Understanding the purchase intention in live streaming from the perspective of social image

The paper examines how e-commerce live streaming (ELS)—a rapidly growing sales format in China—shapes consumer purchase intention through the social image that streamers intentionally construct. Drawing on dramaturgical theory, the authors argue that streamers strategically manage their public persona to gain acceptance, trust, and popularity. Social image is distinguished from self-image and is defined as how streamers are perceived by others based on appearance, behavior, and language. Prior work has identified numerous streamer characteristics, but few studies have analyzed them from the social image perspective. The study addresses three research questions: (1) How do social image characteristics manifested by ELS streamers with different motivations impact trust in ELS shopping? (2) What is the effect of social image built by ELS streamers on purchase intention in ELS shopping? (3) Does trust mediate the relationship between social image factors and purchase intention? Using the SOR framework, the authors conceptualize social image stimuli across three intentional dimensions—being attractive, being the right professional endorser, and being an interesting/effective communicator—predicting trust (organism) and purchase intention (response). The study aims to inform both academic understanding and managerial practice in ELS.

The theoretical background integrates dramaturgical theory and the SOR framework to model how externally observable social image characteristics (stimuli) shape internal states (trust) and behavioral responses (purchase intention). Streamer persona literature suggests a constructed public image shaped by performance, platform features, and style. The authors categorize social image characteristics into three motive-based dimensions: (1) First-impression appeal: physical attractiveness (appearance, makeup, dress, voice, humor, environment). Prior studies show attractiveness drives favorable impressions and persuasion, shaping attitudes and purchase intentions. Hypotheses: H1 (attractiveness→trust), H2 (attractiveness→purchase intention). (2) Professional endorsement quality: matchup congruence (fit between streamer and product/viewer), authenticity (genuineness, consistent persona, credible information, realistic scenes), and expertise (knowledge, advice, problem-solving). Literature indicates these enhance credibility, trust, and purchase intentions. Hypotheses: H3 (matchup→trust), H4 (matchup→purchase intention), H5 (expertise→trust), H6 (expertise→purchase intention), H7 (authenticity→trust), H8 (authenticity→purchase intention). (3) Communicative interaction quality: responsiveness (timely, considerate replies) and entertainment (hedonic value, fun, diversion). Prior research often links these to satisfaction, trust, and intentions, though effects can vary by context. Hypotheses: H9 (responsiveness→trust), H10 (responsiveness→purchase intention), H11 (entertainment→trust), H12 (entertainment→purchase intention). Trust is posited to directly affect purchase intention (H13) and to mediate the effects of the six social image characteristics on purchase intention (H14a–H14f).

Measurement development: Constructs were adapted from validated scales. Physical attractiveness (4 items: Karandashev et al., 2020; Yuan et al., 2016); matchup congruence (4 items: Chen et al., 2022; Parmar et al., 2020), including streamer–product and streamer–viewer fit; authenticity (4 items: Becker et al., 2019; Morhart et al., 2015); expertise (4 items: Ladhari et al., 2020; Zhang et al., 2022); responsiveness (5 items: Guo and Sun, 2022; Gummerus et al., 2004); entertainment (4 items: Chen and Lin, 2018); trust (4 items: Chetioui et al., 2021); purchase intention (4 items: Ho et al., 2022). All items used 5-point Likert scales (1=strongly disagree to 5=strongly agree). Content validity was refined via focus group (e-commerce professors, two ELS streamers, two ELS managers), bilingual translation/back-translation checks, and a pretest with 15 students. Some items were reworded for context fit. Sampling and data collection: Targeted Chinese consumers who had previously purchased via ELS on platforms such as Taobao/Tmall, Douyin, Kuaishou, and JD. Mixed-mode data collection (online via Wenjuanxing links shared on WeChat, Weibo, Zhihu; offline paper surveys near universities in southern China). Field period: March–April 2023. Initial N=365 (268 online, 97 offline). Screening ensured ELS shopping experience. Attention checks (bogus item and diligence item) and randomization of item order were used. After cleaning for missingness and inattentiveness, valid N=323 (231 online, 92 offline), effective response rate 88.49%. Demographics: 63.2% female; majority aged 18–34; most-used platforms: Taobao/Tmall, Douyin, Kuaishou; primary product categories included fashion, beauty, sports & outdoors, electronics; 67.18% watched ELS 1–3 times/week. Data analysis: Reliability and validity examined using SPSS 26.0. Cronbach’s alpha and composite reliability (CR) exceeded 0.7 for all constructs; total alpha=0.925. Convergent validity: factor loadings >0.7 (p<0.001), AVE >0.5. Discriminant validity via Fornell–Larcker criterion satisfied. Confirmatory factor analysis (CFA) and structural equation modeling (SEM) conducted in AMOS 24. Measurement model fit: χ2=625.3, df=467, p<0.001; CFI=0.985; GFI=0.915; AGFI=0.889; RMR=0.04; RMSEA=0.037. Path coefficients estimated with significance tests. Mediation of trust tested via bootstrapping (2,000 samples) to obtain indirect effect CIs following Nitzl et al. (2016).

Measurement quality: All constructs showed good reliability (alphas ≥0.785; CR ≥0.837) and convergent validity (loadings >0.70; AVE ≥0.557). Discriminant validity supported by Fornell–Larcker. Structural paths (standardized estimates, significance): - Trust (TR) antecedents: - Physical attractiveness (PA) → TR: β=0.154, p<0.01 (H1 supported) - Matchup (MU) → TR: β=0.246, p<0.001 (H3 supported) - Expertise (EP) → TR: β=0.201, p<0.001 (H5 supported) - Authenticity (AU) → TR: β=0.200, p<0.001 (H7 supported) - Responsiveness (RS) → TR: β=0.190, p<0.01 (H9 supported) - Entertainment (ET) → TR: β=0.084, p=0.183 (H11 not supported) - Purchase intention (PI) antecedents: - PA → PI: β=0.113, p<0.05 (H2 supported) - MU → PI: β=0.234, p<0.001 (H4 supported) - EP → PI: β=0.158, p<0.01 (H6 supported) - AU → PI: β=0.154, p<0.01 (H8 supported) - RS → PI: β=0.035, p=0.571 (H10 not supported) - ET → PI: β=0.050, p=0.368 (H12 not supported) - TR → PI: β=0.364, p<0.001 (H13 supported) Effect magnitudes: Matchup congruence was the strongest predictor of both trust and purchase intention among social image characteristics, followed by expertise, authenticity, and physical attractiveness. Entertainment had no significant direct effects on trust or purchase intention; responsiveness significantly increased trust but did not directly influence purchase intention. Mediation (bootstrapped indirect effects, 95% CI): - PA → TR → PI: estimate=0.069, CI [0.017, 0.154] (H14a supported) - MU → TR → PI: estimate=0.086, CI [0.035, 0.172] (H14b supported) - EP → TR → PI: estimate=0.105, CI [0.043, 0.229] (H14c supported) - AU → TR → PI: estimate=0.114, CI [0.049, 0.232] (H14d supported) - RS → TR → PI: estimate=0.076, CI [0.018, 0.194] (H14e supported) - ET → TR → PI: estimate=0.033, CI [-0.013, 0.110], p=0.346 (H14f not supported)

The findings address the research questions by showing that streamer social image dimensions influence both trust and purchase intention, with trust serving as a key mediating mechanism in most cases. First-impression appeal via physical attractiveness boosts trust and, directly and indirectly, purchase intention—effects potentially amplified by cultural emphasis on beauty norms in the Chinese context. The professional endorsement dimension—matchup congruence, expertise, and authenticity—consistently enhances trust and purchase intention, with matchup congruence exhibiting the strongest effects. This may reflect social comparison dynamics and the dual fit considered in the study (streamer–product and streamer–viewer), which strengthens emotional connection and credibility. Interaction-related characteristics show asymmetry: responsiveness increases trust but does not directly affect purchase intention, while entertainment neither increases trust nor purchase intention. Psychological and contextual explanations include the activation of persuasion knowledge when entertainment is perceived as commercially motivated, the one-to-many interaction constraints limiting individualized responsiveness impact, and the primacy of product demonstration and expert guidance over hedonic aspects in ELS shopping decisions. Regulatory transparency requirements may also attenuate persuasive effects of entertainment. Overall, the results underscore trust’s central role (H13) and its mediation for most social image factors (H14a–H14e), indicating that crafting credible, well-matched, knowledgeable, and authentic personas is more consequential for driving purchases than entertainment-oriented positioning.

This study advances understanding of how ELS streamers’ social image affects consumer behavior by modeling three intentional dimensions—attractiveness, professional endorsement quality (matchup, authenticity, expertise), and communicative interaction (responsiveness, entertainment)—within the SOR framework. Empirical results from 323 Chinese ELS shoppers show that matchup congruence, expertise, authenticity, and physical attractiveness significantly increase purchase intention (directly and via trust). Responsiveness strengthens trust but does not directly increase purchase intention; entertainment shows no significant effects on trust or purchase intention. Trust is a robust predictor of purchase intention and mediates most social image effects. Managerially, platforms and e-retailers should prioritize aligning streamers with endorsed products and target audiences, cultivate authenticity, and develop expertise, while maintaining ethical practices regarding diversity and transparency. Streamers should focus on delivering credible, expert, and authentic content and consider attractiveness holistically (appearance, voice, style, environment). Entertainment can aid engagement but is not a primary purchase driver. Future research should test cross-cultural generalizability, examine moderating effects (e.g., product category, discount levels, platform characteristics), incorporate regulatory variables, and investigate longitudinal outcomes such as repeat purchase and returns.

- Generalizability: Data were collected in China; cultural and regulatory contexts may limit applicability elsewhere. Cross-country studies are needed. - Scope: The study does not examine moderating effects of product category or different ELS formats; post-purchase behaviors (repeat purchases, returns) were not captured. - Regulatory and platform factors: Regulations were discussed but not modeled; platform characteristics (e.g., recommendation relevance, visual effects, sociability) were excluded and may moderate effects. - Price/promotions: Discount levels and pricing factors were not included and may influence purchase decisions in ELS. - Interaction constraints: One-to-many live formats may limit responsiveness depth, potentially attenuating its direct effect on purchase intention.