The questionnaire comprised of general questions about the participants, as well as specific questions concerning their preferences at the start of the survey. An overview of these questions is shown in Table 1 . Most questions had predefined answers to choose from, whereas for some an open answer was allowed. These questions were translated into sociodemographic characteristics.

For respondents that indicated they did not purchase filet mignon themselves, the survey ended after the general questions. These respondents were not shown the Discrete Choice Experiment (DCE), as their choices were not deemed representative for consumer behaviour on filet mignons.

Besides general questions, a DCE was integrated in the survey to determine if ‘type of animal health monitoring’, ‘animal welfare label’ and ‘price’ (product attributes) influence the consumers’ preference for filet mignon. Discrete choice experiments are a popular elicitation format to investigate an individual’s preferences (Lusk and Schroeder, 2006), and are often used in agriculture (Čop and Njavro, 2022; Lauterbach et al., 2023). These experiments are classified as ‘stated preference’ approaches, in which respondents make purely hypothetical choices when presented with a set of alternatives (Meenakshi et al., 2012). The approach allows to draw conclusions from previously unarticulated preferences about real conditions using an attribute-based measure in a hypothetical decision-making context (Fecke et al., 2018). However, the disadvantage is that the results may suffer from bias as there is no actual commitment from the respondent to buy the product or service (Meenakshi et al., 2012). The results should therefore be interpreted with care. In practice, choice sets including different alternatives are presented to the participant, who is invited to choose the most preferable alternative. For each presented alternative, the attributes and their associated levels are pre-defined. Furthermore, these attributes are systematically varied in their levels, leading to different choice sets. In this way, the respective influences on the participant’s choices in the DCE can be detected (Louviere et al., 2000; Fecke et al., 2018).

In this survey, respondents were asked to choose one of three alternative filet mignons. Each filet mignon alternative was described by three product attributes, i.e. ‘type of animal health monitoring’, ‘animal welfare label’ and ‘price’. To be as realistic as possible and to avoid potential bias, the attribute levels were chosen based on existing welfare labels, the authors’ experience with PLF, and a small market survey at discounters, retailers and local butchers. The attributes and their different levels are presented in Table 2 . The ‘type of animal health monitoring’ consisted of three levels, i.e. Level 1 being monitoring by the pig farmer together with the veterinarian without support of a specific technology, Level 2 being monitoring by the pig farmer and the veterinarian supported by an animal health monitoring technology at group level (for example, sensors that monitor eating and drinking behavior per pen and that generate an alarm at pen level), Level 3 being monitoring by the pig farmer and the veterinarian supported by an animal health monitoring technology at individual animal level (for example, monitoring pig behavior using electronic ear tags and generating an alarm at pig level). The ‘animal welfare label’ was subdivided in four labels, ranging from ‘No label’ to ‘Label 3’. The label specifications were based on the minimum requirements of the ‘Beter Leven Keurmerk’ deployed in the Netherlands (Beter Leven, 2018a, 2018b, 2018c) concerning the minimum time the piglets stay with their mother before weaning, and the pen area and outdoor area per fattening pig. These specifications are presented in Table 3 . ‘No label’ corresponds to the standard legal requirements for animal welfare. The nine price levels ranged from 8 to 24 €/kg and were selected based on the prices found in the local sales locations during a small market survey. The median price found in the market survey of 12 €/kg was considered as the reference.

In a full-factorial design, all possible combinations of the levels of the different attributes would result in an unpractical large number of choice sets (3 x 4 x 9 = 108). To obtain the smallest possible orthogonal design and to keep the number of respondents needed for the analyses low, the number of choice sets was reduced using Ngene 1.2 (Choice-Metrics, Sydney, Australia). In total, 36 choice sets were made. These choice sets were allocated to 3 blocks of each 12 choice sets, and each participant was shown only one of these 3 blocks. This was done to avoid the questionnaire being too extensive and possibly triggering the participants to end the survey prematurely.

Each decision situation presented to the participants (choice set) contained one reference, which is considered as a standard filet mignon found in most shops. This reference had the same attribute levels throughout the survey, i.e. animal health monitoring by the pig farmer, no animal welfare label, and a reference price of 12 €/kg. An example of a choice set is given in Table 4 . The different levels were explained before the choice sets were given, and respondents could check the description of the levels of the attributes at any time during the DCE.

The random utility theory (RUT) proposed by Thurstone (1927) forms the theoretical foundation of the DCE. This theory was later extended by McFadden and others (e.g. McFadden, 1986; McFadden and Train, 2000). The theory proposes the use of ‘utility’ and assumes an individual attaches a certain utility to each option that can be chosen (Louviere et al., 2010; Van De Gucht et al., 2017). According to this theory, individuals will always try to maximize their benefit (McFadden, 1974), and hence choose the product with the highest utility (or benefit). Through the use of DCE, the individuals’ utility for each tested product attribute, as well as the influence of each product attribute on the probability of choosing a product can thus be revealed (Profeta et al., 2021). These utilities can be summarized by systematic (explainable) components and random (unexplainable) components. Systematic components comprise attributes explaining differences in choice alternatives and sociodemographic characteristics, which are other variables that explain differences between individuals. Random components consist of all unidentified factors that influence the choices (Louviere et al., 2010; Van De Gucht et al., 2017). The utility of a specific option ( U i j ) is given by.

where U i j is the utility of individual j for alternative i, V i j is the systematic component of individual j for alternative i, and ϵ i j is the random component of individual j for alternative i. When incorporating the attributes of the alternatives (X) and the sociodemographic characteristics of the respondents (S), the systematic component, V j , can be written as.

where β 0 is the constant which stands for the intrinsic utility of an alternative irrespective of the attribute and levels, β k j is the weight of the k^th attribute for individual j, X k j is the k^th attribute for individual j, β n j is the weight of the n^th sociodemographic characteristic for individual j, and S n j is the measurement of the n^th sociodemographic characteristic for individual j. In this study, the constant β 0 has no true meaning as all alternatives are pieces of filet mignon, and it is therefore deleted from the models.

The participants’ choices were analyzed with Nlogit 6 (Econometric Software Inc., Plainview, NY) using a multinomial logit model, as described by Van De Gucht et al. (2017). First, a base model that describes the utility of each filet mignon alternative was defined using the 3 attributes from the discrete choice experiment. Type of animal health monitoring and animal welfare label were categorical attributes, hence a coefficient was estimated for each level.

Where β is the attribute constant, Health is the type of animal health monitoring, Welfare is the animal welfare label, and Price is the price of the filet mignon.

In a second step, the sociodemographic characteristics of the respondents were tested in an univariable way by adding them separately to the base model. Binary categories were tested using effect coding. Regression coefficients (β) were calculated, indicating if a variable had a positive or negative contribution to the model outcome, being the utility. Variables were selected for multivariable analysis at significance level P< 0.15. Third, a multivariable model was built with the sociodemographic characteristics that were selected from the univariable model using a backwards stepwise procedure at P< 0.05. Finally, Wald post-hoc tests were performed.

As stated above, the coefficients indicate if the contribution of a specific characteristic contributes in a positive or negative way to the utility. However, the value of the estimated model coefficients and thus the magnitude of the impact on the outcome, depends on the range of attribute levels used in the DCE, and should therefore be interpreted with care. Coefficient estimates should not be compared between attributes or sociodemographic characteristics included in the model, but between utilities assigned to alternatives with specific attribute levels (Hensher et al., 2005; Van De Gucht et al., 2017). However, the utility derived from the choice model as described above only allows comparison with the reference, of which the utility was set to zero. Coefficients higher than zero indicate that the specific level contributes positively to the utility of the alternative, and is thus preferred by the respondents compared to the reference. When a continuous attribute or characteristics (e.g. price) is increased by 1 unit, the utility assigned to the alternative increases with the attribute’s coefficient estimate in case of a positive estimate, and vice versa, provided that all other attribute levels are kept constant.

Discrete Choice Experiments are an appropriate tool for calculating a measure of WTP (Hensher et al., 2005). The WTP is the maximum a customer is willing to pay for a proposed product with its specific characteristics. It indicates how much more a product can cost if a an attribute changes to another level. For the different attribute levels, the WTP was calculated by dividing the previously estimated attribute coefficient by the price coefficient of the univariable model. As the calculation of the WTP is a ratio of two estimated coefficients depending on the range of attribute levels used in the experiment, the WTP is also an estimation depending on these attribute levels. Therefore, the selection of the price levels should be done with care in order to become a realistic measure for WTP. The levels must accurately capture the range of preferences for most of the respondents. It is important to ensure the price levels are not too high or too low for attributes under evaluation, in this case health monitoring and welfare label. Typically, ranges around the mean costs are included. Price levels in this study were selected based on market prices. A market study was conducted, including different market channels from discount supermarkets to short chain farm butchers. Nine levels from 8 to 24 €/kg were chosen around the median price of 12 €/kg (see Table 2 ).

The first goal of this study was to investigate whether the type of animal health monitoring, the animal welfare label and the price of filet mignon are important to consumers. Animal welfare label and price significantly affected the utility. They are therefore important attributes to the respondents that influence their choice of filet mignon. This finding is supported by the survey results in which 70% of the respondents acknowledged the need for an animal welfare and health label and 83% of this group indicating to be willing to pay extra for such a label. The utility increased with increasing animal welfare label level, although not significantly between Label 2 and 3. With regard to the label requirements, there is only little difference between ‘No label’ and ‘Label 1’. More specifically, within ‘Label 1’ the time the piglet can stay with the mother before weaning is between 23 and 28 days, which corresponds to potentially two additional days with the mother compared to the 21 to 28 days specified in the minimum ‘No label’ requirements. Within ‘Label 1’, the pen area per fattening pig should also be slightly larger than that of ‘No label’, i.e. 1.0 m² vs. 0.8 m². These small differences in requirements suggest that simply having an animal welfare label is sufficient to increase the utility of the meat. The results furthermore suggest that an outdoor area increases the utility, as this is what mainly distinguishes ‘Label 2’ and ‘Label 3’ from ‘Label 1’. Besides the outdoor area also the other two welfare requirements slightly differ between ‘Label 1’ and the higher level labels. However, the seven additional days of the piglet with the mother, the 0.2 m² extra pen area and the 0.3 m² extra outdoor area per pig, did not significantly increase the utility of ‘Label 3’ compared to ‘Label 2’. These findings support the idea that consumers may see the outdoor area as the most important added value for animal wellbeing. However, the set-up of this survey with multiple requirement differences between the levels of the animal welfare label (i.e. time with mother before weaning, pen area and outdoor area per fattening pig) does not allow to draw conclusions regarding the relationship between the specific requirements of each level and the consumers’ preferences. More studies are needed to determine the importance of the animal welfare label, as well as its specific requirements, with regard to consumers’ preferences.

The type of animal health monitoring did not have a significant effect on the utility of the filet mignon, suggesting that the respondents pay little attention to type of health monitoring or they are neutral towards the monitoring technology. Since no added value of monitoring technology was observed, consumers might believe that monitoring by the farmer and the veterinarian is sufficient and good animal health is already assured in the current system. Comparable, as no adverse effect of health monitoring technology was found either, there are no indications for consumers having a negative attitude towards health monitoring technology. Focus group discussions in three European countries (i.e., Finland, Spain, the Netherlands) had previously identified some major concerns of consumers regarding the implementation of PLF technologies which may affect their acceptance (Krampe et al., 2021). These concerns included the fear that integration of PLF technologies will introduce more industrialization into livestock farming production, that PLF technologies and data are vulnerable to misuse and cyber-crime, and that PLF information is not communicated adequately to allow informed purchase decisions. In this survey, the respondents were not explicitly asked to indicate their concerns on the implementation of PLF, as to not influence their choices. However, as indicated above, even if these concerns could have been present in these respondents, the utility of the meat in this survey was not influenced.

A more in-depth study on the importance of health monitoring technology (independent of animal welfare) on the utility of pork, and other animal products in general, is needed to better assess the possible added value of this technology, as well as to identify potential means to improve the perception of the technology. Public campaigns for example might improve the general awareness and appreciation of PLF technologies.

Not surprisingly, an increasing price of the meat had a negative effect on the utility. Increasing the price of the meat with 1 € decreased the utility by 0.191. Price was found to be the most important information customers look for before they purchase a product (Kumar and Kapoor, 2017). Profeta et al. (2021) also reported a significant, negative effect of increased price on the choice for four meat products (mortadella, salami, chicken nuggets, meat balls) in Germany and Belgium based on a DCE. However, the importance of price generally depends on the consumer profile, with ‘indifferent meat consumers’ being strongly price oriented, while this is significantly less so for ‘cautious meat lovers’, as found in a study performed by Verbeke and Vackier (2004) based on data collected in Belgium. Economic motives, such as the lack of spending power or willpower to purchase good quality meat, may have triggered the observed indifferences of the first consumer group. Future studies on PLF technologies should take into account the different consumer profiles to identify and gain insight into the market segments with highest preference for PLF products in order to better position these products through the development of effective communication and marketing strategies.

The second aim of this study was to determine how sociodemographic characteristics influence the consumers’ preference for filet mignon. The preferences were found to be influenced by gender, frequency of buying filet mignon, and the attitude towards organic and animal welfare and health labels. Female respondents assigned a higher utility to a specific filet mignon than their male counterparts and hence also higher measures of WTP. Various studies have previously demonstrated significant gender differences with regard to food, including food preferences, intake, nutritional knowledge, and shopping behaviour (Beardsworth et al., 2002; Kumar and Kapoor, 2017; Manippa et al., 2017; Lombardo et al., 2020), but to the authors knowledge no studies have been performed on the effect of gender on the utility of food products. In this study, the higher utility assigned by women may be due this gender being generally more involved in food shopping and meal preparation. For example, a study performed in the Netherlands, reported 40% of (native) adult women to be involved in food shopping, compared to 36% of men, whereas the involvement in meal preparation was 51% for women but only 24% in men (Schösler et al., 2015).

The increasing utility for filet mignon with decreasing frequency of buying suggests that consumers who buy the product less frequent value it more. These consumers might be more conscious in their buying behaviour and their demands. In recent years, general awareness regarding a healthy and sustainable diet has led to a decrease in total meat consumption per capita in Belgium (Dagevos and Verbeke, 2022). It is becoming more evident that purchasing behaviour of consumers is not only influenced by their personal pleasure but also by more altruistic reasons, such as environmental sustainability and social and economic justice (Gracia et al., 2012). Vanhonacker and Verbeke (2009) found that in Flanders lower chicken meat consumption was associated with higher frequencies of buying pro-welfare chicken products. However, McEachern and Schröder (2002) suggested that lower meat consumption was mainly a compensation for the price premium of higher welfare products. Nevertheless, it is not surprising that consumer preferences are associated with factors related to sustainability and animal welfare, and these associations may at least partially explain the relationship between utility for pork meat, buying frequency, attitude towards animal welfare and organic labels. Indeed, a positive attitude towards animal welfare and health labels had a positive effect on the utility of filet mignon in this survey. In addition, respondents who consciously buy organic meat also assigned a higher utility to filet mignon, irrespective of the organic label itself. Using a DCE, Profeta et al. (2021) found that the presence of an organic label exerted a positive effect on the choice for mortadella, chicken nuggets, salami and meat balls in Germany and Belgium. Such a direct effect was however not investigated in this study.