Exploring tourist shopping from the perspective of duty-free shopping: An analysis of online reviews

Online reviews

Online travel agency (OTA) and social media sharing platforms have grown rapidly alongside the rapid expansion of the Internet. Online reviews are a type of UGC becoming increasingly prevalent on Internet platforms (Alzate et al., 2022; Boccali et al., 2022). Online reviews are texts that people post on websites and include their thoughts. They are crucial sources of data for learning about customers’ shopping experiences (Park et al., 2021; Zhang et al., 2021b). The experience of the reviewer in using a particular product or service, the sentiment message included in the review, and opinions about the product or service are typical components of online user reviews (Litvin et al., 2008). Online product reviews, a type of electronic word-of-mouth, can have a tremendous impact on customers (Gunasekar et al., 2021; Racherla and Friske, 2012). To eliminate ambiguity in the product purchase process, consumers frequently consult pertinent online reviews before making judgments (Liu and Chen, 2022; Vermeulen and Seegers, 2009). For instance, most customers consult online reviews before reserving at travel-related establishment (Su et al., 2022). Word-of-mouth is the primary factor influencing consumers’ purchasing decisions, and actual consumer reviews are the most important components of good or bad word-of-mouth (Gunasekar et al., 2021).

Online reviews offer a variety of information, encouraging academics to analyze them to determine their potential usefulness (Alzate et al., 2022). Real online reviews have a remarkable impact on whether customers decide to make a purchase, which, in turn, affects the profitability of the relevant operators (Liu and Chen, 2022; Xie et al., 2011). Marketing managers can identify user issues and enhance their marketing strategies by examining their Internet evaluations. Relevant operational managers can target their service strategies after evaluating the quality of their goods and services using online review data (Liu and Chen, 2022). Consumers are more likely to believe in negative Internet evaluations, and it is possible to determine what characteristics of goods or services the customer dislikes by analyzing bad reviews (Kusumasondjaja et al., 2012; Ruytenbeek et al., 2021).

Duty-free shopping

Duty-free is a term used to describe the tourism retail industry, in which some countries and regions allow one or more businesses to set up shops in the ports, or cities to sell duty-free goods to inbound and outbound passengers. In 1947, Shannon Airport, in Ireland, was the birthplace of duty-free shopping. The global duty-free shopping industry has evolved over the past 70 years. Its theoretical foundation is that overseas personnel do not receive corresponding welfare in their home country. Consequently, consumption taxes such as VAT and CGT should be refunded in the form of duty-free shops at airports. The growth of the duty-free shopping industry can boost consumption, as well as urban tourism, transportation, logistics, packaging, and other industries. Currently, duty-free shops include those at entry and exit ports, shops on outlying islands, and shops in cities. Compared to duty-free shops at ports, duty-free shops in cities generally have a larger business area, more complete product categories, more appropriate space layout, and consumers’ shopping time is not limited. As a result, authorities are able to distribute more high-priced goods, expanding the market for big-ticket luxury items, such as clothing, watches, and bags.

Duty-free shopping involves a unique buying pattern that cannot be overlooked because it promotes consumption and economic growth. Previous research on duty-free shopping can be divided into three main areas: shoppers’ motivation (Chung et al., 2013), shoppers’ behavior (Lee et al., 2021), and shoppers’ experience (Kim et al., 2016). For the motivation of duty-free shopping, Choi and Park (2017) found that customers are mostly influenced by two aspects, attitude and desire, in their decision-making process for duty-free purchasing. Additionally, for those who shop at airports, the environment has a remarkable impact on their motivation to purchase (Chung et al., 2013). Regarding consumer shopping behavior at airports, the two categories of impulse-caused purchasing behavior proposed by Lee et al. (2021) are cognitive and affective, with an association between customers’ affective impulse purchase behavior and post-purchase satisfaction.

Researchers have used various levels of investigation to analyze duty-free shopping experiences. Xu and McGehee (2012) found that gender disparities affect how satisfied people are with their shopping experiences. Kim et al. (2016) separately evaluated the two stages of a customer's duty-free shopping experience. The impact of emotional disparities on service quality and brand attitudes was investigated by comparing and contrasting consumers’ emotions before and after entering duty-free shops. Martín et al. (2019) conducted an analysis showing that the determinants affecting airport shoppers’ satisfaction were primarily related to variances in their places of residence and travel destinations, and that satisfaction was more elastic in terms of brand and type of consumer products. Human touch, reputation, and value were shown to be the three key factors influencing shopping destinations when Hung et al. (2021) investigated the preferences of Chinese high-end travelers. Additionally, the way in which duty-free shops are landscaped affects how clients feel about shopping there (Kwon et al., 2015).

Shopping tourism and tourist shopping

Shopping has become a general activity in the tourism process as tourism has grown and the number of tourists has increased (Timothy and Butler, 1995; Yüksel, 2007). People's strong desire to shop began to emerge, and tourists with shopping as their primary motivation for travel began to emerge (Kim et al., 2011). Tourist shopping has a substantial impact on the economies and cultures of destinations. Furthermore, shopping tourism provides tourists with travel experiences and memories (Josiam et al., 2005). Currently, there are two major theories regarding the origins of shopping tourism (Choi et al., 2016). The first school of thought holds that shopping has evolved into one of the primary activities of tourism, rather than merely an incidental activity that occurs during the tourism process (Jansen-Verbeke, 1991). The second viewpoint is that shopping tourism is the result of tourists combining tourism-related activities with cross-border shopping. Tourist shopping refers to the purchasing habits of tourists who engage in various tourism activities. Tourist shopping expenditure planning is inadequate or nonexistent, and consumption elasticity is high.

As a major theme in tourism, shopping tourism is critical for tourism research. Timothy (2005) was the first to introduce the concept of shopping tourism. However, the definition of shopping tourism remains controversial. The current academic consensus is that “tourism behavior and activity with the primary purpose of shopping” is a valid definition. Under this definition, it is unclear how to define whether tourists’ main purpose is shopping. Existing research on shopping tourism focuses primarily on the following topics: shopping behavior (Yüksel, 2007), shopping destination (Yeung et al., 2004), satisfaction (Turner and Reisinger, 2001), and shopping risk perception (Yüksel and Yüksel, 2007). Hobson (2000) used case studies to investigate the content of tourism retail and duty-free shopping. Turner and Reisinger (2001) demonstrated that tourist shopping costs are a significant factor in determining public shopping satisfaction. Yeung et al. (2004) discovered that shopping significantly contributes to a destination's local economy. In many shopping indicators, such as price, variety and supply of products, and the convenience of shopping stores, Hong Kong outperforms among the two shopping destinations of Hong Kong and Singapore. Yüksel and Yüksel (2007) investigated the impact of shopping risk perception on tourists’ mood and satisfaction.

Existing studies on shopping tourism in Asia are especially abundant in Hong Kong, which may be attributed to Hong Kong's global reputation as a “shopping paradise.” Few studies have examined why tourists enjoy shopping activities while on vacation, tourists’ emotions while shopping, and the perceived value of shopping tourism experiences (Choi et al., 2016). Furthermore, the existing research on shopping tourism lacks research on shopping websites and other aspects, and future shopping tourism research scholars should investigate other tourist destinations (Choi et al., 2016).

Perceived value

The marketing and management industry has studied perceived value as a key element affecting consumer purchasing decisions in great detail (Swait and Sweeney, 2000; Tsiotsou, 2006). Zeithaml (1988) believes that perceived value is a one-dimensional concept that provides a comprehensive evaluation of product effectiveness by customers. According to other academics, perceived worth has multiple dimensions, and different topic areas use different definitions for what those dimensions are (Lee et al., 2007). Efficiency and quality are two variables that constitute perceived value in the economic sector (Sánchez-Fernández and Iniesta-Bonillo, 2009).

The perceived value also differs depending on the customer's situation. Sweeney and Soutar (2001) examined value dimensions in a retail scenario to investigate how consumers perceive the worth of consumer durables. Consumers’ perceived value is a critical factor in their satisfaction with their shopping experience and subsequent repurchase behavior (Chang and Tseng, 2013). The perceived value of shopping websites is significantly influenced by high-quality online services and website content (Molinillo et al., 2021). Consumers perceive their shopping environment and location before engaging in shopping behavior, and perceived value has a non-negligible impact on shopping satisfaction (El-Adly and Eid, 2016). Perceived value is a significant element that influences customer loyalty, even though the precise aspects of perceived value are evaluated differently in different settings (Molinillo et al., 2021; Zhang et al., 2021a, 2021b). According to the consumer perceived value (CPV) model, perceived value consists of four primary components: price, quality, social, and emotional (Chi and Kilduff, 2011). Price value is considered the most important of these four factors (Chi and Kilduff, 2011).

Despite the several studies on perceived value, there are very few studies on perceived value within the knowledge system of tourist shopping. Consequently, the perceived value of tourists should be investigated under this theme to provide theoretical contributions to the improvement of the tourism knowledge system and practical effects for the rapid development of tourism (Choi et al., 2016). Through an analysis of online review data, our study examines the precise factors that affect consumers’ perceptions of the value of duty-free shopping. The following three issues were addressed in our study:

What are the factors that affect tourists’ perceived value during duty-free shopping?

Data collection and pre-processing

From 2018 to mid-2022, we used Python crawler and octopus collector to collect 11,304 comments from the Duty-free Centers of Haikou Meilan International Airport and Sanya International Duty-Free City. The above two duty-free shopping places, one representative of airport duty-free shopping and the other of city duty-free shopping, have a large flow of people. The shopping reviews of the two duty-free shopping locations effectively represent the duty-free shopping experience of Hainan's outlying islands. The fundamental process in text mining is data preprocessing. Online review datasets frequently include considerable noise and irrelevant information, which is useful for topic modeling following data pre-processing. First, 9386 legitimate comments were obtained after the gathered comments were cleaned and de-duplicated. The words were deactivated after the text data were cleared. Deactivated words are words that are removed before natural language processing. These words, which include conjunctions, prepositions, pronouns, etc., are the most frequently used words in all languages and do not add much to the text's content. To begin processing deactivated words, a collection of stop words composed of typical Chinese punctuation and other senseless conjunctions was created. Word-frequency analyses were performed on the topic terms following data processing. The stop-word list was then expanded to include any unidentified or meaningless terms that topped the word frequency statistics. This was done to eliminate meaningless terms in the following editions.

Text analysis and LDA

Blei et al. (2003) introduced Latent Dirichlet Allocation (LDA) for the first time. It is a topic modeling-focused unsupervised machine learning model. The mechanism of this topic prediction model is that each text word is chosen for a particular probability subject, and then a word is chosen with a particular probability from that topic. A topic model can utilize LDA based on the probabilistic modeling of discontinuous datasets to identify prospective themes in a text. The LDA model, commonly referred to as a three-level Bayesian probability model, can accurately model text because it includes document (d), topic (z), and word (w) structures. More importantly, the LDA model incorporates probabilistic information in addition to conventional spatial vector models. Prospective subjects of the text dataset can be investigated using the LDA topic model, which can then be used to examine the main themes of the text dataset and the associated feature words. Currently, LDA models have a wide range of applications, primarily in literature analysis (Tomojiri et al., 2022), news opinion analysis (Pu et al., 2022), online reviews (Guo et al., 2022), and computer applications (Ning et al., 2022).

Compared to conventional keyword extraction methods, LDA is an unsupervised machine-learning algorithm that can describe the material more thoroughly and precisely. Reviews of duty-free shopping often include brief writings of 5–200 words. Since these texts are always brief and highly colloquial, they cannot be classified using conventional machine-learning techniques. The vast amount of review data also places increased demand on the efficiency of extraction methods owing to the abundance of textual reviews. Consequently, the topic extraction method used in this study (LDA) is appropriate. From online review text data, we extracted probable topics using LDA, and inductively describe the clustered topic terms as dimensions of CPV.

Perplexity analysis

In information theory, perplexity is a conceptual term describing how well a probability distribution or model predicts a sample. It compares two probability distributions or models, with the one having the lowest degree of confusion better at predicting the sample. Perplexity is a technique used to gauge the strength of linguistic probability models in natural language processing. A probability distribution over complete phrases or sections of text can be considered a linguistic probability model. The specific formula is as follows:

Sentiment analysis

Pang et al. (2002) were the first to establish the idea of sentiment analysis, starting the formally organized study in the field. The process of evaluating, processing, generalizing, and deducing subjective texts with emotional undertones is known as text sentiment analysis. Based on user interactions, the Internet generates many comment text messages. These texts represent people's feelings and emotional dispositions toward many individuals, occasions, and things, including joy, happiness, rage, sadness, and praise. Potential customers or users can read these subjective reviews to understand the public perception of a service or good. Sentiment analysis of online comments is a process that analyzes the internal meaning and elements of the text to determine their emotional tendencies. Currently, there are two methods for text sentiment analysis. The first strategy is based on an organized lexicon. Using this technique, text sentiment analysis was based on sentiment dictionaries. A text corpus can be subjected to sentiment classification by labeling sentiment words with polarity and intensity using text-sentiment dictionaries. Words in a sentiment dictionary are typically divided into sentiment adjectives, degree adverbs, and negatives. Subsequently, word nature and sentiment polarity can be used to estimate the sentiment score of the text. Machine learning forms the foundation of the second sentiment analysis technique. This is used to extract features from a large corpus of labeled or unlabeled texts, and sentiment analysis is performed to produce the results. In our study, online reviews of duty-free shopping were subjected to sentiment analysis using a machine learning approach.

Analysis of high frequency words

Word frequency statistics were collected after subscribing to all review materials and filtering topic terms that were unrelated to the evaluation perspective. The top 15 topic words are shown in Figure 2. From the figure, we can see that consumers are most concerned about the price of duty-free products, and the brand is the second most concern. Since the main consumer group for duty-free shopping is women, concern for cosmetics ranks third. Other factors that affect the duty-free shopping experience include queuing time, shopping environment, activity strength, service attitude, brand, and the variety of duty-free products. In terms of the choice of duty-free shopping products, consumers are mostly concerned with cosmetics and skin care products. Through word frequency analysis, we find that price and brand are the two factors that tourists care about most in the duty-free shopping process.

OPEN IN VIEWER

Figure 2.

Top 15 word frequency statistics.

Theme analysis

To guarantee that the topics are independent of one another, we calculated the ideal number of topics using a confusion curve. The confusion curve graph was combined, and the final number of topics was found to be five (Figure 3). Using the PyLDAvis package for Python, the interactive display of the duty-free shopping review subject class clustering findings is shown in Figure 4. The topics were shown as a circle graph using PyLDAvis, an interactive Python program for LDA visualization. When subjects are mapped from a multidimensional space to a two-dimensional space, the size of each circle represents how frequently the topic appears, and the length between circles indicates the proximity between topics. The size of each circle in Figure 4 denotes the degree of the topic's aggregation; the smaller the circle, the better is the topic's aggregation impact, and the overlap between the circles denotes the closeness between the topics.

OPEN IN VIEWER

Figure 3.

Confusion curve.

OPEN IN VIEWER

Figure 4.

Theme visualization.

The results of the topic distribution after LDA using Python are shown in Figure 5. A topic of zero indicates that the crawled reviews did not belong to the five topics classified. The clustering results for the perception of duty-free shopping experience are presented in Table 1. In this study, the factors affecting tourists’ duty-free shopping experience are divided into five themes; theme 1 is summarized as consumers’ perceptions of duty-free brands and types, and the theme terms related to it are: “brands,” “cosmetics,” “ skin care products,” “watches,” “perfumes,” etc. Theme 2 is summarized as the consumers’ perception of the online shopping process, and the related thematic terms are: “online shopping mall,” “after-sales,” “process,” “order,” “membership,” etc. Theme 3 is summarized as consumers’ perceptions of service during shopping, with related thematic terms: “attitude,” “service,” “checkout,” “enthusiasm,” “membership,” “waiter,” etc. Theme 4 is summarized as the consumers’ perception of the environment during the shopping process, and the theme words related to it are: “store,” “scale,” “overall architecture,” “operation,” etc. Theme 5 is summarized as consumers’ perception of the cost of purchasing duty-free products, and the theme words related to it are: “discount,” “strength,” “price tag,” “price,” “saving,” “discount,” etc.

OPEN IN VIEWER

Figure 5.

Percentage of each theme.

OPEN IN VIEWER

Table 1.

LDA theme analysis.

Topic	Topic vocabulary	Manual coding
1	“brands,” “cosmetics,” “ skin care products,” “watches,” “perfumes,” etc.	Perceived variety and functionality.
2	“online shopping mall,” “after-sales,” “process,” “order,” “membership,” etc.	Perceived online process.
3	“attitude,” “service,” “checkout,” “enthusiasm,” and “membership,” “waiter,” etc.	Perceived service.
4	“store,” “scale,” “overall architecture,” “operation,” etc.	Perceived environment.
5	“discount,” “strength,” “price tag,” “price,” and “saving,” etc.	Perceived cost.

Unlike previous studies, this study found that the online process is also a key factor affecting visitors’ purchase experiences during duty-free shopping. Through emotion analysis, we found that the online shopping process for duty-free goods needs to be further improved, which has not received much attention from scholars in previous studies on the shopping experience of duty-free goods.

Sentiment analysis

SnowNLP is a Python class library used to process Chinese text. The models trained using SnowNLP were based on product shopping data. In general, models need to be retrained as they are used to better match the data. The original reviews were split into two datasets of positive and negative sentiments using the SnowNLP text classification algorithm; the return value was the likelihood of the sentiment. The sentiment index ranged from 0 to 1. The more positive the emotional performance, the closer the result was to one. In addition, the more unfavorable the emotional performance, the closer the result is to zero. Figure 6 displays the percentages of positive and negative comments for each topic after predicting the sentiments of all the comments. Figure 7 displays the results by averaging the probabilities of positive and negative sentiment separately. For Theme 1, “Perceived variety and function,” 79% of the comments were positive and 21% were negative. The average probability value of positive emotions was 90% and the average probability value of negative emotions was 10% for this theme. For Theme 2 “Perceived Online Processes,” 66% of all comments were positive and 34% were negative. The average probability value for positive emotions was 85% and for negative emotions was 10% for this theme. For Theme 3 “Perceived Service,” 76% of all comments were positive and 24% were negative. The average probability value of positive emotions was 89% and the average probability value of negative emotions was 10% for this theme. For Theme 4 “Perceived Environment,” 75% of all comments were positive and 25% were negative. The average probability value of positive emotions was 87% and the average probability value of negative emotions was 12% for this theme. For Theme 5 “perceived cost,” 80% of all comments were positive and 20% were negative. The average probability value of positive emotions was 88% and the average probability value of negative emotions was 14% for this theme. Among the five themes analyzed in this study, Theme 5 “perceived cost” and Theme 1 “perceived variety and function” had a higher percentage of positive comments, Theme 2: “Perceived online process” has the lowest percentage of positive comments.

OPEN IN VIEWER

Figure 6.

The emotional share of each theme.

OPEN IN VIEWER

Figure 7.

Sentiment analysis of each theme.

Theoretical insights

As an activity in the process of tourism, duty-free shopping plays an important role in both tourist shopping and shopping tourism. However, scholars have paid insufficient attention to duty-free shopping. Based on the perspective of the duty-free shopping experience, our study enriches the research on the tourism shopping experience and provides new insights for tourism destination marketing organizations. By mining the text and clustering the topic of comments and sentiment analysis, we propose corresponding marketing suggestions. The theme modeling results can be used to summarize consumers’ duty-free shopping experiences into five basic dimensions: perceived diversity and function, perceived online process, perceived service, perceived environment, and perceived cost. Perceived brand reputation, product categories, product quality, and other factors reflect perceived diversity and functionality. Perceptions of online shopping centers, online customer service, after-sales service, and other factors are influenced by the online process of perception. Perceived service is represented by the shop assistants’ perceived service attitudes, service levels, and business accomplishments. The overall scale, infrastructure, and decor of the duty-free shopping center reflect the sense of the environment. Perceived cost is embodied in factors such as perceived promotion, product price, and time cost. Consumers form perceptual judgments regarding the elements listed above and exhibit positive, neutral, and negative emotions.

Through the findings of our study, we divided the tourists’ perceived value of duty-free shopping into three stages. The first is the cognitive stage. In this stage, tourists’ perception is mainly manifested in five aspects: perception of diversity and function, perception of the online process, perception of service, perception of environment, and perception of cost to form the perceived value of duty-free shopping. The second stage mainly manifests as emotional changes in tourists. Tourists conduct a comprehensive analysis of the shopping experience and form positive, neutral, and negative emotions, thereby affecting their satisfaction. The third stage is tourist behavior. Satisfaction with a previous shopping experience affects the loyalty of tourists, and tourists decide whether to visit or buy again.

In other words, this study enriches the knowledge system of consumer shopping behavior. The primary consideration for Chinese consumers when considering duty-free shopping, according to our findings, is whether pricing is favorable. Whether this finding holds true for customers at other tourist destinations needs to be confirmed in future studies (Figure 8).

OPEN IN VIEWER

Figure 8.

Research conclusion model.

Practical implications

By analyzing online reviews of duty-free shopping in Hainan, this study provides insights for duty-free shopping operators to improve their services and marketing strategies from the perspectives of destination marketing and management. Managers can improve their duty-free shopping business and services in the following areas in the future to enhance consumers’ shopping experiences.

First, the abundance of goods is a key factor that affects the consumer experience in terms of the perceived value of the perceived brand and its functionality. If the product selection is single or the supply is not timely, consumers will not be able to choose their favorite products, which affects their shopping experience. Therefore, relevant operators need to further optimize the variety of duty-free products for the outlying islands. By combining consumer interests in product selection, the richness of duty-free products can be improved to meet consumer shopping needs.

Second, regarding the perceived value of online processes, the convenience and intelligence of online processes as well as the efficiency of online customer service all play a significant role in influencing the consumer experience. Therefore, relevant operators need to further optimize the online shopping mall process and strengthen the training of online customer service. For problems encountered during the operation of the online shopping mall (such as differences between the display of product discounts and actual orders, refunds of products, and after-sales services), the responsible persons should provide timely feedback and solutions and conduct return visits or questionnaire surveys with active users. In addition, the introduction of duty-free products in online shopping malls requires further improvement. Perfect information about the function of goods is helpful in stimulating consumers’ willingness to consume and increase the purchase rate of goods.

Third, in terms of service value perception, service attitude has a significant impact on how consumers perceive the overall service. The analysis of duty-free shopping reviews revealed that a warm attitude enhanced consumer satisfaction during the shopping process. Therefore, managers must strengthen the training of industry service personnel to improve their service levels. To improve the duty-free shopping experience and increase consumers’ willingness to repurchase, it is critical that relevant practitioners handle consumer needs in a timely and effective way.

Fourth, regarding the perception of environmental value, whether the relevant infrastructure is convenient, is a factor that cannot be overlooked in influencing consumer experience. Relevant operators need to improve the construction of infrastructure near duty-free stores, such as parking lots, restaurants, and leisure and entertainment venues. Consumers’ shopping experiences can be improved by creating a world-class duty-free shopping environment.

Fifth, in terms of the perceived cost value, economic and time costs are the two major concerns of consumers. Operators can increase the price advantage of duty-free products by organizing large-scale promotional activities and issuing coupons to reduce the economic cost of purchasing duty-free products, thereby attracting consumers. In addition, it is important to reduce the time cost (e.g., queuing time) during the shopping process and airport pickups.

Limitations and future prospects

First, this study uses online reviews for a duty-free shopping experience analysis. Since online reviews do not provide information such as the residence, age, and occupation of the evaluators, they can be combined with questionnaires and other methods to compensate for this in subsequent studies. Second, future research can use other methods to validate the duty-free shopping experience model proposed in this study, draw conclusions, and conduct further in-depth analyses. Finally, the results of the current study put forward five major factors based on the perception of the duty-free shopping experience, but failed to identify the weight of each of these factors on the duty-free shopping experience; future research can be conducted on this.