Consumer Assessment of Sustainability Traits in Meat Production. A Choice Experiment Study in Spain

: Consumers are increasingly concerned about the way their food is produced. This is particularly relevant in the case of meat, due to the impacts that its production methods can have on greenhouse gas emissions and its role in climate change. In relation to this issue, the purpose of our research is to obtain more information on the consumer decision-making process for beef, in order to determine the relative importance of sustainability claims and traditional attributes, and identify consumer proﬁles with similar perceptions and intentions. A choice experiment was used to assess the inﬂuence of these attributes on consumers’ purchasing decisions. The results reveal that the best purchase choice for the consumer would be organic beef, produced in Spain, with an animal welfare label and eco-labelled. Later on, a cluster analysis was carried out using consumer beliefs and attitudes towards meat consumption as inputs, together with purchasing behaviour variables. A solution was obtained with three well-deﬁned consumer segments showing di ﬀ erent preference patterns: Cluster 1 (Male millennials indi ﬀ erent towards environment or sustainability), Cluster 2 (Sustainability-concerned mature women) and Cluster 3 (Middle-aged meat eaters with established families). The results of this study are relevant to develop more appropriate strategies that may be adapted to the behaviour and expectations of eco-friendly food consumers.


Introduction
During the recent years, there has been a growing public interest in food products produced using sustainable or ethical production methods [1]. In this regard, the increased demand for environmentally-friendly food products is associated with a growing interest in the sustainable use of resources and thus, in future wellbeing [2]. In general, consumers are increasingly aware of the fact that sustainable consumption is fundamental to protect the natural environment, counteract ongoing climate change, and ensure social justice [3,4]. In this sense, livestock products largely contribute to greenhouse gas emissions (GHG) and climate change [5]. Moreover, consumers prefer products identified as sustainable to conventional products based on animal welfare, environmental, or social reasons [6,7]. On the other hand, the extent to which consumers value and respond to environmentally-friendly food products through value-consistent behaviour still remains a questionable point [8].
Such concerns should ideally lead consumers to be willing to pay higher prices for products that have been produced in respect of ethical/environmental attributes, such as animal welfare, health-related features, or aspects related to environmentally-friendly production systems [9]. This is especially relevant in the animal production sector, where many consumers expect animal welfare

Literature Review
Literature on environmental sustainability labels has improved the understanding about what may lead consumers to choose such labels and the corresponding products. For example, researchers [34] reported that a sustainability label is the most efficient method to increase consumer ability to choose an environmentally friendly food product. Other studies [35,36] also showed that many consumers are displaying increased awareness and preference for environmental sustainability, as well as a greater willingness to pay for socially and environmentally-responsible labelled products.
Ecolabelling is an increasingly used tool being applied to differentiate food production and stimulate informed purchasing decisions, thus creating economic incentives for producers to adopt environmentally friendlier technologies. Food products with eco-labels have been found to be preferred by consumers [37,38]. Moreover, a study [39] has shown that consumer stated preferences for eco-labelled goods increase with environmental consciousness and decrease with price-orientation. In addition, it was revealed [40] that consumer desire to preserve the environment is a key concern when choosing eco-labelled products.
In this context, over the last three decades various product standards certifying sustainable production and labels communicating sustainability-related information about the production method of food products have been put in place [41]. The standard of organic production has been one of the most widely known to consumers, both in Europe and the rest of the world, with organic food products becoming suitable examples of sustainable food due to their lower environmental impact [42,43].
Researchers [21] found that consumers place increasing importance on the extrinsic quality attributes of meat products in response to rising concerns on safety, health, convenience, ethical factors, etc. Nevertheless, a study carried out with European consumers on their awareness of sustainability issues-such as carbon footprint, animal welfare, or Fair Trade-found that consumer concern does not necessarily translate into purchasing behaviour, due to the various trade-offs consumers need to consider when shopping. Furthermore, a lack of transparency, credibility, and availability of information about ethical characteristics of production can also reduce the role of ethical product attributes in decision-making [44].
In parallel to the development of markets for sustainable food products, the geographical origin of food has become more important for consumers. Consumer preferences for the country of origin (COO) have been studied in various contexts [45][46][47]. Moreover, a recent study showed that COO might be an important product attribute to target a wider range of consumers and to differentiate markets for sustainable food products [1]. However, some consumer segments that are interested in sustainable production have also been found to place little importance on COO [1,48].
In addition, there is plenty of evidence that COO affects consumer food choices and that consumers are more willing to buy food products originating in some countries than in others [49,50]. Nonetheless, knowledge about the COOs that are preferred by consumers for their food remains insufficient and seems to differ by food product type [51].
Furthermore, previous studies suggested that animal welfare was a significant determinant and an important reason for the purchase of eco-friendly produced foods [52,53]. Animal welfare is receiving increased attention by consumers and is seen as an emerging quality attribute that is linked to specific food products such as beef, with many exporting countries often having this type of certification available [47]. Consumers are concerned about the way animals are bred, fed, and how animals are taken care of. However, animal welfare is often not the most important choice attribute in meat [54].
Finally, price is another important extrinsic factor that can affect consumer purchase decisions [55]. Moreover, price sensitivity has been reported to be of great influence in consumer decisions, as people with higher price sensitivity may compromise their environmental concerns to choose cheaper but less environmentally-friendly products [56]. However, a recent study showed that consumers who were less sensitive to price, shifted their attention from product appearance and price to environmental aspects, when evaluating eco-labelled products [57].

Methodological Procedure and Data Collection
The data used in order to develop the present study were obtained from an online survey carried out in Extremadura, a region in southwest (SW) Spain. The region was selected to be representative of Spain, because its socio-demographics are similar to the Spanish Census of Population and also because it has some characteristics that were considered of potential interest, as it is one of the main beef production areas in Spain and with highly relevant extensive and sustainable livestock production systems. Figure 1 presents the methodological procedure followed for this piece of research. The survey was designed and distributed using Google Forms (www.docs.google.com), which was chosen for its flexibility and benefits for the development of surveys. Even though, according to a study [58], online surveys present important advantages (i.e., lower research costs and little timeconsuming for respondents), this data collection technique might introduce some bias in terms of overrepresentation of some socio-demographic characteristics, which limits the potential inference of its results.
The sample was designed as a random stratified model, proportionally weighted against the gender and age of the population in Extremadura, with the final sample including 285 valid completed questionnaires. The maximum margin of error was 5.9% at a 95% confidence level.
Data were collected during March 2020, and respondents were recruited from databases that had been created from previous marketing studies conducted by the research team. The online questionnaire was organised in five sections to measure the following aspects: (a) purchasing habits of meat; (b) personal concerns for the environment; (c) lifestyle and willingness to pay for meat; (d) choice experiment and consumer preferences; and (e) socio-demographics aspects. The questionnaire was pre-tested on a sample of 12 individuals (not included in the final sample) to detect any possible misinterpretation, error, or duplication. Adjustments were made to the final questionnaire based on The survey was designed and distributed using Google Forms (www.docs.google.com), which was chosen for its flexibility and benefits for the development of surveys. Even though, according to a study [58], online surveys present important advantages (i.e., lower research costs and little time-consuming for respondents), this data collection technique might introduce some bias in terms of overrepresentation of some socio-demographic characteristics, which limits the potential inference of its results.
The sample was designed as a random stratified model, proportionally weighted against the gender and age of the population in Extremadura, with the final sample including 285 valid completed questionnaires. The maximum margin of error was 5.9% at a 95% confidence level.
Data were collected during March 2020, and respondents were recruited from databases that had been created from previous marketing studies conducted by the research team. The online questionnaire was organised in five sections to measure the following aspects: (a) purchasing habits of meat; (b) personal concerns for the environment; (c) lifestyle and willingness to pay for meat; (d) choice experiment and consumer preferences; and (e) socio-demographics aspects. The questionnaire was pre-tested on a sample of 12 individuals (not included in the final sample) to detect any possible misinterpretation, error, or duplication. Adjustments were made to the final questionnaire based on this test. Table 1 shows the socio-demographic characteristics of the final sample compared with those of the population of Extremadura.

Choice Experiment
In order to analyse consumer preferences for meat attributes, a choice experiment approach was carried out in this piece of research. CE is a technique used to analyse consumer preferences for various product attributes [60]. Due to its potential, CE has been widely applied in studies dealing with the analysis of individual preferences for food, including meat [28][29][30][31][32][33]61,62].
CE is based on Lancaster's consumer theory [63] and assumes that the utility that a consumer obtains from a product derives from the attributes that make it up, and which, therefore, affect his/her purchasing decisions. In a CE, participants are provided with alternative configurations of the product being assessed and are asked to choose between the range of options. The selection of the attributes-and their levels-that will define the product is a key stage which must display the product characteristics and dimensions that are most relevant regarding the consumer's purchase decision process.
Compared to other methods used to analyse consumer preferences (e.g., contingent valuation, experimental auctions), CE has the advantage of resembling a real purchasing situation [36,62]. The CE approach also allows to explore how attributes are related to each other, even though respondents are not asked to directly answer how important each attribute is for them [47].
For this piece of research, the attributes and levels to be used in CE were selected after a review of the existing literature in consumer preferences for meat [12,13,15,30,33,45,61,64] and according to preliminary research. The attributes finally selected for the study and their corresponding levels are presented in Table 2. Although the study had a main focus on meat in general, it was considered that it should be narrowed down to a specific type of meat that the participants could assess more easily. Therefore, the product presented to the participants in the CE was pre-packaged sliced beef (500-g tray). Beef was chosen not only for being one of the most popular meats in Spain in terms of consumption, but also because it is one of the most targeted meats for its contribution to climate change. Therefore, the improvement of sustainability in its production processes could be highly valued by consumers.
After the selection of the attributes and levels, these are merged to create hypothetical products, which are later on combined in pairs to create a choice set. Each choice set consisted of three alternatives: alternative product A, alternative product B, and the "no purchase" option, which allowed the participants to choose none of the products offered (see Figure 2 for an example). Based on the chosen attributes/levels, the total number of hypothetical products would have been 48 (2 × 3 × 2 × 2 × 2) with 2256 (48 × 47) possible combinations. As this figure was considered to be too large for participants to evaluate, an orthogonal fractional factorial design was used to reduce the number of comparisons, with 8 choice sets being finally presented to the participants. The order of presentation and allocation to respondents of the various choice sets was randomised. Figure 2 presents an example of choice set.
Sustainability 2020, 12, x FOR PEER REVIEW 6 of 16 (2×3×2×2×2) with 2,256 (48×47) possible combinations. As this figure was considered to be too large for participants to evaluate, an orthogonal fractional factorial design was used to reduce the number of comparisons, with 8 choice sets being finally presented to the participants. The order of presentation and allocation to respondents of the various choice sets was randomised. Figure 2 presents an example of choice set.

Conditional Logit
Conditional logit, a model based on the Random Utility Theory [25][26][27], has been applied in this paper to assess consumer preferences. The clogit module of R statistical package version 3.6.3, was used, following the guidelines described previously [65]. Base levels have been defined for all the qualitative attributes, which allow to set a zero-utility level with respect to the other levels of the attribute. The selected base levels were "Imported" (for the attribute Country of origin), "Conventional" (for Production method), "Without" (for Animal Welfare Label), and "Without" (for Eco-label claim).
The econometric specification is therefore defined as follows: The inclusion of price as an attribute in a choice experiment allows to calculate the marginal substitution ratio between a coefficient and the price, that is, the willingness to pay for the specific attribute. WTP is calculated as follows:

= −
(2) Therefore, WTPk reflects the amount of money that a consumer would be willing to pay to go from the base level to the level of the attribute k provided by the product.

Cluster Analysis
The identification of subgroups of consumers with similar preference behaviour towards sustainable meat was considered to be a valuable finding of this research. Therefore, a k-means cluster analysis was carried out with the Cluster module of IBM SPSS Statistics v 21.
The inputs used were variables related to beliefs and attitudes towards meat production/consumption and environment, together with others measuring purchasing habits and following the procedure developed previously [66]. Although various solutions were checked, a 3-

Conditional Logit
Conditional logit, a model based on the Random Utility Theory [25][26][27], has been applied in this paper to assess consumer preferences. The clogit module of R statistical package version 3.6.3, was used, following the guidelines described previously [65]. Base levels have been defined for all the qualitative attributes, which allow to set a zero-utility level with respect to the other levels of the attribute. The selected base levels were "Imported" (for the attribute Country of origin), "Conventional" (for Production method), "Without" (for Animal Welfare Label), and "Without" (for Eco-label claim).
The econometric specification is therefore defined as follows: The inclusion of price as an attribute in a choice experiment allows to calculate the marginal substitution ratio between a coefficient and the price, that is, the willingness to pay for the specific attribute. WTP is calculated as follows: Therefore, WTP k reflects the amount of money that a consumer would be willing to pay to go from the base level to the level of the attribute k provided by the product.

Cluster Analysis
The identification of subgroups of consumers with similar preference behaviour towards sustainable meat was considered to be a valuable finding of this research. Therefore, a k-means cluster analysis was carried out with the Cluster module of IBM SPSS Statistics v 21.
The inputs used were variables related to beliefs and attitudes towards meat production/ consumption and environment, together with others measuring purchasing habits and following the procedure developed previously [66]. Although various solutions were checked, a 3-cluster solution was finally chosen, due to the groups generated being adequate in size and being of the highest statistical significance. Table 3 shows the aggregate results of the conditional choice model. The value of the coefficient of each level indicates the utility added (positive sign) or detracted (negative sign) to or from the reference level. The results in Table 3 show that all the attributes (except price) have a positive impact with respect to their reference levels on the utility of the respondents. For example, the results for the "Country of origin" attribute indicate that consumers obtain more utility when choosing beef produced in Spain than imported beef. Something similar happens with organic production or the presence of animal welfare and eco-labels, which have a positive impact on the preferences of respondents compared to their baseline reference levels.

Choice Experiment Model
These results are in consonance with those found by other researchers [67], who reported that consumers showed greater preference for local food products than for imported ones. Researchers [68] mentioned that one of the main drivers of consumer preferences and attitudes towards foods is the country of origin. Moreover, another study [21] found that the origin of beef was the most important piece of information demanded by European consumers.
On the other hand, there are studies which indicate that consumer concern about animal welfare is the most important factor when purchasing meat products [69,70]. In a similar way, some researchers [71] found that organic production is very relevant for consumers when buying meat, while others [72] found that 74% of European participants preferred animal-friendly meat to conventional meat.
In addition, the fact that the sign of the price attribute is negative indicates that, as the price of beef decreases, its utility for consumers increases. This means that the probability of choosing a product with a lower price is higher, a result that is consistent with the habitual behaviour of demand. These results are in line with those reported by other researchers [73], who found that the higher price of eco-friendly food products is often perceived as a limiting factor for the purchase of these products. Table 4 presents the detailed socio-demographic characteristics and purchasing habits of the three segments that were generated by the cluster analysis, together with those of the general sample. It also shows the results of Chi-squared tests carried out to look for significant differences between the clusters.  Table 5 shows the data regarding beliefs and attitudes towards meat consumption/production and environment of the three clusters and the general sample. This information is complementary to that presented in Table 4 and allows to better define consumer typologies:

Consumer Segmentation
• Cluster 1 (Male millennials indifferent towards environment or sustainability). This segment has the highest percentage of men and of people under 36 years old. This group has the lowest level of consumption of environmentally-friendly products, but also has the poorest scores on all variables relating to environmental and sustainable behaviour, despite its high level of education.
• Cluster 2 (Sustainability-concerned mature women) displays the largest percentages of women and mature consumers (over 50 years of age). This cluster has the highest frequency of consumption of environmentally-friendly products, and it is also the one that gives the greatest scores to aspects relating to the environment and sustainability. It is also the group with the lowest consumption of meat.
• Cluster 3 (Middle-aged meat eaters with established families) presents the highest percentages of consumers between 36 and 50 years and of larger family units. It also has the highest frequency and level of meat consumption (which can be related to family size). Their members are middle ground, in terms of their attitudes towards sustainability and the environment.

Consumer Preferences Within Each Cluster
Once the clusters had been identified, CE was carried out again for each of the consumer groups. Table 6 lists the results of the choice model for each cluster.  Consumers in Cluster 1 are characterised by showing no significant preference between organic or conventional beef production. They look for beef with animal welfare label although they show the lowest preference for beef with eco-labels. Noticeably, they are the most price-sensitive group of the whole sample, a fact that could relate to their lower income level. These results are consistent with those reported by other researchers [44] who found that price-sensitive segments are slightly overrepresented by men. Additionally, a previous study [74] mentioned that price seems to have a higher impact than logo and/or label when purchasing meat products. It is also well known that attitudes and behaviour towards sustainable food consumption differ in multiple ways between genders [75].
Cluster 2 shows the highest preference of the three clusters for organic production and for the purchase of beef with animal welfare label and eco-label; these results are consistent with the sustainability concerns of the cluster. Consumers in this group also present the lowest sensitivity to price of the entire sample. These results are in agreement with those found by other researchers [76] who stated that women generally show higher willingness to pay for environmentally-friendly food products. These findings are also in line with previous research where female participants declared higher levels of sustainable consumption compared with male participants [75]. Moreover, a previous study [77] found that, with higher levels of education, men revealed increased meat consumption, while women showed reduced consumption. However, and contrary to other studies conducted in other European countries, a study [78] found that men showed more environmental concern and more positive outlook towards green purchase compared with women. Finally, Cluster 3 shows similar preferences to those obtained for the overall sample. In this group, consumers have the highest preference for the Spanish origin, a behaviour that could be related to their age structure (middle-aged consumers) and to the fact that they have children in their families. As many prior studies have noted, the preference for national/local food products is quite common, so that consumers are usually willing to pay a premium price for domestic and local food [49,79].
When making food purchasing choices, consumers can rely on those attributes that are most important to them or make trade-offs amongst a range of attributes [80]. Moreover, they also need to make trade-offs between both positive benefits-such as animal welfare-and (additional) price [81,82]. In addition, it has been found that consumers who show preference for food products associated with ethic-related claims are those who are also genuinely concerned about environmental sustainability [83]. Moreover, a prior study [84] mentioned that providing detailed information about beef processing technology increased consumer acceptability of beef products. It seems clear that label information does more than just provide product-related knowledge to consumers, as it affects their acceptance and purchase intention with respect to the corresponding beef products. Finally, another study [85] showed that beef labels are important sources of information about meat quality for consumers.

Willingness to Pay
One of the most interesting aspects of choice experiments, when price is included as an attribute, is the possibility of determining the WTP or the implicit price for each attribute. The WTP should be understood as the difference in euros between the price the consumer is willing to pay for a particular level, in comparison to the baseline reference level. Table 7 presents the WTP for the overall sample and for each of the clusters. The explanation of the results shown in Table 7 is that the respondents in the overall sample would be willing to pay € 4.71/500 g more for a beef produced in Spain compared to imported beef, or that they will pay € 1.36/500 g more for organic beef compared to beef produced by conventional production means. These results are consistent with those reported by researchers [23] who indicated high consumer preference and WTP for local food. A study [86] reported that consumers were ready to pay premium prices for food products deriving from quality production methods. Additionally, researchers [87] found that consumers in Italy were willing to pay a premium price for organic beef, with similar results also being reported by others [88].
On looking at the figures for the various clusters, we can deduce that they are in agreement with their previously defined characteristics. Thus, consumers in Cluster 1 present the lowest willingness to pay for all the attributes under analysis (they were very price-sensitive), while consumers in Cluster 2 show the highest willingness to pay, which is consistent with the low importance they placed on price. In this context, a recent study [89] revealed that consumers are willing to pay a price premium of approximately 20% for carrots and strawberry jams, if these products are eco-friendly labelled. Similar findings have also been reported for wine [90] and coffee [91]. However, other researchers [92] found that gender affected the understanding of sustainability labels and indicated that young male consumers had a better understanding of such labels.
In this sense, it should be noted that many consumers actually have little understanding of the real meaning of sustainability labels [93]. In this sense, lack of comprehension of the labels could lead to consumers processing information about the claim incompletely or incorrectly [94], resulting in misinterpretation or misunderstanding.

Conclusions
Public interest in sustainability issues has grown significantly in recent years due to growing awareness of climate change and the environment. Thus, an increasing number of consumers are concerned about the environmental, ethical, and animal welfare impact of their food. In this context, this study has explored consumer preferences and willingness to pay for sustainable and environmental attributes of beef, using a choice experiment.
As stated in this paper, one of the contributions of this study is the segmentation of consumers based on their preferences for a wide range of sustainability-related attributes (animal welfare, environmental impact, and production method) and other more traditional product characteristics (country of origin and price). This is in contrast with most consumer studies that have essentially analysed the willingness to pay for animal welfare or sustainability attributes. The selection of these variables was based on a preliminary review of the literature, which identified potential items to include in the survey. Subsequently, those items that were considered most relevant were selected to generate information on the previously mentioned aspects. A conditional logit model showed that consumers preferred beef which had been organically produced in Spain, bearing an animal welfare label as well as eco-labelled. In addition, a cluster analysis was carried out using variables relating to beliefs and attitudes towards meat production/consumption and the environment, along with others that measure purchasing habits, resulting in three well-defined clusters. In this sense, "the Indifferent millennials", the most numerous group, proved to be the cluster that least consumes environmentally-friendly products, as well as having the worst environmental and sustainable behaviour. On the other hand, the "Sustainability-concerned women" presents the highest percentage of mature consumers, with the largest frequency of consumption of environmentally-friendly products, but with the lowest consumption of meat. Finally, Cluster 3 (Middle-aged meat eaters) includes consumers with the highest frequency and level of meat consumption and who are middle ground in terms of their attitudes towards sustainability and the environment.
This study will help develop more appropriate strategies to understand the behaviour and expectations of eco-friendly food consumers. Nonetheless, given the monetary constraints that prevented the study from being carried out nationwide, we regard our findings as not fully inferable to the rest of the population. In this sense, future research should aim at obtaining a deeper insight into certain consumer segments. This would allow an exhaustive analysis of the various environmental labels available in order to measure the willingness to pay and to discriminate the real importance that consumers attribute to the sustainable quality of a product. In the future, these aspects will positively contribute to clarify new patterns of consumption and their incidence in the production systems, as well as in the implementation of incentive policies for a more responsible consumption.