II. Literature and hypotheses

Previous research has demonstrated that the phonetic patterns of names affect product information, consumer behavior, and even pricing (e.g., Klink, Reference Klink2001, Pogacar et al., Reference Pogacar, Plant, Rosulek and Kouril2015). The association of phonetic patterns and inherent product qualities is known as phonetic symbolism (Keller and Lehmann, Reference Keller and Lehmann2006). Market participants unconsciously process these associative perceptions, which affect their judgment of names without conscious awareness (e.g., Yorkston and Menon, Reference Yorkston and Menon2004, Pogacar et al., Reference Pogacar, Shrum and Lowrey2018). The association of specific product attributes over phonological properties has been investigated mainly in experimental studies (e.g., Arora et al., Reference Arora, Kalro and Sharma2022). Following Wu et al. (Reference Wu, Klink and Guo2013), phonetic symbolism creates brand personalities in that names with front vowels indicate a more feminine brand personality, while names with back vowels construct a masculine brand personality. Moreover, Klink (Reference Klink2009) shows that females prefer the two front vowels [i] and [e], while males prefer the two back vowels [o] and [u]. These studies apparently reveal specific gender-specific preferences for names. At the same time, product characteristics can be evoked by these phonetic structures, which can also be transported as quality-indicating characteristics in wine consumption. For instance, we assume that qualities such as lightness could also be associated with a particular grape variety or wine color.

a. Phonetic gender score

Men’s and women’s names have distinguishable patterns; thus, individuals learn to discern linguistic gender indicators in names (Barry and Harper, Reference Barry and Harper1995). Barry and Harper (Reference Barry and Harper1995) devised a phonetic gender score, a quantitative scale based on a name’s length, sounds, and stress that has been used to measure the degree to which a name is feminine or masculine. The phonetic gender score helps determine if a novel or unfamiliar name is feminine or masculine. The phonetic gender score is the sum of two quantitative scales, ranging from −2 (very masculine) to +2 (very feminine). Names with positive scores are predominantly feminine. Names with negative scores are predominantly masculine. One scale measures the phonetic traits of the entire name. The other scale measures phonetic traits of the last phoneme (i.e., the characteristics of a single sound). This normative scale is based on American male and female given names. The phonetic gender scores can be applied to a wide variety of names, other ethnic groups, and different languages (e.g., Latin, French, Spanish, German, and Russian), and consumer products, boats, houses, streets, cities, etc. (Barry and Harper, Reference Barry and Harper1995). With later research, it has conceptually been validated (e.g., Cassidy et al., Reference Cassidy, Kelly and Sharoni1999, Slepian and Galinsky, Reference Slepian and Galinsky2016, Whissell, Reference Whissell2001). Table 1 summarizes the phonetic gender score utilized in the present study (see Barry and Harper Reference Barry and Harper1995).

Table 1.

Phonetic gender scoring system (based on Barry and Harper Reference Barry and Harper1995)

Phonetic gender score = the average of scales one and two.

A study by Pogacar et al. (Reference Pogacar, Angle, Lowrey, Shrum and Kardes2021) examines how brand names express gender connotations and how this affects performance, choice, and attitudes. It shows how linguistically feminine names improve perceived comfort and brand performance. However, they show that the benefits of feminine brand names are diminished when the products are utilitarian and the average user is male. Additionally, some studies illustrate how gender implications affect names and inherent qualities, in which femininity is associated with attributes such as softness, lightness, and mildness. In contrast, masculine names are associated with opposite qualities (e.g., Klink, Reference Klink2000). Moreover, Guevremont and Grohmann (Reference Guevremont and Grohmann2015) have shown that name consonants affect consumers’ perceptions. Pogacar et al. (Reference Pogacar, Angle, Lowrey, Shrum and Kardes2021) have shown that feminine brand names are positively associated with more favorable attitudes and increased consumer choice. Masset et al. (Reference Masset, Terrier and Livat2023) have analyzed over 1400 expert tasting notes and found that more feminine wines receive similar ratings and sell for similar prices as their more masculine counterparts but are perceived as having limited aging potential.

Names are frequently employed in the wine industry to inform customers about the qualities and features of various wines. Motivated by the results of Pogacar et al. (Reference Pogacar, Angle, Lowrey, Shrum and Kardes2021), we test whether linguistically feminine names—often associated with positive qualities—receive higher ratings. As a result, we formulate the following hypothesis:

Hypothesis H1: Higher phonetic gender scores, and therefore more feminine names, will receive higher ratings.

III. Data and model

The dataset comprises wine reviews from wine-tasting experiments conducted for Wine Enthusiast Magazine. It includes specific data on each rated wine bottle’s country, description, winery, designation, rating points, reviews, wine price, province, and region. We obtained data on 18,609 wines professionally rated in blind tasting experiments between 2011 and 2016, yielding a sample of 31,058 individual experiment observations to objectively evaluate the influence of phonetic gender scores on quality outcomes measured by wine experts’ ratings.

The Wine Enthusiast Magazine was first published in print form in 1988. Since then, it has become a digital platform with the website www.winemag.com and various content, including podcasts, buying guides, and expert opinions. With 4.1 million readers focusing on journalism on wine and beverages, the magazine is regarded as one of the top seven wine publications in the U.S. (Storchmann, Reference Storchmann2012). The magazine also provides expert reviews of wines using a predetermined point system, which professional critics provide. According to the magazine information on their website, over 25,000 wines are tested annually as part of this professional grading system run by a network of 25 experts with specialized country-specific knowledge. The reviewers are also allocated as primary reviewers for wine regions, which elevate their status to that of authorities with in-depth knowledge of such regions.

Each bottle of wine tested receives a score between 80 and 100, where 100 is the highest quality rating. The producers of the wines and the prices are kept a secret; the blind tastings are conducted in groups of five to eight wine samples each. Reviewers may be aware of broad details about a sample, such as the vintage and variety. Still, they are unaware of any given option’s producer or suggested retail price. We are, therefore, able to track preferences solely based on the intrinsic qualities of the wine because the wines were served blind, which prevented participants from being swayed by label, price, or origin information (Wang and Prešern, Reference Wang and Prešern2018, Almenberg and Dreber, Reference Almenberg and Dreber2011).

Hence, the blind tastings possess all the characteristics of a quasi-experiment to establish a cause-and-effect relationship (see Figure 1). The reviewers are randomly assigned to different groups, which ensure that any observed effects are not due to pre-existing differences between the groups. In addition, blind tastings are standardized procedures to ensure all participants are treated equally. This helps to eliminate extraneous variables that could affect the results. As the blind tastings are conducted yearly, the results are reliable and can be generalized. Thus, the femininity and/or masculinity of the variety’s name can lead to implicit associations with wine-tasting characteristics.

Figure 1.

Blind tasting process.

We primarily focus on the femininity and/or masculinity of the variety’s name and the relationship between the ratings from professional reviewers, their gender, and the wine color characteristics. Using a technique by Barry and Harper (Reference Barry and Harper1995) (see Table 1), we determined the phonetic gender score of wine types. With values ranging from −2 (extremely masculine) to +2 (highly feminine), the phonetic gender score quantifies how much a name is masculine or feminine based on its length, sounds, and stress. This normative scale is based on male and female names from the United States and has conceptual support from research (e.g., Cassidy et al., Reference Cassidy, Kelly and Sharoni1999, Slepian and Galinsky, Reference Slepian and Galinsky2016, Whissell, Reference Whissell2001).

The complete list of model variables included in the regression analyses explains that brand outcome, measured by ratings for wine title i at time t, is

Model I

(1)\begin{align}Rating{s_{it}} & = \alpha + \beta\, Phonetic\,Gender\,Scor{e_i}\nonumber \\ & \quad + \vartheta\, Control{s_i} + {\varepsilon _{it}}\end{align}

Model II

(2)\begin{align} Ring{s_{it}} & = \alpha + \beta\, Phonetic\,Gender\,Scor{e_i} + {\gamma _1}Female\,Reviewe{r_i}\nonumber \\ & \quad + \vartheta\, Control{s_i} + {\varepsilon _{it}} \end{align}

Model III

(3)\begin{align} Rating{s_{it}} & = \alpha + \beta\, Phonetic\,Gender\,Scor{e_i} + {\gamma _1}Female\,Reviewe{r_i} + {\delta _1}White\,Win{e_i}\nonumber \\ & \quad + \vartheta\, Control{s_i} + {\varepsilon _{it}} \nonumber \\ & \quad \end{align}

Model IV

(4)\begin{align} Rating{s_{it}} & = \alpha + \beta\, Phonetic\,Gender\,Scor{e_i} \nonumber \\ & \quad + {\gamma _1}Female\,Reviewe{r_i} \nonumber \\ & \quad + {\gamma _2}\left( {Phonetic\,Gender\,Scor{e_i}x\,Female\,Reviewe{r_i}} \right)\nonumber \\ & \quad + \vartheta\, Control{s_i} + {\varepsilon _{it}} \end{align}

Model V

(5)\begin{align} Rating{s_{it}} & = \alpha + \beta\, Phonetic\,Gender\,Scor{e_i} + {\delta _1}White\,Win{e_i} \nonumber \\ & \quad + {\delta _2}\left( {Phonetic\,Gender\,Scor{e_i}x\,White\,Win{e_i}} \right) \nonumber \\ & \quad + \vartheta\, Control{s_i} + {\varepsilon _{it}} \end{align}

Model VI

(6)\begin{align} Rating{s_{it}} & = \alpha + \beta\, Phonetic\,Gender\,Scor{e_i} + {\gamma _1}Female\,Reviewe{r_i} + {\delta _1}White\,Win{e_i} \nonumber \\ & \quad + {\gamma _2}\left( {Phonetic\,Gender\,Scor{e_i}x\,Female\,Reviewe{r_i}} \right) \nonumber \\ & \quad + {\delta _2}\left( {Phonetic\,Gender\,Scor{e_i}x\,White\,Win{e_i}} \right)\; \nonumber \\ & \quad + \vartheta\, Control{s_i} + {\varepsilon _{it}} \end{align}

where the definitions of each of these individual vectors of variables are consistent with the categories of variables reported in the descriptive statistics shown in Tables 2 and 3.

Table 2.

Descriptive statistics

Table 3.

Descriptive statistics

Controls consist of country of origin, province, and winery dummies representing the country of origin and growing region of the wine i. Wine is produced in many countries with unique geographic, climatic, and growing conditions that determine product attributes. Consequently, the control dummies account for the effect of product-inherent characteristics on the taste and quality of the wine and thus on the wine ratings (e.g., Horowitz and Lockshin, Reference Horowitz and Lockshin2002).

Specifically, the values of the β coefficient allow us to test hypothesis H₁, the γ coefficients allow for the testing of H₂, and the δ coefficients allow for the testing of H₃.

Using appropriate variables capturing product characteristics, we present correlation coefficients between the variables used in the estimations in Table 4. The highest level of correlation can be observed between the wine colors and the phonetic gender score, with coefficients of −0.15 and +0.15. Consequently, these correlation levels are well below the threshold, whereby multicollinearity would significantly affect our results.

Table 4.

Pairwise correlations

* p < 0.1.

Figure 2 shows the distribution of phonetic gender scores for grape varieties and colors, respectively. We see that the distribution of scores for grape varieties is skewed to the left, with 14 grape varieties having a phonetic gender score of 1.5. However, we see that the phonetic gender scores are equally distributed across red (dashed line) and white (solid line) varieties.

Figure 2.

Distribution of phonetic gender scores across grape varieties (with indication of average scores).

So are the average scores (seen in the tiny bubbles), which range from 87 to 92. In addition, we present histograms to provide visual representations of the distribution of ratings in different scenarios, allowing us to explore how ratings are distributed overall, across different wine colors, and based on phonetic scores (see Figure 3). This grouped histogram shows the distribution of ratings separated by different wine colors. The histograms for phonetic scores compare the distribution of ratings between feminine and masculine sounding wines (i.e., phonetic scores less than 0 are considered masculine and scores greater than 0 are considered feminine).

Figure 3.

Histograms of the average rating scores.

The following section presents the estimation results and a series of alternate model specifications as a robustness check to explore a range of different interactions of signals both separately and jointly. Additionally, we check our results for validity, significance, and robustness.

IV. Empirical analysis and results

In the empirical analysis that follows, we use panel data regressions. Panel data regressions have more degrees of freedom and sample variability than cross-sectional data regressions and include intertemporal dynamics that allow for the control of missing or unobserved variables (Hsiao, Reference Hsiao2007). Since our outcome of interest, the rating points for wine i at time t, is time-varying, and our explanatory variables such as phonetic gender score, reviewer gender, or wine color are time-invariant variables, we focus on random-effects regressions to estimate variation within and between wines (Hsiao, Reference Hsiao2007). In this way, we take advantage of panel data and can deal with possible omitted variable bias due to heterogeneity in the data.

We consider six models to test our hypotheses H₁, H₂, and H₃ empirically. First, we examine the impact of phonetic gender score on the respective wine rating (Model I), also considering the respective gender of the reviewer (Model II), before explicitly including the wine type, divided into red and white wine (Model III) in the estimation. In the following estimation (Model IV), we introduce the interaction term of phonetic gender score and female reviewer to test for the collective impact on rating points. Similarly, this approach is used in Model V by including the interaction term of phonetic gender score and white wine with the reference category red wine into the model to capture the relationship of phonetic gender score and wine category, controlling for product quality. Finally, Model VI is conducted as a robustness check to confirm the validity, robustness, and significance of the previously estimated Models I–V. Using 31,058 observations and 18,609 different wines, we control in all models for wine quality by using country, province, and winery dummy fixed effects representing wine characteristics and winery reputation for separating the effects of phonetic gender scores on rating points (see Table 5).

Table 5.

Regression results

Robust standard errors are in parentheses.

*** p < 0.01, **p < 0.05, *p < 0.1.

Models I–VI reveal that the phonetic gender score significantly negatively impacts the average rating points. Model I demonstrates that a one-unit increase in the phonetic gender score decreases the average rating by –0.067 points. Accordingly, a one standard deviation increase in the phonetic gender score reduces the overall rating by –0.089 standard deviations. The results in Model VI indicate that increasing the phonetic gender score by one unit decreases the rating by −0.082 points on average, with a significance level of 1%. The effects remain stable across all estimated models. In conclusion, these results suggest that a higher femininity score in names appears to have a highly significant negative impact. In comparison, a higher masculinity score in names appears to have a highly significant positive impact on professional evaluations.

At the same time, Models II, III, IV, and VI expose that female reviewers score lower on average than their male counterparts. We control these models for wine quality and follow a stepwise approach by first including the phonetic gender score and reviewer’s gender (Model II), adding the wine category (Model III) before interaction terms are included (Models IV and VI). Model VI shows the most significant effect while female reviewers score on average −0.247 points lower than male reviewers. The results are highly significant across the models at the 1% significance level and remain robust across all estimated models. We include the wine category in Model III. We control for phonetic gender score, reviewer gender, and product quality, resulting in the wine category white wine receiving, on average −0.596 rating points less than red wines.

To test for a possible feminine name advantage of wine varieties based on the reviewer’s gender and wine color, we capture the relationship between female reviewers and the wine category with gender score and introduce interaction terms. We find evidence that suggests that some degree of interaction is essential in explaining variations in the ratings made by professional critics, especially between a feminine variety name and white wines. The interaction of phonetic gender score and white wine shows that an increase in phonetic gender score by one unit increases the rating by 0.042 points compared to the reference category of red wine. This result remains stable in Model VI. Heterogeneity analysis will reveal that this effect is driven by male reviewers, while feminine-sounding white wines are penalized by female reviewers.

Model IV shows that an increase of gender score by one unit increases the rating points given by female reviewers by about 0.032 points, respectively 0.022 points in Model VI in contrast to male reviewers. However, the results are statistically insignificant. Therefore, we estimate separate regressions for men and women to analyze heterogeneity better. By isolating the effects, we are able to test whether the effects differ significantly by gender and to test the identity hypothesis more effectively (see Table 6).

Table 6.

Separate regressions for female and male reviewers

Robust standard errors are in parentheses.

*** p < 0.01, **p < 0.05.

Including the interaction between phonetic gender score and white wine absorbs a large part of the effect of phonetic gender score, showing that female reviewers penalize feminine-sounding wines only when they are white. Instead, male reviewers penalize feminine-sounding wines only when they are red (the sum of −0.097 and 0.126 gives approximately 0 for white wines).

Heterogeneity analysis shows that this effect is driven by white wines for female reviewers and by red wines for male reviewers. The interpretation of these effects requires further investigation. One possible explanation is that each gender penalizes the femininity of the wine they know best—provided that female reviewers know better white wines and male reviewers know better red wines.

V. Discussion of results

Previous literature has examined names, linguistic properties, and product qualities imparted by inherent phonetic structures (e.g., Pogacar et al., Reference Pogacar, Plant, Rosulek and Kouril2015). In this context, the concepts of phonetic symbolism (e.g., Sapir, Reference Sapir1929), which ascribes specific (product) properties to phonetic patterns, and of names associated with gender and preferences are phenomena that have been investigated in mainly experimental studies so far (e.g., Klink, Reference Klink2000, Reference Klink2001, Reference Klink2009). Names are one of the first features consumers see or hear and are able to transmit information, meaning, trust, and reputation (Aaker and Keller, Reference Aaker and Keller1990).

However, little is known about how intrinsic product variety and phonetic gender effects affect professional evaluations, rankings, and ratings. These professional ratings are often determined by blind tastings to guarantee unbiased evaluations, where professional tasters are only provided with some information, such as the grape variety of the wine (e.g., Wang and Prešern, Reference Wang and Prešern2018). Particularly in the wine industry, wine tastings and professional ratings serve as trustworthy quality signals that consumers and prospects rely on (e.g., Storchmann, Reference Storchmann2012). To educate themselves, novice or infrequent wine drinkers may rely on professional rating scores (Yang et al., Reference Yang, McCluskey and Ross2009).

Wine differs depending on several factors. Some properties, like color or grading, are simple and affordable to measure. Others, like sensory or taste attributes, are challenging to quantify before consumption. The qualitative attributes of a wine are supposedly summed up by expert opinions (see Cardebat et al., Reference Cardebat, Figuet and Paroissien2014). Thus, the degree to which a person appreciates a wine depends on their expectation of quality, which is determined by a well-known label, the grape variety, or a wine critic’s ratings (Postman, Reference Postman2010). According to Friberg and Grönqvist (Reference Friberg and Grönqvist2012), a favorable review can result in a 6% rise in demand the week after publication. Kaimann et al. (Reference Kaimann, Spiess Bru and Frick2023) find that prices and product ratings are significantly related and that review consistency evolves and even determines current evaluations. Combris et al. (Reference Combris, Lecocq and Visser1997) state that objective cues (such as vintage and expert rating scores) significantly impact wine prices. However, sensory variables like tannin content and other quantifiable chemicals have not.

Therefore, ratings and tastings may show some limitations. The distinction between two wines may be so marginal that not even the most seasoned reviewer and taster could tell them apart (Barberà et al., Reference Barberà, Bossert and Moreno-Ternero2023). A similar scenario may occur when the wines are compared to a third wine. Postman (Reference Postman2010) spotlights that the results might be skewed simply by the wines’ presentational sequence. The personality of wine number four may easily moderate one’s enjoyment of wine number five. Wines also alter in the glass. For starters, even if the temperature was ideal when they were poured initially, it will quickly warm up to room temperature. Sensory adaptation, the process by which the strength of incoming stimuli is lowered with exposure over time, may mute other facets of the wine’s flavor. As a result, the wine that is tasted last will taste very differently from the wine that was tasted first. The results’ repeatability is another problem with blind tasting. Hodgson (Reference Hodgson2008) found that the judges could not duplicate their results when served the exact wine numerous times.

Despite these limitations, we are confident to shed light on how gendered language, specifically masculine or feminine names, affects perceptions and evaluations. On average, products indicating masculine gender-scored names tend to be better rated compared to feminine gender-scored names. Since genders may be drawn differently to masculine or feminine names, we have presumed that the taster’s gender is another factor that needs further analysis to understand what influences tastings and ratings. Spielmann et al. (Reference Spielmann, Dobscha and Lowrey2021) show how different brand associations lead men and women to focus on gendered representations differently. In particular, men experience a brand relationship with brands that contain masculine representations, even for neutral products such as champagne, board games, or potato chips. Women do not relate to gendered brands differently from men. In our study, female reviewers give lower evaluations than their male counterparts, indicating gender differences in review behavior. Our results imply that products with feminine name scores are likelier to be rated lower by both women and men. This effect especially occurs for white wines for female tasters and for red wines for male tasters. These results emphasize the significance of considering gender bias when tasting and rating wines, and they point to the necessity to clarify the relationship between gender behavior and wine preferences.

VI. Conclusion

This study examines the relationship between gender characteristics and the linguistic cues of wine varietal names, and how this affects the quality judgments of experts who evaluate and rate wines. We find a significant phonetic gender difference in wine ratings, indicating that feminine names receive lower ratings than masculine names. A phonological gender difference is revealed by heterogeneity analysis based on reviewer gender, with female reviewers giving lower ratings for white wines and male reviewers giving lower ratings for red wines. This implies that the type of wine that people are more likely to recognize or enjoy—red for men and white for women—has a phonological penalty that is applied by both genders. We also observe that, on average, women rate wines lower than men.

As a result, the study reveals a gender bias in blind tasting evaluations and provides insightful information for wine critics, enthusiasts, and producers, as well as useful advice for industry participants. It helps companies negotiate consumer preferences and improve wine quality and reputation by providing insight into the unconscious processes of wine tasting. The study promotes a more sophisticated and informed conversation about wine evaluation by encouraging a reevaluation of the variables that influence expert judgments.