Dry-cured ham is one of the most characteristic products of the Spanish gastronomy. Within the dry-cured ham, the Iberian ham (i.e., that coming from the Sus scrofa mediterraneus pig breed) is probably the most widely-known, particularly due to its higher organoleptic quality and recognized identity (i.e., traditionally produced in extensive production systems using transgenerational knowledge). The feeding and management system determines the category of the Iberian ham, basically distinguishing between two main categories: acorn-fed (jamón ibérico de bellota) and fodder-fed (jamón ibérico de cebo). The former involves animals extensively managed, slaughtered after foraging for acorns, grass, herbs, wild legumes, and other natural resources from the Dehesa (or Montado in Portuguese) agricultural ecosystem without external feed input in the final months of the animals’ life. The latter involves animals intensively managed, entirely fed with commercial feed¹. As shown in the dry-cured ham quality standard (BOE, 2014), sub-categories of Iberian ham result from combining these types of management and feeding systems with different levels of breed purity.

The acorn-fed Iberian ham is the most valued by consumers (Díaz-Caro et al., 2019; Granado-Díaz et al., 2021). This is due to its superior organoleptic properties (especially taste, flavor, texture, and color) compared to the fodder-fed Iberian ham. This higher quality coupled with lower yields compared to the latter give rise to significantly higher price levels of the acorn-fed Iberian ham. In particular, while fodder-fed Iberian ham can often be purchased in the market at €5-10/100-g-package, acorn-fed Iberian ham prices can easily reach over €20/100-g-package, placing the latter within the premium segment (Mesías et al., 2013).

The acorn-fed Iberian ham is closely connected to the Dehesa agricultural ecosystem, a Mediterranean oak savanna, agro-sylvo-pastoral system unique for its high levels of provision of ES (including provisioning, regulating, supporting, and cultural services) (Garrido et al., 2017; Campos et al., 2019). It is the result of a historical interaction between ecological and social systems, encouraging multiple land-uses (especially cropping, livestock −including swine, ovine and bovine cattle− and timber and non-timber products) within the Dehesa system. This agricultural system covers 3.1 million hectares in the southwest of Spain and Portugal (Moreno & Pulido, 2009), with approximately one-third located in the region of Andalusia.

There is a great social demand for the ES provided by the Dehesa system (Ovando et al., 2016), with Andalusian people showing both use and non-use values. Notable examples of the former are recreation and aesthetics, hunting, and mushroom and berry picking, while among the latter, biodiversity-related services and cultural values linked to cultural heritage, traditional knowledge, and animal welfare are worth noting (Garrido et al., 2017; Granado-Díaz et al., 2021). However, it is unclear to what extent consumers of acorn-fed Iberian ham consider these values in their purchasing decisions and whether this may be subject to different spatial discounting patterns. Among the studies to date that have analyzed consumers’ preferences toward this type of ham, Mesías et al. (2009), Sahelices et al. (2017), Díaz-Caro et al. (2019), and García-Gudiño et al. (2021) provide evidence of preferences for geographical origin of production. The first three studies offer information about consumer preferences toward protected designations of origin (PDO) −arguably related not only to a geographical location but also to quality cues− and the latter focuses on traditional producer regions. However, there has been no explicit focus on the agroforestry production system (Dehesa) in the specialized literature.

The DCE method (Hensher et al., 2015) was used to build a hypothetical market to analyze the impact of Iberian dry-cured ham attributes on consumers’ preferences. This method relies on the Lancaster Consumer Theory (Lancaster, 1966) and the Random Utility Theory (McFadden, 1974) and has been extensively used in the analysis of consumer preferences regarding food products, including meat products (e.g., Gracia & de-Magistris, 2013; Lusk & Tonsor, 2016).

Table 1 shows the attributes and levels included in the DCE. It includes four non-monetary attributes (Breed purity, Slicing type, Dehesa origin claim, and Pictorial representation of Dehesa) and the monetary attribute Price. The selection of the former is based on previous studies (Mesías et al., 2009; Sahelices et al., 2017; Díaz-Caro et al., 2019) and direct observation of retailer meat shelves. Considering the differences in price and quality encountered for acorn-fed and fodder-fed Iberian hams, a labeled DCE design was employed, using the same attributes and levels except for the Price attribute. This design was used under the assumption that the Iberian ham type (the “label”) already provides the consumer with relevant additional information beyond that included in the attributes and levels (Hensher et al., 2015). Hence, unlike for unlabeled designs, effects specifically related to the Iberian type can be estimated for each attribute. We describe them in more detail below.

The Price attribute was set taking into account market prices of Iberian hams, distinguishing between acorn-fed and fodder-fed. Four levels were set for the Price attribute. They range from €13 to €22/100-g for the acorn-fed ham alternative and from €4 to €13/100-g for the fodder-fed ham alternative, with price differences of €3 between consecutive levels in both cases. Fig. S1 [suppl] presents an example choice card, where it can be observed that prices were displayed at the bottom of the card, which resembles a supermarket shelf, with the price given in euros per 100-g package (in large font size) and in euros per kilogram (in small font size).

The non-monetary attributes included two directly related to the product quality (Breed purity and Slicing type) and two related to the agricultural production system (Dehesa origin claim and Pictorial representation of Dehesa), with the latter being key for the spatial analysis carried out here. The Breed purity attribute included three levels, reflecting the dry-cured ham standard regulation (BOE, 2014): 50% (minimum legal level to be labeled Iberian), 75%, and 100% Iberian breed. The Slicing type attribute comprised two levels: a mention of the “Hand-sliced” claim in the packaging design and no mention of it (i.e., implicitly indicating that the ham was “Machine-sliced”). These attributes were selected because they were identified as relevant determinants of the product price: Iberian hams from higher breed purity and/or sliced by hand are more expensive.

To test the origin feature related to the Dehesa system, the two attributes included in the DCE were: the Dehesa origin claim, with two levels, one where the claim “Produced in the Dehesa” was included on the packaging and the other where it was not; and the Pictorial representation of Dehesa, where a pictorial representation of the agricultural system was incorporated at the bottom of the package. The latter attribute included four levels, related to four different images: 1) No background (no picture), 2) Background-Agricultural system (picture of a common agricultural system −cereal cultivation), 3) Background-Dehesa system (picture of the Dehesa system), and 4) Background-Dehesa system with pigs (picture of the Dehesa system with Iberian pigs eating acorns).²

As shown in Fig. S1 [suppl], the choice cards reproduced the front³ of two 100-g packages (alternatives) of acorn-fed and fodder-fed Iberian ham. This product format was used since this type of package is highly standardized, which also facilitated the graphical design of the alternatives (reflecting the most typical layout encountered on the market). In addition to the two ham alternatives, the no-purchase option was added to make the choices more realistic.

The research on spatial heterogeneity of individuals’ preferences has long evidenced distance-decay effects, especially using case studies focusing on spatially concentrated sites (i.e., sites located within a spatially-contiguous site) (Hanley et al., 2003; Bateman et al., 2006). To show these effects, the use of the distance to the nearest point (i.e., Euclidian distance) between the site under study and the household has been widely used (Schaafsma, 2015).

However, when dealing with scattered environmental sites (over multiple non-contiguous locations), as is the case of most agricultural systems (like Dehesa here), quantity-within-distance approaches are more suitable compared to the traditional distance-to-the-nearest-point approaches (see Granado-Díaz et al., 2020, for a comprehensive theoretical framework and empirical comparison). In the quantity-within-distance approaches, not only the distance but the area of the site within a certain distance from a respondent’s place of residence are assumed to determine respondent’s preferences (Glenk et al., 2020). Previous studies show the suitability of using quantity-within-distance approaches for the case of scattered sites such as natural areas (De Valck et al., 2017), forest areas (Yao et al., 2014), river bodies in a certain watershed (Holland & Johnston, 2017), and agricultural systems (Granado-Díaz et al., 2020). To the authors’ knowledge, the current paper is the first to provide empirical evidence of the use of both, quantity-within-distance and only distance-based approaches, to analyze spatial heterogeneity of preferences for food product attributes. For it, one distance-based spatial index and three area-based spatial indexes have been used. We turn to describe the spatial indexes used.

Following the theoretical framework of Granado-Díaz et al. (2020), the spatial discounting effect for spatially-dispersed sites can be explained by the following expression:

where u_kh represents the utility provided by the site to the household h; p(x,y) is a dummy spatial variable taking the value 1 when the point with coordinates (x,y) is included within the limits of the site under analysis and 0 if located elsewhere; δ(x_i,y_i) is the distance function determined by the difference in the Cartesian coordinates of the site i related to household h; and X⁺, X^-, Y⁺, and Y^- are latitudes and longitude defining the region to be analyzed.

This approach is adapted to the case study by using a 1-km grid for the whole Andalusian region (N= 88,792 grid units). This grid was intersected with the Andalusian Dehesa cartography to obtain the area of Dehesa contained in each 1-km grid square and the percentage of its area represented by this system⁴. Moreover, we calculated the distance from every household location represented in the sample to the centroid of every 1-km grid square with the Dehesa area as the Euclidean distance.

Equation (1) was then approximated in two ways: first, by substituting p(x,y) with the percentage of Dehesa in each grid square, to account for the relative presence of this system throughout the space, and second, by substituting the double integral with the sum of the percentage of Dehesa areas in each grid square, discounted by different functions calculated from the distance between each grid square to each sample location (see below). In Fig. 1 this procedure is depicted graphically.

Using this approach, three different spatial indexes were calculated for each consumer sampled considering his/her place of residence: Area_Inv, in which the area is discounted by the inverse of the distance, Area_In2, in which the area is discounted by the square of the distance, and Area_InL, in which the area is discounted by the natural logarithm of the distance. The expressions used for each index are the following:

The indexes are multiplied by a factor of 10, 10,000, and 1/1,000 to scale them up to approximately one unit on average.

To test whether this approach outperforms indexes based only on distance, a simple inverse-of-the-distance index was also included in the analysis, calculated from the distance of each location to the nearest Dehesa plot.

The investigation is based on a face-to-face survey administered in Andalusian food retail establishments where packages of acorn-fed and fodder-fed Iberian ham were sold. Interviewees responded to the questionnaire voluntarily and there was no reward for it. The questionnaire was administered to a total of 1,158 respondents aged 18 years old or over who had bought Iberian ham (acorn- or fodder-fed) at least once in the last year.

A quota sampling procedure was used in proportion to residence in acorn-fed Iberian ham producing and non-producing provinces. As the Iberian ham consumer population is unknown, we could only consider the whole adult population in Andalusia. The sampling design ensured that a relevant number of rural and urban municipalities were included in the survey. Interviews were made following random routes across shops and supermarkets selling Iberian ham at different times of the day. The socio-demographic characteristics of the sample, including its spatial distribution (using the Area_Inv index), are shown in Table S1 [suppl].

The questionnaire included four blocks, comprising questions relative to: i) the interviewees’ habits relating to the consumption of Iberian ham; ii) the DCE; iii) the interviewees’ knowledge about the Dehesa system and opinions on the environmental and social values associated with it; and iv) the interviewees’ socio-demographic characteristics. In block three, the interviewees were asked to assess their perception about the main ES provided by the Dehesa using a Likert scale, the descriptive parameters of which are shown in Table S2 [suppl].

For the experimental design, valuation results from other surveys were employed to foresee the order of magnitude of the parameters of the consumer utility function. Based on this and following a conservative approach, positive close-to-zero priors for the non-monetary attribute parameters were set, whereas that for the monetary attribute was anticipated to have a negative sign with an order of magnitude two and three times higher for fodder- and acorn-fed alternatives, respectively, compared to the other attributes. An efficient design minimizing the D-error employing Bayesian techniques was used (Rose et al., 2011). The final design included 24 choice sets, containing 4 blocks of 6 sets and presenting a D-error of 0.436. For this purpose, NGene 1.2.1 was used. A pre-test of 40 interviews was implemented to validate the appropriateness of the questionnaire and the experimental design.

We used the mixed logit model (MXL) in preference-space⁵, which improves the standard conditional logit models by accommodating panel data and allowing for flexible substitution patterns to better model preference heterogeneity across individuals (Hensher et al., 2015). MXL has been extensively used to analyze individuals’ preferences toward food consumption (Kallas et al., 2019), as well as environmental services (Yao et al., 2014). Model formulation accounts for the utility that individual i derives from choosing alternative j on purchase occasion (or in choice scenario) s (U_isj) as follows:

where χ_isj is the vector of the attributes and levels, β_i is the vector of coefficients, C_ij is a specific component of the alternative and ε is the random error term (assumed to follow a Gumbel distribution). The joint probability of individual i choosing alternative j in each of the choices s is:

where y_i is the sequence of choices of individual i. This integral requires an iterative process as it has no closed-form solution (Train, 2009). Estimation results described below were produced using 1000 Halton draws to simulate the likelihood functions to be maximized (Train, 2009). All attribute parameters were assumed to be normally distributed.

The modeling procedure was carried out as follows. First, an MXL model considering only the DCE attributes (i.e., Base Model without any interaction) was run. Second, MXL models exploring heterogeneity in the mean were run, including interactions of all the non-monetary attribute parameters related to the agroecosystem (the “Produced in the Dehesa” claim, the three levels of the Pictorial representation of Dehesa attribute, and the alternative specific constants, ASCs) with the different spatial indexes considered. To test for differences in goodness-of-fit of models with different spatial indexes, the Vuong’s (1989) test for non-nested specifications was used. The spatial index with the best performance, as indicated by this test, was then selected for the subsequent analysis, and for the sake of parsimony only statistically significant interactions were kept in the model. This was done following a sequential procedure, each time eliminating the interaction with the lowest level of significance. The resulting model was termed the Spatial Model. With this model, spatial discounting effects can be evidenced, though we cannot separate to what extent they relate to the consumers’ preferences toward conserving the bundle of the ES provided in the agroecosystem of origin. Thus, to test for this hypothesis and check whether spatial discounting is determined by the intensity of these preferences, consumers’ perceptions about the provision of ES in the agroecosystem of origin were incorporated into the Spatial Model.⁶ This was done by using the variable PerceivedES, which incorporates the consumers’ average perception about the ES provided by the Dehesa system (those related to biodiversity, forest fires prevention, animal welfare, soil conservation, landscape aesthetics, cultural heritage, as well as provisioning services linked to rural vitality)⁷. The descriptive statistics of the variable PerceivedES are shown in Table S2 [suppl], along with those of the variables related to the individual ES comprising the former variable. The resulting model was termed the ES-Spatial Model, and included the interactions between attributes and the spatial index or the variable PerceivedES, and double interactions between attributes and the latter two variables. In any case, the final model reported here only includes the significant interactions. This parsimonious model was built following the same sequential procedure proposed for the Spatial Model.

Finally, to compare the goodness-of-fit of the three models built (Base Model, Spatial Model, and ES-Spatial Model), likelihood-ratio tests were used.

The results obtained in this paper point to the existence of spatial discounting in the purchase of food products linked to the agroecosystem in which they are produced. In previous literature, limited evidence of such effects was found, by showing the influence of political boundaries (Lim & Hu, 2016; Farris et al., 2019), local labels (Ay et al., 2017), or, as best, discrete distance intervals between production and consumption (de-Magistris & Gracia, 2016) on consumers’ WTP for food products. Our study goes beyond these studies by including direct estimations of spatial discounting associated with the combination of distance to and area of scattered producing systems in consumers’ WTP. Said effect has previously been demonstrated in environmental valuation (see Schaafsma, 2015, for a literature review), although there is a paucity of studies providing evidence for scattered ecosystems (Czajkowski et al., 2017; Holland & Johnston, 2017). Of those that do, only one is focused on agroecosystems (Granado-Díaz et al., 2020). Therefore, to the authors’ knowledge, this is the first study to prove this effect in a valuation context focusing on marketed food products. In this regard, it is worth noting that our results mirror those of Granado-Díaz et al. (2020) in that area-based indexes are found to outperform simple distance-based indexes, better accounting for spatial discounting in scattered agroecosystems.

In particular, spatial discounting is evidenced by significant interaction terms of the spatial index with attributes related to the agroecosystem of origin, specifically written prompts referring to the Dehesa (i.e., the “Produced in the Dehesa” claim). This claim is associated with significant effects of spatial discounting for both acorn- and fodder-fed Iberian ham. Taking into account that the latter product was only loosely related to the Dehesa, we interpret this finding as reflecting consumers’ cultural identity linked to the agroecosystem and their willingness to support the local economy and communities, in keeping with previous literature on local food products (see Feldmann & Hamm, 2015, for a review). Consequently, consumers living surrounded by Dehesa areas would be willing to pay more for food products that claim to be produced there.

The results also suggest that spatial discounting is affected by consumers’ perception of the ES provided by the producer agroecosystem. We found that consumers who actually perceive that the Dehesa provides such ES are willing to pay more for food products directly coming from it (in our case acorn-fed Iberian ham from the Dehesa) independently of their location with respect to the agroecosystem; as such, they do not exhibit spatial discounting. Previous literature regarding environmental valuation suggests that the lack of spatial discounting could be related to the existence of non-use values attached to these environmental sites (e.g., Bateman et al., 2006; Granado-Díaz et al., 2020). Our results suggest that a similar pattern may arise with food products linked to agroecosystems with high levels of ES provision, as is the case of the Dehesa system. Therefore, consumers who perceive these ES might be taking into account non-use values—which are enjoyed regardless of their place of residence with respect to the Dehesa—incorporating them into the product value. On the contrary, a different spatial discounting pattern was found for consumers who did not perceive these ES, as our results indicated that their preferences revealed the effects of spatial discounting. As little or no relation with the provision of ES could be identified for these consumers, the existence of this effect could simply be related to “sense of place” or “spatial identity”, as suggested by Hanley et al. (2003). In addition, consumers perceiving the ES of the Dehesa presented a negative spatial discount for fodder-fed ham, which could be linked to greater knowledge of the different Iberian ham products and their actual relationship with this agroecosystem. This result shows a more accurate spatial discounting effect as compared to that in Villanueva et al. (2021), who, using the same survey data, found a significant negative relationship between WTP for using pictures of Dehesa as background of fodder-fed Iberian ham packages and living in a Dehesa municipality and/or province where Dehesa landscape is prominent. In any case, we consider that further research is needed to shed light on how consumers’ perception of the ES provided by the producing system shapes spatial discounting, especially by identifying use and non-use values attached to these services. Moreover, the role of the consumer’s identity on this effect should clearly be acknowledged.

All the above-mentioned findings support most of the assumptions of the theoretical model developed by Yamaguchi & Shah (2020) to explain the spatial discounting of ES. In particular, our results evidence that spatial discount rate for ES is heterogeneous among individuals, showing it depends on key parameters identified in such a theoretical model, such as pure rate of spatial preference (i.e., the Area_Inv index), consumption change (i.e., quality cues variables: Iberian breed purity, Slicing type, Dehesa origin claim, and Pictorial representation of Dehesa), and perceived ES (i.e., PerceivedES variable).

Relevant insights can be gained from the results, especially concerning marketing strategies. For instance, the results suggest that information campaigns aimed at increasing consumers’ awareness of the ES provided by the agroecosystem in question would result in higher WTP for the food products coming from it (Pirard & Lapeyre, 2014). In turn, this would help to incentivize farmers’ ES provision as consumers would become more sensitive to differentiation strategies relying on commodification (as highlighted by Salazar-Ordóñez et al., 2021). For the specific case of Iberian ham coming from the Dehesa system, this would imply that marketing and information campaigns should focus on informing consumers of the different services provided by the Dehesa system, its intimate relationship with acorn-fed Iberian ham production and, therefore, the role of this production in sustaining a high natural value system. In addition, the positive results encountered for written prompts referring to the Dehesa system hint at the potential benefits of developing a specific label related to this agroecosystem. In this sense, the Iberian ham sector (especially the acorn-fed producers) would greatly benefit from combining such informational cues with forthcoming quality certifications informing of environmental performance (e.g., eco-score), once these certifications are fully implemented. Given the importance of the sociocultural services provided by this agroecosystem, as for many other traditional food production systems, the inclusion of the sociocultural performance into the calculation of such a score would be recommended.