were also highly valued in all countries. The organic label was the second most important attribute for most countries, matching previous studies' results [51,99]. However, the WTP for organic meat varied from one country to another. In this case, France, followed by Greece, presented the highest WTP for organic meat, making both countries the most attractive market for organic meat products. On the other hand, Spain, Turkey and the UK had the lowest WTP for organic meat but with a high and significant standard deviation. The reported results imply a high heterogeneity of preferences in these latter countries regarding the organic attribute, confirming Koistinen et al. [9] results. However, such a variety of preferences also implies that there might be space for premium niche markets interested in organic meat. Further research is needed to identify possible organic segments within each of these countries and their willingness to pay for organic products.
The carbon footprint label was also positively valued by consumers in most countries (France, Greece, Spain and Turkey). Nevertheless, the WTP for this attribute was lower than other attributes, matching results from previous studies [33,56]. The result is not surprising, as currently the label is used mainly in the UK but for some other food products (e.g. cow milk). As the label has not been adopted for meat, consumers are not used to it. Additionally, Mandolesi et al. [78] suggest that most lamb/goat consumers believe that the production systems for sheep/goat meat are less intensive, while more environmentally friendly than those implemented in the beef industry. Such perception might bias consumers to perceive the Carbon footprint label as not highly informative. Previous research [52,55] supports this idea as the carbon footprint label was ineffective in influencing consumer behaviour.
Although previous research has identified halal as a quality attribute even for non-Muslims consumers [62], the results of the current study results showed little relevance of the label in the investigated countries. Except for consumers in Turkey and Greece (where the Halal attribute was not significant), consumers in the other countries negatively perceive the Halal label. However, the standard deviation estimate for Halal is significant and relatively high for all countries, meaning that there is a high heterogeneity of preferences within countries. As the survey sample was not representative of religion (the Muslim population in each country is frequently under-represented), these results might imply the possibility of niche markets within each country. Further research is required to determine the existence of non-Muslim consumer segments interested in the Halal attribute and their WTP for that attribute.
The lack of visible fat was appreciated only among French and Italian consumers, while the other countries were indifferent to this attribute. Fat content in meat is frequently used by consumers as a health indicator [23], leading them to choose leaner meat cuts, i.e., with less visible fat content [9,52,100]. Scozzafava et al [34] concluded that Italian consumers presented a lower price sensitivity for marble steak only when the meat cut was bought for special occasions and not for everyday consumption. As respondents in this study were requested to make their choices thinking of an ordinary day, the results reported above do not disconfirm their findings.
Different consumer knowledge may also explain the heterogeneity of preference on visible fat between countries. Grunert [101] identified a dysfunctional preference of consumers towards the fat content of meat. The author concluded that consumers believed a certain degree of marbling detracted juiciness and taste from their steak. Baba et al. [13] concluded that the fat content is a relevant attribute mainly for uninformed consumers. Given that the standard deviation for fat content is significant for all countries, it is possible to conclude heterogeneous preferences within each country concerning this attribute, which might be linked with the variability of preferences and knowledge among consumers.
The high protein attribute was statistically significant and negative only in Finland, Greece and Turkey. Such a result implies that consumers were either indifferent or rejected the attribute. Although consumers tend to value additional nutrients positively [102], this is not the case for the protein content. However, Van Wezemael et al. (2014) highlights the importance of how the claim is written. Teratanavat and Hooker [53] support this thesis by warning that claims referring to naturally-sourced nutrients are preferred over fortification or artificial enrichment. In this study, the protein claim did not explicitly specify the origin of the proteins; consumers might have perceived them as a fortification. Moreover, consumers may require additional information on the role of proteins for human health to appreciate such a claim. Further research is needed on the role and preference of proteins in meat products to understand how these claims are perceived and interpreted by the respondents.
The ready-to-cook attribute was significant and negative only in France, Greece, and Spain, implying that consumers from these countries prefer fresh products without any additional spices or preparations. However, in all nations, the attribute presented a significant standard deviation, which implies a high heterogeneity of preferences within countries. Such results suggest that there is a variety of different segments in each country, in which some respondents might appreciate the "ready-to-cook" preparations. Such findings match with the results from Pouta et al. [38], that concluded that consumers presented opposite views regarding the desirability of this attribute [38]. Further research is needed to identify the variety of preferences within countries. Caracciolo, F.; Cicia, G.; Del Giudice, T.; Cembalo, L.; Krystallis, A.; Grunert