Uncertainty and Tourism Consumption Preferences: Evidence from the Representative Chinese City of Shenzhen

: As the fastest growing city in China, Shenzhen, a pioneer of reform and opening up, seems to have exhausted its advantages of geographical convenience, cheap land, and cheap labor; thus, the tourism industry has been adopted as a new economic strategy. However, the Shenzhen tourism market has received little attention from scholars and few studies have been done to investigate the relationships between uncertainty and tourism consumption preferences in different cultures. Therefore, this paper attempts to study Shenzhen as a travel destination to explore the consumption preferences of domestic and international tourists, specify uncertainties in tourism consumption activities, and examine their impacts on preferences. The above aims are achieved using a survey method and a new theoretically proposed preference uncertainty model inspired by combining the modiﬁed virtual expert preference approach and the ordered probit model (MVEP-OPM), in which three major components of tourism consumption (food, accommodation, and shopping) are estimated. (1) The results show that there are signiﬁcant differences in tourism consumption preferences between domestic tourists and international visitors. (2) Inexperience, unfamiliarity, imperfect knowledge, and policy uncertainty are major uncertainties in tourism consumption activities, which affect tourist preferences toward consumption behavior. (3) Uncertainty plays a different moderating role in product-related inﬂuential factors, such as sanitation, safety, and decoration.


Introduction
Tourism is becoming a promising industry due to its diverse potential, playing a decisive role in adjusting industrial structure, driving local consumption, stimulating economic growth, increasing revenue, and providing employment opportunities during the process of urban transformation and upgrading [1]. Therefore, the characteristics [2], attitudes [3], behaviors [4], preferences [5], perceptions [6], motivations [7], and satisfaction [8] in tourism consumption have become focuses of attention. Preferences and motivations sometimes overlap, but motivations play a guiding role rather than a decisive role in consumer satisfaction, while preferences refer to a clear attitude towards making choices after detailed comparisons of various products with similar attributes and functions [9], usually accompanied by uncertainties. In other words, uncertainties exist in tourism consumption preferences [10]; for example, when a tourist books accommodation, they are buying an intangible service for which an evaluation has to be made after it has been experienced. Consequently, various methods and calibration techniques have been proposed to measure uncertainty, including: (1) the numerical certainty scale (NCS), using a follow-up question to elicit the degree of respondent uncertainty on a probabilistic scale ranging from 0 to 100 percent [11] or on an ordinal scale ranging from 1 (very uncertain) to 10 (very certain) [12]; (2) the polychotomous choice (PC) approach, introducing a postevaluation question by providing response categories, such as "definitely yes", "probably yes", "maybe yes", "maybe no", "probably no" and "definitely no", reflecting respondents' coastlines, charming beaches, a pleasant climate, a high-tech image, civilized behavior, and metropolitan style. As early as 2010, Shenzhen was rated as one of "31 destinations for global tourists" by the New York Times. In 2019, Shenzhen ranked second in the list of the top 10 best travel cities published by Lonely Planet. Despite frequent appearances on global travel lists, few scholars have focused on the Shenzhen tourism market.

Tourism Consumption Preferences
According to the 2018 China Statistical Yearbook, food ($10.31 billion, 8.4%), accommodation ($12.21 billion, 9.9%), and shopping ($22.99 billion, 18.6%) are major tourism consumption expenditures for international visitors. Therefore, research on tourism consumption preferences relating to uncertainty will focus on food, accommodation, and shopping.

Food
Local food provides social and leisure attributes to tourism and cultural activities, which are driving forces for tourism consumption and are often described by preference, liking, and choice. Food preference refers to the most satisfying choice within the option range, while food liking is the cause of preference and food choice is a way to express preference [22]. Due to the importance of food preference, its influencing factors have been extensively explored and summarized at the culture level in terms of individual preference, food-related, environment, and experience components [23,24]. The impacts of culture and religion on consumer food preferences are deeply rooted; for example, Muslims are forbidden from eating pork and Western food culture differs from Chinese dining etiquette. For individuals, sociodemographic factors such as age, gender, marital status, and household income are the basis for the diversity of preferences [25]. Flavor, variety, luster, sensory appeal, and other attributes of food can stimulate and delight visitors' senses of smell, taste, and sight [26]. The dining environment is closely related to internal conditions (service level, price, hygiene, and atmosphere) [27] and the external atmosphere (location, evaluation, and recommendations). Regarding the experience, this is complex due to subjective judgments involving health, availability, and convenience, as well as objective factors related to food attributes [23]. Based on the above explanation, taste, variety, service quality, price, location, hygiene, and evaluation are selected as drivers of food tourism consumption preference in this study.

Accommodation
Accommodation is considered to be the main requirements for luxury, leisure, health, and business travel; for example, American tourists spend more than 30% of their total budget on accommodation [28]. Since hotel reservations take precedence over other activities, the selection criteria have been largely explored. For example, Prayag and Lee [7] examined the moderating effects of service interactions, nationality, and hotel star rating on the relationship between tourist motivation and place attachment, while Lai and Hitchcock [29] compared the service quality attributes for stand-alone and resort-based luxury hotels in Macau by surveying 299 overnight visitors. Furthermore, Sohrabi et al. [30] identified the key consumption drivers for hotels in Tehran, including security and protection, promenades, staff and their services, cleanliness and room comfort, networks, pleasure, room facilities, car parking, and expenditure. Additionally, Xu et al. [31] outlined the dimensions of housing, location, host-guest interaction, facilities, and overall sensations as contributing to accommodation sharing through a data-driven approach. One study Hsieh et al. [32] suggested that the distance to commercial districts and sightseeing locations was the most valued aspect by visitors when considering accommodation, while another study Chu and Choi [33] emphasized that both business and leisure travelers showed the same attitudes towards value, room and front desk, food, and recreation components. Furthermore, Rivers [34] highlighted that long-stay guests preferred hotels with memberships and promotions. In addition, Kim et al. [35] suggested that online marketing efforts should highlight a hotel's identity and accessibility. In view of the above, regarding accommodation tourism consumption, widely used factors are identified in this research, including safety, service, sanitation, price, location, and free breakfasts.

Shopping
Shopping can bring about memorable experiences for tourists due to providing relaxation and pleasure, especially for East Asian tourists, who were found to regard shopping as a pleasurable activity that enriches lives [36]. Duty-free stores and famous shopping destinations such as Hong Kong and Hawaii are preferred and attract crazy purchases. Factors influencing shopping have been constantly studied, for example women prefer buying clothes, shoes, jewelry, and cosmetics, while male visitors prefer tobacco, cigarettes, and wine [37]. Married tourists tend to spend more on shopping [2], but low-income groups usually consider their actual needs [38] and are attracted by lower prices. In addition, brands with a long history and reputation are more attractive, despite the higher prices [39]. Product quality provides users with functional or practical benefits, but the service level contributes to consumer satisfaction and commemorative significance contributes to joyful memories [40]. Furthermore, the in-store environment can guide tourists' shopping behavior. For example, color affects perceptions of products [41] and a cluttered environment brings negative emotions to consumers [42]. Store popularity, well-known shopping regions, thematic shopping, and celebrated shopping malls all increase shopping expenses [43]. In addition, promotions largely determine the attitudes of consumers [44]. Therefore, regarding shopping, brand, price, service, quality, packaging, memorable experiences, decorations, practical needs, store popularity, and promotions are selected herein due to their representative nature.

Uncertainty in Tourism Consumption
Uncertainty associated with travel is highlighted as reflecting tourists' concerns and consumption habits, especially for international tourists, for whom uncertainty issues are often amplified [20]. A previous study Quintal et al. [45] discussed the effects of uncertainty on tourism decisions and found that increased uncertainty in Australian visitors negatively affected their motivation to take vacations. To reduce uncertainty, official websites and well-known travel agencies are generally relied on rather than individuals when obtaining information about the travel destination, and tourists tend to travel in groups, picking shorter stays and fewer visits [19]. However, there is still no consensus on how to measure uncertainty. Regarding the NCS, the theoretical support is insufficient in terms of determining the appropriate cutoff points, because numerical probabilities and the levels of an ordinal scale can be interpreted differently by diverse individuals. As for the PC approach, it can be influenced by the words used to elicit respondent uncertainty [15], which might generate higher rates of "yes" responses. Therefore, a new method is proposed, whereby uncertainty is considered a latent variable defined by a set of influential factors and integrated into the virtual expert model [46] to replace the role of virtual experts. Before implementation, it is necessary to explore sources of uncertainty and their impacts on tourism consumption preferences.

Inexperience
Insufficient experience and infrequent purchases inevitably increase the level of uncertainty in tourism consumption [21]. Generally, travel experience has a positive effect on consumer well-being and loyalty, which is conducive to reducing the intertemporal instability of uncertainty preferences [47]. Two criteria were available to measure tourism experiences and familiarity in this study, namely the number of trips and the duration of the stay [48]. These two indicators reflect the degree to which a person utilizes opportunities to accrue experience through time, events, and activities, providing an accurate means to distinguish individuals with different levels of travel experience.

Imperfect Knowledge
Uncertainty entails a limitation of knowledge on the tourism destination and future conditions, making the outcome of an action impossible to predict. Imperfect knowledge partly stems from the cultural differences, daily habits, and lifestyles of tourists [49]. In other words, familiarity with diverse cultures and destinations is informed by tacit or personal knowledge, which is particularly true for tourists visiting unfamiliar places [20]. Consequently, imperfect knowledge may lead to relatively high uncertainty; the higher level of education, the lower the degree of knowledge imperfection.

Policy Change
Subjective policy uncertainty refers to the probability of a policy intervention being effective [50]. Additionally, respondents may not fully understand a proposed policy or the way it is to be implemented. Several preference studies have investigated the role of policy change in the willingness to pay for non-marketed goods [51], showing that policy uncertainty negatively influences individual valuations. Since the announcement of a pilot duty-free shopping policy for offshore tourists in Hainan, the Shenzhen government has been considering whether to follow suit, leading tourists to hesitate when shopping. Therefore, this paper examines the duty-free shopping policy to study the influence of policy changes on tourists' shopping preferences.
Despite increasing studies on tourism consumption covering a wide range of topics, there is still a gap in the knowledge in terms of the impact of uncertainty on tourism consumption preferences in the Shenzhen tourism market. Uncertainty is inherent and inevitable in tourism consumption activities, however the NCS and PC approaches fail to accurately estimate it. To bridge this knowledge gap, a new consumer preference uncertainty model is theoretically developed in this study, in which uncertainty is considered a latent variable defined by a set of influential factors. Moreover, summarizing the above literature, a conceptual framework illustrating the relationships among uncertainty (imperfect knowledge, inexperience, unfamiliarity, and policy change) and tourism consumption preferences (food, accommodation, and shopping) and corresponding influential factors is proposed ( Figure 1).

Model
The proposed model in this paper was inspired by the combination of the modified virtual expert preference approach and the ordered probit model (MVEP-OPM). The probit model is widely applied for ranked data, which decomposes the overall preferences provided by the respondents into factor-related utility weights [52]. The original virtual

Model
The proposed model in this paper was inspired by the combination of the modified virtual expert preference approach and the ordered probit model (MVEP-OPM). The probit model is widely applied for ranked data, which decomposes the overall preferences provided by the respondents into factor-related utility weights [52]. The original virtual expert preference (OVEP) approach was first developed by [46] as a general consumer preference model for experience products and overcomes the limitations of choice models, especially when significant information on non-quantifiable attributes is missing. Due to the heavy computational load, a modified VEP (MVEP) was proposed, in which uncertainties are used instead of virtual experts to estimate the unobservable product utility. In the MVEP model, Y ij represents the stated preference of consumer i for product j. R denotes the number of points on the ordered scale and r represents a specific response, which is modeled by postulating the existence of R − 1 thresholds for the latent random utilities (U ij ), as below: The latent product utility U ij is decomposed into the effects of certain (U ij ) and uncertain (U * ij ) components as: Here, U ij is measured by observed product attributes X and individual characteristics X * as: The intercepts α i0 and α 0j capture the main effects of consumer i and product j, respectively; X(ΨX * ) reveals the interaction effect of consumer characteristics and product attributes; Ψ is the corresponding parameter matrix, indicating the effects of observed individual characteristics on consumption preferences; ξ βi is the error term, accounting for unobserved heterogeneity across consumers.
Here, U * ij is estimated by representative uncertainty influencing factors in tourism research as: where Z is the uncertainty factor; α 3 , α 4 are corresponding parameters. If consumer i response choices are related to preference, the probability P ij is given by: where Φ(·) is the standard normal cumulative distribution function. The log-likelihood function is: The individual-specific cutoff point C ir accounts for scale usage heterogeneity, enabling us to cluster people based on preference similarity. The log-likelihood in Equation (6) is maximized with respect to the elements of α 0 , α 1 , α 2 , α 3 , α 4 along with the cutoff points C i1 , C i2 , . . . , C i,r−1 using an iterative procedure to give maximum likelihood estimates of both sets of parameters. The definitions of variables in the model are explained in Table 1. PartI: Food all independent variables 4 (friend/relative home) duration of stay X 15 (s_price) unfamiliarity 1 (less than 3 days)

Data Collection
The data were obtained through questionnaires (Chinese version, CV; English version, EV), in which 35 items (3 preferences, 4 uncertainties, 4 consumer characteristics, 24 product attributes) were designed and collected by asking tourists to assess their tourism consumption preferences in terms of food (partI), accommodation (parII), and shopping (partIII) at 14 passenger gathering places in Shenzhen, including East Gate Pedestrian Street, Splendid China, Window of the World, Lotus Hill, Shenzhen Airport, Overseas Chinese Town, Coastal City, Shenzhen North Railway Station, Shenzhen Library, Shenzhen University, Peking University Shenzhen Graduate School, Yitian Holiday Plaza, Shekou Sea World, and Overseas Chinese Town Contemporary Art Terminal. In designing the questionnaire, the survey was initially presented as the Chinese version (CV) and then translated into the English version (EV). The double back-translation method, double proofreading process, and a pilot test were completed prior to data collection. In addition, at the questionnaire collection site, for each target international tourist, the questionnaire was further interpreted by two independent professional translators to avoid ambiguity caused by language barriers due to different cultural backgrounds. Respondents were free to respond to three parts, since each part is independent.
Uncertainty and preference: Table 3 demonstrates the uncertainty, preference, and comparative analyses between domestic and international visitors. (1) Food: The degree of uncertainty for this component is generally at a moderate level and "local flavor" restaurants are preferred, since the mean values of uncertainty 1 , uncertainty 2 , uncertainty 3 , and food preference are 2.56, 2.16, 2.00, and 2.71, respectively. Influential factors such as flavor (4.14) and hygiene (4.26) are considered important. However, the analysis reveals significant mean differences between the two groups in the uncertainty category, since the degrees of uncertainty caused by imperfect knowledge (Mean DT = 2.19 vs. Mean IT

Food Consumption Preference Estimation Results
Model 1 shows the results for domestic tourists' food preferences. Uncertainty 1 (Coef. = −0.5697) and uncertainty 2 (Coef. = 0.2693) are significant at 0.1 and 0.05 respectively, indicating that domestic tourists prefer fast food when facing a higher level of uncertainty 1 (a lack of knowledge of indigenous food) and tend to seek local cuisine under lower degrees of uncertainty 2 when they have experience due to an increasing number of trips, because imperfect knowledge makes it difficult for tourists to ascertain whether their expectations can be satisfied when dining in a high-class restaurant. Both flavor (0.1314) and variety (0.3483) exhibit positive effects at the 0.1 significance level, indicating that domestic tourists tend to prefer "local flavor" restaurants, confirming the statistical analysis results shown in Table 3. As local cuisine can satisfy tourists' expectations for vision, taste, and smell, enjoying local food (night market delicacies and street snacks) is a sensory pleasure event and part of a holiday experience, especially for females (Coef. = −0.4670, p < 0.01) and single (Coef. = −0.7613, p < 0.05) tourists. However, married visitors dislike fast food, night markets, and street snacks, which can be attributed to health concerns, since the hygiene coefficient is −0.5775 at the 0.1 significance level; this is particularly true for family tours with children. In addition, an interaction between uncertainty 1 and hygiene was estimated and found to be significant (Coef. = 0.1327, p < 0.1), indicating that uncertainty 1 plays a regulatory role in terms of the impacts of hygiene on food consumption preferences. Figure 2a shows the regulation mechanism and reveals that the lower the degree of uncertainty 1 , the larger its effect on the impacts of hygiene on food preference.  Model 2 reveals the results for international tourists' food preferences. The coefficients of uncertainty2 and uncertainty3 are −0.2060 and −0.7917, respectively, which pass the 0.1 significance level test, indicating that local indigenous food loses its appeal to international tourists visiting Shenzhen with increases in the number of trips and duration of stay, differing from domestic tourists. Taking uncertainty3 as an example, as seen in Figure 3b, the second derivative of food preference with respect to uncertainty3 is positive and the inflection point of uncertainty3 is 2.63; that is, the duration of stay is about 11 days (see definitions in Table 2). Before the turning point, the dietary preferences of foreign tourist gradually shift from high-class restaurants to local snacks. However, when a tourist stays at a destination for more than 11 days, their food preference once again transfers to regular restaurants rather than local flavors or fast food. Compared with domestic tourists, in addition to variety (Coef. = 0.1823, p < 0.5) and sanitation (Coef. = −0.2452, p < 0.01) discussed in model 1, international tourists are still concerned about the location (Coef. = −0.2546, p < 0.05) and evaluation (Coef. = −0.1925, p < 0.05) of restaurants, as it is inconvenient and troublesome for them to look for delicacies due to language problems. Consequently, conveniently located and highly recommended restaurants are favored by international tourists. Model 2 reveals the results for international tourists' food preferences. The coefficients of uncertainty 2 and uncertainty 3 are −0.2060 and −0.7917, respectively, which pass the 0.1 significance level test, indicating that local indigenous food loses its appeal to international tourists visiting Shenzhen with increases in the number of trips and duration of stay, differing from domestic tourists. Taking uncertainty 3 as an example, as seen in Figure 3b, the second derivative of food preference with respect to uncertainty 3 is positive and the inflection point of uncertainty 3 is 2.63; that is, the duration of stay is about 11 days (see definitions in Table 2). Before the turning point, the dietary preferences of foreign tourist gradually shift from high-class restaurants to local snacks. However, when a tourist stays at a destination for more than 11 days, their food preference once again transfers to regular restaurants rather than local flavors or fast food. Compared with domestic tourists, in addition to variety (Coef. = 0.1823, p < 0.5) and sanitation (Coef. = −0.2452, p < 0.01) discussed in model 1, international tourists are still concerned about the location (Coef. = −0.2546, p < 0.05) and evaluation (Coef. = −0.1925, p < 0.05) of restaurants, as it is inconvenient and troublesome for them to look for delicacies due to language problems. Consequently, conveniently located and highly recommended restaurants are favored by international tourists.

Accommodation Consumption Preference Estimation Results
Model 3 shows the results for domestic tourists' accommodation preferences, showing negative significant effects of uncertainty1 (Coef. = −0.1895, p < 0.01), uncertainty2 (Coef. = −0.3126, p < 0.05), and uncertainty3 (Coef. = −0.3140, p < 0.01), revealing that a high degree of uncertainty leads to a tendency of domestic tourists to lodge in the homes of friends or relatives. However, when the uncertainty is reduced-that is, through more frequent trips and a longer duration of stay-domestic tourists like to find accommodation outside (hotel or motel) rather than overnight at a friend's home, because for most families in Shenzhen an extra bedroom is not available for long-stay guests due to expensive housing prices. In addition, both safety (Coef. = 0.2429, p < 0.1) and freebreakfast (Coef. = −0.1025, p < 0.1) show significant associations with accommodation preferences, indicating that safety-conscious tourists dislike staying outside, and that the more they value free breakfasts, the greater their preference for "star" hotels.
Model 4 shows the results for international tourist accommodation preferences. Uncertatinty2 is negatively significant at the level of 0.1, indicating that international visitors favor higher class hotels under high uncertainty conditions caused by inexperience. If uncertatinty3 (Coef. = −0.8032, p < 0.1) and its square item (Coef. = 0.1859, p < 0.05) are significant, an uncertatinty3 of 2.1 would result in a minimum value for the accommodation preference, as shown in Figure 3d. In other words, for travel within 7 days, accommodation preferences gradually decline from "star" hotels to youth hostels, but accommodation options will be reversed if the stay is extended. Regarding influential factors, safety (Coef. = 0.4333, p < 0.05) and its interaction with uncertainty2 (Coef. = −0.1770, p < 0.05) are significant, demonstrating that uncertainty2 regulates the effects of safety (see Figure 2b). Specifically, safety shows stronger and weaker impacts on accommodation preference, respectively, under high and low levels of uncertainty2. Moreover, unlike domestic tourists, international visitors are sensitive to sanitation (Coef. = −0.3441, p < 0.01)

Accommodation Consumption Preference Estimation Results
Model 3 shows the results for domestic tourists' accommodation preferences, showing negative significant effects of uncertainty 1 (Coef. = −0.1895, p < 0.01), uncertainty 2 (Coef. = −0.3126, p < 0.05), and uncertainty 3 (Coef. = −0.3140, p < 0.01), revealing that a high degree of uncertainty leads to a tendency of domestic tourists to lodge in the homes of friends or relatives. However, when the uncertainty is reduced-that is, through more frequent trips and a longer duration of stay-domestic tourists like to find accommodation outside (hotel or motel) rather than overnight at a friend's home, because for most families in Shenzhen an extra bedroom is not available for long-stay guests due to expensive housing prices. In addition, both safety (Coef. = 0.2429, p < 0.1) and free-breakfast (Coef. = −0.1025, p < 0.1) show significant associations with accommodation preferences, indicating that safety-conscious tourists dislike staying outside, and that the more they value free breakfasts, the greater their preference for "star" hotels.
Model 4 shows the results for international tourist accommodation preferences. Uncer tatinty 2 is negatively significant at the level of 0.1, indicating that international visitors favor higher class hotels under high uncertainty conditions caused by inexperience. If uncertatinty 3 (Coef. = −0.8032, p < 0.1) and its square item (Coef. = 0.1859, p < 0.05) are significant, an uncertatinty 3 of 2.1 would result in a minimum value for the accommodation preference, as shown in Figure 3d. In other words, for travel within 7 days, accommodation preferences gradually decline from "star" hotels to youth hostels, but accommodation options will be reversed if the stay is extended. Regarding influential factors, safety (Coef. = 0.4333, p < 0.05) and its interaction with uncertainty 2 (Coef. = −0.1770, p < 0.05) are significant, demonstrating that uncertainty 2 regulates the effects of safety (see Figure 2b). Specifically, safety shows stronger and weaker impacts on accommodation preference, respectively, under high and low levels of uncertainty 2 . Moreover, unlike domestic tourists, international visitors are sensitive to sanitation (Coef. = −0.3441, p < 0.01) and price (Coef. = 0.2726, p < 0.01) as they relate to accommodation; that is, the more they pay attention to sanitation, the greater their preference for superior accommodation, such as star hotels, while price-sensitive visitors show preference for cheap accommodation, such as hostels.

Shopping Consumption Preference Estimation Results
Model 5 provides the results for domestic tourists' shopping preferences. It shows that uncertainty 2 is negatively significant at the 0.01 level, as it is well known that the more frequently a person travels to a destination, the less willing they are to shop. However, if the duty-free policy were be implemented, tourists' shopping enthusiasm would be rekindled, as shown by the estimation results for uncertainty 4 (Coef. = 0.1358, p < 0.05), which show that a higher tax rebate rate would result in a larger proportion of shopping expenses. Although tourists with considerable income have the ability to pay, this process is subject to uncertainty 2 due to the significant interactions (Coef. = 0.2936, p < 0.01). Popular brands with great reputations attract and stimulate spending (Coef. = 0.1347, p < 0.1), while the price factor is no longer a focus for domestic tourists from different financial backgrounds, as they can afford their favorite brands. Portability is becoming a key factor and items that are easy to carry increase shopping consumption (Coef. = −0.2785, p < 0.01). Furthermore, the memorable coefficient is −0.4297 in model 5 and passes the significance level of 5%, while its interaction with gender is positive (Coef. = 0.2364, p < 0.05), showing that commemorative shopping preferences will undermine consumer spending, especially for female tourists.

Marginal Utilities
The marginal utility of preference is an important indicator that measures the sensitivity of preference in response to uncertainty and influential factors. The marginal utility of preferences is discussed only by taking shopping consumption as an example. For food and accommodation, the results are shown in Table 5. Considering that the value of shopping preference is 1, increasing uncertainty 2 by 1% would lead to a 31.16% increase in domestic tourist shopping expenditure, while also resulting in a 15.92% decrease for international visitors. Moreover, with a 1% increase in uncertainty 4 , the shopping expenditure of domestic and international tourists will likely decrease by 4.19% and 26.16%, respectively. Domestic and international visitors also exhibit different levels of sensitivity to the changes of influential factors, such as brand, packaging, carrying, memorable, decoration, and needs components, due to their significant marginal utilities. For example, domestic tourists are highly sensitive to the brand (−0.0416) component in the model, while the practical needs (−0.0990) component is more flexible in terms of expenditure for international visitors, further confirming that domestic tourists regard shopping as a necessity.  I  D  I  D  I  D  I  D  I  D  I  D  I  D  I  D  I  D  I  D  I  C  Note: n.a. indicates non-significant marginal effect; ***, **, * indicate significance at the 1%, 5%, and 10%, respectively; D and I represent domestic and international tourists.

Conclusions
In this study, we applied a new proposed MVEP-OPM model to explore the tourism consumption preferences under uncertainty conditions in Shenzhen. In the MVEP-OPM, the virtual expert preference approach was replaced by uncertainty to efficiently research the unobserved components of certain utilities, whereby uncertainty was considered a latent variable and measured using a set of influential factors. In addition, three key tourism activities, namely food, accommodation, and shopping, were examined using sample responses from domestic and international tourists, providing insights into the tourism consumption preferences in response to changes in uncertainty and influential factors.
Some of significant findings were the overall higher degree of uncertainty caused by imperfect knowledge, inexperience, unfamiliarity, and policy changes; the greater impacts of uncertainty on tourism consumption preferences; and the more obvious regulating effects on other influential factors. The empirical results also revealed different types of tourism consumption preferences in models 1 to 6. Regarding food preferences, generally both domestic and international tourists preferred "local flavor" restaurants; however, domestic tourists thought highly of flavor, while international tourists paid more attention to recommendations and location. Moreover, the lower the degree of uncertainty, the larger its regulating effect on food hygiene preferences (see Figure 2a). For accommodation preferences, sanitation and price were deemed as important influential factors by international visitors, but they were not significant for domestic tourists. With reference to shopping preferences, domestic tourists regarded shopping activities in Shenzhen as a necessity, while international tourists preferred accommodation experiences and their expenditure elasticity for shopping was much higher than for other product categories. Furthermore, domestic tourists were interested in famous brands and merchandise that was easy to carry, while international tourists were more concerned by packaging and in-store decorations. The moderating role of uncertainty on decoration was regulated by the degree of uncertainty caused by inexperience (see Figure 3c). Finally, regarding the differences observed in tourism consumption preferences above, these can be largely attributed to cultural background, income, marital status, gender, and the degree of uncertainty.

Implications
Understanding the various types of tourism consumption preferences under uncertainty conditions is of great significance to the marketing strategy for Shenzhen, such as in market segmentation and the implementation of a duty-free policy. Based on the above empirical results and discussions, the following recommendations are made: For food consumption, since both domestic and foreign tourists attached great importance to food hygiene and variety, the Shenzhen Food Safety Bureau, as the main supervision department, should actively play a guiding role, focusing on improving food safety and hygiene in the catering industry in tourism areas, for example by strengthening the governance, implementing dynamic supervision, and ensuring the disclosure of food safety information. The main reason for visitors losing confidence in local food consumption is that in the era of rapid socioeconomic development, the population flow has led to the spread and homogenization of food, resulting in the food being similar in different regions. Therefore, regarding food operators in scenic spots or business districts, in addition to focusing on sanitation conditions and the dining environment, food diversity should be improved in order to improve the dining experience and relieve taste fatigue. In addition, due to uncertainty caused by imperfect knowledge, inexperience, and unfamiliarity, the food consumption preferences of domestic and international tourists are quite different. Domestic tourists focus on flavor, while international visitors focus on the location and recommendations. According to the experimental results, the more importance tourists place on flavor and variety, the greater their preference for local restaurants and street snacks. However, greater consideration of location and recommendations will lead to larger opportunities for high-class restaurants. Therefore, local restaurants and street snacks are suggested for domestic tourists as key consumer groups, while senior restaurant chains with convenient locations and that come highly recommended should be prepared to better serve international tourists. Meanwhile, to minimize the degree of uncertainty arising from imperfect knowledge, inexperience, and unfamiliarity, it is recommended that the Shenzhen Tourist Administration strengthen the accessibility for international tourists, such as by advertising restaurant logos in tourist areas, promoting official websites, and providing English labeling at restaurants; Travel accommodation is usually accompanied by certain population movements, so safety was deemed important by both groups. To prevent fraud, theft, and cheating, it is advised that security departments strengthen their supervision over the standardization of hotels near tourist areas, and that hotel managers regulate the management of information related to tourists. In addition, based on empirical results, if international tourists travel to Shenzhen with a high degree of uncertainty caused by unfamiliarity, their accommodation preference tends to be for star hotels. Specifically, their hotel preference will gradually drop from star hotels to hostels within a 7-day period. However, with the extension of this stay, the situation reverses. Therefore, hostels in Shenzhen should attract international tourists for 7-day stays, while star hotels should advertise to short-term tourists (less than 3 days) or long-term tourists (more than two weeks); For shopping consumption, facing fierce competition in Guangdong-Hong Kong-Macao Greater Bay Area, Shenzhen's tourism industry must enhance its competitive advantages to prevent the outflow of consumption potential. The empirical results showed that a lower degree of uncertainty arising from policy changes would contribute to a greater proportion of shopping spending, providing guidance for public agencies in Shenzhen in formulating and evaluating the effectiveness of tourism policies. Therefore, it is necessary to implement a duty-free policy in Shenzhen and further improve the Qianhai-Shenzhen-Hong Kong Modern Service Industry Zone. For the majority of domestic tourists, Hong Kong is their main shopping detination. Since the multiple entry permit policy was established in 2009, tourists and Shenzhen citizens have enjoyed convenient travel and shopping in Hong Kong, which has affected Shenzhen's tourism industry to some extent. The implementation of a duty-free policy could ensure that quality imported goods are available at competitive prices, meaning that tourists do not have to travel to Hong Kong to access global products. In addition, retail stores in Shenzhen should consider innovative marketing, such as boosting brand awareness, focusing on product packaging, making items easy to carry, and improving in-store environments. However, price cuts are unlikely to be effective, as this study showed that most tourists perceive retail products as price-inelastic.
Despite applying a new method to measure uncertainty and developing a model to examine the relationships between tourism consumption preferences and uncertainty, there are still some limitations. (1) Due to indicator restrictions in measuring uncertainty, this study only examined tourism consumption preferences under uncertainty conditions related to imperfect knowledge, inexperience, unfamiliarity, and policy changes. In fact, as mentioned in the literature review, age, income, war and political instability, health, crime, and terrorism are also factors causing uncertainty. In future research, these various uncertainty factors could be investigated as they relate to tourism consumption activities, such as transportation and entertainment. (2) For the questionnaire survey used for international tourists, although the double back-translation method, double proofreading process, and a pilot test were adopted, these measures only ensured the rigor and standardization of the process, but did not mitigate situations in which both the investigator and respondent are not native speakers of English, such as Japanese and Korean tourists.
In addition, it needs to be emphasized that one of the main challenges for Shenzhen's tourism industry is how to prepare for the upcoming generational shift in order to pursue sustainable development and transformation. In fact, a large number of well-known tourist attractions in Shenzhen, such as Splendid China and Windows of the World, were built in the 1990s, carrying the dreams of the previous generation during that special era in the early stage of China's reform and opening up. However, these attractions do not interest the new generation of young people. In the future, it would be worth verifying of proposed MVEP-OPM model from the perspective of transgenerational analysis [53]. Regarding the response to the coronavirus disease 2019 (COVID-19) pandemic and the regeneration of post-COVID-19 urban economies, the scope of the models applications will be expanded in conjunction with urban tourism planning.