Factors affecting consumers’ impulsive buying behavior in tourism mobile commerce using SEM and fsQCA

Impulse buying and urge to buy impulsively

Impulse buying is “a purchase that is unplanned, the result of an exposure to a stimulus, and is decided on the spot” (Piron, 1991, p. 512). Unlike the rational and reasonable consumers’ decision-making process of evaluating and comparing alternatives before consumption (Häubl & Trifts, 2000), impulsive buying behavior arises from a sudden burst of hedonic and positive affect (Chung, Song, & Lee, 2017; Dhurup, 2014). Consumers can act impulsively due to external stimuli and, therefore, are more likely to make impulsive purchases when exposed to an environment in which they can buy goods directly (Floh & Madlberger, 2013). Impulse purchasing is characterized by a short and spontaneous decision-making process, and as mentioned above, it can be based on unplanned behavior and influenced by an emotional state (Verhagen & Van Dolen, 2011).

Consumer impulse purchases differ from habitual purchasing or instinctual conditions (Rook, 1987). Therefore, not only the external stimuli, but also the individual's mental state should be considered. This can be viewed in the same way that impulse buying can be linked to individual pleasurable experiences or values (Rook, 1987). Therefore, because it is difficult to directly measure the impulse buying behavior itself, we attempt to estimate impulse buying behavior by identifying the impulse buying intention of consumers (Chung et al., 2017).

Urge to buy impulsively is an emotional state enabling consumers to purchase impulsively (Parboteeah, Valacich, & Wells, 2009; Verhagen & Van Dolen, 2011). This urge is defined as “the state of desire that is experienced upon encountering an object in the environment” (Beatty & Ferrell, 1998, p. 172), although not all urges to buy impulsively result in impulsive buying behavior (Chung et al., 2017). Accordingly, the impulsiveness of the individual should be considered first when confirming a person's impulse. Many studies have examined impulse buying based on various personality and demographic characteristics (Turkyilmaz, Erdem, & Uslu, 2015). However, impulse buying behavior is only partially explained by a person's temperament, and external factors due to the changing environment should be explored as well (Wells, Parboteeah, & Valacich, 2011). Therefore, in addition to individual characteristics, this study also considers variables set from the external environment.

Originally, impulse buying research was framed as studies of habitual purchase or unplanned behavior (Rook, 1987). Then, impulse purchasing research was conducted according to purchasing environment or the buying stimulus of sellers (Xu, 2007). Impulse buying was further studied as product categories or with consumer tendencies and basic information. Many empirical studies have demonstrated how various online cues (like interface design and scarcity messages) can stimulate consumers’ urge to buy impulsively (Dawson & Kim, 2009; Floh & Madlberger, 2013; Liu, Li, & Hu, 2013). As mobile transactions are increasing, research on impulse purchasing in online shopping naturally leads to mobile commerce.

Impulse buying in mobile commerce

The study of impulse buying in mobile commerce has proceeded as follows. According to Alliance (2008), mobile payment systems induce consumers to buy impulsively. The researchers then investigated consumers’ impulse buying behavior according to the effect of advertising provided by mobile commerce (Drossos & Fouskas, 2010; Drossos, Kokkinaki, Giaglis, & Fouskas, 2014). Wu and Ye (2013) empirically demonstrated the influence of consumers’ immersion and pleasure while also examining the impact of mobile shopping on impulse buying based on flow theory. T. Lee, Park, and Jun (2014) explored how mobile transaction features can drive impulse buying and demonstrated that impulsive buying behavior can be closely related to consumer regret. C.-C. Chen and Yao (2018) suggest that the characteristics of mobile transactions are strong precursors for consumers’ impulse purchases, as compared with the transactions on other buying channels. Meanwhile, Dewan and Benckendorff (2013) assessed influences by dividing users’ impulsiveness and the degree of technology use on mobile usage with regard to destinations. According to the results of this study, impulsiveness does not have a significant influence on consumers determining a destination while using mobile devices.

Impulse buying in tourism

The study of impulse buying in the tourism and hospitality field can be divided into two parts depending on the situation of impulse buying. One studied the behavior of customers in specific situations of tourism or using hospitality services. In this case, a person can engage in impulse buying behavior because of their mood or external stimulus. Several scholars studied the situation in settings like casino (Prentice & Wong, 2016), cruise tourism (Ahn & Kwon, 2020), festivals (Chang, Stansbie, & Rood, 2014), and airport (Geuens, Vantomme, & Brengman, 2004). The other is the behavior of customers in a situation of transaction via an online environment, such as the web or mobile settings. In this case, discussions were mainly conducted about the stage of exploring or planning a product or service. Stern (1962) classified impulse buying that may occur in this process as planned impulse buying. This is because, in a person's pre-trip stage, the product and service search and planning are somewhat different from the nature of impulse purchase. Nevertheless, researchers have examined impulsive behaviors among people in the tourism and hospitality industries that may be caused by people's temperament and the efforts of suppliers. Recently, in addition to sites that provide tourism products or services, research has examined a person's impulse purchase in social media by set sharing travel experiences as a stimulus. Chih, Wu, and Li (2012) applied an individual's hedonic persuasion and impulsiveness-related tendency as the main research factors to check how customers’ impulse purchases occur on online travel agencies’ websites. Floh and Madlberger (2013) verified how the pleasure of shopping induced by website features affects the impulsive buying behavior of customers by applying the framework of stimulus and response. Experimental studies have also been performed to check how impulsive behavioral intentions are among people exposed to social media post types and the influence of the poster (Szymkowiak, Gaczek, & Padma, 2021; Yao, Jia, & Hou, 2021). Those kinds of exposure and stimulation of information in the online environment can be further strengthened by mobile technology (Hwang, 2010).

Tourism mobile commerce

The advent of smartphones and applications has made it possible for consumers to search for information and purchase activities with little time and place restrictions (Kim, Koo, & Chung, 2021). This enables individuals to easily acquire, share, and purchase information before and during travel (Um & Chung, 2021). Several previous studies verified which factors stimulated consumers’ intention to purchase mobile tourism products and use mobile services. Lu, Mao, Wang, and Hu (2015) used the service characteristics that users expect from mobile use as factors based on social cognitive theory. In addition, Mahrous and Hassan (2017) conducted an empirical study by applying the psychological characteristics of consumers to confirm the motivation for using mobile as a tourist product purchase channel. These included consumer general characteristics and shopping habits. Rodríguez-Torrico, Prodanova, San-Martín, and Jimenez (2020) confirmed that an individual's attachment to a mobile phone leads to travel-related shopping through mobile. Their study applied the concept of place attachment theory to mobile. Hatamifar, Ghaderi, and Nikjoo (2021) studied the intention of tourists to use mobile apps using the characteristics of mobile apps and technology acceptance models. These studies derive factors based on theories used to study human behavior or perception. Also, the characteristics of mobile technology were applied as variables. However, little study has been conducted to examine the variables by directly determining the reason for using mobile for tourism products or services. Additionally, research on impulse buying on mobile devices in a tourism context is still in an early stage (Wu & Ye, 2013). Therefore, we aim to extract mobile usage motives through qualitative analysis and to grasp the influence they have on consumers’ impulse-buying decisions.

Qualitative research (interview)

The qualitative methodology of Study 1 consisted of interviews with self-reported tourism mobile commerce users. A sample of thirteen participants who had experience with purchasing tourism products in mobile environments was recruited. We enlisted interviewees with at least one year of mobile commerce experience via an online community. We selected the final 13 participants with consideration for age and gender. According to Vasileiou, Barnett, Thorpe, and Young (2018), a sample of 15 to 30 participants is a median sample size for interviews. Nonetheless, our interviews are applied to play an auxiliary role in studies with simple questions and single cases (i.e. tourism), so we considered that 13 is proper sample size for Study 1 (Malterud, Siersma, & Guassora, 2016). To reduce bias that may arise from various types of tourism products, we asked participants to respond to the interview, taking into account only accommodation and airline ticket purchases. We organized the interviews using aliases to protect the respondents’ identities. The key question asked was, “Why did you use tourism mobile commerce?” In addition, information about respondents’ purchasing behavior was obtained through questions about the mobile commerce companies they used and the tourist products they typically purchased. Finally, respondents were also asked about their gender and age.

Responses to the key question, “Why did you use the tourism mobile commerce?” were coded into three stages. Interview was conducted in Korean, coding process for keywords extraction was conducted in English. First, for the data collected from the 13 interviews, two coders summarized the motivation type in the form of a keyword for each response. Subsequently, the coder-identified motivations were compared, and the keywords were either matched or similar keywords were compiled and reorganized. Finally, the eight keywords that appeared the most were selected, and the coding was completed. A list of motivations finalized by agreement between the two coders listed in order of frequency is shown in Table 1. Detail information for interviewees is presented in Appendix A.

OPEN IN VIEWER

Table 1.

Coding table of interviewees’ mobile commerce usage motivations (n = 13).

Motivation	# of mentioned
convenience	10
perceived value	10
ubiquity	8
notice of special promotion	4
timeliness	1
easy to change	1
trust	1
time killing	1

Except for the top four keywords identified, the remaining motivations were not utilized in subsequent analyses, judging that the frequency mentioned was low and unimportant. The most common motivational factors mentioned were convenience, perceived value, ubiquity, and special promotion. The motivations of easy to change, trust, and time killing were excluded. Furthermore, in this study, the motives of convenience and ubiquity can be grouped together as technology-specific perceptions, while the motives of perceived value and special promotion are considered transaction-specific perceptions.

First, for the technology-specific motive of convenience, the explanations provided by the interviews included searching for real-time information. Other responses that were coded as a convenience motive included “Easier to operate than PC” and “Easy to pay and cancel.” Most respondents said that it is convenient to search, book, and purchase while traveling. Examples of interview responses related to convenience are as follows.

The second most common motivation for mobile use is perceived value. Perceived value is related to transaction costs through a mobile device, and it mainly considers that using a mobile device is better than a PC in terms of time, effort, and price while trading. This can be applied more broadly than other variables, depending on what users value more in mobile transactions. An example of a response to perceived value from the interviews is as follows.

The third most common motivation for mobile use is ubiquity. Closely related to convenience, ubiquity is a specific characteristic of mobile commerce referring to the ability of mobile devices such as smartphones to facilitate transactions anytime and anywhere. This means that it is possible to complete tasks such as purchasing goods and making reservations without being limited by time, place, and space. Examples of responses related to ubiquity are as follows.

The last motivational factor to be used in this study is the notice of special promotions. This motivational factor was seen as one of the characteristics that appears in transactions through mobile devices with perceived value. Mobile users can carry mobile devices all the time, unlike their desktop, and, if they set up alarms, they can receive information on special promotions for tourism products in real time. Marketing that characterizes these mobile transactions not only drives people's impulse to travel but also drives a desire to buy tourism products. The following are examples of interviews related to this.

The analysis of Study 1 identified eight motivations for mobile use. Motivations mentioned only once were removed, judging that they were not important motivators. The remaining four motivational factors were identified as core constructs for a research model of mobile tourism impulse buying intention. Based on insights gained from Study 1 and further literature review, several hypotheses where developed, such that the research model analyzed in Study 2 included two mobile characteristics and two transactional characteristics confirmed through Study 1 and, additionally, one individual characteristic was used.

Research model

To summarize the research model (see Figure 2), the technology-specific characteristics of mobile commerce (convenience and ubiquity) are expected to result in impulsive buying through perceived value. Meanwhile, the direct impacts on impulsive buying are also investigated, considering that the transactional characteristics (i.e. special promotion and perceived value) could have a direct effect on impulsive buying. This study also set the impulse tendency of individuals as an independent variable, since this psychological characteristic has been found to have a great influence on impulse buying in previous studies (Bauer et al., 2006; Wells et al., 2011).

OPEN IN VIEWER

Figure 2.

Research model.

Quantitative research (survey)

In order to test the research model in Study 2, we constructed the questionnaire by adjusting the measurement items of the variables used in previous studies to the domain of this research. In order to verify the hypotheses, we confirmed the reliability and validity of the research model through measurement model analysis.

Data collection

The online survey conducted among the potential respondents were extracted from a customer membership database of one of the largest Korean travel agencies. All participants in this survey were contacted and solicited using e-mail, and then they responded via survey websites. Respondents were requested to first answer a screening question about whether they have bought tourism products using a mobile application or website, and only those who answered “Yes” could proceed to the remaining questions. A total of 426 valid responses were obtained to analyze the research model. The demographic information of the sample is shown in Table 2.

OPEN IN VIEWER

Table 2.

Demographic characteristics of respondents (n = 426).

Variable	Percent	Variable	Percent
Gender		Job
Male	30.3	Student	7.5
Female	68.8	Office manager	31.9
Not Answered	0.9	Sales service	8.2
Age		Technician	9.4
Under 20	0.2	Professional	21.1
20–29	30.5	Self employed	4.7
30 −39	45.3	Public servant	4
40–49	16.9	Housewife/husband	6.8
50–59	6.1	Others	6.3
Over 60	1	Monthly average income
Education		Below ￡690	8.2
Less than High School	8.2	￡690 - ￡1379	16.7
College or University	11.7	￡1380 - ￡2069	32.4
Graduate	65	￡2070 - ￡2759	17.1
Masters	3.1	￡2760 - ￡3448	11.5
Doctoral	11.7	Over ￡3448	14.1
Other	0.2	Total	100

A test for non-response bias effects was conducted by separating the top 10% of those who initially participated in the response for the entire response period from the top 10% who participated in the survey at the end, and the mean difference between groups was calculated (Armstrong & Overton, 1977). No statistically significant difference between the two groups was detected, indicating that this study demonstrates a relatively weak non-response bias.

Measurement model

The measurement model was tested with confirmatory factor analysis using AMOS. For a good model fit, χ2/d.f should be less than 3.0 (Bollen, 1989). We also checked the goodness of fit index (GFI), adjusted goodness of fit index (AGFI), normed fit index (NFI) and comparative fit index (CFI) which are all recommended to be more than 0.9 (Doloi, Iyer, & Sawhney, 2011). The results of these indices reported in Figure 2 suggest an acceptable model fit. Furthermore, RMR is 0.039 and RMSEA is 0.048 which are good for under 0.05 (Xia & Yang, 2019). TLI is 0.967 and IFI is 0.972 which are good for over 0.9 (Bentler & Bonett, 1980).

In order to confirm the validity of constructs, we checked the convergent validity and discriminant validity as follows. First, all standardized loadings of measurements should exceed 0.5, and values of composite reliability (CR) should exceed 0.7, average variance extracted (AVE) should exceed 0.5 and Cronbach's α for each construct is required to be larger than 0.7. Table 3 shows that the values of CR and Cronbach's α for each construct meet these criteria.

OPEN IN VIEWER

Table 3.

Results of convergent validity testing.

Construct	Item	Standardized loadings	Construct Reliability (CR)	Average Variance Extracted (AVE)	Cronbach's α
Convenience	conv1	0.798	0.911	0.719	0.910
	conv2	0.901
	conv3	0.851
	conv4	0.839
Ubiquity	ubi1	0.819	0.896	0.743	0.890
	ubi2	0.845
	ubi3	0.893
Perceived Value	Per_val1	0.747	0.890	0.671	0.878
	Per_val2	0.715
	Per_val3	0.901
	Per_val4	0.897
Notice of special promotion	noti1	0.763	0.885	0.660	0.885
	noti2	0.759
	noti3	0.842
	noti4	0.878
Impulsiveness	impul1	0.820	0.915	0.729	0.914
	impul2	0.883
	impul3	0.837
	impul4	0.873
Urge to buy impulsively	urgetobuy1	0.819	0.889	0.727	0.888
	urgetobuy3	0.845
	urgetobuy4	0.893

For discriminant validity of the measurement model, we compared the square root of the AVE for each construct and the correlations between that construct and the others. If the square root of the AVE is larger than the correlations between that construct and other constructs, then this value fulfills the conditions for discriminant validity (Fornell & Larcker, 1981). As shown in Table 4, the minimum value of square root of the entire AVE of constructs (0.812) is higher than the maximum value of correlations among the constructs (0.757). Moreover, if the value of MSV and ASV is lower than AVE, discriminant validity for constructs is secured (Hair, Anderson, Babin, & Black, 2010). Therefore, discriminant validity was secured.

OPEN IN VIEWER

Table 4.

Correlation and descriptive statistics.

	mean	SD	AVE	MSV	ASV	Impul	Conv	Ubi	Per_val	Urgetobuy	Noti
Impul	2.75	1.03	0.729	0.311	0.072	0.854
Conv	3.46	0.80	0.719	0.573	0.269	0.082	0.848
Ubi	3.56	0.82	0.743	0.573	0.267	0.117	0.757	0.862
Per_val	3.54	0.75	0.671	0.517	0.279	0.073	0.666	0.719	0.819
Urgetobuy	2.89	1.00	0.727	0.311	0.158	0.558	0.295	0.229	0.335	0.853
Noti	3.42	0.85	0.660	0.316	0.195	0.157	0.483	0.421	0.562	0.474	0.812

Note: SD = Standard deviation, AVE = Average variance extracted, MSV = Maximum shared variance, ASV = Average shared variance, Digits with italic and bold mean the square root of AVE.

Hypothesis testing

The results of hypothesis verification are summarized in Figure 3. As can be seen in Figure 3, all hypotheses were supported. 61.1 percent of the variance in perceived value was explained by the model. The model also explains 46.7 percent of the variance in urge to buy impulsively.

OPEN IN VIEWER

Figure 3.

The estimated structural model.

Hypotheses 1 addresses the structural relationships between convenience and perceived value for mobile usage. Convenience has a positive effect on Perceived Value for mobile usage (b = 0.187, t-value = 2.842) and was statistically significant at p < 0.01, supporting Hypothesis 1. Hypotheses 2 shows the cause and effect relationships between Ubiquity and Perceived Value for mobile usage. Ubiquity has a positive effect on Perceived Value (b = 0.454, t-value = 6.616, p < 0.001), supporting Hypothesis 2. Notice of Special Promotion has a positive effect on Perceived Value (b = 0.282, t = 5.927, p < 0.001). Therefore, Hypothesis 3 was supported. To sum up, both mobile characteristics and transactional characteristic (represented by Notice of Special Promotion) have statistically significant positive relationships with Perceived Value.

The variables Impulsiveness (b = 0.499, t-value = 10.412, p < 0.001), Perceived Value (b = 0.108, t-value = 1.984, p < 0.05), and Notice of Special Promotion (b = 0.326, t-value = 5.613, p < 0.001) were each found to have a significant positive effect on Urge to Buy Impulsively in mobile commerce. Thus, the results support H4, H5 and H6.

Fuzzy-set qualitative comparative analysis (fsQCA)

Using fsQCA, researchers can find some patterns or interrelationship between antecedents and dependent variables (Kent & Olsen, 2008). Compared to SEM, fsQCA is suitable for complex and non-linear models. In addition, this method is useful in exploring the patterns in combinations of variables rather than proving processes that focus on the directionality between variables. That is, fsQCA is used to determine the optimal combinations of variables, not a causal relationship between variables. In addition, more sophisticated and diverse research results can be obtained by presenting the partial influence of each variable upon the other variables included in a study, unlike the clustering analysis method (Ragin, 2008).

fsQCA results

We examined the combination and influence of variables that can stimulate consumer's urge to buy impulsively in mobile commerce. For this, fsQCA was performed based on five causal conditions: convenience, ubiquity, perceived value, notice of special promotion, and impulsiveness. Table 5 shows the two patterns derived from the fsQCA analysis, a complex solution and a parsimonious solution, which show how the combination of causal variables explains the outcome condition. Unique coverage refers to the degree to which the constituent combinations of causal variables overlap with other combinations (Ragin, 2008). Consistency, on the other hand, is generally judged to have an appropriate level if it has a value higher than 0.75 (Schneider & Wagemann, 2013). Regarding the analysis results, Type 1 shows that urge to buy impulsively results from low convenience, ubiquity, perceived value, and notice of special promotion and high impulsiveness. The Type 1 pattern has a consistency of 0.75 and explains a good number of cases (raw coverage = 0.57). Type 2 indicates high convenience, ubiquity, perceived value, notice of special promotion, and super-high impulsiveness. This pattern type explains the “less is more cases” (raw coverage = 0.51) and has a consistency of 0.89. In other words, Type 1 can be considered to be a group heavily influenced by internal motivation. Type 2, on the other hand, is a group that has impulsive desire under the influence of multiple factors. To sum, both types contain the high or low conditions of mobile, transactional and individual characteristics, which implies that these conditions are not mandatory for impulse buying except for impulsiveness. Table 5 shows the detailed output from fsQCA.

OPEN IN VIEWER

Table 5.

Configurations leading to urge to by impulsively.

Type	Causal conditions					Raw coverage	Unique coverage	Consistency
	Conv	Ubi	Per_val	Noti	Impul
1	⊗	⊗	⊗	⊗	●	0.576684	0.131574	0.75636
2	•	•	•	•	●	0.512997	0.067886	0.89064

solution coverage: 0.644571, solution consistency: 0.764043, consistency cutoff: 0.75636

Note: Black circles indicate the presence of a condition, white indicate the absence of condition

Large circles represent major elements, small circles represent peripheral elements

Theoretical implications

Our research has several theoretical implications. First of all, we have explored the impulse-buying behavior of consumers in purchasing tourism products through mobile commerce, which has not been fully achieved before. While many researchers have traditionally focused primarily on the planned behavior of consumers in tourism studies, this study focuses on the consumption patterns of unplanned behavior. This study combined the qualitative research methods with empirical analysis and presented the in-depth study results. This methodological progress could be helpful for future studies on impulse buying.

Second, our research has led to theoretical advances in mobile research by qualitatively exploring why consumers use mobile applications in purchasing travel-related products. In previous studies, the motivation for using mobile was taken from the characteristics of mobile commerce or selected from the results of previous studies. However, in this study, interviews were conducted to extract variables. Our research will provide an additional basis for future mobile commerce research.

Finally, we have identified the types of consumers who show impulse-buying behavior through fsQCA and have presented the characteristics of each type of consumer pattern. By doing so, our study has overcome some of the limitations of many previous impulse purchasing studies by sufficiently identifying a combination of psychological and external environmental factors influencing impulse purchase behavior. In other words, we have supplemented the limitations of causal verification of many existing empirical studies through further analysis.

Practical implications

For practical implications, this study can be used as a resource to build a sales strategy for mobile commerce of OTA practitioners. Practitioners should devote attention not only to the four major motivations for mobile use, but also to the other factors mentioned. Although not directly highlighted in this study, practitioners should also consider variables such as ease of change and time-killing behavior. Considering these motivations, the company should create an environment for optimizing mobile apps and include fun elements in the app itself that will elicit users’ desire to buy (Shankar, O'Brien, & Absar, 2018). In addition, they can confirm the details of the interview through our research. In other words, our research is an appropriate study for practitioners at a time when mobile use is increasing in tourism.

Second, our research can provide some practical implications for OTAs in setting up a sales strategy to drive impulse buying by consumers. To stimulate consumers’ impulsive desire, practitioners should focus on the value consumers can have by using mobile. They should also make active use of special programs, because such promotions are not only a means of conveying value to consumers but also a means of arousing a desire to purchase directly. Additionally, since impulsiveness is the greatest motivation for consumers’ impulse buying, practitioners should operate sales strategies that drive customers’ internal motivation and psychological factors.

Third, our study provides practical implications for tourism industry marketing strategies. Practitioners need to use the seasonality of tourism to stimulate consumers’ impulse buying by concentrating on a point of time in which tourism is activated, right before the point of carrying out marketing activities. These practitioners’ marketing strategies are important because they can lead to immediate purchasing activities among consumers. Therefore, practitioners should be aware that, along with the idea that travel-related purchasing or browsing activities will be carried out over a long period of time, immediate and impulsive purchasing activities can also be activated through mobile transactions.

In addition, practitioners are encouraged to provide a flexible transaction policy that can support consumers’ immediate product purchase or search during travel. Through this support, practitioners can build trust even in the consumers’ impulsive decision making.

Limitations

Our research has the following limitations. First, the variables obtained from the qualitative research in this study were not sufficiently considered for the variables previously identified in studies of impulse buying behavior. Besides, only impulsiveness was considered in this study as an individual characteristic to identify impulse buying desires. In other words, future studies require a holistic view that considers the variables identified in impulse buying studies. Second, we divided the types of consumers who showed impulsive buying behavior through fsQCA, but only two groups were identified. Therefore, future work needs to find out the results by applying the fsQCA in a wider area.