II. Literature review

Studies on wine demand have broadly focused on generation differences (Thach and Olsen, Reference Thach and Olsen2006), marketing strategy (Thach, Reference Thach2009), local wine (Kolyesnikova, Dodd, and Duhan, Reference Kolyesnikova, Dodd and Duhan2008; Woods et al., Reference Woods, Deng, Nogueira and Yang2015; Farris et al., Reference Farris, Malone, Robison and Rothwell2019), behavior dynamics and sensory preferences (Bruwer, Saliba, and Miller, Reference Bruwer, Saliba and Miller2011), wine consumption and preference (Hussain, Cholette, and Castaldi, Reference Hussain, Cholette and Castaldi2007; Stanco, Lerro, and Marotta, Reference Stanco, Lerro and Marotta2020; Gustavsen and Rickertsen, Reference Gustavsen and Rickertsen2020), wine labels (Loureiro, Reference Loureiro2003; Mueller et al., Reference Mueller, Lockshin, Saltman and Blanford2010; Eustice, McCole, and Rutty, Reference Eustice, McCole and Rutty2019), wine knowledge (Gustafson, Lybbert, and Sumner, Reference Gustafson, Lybbert and Sumner2016), as well as health benefits of wine (Yoo et al., Reference Yoo, Saliba, MacDonald, Prenzler and Ryan2013). These studies emphasize wine itself but do not mention much about the role of non-wine products in the context of direct purchases from wineries. Complementary non-wine products are often additionally offered by wineries and can include food products, vineyard tours, merchandise in wineries, and wine festivals. Some research has highlighted the importance of other factors production addition to wine that can influence the visitor’s purchase motivation and decision, including engagement with regions, tourist preference, cellar visits, festivals and events, and societal stability, in sustaining the business and increasing future patronage (Gaetjens, Corsi, and Plewa, Reference Gaetjens, Corsi and Plewa2023; Gómez, Pratt, and Molina, Reference Gómez, Pratt and Molina2019; Mitchell, Hall, and McIntosh, Reference Mitchell, Hall, McIntosh, Hall, Sharples, Cambourne and Macionis2000; Gergaud, Livat, and Song, Reference Gergaud, Livat and Song2018). Wineries frequently provide tasting events and other wine promotions to attract visitors. Understanding the scale and determinants of non-wine purchases during winery visits showcases the potential for owners, marketers, and managers to promote business growth in wine hospitality.

III. Data and empirical model

This study focuses on winery consumers in selected Northern Appalachian states, including Pennsylvania, Ohio, Kentucky, and Tennessee. All respondents were required to be 21 years old. A total of 1,609 wine consumers completed a survey of wine-related purchase experiences in September 2012. This dataset is the same as in Woods et al. (Reference Woods, Deng, Nogueira and Yang2015). The sampling method was managed by SurveyMonkey, Inc. Respondents self-identified as wine drinkers. This dataset uniquely explores both wine and non-wine expenditures, presenting an opportunity for better understanding their determinants with a view toward strategic merchandising. A limitation of these data is that they are self-reported purchase activities based on wine consumption and winery visit recall rather than winery intercept sales.Footnote ¹ The analysis, however, provides insight into important purchasing patterns from venues where these data may be otherwise difficult to gather.

Following the market segmentation model adapted from the Hartman Organic Lifestyle Shopper Study 2000 (Hartman Group, 2000) and the framing of Wells and Haglock (Reference Wells and Haglock2008), who segmented consumers of health and sustainable foods, wine consumers are segmented into Core (purchased wine at least once per week), Mid-level (at least once per month), and Periphery (at least once per year). Wine consumption frequency, local wine expenditure, winery purchase activity, and knowledge can then be analyzed by segment. A similar segmentation model is currently used by the Wine Market Council (McMillan, Reference McMillan2023). A random utility theory with interval regression models is to elicit the estimated consumer spending (ECS) for local wine monthly purchasesFootnote ² and non-wine productsFootnote ³ in a winery visit. There are 24 independent variables used to explain the monthly average local wine ECS and non-wine product ECS in a winery visit. In order to decrease the hypothetical bias, the true ECS is assumed and can be observed by the latent variable $y_i^ * $. The model can be set as Equation (1):

(1)\begin{equation}y_i^ * = x_i^\prime \beta + {u_{i\;}}{\textrm{and}}\;{{\textrm{y}}^ * }\left| {x \sim Normal\left( {{x^\prime }\beta ,{\sigma ^2}} \right)} \right.\end{equation}

where y_i = 1 presents the range of ECS that is chosen by respondents, x_i represents the independent variables including social-demographic, consumer background, and wine preference, β exhibits the coefficient of the variable, u_i represents the error term, and the normal distribution is assumed in the interval regression. The empirical models of monthly average local wine ECS and non-wine product ECS are as follows:

(2)\begin{equation}Local\;Wine = y_{LW}^ * = {\alpha _0} + {\alpha _1}{X_1} + {\alpha _2}{X_2} + \cdot \cdot \cdot + {\alpha _{24}}{X_{24}} + \varepsilon \end{equation}

(3)\begin{equation}Non\_Wine\_Products = y_{NWP}^ * = {\beta _0} + {\beta _1}{X_1} + {\beta _2}{X_2} + \cdot \cdot \cdot + {\beta _{24}}{X_{24}} + \varepsilon \end{equation}

Thus, the ECS differences between local wine and non-wine products can be a potential indicator to winery owners of the relative magnitude and importance of the non-wine product business. A correlation of independent variables is performed and presented in Table 1. Most variables have low correlation, suggesting less concern for multicollinearity.

Table 1. Correlation matrix for independent variables

IV. Empirical resultsFootnote ⁴

Wine consumers in different consumption frequency classes are expected to behave differently with respect to non-wine purchase behavior during a winery visit. In order to define the wine consumer via the market segmentation model, three consumer groups, that is, core consumers, mid-level consumers, and periphery consumers, are identified based on the frequency of their wine purchasing in a year. Results in Figure 1 show that the core consumers (about 12.1% of total respondents in the region) drink wine more than 52 times in a year; mid-level consumers (about 45.5%) roughly drink wine about 12 to 52 times in a year; and periphery consumers (about 42.4%) drink wine less than 12 times in a year. In other words, more than half of consumers in the region at least drink wine once per month.

Figure 1. The definition of wine consumers based on the frequency of wine consumption.

The spending between local wine and non-wine products is further compared based on the market segmentation model. Figure 2 shows that core consumers on a monthly average spent about $69.87 for local wine, which is about two times higher than the overall monthly average of $34.62. Meanwhile, core consumers spent, on average, about $44.16 on non-wine products at their last winery visit. Recognizing the nominal differences in wine and non-wine products across segments, it is helpful to explore ECS potential determinants to better understand marginal effects based on the model specification.

Figure 2. The spending comparison between local wine and non-wine products.

The definitions and sample statistics of variables are presented in Table 2. Only a partial share of wine consumers from the region (n = 627) reported buying local wine from all retail sources at an average of $34.62 monthly. Of those respondents who indicated having visited a local winery within the past 12 months (n = 712), they reported purchasing an average of $25.91 for non-wine products in their previous visit. These two groups are not fully identical since not all respondents who have spent on non-wine products have purchased local wine before. Most respondents in this study overall (all wine consumers in the region) are female (about 69%), and the average age of respondents is about 52 years old. Most respondents are white. The annual average income of respondents is $67,340. Roughly 63% of respondents are urban residents. About 76% of respondents watch a food channel. Respondents indicated that, on average, they visited a local winery about 1.26 times in the past three years. Average bottle prices purchased indicate that respondents most frequently purchase in the Super ($7–$14/bottle) wine category (71%). Among the types of wine, most respondents (52%) buy red wine. In terms of sugar content (dry/sweet), respondents prefer dry and sweet approximately equally.

Table 2. Definitions and sample statistics of independent variables (n = 1,609)

The ECS for local wine and non-wine products interval regression model is estimated and presented in Table 3. Results show that these two models received valid outcomes from the Likelihood Ratio (LR) χ² test. The estimated parameters in the interval regression model reflect the actual value of spending in U.S. dollars. Regarding the monthly average ECS of local wine, respondents who are from Pennsylvania, have more wine drinkers in a household, represent core and mid-level wine consumers, are wine experts, more frequently visit local wineries, prefer to buy Luxury wine, and prefer more sweet wine are more likely to report a higher average monthly spending for local wine. Interestingly, respondents who self-rated themselves as wine experts (i.e., above average and expert level) have significantly higher local wine spending compared to those who report a lower wine knowledge level in the region. Johnson and Bastian (Reference Johnson and Bastian2007) also point out that wine knowledge is an important expenditure indicator for wine generally. This study extends this outcome, suggesting that consumers with higher wine knowledge spend more specifically on local wine.

Table 3. The ECS for local wine and non-wine products

Note: Asterisks indicate levels of significance:

* = 0.10, ** = 0.05, and *** = 0.01.

The ECS for non-wine products uses similar determinants to explore marginal effect but points to different spending relationships. Male respondents with higher income, respondents that have kids at home, are from Pennsylvania and Ohio, are from an urban area, include more wine drinkers in a household, are core consumers, those who watch food channels, identify as wine experts, more frequently visited local wineries, preferred to buy Popular ($4–7/bottle) and Luxury (>$25/bottle) sparkling wine, and preferred more sweet wine are more likely to spend more money at wineries for non-wine products. It is interesting to see that male consumers’ spending on non-wine products is positive, especially where it is not significant in local wine purchases. Other variables, that is, income, have kids at home, urban, food channels, Popular ($4–7/bottle), and sparkling, are also important for the ECS of non-wine products and reflect different impacts on non-wine purchases compared to local wine. These characteristics identify a distinct consumer group, indicating a positive tendency toward non-wine products, and would justify a potentially different approach to the marketing of these products.

These non-wine purchases provide a strong indication that there are heterogeneous preferences around both local wine and non-wine purchase activities that need to be considered for wineries. The ECS of non-wine products is difficult to elicit since most respondents can remember how much they spent on non-wine products in their previous visit rather than their monthly or yearly total spending. Although the aggregated ECS of non-wine products can be calculated in this study, the $76.19 should be used with caution. It implies that preferences and spending are likely to be highly heterogeneous, depending on the visitors. There may likely be helpful corresponding marketing strategies that could subsequently be effective in raising non-wine spending.