Beef animal welfare , attitudes and Willingness to Pay : A regional comparison across the Pyrenees

Attitudes towards beef animal welfare (AW) and Willingness to Pay (WTP) for AW certification are investigated among consumers in two Spanish and two French regions located on both sides of the Pyrenees (n=1213). Attitudes were measured through a scale of 11 animal practices, on which, consumers report their degree of concern and trust on the supply chain compliance. Attitudes significantly differed across regions, especially with respect to those AW practices carried out by farmers, while trust lies behind concerns. Three segments based on individual consumer attitudes are defined by opposing those consumers who are more concerned and who trust more on the compliance with AW standards (n=264, 22%) to those less concerned and who are more uncertain about stakeholders ́ compliance with AW rules (n=356, 29%). Consumer location, gender, age and education significantly differed across attitudinal clusters. Results from a contingent valuation survey show that WTP for certified animal friendly beef ranged between 20.6% and 22.6% over the average market price of standard beef, in Spain and France, respectively. Both, consumers’ socio-demographic characteristics and habits regarding beef meat purchasing and attitudes towards farmers influenced this WTP (the more consumers trust in farmers’ involvement in animal welfare, the highest is their WTP), while a negative overall attitude significantly reduced WTP. Additional key words: meat; animal welfare labelling; WTP; cross-regional. Abbreviations used: AW (Animal Welfare); CI (Confidence Interval); CV (Contingent Valuation); EU (European Union); LLR (Log-Likelihood-Ratios); PDO (Protected Designation of Origin); WTP (Willingness to Pay). Citation: Sans, P.; Sanjuán-López, A. I. (2015). Beef animal welfare, attitudes and Willingness to Pay: A regional comparison across the Pyrenees. Spanish Journal of Agricultural Research, Volume 13, Issue 3, e0105, 14 pages. http://dx.doi.org/10.5424/ sjar/2015133-7273. Disclaimer: The views expressed are purely those of the authors and may not in any circumstances be regarded as stating an official position of the European Commission. This work has one supplementary figure that does not appear in the printed article but that accompanies the paper online. Received: 18 Dec 2014. Accepted: 21 Jul 2015 Copyright © 2015 INIA. This is an open access article distributed under the Creative Commons Attribution License (CC by 3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Funding: This study is funded by EU-FEDER Operative Programme of Territorial Cooperation Spain France Andorra, 2007-2012 (POCTEFA), Project OTRAC. Competing interests: The authors have declared that no competing interests exist. Correspondence should be addressed to Pierre Sans: p.sans@envt.fr


Introduction
Since the 80's, developed countries have put in place regulations to guarantee an acceptable level of animal welfare (AW).For instance, the European Union (EU) legislation establishes a set of standards, mainly affecting intensive production systems, which cover animal rearing, transportation and slaughtering1 , while a new strategy for the period 2012-2015 has been designed in order to better address detected weaknesses in the legislation and their enforcement, as well as to better inform the consumer (EC, 2012).These regulation initiatives aim at matching increasing consumers' concerns, impelled by communication campaigns carried out by animal protection organisations, food scares, and the frequent presence of animal welfare issues on the media.Several papers argue that animal welfare play an increasing role in consumers' purchasing behaviour of food equivalent to health, environmental and other ethical concerns (Napolitano et al., 2010b;Fernqvist & Ekelund, 2014).Nevertheless, to what extent these concerns affect the purchase decisions is an empirical question.2 attribute into a search attribute (Darby & Karni, 1973).Trust on the information delivered relies not only on the credibility of regulatory authorities but also on the perceived commitment of meat chain stakeholders in complying with AW superior standards (de Jonge et al., 2008;Vanhonacker & Verbeke, 2014).This latest aspect has been studied by Nocella et al. (2010Nocella et al. ( , 2012) ) who, in their cross-country study, found that WTP for AW labelling is more influenced by trust toward farmers' compliance with AW practices than other stakeholders (i.e.transporters).
Interestingly, the pan-European study by Nocella et al. (2010Nocella et al. ( , 2012) also highlights differences in terms of attitudes and WTP for animal welfare certification schemes between Northern (Germany, UK and France) and Southern EU countries (Italy and Spain), where the former group manifest a higher level of trust in stakeholders.However, this 'better' attitude in the Northern countries does not translate into a higher WTP.Likewise, Ipsos-LE ( 2013) also provides different WTP estimates for AW certification in different countries in the EU.In particular, they report 24% and 28% premia for broilers and eggs, respectively, in France, and 27% and 30% in Spain.On the contrary, some other studies consider that the country of residence is not among the most influential variables (Vanhonacker & Verbeke, 2014).Although cross-regional studies are encouraged by some authors to better account for regional patterns in consumer purchasing behaviour (Mittal et al., 2004;Thelen et al., 2006), these are scarce in the literature -an exception is Sanjuán et al. (2012) -while to the authors knowledge there are not regional studies in the AW field.
This paper investigates consumers' WTP for animal friendly labelling in beef production with a contingent valuation study, focusing on cross-regional/country comparisons and the role of attitudes towards AW and trust in stakeholders.The locations are two Spanish regions (Aragón and Catalonia) and two French regions (Midi-Pyrénées and Languedoc-Roussillon) contiguous to the Pyrenees mountains.Beef rearing is an important economic activity in the mountain regions, and the Pyrenees in particular, accounting for 24 and 17% of domestic beef sales, in Spain and France, respectively (Santini et al., 2013).Despite the economic importance of this sector, but probably due to its relatively more extensive production system, AW of cattle has not been analysed as much as other species but still 18% of EU citizens consider that beef cattle conditions should improve remarkably (Eurobarometer, 2007).Previous research on beef AW has been conducted by Carlsson et al. (2005) in a multi-attribute context, and Benett et al. (2012) in a multi-species basket, both applying hypothetical choice experiment techniques.To the A classical way to evaluate consumers purchasing behaviour is to measure their willingness to pay (WTP) a premium for food products that guarantee a superior level of AW than that provided by the standard regulation (Lagerkvist & Hess, 2011).Bennett (1997) and Bennett & Blaney (2003) estimate WTP on the ban of battery cages in egg production in the UK; Moran & McVittie (2008) on changes in stocking and other husbandry practices in broilers production in the UK; Tonsor et al. (2009) on the ban of pig gestation crates in US; and Taylor & Signal (2009) on compliance with generic farm AW standards in Australia.All these studies have used the contingent valuation (CV) method, while alternative methodological approaches include hypothetical choice experiments, where either different levels of AW are jointly considered (Carlsson et al., 2005(Carlsson et al., , 2007a,b;,b;Lagerkvist et al., 2006;Liljenstolpe, 2008;Bennett et al., 2012) or AW relieving practices enter as an additional attribute into a set of product or production characteristics, such as organoleptic characteristics and organic production process (Scarpa et al., 2013).
Animal welfare is credence good as the consumer cannot be certain about the production, transport and slaughter conditions even after consuming the food product.In this context, a label or certification becomes a useful tool of information and guarantee to the final consumer.Currently, there is not an EU global AW certification in place although its creation is envisaged (EC, 2012).A few papers have studied WTP towards the hypothetical EU Animal Welfare label in order to assess the benefits of the labelling proposal.Gracia et al. (2011) applied an experimental auction method to dry-cured ham in Spain, and estimated that WTP for an hypothetical EU-certification scheme that guarantees superior AW standards varies between 19% and 23% with respect to a non-labelled product, depending on the way the information is delivered.Kehlbacher et al. (2012) conducted a study in the UK for generic AW, and found that 94.3% of consumers would like a 'welfare scoring system' to label food products.Depending on the level of AW guaranteed, willingness to increase the monthly expenditure on meat varies between 26 and 34%.Van Loo et al. (2014) applied an hypothetical choice experiment to investigate preferences and WTP for an array of sustainability labels (i.e.organic, carbon footprint) in chicken amongst Belgium consumers and found that the EU animal welfare label ranks second in terms of WTP after free range claims, with premia of 26% for lower income and 39% for higher income consumers.
Provided that consumers rely on the certification or labelling scheme of AW, the information conveyed by the certification is a way to transform this credence 3 Beef animal welfare, attitudes and Willingness to Pay the first component of the attitudes or the 'outcome belief' can be interpreted as the importance attached to an array of AW issues (I) and the second component or 'belief strength', as the trust in stakeholders to comply with the AW rules (T).
As a result, a multiplicative formula between both components is applied to measure attitudes: where the sub-index i stands for each attribute or practice that compose the AW concept (i=1,...,W).
In our application, consumers were informed that farmers and other stakeholders involved in rearing and transport of cattle can improve the AW in different ways, while compliance with these animal friendly practices could be guaranteed through a certification or a label issued by public authorities.A total of 11 practices or attributes (Table 1) were considered, from which 8 coincide with those employed by Nocella et al. (2010).After verifying this list with experts in AW personally involved in their implementation3 , two additional items were added ('avoid that the animal suffers from fear and stress' -STRESS -and 'allow the animal to manifest its natural behaviour' -NATBEH), one was dropped because of difficulties to be understood by a focus group ('focus on the selection of animals to get a better productivity'), and one ('prohibit mutilations') was divided in two: MUTI ('prohibit mutilations, such as cutting horns or ears') and CAST ('prohibit castration'), to differentiate practices that may evoke a different degree of rejection among consumers.
For each of these items, both, the degree of importance or concern and trust in stakeholders to comply with those standards were investigated.Concerns were evaluated asking the consumer about the degree of importance assigned to the guarantee of each AW item, in a three point scale ("How important is it for you that…"), from 1 'Not at all important' to 3 'Very important'.Trust was evaluated asking the consumer about how likely they consider that the stakeholders involved in each stage will comply with those standards ("To what extent do you think is likely or unlikely…"), in a three point scale, from 1 'Very unlikely' to 3 'Very likely'.Due to the low variability observed by Nocella et al. (2012) with a five point scale, a three point scale was chosen to simplify the questionnaire.authors´ knowledge, however, WTP for the EU-AW labelling on beef has not investigated yet.An important result of Carlsson et al. (2005) is that the relative importance of AW attributes is animal specific, which justifies focusing on one particular species.In order to make comparisons across countries feasible, WTP is expressed as the premium over the price of beef meat currently available at the market, and which does not carry any AW labelling or claim.Special emphasis is also put on the identification of relevant personal traits that help to define the most receptive consumers' segment toward AW labelling in beef.

The survey
A representative sample of the regional population in terms of gender and age was recruited between September 2010 and April 2011, in the main cities of four regions located on both sides of the Spanish-French border: in Spain, Zaragoza in Aragón and Barcelona in Catalonia; in France, Toulouse in Midi-Pyrénées, and Montpellier and Perpignan in Languedoc-Roussillon2 .The survey was addressed to regular consumers of beef and involved in food shopping.The final sample was composed by 1213 consumers: 294 in Aragón, 304 in Catalonia, 317 in Midi-Pyrénées and 298 in Languedoc-Roussillon.The AW attitudes and contingent valuation questions were complemented with information on socio-demographics and consumption and purchasing habits of beef.The survey was administered face-toface, in a closed environment, where no more than 12 respondents participated simultaneously.

Measurement of attitudes towards animal welfare
The approach we follow in this paper to measure attitudes towards AW follows closely Nocella et al. (2010), which in turn is inspired by the expectancy value model (Fishbein & Ajzen, 1975).The expectancy value model states that a person's attitude towards an object depends on the 'outcome belief' or evaluation of the attributes associated with the object, and the 'belief strength' or subjective probability that the object possesses that attribute.Thus, for example, 4 where β 0, β 1 and β k are parameters to estimate and ε n are random errors, collecting the non-observable influences on WTP.With a double bounded approach, four outcomes or sequences of responses are possible, each one, with a probability that depends on the statistical distribution assumed for the errors, being the logistic one of the most commonly employed (Markosayan et al., 2009).Estimation is then carried out by maximum likelihood.
In order to avoid the starting point bias (Flachaire & Hollard, 2007), five different initial bids were employed in this study with a gap of ±5%.The subsequent bids were modified by ±10%, except when the subsequent bid would result in a price lower than market price, in which case the bid was reduced by 5%.Specific market prices were used for each location as a reference: 12.76 €/kg in France (MAAPRAT, 2009);9.25 € in Aragón and 9.85 € in Catalonia (MAGRAMA, 2010).Nevertheless, WTP estimates may still be upward biased as a result of the presence of the 'warm glow' and 'part hole' biases (Bennett & Blaney, 2003).Warm glow means that the respondent is willing to pay for moral satisfaction (i.e.contributing to a good cause), not for the good itself (Andreoni, 1990;Kahneman & Knetsch, 1992), and 'part hole' bias implies that the respondent is not only answering for the good under consideration (i.e.beef AW) but for a larger category (i.e.farm AW) (Carson & Mitchell, 1995).Likewise, in order to mitigate the hypothetical bias consumers need a reminder of their budget constraint (Sans et al., 2011), which was included in the text of the question and remarked orally.Suppl Fig S1 [pdf online] indicates, as an example, the WTP question formulated in Aragón.

The contingent valuation method
Contingent Valuation (CV) is one of the stated preference methods available to assess the value of goods or characteristics that either have a public utility and/ or are not available in the market.In comparison to other stated preference methods, like choice experiments, CV is especially suitable in the case of bans on certain characteristics or labelling (Carlsson, 2013).This probably explains the number of studies that apply the CV approach in relation to AW, either to value AW enhancing practices, including bans (e.g., Bennett & Blaney, 2002); or to get the economic value of a certification to guarantee animal friendly practices superior to the legal minimum standards (Bennett & Blaney, 2003;Nocella et al., 2010).
Among the different formats to apply a contingent valuation study, the double-bounded approach (Hanemann et al., 1991) has become the workhorse as is recognized as more efficient asymptotically (Hanemann et al., 1999).In this approach, each respondent faces two bids (i.e.price) and elects whether he/she is willing to pay the specified price.If the first bid is accepted, then a second one is offered of higher value.Conversely, if the first bid is rejected, the second is of lower value.
The latent (i.e.only partially observed) WTP of individual n is represented as a linear function of the last bid faced by individual n (Bid n ) and optionally some additional explanatory variables (X k,n ), such as attitudes or socio-demographics:  2).People older than 54 years old account for 26% of the sample, with the highest representation in Aragón (32%) (Table 2).
With respect to education levels, 36% of the respondents have a university degree, ranging between 28.6% in Catalonia and 49.6% in Aragón (Table 2), while 8.3% and 56% of the total number of respondents have reached primary and secondary studies, respectively.

Sample profile 4
The consumers interviewed are regular eaters of beef: 88% consume beef at least once a week at home.Besides, 34% of respondents also consume beef away from home on a regular basis, with percentages up to 41-43% in France.Consumers interviewed are mainly 4 For saving spacing reasons, sample characteristics are not included in a specific table.Some socio-demographics are reported in Table 2, while other purchase and consumption habits can be consulted in Sanjuán et al. (2012).a Definition of segments is explained in the section "Attitudes towards animal welfare: concerns and trust", and the segment profile is described in Table 4. b A F and A O are described in the section "Measurement of attitudes towards animal welfare".c Clusters definition and profiles are described in the section "Attitudes towards animal welfare: concerns and trust".

6
The multiplicative scores, although well-grounded in the theory, do not allow disentangling the influence of trust, which is one of our main focuses.Thus, two consumers may have the same A F value, for instance, but as a result of the combination of high importance/ low trust or the opposite.To answer this question, we have defined three groups of consumers as follows: first, individual scores for each of the 11 items for importance (and trust), were added (i.e. the total score for importance (trust) of an individual can vary between 11 and 33).Second, the median (29 for Importance and 21 for Trust) was used as a segmenting value.Third, segments were built up: consumers more concerned (with importance scores higher than the median) and who trust more (trust scores higher than the median) on the compliance with AW standards belong to the first segment with "relatively more favourable attitude" (N=264; 21.8%).The second segment, identified as "relatively more unfavourable attitude" (N=356; 29.3%) groups the consumers that are less concerned (importance score lower than the median) and who also trust less on the compliance with AW standards (trust score less than the median).Finally, remaining respondents belong to the "Mixed attitude" segment (N=567; 46.7%), which is mainly composed by people who do not trust as much as they are concerned (i.e. more than 50% of the respondents in this segment show a high degree of concern (above the median), while the majority (52%) are below the median trust.We will refer to this segmentation approach as "ad-hoc".
Complementarily, a k-medians clustering has been applied in order to be more flexible in the definition of the groups.Three clusters were also considered, and the initial centres were chosen randomly.A chi-square test shows a significant association between both segmentation approaches (p<0.001).The sizes of the segments are comparable although with favourable and unfavourable segments less polarized, as expected ("relatively more favourable attitude" 27.3%; "relatively more unfavourable" 26.2%; and "mixed attitude" 46.5%).

Attitudes towards animal welfare: concerns and trust
With the exception of the item "Prohibit castration" (CAST), a majority of consumers consider that the proposed AW attributes are very important.Depending on the attribute under consideration, between 77% and 98% of the consumers consider the proposed items "rather or very important".The highest rates are observed for the items directly linked with the living conditions of the animal ('avoiding unbalanced diet' -DIET, 'freedom of movement in stalls' -FREE, 'avoid that the animal suffers from terror and stress' -STRESS and 'providing sufficient space during transport' -TRAN).
In contrast, trust about stakeholders' commitment with AW is more polarized: no more than 40% of the respondents consider 'very likely' that stakeholders will comply with any of the AW attributes.The lowest trust is observed for the 'transport conditions' (TRAN), 'rest before slaughtering' (SLAU), 'stress prevention' (STRESS) and 'expression of a natural behaviour' (NATBEH).
Equation [1] is then calculated with all the 11 practices to provide a total measurement of attitudes (A), and also, differentiating between those practices carried out by farmers Using an ANOVA one-way test, attitudes are significantly different across regions in global (p<0.01), and with respect to those practices put in place by farmers (p<0.001) and stakeholders (p<0.10)(Table 3).In particular, consumers in Languedoc-Roussillon have significantly lower A, A F and A O values, and consumers in Catalonia show the most favourable attitude towards farmers.Attitudes towards stakeholders are very close in both Spanish regions and Midi-Pyrénées.Interestingly, the clusters defined by k-medians share the same socio-demographic profile, with the only difference that income becomes significant (i.e. consumers with a better attitude also enjoy a significantly higher level of income).Minor differences are found in terms of purchasing habits (e.g. the place of purchase is not significantly different across clusters), while those consumers with a more favourable attitude not only tend to buy more EU quality labels but also other branded beef meat.Therefore, both segmentation techniques provide segments of consumers, not identical, but very consistent in terms of size and composition.

Willingness to Pay
In Table 2, the explanatory variables used in the estimation of WTP are described.Originally, additional variables describing the respondent's consumption and purchasing habits of beef meat were also considered to By using the Chi-square statistic, we tested for the association between the segment of membership and an array of personal characteristics.Only those variables that are found to be significantly different according to the first segmentation approach are shown in Table 4 5 .In terms of socio-demographics, location, gender, age and education are found to be significantly different across attitudinal clusters.In particular, in the cluster with "relatively more favourable attitude", there is a higher proportion of respondents with primary studies (p<0.001),females (p<0.001),older than 35 years old (p<0.001),and residents in Catalonia and Midi-Pyrénées (p<0.01).Consumers of this cluster consider themselves as more experienced in relation to purchasing beef (p<0.05), while they use less supermarkets/hard discount stores (p<0.10) for buying beef.Moreover, EU quality labels, such as specific Protected Geographical Indication or Protected Designation of Origin (PDO) in each region, are bought by a higher proportion of members of this cluster than non-members (p<0.05). 5Results on the second segmentation approach are available from the authors upon request.8 titudes towards AW.Model 1 uses the multiplicative formula of importance and trust, segmented into attitudes towards farmers and other stakeholders (variables A F and A O ).Model 2 includes ad-hoc segments of consumers, defined relatively more or less favourable attitude towards AW according to the total scores on concerns and trust with respect to the median value (segments 1 and 2 in Table 4).Model 3 includes clusters of consumers identified as relatively more or less favourable according to the k-medians clustering technique.A specific constant is included in each model to account for special characteristics of respondents in each location that are not accounted by the remaining explanatory variables6 .
test if a higher involvement with the meat product translates into a more intense commitment with AW.In particular, the frequency of consumption at home and away from home, quantity consumed, and formats of purchase (carved on request, packaged, frozen) were not significant.Among the alternative distribution channels employed by the respondents to purchase beef (i.e.butcher's, specialized supermarkets, super-and hypermarkets, discount shops), only purchasing directly from the producer was found to have a significant effect.Finally, other socio-demographic variables like the presence of children in the household were not significant either.
In Table 5, three different model specifications are presented which only differ in the measurement of at- a LL0: value of the restricted log-likelihood in a model with intercepts and bids.b LL: maximum value of log-likelihood function with all explanatory variables.c LLR: Log-likelihood ratio to test for the joint significance of all but the intercepts and bid (i.e.uses LL0) as benchmark (Herriges, 1999).Critical value is 15.51 (chi-square distribution with 8 degrees of freedom).p-values in parentheses.d Wald test on the joint significance of all coefficients except for the intercepts (Harpman & Welsch, 1999).Critical value is 16.92 (chi-square distribution with 9 degrees of freedom).9 Beef animal welfare, attitudes and Willingness to Pay where X k (k=1,..,8) refers to each of the explanatory variables besides the bid and constants (valued at their mean regional levels, reported in Table 2), and β k the accompanying coefficients.To obtain mean values and a confidence interval (CI), the Krinsky & Robb (1986) parametric bootstrap technique is applied (following Park et al. (1991)) 9 .Results of WTP for each region are presented in Table 6.Model 1 has been used in the calculations, while Model 3 results are of very similar magnitude 10 .
Both statistics, Log-Likelihood-Ratios (LLR) and Wald test, support the joint significance of the explanatory variables, and the fit is slightly worse in Model 2 (i.e.LL is the lowest in Model 2).The coefficient estimates are also very similar in all models.The bid coefficient is significant and negative, thus, the probability of accepting a price diminishes when the bid size increases, as expected in a normal demand function.All the regional specific constants are positive and significant but of a different magnitude, which is confirmed by a Wald test 7 .
Women and older people (> 54) are willing to pay more for animal friendly certification than men or younger consumers.A positive relationship is also found between higher income levels and WTP for beef welfare certification, although this effect is only significant at 10% (5% in Model 3).Education, however, has a negative impact on willingness to pay, and people with university degrees are more reluctant to pay a premium for the certification of AW in beef.Interestingly, purchasing directly from the producer affects positively WTP, while experience with beef preparation stages affects negatively, although not significantly WTP.
With respect to attitudinal variables, a positive and significant effect is found for attitudes towards AW practices accomplished by farmers (A F ) while attitudes towards practices implemented by stakeholders (A O ) reveal a negative albeit insignificant impact on WTP for animal welfare certification.This result concurs with the pan-European study by Nocella et al. (2010) in the general context of farm animals.In contrast with the aforementioned study, however, we found a significant influence of attitudes towards farmers also in the Spanish regions, and not only in France 8 .In Model 2, the segment of consumers more concerned and who places a higher level of trust in stakeholders (variable Fav_Att) are also willing to pay more, while a negative attitude (variable Unfav_Att) clearly has a negative and significant impact on WTP.The latter is even sharper in Model 3. Therefore, the results confirm the relevance of attitudes as precursor of purchasing behaviour.
WTP estimates are then calculated using the coefficients of the model as follows (Hanemann et al., 1991): 7 In particular, the Wald test (W) finds significant differences between the specific constant for Languedoc-Roussillon and Aragón (W=4.201,p-value=0.040),and Catalonia (W=3.93,p-value=0.048).
8 Model 1, applied separately to the French and Spanish samples, reveals a positive and significant coefficient for the AF variable. 9500 random draws were taken from a normal distribution, with mean, the estimated coefficient vector, and variance, the estimated variance-covariance matrix.For each draw, WTP is calculated and then, the corresponding 0.05 and 0.95 percentiles are obtained. 10Results are available from the authors upon request.Mean WTP for certified animal friendly beef ranges between a minimum of 20.6% in Languedoc-Roussillon and Catalonia, and 22.10 and 22.5% in Midi-Pyrénées and Aragón, respectively, over average market price (regional in Spain and national in France).Thus, we found more similarities among neighbouring regions 10 besides an array of socio-demographic and consumption and purchasing habits.
We found that attitudes not only differ across countries but also across regions within the same country, and that this difference is mainly driven by a different attitude towards farmers' compliance with AW. Results are conclusive in highlighting that there is a general lack of trust in stakeholders' compliance, in particular, practices that take place beyond the farm-gate.In contrast with Nocella et al. (2010), however, there is not a clear division North-South with respect to attitudes.
WTP for animal welfare certification is not affected equally by attitudes towards farmers and other stakeholders, where the former have a positive and significant effect and the latter an insignificant effect.This result concurs with the pan-European study by Nocella et al. (2010Nocella et al. ( , 2012) ) but, in contrast with the aforementioned study, we found a significant influence of attitudes towards farmers also in the Spanish regions, and not only in France.Furthermore, WTP are slightly different across regions, ranging between an average of 20.6% and 22.5%, with a maximum gap of about 4% taking into account the empirical distribution, while not a clear pattern across countries emerge (i.e.maxima WTP are found in Aragon and Midi-Pyrenees, and the minimum in Languedoc-Roussillon).
Interestingly, despite the hypothetical nature of our methodological approach, our range of results concurs with those of previous studies based on auctions (e.g.Gracia et al. (2011) report 19-23% premia for dry-cured ham), or using a different payment vehicle like increase in the monthly expenditure in meat (Kehlbacher et al. (2012) report 26-34% for generic AW).It is also remarkable that our estimates for AW certification in beef move in the range of previous literature despite the different sector coverage (besides the above mentioned examples, Ipsos-LE (2013) estimates in France and Spain move in the range 24-30% for broilers and eggs).
Importantly, attitudes are found to trigger the biggest WTP gap between segments of consumers: the group with higher concern and who exhibit more trust in stakeholders (22% of respondents) is willing to pay up to 4% more than those consumers who exhibit lower trust and concerns (29% of respondents).Consumers in the most favourable attitude segment are mainly women, older than 35 years old, with up to secondary education level, use less big distribution chains, while feel more inclined to buy EU food Quality labels (i.e.PDO), and are mainly residents in Catalonia and Midi-Pyrénées.Therefore, the market of animal welfare differentiated products is segmented mainly by attitudes that share an international border, than within the same country.The 90% CI provides a range of variation of about 3% in every region.
Next, WTP for specific segments of consumers was calculated11 , by replacing mean values in Eq. [3] by the presence (1) or absence (0) of the characteristic.Interestingly, the segmenting variable that triggers a bigger WTP gap between groups of consumers is the attitude towards AW, with a difference in 4.4%.In other words, those consumers with higher concerns and that simultaneously trust more the stakeholders as guarantors of AW, are willing to pay around 24% over the market price, while this premium falls to 20% among those consumers with both, a lower interest and trust.When the AW attitude is measured using the multiplicative score (Model 1), results suggest that a maximum of 31% premium could be achieved from those consumers with a maximum level of trust in farmers.Consumers who use short-circuits to buy beef are willing to pay 3.6% more than those who never use this distribution channel.
From the biggest to the smallest gap across sociodemographic segments, we can sum up the results as follows: older than 54 years old consumers are willing to pay up to 3% more than younger consumers; consumers with university education are willing to pay 2.3% less than those with lower education (around 25%); women are willing to pay up to 2% more than men (21%), and the same difference is found between consumers with higher and lower income (22%).

Discussion
Voluntary labelling schemes of animal welfare enhancing practices additional to the standard regulation are already in place in the EU using public, private or self-declaration attestations, although their scope and application is still limited, both geographically and by animal species (Ipsos-LE, 2013).In this paper, we have tried to shed light on WTP for such labels, in particular one guaranteed by a public authority, with four main differences with respect to previous literature: we focus on one of the least studied species, beef meat; we carry out a cross-country and cross-regional study; we measure attitudes towards AW using a tested scale that combines the degree of concern and trust in stakeholders towards and array of AW enhancing practices to implement on the farm and at the transportation stages; and we test for the influence of these attitudes on WTP 11 Beef animal welfare, attitudes and Willingness to Pay higher levels of quality (Sanjuán et al., 2012).The personal contact between the consumer and the producer reduces uncertainty regarding food safety and other quality attributes (Sans et al., 2008).Our results also suggest that short circuits contribute to build up a favourable attitude towards AW, enhance trust on farmers and other stakeholders, and reinforce the commitment with AW by willing to pay a premium for animal welfare enhancing practices and its labelling.Second, experience with beef preparation stages affects negatively, although not significantly WTP.The negative sign of experience may be reflecting a conflict between AW and sensory experience (i.e. consumers mainly driven by taste place animal welfare in a second order (Vanhonacker et al., 2007;de Barcellos et al., 2011;Gracia & Zeballos, 2011;Lusk, 2011), while higher welfare meat is more strongly associated with higher quality and better health than with a better taste ( Vanhonacker & Verbeke, 2014).
Our study has several limitations.First, the hypothetical nature of the contingent valuation method does not involve money, and accordingly, respondents may overstate their WTP (Napolitano et al., 2010a).The closeness of our results to those obtained with experimental auctions compatible with economic incentives (Gracia et al., 2011), however, may be indicating that, if such an upward bias exist is probably more related to other sources, like 'warm glow' and 'part hole' or other specific details in the experimental design rather than the elicitation method of WTP, as concluded by Lagerkvist & Hess (2011) in their meta-analysis.Anyway, any eventual over-statement is less relevant when the focus lies on identifying the relevant explanatory factors of WTP as well as differences across consumers' segments, which is one of the main purposes of our paper.
Second, stated preferences evaluated with CV may only reveal imperfectly purchasing decisions in a real choice situation where animal welfare issues enter as an additional attribute in a complex and multi-attribute bundle.However, our empirical approach was unidimensional and consumers where not asked to rank AW among other key-decision attributes.Moreover, in reallife, the shopper is often placed in a time-pressed and information-overloaded situation and his/her decision may be different from that stated under more relaxed conditions (de Jonge & van Trijp, 2013).
Finally, we have tried to measure consumers' attitudes towards AW and their WTP for AW certification without actually knowing consumers' prior knowledge towards AW measured by the interaction of concern and trust, but indirectly, by some relevant socio-demographic variables and purchasing habits.De Backer & Hudders (2015) recently show that AW attitudes of Belgium consumers can even partially predict the diet choices of meat eaters, flexitarians and vegetarians.
Interestingly, consumers already familiar with other EU certification schemes are also more receptive to animal welfare certification.Likewise, WTP could go up to 31% among those consumers who exhibit the highest concern and trust in farmers' compliance.Thus, trust building would help to foster the mean WTP necessary to compensate for higher costs of implementation, while at the moment, all the weight seems to fall on the farmer, and better communication efforts should be channelled to improve trust on other stakeholders' performance (i.e.transporters).
Regarding WTP estimation and the influence of socio-demographics, our results concur with most of the literature, finding a positive relationship with the level of income (Bennett & Blaney, 2003;Lagerkvist & Hess, 2011) and being female (Nocella et al., 2010;Gracia et al., 2011;Toma et al., 2011Toma et al., , 2012)).With respect to education and age, however, we find a negative and positive effect, respectively, that contrast with previous research (Nocella et al., 2010;Gracia et al., 2011;Lagerkvisst & Hess, 2011;Toma et al., 2011Toma et al., , 2012)).That is, older people than 54 years old and with no university degree are willing to pay more for the certification of AW in beef.A different treatment of age (i.e.cutting point to select the age segments, continuous versus intervals) as well as a certain upward education bias in our sample, could explain this outcome12 .Likewise, although a significant gap in WTP was found across socio-demographic groups, this is always lower than the one found by contrasting attitudes.As some authors argue (Kehlbacher et al., 2012), AW has reached a certain overall level of consensus in society, and accordingly socio-demographics have lost influence in favour of attitudes and personality traits.In this sense, personal values, beliefs and life-styles are often claimed as better descriptors of sustainable and/or ethical food consumers (Verain et al., 2012;Vanhonacker & Verbeke, 2014).
This paper also provides some other interesting results rarely treated previously in the literature.First, the place of purchase of beef does not exert any direct influence on WTP, with the exception of purchasing directly from the producer that affects positively WTP.Short circuits are associated with the provision of on the subject.Even though respondents were provided with identical information at the moment of the survey prior knowledge may influence responses.Besides, the question was very specific with respect to the public authorities' endorsement of the certification.Consumers do not usually know much about how farm animals are raised, or the changes (and costs involved) in the production process necessary to accommodate their preferences (Olynk & Ortega, 2013) or new legislation (Blokhuis et al., 2008).Moreover, trust is usually higher when the certification is managed by public authorities than by stakeholders (retailers for example) (Heerwagen et al., 2015) and accordingly, this may have pushed up the premium for labelling.
In summary, this paper investigates consumers´ WTP for a label that guarantees compliance with superior animal welfare standards in beef, focusing on trust in stakeholders (farmers and transporters) as a way to transmit reliability of the information conveyed in the label.Our study proposes a cross-regional approach that allows capturing within-country heterogeneity, which otherwise is masked, either by the use of samples representative at the country level (normally in crosscountry studies) or the assumption that the results obtained with a local/regional sample can be perfectly extrapolated to the country level.In this sense, this study provides evidence about the difficulties of establishing a North-South dichotomy with respect to AW attitudes, trust, and WTP, as regions in different countries (i.e.Aragón and Midi-Pyrénées) seem to be closer than regions within the same country (i.e.Midi-Pyrénées and Languedoc-Roussillon).
Results show clearly a need to enhance trust on stakeholders' compliance with AW practices, especially beyond the farm gate, as mistrust is counterweighting a positive influence of a superior degree of awareness and concern.This would lead to a higher level of commitment and actual purchase of AW guaranteed label by a larger proportion of the population.Currently, only around one fifth of the respondents in the selected regions show a distinctive major concern and trust in stakeholders, which translates into a significant WTP differential.
(A F ) (STRESS, NATBEH, FREE, INSP, DIET, MUTI and CAST) and those carried out by other stakeholders (A O ) (TRAN, PERS, VEIC and SLAU).The Cronbach's alpha indicates a satisfactory level of reliability: 0.82, 0.73 and 0.75, for A, A F and A O , respectively, which are in line with the figures reported by Nocella et al. (2010) (0.71 and 0.75 for A F and A O , respectively).

Table 1 .
Items used in the scale to measure attitudes towards Animal Welfare Beef animal welfare, attitudes and Willingness to Pay women (53.6%), ranging between 51.7% in Aragón and 59.4% in Languedoc-Roussillon (Table

Table 2 .
Description of explanatory variables in WTP estimation

Table 3 .
Attitudes towards animal welfare by region: total (A), farmers (A F ) and other stakeholders (A O ).Standard deviation in parentheses

Table 4 .
Profile of segments based on the degree of importance and trust attached to animal welfare ***,**,*: significant differences between the three segments at 1, 5 and 10% significance level, respectively, based on a Chi-square statistic.

Table 5 .
Estimation results of the double-bounded model

Table 6 .
Willingness to Pay (WTP) for animal welfare (AW) guaranty across regions a

Mean estimate of WTP across consumers segments
Based on coefficient estimates in Model 1. Model 3 provides very similar results that are available from the authors upon request.WTP among consumers with 'More favourable attitude towards AW', uses coefficients in Model 2. b Evaluated at the maximum value of the scale (63). a