II. Background

III. Online experiment

We designed a 2 x 5 (i.e., 10 wines) between-subject randomized online experiment. We presented two different label designs, that is, traditional and modern, and five sets of information about the producer, including no producer information, which we used as a reference point.

Participants were French-speaking respondents because French is a gendered language. French grammar distinguishes nouns by gender, most of which are associated with a male or female gender tag. Anything that is attached to a noun (pronouns, adjectives, and verbs) will also reflect the assigned gender (DeFranza, Mishra, and Mishra, Reference DeFranza, Mishra and Mishra2020). For instance, vigneron is the French noun for a male winemaker, and vigneronne is the feminine variant; the spelling as well as the pronunciation are different. This is why we decided to run the online experiment in France and in the French-speaking part of Belgium (Wallonia). In addition to language, using these two countries enabled us to gather data from a wine-producing country (France) and a non-producing country (Belgium), as well as examine any cultural differences.

The labels used for the experiment contain exactly the same information (winery name, appellation of origin, vintage, and other mandatory information), except for the information related to the winemaker. Two styles of labels were also used (traditional vs. original; see Celhay and Passebois, Reference Celhay and Passebois2011). Because we use fictitious labels and producer names, our online experiment relies on a kind of deception. If deception appears as standard in the psychology and marketing fields, it is often thought that it should be banned in experimental economics (Cooper, Reference Cooper2014). Indeed, control of the experiment is lost if subjects doubt experimental instructions and material and form their own theories about payoffs. However, Cooper (Reference Cooper2014, p. 113) notes that deception can be considered acceptable when the following four rules are met: “(1) The deception does not harm subjects beyond what is typical for an economic experiment without deception. (2) The study would be prohibitively difficult to conduct without deception. (3) Subjects are adequately debriefed after the fact about the presence of deception. (4) The value of the study is sufficiently high to merit the potential costs associated with the use of deception.” Here, rules (1) and (2) are met. The kind of ambiguity used here is found in most wine experiments, as observed by Masset and Raub (Reference Masset and Raub2023). Regarding the last rule, Cooper (Reference Cooper2014) notes that the value of a study lies in the eye of the beholder. But he also mentions that “most economists would argue that the discrimination studies using deception were sufficiently important […] that the use of deception was justified” (p. 114). Given that our aim is to analyze if WTP varies with information about the gender of the producer, we believe that deception is necessary to investigate this kind of discrimination. Rule 3 is trickier: given the number of respondents, who remain anonymous in an online setting, it is hard to organize a debrief. A professional panelist who sent out the survey mentioned that those who participate in online studies might use some kind of misinformation and/or ambiguity. On the one hand, study participants have to certify that they are committed to answering spontaneously and independently from previous studies they might have participated in. On the other hand, the participant manages the frequency of survey participation requests to avoid any learning process, generating the belief that experimenters are liars.

Each wine (except the reference points) appears with a sentence indicating who made the wine or whether the producer belonged to a particular producer group, as shown in Table 1. Wines 1 and 6 are the reference points, as no information about the winemaker is provided. Wines 2 and 7 are made by winemaker Georges Cadieux, and wines 3 and 8 are made by winemaker Nathalie Panetier. Both names come from a randomly generated list. In French, the first names Georges and Nathalie are, respectively, masculine and feminine. Moreover, their phonetic structure suggests a male versus female personality, respectively, as shown by Wu, Klink, and Guo (Reference Wu, Klink and Guo2013). On wines 4 and 9, the sticker Fémivin refers to a group of female winemakers only. To our knowledge, in France, no such producer group exists for male winemakers. However, Vignerons Indépendants is a French winemaking trade association that includes the noun “male winemakers” in its name (vignerons), even if it is not dedicated to male winemakers only. We use its sticker on wines 5 and 10 as a reverse of the female winemaker group.

Table 1.

2 x 5 between-subject online experiment, WTP (euros), and perceived quality for each wine (N = 1,351)

The survey took place from March 25, 2021, to April 2, 2021, under almost unrestricted circumstances (i.e., no strict lockdown due to the COVID-19 pandemic in France and Belgium). Respondents from France and Belgium were randomly drawn from a panel of consumers selected by the company Respondi, a professional panelist that sends out survey links by email. Respondents are rewarded only if they fully complete the survey with points that are transformed into gift cards usable in several stores. A screening question enables the selecting of respondents who have consumed wine over the last three months. Each respondent was assigned randomly to one of the ten conditions. The final sample consisted of 1,000 French and 500 Belgian respondents, with a distribution among experimental conditions that ranged from 9.33% to 10.51% (see Table 1). We excluded respondents who preferred not to declare their income, which left us with 1,351 observations.

The survey started with a randomized experiment. Respondents were asked first to assess how they perceive the quality of the wine over a 5-point scale (from 1—poor quality to 5—excellent quality) and if they were willing to purchase the wine using a 5-point Likert scale (from 1—extremely unlikely to purchase to 5—extremely likely to purchase). Then, cheap talk was used right before the valuation question, when respondents were invited to give their WTP, in euros, for the same bottle of wine and to indicate 0 if they were not willing to pay for the wine; 100 euros was the maximum possible amount. Average WTP, perceived quality (PQ), measured on a 5-point scale, and WTP measured on a 5-point Likert scale are shown in Table 1 for each wine. An attention check question was also included in this part of the questionnaire.

Sociodemographics, as well as wine-related characteristics and opinions of respondents, were questioned in the final section of the questionnaire. We know the respondents’ wine consumption frequency and if they were used to consuming red wine. Objective wine knowledge was reflected in a score of over 5 points obtained from a set of five questions inspired by Velikova, Howell, and Dodd’s (Reference Velikova, Howell and Dodd2015) scale. Respondents received 1 point for each correct answer, 0 otherwise. The questions are presented in Appendix 1. The survey also included a question about the perceived quality of Bordeaux wines as a whole as a proxy for the collective reputation of the wine industry (Landon and Smith, Reference Landon and Smith1997, Reference Landon and Smith1998).

Summary statistics of sociodemographic and wine-related characteristics and opinions are presented in Tables 2 and 3. The sample is in line with the age and gender proportions of the population of interest (in this case, wine consumers) in the two countries studied (Wine Intelligence, 2022, 2023). We note that the average amount of money usually spent for a bottle of wine is similarly distributed across all conditions (details available from the authors on request).

Table 2.

Descriptive characteristics of respondents (N = 1,351)

Table 3.

Wine characteristics and opinions of respondents (N = 1,351)

IV. Empirical analysis

To analyze the effect of the winemaker’s gender on the consumer’s WTP, we estimated a series of regressions in which the WTP was a function of the presented label, controlling for sociodemographic and wine-related characteristics and opinions of the respondents. Two empirical issues required additional attention. First, WTP is a left- and right-censored variable because respondents declared 0 euros when they were not willing to pay for the bottle of wine, and 100 euros was the maximum amount they could declare. Said differently, our dependent variable has a corner at 0 and, in such cases, the distribution of WTP exhibits a spike at 0 with about 5% of respondents, as shown in Figure 1. This implies that we should use some censored regression model, like the Tobit-type models.

Figure 1.

Distribution of WTP (N = 1,351).

Second, our sample might include protest responses as well as out-of-market respondents. Indeed, protest responses are identified with respondents who claimed that they were willing to purchase the good but then declared a 0 WTPFootnote ¹; some other respondents opted out of the market, claiming that they were not willing to purchase the good, but some declared a non-0 WTPFootnote ² (see Figure 2). In that case, a spike-based modeling approach à la Kriström (Reference Kriström1997) can be implemented. It uses a hurdle model that combines discrete and continuous parts (Brown and Taylor, Reference Brown and Taylor2000), that is, a set of two equations: a discrete component models the participation decision, estimating the probability that a respondent will state a positive WTP, and a continuous component models the contribution decision, with WTP as a dependent variable only for those respondents stating a positive WTP. These two equations can be estimated separately in two stages (Reiser and Shechter, Reference Reiser and Shechter1999; for an application, see Brown and Taylor, (Reference Brown and Taylor2000)). Such an empirical strategy excludes respondents willing to purchase the good but only willing to pay 0 because of current circumstances or characteristics. Moreover, it can overestimate the effect because it produces higher values than other models (Bengochea-Morancho, Fuertes-Eugenio, and Del Saz-Salazar, Reference Bengochea-Morancho, Fuertes-Eugenio and Del Saz-Salazar2005). Along the same vein, another possibility is to estimate a double-hurdle model (Cragg, Reference Cragg1971): the first hurdle refers to the participation decision, the second one to the contribution decision, and both decisions are modeled simultaneously. The double-hurdle model enables distinguishing among respondents declaring a WTP of 0 because they do not want to purchase the good and those for whom current circumstances or characteristics dictate a WTP of 0 but would like to purchase the good, as suggested by Engel and Moffatt (Reference Engel and Moffatt2014). In Figure 2, we present only the second part of the double-hurdle model, that is, the contribution decision, to explain differences in WTP.

Figure 2.

Sequence of valuation questions with percentage of respondents.

Clinch and Murphy (Reference Clinch and Murphy2001) recommend using a range of models to infer WTP. In Table 4, we present robust estimation results using the Tobit and the double-hurdle modeling approaches for two empirical specifications: model 1 includes a dummy variable for each label, and model 2 assesses each label against the gender of the respondent. All specifications include the same set of sociodemographic characteristics of respondents: gender, age, country (France or Belgium), and monthly net household income. They also contain the following wine-related indicators: wine consumption frequency, if the respondent is the primary wine shopper, level of objective wine knowledge, and the average amount of money usually spent on a bottle of wine. We also include a collective reputation indicator and control for the perceived quality of each bottle of wine because it is traditionally a determinant of WTP for wine (e.g., Lunardo and Rickard, Reference Lunardo and Rickard2019) and for the kind of label (traditional vs. modern). In Table 4, we report only the coefficients for the variables of interest. More detailed results are available in Appendix 2.

Table 4.

Estimation results

Robust standard errors in parentheses.

*** p < 0.01, **p < 0.05, *p < 0.1

As expected, the perceived quality of each specific label increased the WTP significantly, and the collective reputation of Bordeaux wines, such as the label’s style, did not prove significant. We also got a negative effect from age, from France compared to Belgium, and from the level of objective wine knowledge. Unsurprisingly, the effect of the average amount of money spent on a bottle of wine was positive.

The Tobit estimation of model 1 highlights a price reduction associated with the group of female winemakers (–1.17 euros, significant at 10% only), although it is not significant in the double-hurdle model. When assessed against the gender of the consumer (model 2), our estimation results suggest that male respondents ask for a price reduction when the wine is produced by a female producer group (–1.4 euros with the Tobit model, significant at 10%, almost –3 euros with the double-hurdle model, significant at 5%). The Tobit estimation of model 2 also highlights that female consumers devalue wines when the name of a female winemaker appears on the label (–1.8 euros, significant at 5%), but this is not confirmed by the double-hurdle model. Here, we get a result in line with the empirical evidence provided by Tak, Correll, and Soule (Reference Tak, Correll and Soule2019): female products, produced in a male-typed industry, are disadvantaged by male consumers. Our results show that this is the case when women form an all-female producer group. As such, the study respondents react differently to the group of producers than to a single female producer.

V. Discussion and conclusion

Our results suggest that the gender of the winemaker matters in wine valuation. Specifically, the wine industry is still male-dominated, and it seems that collective strategies supporting women are not appreciated by consumers. Interestingly, we do not find any significant difference in the WTP between products made by male and female winemakers when the gender information is communicated through the producer’s first name. This can suggest that the winemaker’s name is associated with a craftsperson genuinely making the wine. In that case, the name is a means to build authenticity (Maguire, Reference Maguire2018), whatever the gender of the producer, which makes no difference in terms of WTP.

We get different results for a group of female producers only. When a female winemaker claims to belong to a group of women making wine, consumers, especially male consumers, want a price reduction for the group’s wine. Mentoring and networking can help female winemakers address several challenges not considered by traditional professional organizations (Le Brun, Guétat-Bernard, and Annes, Reference Le Brun, Guétat-Bernard and Annes2019). The Women Do Wine group (France), the Diversity in Wine Leadership Forum (North America), the Australian Women in Wine Awards, and the Women in Wine Leadership Symposium (U.S.-based but international) seek to support and mentor women and highlight their role in the wine industry, but these groups of women are new and operate in only a few wine-producing countries (Livat and Jaffré, Reference Livat, Jaffré, Charters, Demossier, Dutton, Harding, Smith Maguire, Marks and Unwin2022). We think that they are not well enough known at this time and that there is a lack of familiarity and fluency on the consumers’ side, which can produce the desire for a price reduction. Moreover, because they challenge professional institutions and do not align with traditional communication strategies, these groups of female producers can signal much more than information about the producer: they can draw a political statement and aim at lobbying (English, Reference English2022). Hence, belonging to these organizations can be perceived as activism, associated with negative stereotypes (e.g., eccentric and militant; see, for instance, Bashir et al. (Reference Bashir, Lockwood, Chasteen, Nadolny and Noyes2013)), and generate a backlash discourse (Dyer and Hurd, Reference Dyer and Hurd2018), especially on the male consumer side.

Previous research has shown that, in California, female winemakers are more highly acclaimed by experts proportional to their presence in the field than male winemakers (Gilbert & Gilbert, Reference Gilbert and Gilbert2015), suggesting some inverse discrimination. However, transactions on product markets can be affected by biased judgments and decisions. If the quality of wines made by female winemakers does not matter from the experts’ point of view, consumers might demand a price reduction, which reflects negative discrimination. This is highlighted by our results in the case of male consumers of products from female winemaker groups. This specific discrimination might also come from their peers, that is, female consumers, which deserves further research.

The study could be replicated in other countries with a gendered language but a high gender balance, such as the French-speaking part of Switzerland, a country with one of the lowest UN Gender Inequality Index rankings (United Nations, 2022). English-speaking countries, that is, those using a non-gendered language such as the United States, should also be studied. In the same way, a laboratory experiment could be carried out to overcome the hypothetical situation. This would involve a blind tasting of the same wine presented as having been made by a male winemaker on the one hand and a female winemaker on the other. Complementary research could investigate in depth the place and role of women in the wine industry and, more generally, how the growing diversity affects the industry. For example, some field experiments would be helpful to analyze if and how a female wine salesperson compared to a male one influences purchasing behavior in the same wine store. It could be used to study how a female wine tour guide or a female wine steward in a restaurant, compared to a male one, affects customer satisfaction. These questions also make sense for other minorities in the wine industry, for example, people of color, people who are disabled, and LGBTQIA+ people, because the industry is mainly dominated by white people (Inglis and Ho, Reference Inglis, Ho, Charters, Demossier, Dutton, Harding, Smith Maguire, Marks and Unwin2022).