Cumulative information on quality and willingness to pay: a study on wine evaluation

ABSTRACTAvailability of information about a good with uncertain quality can influence the way consumers perceive its quality, hence, their willingness to pay (WTP) for it. We present a study to investigate whether and to what extent WTP is impacted by the degree of information available to consumers who are exposed first to extrinsic and then intrinsic information regarding a variety of Italian wines. We implement linear mixed models in a Bayesian framework, which provides a flexible tool to account for different sources of heterogeneity, e.g. correlation within groups of observations and spatial correlation between participants sitting nearby. Based on data collected in Italy, results show that the availability of extrinsic and intrinsic information yields relevant changes in WTP, but this effect also depends on age, gender, drinking habits, wine quality, and connoisseurship of the agents. According to the findings, the analyzed wines cannot be considered search goods, although this might not hold for more experienced consumers.KEYWORDS: Wine qualitywillingness to paylinear mixed modelinformation asymmetryINLA Disclosure statementNo potential conflict of interest was reported by the author(s).Ethical statementThe participants to the study had to book it (for free) on a website, knowing that they would have been part of a data collection process and to a wine tasting event, with an expert speaking. During the event, they got to know that their data will be used for a statistical analysis. They were able to leave the event whenever they wanted.Supplementary dataSupplemental data for this article can be accessed online at https://doi.org/10.1080/15378020.2023.2259315Notes1. Though in literature it is unclear which good characteristics consumers find most relevant to evaluate its quality, consumers’ quality perception is modeled within several contexts such as marketing, psychology, business, and economics. For instance, within the economic literature, theoretical foundations of the approach based on the characteristics of the good have been provided by Lancaster (Citation1966), Muth (Citation1966), and Becker (Citation1965), and this approach has been empirically tested in several studies using hedonic techniques (e.g., Muellbauer, Citation1974).2. A search good is a good whose quality can be evaluated by the consumer before its consumption (Stigler, Citation1961). An experience good is a good whose quality can only be evaluated by the consumer after its consumption (Nelson, Citation1970). A credence good is a good whose characteristics cannot be evaluated by the consumer even after its consumption, and the evaluation of its quality requires further information that could be costly to obtain and can be collected from experts (Darby & Karni, Citation1973).3. Similar to the market for lemons (Akerlof, Citation1970), the negative effect of (ex-ante) asymmetric information is due to the adverse selection mechanism, whereby an agent who possesses more information about the quality of the good than others exploits his informational advantage to the detriment of others.4. The previous literature has also examined the influence of information on WTP for restaurants. For instance, researchers have explored how consumers may deduce food quality from cues like the physical ambiance and ethnic authenticity, as demonstrated in the work of Lin and Jiang (Citation2022). Similarly, factors such as food presentation can impact perceptions, as illustrated by Kuo and Barber (Citation2014) investigation into the material of dishware used. Also fine-dining products’ engrossment and acquaintance play a role in shaping consumers’ WTP (Gupta et al., Citation2022). The presence of uncertainty extends beyond quality and encompasses other facets of the consumption experience, including safety and health considerations. This was underscored by Belarmino and Repetti (Citation2022) study, which identified an effect associated with the usage of masks within restaurant environments during the COVID-19 pandemic.5. The former is an indivisible part of the product, such as its color, while the latter is external to the product, such as its brand and its packaging.6. Information can originate from various external influences like marketing campaigns, which can subsequently impact consumers’ WTP. For instance, Remar et al. (Citation2016) demonstrated that local food marketing plays a significant role in influencing consumers’ WTP and, consequently, their purchasing choices for local food products.7. Experimental auctions are a form of non-hypothetical valuation, while laboratory and field experiments are the other forms of experimental methods to measure the WTP. See, for example, Lusk (Citation2003), Huffman et al. (Citation2003), Didier and Lucie (Citation2008), and De Groote et al. (Citation2011). For a review of methods for measuring WTP, see Breidert et al. (Citation2006).8. To understand the nature of wine and whether it is a search good or not, in Appendix A we formally represent a theoretical framework to conceptualize this process.9. The event had a total of 40 available seats for participants. At the time of booking their seats, all participants were informed about the data collection process, the presence of an expert, and the tasting experience. They could leave whenever they wanted, but all 38 participants remained until the end of the event. The characteristics of the participants, including their demographic information and past experience, are detailed in Table 3.10. Table B1 in Appendix B provides the prices of the wines obtained from a restaurant wine list, as well as the ranking of the quality of each wine. To ensure the accuracy of the rankings, we cross-referenced the expert ratings from several renowned Italian wine guides. The rankings aligns consistently with these sources. For instance, when considering the expert ratings as an indicator of quality, Chardonnay and Brunello receive the highest scores according to Gardini Notes Wine Ranking by Luca Gardini. On the other hand, Verdicchio and Sangiovese receive the lowest scores according to Sensorial Synesthetic Wheel by Luca Maroni. Nebbiolo and Malvasia receive intermediate ratings according to Bibenda by Fondazione Italiana Sommelier. By consulting multiple sources and comparing our rankings, we ensured the reliability and validity of our assessments of wine quality.11. The concept of “local” in this paper does not refer to the region of origin but rather to the proximity of the production location and Rimini in the Emilia-Romagna region. While the Verdicchio wine that is part of this study is produced in the Jesi area of the Marche region, it is located less than 80 km in a straight line from Rimini. The other wines in this study are produced at greater distances from Rimini, as we reported in Table B1 in Appendix B. Concerning distance and its relationship with consumers’ WTP for wines, Ay et al. (Citation2017) find that the premium associated with organic wines decreases as the distance between the consumer’s home and the vineyard increases, suggesting a role of proximity in the evaluation of wine. The contribution of local products to shaping consumers’ WTP has also been investigated in the context of restaurants and foodservice (e.g., Remar et al., Citation2016; Shin et al., Citation2018).12. Table 2 provides information on the content of each label.13. All the participants were involved in all six rounds of the study, meaning that the information accumulation process was consistent for all wines and participants. As a reward for the participation, a final lottery was conducted, where six prizes, each consisting of a bottle of wine, were awarded. In the first round of the lottery, one of the participants was randomly selected as the winner. This winner was then removed from the list of potential winners for the remaining rounds. This ensured that each participant had the opportunity to win up to one bottle of wine as reward for their participation.14. From Figure D1 in Appendix D it can be noticed that all six wines present variability in the declared WTP at the three information levels (the lines are rarely stable throughout the three steps), often showing growing or inverted U-shaped dynamics.15. The sample of participants in this study appears to be heterogeneous, although it is not representative of the entire population, which includes individuals who do not drink wine. It is important to note that nondrinkers were unable to participate in the event as one of the steps involved tasting the wines. Additionally, it is likely that nondrinkers would not have had the inclination to participate in the study, since the participants had to voluntarily enroll in the study. Therefore, the sample primarily consists of individuals who have an interest in wine and are willing to engage in wine-related activities, and the results will reflect this type of consumers’ behavior.16. This constraint does not change the model, but it is only applied to facilitate interpretation of the results.17. For instance, from Figure D1 we see that WTP measurements taken on a single subject are correlated, thus we need to deal with not i.i.d. errors.18. The existence of a spatial structure might be observed when participants that are near one another are part of the same social group (e.g., a group of friends or family).19. The matrix R is rank-deficient and the notation R−1 in Eq. (4) indicates the generalized inverse; the model by Besag (Citation1974) is denoted in the statistical literature as intrinsic conditional autoregression (ICAR).20. See Figure D2 in Appendix D, displaying the estimated smooth effect of age on WTP, where a decreasing (almost linear) trend can be observed, however suggesting there is no clear relation between WTP and age.21. We estimated all models using the R package INLA (Rue et al., Citation2017). In Appendix E, we present the main advantages of using INLA.22. A credible interval summarizes the posterior distribution of a parameter and is the main inferential output of a Bayesian analysis.23. For the sake of simplicity and given the exploratory purpose of the work we avoid formal hypothesis testing and look at the posterior credible intervals of the estimated effects to detect “statistically significant” results. For instance, we denote a change between WTP at different cumulative information levels, e.g., β2−β1, as significant if the 95% credible intervals of β2 and β1 do not overlap; when the intervals partly overlap we refer it to poor evidence of a significant change. Thus, we use the words “statistically significant” in a loose sense here. Commenting the credible intervals in this way is safe and practical way to interpret the Bayesian model output without incurring overstatements (i.e. statements that are not supported by the data). A formal Bayesian testing procedure would be possible by computing the posterior probability that β2>β1 and rejecting H0:β1=β2 if this probability exceeds some pre-defined threshold.24. Credible intervals for the participants who stated that they attended at least a wine-tasting course are wider with respect to those of other participants. This is due to the low number of participants (6) who answered positively.25. We would like to express our gratitude to the anonymous referee who brought attention to this limitation, thus paving the way for future research to address and overcome it.Additional informationFundingThe study was part of one of the events of the Rimini Campus of the University of Bologna ‘European Researchers’ Night’, the funding for that event come from University of Bologna’s Rimini Campus funds. The funding source had no role in the design of the study, in the collection, analysis, and interpretation of data, in the writing of the report, and in the decision to submit the article for publication.