A Meta-Analysis of Online Impulsive Buying and the Moderating Effect of Economic Development Level

Impulsive Buying and Online Impulsive Buying

According to Rook (1987), impulsive buying is defined as “a sudden, often powerful and persistent urge to buy something immediately”. Based on the definition, Beatty and Ferrell (1998) extended the definition as “a sudden and immediate purchase with no pre-shopping intentions either to buy the specific product category or to fulfill a specific buying task”. Although the definitions of impulsive buying varied in detail, the nature of it remains the same — unplanned. Previous studies have investigated it from various perspectives, including environment (Chang et al., 2011; Mattila & Wirtz, 2008; Mohan et al., 2013), individual (Peck & Childers, 2006; Sharma et al., 2010; Verplanken & Herabadi, 2001), product (Bellenger et al., 1978; Kacen et al., 2012; Liao et al., 2009). To integrate diverse findings and to provide a comprehensive overview of impulsive buying, researchers have made considerable efforts to review the literature (Amos et al., 2013; Iyer et al., 2019; Muruganantham & Bhakat, 2013).

In recent years, the advancement of information technology contributes to the fast growth of e-commerce, which amplified impulsive buying behavior in the online context. From the view of facilitators, taking advantage of information technology, consumers are experiencing a much smoother buying decision process in the online context. First, navigation and search functions help consumers accelerate their searching process, and some people may come to a quick decision (Moe, 2003). Second, personalized recommendation efficiently optimizes consumers’ product discovery process, which drives impulsive buying (Smith & Linden, 2017). Third, one-click buying online makes the path to purchase shorter and easier, which both increases the conversion rate and the incidences of impulsive buying (Verhagen & Dolen, 2011). In general, with all these IT-facilitated features, consumers’ online shopping experience is smoother and the likelihood of impulsive buying may increase (Stern, 1962). However, from the view of prohibitors, consumers may have security concerns. That is, making consumers feel secure is the prerequisite for online shopping. Besides, people with little Internet experience may have higher shopping costs online. Even experienced online shoppers may get frustrated if the website is hard to use. In this regard, optimizing the website and making it easy to use is necessary. These are specific factors that influence consumers’ online shopping experience dramatically. Therefore, exploring how different factors, especially website-related attributes, affect online impulsive buying is worthwhile (Liu et al., 2013; Parboteeah et al., 2009; Turkyilmaz et al., 2015).

Website-related attributes, supported by information technology, serve as environmental cues for online impulsive buying and they can be categorized into task-relevant cues and mood-relevant cues (Parboteeah et al., 2009). Task-relevant cues, such as navigability and search functions, can help consumers achieve their online shopping goals. On the contrary, mood-relevant cues, such as website visual appeal, mainly influence how much users enjoy browsing a website, but they do not directly support specific shopping goals (Parboteeah et al., 2009). Focusing on the website attributes as stimuli, Liu et al. (2013) found that website ease of use, website visual appeal, and product availability (scarcity) are crucial antecedents of online impulsive buying. Despite the meta-analysis of impulsive buying, to our knowledge, there is no meta-analysis on online impulsive buying, and hence these unique website stimuli were not emphasized. Therefore, this study aims to bridge this research gap and provide an amalgamation of the findings of online impulsive buying.

Traditional impulsive buying studies have divided influencing factors into two categories: internal and external ones (Iyer et al., 2019; Kalla & Arora, 2011; Wansink, 1994; Xiao & Nicholson, 2013). The most widely studied internal factors are consumer-related factors, like impulsive buying tendency and pre-purchase mood (Ozer & Gultekin, 2015). Take a step further, researchers explored more into consumers’ emotion-related factors, which are defined as affective stimuli factors. As for external factors, environmental factors, such as window displays and in-store design are widely studied (Gudonavičienė & Alijošienė, 2015).

When it comes to the online context, internal factors remain unchanged. The main internal factors are still personality, hedonic motivation, trust, and so on. However, external factors are different from those of offline, mainly because online impulsive buying happens when a consumer interacts with the website (Verhagen & Dolen, 2011). That is, websites act as an intermediary between consumers and products. Consequently, online impulsive buying researchers have seen website-related attributes, such as IT-facilitated website attributes and media format (Adelaar et al., 2003; Liu et al., 2013), as the primary external environmental factors. For website attributes, perceived ease of use, visual appeal, and product availability (scarcity) are crucial cues of online impulsive buying (Liu et al., 2013). For media format, Adelaar et al. (2003) found that the media format of items’ information presented in the online shopping environment could increase impulsive buying behavior.

To summarize, the key distinction between online and traditional impulsive buying is that e-commerce is full of IT features, and the external stimuli in online impulsive buying are website-related factors.

The Factors for Meta-Analysis

Based on the research emphasis of the collected online impulsive buying literature and following the guidelines of prior research,1 this study included 13 main factors (i.e., website security, website navigability, website visual appeal, interactivity, ease of use, scarcity, novelty, price, promotion, arousal, pleasure, positive emotion, and negative emotion) that have been explored most in previous research and they fall into three categories of stimuli according to the Stimulus-Organism-Response (SOR) framework: two from external factors (i.e., website stimuli (Liu et al., 2013; Parboteeah et al., 2009; Wells et al., 2011)) and marketing stimuli (Chan et al., 2017; Park et al., 2014; Shim & Altmann, 2016)), and one from internal factors (i.e., affective stimuli (Dawson & Kim, 2009; Huang, 2016; Rook & Gardner, 1993)).

Website stimuli, as the important external factors that distinct online and offline impulsive buying, were extensively examined in online impulsive buying research. Therefore, website stimuli were included first in this meta-analysis. Besides website stimuli, another important external factor is the marketing stimuli. Moreover, as part of internal factors, factors related to individuals’ emotions or moods were also included in the meta-analysis due to the importance of the affective process of online impulsive buying.

Website Stimuli

Marketing Stimuli

Affective Stimuli

Moderating Variable: Economic Development Level

Consider budget constraint, consumption level tremendously affects consumers’ buying decision (Tian & Liu, 2011). Also, it affects consumers’ price sensitivities. In this regard, consumption level might moderate consumers’ online impulsive buying behavior. However, only the countries and regions were recorded in previous studies. To deal with this data availability issue, we chose the economic development level as a proxy of consumers’ consumption level. The solution is supported by the following reasons:

First, economic development level is usually positively correlated to consumption level. Statistically, consumers who live in countries or regions with a higher economic development level usually also have a higher consumption level. On the one hand, people in developed countries or regions are more likely to have higher disposable income. Hence, when they are faced with a buying decision, they are less likely to hesitate due to budget constraints. On the other hand, in terms of price sensitivity, the results of a PayPal study indicated that 56.0% of consumers in the USA shop online out of the price advantage as compared to 68.0% in India and 83.0% in China (Saxena, 2019). Besides, Kübler et al. (2018) analyzed the sensitivity of sales to price and user ratings across developing and developed countries. The result indicated that countries with a lower level of income inequality are more sensitive to rating volume when it comes to economic factors. To some extent, this demonstrates the difference in price sensitivity between consumers in developed and developing countries or regions. Second, the economic development level is also positively related to the development level of information technology. Online shopping has been widely accepted in many developed countries while it is still in the primary stage in many developing countries. In this regard, the online shopping experience, from the searching stage to the delivery, may be markedly different in developed and developing countries. That is, people from a country or region with higher development usually have a better online shopping experience, and it may further lead to online impulsive buying. Given these reasons, we decide to use the economic development level as a proxy for consumption level, and we suppose that it may affect online impulsive buying.

The source of the sample in articles that we chose to conduct the meta-analysis includes both developed and developing countries and regions (Zhao et al., 2019). Based on the existing online impulsive buying literature, we speculated that the influences of the antecedents of online impulsive shopping are different between developed and developing countries and took the economic development level as a moderator. Hence, the following hypothesis is proposed:

• H14. Economic development level has a significant moderating effect on the relationship between the website stimuli, marketing stimuli, and affective stimuli and online impulsive buying.

Figure 1 presents the proposed research model.

Data Collection

To ensure the accuracy of meta-analysis results, we made considerable efforts to search relevant literature, including published journals, conference proceedings, and dissertations. Following Webster and Watson (2002), “online impulsive buying”, “online impulsive purchasing”, “online impulsive shopping” were used as keywords and we added keywords such as “website stimuli”, “marketing stimuli” and “affective stimuli” to list the search formula with Boolean logic in the systematic retrieval of relevant articles in the following databases: Google Scholar, Web of Science, Science Direct, SpringerLink, etc. The preliminary search found 1345 initial papers. Then, we read the title and abstract of each paper carefully to check whether it is related to online impulsive buying and dropped the repeated articles. Finally, 121 articles were included.

Meta-analysis requires papers to meet the following criteria:

1. The paper must be an empirical study of online impulsive buying and quantitatively tested relationships between antecedent factors and online impulsive buying tendency or behavior.

2. The paper must have reported correlation coefficients or other values (e.g. F-value) that could be converted to correlation coefficients.

3. The paper must have reported the sample size.

Besides the above screening criteria, to ensure the independence of the research, we also excluded relevant research conducted by the same research team using the same sample.

Finally, 54 articles met all the above criteria and were used for the meta-analysis. Among the 54 articles, 37 are journal articles, 6 are published in conference proceedings, and 11 are dissertations. These studies were published during 2006–2020. In specific, 2 studies were published before 2010 (during the second wave of e-commerce), and 52 studies were published between 2010 and 2020 (during the third wave of e-commerce). The total sample size of the articles is 19,085 and the average sample size is 353. Figure 2 shows the paper selecting process. Figure 3 demonstrates the world distribution and coverage of the studies included in the meta-analysis. Selected papers are listed in Appendix 2 (Table 7).

Coding Procedure

Each article was scrutinized to extract key data to be used in the study. The key data include: author name, publication date, publications, investigated countries or regions, sample size, key constructs, and reported effect sizes. Considering that there are many constructs with different names expressing similar meanings, we merged constructs with similar meanings. For example, perceived enjoyment is similar to entertainment, and visual appeal is similar to aesthetic appeal. According to the guidelines of Rana et al. (2015), we only selected those relationships that have been explored three or more times in the literature in the meta-analysis, and finally, we got a total of 13 relationships.

To conduct the moderator analysis, all articles included in the meta-analysis were divided into two groups based on the economic development level of the country or region.2

A small number of studies did not report the correlation coefficient, but they reported the standard regression coefficient. In this case, we treated the standard regression coefficient as effect size. Very few studies reported F-value. For these studies, we used the formula proposed by Wolf (1986) to calculate the effect size:$r=\sqrt{\frac{F}{F+\mathit{df}}}$, where F is the F-value of the path, and df is the degree of freedom.

In order to ensure the accuracy of data coding, two researchers in this study conducted back-to-back coding on the literature samples according to the coding specifications proposed by Lipsey and Wilson (2001), and then cross-checked the coding results. The consistency ratio was 93.5%. Finally, the research team discussed the inconsistent results carefully and referred to previous classifications to reach agreements.

Analysis Procedure

First, we provided descriptive statistics of each antecedent factor to roughly observe the impact of each antecedent factor on online impulsive buying.

Second, we calculated the combined effect of each pair of relationships (Fleiss, 1993). Besides, to ensure the normal distribution of the correlation coefficient of each pair of relationships, we made the Fisher r to z transformation.

Then, we used a heterogeneity test (Q-test) to test the heterogeneity of the distribution of effect sizes and find potential moderator effects. The economic development level in each study was used as the categorical moderator factor. A forest plot was used to visualize the results of the subgroup analysis.

Finally, to avoid publication bias, we calculated the fail-safe N for each relationship. Publication bias refers to the phenomenon that in academic research, researchers tend to report significant results and avoid reporting results that are not statistically significant (Kraemer & Andrews, 1982).

Descriptive Statistics

This research examined 13 antecedent factors. The average sample size of each path is over 200. The descriptive statistics of each relationship are shown in Table 2.

Correlation Analysis

In this section, we used the correlation coefficient and sample size to calculate the relationship between website stimuli factors, marketing stimuli factors, affective stimuli factors, and online impulsive buying.

It is worth mentioning that the choice of the fixed-effect model or random-effect model is very important. According to Borenstein et al. (2007), when there is a single effect in the hypothesis sample, the fixed-effect model should be selected. Otherwise, the random-effect model should be selected. Given the differences in the samples, the random-effect model is selected to calculate the combined effect.

The results of the meta-analysis are shown in Table 3. Except for the combined effect sizes of the relationship between website security, price, negative emotion, and online impulsive buying, all 95% confidence intervals of combined effect sizes exclude zero, and all Z-scores are significant, indicating that these combined effect sizes are statistically significant.

According to Cohen (1988), the combined effect sizes can be categorized into weak (around 0.1), moderate (around 0.3), and strong (around 0.5). The findings indicated that the website stimuli factors have a significant and positive relationship with online impulsive buying except for the website security factor. Among them, the correlation between interactivity and online impulsive buying is the weakest and the combined effect size of it is only 0.17. It can be inferred that in previous studies, the interactivity of websites is not the main factor affecting online impulsive buying. With regard to the marketing stimuli factors, price, novelty, and promotion are crucial factors affecting online impulsive buying, with their combined effect size over 0.3. In addition, the relationships between price, novelty and online impulsive buying are not significant. As for affective stimuli factors, although the relationship between the negative emotion factor and online impulsive buying is insignificant, the rest of the affective factors are significantly and positively related to online impulsive buying at a moderate level. By observing the results of correlation analysis, it is easy to find that the number of studies with insignificant relationships is no more than five, and the maximum and minimum correlation coefficients for each study varied greatly. Therefore, the limited number of studies and the large differences in the effect sizes reported in the selected studies may account for the insignificant results of these four relationships.

The heterogeneity test of the effect size of our study was estimated using the Q statistic to determine whether each effect size can be merged into a new value. The result of the Q-test in Table 3 revealed that the heterogeneity of the effect sizes is significant and this confirmed the validity of choosing the random-effect model in this study. Also, the Q-values in Table 3 are significant, indicating that all the relationships have significant heterogeneity. Therefore, we can further examine the existence of moderators that affect each pair of relationships.

Publication Bias

The fail-safe N is used to test the publication bias. In Table 5, the fail-safe N of all the relationships is greater than the corresponding “5*K + 10” (K is the number of studies) standard, indicating that publication bias is not a concern (Rosenthal, 1979).

Theoretical Implications

In recent years, impulsive buying has received wide attention from consumer behavior researchers. However, the current meta-analysis research on impulsive buying mainly focused on the offline market, and few studies have involved the unique factors of the online market. This study fills this literature gap by focusing on the factors that influence online impulsive buying, especially the IT-supported website stimuli. Meanwhile, this study can extend the influence of different factors on the impulsive buying of digital products. Digital products can only be sold through online channels. Thus, the conclusions of research on offline impulsive buying behavior may not apply to these products. Therefore, compared with the traditional offline impulsive shopping behavior research, this study can be applied to a wider range of product types.

To address the research bias caused by inconsistent findings in the existing research results on online impulsive buying, this study conducted a meta-analysis to integrate 54 empirical studies and proposed a comprehensive framework for studying the influencing factors of online impulsive buying based on quantitative statistical analysis. In this research, 54 relevant empirical studies were analyzed. From these articles, we selected the factors that have been explored three times or above, and finally focused on 13 main factors. Particularly, factors were classified into the following three categories: website stimuli (website security, website navigability, website visual appeal, interactivity, and ease of use), marketing stimuli (scarcity, novelty, price, and promotion), and affective stimuli (arousal, pleasure, positive emotion, and negative emotion). According to the results of the study, these factors are significantly and positively related to online impulsive buying except website security, price, novelty, and negative emotion.

In addition, this study also verifies the applicability and effectiveness of meta-analysis methods in the field of information systems, providing new ideas and methods for related research. To our knowledge, this study is the first attempt to conduct a meta-analysis to study online impulsive buying. This study takes full advantage of the meta-analytic approach to expand the limited small sample of a single independent study into a large sample of data to verify the relationships of variables in existing empirical studies at a higher logical level. It provides effective research ideas and methods to clarify the sources of heterogeneity, avoid potential measurement errors, and then propose more credible and robust research conclusions. The results of our study will provide a theoretical basis for future research on online impulsive buying.

Further, this study selected economic development level as a moderator and categorized the sample into developed and developing countries or regions to explore the moderating effect on the relationship between influencing factors and online impulsive buying. The results of the moderator analysis illustrate that the economic development level has a significant moderating effect on the relationship between website visual appeal, ease of use, price, promotion, pleasure, positive emotion, and online impulsive buying. Specifically, in terms of website stimuli factors, consumers in developed countries or areas attach more importance to website visual appeal. Nevertheless, for consumers in developing countries or areas, the ease of use of websites is a more important factor to trigger their impulsive buying intention or behavior compared with consumers in developed countries or regions. With regard to marketing stimuli factors, consumers in developed countries or areas pay more attention to promotional factors, while consumers in developing countries or areas are more susceptible to price factors. For affective stimuli factors, this study confirmed that consumers in developing countries or areas are more likely to be stimulated by pleasure and positive emotion factors and induce online impulsive buying. With the moderator analysis, this study provides implications for future research on online impulsive buying behavior across different economic development levels.

Practical Contributions

This study also provides managerial insights for practitioners of e-commerce websites. First, managers of online stores should pay attention to the website stimuli, marketing stimuli, and affective stimuli, and take steps to optimize consumers’ online buying experience. Specifically, in terms of website stimuli, IT capacity should be enhanced. To enhance the visual appeal of websites, the navigability, interactivity, and ease of use of websites, e-commerce websites should improve the interactive design of websites and website performance. Online shopping websites can use VR and AR technology to provide a more interactive consumer experience so as to stimulate consumers to buy. For example, many online clothing retailers have begun to launch AR/VR applications such as AR shoes and virtual reality fitting mirrors. This allows customers to try on clothes virtually, greatly improving customers’ shopping experience. Besides, checking the updating of the system and software patch regularly can help to improve the consumer experience, and it would be helpful to upgrade the professional network security firewall to improve website security. With regard to marketing stimuli, besides using traditional promotion activities to persuade consumers into consumption, managers can use IT-enabled inventory availability to underline product scarcity as well as use big data techniques to conduct precision marketing. Therefore, e-commerce websites may employ hunger marketing to induce impulsive buying, use inventory availability information to emphasize the scarcity effect and maximize the marketing effect by targeting consumers. Specifically, hunger marketing is to give a surprisingly attractive price with limited quantity, creating the illusion of hot sales in short supply. Thus, companies can not only benefit from the raised price but also make the brand more appealing. Finally, for affective stimuli, managers should: (1) keep the purchasing flow smooth to provide consumers with a pleasant online buying experience, (2) design an easy-to-use interface to keep users in a good mood when shopping online, and (3) provide customized service and precision marketing to increase consumers pleasure or arousal level. For example, to provide a better shopping experience, online stores should provide quick and valid responses to consumers’ questions and requests to improve the feedback efficiency and recommend products according to consumers’ consumption history and characteristics.

Second, the moderator analysis can help cross-border e-commerce. Concretely, cross-border e-commerce practitioners should pay more attention to website visual appeal and promotion activities in developed countries or regions to increase the likelihood of online impulsive buying. In contrast, in developing countries or regions, to provide users with a pleasant buying experience, more efforts should be put into website performance optimization and product pricing strategy, simplifying the website operation process and improving user satisfaction. In short, managers should develop targeted strategies to stimulate consumption according to the economic condition of the area and the consumption level of the consumers.

Finally, with the rapid development of Internet technology and the prosperity of e-commerce, the proportion of online shopping in retailing is increasing steadily. Meanwhile, in the context of the COVID-19 pandemic, to avoid cross-infection, more and more consumers tend to buy goods online. Our research focuses on the factors that influence online impulsive shopping by improving consumers’ shopping experience, which has strong significance for serving consumers better in the context of the COVID-19 pandemic.

Limitations and Directions for Future Research

Despite attempts to conduct this meta-analysis rigorously, there are still some limitations. First, the number of relevant studies that can be used for the meta-analysis of online impulsive buying is limited. Moreover, some online impulsive buying studies were excluded because they did not provide the necessary data for statistical calculation.

Second, this study emphasized website-related factors. Although we have included the most investigated antecedents of online impulsive buying in the meta-analysis, there are other factors that may also have a significant effect. Future research could include more factors, such as consumer personality.

Third, due to the data availability issue, we took the economic development level as the proxy of consumers’ consumption level, which may not be as accurate as the real individual-level data. If possible, future research can employ individual-level data.

Fourth, due to the fact that some articles included in the meta-analysis suffer from poor data quality, for example, the collected data may be sparse, or the sample is biased, we do not have enough data to analyze the moderating effects of age, product type, context, etc. However, the moderating effects of these factors are worth studying. Therefore, in the future, researchers can conduct more in-depth and detailed studies on these factors to find boundary conditions.

Finally, only quantitative studies were used in the meta-analysis. Future research could consider weight-analysis that allows the inclusion of qualitative studies while evaluating the strength between antecedents and consequences. Moreover, researchers can attempt to use the structural equation modeling technique to test the relationship in and out of the study.