Investigating the Effects of Video-Based E-Word-of-Mouth on Consumers’ Purchase Intention: The Moderating Role of Involvement

Abstract

Short videos have been increasingly prevalent around the globe and have become an important channel for users to share product and service information and for marketers to attract potential customers. However, rarely have studies empirically examined the impact of product review videos posted on short video platforms on consumers’ purchase intention. Grounded in the elaboration likelihood model, this study proposes a research model to investigate how the product review video features (i.e., video information quality, product information visualization, video emotion polarity, and video publisher credibility) influence consumers’ purchase intention. Moreover, the moderating role of involvement (i.e., product involvement and video involvement) in the above-mentioned relationships have also been examined in this new research context. We empirically validate the research model with survey data. It is interesting to find that product information visualization and video publisher credibility are significantly and positively related to purchase intention. Video involvement negatively moderates the relationship between video publisher credibility and purchase intention. Furthermore, video emotion polarity negatively moderates the relationship between product information visualization and purchase intention. Both theoretical and practical implications are discussed.

1. Introduction

With the development of sharing economy, short videos have become increasingly popular among individuals of all age groups around the world. Given the prevalence of consumer participation and interaction, the size of consumers of short video platforms has also rapidly expanded. Short videos have thus been a breakthrough point for a new round of economic growth [1]. Short-form mobile videos refer to video applications (e.g., Instagram, Snapchat, TikTok, and Kuaishou) that enable users to shoot, edit, upload, and share short videos [2]. Users on these platforms can not only edit and upload their own created videos but also can watch, comment, forward, and like the videos shared by other users [3]. Many videos on these platforms are created by users to express their opinions or personal experiences of using a product [4]. These product review videos, also known as video-based e-Word-of-Mouth (vWOM), provide vivid demonstrations and rich product information and contain deeper emotions than other kinds of online reviews [3]. Potential customers thus would like to watch such video reviews of a product before purchasing it, especially for technical products such as smartphones, computers, and household appliances [3]. vWOMs have thus aroused increasing attention because of the large potential business values.

Existing studies suggested that online review video platforms are ideal for the eWOM of a product because those product review videos are more authentic and persuasive [3]. More specifically, many prior studies focused on investigating how the viewers’ post-video behavior (e.g., like or dislike and comments) influenced their engagement with the vWOM [5,6]. Recent studies further expanded this research area by exploring the antecedents of product review video views on video platforms [7]. However, very few studies in the literature have examined how individuals interpret the vWOM and are influenced by such vWOM. Therefore, the effects of vWOM on potential consumers’ purchase intention are still unclear. Furthermore, as a new global industry, short video platforms have been paid rarely attention from academics, though plenty of eWOM videos created to share personal experiences. Therefore, the first objective of this study is to examine the effects of vWOM on consumers’ purchase intention in the context of short videos.

Existing studies demonstrated that consumers’ purchase decisions in the purchase process could be influenced by their involvement degree [8]. More specifically, to what extent consumers gather and process product information during the purchase process depends on their involvement [9]. When consumers’ involvement is high, they would actively search for more product information and carefully process the gathered information in order to make a satisfactory decision [10]. Furthermore, existing studies have identified two kinds of consumer involvement: product involvement and contextual involvement [11]. However, rarely studies have been conducted to examine the role of the two kinds of involvement on consumers’ purchase intention in the context of vWOM. The second objective of this study is thus to investigate the moderating role of product involvement and contextual involvement (i.e., video involvement) on the relationships between vWOM features and consumers’ purchase intention.

To fulfill the above two research objectives, we proposed the following two research questions: (1) which features of vWOM exist, and how will they influence consumers’ purchase intention? and (2) how do the product and video involvement impact the relationship between vWOM features and consumers’ purchase intention? To answer these two research questions, we drew on the elaboration likelihood model (ELM) [12] to propose a research model. More specifically, we identified four vWOM features, including video information quality, product information visualization, video emotional polarity, and video publisher credibility. We then categorized them as a central route and peripheral route, respectively, based on ELM. We thus investigated the direct effects of the four features on consumers’ purchase intention and the moderating role of product and video involvement.

This study is organized as follows. In Section 2, we introduce the literature review and theoretical background. The research model and hypotheses development are illustrated in Section 3, followed by the research methodology. We analyze the data and report the results in Section 5. We discuss the implications of this study before concluding the paper.

2. Literature Review and Theoretical Background

2.1. Video-Based Electronic Word-of-Mouth and Purchase Intention

Electronic word-of-mouth (eWOM) refers to “any information, including not only customers’ own statements but also shared/forwarded posts from retailers or other published sources, which are exchanged among potential, actual, or former customers about a product or company available to a multitude of people and institutions via the Internet” [13]. Prior studies suggested that there were different formats of eWOM, such as online text reviews, picture reviews, and video reviews [14]. With the globally prevalent use of short video platforms, consumers are engaged in eWOM communication by sharing their opinions and experiences of products and services on these platforms [15]. Video-based eWOM (vWOM) has thus aroused increasing attention for digital marketing. However, consumers’ interpretations of these video reviews and the subsequent effects on their purchase intention have rarely been empirically explored.

This study defines vWOM as the video-based product reviews created by users and published on short video platforms [16]. Existing studies suggested that video reviews are more persuasive than other kinds of eWOM because of the multimedia and vivid presentation that enable video creators to share their product experiences to the audiences in a direct and detailed way [17].

Existing studies in the eWOM literature mainly investigated three aspects of eWOM, viz.: generation, adoption, and diffusion [14]. eWOM generation refers to creating and posting new product reviews; eWOM adoption refers to accepting the information of such product reviews; eWOM diffusion refers to sharing and passing along the product reviews. This study anchors into the eWOM adoption research area, where potential consumers may make their purchase decision after viewing the review videos. In this stream, many studies have examined how various eWOM factors affect consumers’ eWOM adoption and their attitudes toward the product, such as eWOM attributes (e.g., volume, valence, and quality), individual relevance, and viewers characteristics [18,19]. Among them, the most frequently examined review features are length, readability, review valence, and reviewer credibility [20]. In line with prior studies, we identified four vWOM features that may influence consumers’ attitudes and behaviors, including video information quality, product information visualization, video emotion polarity, and video publisher credibility.

It is worth noting that existing literature has prevalent refer reviews sentiment to describe the positive and negative emotions of the reviewers [21]. In this research area, prior studies have widely employed sentiment-mining tools to score the overall sentiments expressed in each text-based review [22]. In the review videos context, there are also sentiment cues for consumers, such as video title and video content [7]. The video title can capture viewers’ attention and motivate them to click on the video, which is essentially emotional [23]. Furthermore, the presenter in the video may also directly demonstrate their emotion (e.g., favorite or dislike) [16]. We thus adapted the video emotion polarity as a peripheral route factor to describe the extent to which consumers form their attitudes towards the products based on the emotional polarity expressed in the video cues.

Prior literature has identified readability as an important feature of online text reviews and has also prevalently examined its effects on consumers’ purchase intention [21]. Review readability refers to “the reading ease that improves the comprehension as well as the retention of the textual material” [24]. The degree of readability indicates the required amount of cognitive effort for a consumer of a certain age and education level to understand a text-based review [24]. In line with this research area, in the short video context, we argue that review videos conveying factual product information and high-arousal cues should be comprehended by potential consumers [7]. We thus conceptualize product information visualization by adapting the review readability into the review videos context. Typically, we define it as the extent to which consumers are able to comprehend the product information expressed in the review videos [25]. We summarized the definitions of the four features of vWOM in

Table 1. Definitions of vWOM features.

Construct	Definition
Video information quality	The extent to which the video information is perception of precision, credibility, relevance, comprehensibility, and timeliness [26].
Product information visualization	The extent to which consumers are able to comprehensively comprehend the product information expressed in the review videos [25].
Video emotion polarity	The extent to which consumers are attracted or affected by the emotion polarity (positive emotion) toward the product in the video [27].
Video publisher credibility	The extent to which consumers perception of the video publisher is credible [28].

.

Purchase intention demonstrates the possibility of a consumer willing or planning to purchase a product or service [29]. Prior studies indicated that a high level of consumers’ purchase intention leads to a high possibility of actual buying [30]. In the context of short videos, purchase intention refers to the degree to which consumers desire to make a purchase after viewing the review videos [31].

Most of the existing studies in the eWOM literature focused on the effects of eWOM on consumers’ purchase intention [30]. One stream of studies examined the direct effects of eWOM on purchase intention and found that positive (negative) eWOM increases (reduces) purchase intention [32]. Moreover, prior studies found that negative eWOM has a more significant influence on customers’ purchase intention than positive eWOM [33]. Another stream of studies investigated the indirect relationships between eWOM and purchase intention by identifying various mediators such as value co-creation, trust, and flow experience [12,30,34]. The third stream of studies explored the antecedents of eWOM that further influence purchase intention (e.g., personal value) [35]. However, previous studies have rarely conducted research in the product review videos context.

2.2. Elaboration Likelihood Model (ELM)

ELM provided a solid theoretical framework to explain the underline processes of persuasive communications [36]. More specifically, ELM suggested two routes to persuasion. The first route is the central route that individuals take careful and thoughtful consideration of the truth of information presented in the online reviews, which demands high cognitive effort and has a persistent impact on individual attitudes towards the arguments. The other route is a peripheral route in that the individual intuitively judges the information presented without the necessity of scrutiny of the truth, which is automatic and heuristic and has a short-term influence on individuals’ attitudes. Furthermore, both central and peripheral routes process the received information concurrently and jointly influence individuals to be persuaded or not [37].

Based on their systematic review, Chou et al. [20] demonstrated that quality-related features of online reviews (e.g., reviews readability) are processed by the central route, and the non-quality related features (e.g., review sentiment) are processed by the peripheral route. Consistently, in their systematic review, Shahab et al. [38] also suggested that the factors influencing the central route are information-quality-related, and the factors influencing the peripheral route include source credibility, the attractiveness of sources, and arguments numbers.

In line with prior literature, we thus identify video information quality and product information visualization as the factors of the central route. We identify video publisher credibility and video emotion polarity as the factors of the peripheral route, whereas the video publisher credibility is a factor of source credibility. It is worth noting that the video publisher may be the creator of the videos or the person who forwards the videos from others on the short video platform.

2.3. Consumers Involvement

Based on ELM, consumers may process the same information differently in terms of their involvement [39]. Involvement refers to the degree to which consumers perceive the relevance of an object based on their values, needs, and interest [40]. Therefore, involvement is related to consumers’ motivation to process information. Existing studies suggested that there are two kinds of involvement: enduring product involvement and situational/contextual involvement [11,41]. Enduring product involvement indicates the consumers’ personal needs or interest in the product, which is a stable process over a long period. Situational/contextual involvement refers to a temporary and fluctuated elevation of personal interest within a short time frame of a purchasing decision.

Prior studies have examined the moderating effects of product involvement on the relationships between eWOM and consumers’ attitudes toward the products [42,43]. Product involvement is associated with personal characteristics (e.g., values) and thus can be different even for the same product. Furthermore, existing studies also demonstrated the important role of situational involvement (e.g., online involvement) in the relationships between eWOM and consumer attitude and purchase decisions [39,44].

Based on the existing literature, this study thus takes both product involvement and situational/contextual involvement into consideration and further examines their moderating roles in the relationships between vWOM features and consumers’ purchase intention. More specifically, the features of video (e.g., auditory loudness, hue, and music style) may influence consumers’ interpretation of the video content information, which further influences their motivation to view and be persuaded [7]. We thus identify video involvement as a distinct situational factor in the short video context and define it as the degree to which consumers perceive that the video cues are relevant to their preferences and interests. When video involvement is high, consumers are more motivated to spend more cognitive effort to process the video information. In the next section, we illustrate the proposed research model and hypotheses development.

3. Research Model and Hypothesis Development

Based on ELM and the literature review in the preceding section, this study investigates the features of product review videos on consumers’ purchase intention and the moderating role of both product and video involvement in the above-mentioned relationships. The proposed research model is shown in Figure 1.

3.1. Relationships between Central Route Factors and Purchase Intention

Existing studies in the context of online reviews suggested that high-quality reviews have a greater effect on consumers’ purchase intention than low-quality reviews [45]. Furthermore, high-quality information makes consumers perceive of narrowed social psychological distance between them and information sources, which subsequently improves their purchase intention [46].

The video information quality feature of vWOM refers to the extent to which the video information is the perception of precision, credibility, relevance, comprehensibility, and timeliness [26]. Prior studies demonstrated that video reviews provide story-based communication and are more comprehensive and timeliness in nature than other kinds of reviews [3]. More specifically, product review videos can provide richer and more realistic information about the products because these review videos show consumers’ own consumption experiences [47]. Therefore, the user-generated review videos are believed to be credible and persuasive [48]. As such, consumers would be more motivated to view and spend cognitive resources to find and process the information presented. The more high-quality information they obtain, the more they are willing to purchase. We thus hypothesize that:

Hypothesis 1 (H1).

Video information quality positively influences consumers’ purchase intention.

In the review videos context, product information visualization refers to the extent to which consumers are able to comprehensively comprehend the product information expressed in the review videos [25]. We argue that product information visualization leads to a higher level of perception of credibility, helpfulness, and persuasiveness and consequently positively affects consumers’ purchase intention. Prior studies suggested that online reviews presented in a different format (i.e., text, image, and video) have various effects on consumers’ perception of credible, helpful, and persuasive [49]. As a rich media, the video format could provide high-quality information, subjective knowledge, and vivid emotional cues than the text format, which in turn significantly influences consumers’ purchase intention [50]. Moreover, recent empirical evidence has shown that video-based eWOM had the greatest impact on consumers’ purchase intention due to meeting the needs of consumers for more product information, followed by image-based and text-based eWOM [51]. As such, product review videos provide intuitive and multisensory information that enables consumers to easily comprehend and make purchase decisions. We thus hypothesize that:

Hypothesis 2 (H2).

Product information visualization positively influences consumers’ purchase intention.

3.2. Relationships between Peripheral Route Factors and Purchase Intention

Video emotional polarity refers to the extent to which consumers are attracted or affected by the emotional polarity (positive emotion) toward the product in the video [27]. Existing studies suggested that positive eWOM would significantly increase consumers’ purchase intention, and negative eWOM would significantly reduce it [52]. Moreover, the proportion and quality of negative eWOM could be a central cue to high-involvement consumers [53]. In the context of vWOM, Agrawal and Mittal [3] analyzed the text comments of popular review videos on YouTube and found that the over-sentiment expressed in product review video comments significantly influenced consumers’ purchase intention. Moreover, a positive sentiment expressed in review video comments positively affected purchase intention, and negative sentiment adversely influence purchase intention. We thus argue that when consumers perception of high levels of video emotion polarity, measured as the extent to which they attracted by the positive emotion expressed in the reviews video, the consumers may have high levels of purchase intention. We thus hypothesize that:

Hypothesis 3 (H3).

Video emotional polarity positively influences consumers’ purchase intention.

Existing studies indicated that information source credibility was more persuasive and had a more positive influence on consumers’ attitudes [54]. Video publisher is related to source credibility and further influence consumers’ perception of review helpfulness and their attitude towards the products [55]. More specifically, when consumers perceived that the review publisher’s expertise was high, they would more actively seek eWOM from this publisher and further increase their purchase intention [56]. We thus argue that perceived video publishers’ expertise and trustworthiness would increase potential consumers’ purchase intention.

Hypothesis 4 (H4).

Video publisher credibility positively influences consumers’ purchase intention.

3.3. Moderating Role of Product Involvement

Product involvement plays a crucial role in understanding consumers’ information processing model [57]. Based on ELM, consumers with a high level of product involvement would be more familiar with the product information and pay more attention to the persuasive content. In this situation, the central route would be salient in influencing consumers’ purchase intention. When consumers had a low level of product involvement, they would not spend much cognitive effort on analyzing and processing the information content but would focus on the peripheral cues of the information, such as attractiveness, credibility, and reliability of the presented information [10,58]. As such, individuals with high product involvement evoke more elaborate cognitive processing and consequently form positive perceptions of the product [59]. When the product involvement is low, customers make purchase decisions mainly according to non-compensatory incomplete processing of attributes and investing less cognitive resources [8]. We thus hypothesize that:

Hypothesis 5a,b (H5a,b).

Product involvement positively moderates the relationships between video information quality (a)/product information visualization (b) and purchase intention; such that the higher the product involvement, the stronger the relationship between video information quality (a)/product information visualization (b) and purchase intention.

Hypothesis 5c,d (H5c,d).

Product involvement negatively moderates the relationships between video emotion polarity (c)/video publisher credibility (d) and purchase intention; such that the higher the product involvement, the weaker the relationships between video emotion polarity (c)/video publisher credibility (d) and purchase intention.

3.4. Moderating Role of Video Involvement

Although much attention has been paid to the moderating role of product involvement [53], recent studies also suggested that consumers’ purchase decisions may also be influenced by the context/situation involvement [44]. In the context of product review videos, we argue that video involvement would moderate the relationships between vWOM features and consumers’ purchase intention.

Potential consumers may collect information cues in the video to reduce their perception of purchase risks [11]. Consumers with high video involvement would thus pay more cognitive resources on and more carefully process the video information [44]. As such, they would be more able to interpret the video information cues, such as video information quality and visualized product information, which may reduce their perception of risks and improve their purchase intention. Furthermore, consumers with a lower level of video involvement would have a high level of concern for the quality of the vWOM, and their attention to vWOM is also low. As such, consumers may make purchase decisions according to the peripheral cues, such as the credibility of sources and their perception of video emotion polarity. Therefore, we hypothesize that:

Hypothesis 6a,b (H6a,b).

Video involvement positively moderates the relationships between video information quality (a)/product information visualization (b) and purchase intention; such that the higher the video involvement, the stronger the relationships between video information quality (a)/product information visualization (b) and purchase intention.

Hypothesis 6c,d (H6c,d).

Video involvement negatively moderates the relationship between video emotion polarity (c)/video publisher credibility (d) and purchase intention; such that the higher the video involvement, the weaker the relationship between video emotion polarity (c)/video publisher credibility (d) and purchase intention.

4. Research Methodology

4.1. Measurement Development

All constructs and measures in this study were adapted from validated research in existing literature. Video information quality was measured using a four-item scale adapted from Gao and Bai [60], and sample times were “The information presented in the product review videos is what I need” and “The information presented in the product review videos is comprehensive”. The four items of product information visualization were adapted from Gefen and Straub [61] and Dutta-Bergman [62]. One sample item was “There are pretty much video publishers’ product usage experiences presented in the review videos”. Video emotional polarity was measured with two items adapted from Zhu et al. [27], including “I’m more likely to be attracted to review videos that are more positive about a product” and “I pay more attention on the review videos that are more negative about a product”. Video publisher credibility was measured using a four-item scale adapted from Ohanian [28], and one sample item was “I think the video publisher is trustworthy”. We measured product involvement using a four-item scale adapted from Zaichkowsky [63], and one sample item was “I think the product is important to me”. Video involvement was measured using a four-item scale adapted from Dutta-Bergman [62], and one sample item was “I put a lot of effort in evaluating the arguments presented in the review videos”. Purchase intention was measured using three items adapted from Zeithaml et al. [64], and one sample item was “I consider buying the product introduced in the review videos while viewing them”. Following the approach suggested by [65], we translate the English questions into Chinese for final data collection. The seven-point Likert scale (ranging from 1 = strongly disagree to 7 = strongly agree) was used to measure the items.

4.2. Data Collection

To verify the proposed research model, we collected data using survey approach. We employed the online services of a large survey company in China (http://www.wjx.cn, accessed on 18 May 2022) to develop the questionnaire link. The respondents of this study are TikTok short video application users from China who are consumers of technical products (e.g., smartphones and laptops). TikTok has operated in 155 countries around the globe [66]. It reported that TikTok, together with its China counterpart Douyin, had roughly one billion monthly active users and became the top rank of global digital platforms [66]. Its commercial success thus becomes critical to the long-run corporate value. This study is thus focused on the context of TikTok, which is also consistent with our research objectives. We thus firstly sent the questionnaire link to several targeted TikTok users. Then, the questionnaire link was disseminated using snowball method. Snowball sampling has been widely used in existing literature, and its advantages are that it can accurately target the respondents and thus will reduce the difficulty of locating the respondents and, in turn, substantially reduce the survey cost [67]. The snowball sampling is appropriate for this study because we target the TikTok users in particular rather than a sample that includes users from multiple short-video platforms. To further ensure the respondents are TikTok short video platform users, we added the screening question (“Are you a short-video platform user (TikTok)?”) before the formal question survey and only those with prior short-video experiences were eligible. After three weeks, we received 227 responses in total. We dropped the data that answered carelessly, resulting in a total of 169 valid responses.

Table 2. Demographic characteristics (N = 169).

Category		N (%)
Gender	Male	110 (65.1)
	Female	59 (34.9)
Education	High school or below	12 (7.1)
	College	12 (7.1)
	University or above	145 (85.8)
Age	Under 20	9 (5.3)
	21–25	107 (63.3)
	36–40	2 (1.2)
	41 and above	51 (30.2)

lists the demographic characteristics of the sample.

Following the suggestions of Armstrong and Overton [68], we estimated non-response bias. For all the constructs, we compared the early 25% responses and late 25% responses. Results showed that there were no significant differences between the two groups in terms of construct means (p > 0.10), indicating that non-response bias was not a serious concern in this research.

4.3. Common Method Bias (CMB)

Two methods were used to test CMB. First, Harman’s single-factor test was employed. The results showed that the eigenvalues of all the constructs were higher than the threshold value of 1.0. The first construct accounted for 19.50% of the total variance, which was less than the cut-off value of 50% [69].

Second, following the procedure suggested by Podsakoff et al. [70] and Williams et al. [71], the common method factor approach was used to further test CMB. More specifically, we evaluated how each variance of indicators was substantively explained by all principal constructs and the method factor, and the loadings of indicators on the method factor and on their substantive factors. As the results show in

Table 3. Results of common method bias test.

Construct	Indicator	Substantive Factor Loading (R1)	R12	Method Factor Loading(R2)	R22
Video Information Quality (VIQ)	VIQ1	0.883 ***	0.780	−0.079	0.006
	VIQ2	0.935 ***	0.874	−0.060	0.004
	VIQ3	0.738 ***	0.545	0.025	0.001
	VIQ4	0.690 ***	0.476	0.125	0.016
Video Emotional Polarity (VEP)	VEP1	0.927 ***	0.859	−0.008	0.000
	VEP2	0.918 ***	0.843	0.008	0.000
Video Information Visualization (VIV)	VIV1	0.821 ***	0.674	−0.022	0.000
	VIV2	0.746 ***	0.557	0.041	0.002
	VIV3	0.897 ***	0.805	−0.016	0.000
	VIV4	0.821 ***	0.674	0.000	0.000
Video Publisher Credibility (VPC)	VPC1	0.842 ***	0.709	0.041	0.002
	VPC2	0.902 ***	0.814	0.013	0.000
	VPC3	0.877 ***	0.769	0.021	0.000
	VPC4	0.988 ***	0.976	−0.072	0.005
Video Involvement (VIN)	VIN1	0.837 ***	0.701	0.017	0.000
	VIN2	0.917 ***	0.841	−0.065	0.004
	VIN3	0.863 ***	0.745	0.063	0.004
	VIN4	0.892 ***	0.796	−0.017	0.000
Product Involvement (PIN)	PIN1	0.945 ***	0.893	−0.035	0.001
	PIN2	0.946 ***	0.895	0.004	0.000
	PIN3	0.874 ***	0.764	0.032	0.001
Purchase Intention (PUI)	PUI1	0.841 ***	0.707	0.116 **	0.013
	PUI2	0.981 ***	0.962	−0.050	0.003
	PUI3	0.999 ***	0.998	−0.065	0.004
Average		0.878	0.777	0.001	0.003

Note: ** p < 0.01, *** p < 0.001.

, the substantive factor explained 77.7% of the variance on average, and the average method factor variance of the indicators was 0.3%. The ratio between the average substantive factors and method factor variance is very large. Therefore, CMB was not a serious issue in our study.

5. Data Analysis and Results

5.1. Measurement Model

We employed confirmatory factor analysis to verify the reliability and validity of the constructs. Cronbach’s alpha and composite reliability were used to assess the reliability of all the constructs [72]. As the results show in

Table 4. Results of confirmatory factor analysis.

Construct	Indicator	Loading	Cronbach’s Alpha	Composite Reliability	Average Variance Extracted
Video Information Quality (VIQ)	VIQ1	0.814	0.830	0.887	0.662
	VIQ2	0.868
	VIQ3	0.744
	VIQ4	0.825
Video Emotional Polarity (VEP)	VEP1	0.916	0.825	0.919	0.851
	VEP2	0.929
Product Information Visualization (PIV)	PIV1	0.791	0.840	0.893	0.677
	PIV2	0.763
	PIV3	0.894
	PIV4	0.837
Video Publisher Credibility (VPC)	VPC1	0.865	0.924	0.946	0.815
	VPC2	0.915
	VPC3	0.904
	VPC4	0.925
Video Involvement (VIN)	VIN1	0.829	0.900	0.930	0.769
	VIN2	0.867
	VIN3	0.917
	VIN4	0.892
Product Involvement (PIN)	PIN1	0.917	0.911	0.944	0.850
	PIN2	0.946
	PIN3	0.902
Purchase Intention (PUI)	PUI1	0.937	0.935	0.958	0.885
	PUI2	0.939
	PUI3	0.947

, Cronbach’s alpha scores ranged from 0.825 to 0.935, and the composite reliability ranged from 0.887 to 0.958, which are all higher than the threshold value of 0.7 [72]. These results indicated the goodness of reliability.

Both convergent and discriminant validities were tested. As shown in Table 4, all the factor loadings were above the suggested value of 0.7, and all the values of average variance extracted (AVE) were above the benchmark value of 0.5 [72]. These results demonstrated the good convergent validity of the measurement model. As the results show in

Table 5. Means, standard deviation, correlations matrix and Heterotrait–Monotrait ratios.

	Mean	S.D.	VIQ	VEP	PIV	VPC	VIN	PIN	PUI
VIQ	4.997	1.062	0.814
VEP	4.787	1.320	0.320 (0.394)	0.922
PIV	4.644	0.981	0.623 (0.757)	0.310 (0.377)	0.823
VPC	5.260	1.082	0.686 (0.782)	0.469 (0.539)	0.636 (0.718)	0.903
VIN	4.231	1.251	0.401 (0.461)	0.410 (0.484)	0.384 (0.444)	0.378 (0.409)	0.877
PIN	4.357	1.049	0.522 (0.605)	0.248 (0.284)	0.465 (0.532)	0.396 (0.432)	0.635 (0.708)	0.922
PUI	4.836	1.303	0.537 (0.599)	0.315 (0.357)	0.617 (0.692)	0.677 (0.721)	0.492 (0.529)	0.443 (0.479)	0.941

Note: VIQ represents video information quality; PIV represents product information visualization; VEP represents video emotion polarity; VPC represents video publisher credibility; VIN represents video involvement; PIN represents product involvement; PUI represents purchase intention.

, the square roots of AVEs for all constructs in the diagonal row were higher than the inter-construct correlations [72]. In addition, we further tested discriminant validity by calculating the Heterotrait–Monotrait (HTMT) ratios [73]. As shown in Table 5, all the HTMT values were lower than the suggested score of 0.85, indicating the discriminant validity was reaffirmed. These results demonstrated the good discriminant validity of the measurement model.

Considering that a few correlations between constructs were greater than 0.6, we then tested the potential multicollinearity problem. According to Mason and Perreault [74], the variance inflation factor (VIF) was used to test collinearity: when VIFs are above 10, multicollinearity exists. The results showed that the value of the highest VIF was 4.399. Thus, multicollinearity was not a serious concern.

5.2. Structural Model

The hierarchical ordinary least squares regression analysis was employed to test the structural model. Hierarchy regression analysis is more suitable for models with multiple moderating effects [75]. In addition, hierarchical regression analysis may overcome the drawbacks of PLS in that the strength of the relationships is overestimated, and the significance of the relationships is underestimated [76]. We mean-centered the data before analysis to minimize the potential multicollinearity issue [77].

The hierarchical regression analysis results were summarized in

Table 6. Hierarchical regression analysis results of direct effects.

Construct	DV = Purchase Intention
	Model 1	Model 2	Model 3
Gender	0.216 **	0.291 ***	0.272 ***
Age	0.027	0.013	0.002
Education	0.144	−0.031	−0.021
VIQ (H1)		0.077	0.137
PIV (H2)		0.382 ***	0.449 ***
VEP (H3)		0.057	0.030
VPC (H4)		0.327 ***	0.212 *
VEP × VIQ			0.358 ***
VEP × PIV			−0.359 ***
VPC × VIQ			−0.227
VPC × PIV			0.175
R2	0.079	0.579	0.630
△R2		0.499	0.051
F (p-value)	0.079 **	47.703 ***	5.388 ***

Note: VIQ represents video information quality; PIV represents product information visualization; VEP represents video emotion polarity; VPC represents video publisher credibility. * p < 0.05, ** p < 0.01, *** p < 0.001.

and

Table 7. Hierarchical regression analysis results of moderating effects.

Construct	DV = Purchase Intention
	Model 1	Model 2	Model 3	Model 4
Gender	0.216 **	0.291 ***	0.274 ***	0.231 ***
Age	0.027	0.013	−0.019	0.009
Education	0.144	−0.031	0.026	0.050
VIQ		0.077	0.031	0.034
PIV		0.382 ***	0.302 ***	0.252 ***
VEP		0.057	−0.031	−0.022
VPC		0.327 ***	0.334 ***	0.290 ***
PIN			0.010	0.077
VIN			0.280 ***	0.351 ***
PIN × VIQ (H5a)				−0.310 **
PIN × PIV (H5b)				0.038
PIN × VEP (H5c)				−0.141 *
PIN × VPC (H5d)				0.159
VIN × VIQ (H6a)				0.203 *
VIN × PIV (H6b)				0.045
VIN × VEP (H6c)				0.112
VIN × VPC (H6d)				−0.376 **
R2	0.079	0.579	0.633	0.725
△R2		0.499	0.054	0.092
F (p-value)	4.742 **	47.702 ***	11.812 ***	6.306 ***

Note: VIQ represents video information quality; PIV represents product information visualization; VEP represents video emotion polarity; VPC represents video publisher credibility; VIN represents video involvement; PIN represents product involvement. * p < 0.05, ** p < 0.01, *** p < 0.001.

. In Table 6, we focused on examining the direct effects of the central route and peripheral route factors on purchase intention and also tested the moderating effects of peripheral route factors on the relationships between central route factors and purchase intention. The control variables were included in model one, the independent variables and moderators were included in model two, followed by the interaction terms in model three, respectively.

As the results show in Table 6, the identified four vWOM features accounted for 57.9% of the total variances. Product information visualization was significantly and positively related to consumers’ purchase intention (β = 0.382, p < 0.001), supporting H2. Video publisher credibility was significantly and positively related to purchase intention (β = 0.327, p < 0.001), supporting H4. However, the relationships between video information quality (β = 0.077, p > 0.05) and video emotional polarity (β = 0.057, p > 0.05) and purchase intention were nonsignificant, rejecting H1 and H3. Furthermore, we also tested the moderating role of video emotional polarity and video publisher credibility on the relationships between video information quality and product information visualization. It was interesting to find that video emotional polarity significantly and negatively moderated the relationship between product information visualization and consumers’ purchase intention (β = −0.359, p < 0.001).

In Table 7, we tested the moderating effects of product involvement and video involvement on the relationships between central/peripheral route factors and consumers’ purchase intention. The control variables were included in model one; the independent variables were included in model two; the moderators were added in model three; followed by the interaction terms in model four, respectively.

As the results show in Table 7, by adding the interaction effects of product/video involvement into the model, 72.5% of the total variance of purchase intention was explained. The moderating effect of video involvement on the relationship between video publisher credibility and purchase intention was significantly negative (β = −0.376, p < 0.01), supporting H6d. However, although the interaction effects between product involvement and video information quality/video emotion polarity and between video involvement and video information quality were significant, the direct effects between video information quality/video emotion polarity and purchase intention were statistically non-significant, thus rejecting H5a, H5c, and H6a. In addition, H5b, H5d, H6b, and H6c were also rejected due to the non-significant interaction effects.

To further interpret the results, we plotted the moderating effects of video emotion polarity and video involvement. A high and low level of a moderator (i.e., video emotion polarity and video involvement) are indicated by the scores one standard division above and below the mean [77]. The moderating effect of video emotion polarity on the relationship between product information visualization and purchase intention was plotted in Figure 2. The effect of product information visualization on purchase intention is stronger with a low level of video emotion polarity than that with a high level of video emotion polarity. The moderating effect of video involvement on the relationship between video publisher credibility and purchase intention was plotted in Figure 3. The effect of video publisher credibility on consumers’ purchase intention is stronger with a low level of video involvement than with a high level of video involvement.

5.3. Qualitative Results

To further interpret the above results, we conducted a follow-up interview. More specifically, we aim to further understand the following two issues in this interview: What are the reasons for the insignificant effects of video information quality and video emotion polarity on consumers’ purchase intention (i.e., H1 and H3)? and What are the reasons of the insignificant moderating effects of product involvement (i.e., H5a–d)?

To answer the above two research questions, we conducted a semi-structured interview with 17 respondents randomly selected from the previous survey study sample. The sampled interview questions include “When you want to by a technical product (e.g., smartphone, laptop), will you watch the related product review videos on the short video platforms? How do you feel about the video information quality? How and to what extent the review videos related to your needs, values, and interest of the products?”

First, more than half of the respondents said that the product review videos they viewed were recommended by the short videos platform rather than searched by themselves, and thus they perceived that the information quality of the review videos is uneven. One of the respondents said that “I usually watch the review videos recommended by the platform, but I would trust more on the product information that I searched by myself”. These results suggested that potential consumers are more reliant on their own searched information rather than the recommended information by the short videos platform, which may result in the non-significant effects of video information quality and video emotion polarity on purchase intention.

Second, twelve of the respondents claimed that when the product is important to them, or they are very interested in it (that is, the degree of product involvement is high), the product review videos on the short video platforms were not the main references of their purchase decision. One of the respondents pointed out that “when I am interested in the product, I would like to firstly search the information online and also carefully read the online text-based reviews in a particular online store (e.g., the brand official online store). After that, I usually do not need to view the reviews videos”. In addition, fourteen of the respondents suggested that when they were highly involved with the products, they usually had a better understanding of the products in terms of basic functions and characteristics of the product. These results indicated that consumers with high product involvement may have already obtained familiarity with the product and thus tend to search text-based information, which requires high cognitive effort for further purchase decision-making, rather than rely on the shared experiences from others, consequently leading to the insignificant moderating effects of product involvement in the review videos context.

6. Discussion, Implications, and Future Research

6.1. Key Findings

This study has the following key findings. First, both product information visualization (central route factor) and video publisher credibility (peripheral route factor) positively and significantly influence consumers’ purchase intention. These results are consistent with existing literature that eWOM source credibility and the trustworthiness of product information positively affected consumers’ intention to buy [78]. However, it is interesting to find that the direct impacts of video information quality (central route factor) and video emotional polarity (peripheral route factor) on purchase intention are non-significant. Prior studies suggested that the positive sentiment expressed in product reviews video comments significantly and favorably influenced consumers’ purchase intention, and the negative sentiment expressed in product review video comments significantly and adversely affected consumers’ purchase intention [3]. This study thus complemented this research area by examining the effects of perceived video emotional polarity and providing nuanced evidence.

Second, we found that video involvement negatively moderates the relationship between product information visualization and purchase intention. This result is consistent with the findings in prior studies that suggested providing a pleasant shopping environment with rich media (e.g., review videos) increased consumers’ perceived amount of information. Consumers who perceive more information may perceive a high level of external environmental stimulation and, subsequently, distract their attention capability for the decision tasks [79]. Moreover, the results showed that video emotional polarity negatively moderated the relationship between product information visualization and purchase intention. That is, the consumers who are rational rather than emotional would focus more on comprehensively comprehending the product information in the video, which facilitates their purchase intention [51].

Third, it is also interesting to find the non-significant effects of video information quality and video emotion polarity and the non-significant moderating effects of product involvement. We further examined these effects by conducting follow-up interviews with the survey respondents. The possible explanations are: consumers may perceive that the reviews videos they view were recommended by the platform and thus may attribute them as discredited commercial advertisements. Moreover, consumers with high product involvement tend to search text-based information by themselves to make a purchase decision, leading to the insignificant moderating effects of product involvement. These results also demonstrated that the product review videos might play different roles in different stages of consumers’ purchase decisions, which are worthy of future exploration.

6.2. Theoretical Implications

This study contributes to the existing literature in the following aspects. First, our study identified four features of product review videos and examined their impacts on consumers’ purchase intention. This research is thus responding to the research calls in the existing literature that more research on user-generated product review videos and how they influence consumers’ purchase behavior [47,49].

Second, this study identified two kinds of involvement in the review videos context, viz.: video involvement and product involvement, and further investigated their moderating effects. We found that when adding video involvement into the model, the moderating effects of product involvement are non-significant. This study thus contributes to existing eWOM literature by demonstrating the distinct role of contextual involvement (i.e., video involvement) and was also provided fresh and nuanced empirical evidence for understanding the boundary effects of review videos [17,51].

Third, this study also contributes to the ELM model by examining the moderating effects of the peripheral route factor (i.e., video emotion polarity) on the relationship between the central route factor (i.e., product information visualization) and consumers’ purchase intention. Prior studies demonstrated that central and peripheral routes jointly influenced consumer purchase decisions [37]. Rarely have studies investigated the interaction effects of these two routes on consumers’ purchase intention, and this study thus provided empirical evidence of the interaction effects in the review videos context.

6.3. Practical Implications

The findings of this study have several implications for practitioners. First, this study found that the total variances of purchase intention explained by product review video features are large. Therefore, we suggest that rather than only focusing on attracting potential consumers, practitioners should also find effective ways to improve the transformation rate from attracted potential consumers to actual customers [3].

Second, our results also suggested the crucial effects of product information visualization and video publisher credibility on consumers’ purchase intention. We thus suggested that marketers visualize the product information by matching consumers’ preferences. Furthermore, we also suggested that short video platforms strengthen the censorship mechanism for video publishers.

Third, we found that video involvement played a more important role in influencing consumers’ purchase intention than product involvement in the product review videos context. We thus suggested marketers and video creators should effectively link their video content with consumers’ experiences and the needs of products to improve the market value of these product review videos [7].

6.4. Limitations and Future Research

Despite the findings of this study deeper our understanding of video-based word-of-mouth, there are several limitations that should be addressed in future studies. First, this study validated the proposed research model with cross-section data points. However, consumers’ purchase intention on the short video platforms is an ongoing phenomenon, given the dynamic and vivid nature of review videos. Future studies could develop a longitudinal research design to further test the research model and capture the dynamic relationship between vWOM and purchase intention. Second, this study collected data through snowball sampling that may lack randomness. Future studies could collect more data from multiple sources to validate the research model. Third, other remaining unexplained features of vWOM (e.g., video optimization practices feature) and consumers’ characteristics may impact the proposed relationships. Future research thus could explore additional moderating effects [17].

7. Conclusions

This study investigated how the key features of vWOM influence consumers’ purchase intentions and the moderating role of video and product involvement on the relationships between vWOM features and purchase intention. The results indicated that product information visualization and video publisher credibility exert significant positive effects on purchase intention. Moreover, video involvement negatively moderates the relationship between video publisher credibility and purchase intention, and video publisher credibility negatively moderates the relationship between product information visualization and purchase intention. These interesting findings provided nuanced empirical evidence and enriched our understanding of video-based word-of-mouth.