Harnessing Empathy: The Power of Emotional Resonance in Live Streaming Sales and the Moderating Magic of Product Type

Abstract

The emotional expressions in live streaming e-commerce possess a strong contagious effect, enabling viewers to easily resonate with the specific emotions conveyed by the streamers and consciously build an empathy transmission chain. This study constructs a regression model based on the emotional contagion theory and explores the impact of empathy between streamers and viewers on sales performance. Using data from 30 live streams, totaling 22,707 min, from one of China’s most popular live streaming rooms, “East Buy”, between February and April 2024, we demonstrate the significant positive impact of empathy between streamers and viewers on sales. Additionally, product type positively moderates this relationship. The unexpected thing is that live streaming time does not significantly affect the relationship between empathy and sales. This study employs text sentiment analysis methods to extract emotional features from the streamers’ speech and real-time comments from viewers. Our research extends the application of emotional contagion theory to the context of live-streaming e-commerce, enriches the literature on emotional interaction in service marketing, and provides practical insights for live-streaming platforms and streamers. Streamers can optimize marketing strategies and achieve sales goals by creating a more engaging and empathetic live-streaming experience.

1. Introduction

Live commerce is a new e-commerce marketing model where streamers promote products or services to users through interactive videos on online platforms [1]. In recent years, traditional overseas e-commerce platforms, social media platforms, and Chinese internet companies expanding abroad have launched live commerce features. Multiple overseas platforms have seen explosive growth in live commerce traffic, with users in regions such as North America, Latin America, and Southeast Asia developing a habit of shopping via live streams [2]. In 2023, live commerce platforms’ global market size and revenue share were approximately USD 908 million, expected to reach USD 4.88 billion by 2032 [3]. For example, the live commerce market size reached CNY 4.9 trillion in China, with a year-on-year growth rate of 35.2%. After five years of rapid development, the industry has entered a stable growth stage [4]. Unlike traditional commerce models, live commerce’s high interactivity and immediacy enhance consumer engagement, improve purchase decision efficiency, and provide an immersive shopping experience. By watching live streams, consumers can intuitively understand product features and real-world performance [1,5]. Streamers and viewers interact in real-time through various media forms, such as video and text, offering merchants a new way to connect with users and instantly promote products [6]. As a convergence point of technology and marketing, live commerce constructs a virtual consumption space that includes streamers, viewers, and products.

As live commerce rapidly evolves, content-oriented live streaming rooms increasingly demonstrate their unique significance. Compared to sales-driven live streaming rooms, content-oriented ones prioritize providing valuable content and attracting viewers through knowledge sharing, entertainment interactions, and life experiences, thereby indirectly promoting product sales. The focus on “content” in live commerce signifies the refinement and differentiation of the live economy [7]. The synergy between content and live commerce extends beyond the live streaming room and permeates nearly every stage and scenario. The industry’s logic shifts from the “attention economy” to the “sharing economy”.

Empathy is how individuals comprehend and infer others’ emotions and feelings while experiencing similar emotional responses [8]. Empathy involves emotional contagion, which encompasses the alignment with another person’s emotional state [9]. For instance, when a colleague blushes due to a mistake, sadness is felt on their behalf. When a friend rejoices at receiving praise, pride is felt for them. Emotional contagion theory suggests that individuals unconsciously imitate others’ emotional expressions, such as facial expressions, tone of voice, posture, and behavior. This mimicry is a pervasive natural reaction, leading to uniformity in the perception and expression of emotions among humans [10]. In service contexts, empathy assists employees in better understanding customer needs, enhancing group relationships, and influencing potential user behavior [11,12]. As a modern form of interpersonal communication, live streaming allows hosts to interact with viewers in real time. Hosts with high empathy skills are better equipped to understand and respond to viewers’ emotional needs, facilitating efficient communication, fostering closer emotional bonds, and creating a virtuous cycle.

The theory of synchronicity posits that physiological and psychological states follow rhythmic and cyclical patterns. Optimal cognitive and behavioral performance occurs during circadian rhythm peaks or when these peaks align with decision-making times [13]. Differences in live-streaming experiences, attention, emotions, and decisions between day and night are attributed to the human biological clock and individual time allocation. For instance, the desire for variety varies throughout the day, increasing with rising arousal levels [14]. This rhythmic and cyclical variation significantly affects live streaming experiences. During the day, viewers possess ample attention and cognitive resources, making them more inclined to engage with information-dense content, thus facilitating rational analysis and decision-making. At night, viewers’ physiological and psychological states shift; attention and cognitive resources may diminish, while emotional needs and desires for instant gratification increase [15]. Consequently, viewers tend to select content that offers emotional resonance and immediate satisfaction. Hosts can attract viewers and promote purchases by enhancing emotional interactions and creating a relaxed, enjoyable atmosphere.

Additionally, products are an indispensable factor influencing consumer purchases in live streaming. The categorization of products into search goods and experience goods based on their attributes has been widely discussed and accepted in previous studies [16,17]. Search goods are those for which consumers can obtain information about performance and suitability before purchase. In contrast, experience goods have vital attributes that are unknown or difficult to assess without direct experience [18]. In the live streaming environment, the mechanisms by which search goods and experience goods influence consumer purchasing behavior differ. For search goods, consumers typically gather and compare extensive information before purchasing to ensure the product meets their needs. Live streamers can help consumers obtain the necessary information and boost their confidence in purchasing by providing detailed product demonstrations, performance explanations, and user reviews.

In contrast, the purchase decisions for experience goods rely more on consumers’ emotional experiences and subjective perceptions. Without direct experience, consumers find it challenging to understand the quality and suitability of the product thoroughly. Therefore, live streamers can aid consumers in better perceiving the characteristics and value of experience goods through vivid descriptions, live demonstrations, and emotional interactions. Thus, the research questions posed in this study are:

RQ1: How does emotional resonance between streamers and viewers affect sales performance?

RQ2: How do streaming time and product type moderate the relationship between streamer–viewer emotional resonance and sales performance?

This study investigates the application of emotional contagion theory in live-streaming e-commerce by analyzing data from the “East Buy” live streams conducted between February and April 2024. It aims to explore the impact of the emotional field co-created by streamers and viewers on consumer purchasing behavior and the roles of live-streaming time and product type. The dataset includes 30 live streams, 11,009 text interactions between streamers and viewers, and 22,707 min of real-time trend data. The explanatory variable empathy was constructed through sentiment analysis of the interaction texts. This variable was then used in a regression model with live-stream sales performance to derive the research findings. The results show that increased empathy between streamers and viewers correlates with higher sales increments. Analysis of the moderating variable, live-streaming time, reveals no significant impact on the empathy–sales relationship. Conversely, product type positively moderates the relationship between empathy and sales.

This study makes several theoretical contributions to the field of live-streaming e-commerce. First, measuring the emotional expressions of streamers and viewers within short time frames enriches the application of emotional contagion theory in live-stream sales scenarios. Second, emotional resonance between streamers and viewers is measured using advanced extensive data text analysis methods. To our knowledge, this study is the first to use big data techniques to measure this variable from the linguistic expressions of live stream participants and align it with real-time sales data to explore the transmission mechanisms. Third, this research extends the literature on emotional interactions in service marketing within live streaming, demonstrating that the emotional value provided by streamers during broadcasts enhances viewers’ trust in products and services.

The remainder of the paper is organized as follows: Section 2 presents the theoretical framework, reviewing the literature on emotions and empathy in live streaming and applying text analysis methods in this context. Section 3 introduces the hypotheses and models, establishing the overall framework of the study. Section 4 details the data and variables, including sources and preprocessing steps. The study’s variables encompass explanatory, dependent, moderating, and control variables. Section 5 reports the research findings, validating the study’s hypotheses. Section 6 discusses the research implications and offers suggestions for future research directions.

2. Literature Review

2.1. Emotional Contagion Theory

Early research suggested that emotional contagion is a process by which primitive sympathetic nervous responses directly induce emotions [19]. American psychologist Hatfield later elaborated on Emotional Contagion Theory, positing that emotional contagion is an automatic and unconscious process. Individuals naturally mimic and synchronize with others’ facial expressions, voices, postures, and movements, resulting in emotional convergence and alignment between the interacting parties. This represents the earliest definition of primitive emotional contagion [10]. Barsade [20] expanded this concept to include conscious emotional contagion, which involves a more deliberate and active process of transmitting emotions through mimicry and cognitive and emotional regulation. Research on emotional contagion primarily originates from psychology, with two distinct mechanisms influencing its impact: “Mimicry-Feedback” and “Category Activation”. The former suggests that individuals unconsciously imitate others’ emotional expressions, such as facial expressions, tone of voice, and behaviors, enabling humans to instantly reflect subtle emotional changes they observe, forming the basis of social synchronization. The latter involves exposing individuals to emotions through textual exchanges or other forms of communication without necessarily involving behavioral mimicry [21].

With the proliferation of social media, emotional contagion has become more pronounced in online virtual consumption spaces [22]. Live streaming significantly reduces perceived spatial distance, creating a sense of immediacy and intimacy between streamers and users, thereby shortening the emotional transmission chain. Meng et al. [23] demonstrated from the perspective of emotional contagion that performances by influential streamers can evoke consumers’ emotions, enhancing their willingness to purchase recommended products. In live commerce, multidimensional emotional displays by streamers influence the valence and arousal of viewers’ emotions, fostering an atmosphere of emotional contagion [24]. Live streaming provides a platform for consumers to share information, express emotions, and interact with streamers and other viewers. Additionally, the stories and narratives told by streamers serve as crucial vehicles for emotional contagion, with viewers resonating emotionally when they perceive the authenticity and intensity of the streamer’s emotions, ultimately leading to purchasing behaviors in the consumption context.

2.2. Emotions During Live-Streaming

Emotion is a person’s feeling or reaction to a particular event [25]. It is pervasive in everyday interpersonal interactions and is expressed through voice, body posture, or facial expressions [26]. Language also conveys emotions. The PAD emotion model captures an individual’s emotional state across three dimensions: pleasure, arousal, and dominance. This model describes the subjective experience of emotions and corresponds well with physiological arousal and external manifestations of emotions [27]. Russell [28] proposes that the valence–arousal model, which describes and classifies emotional states, is a significant theoretical framework in emotion research. Valence reflects the perceived pleasantness of an emotion, while arousal captures the intensity and level of emotional activation. Ekman [29] also put forward discrete emotion theory, which considers emotions as relatively independent entities and provides a nuanced understanding of how emotions influence behavior, potentially offering a more detailed description than the valence–arousal dimensions.

The immediacy and interactivity of live streaming create a unique emotional communication environment. Compared to traditional commerce, the context and activities of live e-commerce more readily trigger emotional contagion. Transiting a host’s emotions is a fundamental skill that drives consumer behavior through continuous emotional interaction [30]. Positive emotions help reduce the complexity and vulnerability that consumers experience when shopping online [31]. Tong et al. [32] suggest that consumer emotions of pleasure and arousal mediate the relationship between the visual complexity of a live stream and consumer purchase intention. Wang et al. [24] investigate how the host’s body movements, facial expressions, voice, and interactions capture and influence group emotional valence and arousal. Researchers are exploring the impact of specific emotional displays on consumer engagement and behavior to understand the genuine intentions behind consumer purchases. Emotional expressions in real-time comments affect observers’ cognitive and behavioral responses in short video contexts [33]. In live streams, the host’s joyful emotions can be transferred to viewers, prompting increased tipping behavior [34]. Luo et al. [35] examine the impact of travel product endorsers’ two emotional displays, excitement and calm, on impulsive tourist purchases. Bharadwaj et al. [36] use machine learning algorithms to analyze hosts’ facial expressions, such as happiness, sadness, and surprise, concluding that facial emotion displays have an inverted U-shaped effect on sales performance, indicating that a host’s emotional expression influences viewers’ perceived product value. However, a study on B2B live streaming found that host emotional expressions can reduce sales performance, potentially due to buyers’ skepticism towards overly emotional persuasion [37]. Most existing research focuses on the influence of the host’s or the viewers’ emotional displays on user engagement or purchasing behavior in live streams [38]. It remains unclear whether the emotional resonance between the host’s conveyed emotions and the implicit emotions in viewers’ real-time comments, the effectiveness of emotional marketing strategies, and the mechanisms by which different types of emotional displays by hosts elicit synchronous responses from viewers to achieve live streaming goals.

2.3. Research on Empathy

Empathy is how individuals understand and infer others’ emotions and, in turn, experience similar emotional reactions. Empathy is crucial in various aspects of human behavior and interpersonal relationships. Empathy can be categorized into cognitive empathy and emotional empathy [11]; cognitive empathy involves inferring and understanding another’s emotional state, while emotional empathy is an indirect emotional experience triggered by another’s emotions [39]. In professional service research, cognitive empathy is often described as perspective-taking (viewing a situation from another’s point of view). In contrast, emotional empathy is more accurately characterized by emotional contagion and concern [40]. The core of empathy lies in emotional contagion, which involves matching another person’s emotional state [9], such as feeling tense when observing someone in pain [41].

Empathy is generally considered a key determinant of positive customer–employee relationships. When employees exhibit empathy towards customers, the customers are more likely to positively evaluate the employees’ performance and have a better brand experience. However, Iglesias et al. [42] found that empathy between employees and customers negatively moderates the effect of sensory brand experience on brand equity in the banking industry. With advancements in robotics, empathy is deemed an essential skill for successful interactions between users and social robots [43]. Hotel service providers may benefit from empathy in human–robot interactions, leading to repeat purchases through long-term customer loyalty [44].

Moreover, research on empathy has recently expanded into other areas of social life. Choi et al. [45] explored how empathy among different diners affects the emotional experience of solo diners, finding significant impacts regardless of whether the emotional valence was positive or negative. Jung and Im [46] posited that social media influencers (SMIs) build closeness with their followers through empathy. Central influencers express their viewpoints and share immediate feelings via posts, influencing consumers’ empathetic responses. Empathy facilitates the exchange and resonance of emotions between parties in virtual social contexts. Hossain and Rahman [12] employed machine learning to investigate the empathetic behaviors of potential consumers based on consumer reviews. When potential customers rated reviews as helpful, cool, or funny, it indicated the presence of empathy between current and potential consumers. The audience is not simply passive viewers but active participants engaging in the bilateral communication process. Participatory culture, enabled by web technology, provides individuals with new forms of expression, increasing consumer engagement in public discourse and user-generated content [47]. Despite the prevalence of emotional marketing, few studies have examined the pathways of emotional resonance between live streamers and viewers in dynamic contexts and its impact on consumer behavior.

2.4. Application of Big Data Text Analysis in Live Streaming

The information reflected in the text can help marketers understand market patterns and use these insights to influence the recipients of the text information [48]. Therefore, text analysis techniques are employed to analyze valuable information from large volumes of unstructured text data for decision-making purposes. In previous consumer research, text analysis has been used to extract underlying information from user reviews. For instance, Jia [49] analyzed ratings and reviews from various restaurant websites to compare the motivations and satisfaction of visitors to Chinese and American restaurants. Kim et al. [50] used topic modeling to analyze movie reviews’ linguistic and non-linguistic features to explore the intentions behind posting genuine versus fake reviews.

As a new form of social interaction, live streaming generates a large amount of timely interactive information and more objectively expresses the thoughts of participants [51]; researchers can use multiple dimensions of language to measure consumer thoughts and interactions [52]. Kang et al. [53] analyzed the interaction text between streamers and viewers to investigate how interactivity affects viewer likes, rewards, and comments through relationship strength. Chen et al. [54] used feature extraction and clustering analysis to analyze text comments in gaming live streams and predict factors influencing live stream viewing. Luo et al. [55] employed the LDA method to classify streamer discourse into informational and emotional types, exploring user behavior regarding behavioral, emotional, and relational engagement. With the rapid development of live commerce, the commercial value of live streaming is further explored. Streamers’ detailed product demonstrations and quick responses to viewer questions prompt consumers to purchase quickly. Yang and Wang [56] studied how sellers address consumers in live sales and concluded that streamers’ use of address techniques promotes suggestive sales. Zhang et al. [1] found that viewers reveal their genuine emotions when sending bullet comments, categorizing real-time comments into entertainment and informational types to explore factors influencing purchasing behavior in clothing live streams.

Furthermore, text analysis techniques can capture emotional transmission between streamers and viewers. Lin et al. [33] used the Linguistic Inquiry and Word Count (LIWC) text analysis tool to measure viewer emotions in live streams, revealing how streamers’ emotional displays affect viewer emotions. Compared to the traditional questionnaire interview method, which relies on the passive recall of emotions, text analysis uses objective data to measure emotions, increasing the accuracy of identifying and categorizing emotions while reducing potential biases from manual analysis.

3. Hypothesis Development

The synchronous nature of live streaming creates a context of co-presence, where time and space converge, allowing streamers to establish emotional connections with viewers through engaging content, personalities, and cultural elements [57]. In the new digital environment, the richness of users’ empathetic expressions and the evolution of empathy logic warrant further investigation. To understand this dynamic, our conceptual model illustrates how the empathy between streamers and viewers, particularly in fine-grained emotional displays, impacts sales performance. It also examines whether the emotional resonance between streamers and viewers has heterogeneous effects on sales, moderated by live streaming time and product type.

3.1. Empathy and Sales Performance

Empathy is understanding and responding to another person’s unique experiences [58]. In service management and marketing, empathy is a fundamental skill for employees, especially those involved in sales or customer interactions [59]. When employees interact with consumers empathetically, they are more likely to convey positive emotions to consumers, leading to more favorable perceptions of the organization [60]. In live commerce, emotions significantly influence user purchasing behavior. Unlike traditional consumption scenarios, live commerce leverages the internet’s ability to collapse perceived spatial distances, fostering emotional resonance between consumers and streamers through high responsiveness [61]. Users who achieve a sense of immersion and experience within the virtual consumption scene constructed by the streamer are more likely to engage in purchasing behavior [57]. Streamers share personal stories and experiences, impart knowledge and values, and significantly influence consumers’ perceptions of products through language and behavior. This interaction triggers consumers’ indirect memories of similar experiences, leading to deeper resonance and emotional alignment, ultimately affecting their purchasing behavior. Therefore, we propose Hypothesis 1:

H1: 

In live commerce, empathy between streamers and viewers positively impacts sales performance.

3.2. The Moderating Role of Time

Human biorhythms significantly influence individuals’ emotions, experiences, and behaviors. Different circadian types affect consumer viewing experiences, attention, feelings, and decision-making. People tend to be more sensitive to their own and other’s feelings in the evening. Studies indicate that positive emotions peak around midday and evening, while negative emotions are most pronounced in the morning and mid-afternoon [62]. As the day progresses, arousal levels gradually increase, enhancing the perception of complex sensory experiences, and this is particularly noticeable in the evening [63]. Research on consumer happiness post-purchase across different chronotypes shows that night-active individuals derive greater happiness from material purchases [64]. During the day, people likely spend more time working or studying, whereas evenings are reserved for entertainment and social activities. These factors influence attitudes and behaviors toward live-stream content and purchase intentions. Thus, Hypothesis 2 is proposed:

H2: 

Compared to daytime, empathy between streamers and viewers positively impacts sales performance during nighttime live streams.

3.3. The Moderating Role of Product Type

In consumer behavior research, product categorization is a crucial topic. Products are typically divided into search and experience goods based on different attributes and consumer decision processes. Search goods are those for which consumers can obtain ample information through search and comparison before purchase, whereas experience goods require personal experience to assess their value [18]. Live streaming offers concrete product information and references, reducing barriers to online purchases [65]. The way streamers explain different types of products varies [36]. When purchasing search goods, consumers prefer information about product attributes.

In contrast, when buying experience goods, they seek information related to experiences [16]. In live streams, streamers’ comprehensive presentations and descriptions of search goods allow viewers to understand the product thoroughly. Additionally, viewers can gather product reviews through interactions with others. In contrast, information received about experience goods is subjective, with significant differences in individual descriptive dimensions, methods, and perceptions, requiring consumers to make independent judgments [17]. Thus, Hypothesis 3 is proposed:

H3: 

Compared to experience goods, empathy between streamers and viewers positively impacts sales performance for search goods.

Furthermore, live shopping has become a preferred choice for many consumers, who may make quick purchase decisions influenced by specific factors. Our research model includes these potential influencing factors on live sales as control variables. By controlling for factors related to live streams, streamers, and viewers, we can eliminate their possible impact on consumer psychology and behavior, allowing for a deeper exploration of empathy’s specific effect on sales. This approach enhances the reliability of the results and provides insight into the potential mechanisms of empathy transmission in live commerce. Based on these hypotheses, the model of this study is illustrated in Figure 1.

4. Data and Methodology

4.1. Data Description

Data from Douyin, a leading live-streaming service provider in China, is collected to validate our model. Douyin, one of China’s largest e-commerce live-streaming and social platforms, has over 700 million daily active users. In May 2018, Douyin launched its shop, and by June 2020, it had established an e-commerce division, initiating streamer competitions and significant promotional events [66]. Users can shop through short videos and live-streaming scenarios. In 2023, the average viewership per live-streaming session on Douyin reached 10,000, with the total gross merchandise value (GMV) exceeding CNY 2 trillion. The platform hosts numerous merchants and viewers, facilitating active transactions of products such as fresh food, clothing accessories, daily necessities, books, cultural and creative items, and virtual products.

We select the “East Buy” live stream room as our research subject. On 28 December 2021, New Oriental announced the launch of its live-streaming e-commerce platform “East Buy”. Six months later, the host “Dong Yuhui” gained significant popularity for his unique sales style, with an average daily sales volume exceeding CNY 20 million, becoming a hot topic on social media. There are three main reasons for choosing the “East Buy” live stream room: First, unlike other live stream rooms that stimulate consumers to purchase through low prices and urgency, the hosts use bilingual teaching to integrate literature, history, and philosophy into product descriptions, significantly increasing user retention time and creating a unique and highly engaging live stream environment. Second, “East Buy” hosts participate in live scene sales as knowledge producers and sharers. Through expression, communication, identification, mediation, and education, they disseminate product-related knowledge, enhance the content richness of the live stream, and foster consumer emotional resonance [57]. Third, the interaction in “East Buy” stimulates the audience’s interest in culture and education, enhances the brand’s added value, and provides an excellent example of emotional marketing in the age of social media.

The data collection was conducted from 27 February to 28 April 2024. We obtained content from 69 live streams hosted by “East Buy”. Most of these streams began at 7 a.m. and lasted 16–17 h. The total duration of the live streams amounted to 60,547 min, which included information such as the host’s speech, audience comments, sales performance, and listed products. Due to the extended duration of live streams, each session typically involves 2–3 hosts taking turns to present or a primary host supported by multiple co-hosts. However, the number of hosts is not the focus of this study, and significant variability exists in the data. Therefore, this variable was excluded from the scope of our research. We segmented the data based on the natural pauses in the host’s speech and matched them with the corresponding live stream data from the same period of time. Redundant, missing, and invalid data were removed, and the original dataset was preprocessed following these steps:

(1)

To ensure the rationality and accuracy of emotion analysis, we segmented each live broadcast into natural pauses during anchor product descriptions and audience interactions. This approach allowed us to extract the anchor’s speech during these intervals and transcribe it into text;

(2)

Audience barrage interaction comments are collected per second, matched with the time segments corresponding to natural pauses in the anchor’s speech, and missing or erroneous values are deleted;

(3)

The sales data are matched according to the time of the above-mentioned segments. If multiple sales data correspond to a specific time segment, the average value is taken;

(4)

Other data related to the live broadcast, such as real-time audience attendance, real-time fan growth, and real-time revenue per thousand views (CPM), are matched with the time of the above segments.

The final dataset used for analysis comprises 30 live streams conducted between 1 March and 26 April 2024, totaling 11,009 host–audience interaction records and 22,707 min of live stream data.

4.2. Variable Measurement

4.2.1. Empathy

Empathy describes the emotional matching between hosts and viewers during live broadcasts. The measurement of empathy involves three main steps: Firstly, we collect textual data from host product descriptions and interactions with viewers, aligning viewer comments with natural pauses in the host’s speech. Emojis from viewer messages are removed to facilitate subsequent analysis. Secondly, the processed host–viewer text undergoes sentiment analysis using the sentiment ontology library developed by the Dalian University of Technology Institute of Information Retrieval (DUTIR), following the approach by [67]. Each text segment is analyzed for word frequency counts related to joy, sadness, surprise, anger, disgust, fear, and liking. Finally, using Python, we represent the emotions generated by hosts and viewers during natural pauses as vectors (denoted as Vector1-streamers and Vector2-viewers, respectively) and calculate their similarity using the cosine similarity algorithm to derive the empathy values (Empathy ∈ [0, 1]) needed. The construction process of empathy is illustrated in Figure 2.

4.2.2. Sales Performance

Sales performance is matched with sales increases corresponding to natural pauses in the host’s speech during live broadcasts. During the data collection phase, we observed inconsistencies in the timing between these two metrics; precisely, multiple sales data points could correspond to a single natural pause in the host’s speech. Therefore, we averaged the sales data points within the same time segment to obtain the mean sales increase for that period. An example of emotion analysis matched with sales data is shown in

Table 1. Emotion analysis.

Time	Streamer	Emotion	Viewer	Emotion	Empathy	Sales
2 March 2024 14:04:15 — 2 March 2024 14:07:19	Thank you for your support of “EastBuy”. Thank you again, then we also ask the teacher to put the lens can be around a circle, our teacher can put the lens around a circle, our Xue Ning teacher, our camera teacher, behind the scenes of each of us to now late at night are still insisting on working, in order to give you to present, this one is actually not professional. But is full of sincerity of a performance. Thanks again to all the friends accompanied us until now, but also once again thanks to all the friends accompanied us through the storms of the Oriental Selection, after so many things happened, or resolutely choose to stick to the side of the Oriental Selection of each live broadcast, we thank you again, and I hope that this evening this is not a professional program, can bring you a smile of the shallow. Then our purpose has been achieved. Thank you, thank you, thank you, thank you. (Excerpt)	Positive 32 Negative 6 Anger 0 Disgust 0 Fear 0 Sadness 0 Surprise 0 Good 23 Happy 9	Makati City Makati City Makati City Makati City Makati City Makoma City Makati City [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] [Likes] Smooth flow against the current Big sedan chair [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] [applause] I’m here with you. Makati City Sedan chair. Apple juice is good. Apple juice is really good. [facepalm] [facepalm] [facepalm] [facepalm] [facepalm] [facepalm] [facepalm] [facepalm] [facepalm] [facepalm] [facepalm] The sedan chair. [Likes] [Likes] [Likes] [Likes] [Likes] Any more mystery guests? Big Flower Bridge (Excerpt)	Positive 92 Negative 1 Anger 0 Disgust 1 Fear 0 Sadness 0 Surprise 1 Good 84 Happy 7	0.93387	178

.

4.2.3. Moderator Variable

More viewers watch live streams at night than during the day, so we categorize live stream time as a binary variable: daytime (0) from 6:00 a.m. to 6:00 p.m. and nighttime (1) from 6:00 p.m. to 6:00 a.m.

Product type is categorized into experience and search products as a binary variable. Considering the definitions of search and experience products, we set categories such as clothing, shoes, and household cleaning products as 1 (search products) and categories like toys, books, and smart home devices as 0 (experience products).

4.2.4. Control Variable

The length of streamer dialogues (Streamer Length) is calculated based on the total number of words the streamer speaks during natural pauses in their speech. The information conveyed in each segment of dialogue by the streamer varies, influencing viewers’ perceptions of the product’s value. Excessive elaboration by the streamer can lead to information overload among viewers, potentially prompting impulse purchases [68]. Conversely, insufficient elaboration may leave viewers with incomplete product information, reducing their inclination to purchase. Therefore, we consider the word count per dialogue segment the streamer speaks as a control variable.

Viewer Comment Length (Viewers Length) is calculated based on the total number of words in viewer comments during natural pauses in the streamer’s speech. The word count of real-time viewer comments reflects audience engagement with the streamer and the product. Generally, audiences who engage more actively in live interactions are more likely to make purchasing decisions.

Popularity in the live stream refers to increased fans within the current broadcast. Audiences are attracted by certain qualities of the host or the live stream itself, which can turn them into fans who make purchasing decisions based on their likes and loyalty.

Streamer’s competence denotes the ability of the host to introduce products and generate income through interactions per minute. Differences in how hosts present products and engage with viewers affect consumers’ perceptions of product value.

Live stream attractive indicates the number of viewers entering the live stream. Viewers who are drawn into the live stream will likely stay longer and potentially make purchases.

Table 2. Variables measurement.

Variables	Definition	Measurement
1. Empathy	The extent to which streamers and viewers emotionally resonate in live streaming	Degree of similarity between streamers’ speech and viewer comments over a given period
2. Sales Performance	Live streaming sales	Increase in live sales during a given period
3. Time	Live streaming start time	Live streaming starts at 1 (6:00 p.m.–6:00 a.m.) at night and 0 (6:00 a.m.–6:00 p.m.) in the daytime
4. Product type	Products on the shelves in the live room	Split into search-based and experience-based products, with 1 for search-based products and 0 for experience-based products
5. Streamers Length	Length of content for product descriptions and audience interaction by the anchor	The total number of words used by the anchor for product descriptions and interaction in a given period
6. Viewers Length	Length of content for viewers’ interaction	Total number of words commented on by viewers in a given period
7. Popularity	Live streaming popularity	Number of followers added to the live stream in a given period
8. Streamer’s competence	Streamer’s ability to present products and interact with viewers	Bandwagon output of anchors introducing products and interacting with viewers in a given period
9. Live stream attractive	The attraction of live broadcasting	Number of viewers entering the live stream in a given period

and

Table 3. Descriptive statistics.

Variables	N	Min	Max	Mean
1. Empathy	11,009	0.000	0.6509	0.3812
2. Sales Performance	11,009	1	242.07	295.434
3. Time	11,009	0	0.47	0.499
4. Product type	11,009	0	0.91	0.292
5. Streamers Length	11,009	0	140.29	62.017
6. Viewers Length	11,009	0	215.89	453.028
7. Popularity	11,009	0	2313.13	3165.239
8. Streamer’s competence	11,009	5400	627,307.53	664,717.77
9. Live stream attractive	11,009	129	8946.12	8672.86

, respectively, present our definitions, measurements, and descriptive statistics for all variables.

4.2.5. Model Description

We employ a linear regression model to measure the impact of empathy on sales performance, as well as the moderating effects of live stream timing and product type. Due to its simplicity and interpretability, the linear regression model is well-suited to the current research problem and helps avoid the overfitting issues that may arise with other models. The model is specified as follows:

In Model 1, we examine the direct impact of empathy on sales performance, incorporating several potential control variables that may influence the results.

Model1: Sales Performance = α₁ + β₁ Empathy + β₂ Streamers Length + β₃ Viewers Length + β₄ Popularity + β₅ Live stream attractive + β₆ Streamer’s competence + ℇ

(1)

Models 2 and 3 are designed to test the moderating effect of live stream duration.

Model2: Sales Performance = α₂ + β₁ Empathy + β₂ Time + β₃ Streamers Length + β₄ Viewers Length + β₅ Popularity + β₆ Live stream attractive + β₇ Streamer’s competence + ℇ

(2)

Model3: Sales Performance = α₃ + β₁ Empathy + β₂ Time + β₃ Time*Empathy + β₄ Streamers Length + β₅ Viewers Length + β₆ Popularity + β₇ Live stream attractive + β₈ Streamer’s competence + ℇ

(3)

Models 4 and 5 aim to investigate the moderating effect of product type.

Model4: Sales Performance = α₄ + β₁ Empathy + β₂ Product type + β₃ Streamers Length + β₄ Viewers Length + β₅ Popularity + β₆ Live stream attractive + β₇ Streamer’s competence + ℇ

(4)

Model5: Sales Performance = α₅ + β₁ Empathy + β₂ Product type + β₃ Product type*Empathy + β₄ Streamers Length + β₅ Viewers Length + β₆ Popularity + β₇ Live stream attractive + β₈ Streamer’s competence + ℇ

(5)

In Models 1–5, Sales Performance is the dependent variable in this study. Empathy is the key explanatory variable, while Time and Product type act as moderating variables. Time*Empathy represents the interaction term. In Models 4–5, Product type is the moderating variable, and Product type*Empathy represents the interaction term. Control variables include Streamers Length, Viewer’s Length, Popularity, Streamer’s competence, and Live stream attractiveness. ℇ denotes the error term.

5. Result

To examine the potential multicollinearity between independent variables and interaction terms, we conducted a Variance Inflation Factor (VIF) test, finding values less than 5. Next, hierarchical regression was used to test the main and moderating effects. Model 1 regresses empathy on sales performance, testing Hypothesis 1. Models 2 and 4 incorporate the moderating variables of time and product type into the regression model. Models 3 and 5 include the interaction terms of empathy with time and product type to verify Hypotheses 2 and 3. All models include control variables. The analysis was conducted using SPSS 26.0. The coefficient of determination, R², is 0.728, indicating that the model demonstrates good fit validity. We calculated the F-values for Models 1, 3, and 5, and the results suggest that the regression models demonstrate strong explanatory power for sales performance. Additionally, the relationship between host–audience empathy and sales performance is significant. The main results and moderating effects are presented in

Table 4. Regressions result.

Variables	Model1	Model2	Model3	Model4	Model5
	Sales Per	Sales Per	Sales Per	Sales Per	Sales Per
Empathy	15.125 *** (3.671)	15.275 *** (3.708)	21.035 ** (3.713)	15.155 *** (3.678)	−28.406 * (−2.035)
Time		8.230 * (2.394)	15.563 * (2.581)
Time*Empathy			−11.458 (−1.481)
Product Type				2.700 (0.534)	−29.907 ** (−2.673)
Product type*Empathy					47.175 ** (3.266)
Streamers Length	0.036 (1.434)	0.045 (1.757)	0.045 (1.771)	0.037 (1.471)	0.038 (1.480)
Viewers Length	0.026 *** (7.463)	0.024 *** (6.975)	0.025 *** (7.036)	0.009 *** (2.102)	0.026 *** (7.473)
Popularity	−0.009 *** (−15.158)	−0.008 *** (−12.298)	−0.008 *** (−12.361)	−0.009 *** (−15.164)	−0.009 *** (−15.142)
Live stream attractive	0.016 *** (60.963)	0.016 *** (60.418)	0.016 *** (60.417)	0.016 *** (60.963)	0.016 *** (60.900)
Streamer’s competence	0.000 *** (61.404)	0.000 *** (58.238)	0.000 *** (58.251)	0.000 *** (61.399)	0.000 *** (61.440)
Obs	11,009	11,009	11,009	11,009	11,009
R2	0.728	0.728	0.728	0.728	0.728
F	4898.28 ***	/	3676.67 ***	/	2944.61 ***

Note: Robust T-statistics in parentheses, *** p < 0.01, ** p < 0.05, * p < 0.1.

.

5.1. Main Effect

From the results of Model 1, it is evident that even after controlling for variables such as streamer length, viewer comment length, live stream popularity, live stream attractiveness, and streamer competence, empathy between the streamer and viewers significantly positively impacts sales (β = 15.125, ρ < 0.01). This result supports Hypothesis 1, indicating that empathy between the streamer and viewers plays a crucial role in live stream sales. Our research further explores the conclusions of Lin et al. [33], indicating that the emotions of streamers and viewers mutually influence each other. This aligns with our view that there is emotional resonance between streamers and viewers during live streams. Similarly, the display of the streamer’s emotions influences viewer participation. In this study, empathy between the streamer and viewers significantly affects viewers’ purchasing behavior. Based on the theory of emotional contagion and the analysis of actual purchasing data, we establish the importance of empathy marketing in live streaming, further advancing the understanding of the live streaming e-commerce field.

5.2. Moderating Effect

5.2.1. Moderating Effects of Time

Models 2 and 3 examine the moderating effect of live stream time on the relationship between empathy and sales. In Model 2, the live stream duration, as an independent variable, significantly impacts sales performance (t = 2.394, p < 0.1). In Model 3, after incorporating the interaction term between time and empathy, the interaction term Time*Empathy has a coefficient of −11.458 (t =−1.481, ρ > 0.1), indicating an insignificant moderating effect. Hence, Hypothesis 2 is not supported. The empathy between streamers and viewers is not moderated by the live stream time, which differs from our previous understanding.

5.2.2. Moderating Effects of Product Type

Models 4 and 5 explore the moderating effect of product type on the relationship between empathy and sales. The interaction term’s coefficient is 45.175 (t = 3.266, ρ < 0.05), indicating a significant positive moderating effect, thus supporting Hypothesis 3. The products’ intrinsic characteristics and perceived value influence viewers’ purchasing decisions during live streams. Factors such as functionality, quality, price, and brand reputation are key in purchasing decisions. Buying behavior is usually need-driven, meaning that viewers’ need for a specific product or service at a given time point drives their purchasing decisions. Therefore, product type is an essential factor influencing viewers’ purchasing behavior during live streams and should be considered in the model. The findings indicate that search-type products have a more positive impact on the relationship between empathy and sales than experience-type products.

6. Discussion

6.1. Main Findings

This study examines the impact of empathy between streamers and viewers on sales performance in live streams and the roles of live streaming time and product type. Specifically, the emotional resonance between streamers and viewers positively and significantly affects sales. When viewers perceive their emotions aligning with the streamers’, they are more likely to purchase products from the live stream. H1 is supported. Viewers unconsciously mimic the emotional expressions they observe during live streams and internalize these emotions. For instance, when a streamer displays enthusiasm and excitement about a product, viewers may experience similar feelings. The internal mechanism of emotional contagion, “empathy and regulation”, suggests that viewers can empathize with a particular emotion from the streamer’s expression, compare and regulate their feelings with those of the streamer, and decide whether to accept these emotions based on cognitive mechanisms such as language-mediated association [69] and active perspective taking. Viewers consciously use reasoning, analysis, and imagination to understand others’ true feelings, thus deciding whether to follow or reject them. Streamers amplify the frequency and intensity of emotional exposure to increase user engagement and activity, leading to emotional contagion [70].

The moderating effect of product type in the model is positive and significant. Viewers receive more information from the streamer’s speech for search products, while experience products emphasize individual perception. H3 is supported. This result can be explained from two perspectives. First, viewers’ emotional needs while watching live streams are diverse, such as relaxation, socializing, and entertainment. These emotional needs may not be significantly influenced by specific periods. Viewers choose to watch live streams based more on their schedules and preferences than the inherent characteristics of the time itself. Second, individual differences play a crucial role in emotional needs. Significant variations in viewers’ emotional needs at different times may obscure the overall effect of time as a moderating variable. The mechanism of empathy’s role in live stream sales may be more complex. Empathy promotes sales by enhancing viewers’ engagement and trust, but this effect may not be directly moderated by the time of day. For instance, viewers may develop increased interest and trust in a product due to a streamer’s empathy at any time.

Contrary to expectations, streaming time does not significantly impact the relationship between empathy and sales. Interaction and emotional resonance between streamers and viewers positively affect sales regardless of whether the live stream occurs during the day or at night. H2 is not supported. This can be explained in two ways. First, search-type products usually have clear performance indicators and functional characteristics involving higher information asymmetry. As an information dissemination channel, live streaming helps consumers acquire more knowledge about the products, reducing the uncertainty in their purchasing decisions. Similarly, the streamer’s detailed explanations and demonstrations help viewers better understand product features and the establishment of empathy enhances the persuasiveness of the information. Second, the emotional resonance between streamers and viewers stimulates viewers’ intrinsic purchasing motivation. By displaying enthusiasm for the product and empathy for viewers’ needs, streamers can arouse viewers’ desire to purchase. Empathy can be seen as a form of consumer empowerment, making consumers feel they are making informed purchasing decisions.

6.2. Theoretical Contributions and Managerial Implication

The theoretical contributions of this study are as follows: First, it enriches research on emotional contagion theory in live streaming. Based on streamer–viewer interactions, advanced textual sentiment analysis techniques capture emotional expressions during live streams. Previous studies have mainly focused on the impact of positive or negative emotions displayed by streamers on viewers’ psychology and behavior [1]; this study delves into discrete emotions to explore how momentary emotional resonance between streamers and viewers during live streams influences purchasing decisions. Second, extensive data methods measure emotional resonance in live streams. Past studies on empathy have primarily relied on self-reported questionnaires or narrative scenarios, where participants express their agreement with specific statements or stories [71]. The mechanism of empathy transmission has changed in the new context of live-streaming sales. Real-time interactions within the live stream allow viewers to ask questions and express thoughts at any moment, enabling the quantification of empathy through the emotions expressed in viewer comments. Third, this study expands the service marketing literature on emotional interactions in live streaming. Consumers watching live streams cannot physically interact with products and must make initial value judgments based on the streamer’s presentations and descriptions. This uncertainty requires streamers to provide more detailed services and demonstrate empathy to discern viewers’ emotions hidden behind screens. Additionally, interactions between streamers and viewers, as well as among viewers, bridge the emotional distance, weakening the traditional buyer–seller relationship and enhancing emotional connections between streamers and consumers [72].

The managerial implications are: First, streamers should strategically display emotions during live streams. Streamers need to understand the meanings behind viewers’ comments and respond effectively to foster a sense of recognition from viewers towards the stream and the streamer. Live purchasing behavior results from viewers’ emotional value being satisfied, and the extent of emotional value provided by the streamer significantly influences viewers’ trust in the streamer, subsequently affecting their purchasing decisions. Second, streamers should focus on individual emotional expressions while guiding viewers to create a positive live-streaming atmosphere. Traditional “consumerist” live streaming induces tension and competition for discounted products, which may lead to impulsive purchases and subsequent returns. Streamers must manage the live stream’s pace and address consumers’ preferences and concerns [73,74]. Third, homepage recommendation algorithms and consumer identification should be more accurate for live-streaming platforms. Platforms must pre-filter information, helping consumers find meaningful content amidst vast information, enhancing engagement in the live stream [75,76]. Further, we used “East Buy” as an excellent case study for live interaction and emotional marketing, optimizing the live room strategy through real-time audience feedback and user-generated content and providing practical management advice for other brands.

6.3. Research Limitations and Future Perspectives

This study has the following three limitations that future research should address. First, objective data from live streams are used for measurement without delving into participants’ subjective perceptions [77]. However, emotional expressions in streamers’ speech and viewers’ comments are analyzed through text analysis; potential psychological measurement scales are not used to validate our assumptions. Future research can employ alternative methods (e.g., questionnaires, case studies, or experiments) to collect data and verify potential psychological mechanisms from multiple angles.

Second, only the impact of live streaming time and product type on consumer purchases is considered. In practice, factors related to the live stream room, streamer, and platform may influence viewers’ short-term emotional perceptions. For instance, the live stream room setup, co-streamers’ participation, and platform push mechanisms may impact consumers’ emotions and purchasing decisions. Future research can explore these related factors to determine whether emotional resonance between streamers and consumers varies under different circumstances.

Third, this study examines the impact of empathy on consumer purchases. Previous research indicates that streamers rarely display negative emotions during live streams; instead, they engage in emotional performance or labor for better live-stream effects [78]. Therefore, the positive impact of empathy on sales may result from the streamer’s unilateral efforts. Future research could consider other attributes of live streaming, such as the participation level of co-streamers and the background setup of the live stream room, to determine their influence on viewers’ emotions and subsequent consumer behavior.