Food is considered an essential part of a culture, and the food preferences or evaluations of tourists are shaped by their cultural background (e.g., Jia, 2020). Food consumption is not just to satisfy our physiological needs but also to social and psychological needs. Dining provides opportunities for individuals to socialize, and to develop a sense of cultural identification, which is also a source for developing cultural links between different countries (Jia, 2020).

The study of Hofstede (1980) initially proposes four cultural dimensions in the cultural framework: “Power Distance,” “Uncertainty Avoidance,” “Individualism-Collectivism,” and “Masculinity-Femininity.” The work of Hofstede (2001) later adds the Confucian dynamic of “Long-term versus Short-term Orientation” and combines the dimension of “Indulgence versus Restraint.” As argued by Hofstede (2001), these cultural dimensions are the shared assumptions among members of a particular cultural group, and people from different cultural backgrounds will have different views about what is appropriate in various situations. Our work focuses on the two dimensions of Uncertainty Avoidance and Individualism-Collectivism.

Uncertainty Avoidance refers to the extent to which uncertainty is undesirable and people attempt to avoid it (Hofstede, 2001). Members of high-UAI cultures feel threatened with uncertain or unknown environments and have a strong need for clarity and security. Formality, written rules, or unwritten social norms provide certainty and a sense of security (De Mooij and Hofstede, 2002). Compared with people from low-UAI cultures, they are less open to new experiences and appear to be conservative. In the context of gastronomic consumption, high-UAI consumers are less likely to try new cuisines, have a limited range of food choices (Tse and Crotts, 2005). However, the research of Seo et al. (2012) find that foreign expatriates in South Korea who are from high-UAI cultures tend to perceive Korean foods as expensive and dine out less often than those from low ones, and they do not find significant differences between the two groups in attitudes, understanding, preference, satisfaction, and behavioral intentions with regards to Korean food or restaurants.

Individualism-Collectivism refers to the degree to which the welfare of the individual is valued more than the group (Hofstede, 2001). Members of the individualistic society are “me-oriented”). Their personal interests are seen as more important than group interests. They see themselves as independent from each other. The social networks in individualistic societies are rather loose. Members value their personal belief and their behaviors are guided more by their own interests and preferences than group ones (Oyserman et al., 2002). In contrast, members of collectivist society are “we-oriented,” they value the interests and social norms of their group. The social networks are fairly tight in a collective society, collectivists value the opportunity of interpersonal interaction and group membership is an important aspect of their identity (Gilboa et al., 2020). In addition to themselves and immediate families, collectivists also care about friends and other members of their group. Their behaviors are usually guided by group expectations and social norms (Torelli and Shavitt, 2010; Cao et al., 2021; Zhou et al., 2021). Collectivistic cultures emphasize the positive emotional value of culinary consumption on social occasions as a driving force of negotiating and fostering social relationships, bridging communication, and reinforcing a sense of belonging (Kim et al., 2020). Specifically, dining out together when traveling abroad can deepen interpersonal affection (Kim and Eves, 2012), and to the benefit of building new social relations and strengthening social bonds (Mak et al., 2012).

Overall, several studies have compared Western tourists and Asian tourists and suggested that Asians are less willing to challenge novel and unfamiliar cuisines while traveling abroad. The study of Cohen and Avieli (2004) indicate that Asians tend to be less novelty-seeking and more dependent on restaurants that provide their home cuisines. Japanese tourists are less inclined to culinary novelty-seeking when compared with American and Canadian tourists (Sheldon and Fox, 1988), and Koreans have a strong preference for their national cuisines (March, 1997). Within the western cultures, French customers (high-UAI) are more accepting of recommendations from waiters than those from English-speaking countries who (low-UAI) prefer to make consumption decisions by themselves (Cohen et al., 2009). Nevertheless, there are variations among Asian or Western cultures and food consumption behavior.

The cultural framework created by Hofstede has been widely applied in various studies of international business and management to explain a range of phenomena, including the difference between countries/regions at national, organizational, and individual levels. From the micro perspective, most studies focused exclusively on UAI and IDV, two dimensions that have become widely accepted in different fields of studies and have generated stable and reliable results at individual outcomes (Sekiguchi, 2006; Shavitt and Barnes, 2020; Watts et al., 2020). In food tourism literature, Jia (2020) found a link between individualism/collectivism and restaurant evaluation as U.S. tourists (individualism) tend to focus on occasion-related topics, while Chinese tourists (collectivism) tend to discuss more food-related topics. From the macro perspective, a series of reviews of the literature suggests that “Power Distance” and “Masculinity-Femininity” have theoretical relevance for the empirical research at the organization or/and country levels (Kirkman et al., 2006, p. 310, 2017), and as this study focuses on the individual level, these two dimensions are not examined in this study. Additionally, “Indulgence versus Restraint” is theoretically relevant at the organization level as well and relatively less applied to empirical studies, and “Long-term versus Short-term Orientation” is most relevant to Asian cultures (Dimitratos et al., 2011), not suited for studies focusing on multinational tourists. Therefore, these two dimensions are not examined in this study, either.

The presence of a travel companion can impact tourist behavior and decision-making in a foreign unfamiliar destination, travel companion can assist in decision-making through providing information and mental support, which reduces risk perception and enhance satisfaction. The travel companion might be friends or peers, family members, and colleagues, etc. Significantly research had examined the differences in behaviors when the consumer is alone versus with someone. In the Marketing literature, the impact of shopping companions has been well examined, including the influence of spending (Yim et al., 2014), shopping experience (Li X. et al., 2021). Additionally, the study of Borges et al. (2010) identified the type of shopping companion significantly impacts on hedonic shopping value, shoppers express more positive emotional effect and hedonic shopping motivation of accompanied by friends in relation to either shopping with family members or alone. In the context of tourism, travel companions have been repeatedly examined as a control variable (Li and Katsumata, 2020). Little research qualifies the mediating or moderating effects of travel companions (Su et al., 2020). More specifically, the outcome of differences among the type of travel companions has been neglected in the tourism literature. Thus, this research questions the moderating effects of travel companions on the relationship between national culture and culinary consumption by qualifying the type of person tourist travels with, e.g., partner vs. family vs. friend vs. colleague.

The breathtaking growth of social media has attracted increasing attention from researchers in marketing and tourism (Jia, 2020; Sun, 2021; Sun et al., 2021). Tourism scholars focus on three fields regarding the content: “What is the mining”; “who post the content”; and “why do they post?” (Leung et al., 2019). Due to the nature of our study, we explore the “who posts the reviews” (what cultures those tourists are from) and “what are the reviews” (various experiences of tourists). Previous studies have proposed multiple frameworks for these two tasks. For example, the study of Vu et al. (2019) utilize restaurant reviews to analyze tourist dining preferences. The recent development of machine learning and artificial intelligence models enables researchers to deal with the limitations of traditional methods. Various Natural language processing models are available as well, such as sentiment analysis (Alaei et al., 2019) and topic modeling (Jia, 2020). However, the results generated from machine learning models are at an aggregated level, losing the fine details at the individual, micro-level, which are valuable to generate in-depth knowledge. Moreover, although the sample sizes of social media contents are normally much larger than those collected from traditional research methods, yet the representativeness of the samples is limited to those who posted the comments, while those who did not are missed out. Therefore, our research integrates both traditional and machine learning methods to complement each other.

A total of 36,091 pieces of User-Generated Content commenting on restaurants in Osaka were retrieved by applying ScrapeStorm (2019), a crawler software. Each review contained restaurant names, cuisines, review comments, user location, whereas topics are not relevant to the restaurant and culinary services were removed, and reviewers whose places of origin are Japan are further removed. After the data cleaning, 7,803 reviews remained which were posted by tourists from 86 countries. The data was divided according to the places of origin of the reviewers into four cultural groups using the two dimensions of UAI and IDV. More specifically, the median of the cultural score calculated from 86 countries in this study, if the assigned score of a country is higher than the median of the assigned culture score, we categorize the country as high-UAI (IDV) group. Table 2 shows the locations and proportion of reviewers, and Table 3 shows the data distribution of these four groups.

The unstructured review data were transformed to generate the corpus and vocab by performing a series of Natural Language Processing procedures as follows. First, we only extracted reviews posted in English as the hierarchical Latent Dirichlet Allocation algorithm could only process English. This step was followed by using the text tokenization algorithm in the Natural Language Processing package (Loper and Bird, 2002), in which the reviews were split into words called tokens. An algorithm was applied to convert all tokens to lower case and the PortStemmer module in the package converted those tokens into root form. For example, “service,” “servicing,” “services,” and “serviced” mean the same thing, accordingly, stemming those words into “service” can improve the performance of the topic modeling algorithm. According to Vu et al. (2019), English noun words normally refer to interested entities, and therefore only nouns have been retrieved from the stemmed tokens. Further, the stop words were removed, which are normally in English sentences, whereas most of the time are meaningless for User-Generated Content analysis, such as “will,” “you,” “I,” etc. The Natural Language Toolkit package (Loper and Bird, 2002) collected usual stop words and we added some extra words into the list, such as the name of the restaurant, “place” and “restaurant.” After that, the tokens were loaded to the corpus and vocab variables and utilized in the topic modeling.

In this study, hierarchical Latent Dirichlet Allocation was adopted to explore User-Generated Content commenting on restaurants in Osaka by the international tourists who reside in different outbound markets. Hierarchical Latent Dirichlet Allocation is a non-parametric extension of Latent Dirichlet Allocation developed by Blei et al. (2010). It is an unsupervised topic modeling algorithm to cluster natural languages into various topics, such as reviews or comments. Hierarchical Latent Dirichlet Allocation has been commonly used in the marketing disciplinary (Toubia et al., 2019; Jia, 2020) and with two advantages compared with the traditional one: (a) instead of manually providing the topic number K in Latent Dirichlet Allocation algorithm, hierarchical Latent Dirichlet Allocation can generate topics automatically without given any topic number parameter; and (b) hierarchical Latent Dirichlet Allocation algorithm can present the topics in a hierarchical structure. In this study, we used the hierarchical Latent Dirichlet Allocation package written by Reades (2019) in Python 3. The final results were then transformed and visualized in the work of Flourish (2020).

By adopting scored cultural dimensions of Hofstede (2015) based on the home cultures of reviewers, we utilized the algorithm to classify the reviews in four different cultural groups: high/low-UAI; high/low-IDV. The distribution of the User-Generated Content is shown in Figures 2A,B.

The data for Study 2 was based on two secondary sources, one data was from the original work of Hofstede (2015), and the other one was obtained from the Osaka Convention and Tourism Bureau covering for 2 years. Data for the primary independent variables, the cultural dimensions of UAI and IDV were adopted from the original work of Hofstede (2015), the survey respondents were assigned the score based on their place of origin. Among those 112 centuries and regions, the score of UAI ranged from 8 the lowest (Singapore) to 112 the highest (Greece).

Data for other variables were obtained from the Osaka Convention and Tourism Bureau (OCTB) who organized seasonal surveys that were administered to international tourists from over 60 countries, from 2016 to 2018 in the international departure hall of Kansai International Airport. The survey was originally written in Japanese and was then translated to five languages by experts (English, Japanese, Simplified Chinese, Traditional Chinese, and Korean), and conducted during eight periods as shown in Table 4. The survey consists of 37 questions, including social demographics (age, gender, and nationality), travel attributes (packaged tour or not, travel experience, length of stay in Osaka, and travel companions), and travel activities (purchased items, culinary consumption, sightseeing spots, and entertainment activities). The self-administered seasonal questionnaire was administered to 9,141 random samples of tourists from different countries/regions to Osaka, Japan. Moreover, 194 samples were eliminated in the empirical analysis due to missing cuisine content or national information, remaining 8,947 samples. Additionally, OCTB also surveyed international tourists whether stayed and dined in Osaka. The culinary consumption behaviors of international tourists tend to be homogeneous, this is caused by the airport context-based dining is fewer selection options compared with center city. Therefore, samples that without staying in Osaka just departing from Kansai Airport were also eliminated in the analysis and finally remained 5,438 usable samples.

To concretely explore the culinary consumption of international tourists, a multiple-choice question with 23 choices encompassing all possible restaurant categories was designed. Those 23 types of restaurants include both Japanese cuisines and non-Japanese cuisines, Japanese cuisines mainly eat separately, while non-Japanese cuisines, such as Chinese cuisines often shared. Considering the type of food that may have influences on food sharing behaviors, to address this critical issue, we conducted the analysis by omitting non-Japanese cuisines to make sure the robustness of the results. Finally, based on the 16 restaurant categories, this study designed three dependent variables (Japanese daily food, Japanese luxury food, and the total number of cuisines consumed). The detailed information of cuisines is presented in the first column of Table 1. If at least one of the cuisines was selected, it was assumed that respondents had consumed the respective food and the value of “1” was assigned for consumed categories, otherwise “0.” Travel companions (partners, family, friends, and others) were treated as moderating variable strings. In addition, we define the composite variable to simplify the discussion. Let companion = 1 if a respondent travels with either partner, family, friends, and/or colleague (other).

Because packaged tour or not, travel experiences, consumption per day, and length of stay are expected to have different culinary consumption patterns and extent of culinary knowledge, we control for these effects a consumption per day (total spending/nights of stay in Japan); packaged tour or not (tour = 1; individually = 0); first/repeat trip to Osaka (first time = 1; repeat = 0); length of stay in Japan/Osaka (0–7 nights = 0–7, 8–13 nights = 10.5; 14 or more = 14). The control variables are also including the age, gender (woman = 1), satisfaction (very dissatisfied = 1; very satisfied = 7), and revisit intention (strongly disagree = 1; strongly agree = 5) of the respondents.

A total of 5,438 usable responses were obtained for data analysis, with 3703 excluded due to missing cuisine content or nationality information, or without staying in Osaka. Table 5 shows the summary statistics of the variables. In the sample, roughly 50% of tourists were from three Asian countries/regions, China (22%), South Korea (17%), and Taiwan (10%) followed by the U.S. (6%). Figure 8 presents the shares of nations/regions of international tourists to Osaka from 2016 to 2018.

Afterward, we applied the Poisson regression model to examine H6 and H7, and the main aim was to examine whether different types of travel companions (partner, family, friend, and others) have varied impacts on the relationship between collectivism and consumption of daily cuisines.

To test the hypotheses H6 and H7, we referred to the results of model M4-1 which was well fitted (lower AIC) for the dependent variables compared with M4-0. H6 was hypothesized that the social presence of a partner significantly moderates the relationship between collectivism and the range of culinary consumption (M4-1, β = –0.07, p < 0.001), and thus H6 was supported. Additionally, H7 was hypothesized that the presence of socially more distant travel companions significantly moderates the relationship between collectivism and the range of culinary consumption (M4-1, β = –0.110, p < 0.01), and thus H7 was supported. Results for the moderating effects of different types of travel companions are shown in Table 7. While the results of all hypotheses are presented in Table 8 and Figure 9.

Our research thus offers several important theoretical implications for advancing food tourism literature by revealing the cultural differences in tourist culinary exploration behaviors. Additionally, this study focuses on how social facilitation impacts national culture and the culinary consumption behaviors of tourists to shed light on how interpersonal factors impact the consumption decision-making of tourists.

First, our study uncovered that the national culture of UAI and IDV has an impact on multiple aspects of tourist culinary exploration, including the consumption of luxury versus daily cuisine, and the range of culinary consumption. Tourists from high-UAI cultures are more likely to consume luxury cuisines, given that they have a stronger need for clarity and security, and formal dining environments provide a sense of assurance (De Mooij and Hofstede, 2002). Tourists from low-UAI cultures are more risk-taking, enjoy adventure-seeking activities, and thus they are more likely to try novelty cuisines (Tse and Crotts, 2005). In contrast, tourists from high-UAI cultures tend to be risk-avoiding, and they tend to avoid unfamiliar cuisines, and specifically distant the street food and the food in night markets. Usually, compared with luxury cuisines, daily cuisines may less secure, especially raw dishes and/or cold dishes. Therefore, tourists from high-UAI cultures are more likely to consume Japanese luxury cuisines than daily cuisines. However, tourists from higher-UAI cultures tend to have a greater range of culinary consumption, which indicates that a high level of food security or high-end offering, such as luxury cuisines and/or high food quality tends to induce the culinary consumption of tourists. Our results contrast with the common assumption that tourists from high-UAI cultures have a smaller range of culinary consumption (Tse and Crotts, 2005), which is a Hong Kong-based context, Hong Kong cuisines were profoundly influenced by Chinese cuisines; thus, tourists from different regions (i.e., the United States, mainland China, Taiwan, and Singapore) come with different levels of prior knowledge on Hong Kong food, and the food culture of tourists greatly influences their food perception and evaluation of destination cuisines (Chang et al., 2011).

Second, the study results indicated that individualism/collectivism has a different influence pattern, in that tourists from collectivistic cultures showed a greater range of food consumption than those from individualist cultures. This finding can be explained by their tendency of collectivist preference to travel with family members, friends, and other travelers rather than alone, and can also be explained by their value for social bonds with travel companions. The results are in line with the collectivist value which emphasizes the importance of dining to maintain or improve relationships with others (Hui and Triandis, 1986).

Third, our findings identified that travel companion moderate the relationship between national culture and culinary consumption. Prior studies examined the novelty culinary-seeking behaviors of the international tourists (Tse and Crotts, 2005) or the uncertainty avoidance of immigrants on their attitudes toward local foods (Seo et al., 2012), while the literature on how cultural background affect the culinary consumption behaviors of tourists is relatively neglected. Our findings shed insight on those tourists from low-UAI cultures are more likely to consume a greater range of local daily cuisines (novelty culinary-seeking behaviors) while dining with travel companions. Additionally, in the presence of travel companions, tourists from collectivistic cultures will have a significant effect on the consumption of a wider range of culinary selections than tourists from individualistic cultures. The social facilitation of eating seems to be working in such a situation (Luo, 2005). Dining together helps to create an atmosphere of relaxation and a positive mood, and food becomes more palatable, leading to greater exploration of the local food (Kim and Huang, 2021). There is also a “time extension” effect when people eat together (De Castro et al., 1990), and social interaction reduces self-monitoring, thus tourists may engage more in the exploration of food (Bellisle and Dalix, 2001).

Fourth, our study contributed to the literature in identifying the moderating effects of different types of travel companions on the relationship between national culture and culinary consumption. The moderating effect of opposite partners and less familiar others are both positive and significant, while the two are distinct in terms of social distance. Firstly, the results of this study indicate tourists from collectivistic cultures show a greater range of culinary consumption when accompanied by opposite partners, the finding is in line with Townsend and Levy (1990), their findings heightened preference for opposite partners with resources is expected to be more attractive. Indeed, females perceived men as more desirable when they exhibit signals to resource access (Dunn and Searle, 2010). In the literature on consumer behavior, Cheng et al. (2013) reveal that an opposite-gender companion is more likely to induce impulse consumptions. Clauzel et al. (2019) report that the proportion of set menus ordered by a couple is mostly maximized, while it is minimized for those who take meals alone. Our study further revealed that those tourists from collectivistic cultures dining with socially more distant travel companions (such as colleagues and unfamiliar travelers) would increase the total number of culinary consumptions. Our findings are consistent with the existing literature of Luo (2005), which states that companions with far social distance less tend to come up with comparable concerns for long-term consequences, thus socially more distant peers may trigger impulsive consumption by satisfying hedonic goals and spontaneity. Our findings also connected with the findings of Jaffe et al. (2019) in that social distance positively and significantly influences the consumption of individuals regarding diversity.

Finally, our study further revealed a statistically significant relationship between travel experiences and culinary exploration behaviors, including positive relationships between travel experiences and novelty-seeking and variety-seeking behaviors on culinary exploration. This is consistent with the findings of Mak et al. (2012) that with more overseas travel experiences, tourists generally have a broader exposure to novel and foreign cuisines.

This study used both online and offline large and real international samples to conduct the cross-cultural comparison of the culinary consumption of tourists (e.g., luxury versus daily versus variety-seeking behaviors). The findings provided destination managers with a better understanding of the culinary preferences of international tourists and choice context and develop marketing communication strategies effectively. This study has several practical implications for destination marketing management.

First, the study findings suggested that luxury cuisines together with excellent service and environment should be leveraged to attract repeat tourists from high-UAI cultures. In the case of Japanese cuisines, tourism marketers should promote Kaiseki, Fugu, crab, and similar traditional local luxury cuisines to fascinate tourists from Belgium, Greece, Portugal, and other high-UAI scored travel markets.

Second, this study could provide practical implications for operators to extend travel companions to induce a greater range of culinary consumption. Our findings reveal that those tourists from collectivistic cultures are more likely to consume a greater range of cuisines when accompanied by partners and other companions (such as colleagues and unfamiliar travelers). Accordingly, marketers should promote local cuisines and food-related activities to tourists from collectivistic cultures based on the types of travel companions. For example, the recommendation of local alcohol or special cuisines to the international tourists accompanied by a couple and or colleagues would be desirable. Therefore, travel destinations that are rich in culinary resources should leverage those resources to attract tourists from collectivistic cultures, including China, South Korea, and Russia. Moreover, tourists from high-UAI cultures also increase the likelihood of daily cuisines in the context of dining with travel companions. Accordingly, increasing the trust perception of local cuisines is also expected to attract the patronization of tourists from risk-avoiding cultures.

Finally, repeat visitors showed a stronger patronization of Japanese daily cuisines than did first-time visitors. Accordingly, marketers should promote more Japanese daily foods to the highly repeated travel markets, including Hong Kong, Korea, Singapore, and Taiwan. More specifically, younger visitors have a clear tendency to enjoy diverse and unfamiliar cuisines. Therefore, this study provides valuable insights to policymakers to increase financial outcomes. Promoters could develop food travel packages based on travel companions, travel experiences, and age data.

Although the current study provided important contributions to applying food tourism as a means of promoting a destination and attracting tourists, there were still certain limitations that should be considered. First, the sample of Study 1 is limited to the User-Generated Content in English. Despite the Chinese tourists with the largest market shared in Japanese inbound tourists, the count of comments by Chinese tourists was much low compared with their proportions. Thus, further research using Chinese social media would provide greater insights. Second, this research focused solely on Osaka Japan, and, therefore, the conclusions might not be generalized. Future research should extend this model to more host countries and regions where food tourism is one of the leading tourism industries to determine the generalizability of the cross-national results. Third, this study was limited to identifying multinational cultures on the culinary consumption of inbound tourists from the macro level, and future research should explore the intrinsic psychological factors of individuals through the additional methodology. Fourth, the cultural distance might significantly impact the culinary knowledge and language familiarity of tourists, and, in turn, impact their culinary consumption, the survey did not cover the information on the host country language skills of respondents and their prior food knowledge of the various culinary offerings. Therefore, future research may include the language skills of the host country and prior food knowledge as controls in the research model. Finally, the design of the survey methodology also needed to be improved. We would be able to obtain more appropriate results by using the share of food consumption, rather than the binary question of “ate” or “did not eat” for luxury food and daily food, as in this study.