INTRODUCTION
⌅In the last decade consumers’ interest in locally grown food has increased (Bir et al., 2019). Encouraged by this increase, the number of local farms supplying local fresh produce has also expanded. One reason for this increase in supply comes from proposed policies that promote local food direct purchase/sales to consumers (Soares et al., 2017). Evidence from previous research show that consumers are also responding through increased demand and are willing to pay (WTP) more for local fresh produce (Enthoven & Van den Broeck, 2021; He et al., 2021). Some reasons for this increase in demand for local food relates to the perceived benefits of high-quality food, nutritional value, and low environmental impact (Fernqvist & Ekelund, 2014; Enthoven & Van den Broeck, 2021; He et al., 2021). However, the term “local” is determined by the “credence” nature of attributes, meaning that the quality of the product cannot be verified either ex-ante or ex-post by consumers.
As pointed out by Darby & Karni (1973) sellers do not only provide food products on the market but presume they are fully aware of consumer’s needs. The problem faced by consumers is the insufficient level of information available to judge whether the quality cues (e.g., local) provided by sellers can be trusted. For this reason, quality labels coming from reputable certification agents (e.g., government agencies) aim at lowering information asymmetry and transforming credence into search attributes (Darby & Karni, 1973). When the information is trusted by consumers based on their subjective knowledge, then credence attributes are considered as important quality cues that affect purchase decisions (Caswell & Mojduszka, 1996; Grunert & Wills, 2007). However, subjective knowledge is heterogeneous among consumers. For example, some studies found that consumers believe organic food products are healthier than non-organic food products (Gundala & Singh, 2021; Yeo et al., 2022). Likewise, consumers perceive local products to be fresher, tastier, and more nutritious than non-local food (Campbell et al., 2013; Meyerding et al., 2019). If consumers know that the food is produced locally and based on their subjective knowledge think local foods are fresher, then adding a “locally produced” label may not affect the decision to purchase the food, since the judgment of freshness can be done by visual observation (at least for unpacked and non-peelable goods that can be visually observed). Yet, if consumers perceive local foods as tastier, then adding the “locally produced” label will likely affect their purchase behavior since “taste” cannot be evaluated by a visual observation. Such notions indicate that credence quality cues directly and indirectly impact consumers’ choices and influence the perceptions of the product’s quality and the attributes that it possesses.
Research on consumer preference for local food is ample. Most scholars have mainly explored local food systems in the context of rural areas and their development (Mundler & Laughrea, 2016), the distribution of local food (Hedberg & Zimmerer, 2020), and consumer behavior (Birch & Memery, 2020), among others. Within consumer behavior, most studies explore the influence of “local” as a credence quality cue on consumer food choices (Bazzani et al., 2017), and/or their interactions (Meas et al., 2014). However, considering just the credence aspects of the product has limitations, especially when it is demonstrated that search and experience attributes influence consumer preferences and purchase decisions (Hughner et al., 2007). Search attributes help consumers determine the quality of a product before purchase by exploring product characteristics (Caswell & Mojduszka, 1996). Experience attributes remind consumers of the positive/negative characteristic from a previous purchase (e.g., the taste). When combined, search (e.g., appearance) and experience attributes (e.g., taste) form two important decision-making sensory attributes that affect the purchase and re-purchase of the food product (Caswell & Mojduszka, 1996; Ballco et al., 2020; Ballco & Gracia, 2020).
Within sensory attributes, it is demonstrated that consumers spend few seconds searching for information when purchasing food for different reasons (e.g., time) (Ballco et al., 2019, 2020). This makes visual appearance a key sensory attribute that attracts consumers’ attention (Orquin et al., 2018; Perkovic et al., 2022). Besides search attributes, previous studies identify experience attributes as crucial predictors of repeated food purchases (Caswell & Mojduszka, 1996; Schleenbecker & Hamm, 2013). Given the importance of search and experience attributes in consumer behavior, several consumer research studies combine sensory analysis with economic methods in a common research design. Some examples include studies that find positive impacts and higher WTP values for origin labels and sensory evaluations (Feldmann & Hamm, 2015; Bekele et al., 2017; Ballco et al., 2020; Ballco & Gracia, 2020; Milford et al., 2021). The overall results of these studies show that consumer preferences are influenced by an interchange of multiple attributes. Of particular importance is the taste of the product. However, consumers cannot taste the food before purchase to acquire information on the experience attributes in most grocery stores. Therefore, it is crucial to investigate the impact of various food product characteristics at different purchase stages where information on multiple types of attributes is presented.
Given the aforementioned, the main objective of our research is to explore consumers’ acceptance and the sensory evaluation of a local fresh potato variety and measure the WTP for the most important attributes. The selection of the product is driven by the interest of local producers in launching the potato variety in the local market of Aragón under a brand name with certain culinary information. Aragón is an important producing region of high-quality potatoes in Spain. The area dedicated to the production of potato in 2021 was approximately 600 ha, with an average yield of 40 tons/ha. The organoleptic quality of the production was high. Yet, the washing quality of the potato was low1. The primary destination of the local potato is mainly the processing industry (i.e., fried, 80%) and consumption (i.e., unprocessed, 20%) (Almunia, 2021).
To reach the main objective we collected data from two experiments performed with consumers in the fifth largest town of Spain (Zaragoza). We selected this city because it is the capital of the region where the potato producers are interested in launching the local potato. In addition, the city is centrally located, no more than 100 kilometers away from any location in the region. This is consistent with the official European Union (EU) definition of local foods, namely “...a food system in which foods are produced, processed, and retailed within a defined geographical area” (depending on the source, within a 20 to 100-km radius approximately).
This study contributes to the literature by determining fresh produce acceptance in the shopping and consumption stages in several ways. Specifically, we combined sensory analysis with experimental auctions to examine consumer preference for a traditional local fresh potato variety. The methodology explored consumer behavior on product information at different decision-making stages (i.e., purchase and consumption) and examined the trade-off between credence (i.e., local), search (i.e., washed vs unwashed, origin), and experience (i.e., appearance, odor, texture, taste) attributes. Various researchers have studied the trade-off for fresh produce attributes, such as local and organic quality labels, brands, and sensory characteristics (e.g., appearance and/or taste), among others, Mascarello et al. (2015); Ceschi et al. (2018), Liu et al. (2019). However, these studies did not include sensory analysis (e.g., taste) in their research. To the best of our knowledge, only the study of He et al. (2021) combined a similar approach as our research. We build on the fundamentals of He et al. (2021); however, instead of using a contingent valuation method we used an experimental auction. Additionally, besides conducting the research based on a literature review and market analysis, the selection of the attributes in our case was developed in close collaboration with local producers who want to launch a differentiated potato in the local market. The local potato variety is ‘Agria’.
The results of this study are expected to help policymakers and local fresh produce growers to tailor their labeling and promotion strategies around consumers’ responses and decisions. Policymakers and marketers may consider the premiums that consumers are WTP for the most important attributes for a more efficient fund allocation and to guide decisions on promotion investments. Local fresh produce growers can use the information of this study to select the most beneficial marketing and pricing strategies. Researchers and scholars in the agri-food sector can benefit from the methodological approaches of this research and replicate them in new/differentiated food products that aim to be introduced in the market.
1 Washing quality refers to the degree to which potatoes are cleaned of dirt and debris during the washing process, which can affect their overall quality and taste. A high-quality wash will remove a significant amount of dirt, debris, and other contaminants from the surface of the potatoes, leaving them looking clean, bright, and uniform. A poor-quality wash may leave behind dirt and debris, resulting in potatoes that appear dirty, uneven, and less appealing. Other factors that can impact the washing quality of potatoes include the water temperature, the duration of the washing process, and the equipment used for washing (Smith, 2019).
MATERIAL AND METHODS
⌅To fulfill the aim of the study two subsequent artifactual experiments with consumers (Harrison & List, 2004) were applied using a within-sample subject design. Fig. 1 summarizes the two experiments and the different tasks.
The experiments were conducted at different halls owned by consumer associations and the local government and were reachable in central locations at the capital’s city center. Experiment 1 mimics a real shopping environment (i.e., to evaluate the product based on the information and then decide to buy/not buy the product). In this experiment, the participants first evaluated the product of reference based on the extrinsic information on the package and compared it with other fresh potato packages available in the local market. Second, the participants were asked to reveal their acceptability of the product through visual appearance rating, purchase intent (PI) and WTP. The objective of experiment 1 was to study the acceptance of the local potato variety in the shopping stage. Experiment 2 corresponded to a sensory ranking task (e.g., taste, smell, texture) conducted at home once the local potato was fried. The aim of this experiment was to analyze consumers’ acceptance at the consumption stage.
Participants
⌅The experiments were conducted in the largest town of the producing region of the potato variety in spring 2018. The experiments were implemented in accordance with the Declaration of Helsinki approved by the independent ethics committee of the Agrifood Center of Research and Technology of Aragón (CITA). Recruited participants were representative of food shoppers and had experience with the reference product. The sample was established with people older than 18 years who were responsible for food purchases and cooking in the household (or share the responsibility). The participants were recruited via consumer associations located in different central neighborhoods and public institution halls in the city where the experiment took place (universities, town hall learning centers, community activity centers, etc.). A total of 141 participants were recruited in experiment 1 (shopping stage). Out of 141 consumers that participated in the first experiment, only 104 provided responses to experiment 2 (consumption stage). Table 1 presents the descriptive statistics of the consumers’ characteristics from the two experiments and compares the sample with the general population of the region. In addition, statistical analyses were conducted to measure for consumer characteristic differences between samples.
| Characteristics | Experiment 1 (n=141) | Experiment 2 (n=104) | Population 1,2 |
|---|---|---|---|
| Gender1 | |||
| Male | 24.3a | 22.1b | 49.1 |
| Female | 75.7a | 77.9b | 50.9 |
| Age (average, standard dev)1 | 54.5a (14.4) | 54.7b (13.2) | N/A |
| 18-34 | 6.8a | 4.8b | 22.8 |
| 35-44 | 16.7a | 18.5b | 20.2 |
| 45-54 | 28.0a | 28.2b | 19.0 |
| ≥ 55 | 48.5a | 48.5b | 38.0 |
| Level of education2 | |||
| Primary | 21.3a | 19.2b | 23.0 |
| Secondary | 25.7a | 26.9b | 48.7 |
| Higher | 53.0a | 53.8b | 28.3 |
| Household income range | |||
| ≤ 1500 €/month | 24.3a | 25.9b | N/A |
| 1501-2500 €/month | 23.5a | 22.1b | N/A |
| 2501-3500 €/month | 19.8a | 22.1b | N/A |
| > 3500 €/month | 10.3a | 9.6b | N/A |
| Do not know/refuse to answer | 22.1a | 20.2b | N/A |
| Household size (average, standard dev) | 2.8 (1.0) | 2.7 (1.0) | N/A |
| Vegetarian | 3.0a | 1.0b | N/A |
| Years living in the region (average) | 49.9a (18.6) | 49.8b (17.7) | N/A |
| Frequency of food shopping | |||
| Always | 53.2a | 53.8b | N/A |
| Often | 42.4 | 42.3 | N/A |
| Frequency of cooking at home | |||
| Every day | 81.3a | 79.8b | N/A |
| Several times a week | 9.3a | 10.6b | N/A |
Table 1 suggests that the sample was not representative of the population in terms of sociodemographic characteristics. Yet, since the sample mainly consisted of consumers who purchase and cook food at home, it is an important representation of the general population that decides what to buy and cook in the households of the target market region. Table 1 shows that participants always or often purchase food (95.6%) and cook every day/several times a week (90.6 %). Most of the respondents are female (76%), with an average age of about 54.5 years, having lived in the region for 49 years. The overrepresentation of females in our sample can be attributed to women being more responsible for cooking than men (Wolfson et al., 2021) but we cannot discard that is driven by the general trend that finds women more likely to participate in surveys than men (Curtin et al., 2000). The average composition of the households is three members, and about 70.0% of the participants live in households without children under 18 years. Regarding household incomes, about 50% of the respondents have a net household monthly income of less than 2,500€ and about 11% of them earn more than 3,500 €/month. Respondents with a secondary level of education are under-represented (26%), and participants with a higher level of education are over-represented (53%). The high proportion of participants with a university level of education is common in many studies because they are more inclined to respond to questionnaires (Verhoef, 2005).
Experimental design and implementation
⌅Product selection
The selection of the product is driven by the interest of local producers who want to launch the potato variety in the local market of Aragón. The traditionally produced local potato variety in the region used in this research is ‘Agria’. ‘Agria’ occupies a large part of the cultivation area in the Spanish territory, both with national and international seed. It is a variety of great importance in the region of Aragón. The different producing areas in this region are in the Valle del Jiloca (Zaragoza and Teruel), Bajo Aragón, Somontano del Moncayo (Zaragoza), and Hoya de Huesca and Pyrenees (Huesca). ‘Agria’ is a late-season variety, known for the strong flavor, and the particularity of not losing color when cooked. This potato is a good source of fiber, and vitamins B and C. ‘Agria’ is best suited to make French fries as they preserve the crispness on the outside and the tenderness on the inside. They are also very versatile with different cooking methods (e.g., baked, mashed, roasted, and boiled) as they do not disperse and have a pleasant taste.
The type of packaging used in the experiment was selected after exhaustive market analysis of potatoes sold in the local market at the time of the research. This resulted in identifying a bag of 3 kg of potatoes as the reference packaging. The lowest market price for 3 kg of potatoes was 2.4€ or 0.8 €/kg, and the highest market price found was 3.6€ or 1.2 €/kg. Besides the local potato, five other potatoes sold in the market with different characteristics were selected. Potatoes were displayed as available in the market, in 3-kg bags with all the extrinsic information (e.g., price, brand, origin, potato variety, producer, etc.). However, as some bags do not allow for visual observation of the product, small bulk displays were set up near the packaged potatoes to allow consumers to visually inspect the differences between products (see Fig. S1 [suppl]). The potatoes were numbered 1 to 6 where 4 was the number given to the local potato. The rest of the potatoes were from the same and different varieties (‘Agria’, ‘Monalisa’ and ‘Kennebec’) produced in other Spanish regions (Castilla y León, La Rioja, Galicia and two national origins labeled as Spanish) (Table 2).
| Potato name | Variety | Origin | Visual appearance Mean (St. Dev.) | Potato choices (%) |
|---|---|---|---|---|
| 1. Del Terruñoa | Agria | Castilla y León | 5.61 (1.83) | 9.4 |
| 2. Denifrieb | Agria | La Rioja | 4.33 (1.72) | 0.7 |
| 3. Freirc | Agria | Spain | 4.93 (1.86) | 4.4 |
| 4. Potaticod | Agria | Aragón | 6.78 (1.57) | 39.1 |
| 5. Cachelosa | Kennebec | Galicia | 5.61 (1.90) | 14.5 |
| 6. Potatod | Monalisa | Spain - France | 6.48 (2.09) | 31.9 |
Experiment 1: Shopping stage
Upon arrival, participants were informed on the nature of the experiment and signed a consent form to participate. An identification number was assigned to guarantee anonymity. The researcher explained the working sessions and provided information in clear written instructions.
The design in experiment 1 consisted of three different tasks where consumers had to evaluate visual appearance rating, declare their PI, and provide their WTP.
First, the participants were asked to visually inspect six different potatoes as evaluated on the shelves of a supermarket, and were asked to choose one bag, the one they most prefer. To have a better visual inspection, small bulk displays were set up near the packaged potatoes to allow consumers to visually inspect the differences between products (see Fig. S1 [suppl]). Then, the participants rated their visual liking of the different potatoes using a 9-point hedonic scale from 1 “dislike extremely” to 9 “like extremely”.
Second, the participants were given information regarding the origin of the local potato. Although the regional origin of the local potato was displayed in the bag in the previous task, some of the participants might have skipped reading this information. Therefore, to ensure that all the participants had been provided the same level of information, the following information was given “This potato is locally produced in Aragón in different counties (several counties were mentioned). The potato has been traditionally produced in the region”. Then, the participants were asked to rate again their visual liking but, from this task and forward we examined the preferences only for the local potato (no. 4) excluding the rest of the potato packages. Ratings were measures using a 9-point hedonic scale from 1 “dislike extremely” to 9 “like extremely”2.
Third, consumers were asked to declare their intent to purchase the local potato using a scale from 1 “very unlikely” to 5 “very likely”. To measure the WTP a Becker-DeGroot-Marschak (BDM) hypothetical auction mechanism was used. The participants were asked to submit a bid for the local potato. They were asked to submit their maximum WTP for the potato and we randomly drew a price; however, the participants were informed that there would not be a transaction due to the lack of product availability. Finally, participants responded to several questions regarding sociodemographic characteristics, purchase and consumption habits (Table 1). After completing the questionnaire, the participants received two pieces of the local potato, and were asked to fry them at home and then taste (Experiment 2). They were instructed to cook the potatoes as they usually do in their household. The researcher explained the sensory analysis questions and informed the participant on how to return the sensory evaluation material. All these explanations were also written and provided with the sensory analysis materials.
2 If participants paid attention to the information on the bag, we expected that this rating and the previous would be similar.
Experiment 2: Consumption stage
The main objective of experiment 2 was to evaluate the experience of the sensory attributes (i.e., taste) of the locally produced potato. The participants were asked to fry the potato as they normally do at home and rate the sensory aspects. First, they were asked to rate the overall liking, then the appearance (color), followed by the odor, texture, and finally rate the overall taste. The overall liking was rated by asking them the following question: “Please rate the overall liking of this local potato. How much do you like it?”. A 9-point hedonic scale was used where 1 indicated “disliking extremely” and 9 “liking it extremely”. The questions regarding the rest of the sensory aspects were similar. More precisely, the participants were asked “How much do you like the overall taste of this local potato?”. An identical question was repeated for odor and texture. Likewise, their rating consisted of a 9-point hedonic scale where 1 indicated “disliking it extremely” and 9 “liking it extremely”.
METHODOLOGY
⌅Experiment 1: Shopping stage
All the statistical analyses for both experiments were conducted in STATA 10.1 (StataCorp L.P., 2007). For the evaluation of the different potato packages (Task 1 in Fig. 1), we report the percentage of respondents choosing each package and the average and standard deviations for the visual appearance liking scores (Table 2). To test if the visual appearance liking scores are different among potatoes, we used t-tests.
The average of the measurements for visual appearance liking, PI and the WTP for the local potato (Task 2 and 3 in Fig. 1) was calculated to study the acceptability of the local potato. To study the relationships among these three measurements, we estimated different single and two regression models. To specify these models, we considered that, although the three measures offer different ways of understanding food products’ acceptability, they assess different aspects of the same phenomenon. Ginon et al. (2014) previously analyzed the relationship between WTP and overall liking scores by calculating the consistency between them (using a Kendall correlation coefficient between mean values). They found consistent results with previous studies (Ginon et al., 2009; Andersen et al., 2019), an overall consistency in variant ranking between hedonic and WTP. Considering these studies, we analyzed the relationship between WTP and either overall liking scores or PI scores, respectively. To do that, we used a set of different regression models. First, we run a regression of WTP on either overall liking or PI. The overall liking scores directly measure the sensory characteristics of the product, reflecting sensory preferences. The willingness to buy and the WTP, for their part, also include other attributes in their evaluation, reflecting purchase preferences. This means that we expect a higher explanatory power of the PI scores on the WTP relative to the explanatory power on the overall liking scores (Ginon et al., 2009). The last two measurements are more similar and related with one another (i.e., the Pearson correlation coefficient should be statistically significant and different from zero). In other words, consumers are generally willing to pay for a product if they previously liked it, which might indirectly affect WTP through the influence on the PI. Therefore, we assume that WTP is directly explained by the PI and this PI is determined by the overall liking. Then, the following two-equation model was specified and estimated:
where OL is the overall liking and WTP is the participants’ WTP. In addition, X is the set of sociodemographic, purchase and consumption habits of consumers, ε and γ are the error terms. We assumed that these two errors were correlated following a joint normal distribution N (0, Ω). Then, a recursive system with correlated errors was estimated (Drukker, 2011). Table 3 shows the estimations for this two-equation recursive model and the variables included in this model are described in Table S1 [suppl].
| PI | WTP | ||||
|---|---|---|---|---|---|
| Average (St. dev.) | 3.88 (0.89) | 2.89 (0.63) | |||
| Variables | Coefficient | t-ratio | Variables2 | Coefficient | t-ratio |
| Constant | 2.37 | 7.03*** | Constant | 2.27 | 10.20*** |
| Visual appearance1 | 0.14 | 3.38*** | PI | 0.20 | 3.77*** |
| Female | 0.40 | 2.41** | Age55 | -0.14 | -1.64* |
| Vegetarian | -1.06 | -2.57** | Low income | -0.28 | -2.84*** |
| Shop_super | 0.44 | 3.08*** | |||
| Adj R2 | 0.20 | Adj R2 | 0.15 | ||
| Adj R2 model only appearance in PI | 0.14 | Adj R2 model only intention in WTP | 0.06 | ||
Experiment 2: Consumption stage
The average and standard deviation for the different sensory likings (OL, appearance, odor, texture, and taste) and the Pearson correlation between them were calculated and reported in Table 4. To test if the different sensory likings were statistically different, t-test analyses were conducted. To measure the contribution of each of the sensory likings in consumers’ evaluation of OL, four single regressions were estimated. The OL scores were the endogenous variables in the four equations and the different sensory liking scores were the explanatory variables in each of them. Then, the adjusted R2 is an indication of how much each of the sensory liking scores explain the OLs. The higher the value the higher the explanation. In addition, the estimated coefficient for each of the sensory liking scores measures the marginal effect of this particular sensory liking on the OL. Then, the higher the marginal effect the higher the influence on the OL. Considering both values, the adjusted R2 and the liking estimated marginal effects, the higher they are, the more important the sensory liking scores in the OL.