Comparing E-Commerce Micro- and Macroinfluencers in TikTok Videos: Effects of Strategies on Audience Likes, Audience Shares, and Brand Sales

Abstract

This article compares how micro- and macroinfluencers promote audience likes, shares, and brand sales in TikTok videos (Douyin in China). Using insights from the extant literature, the study examines how influencers’ charisma strategy (including attractiveness), content strategy (including message quality), and bonding strategy (audience relationship) affect audience likes, audience shares, and brand sales. The findings confirm the core tenet of charisma and content strategies. Some unexpected effects of the bonding strategy that correspond to the two (micro versus macro) influencer groups are also revealed. The salience of these strategies justifies the uniqueness of influencer marketing and helps link it to the celebrity brand endorsement literature. It also shows how microinfluencers, a topic of emerging significance, can thrive. The study provides insights for future research and industry advancement.

Keywords:

• Brand sales
• likes and shares
• microinfluencer
• persuasive strategies
• TikTok in China

Influencer marketing is a refreshing and exciting disruption to the half-century-old practice of celebrity brand endorsement. Empowered by social media and the availability of low-cost, low-barrier setups (e.g., Amazon Influencer Program), many lay brand spokespersons have emerged as effective influencers, as they thrive to produce branded content that engages consumers and shapes their preferences. This disruptive innovation in brand endorsement is unprecedented, as it has spread across both developed and developing economies. Spending in influencer marketing is poised to reach $21.1 billion by the end of 2023, and more than 67% of marketers who allocate budgets for influencer marketing intend to increase their respective expenditures (Santora 2023). In particular, influencer marketing on TikTok has seen explosive growth in recent years. The social media platform was reported to have 1.7 billion monthly active users in 2023, making it a prime platform for brand collaboration (Iqbal 2023). It has now surged ahead as the most popular influencer marketing channel, with 56% of brands opting to promote their products via TikTok, marking its first-ever lead over Instagram (51%) and surpassing Facebook (42%) and YouTube (38%) by a significant margin (Geyser 2023). As these numbers continue to rise, influencer marketing on TikTok will continue to redefine the boundaries of brand communication.

In the emerging literature on influencer marketing, researchers have drawn anecdotal insights from successful influencers (García-Rapp 2017; Hung 2020), proposing conceptual models based on strategies such as charisma (compelling attractiveness that inspires devotion in others), message content, or the relational bonds that influencers apply to gain audience trust, empathy, and likes. Focusing on highly successful influencers allows researchers to illustrate the phenomenon and discover the foundational theories that influencers use to build their success. Yet this approach has unwittingly relegated microinfluencers, who have a smaller audience base, to a topic of lesser importance.

Microinfluencers may appear to be less impactful brand promoters, but they perform exceptionally well. In fact, 80% of marketers featured in HubSpot’s 2023 Social Media Marketing Report indicated their favorability toward collaborations with smaller influencers (Santora 2023). This preference stems from the ease of establishing long-term partnerships with smaller influencers and the cost-effectiveness associated with such collaborations (Santora 2023). Relatedly, in a luxury product campaign that featured both a celebrity and a microinfluencer, the latter outperformed the former by increasing the number of visits, likes, and comments on the promoted brand’s web page (Marques, Casais, and Camilleri 2020). The stronger effect of microinfluencers has been attributed to their perceived authenticity in another study (Park et al. 2021). Together, these studies pointed out that the processes underpinning microinfluencers’ effectiveness may differ from those used by highly successful influencers, the core focus of the influencer literature.

This study is a response to a call to address this research gap (Park et al. 2021). It aims to conceptually investigate and empirically verify the postulates on the success of microinfluencers. We purport that both source-related (charisma, bonding) and message-related (content) strategies are employed by influencers, although macro- and microinfluencers may deploy these strategies differently. In addition, researchers have pointed out that audience likes and shares, though related, are driven by dissimilar processes (Li, Lai, and Lin 2017; Rudat, Buder, and Hesse 2014). Thus, in this study, we examine the effects of three persuasive strategies (charisma, content, and bonding) on audience likes, audience shares, and brand sales.

In contrast to the extant literature, which relies largely on experiments and consumer surveys, this article employed a database of TikTok influencers in China to assess microinfluencers’ (versus macroinfluencers’) effects in a naturalistic setting. Specifically, we deployed performance metrics to study a selected number of TikTok videos in which an influencer promoted a single brand and related them to the three strategies the influencers used. Furthermore, we analyzed the content featured in audience live texts, a form of social listening (Hung, Tse, and Chan 2022), to capture the influencers’ strategy in an unobtrusive manner.

This article aims to achieve several objectives. First, it conceptualizes and empirically validates the strategies (charisma, content, and bonding) microinfluencers deploy to enhance their performance. Second, by empirically comparing micro- and macroinfluencers, the study helps define the strategic options microinfluencers may use to stimulate audience likes and shares, which in turn inform their effects on brand sales. Finally, we conducted our investigation using TikTok videos, a powerful social media and brand promotional platform in China, as our study context.

Literature Review

For more than half a century, firms and advertising agencies have employed celebrities as key spokespersons for their brands. This practice is backed by decades of research that have endorsed movie stars, singers, and sports celebrities as effective spokespersons who can generate handsome returns for a firm’s promotional investment (Pringle 2004). Extant literature has identified source attractiveness (and meanings), source credibility, and relational bonding as the key mechanisms underlying the effectiveness of this strategy (Hung 2021). Interestingly, while noncelebrity endorsements (e.g., in the form of a user-consumer) are used in testimonial ads, they are comparatively less effective than celebrity endorsements (Parmar and Patel 2014; Saeed et al. 2014). This is in contrast to what we now see in influencer marketing.

Research on influencer marketing has been expanding quickly to coevolve with this practice. To validate influencer effectiveness, seminal studies have used celebrity endorsement theories, such as source credibility (Lou and Yuan 2019; Schouten, Janssen, and Verspaget 2020) and source attractiveness (Chae 2018; Morhart et al. 2020). Some have examined new constructs, including knowledge expertise (García-Rapp 2017; Jiménez-Castillo and Sánchez-Fernández 2019) and consumer engagement (Hung 2020; Yuan and Lou 2020), while others have examined different forms and contexts of influencer presentations, namely live shows, short videos, and gaming platforms (Brennan 2020; Hung 2021). Furthermore, influencers have ventured into many consumption domains (e.g., fast-moving consumer goods [FMCG], tourism, and fashion) and topics (e.g., new products, usage experiences, and best buys).

Influencers As Brand Promoters

The contemporary marketplace is fiercely competitive. As firms compete for market presence and growth, they produce an array of products, models, and brand extensions, making the marketplace confusing, even to knowledgeable consumers. As a result, many frustrated consumers seek independent sources of information, such as recommendations made by empathetic laypersons in product/user communities (Hung and Li 2007). Though many layperson influencers and their content may not be at a professional level, their opinions are welcome as authentic and trustworthy (Ye et al. 2021). In short, confusion in the marketplace has given rise to a surging need for objective information provided by laypersons, thus giving birth to a new type of brand spokesperson in the advertising ecosystem.

Among lay spokespersons, successful ones project homophily (Ladhari, Massa, and Skandrani 2020) and authenticity (Ye et al. 2021), attributes which can shorten the social distance between spokespersons and consumers. Thus, some influencers take on the “boy/girl next door” image instead of trying to radiate the glamour that celebrities possess (Hung 2020). In turn, they are perceived as “empathetic friends” (Ye et al. 2021). With growing and unwavering audience support, these influencers receive financial rewards from sponsors and media platforms, thereby elevating influencer marketing into a freelance, self-made career that has attracted many.

Influencer Type

Influencers can be categorized into four tiers based on their reach (Campbell and Farrell 2020). At the top are megainfluencers, with 1 million to 5 million followers. They are followed by macroinfluencers, with 100,000 to 1 million followers; microinfluencers, with 10,000 to 100,000 followers; and nanoinfluencers, with 1,000 to 10,000 followers. While this Instagram-based categorization system is useful, the actual number of followers per type may vary from one country to another and from one platform to another.

China’s influencer market is the largest globally. Worth CNY 1.3 trillion (USD 210 billion) in 2020, it is estimated to reach CNY 6.7 trillion (USD 1.035 trillion) by 2025 (Pjdaren 2023). Given the sheer size of the market, the reach of its top influencers is similarly impressive; for example, “Lipstick King” Li Jiaqi (Austin Li) has more than 100 million followers across various platforms. Thus, the number of followers delineating macro- and megainfluencers especially needs to be adjusted upward to reflect the reality of this country’s context (Robles 2020).

The differences among influencer types (e.g., macro- versus microinfluencers) are not based on reach alone. With a sizable follower base, macroinfluencers possess unique personalities, lifestyles, or product-related knowledge that make them “taste leaders” (McQuarrie, Miller, and Phillips 2012). This is in contrast to microinfluencers, who, as everyday persuaders (Marques, Casais, and Camilleri 2020), rely strongly on a product demonstration approach (Daxue Consulting 2021). Thus, there is a difference in the persuasion techniques used by macro- and microinfluencers.

Although macroinfluencers may seem to have greater reach and influence, a recent Forbes article noted that microinfluencers offer marketers certain advantages, including lower costs and higher availability (Ehlers 2021). Their authenticity also provides an advantage when promoting hedonic (Park et al. 2021) or local (Silalahi 2021) products. Nevertheless, few empirical investigations have specifically examined the persuasion techniques they deploy.

Charisma Strategy

Rather than formal Weberian terminology, we use charisma as a media entertainment concept to refer to the compelling and magnetic qualities (attractiveness and aspirational meanings) individuals (e.g., influencers) possess and draw on to inspire devotion in their followers (Hendriks 2017; Cocker and Cronin 2017). Moreover, such qualities of individual personality are considered extraordinary and endowed with exceptional powers (Cocker and Cronin 2017). In the endorsement literature, celebrities are notable icons who possess the charisma to capture the public’s admiration and fascination. Given their talents and air of glamour, they personify the images and values desired by a materialistic society. As brand endorsers, their attractiveness and the aspirational images they project allow them to achieve a megaphone effect (McQuarrie, Miller, and Phillips 2012). This, in turn, encourages consumers to model their consumption behaviors (Ki and Kim 2019).

Given the success of celebrity endorsers, it is logical for influencers to come off as more charismatic with the aim of improving their brand promotional capabilities. Indeed, studies on influencer marketing have provided evidence for the use of this strategy (Lou and Yuan 2019). However, there is a caveat. While charisma strategy aligns well with macroinfluencers’ “taste leadership” positioning, it presents an obvious challenge to microinfluencers whose main appeal lies in their boy/girl-next-door images. The latter needs to balance being attractive enough to draw attention and admiration and avoiding being perceived as distant from their followers. Thus, microinfluencers may find charisma strategy inapplicable.

Content Strategy

In addition to source factors (e.g., charisma), consumers’ adoption of information is based on the quality of the message. Factors such as message complexity, relative advantage, informativeness, and playfulness can enhance a message’s quality (Sussman and Siegal 2003) and entice consumers to adopt and disseminate the information contained within (Hsu, Lin, and Chiang 2013). Some celebrities are experts in a given field, and their messages are perceived as authoritative, credible, and filled with evidential support (Hung, Li, and Tse 2011). Microinfluencers, who may choose not to deploy a charisma strategy, may compensate by adopting a content strategy. Indeed, microinfluencers are perceived by consumers to possess more product knowledge than macroinfluencers (Kay, Mulcahy, and Parkinson 2020).

Microinfluencers can improve their messaging and content by conducting detailed research on the brands they discuss. By studying the features of different brands, they can present pros and cons using an objective approach and with confidence. Alternatively, they can take on a personal approach, try out the brands, and share their firsthand experiences with followers (Hung 2021). Furthermore, they can provide personalized comments and advice to follower inquiries, a tactic especially effective for microinfluencers thanks to their smaller audience base. However, macroinfluencers usually find it impossible to provide personalized responses, as they have too many followers. Some hire assistants to do the job, but such responses may not be perceived as personal or authentic. A study has pointed out that this is a limiting condition for macroinfluencers (Richardson 2023).

Bonding Strategy

Influencers can enhance their effectiveness by improving their bonds with followers. The shortened social distance between influencers and followers renders the audience more receptive to the messages of influencers and enhances their desire to mimic the influencers (Ki and Kim 2019). The advantage of a bonding strategy has been widely discussed in the parasocial relationship literature, with empirical support for its effectiveness for both celebrities and megainfluencers (Gong and Li 2017; Hung 2020; Jin, Ryu, and Muqaddam 2021).

Perhaps more specific to microinfluencers is their ability to bond well with small, unique groups of followers (Marques, Casais, and Camilleri 2020). Richardson (2023) noted that after influencers gain 100,000 followers, their level of personalized engagement drops by 20%, thus placing microinfluencers in an advantageous position to use this strategy. In China, some influencers from rural areas use their local dialects to better relate to audiences around the region. In Indonesia, microinfluencers have been found to be more effective in promoting local products (Silalahi 2021). In either case, it is likely that their smaller follower base, personalized engagement, and shared affection for and affiliation with the region allow microinfluencers to bond closely with their followers and achieve effectiveness and success.

Figure 1 shows the proposed model that integrates the three strategies that microinfluencers may use to boost audience likes, audience shares, and brand sales. Our study assessed the model using e-commerce influencers of FMCG in China. An investigation of the effectiveness of microinfluencers in FMCG complements past studies that have focused on microinfluencers in luxury and hedonic goods (Marques, Casais, and Camilleri 2020). We chose China because the advancement of social media platforms and the continuous proliferation of products have given rise to many influencers in this market. Collectively, they have turned the market into an influencer economy with several notable characteristics.

Figure 1. Persuasive strategies for microinfluencers.

First, many local and foreign brands are spending more on influencer marketing in China to promote their brands and spur sales. A 2021 survey showed that about two-thirds of advertisers selected influencer promotion as a focus in their social marketing plans (Thomala 2023). As Chinese consumers have become more pragmatic, influencers’ trustworthiness and authenticity have become important elements in successful marketing campaigns. Furthermore, industry studies have revealed that up to two-thirds of consumers were convinced by an influencer review or endorsement video, and many purchased items promoted by influencers (Aventura 2021).

Second, influencers are unwittingly embroiled in a serious competition for success. TOPKLOUT, a media data agency, pointed out that by 2020 China had more than 9 million influencers with more than 10,000 followers; this is more than anywhere else in the world. In 2022, the average number of daily posts published by influencers was 37.5 million. During this year’s online shopping festival 618, eight influencers achieved a gross sales value of over USD 15.4 million individually through the TikTok livestreaming channel alone.

These events showcase that China, as a global forerunner in influencer marketing, is taking the lead in building an influencer business ecology (Wu 2019). This ecosystem is creating new business segments and opportunities, such as paid knowledge, livestream tipping, and gifting, Internet protocol (IP) authorization, influencer content analysis, and specialized supply chains. As a result, the influencer economy formed a novel value chain at its dawn, connecting firms across business sectors and presenting a huge market potential for those involved, whether they be firms, agencies, or other influencers.

Hypotheses Development

Impacts of Charisma, Content, and Bonding Strategies on Audience Likes

Brand endorsement research has shown that charisma can be attained by celebrities in several ways: through their talents (e.g., inspiring admiration through the silver screen/sports arena), appearance (e.g., outstanding looks or air of glamour), staged images (e.g., on TV and social media), and lifestyles. Similar to celebrities, most macroinfluencers have ample financial and professional support to showcase these attributes, allowing them to deploy their charisma strategy and enhance their effectiveness. However, for many microinfluencers, these tools may not be available. Aside from constraints associated with resource availability, they are concerned that deploying these tools may alienate their followers. Should individual influencers choose to improve their charisma, their core appeal may be lost, and that can deal a devastating blow to their microinfluencer career. Thus, we propose the following:

H1(a): A charisma strategy is positively associated with audience likes for macroinfluencers.

H1(b): Microinfluencers are less likely to deploy a charisma strategy to increase their audience likes.

As discussed, the marketplace is highly competitive. Firms eager to engage in brand extension and new product development may rely on microinfluencers who are highly knowledgeable in their product category to help diffuse the latest trends among consumers (Shen 2021). Indeed, many consumers may find microinfluencers to be useful and trustworthy sources of information. Drawing on the theory of information relevance (Zhang and Choi 2022), microinfluencers have used content topicality, novelty, understandability, reliability, authenticity, and interestingness to create content that is useful and desired by the audience (Rungruangjit and Charoenpornpanichkul 2022). By focusing on a well-defined area of expertise, microinfluencers share information with like-minded individuals and bypass the issue of having to bond with “outsiders” (Bazerman 2001). Thus, it is unsurprising that many microinfluencers build their reputations and follower bases using content strategy. This differs from macroinfluencers, who are more likely to use their charisma to promote brands across different product categories. In other words, microinfluencers are more likely than macroinfluencers to distinguish themselves on content-oriented persuasiveness. Here, we propose the following:

H2(a): Content strategy is positively associated with audience likes for microinfluencers.

H2(b): Macroinfluencers are less likely to deploy content strategy to increase audience likes.

A key factor affecting the success of all types of influencers is the bond they cultivate with their followers. Among macroinfluencers, previous studies have confirmed that source characteristics are a strong driver of the parasocial relationships they develop with followers (Gong and Li 2017; Jin, Ryu, and Muqaddam 2021). Without charisma, microinfluencers may use their authenticity to overcome the bonding issue, as highlighted by recent industry practices. An influencer study by Forbes (Ehlers 2021) showed that microinfluencers are perceived to possess a higher level of authenticity; more importantly, they practice deeper engagement (including interactivity) with their followers. In the United Kingdom, a study showed that 82% of consumers are “highly likely” to follow a recommendation made by a microinfluencer because they had formed a relationship built on trust, friendship, and honesty with their followers (Richardson 2023). We posit that despite the different paths they take, both micro- and macroinfluencers are eager to improve their bonds with their audiences. We propose the following:

H3: Bonding strategy is positively associated with audience likes for both macro- and microinfluencers.

Impacts of Charisma, Content, and Bonding Strategies on Audience Shares

We posit that while audience likes and shares are related, they differ in a fundamental way in that audience likes reflect the audience’s personal feelings toward the influencer, while shares denote the audience’s endorsement of the influencer. In addition, audience shares are affected by the relationship between the forwarder and the receiver.

Since the emergence of social media platforms, audience shares have been a significant indicator of success because shared content has a higher chance of being consumed and paid attention. The same applies to influencers. Audience shares allow an influencer’s follower base to grow, increasing the influencer’s impact and, ultimately, his or her financial reward.

We posit that macroinfluencers rely more on charisma strategy than content strategy to enhance audience shares. In contrast, microinfluencers understand their limitations and rely more on content strategy to enhance audience shares. In a field experiment, Viral Nation selected 175 college-aged women with 5,000 Instagram followers each to promote Pink, a subbrand of Victoria Secret (Tatum 2022). The results showed that, though lacking charisma, these microinfluencers were effective in generating shares and enhancing Pink’s acceptance through the content they created. Thus, we propose the following:

H4(a): Charisma strategy is positively associated with audience shares for macroinfluencers.

H4(b): Microinfluencers are less likely to deploy charisma strategy to enhance audience shares.

H5(a): Macroinfluencers are less likely to deploy content strategy to enhance audience shares.

H5(b): Content strategy is positively associated with audience shares for microinfluencers.

Following the previous discussion, we posit that bonding strategy will be highly significant for both micro- and macroinfluencers in enhancing audience shares. For macroinfluencers, followers’ desire to bond with them would likely be aspirational; that is, they aim to bond with macroinfluencers given their charisma. For microinfluencers, the underlying reason for followers’ bonding desire would differ. While some followers may be awed by certain influencers’ product expertise, others may bond with them because of their deeper engagement efforts (Ehlers 2021). Thus, we propose the following:

H6: Bonding strategy is positively associated with audience shares for both macro- and microinfluencers.

Regarding brand sales, for microinfluencers we posit that both audience likes and audience shares would exert significant positive impacts on brand sales due to a homophilous follower base. Reflected in the homophily principle (McPherson, Smith-Lovin, and Rawlings 2021), people with similar interests, backgrounds, and preferences tend to consume similar content or products (Liu-Thompkins 2012; Leung, Gu, and Palmatier 2022). However, for macroinfluencers, whose follower base is broader and more diverse, liking behavior can take on different levels. Aside from liking content in which he or she is genuinely interested, a follower can be in a “like-back” network hoping that other users will reciprocate the likes and increase his or her popularity (Sen et al. 2018). In addition, one can like a post to mean “content being read” or “noted.” Still, others may like posts created by influencers as a supportive gesture, regardless of content (Sen et al. 2018). In measuring influencer marketing effectiveness, the number of reposts/shares (not the number of likes) is employed as an engagement proxy in the macroinfluencer sample (Leung, Gu, and Palmatier 2022). Correspondingly, audience shares that symbolize a higher level of commitment and a referral from friends will exert a significant impact on brand sales for macroinfluencers. Thus, we propose the following:

H7(a): Audience likes and shares are positively associated with brand sales for microinfluencers.

H7(b): Audience shares are positively associated with brand sales for macroinfluencers.

Study Design

We tested the hypotheses using data obtained from an established TikTok-authorized information provider, Douchacha, which specializes in collating and analyzing TikTok video data in China. Douchacha organizes video data according to product categories and influencer types. These data have been used by many domestic and foreign firms (e.g., Miaozhen.com, China’s largest e-commerce service provider) and in recent academic studies (e.g., Hung, Tse, and Chan 2022). The data for this study were collected between June and December 2020 and included both quantitative (e.g., brand sales) and qualitative data (e.g., audience live texts). The unique set of quantitative data included the number of followers an influencer has as well as the length, the number of audience likes, and the number of audience shares of each brand-sponsored video. The price of the promoted product and the number of units sold through the TikTok platform (i.e., when an audience member or friend clicks the link related to a video to complete a purchase) were also captured. For qualitative data, audience live texts were collated by Douchacha. Live texts are voluntary responses the audiences make when they watch a video, and they help us understand the effectiveness of an influencer’s charisma and bonding strategies.

Several criteria were used in selecting the TikTok videos, the unit of analysis. First, we limited the videos to those that promoted a single brand only, as videos promoting multiple brands may have potential confounders. Second, we selected videos in four categories of highly promoted consumer products: fashion, cosmetics, food and drinks, and household products. Third, we restricted the data set to two types of influencers: macroinfluencers (5 million followers) and microinfluencers (30,000 to 100,000 followers). This categorization of microinfluencers is in line with those used in recent studies (Campbell and Farrell 2020), while the categorization of macroinfluencers has been adjusted upward to reflect the blooming influencer market in China (Robles 2020). Fourth, we randomized the selection of videos using these criteria. The final data set included 189 microinfluencer and 228 macroinfluencer videos for a total of 417 videos.

Measures

Table 1 displays the variables and their operationalization. Table 2 provides examples of the TikTok videos used in this study. Table 3 presents the descriptive statistics and correlation matrix of the quantitative variables. The brand-sponsored TikTok videos were brief, lasting from 15 seconds to several minutes. The average length of these videos for the current sample was 46.8 seconds. Brands usually grant influencers a great deal of autonomy in generating their content. The content is therefore unique, authentic, and engaging, with diverse themes, such as product unboxings, demonstrations, and storytelling. The video titles were also carefully crafted by the influencers. Most video titles were long, allowing us to use the title content to understand the content strategy deployed.

Table 1. Variable operationalizations and codebook.

Variable	Definition	Operationalization	References
Dependent variables
Brand sales	Ordering directly from the link provided in video; nondirect sales (audience purchases through other means) not included	Transaction value recorded on Douchacha	Yang, Zhang, and Zhang (2021)
Audience shares	Number of shares of an influencer video	Number of shares recorded on Douchacha	Valsesia, Proserpio, and Nunes (2020)
Audience likes	Number of likes generated by an influencer video	Number of likes recorded on Douchacha	Valsesia, Proserpio, and Nunes (2020)
Independent variables
Influencer charisma	Degree to which audience appeals to the influencer personally	Audience live texts that show the extent of audience appeals to influencer
Message content	Degree to which video title focuses on product and benefits	Extent of video title mentions product and benefits
Influencer–audience bond	Degree to which audience parasocially responds to influencer video	Audience live texts that reflect parasocial response to influencer
Control variables
Video length	Length of video	Length measured in minutes	Rajaram and Manchanda (2020)
Product price	Price of promoted product	Unit price of promoted product (RMB)	Yang, Zhang, and Zhang (2021)

Table 2. Examples of TikTok videos in the study.

Example	Influencer Name	Number of Fans	Product and Video Title (by Influencer)	Duration of Video (in seconds)	Live Texts	Number of Likes	Number of Shares	Price (RMB)	Dollar Sales (RMB)
1 (Macro)	He Dan Tong	558,280	Food/see notes	29	See notes	411,351	18	69	8,639
2 (Macro)	Big LOGO	1,9385,000	Food/see notes	106	See notes	548,000	23,000	539	1,563,100
1 (Micro)	SoSoMeiDa	33,042	Fashion/see notes	10	See notes	2,348	1	159	159
2 (Micro)	Cameraman in Guangzhou	44,000	Fashion/see notes	10	See notes	1,045	25	79	132,404

Note.

1 – Macroinfluencer (food): Video title is 管住嘴迈开腿, 明天我是李宣美#沙雕#迷惑行为大赏#向往的生活; translates as “Control your appetite; Relax; Tomorrow I will be ‘Li Shun Mei’; #Seductive bonus; #Aspired lifestyle.” Top live texts are 真实; 减肥; 看到; 运动; 没错; 知道; 没有; 声音; 可爱; translates as “Real; Weight loss; Can see; Exercise; No mistake; I know; I don’t have it; Sound; Lovely.”

2 – Macroinfluencer (food): Video title is 一定要看到最后！55000块东方明珠空中酒店！带走的东西多到拿不动！ #抖音心动餐厅 #酒店中的劳斯莱斯; translates as “Must see till the end; RMB 55K Pearl Sky Hotel; Bonus beyond you can move; #Douyin heart-warming restaurant; #Hotel’s Rolls Royce.” Top live texts are 想去；地方；睡觉；感觉；喜欢；不好；视频；见识；有钱；拿走; translates as “Want to go; Place; Sleep; Feels good; Like it; Not as good; Video; I know now; Have money; Take the bonus.”

3 – Microinfluencer (fashion): Video title is 过了40岁女人试试减龄裙, 洋气时尚, 有魅力#70后 #减龄穿搭 #谁穿谁好看 #秋季新款 @抖音小助手; translates as “Over 40 reduce your age dress; Foreign style; Attractive; #Post-70s; #Age-reduction mix; #Nice look; #Autumn dress, Douyin.” Top live texts are 好看; 身材; 衣服; 漂亮; 女人; 裙子; 不错; 前提; 没有; translates as “Nice look; Good body shape; Dress; Beautiful; Lady; Dress; Not bad; Don’t have.”

4 – Microinfluencer (fashion): Video title is 昨天广州地铁三号线看到的小姐姐, 还可以捞到嘛, 太好看了趴！！！#把生活拍成电影 #每日穿搭 #看三十秒; translates as “Seen yesterday at Guangzhou MTR #3; Can have it; Excellent; Ordinary life as in movie; #OOTD; #30 Sec.” Top live texts are 可以; 打开; 特效; 喜欢; 帅帅; 浪花; 看到; 姐姐; 线人; 看见; translates as “Acceptable; Open it; Special effect; Like it; Smart looking; Trendy; Seen it; Beautiful lady; Middleman; Saw it.”

Table 3. Descriptive statistics and correlations.

Variables	M	SD	Correlation Matrix
			1	2	3	4	5	6	7
1. Brand sales	438153.92	2428839.99	1
2. Influencer charisma	9.35	14.16	1.08*	1
3. Message content	24.34	18.21	−.007	−.317**	1
4. Influencer–audience bond	3.11	5.12	−.011	−.070	−.107*	1
5. Audience shares	438.26	1176.33	.297**	.115*	.007	.007	1
6. Audience likes	88032.01	180158.45	.347**	.353**	−.151**	−.007	.293**	1
7. Video length	46.80	51.95	.191**	.271**	−.081	−.021	.131**	.359**	1
8. Product price	105.05	179.45	.053	−.031	−.022	−.003	−.004	.066	.021

Note. The statistics are based on the original values of variables (i.e., without natural logarithms).

*p < .05; **p < .01.

The dependent variables (DVs) and control variables used were similar to those used in other studies (for reference, see Table 1). The coding of live texts that made up the independent variables (IVs) was original.

Audience Likes

Douchacha recorded the number of people who clicked “Like” on each video. We used the total number of likes as a DV.

Audience Shares

Douchacha recorded the number of shares that a video generated. We used the total number of shares as a DV.

Brand Sales

Douchacha recorded the transaction value (in RMB) as the audience placed orders directly via the link provided in the video. We operationalized the total revenue generated by a video as brand sales (see also Yang, Zhang, and Zhang 2021). If an audience member purchased the promoted product using other means (i.e., without clicking the link), these “indirect” sales were not included.

We applied natural log transformation to these variables prior to conducting the regression analysis. Natural log transformation is often deployed in studies using data sets, as standardized data can better explain nonlinear functions between DVs and IVs (Harre, Lee, and Pollock 1988; Osborne 2010). As influencer effects often display a nonlinear data pattern (Li, Lai, and Chen 2011), it was appropriate to apply log transformation to the DVs.

Content Analysis to Obtain IVs

When conducting content analysis, we followed the procedures reported in seminal studies (e.g., Tse, Belk, and Zhou 1989) and subsequent discussions of this methodology (Kolbe and Burnett 1991). The team first developed a detailed codebook that collated the items to be analyzed. Two raters coded the items accordingly and checked their scores to determine interrater reliability. Then, they discussed any differences in coding, modified the codebook, and repeated the process until reliability was high. In this study, the tasks of content analysis and classification were relatively simple, and two rounds of modification were carried out. In the first round, with 300 items coded, interrater reliability was 85%. After discussion and modification, the second round of coding reached over 90% interrater reliability, with the reliability of individual constructs ranging from 91.64% to 93.87%.

Message Content (Content Strategy)

To examine how the influencers deployed content strategy to enhance their effectiveness, we analyzed the content found in the titles of the TikTok videos. The titles were developed by the influencers and thus framed the appeals the influencers intended to make in the videos. The titles featured affective (such as “happy” and “heartwarming”), location-specific (“name of a village” or “city”), entertaining (“fun,” “enjoyable”), and most importantly, “product and benefit” elements. We operationalized the video title content featuring products and benefits as the measure of content strategy, as such titles included facts on the product (e.g., “little black dress for spring”), its benefits (e.g., “this is so beautiful, it will make your friends jealous”), and social appeals (e.g., “great for the whole family!”). The more a video mentioned the product and its benefits in the title, the more an influencer was said to rely on content strategy.

Influencer Charisma (Charisma Strategy)

To examine the extent to which the influencers used charisma strategy in the videos, we deployed the live texts made by the audience that referred to the influencers. Douchacha reports the live text that the audience keyed in when they watched the video. The live texts were read by a machine and collated into various high-frequency categories. The live texts included personal addresses (e.g., “Hi, big brother/sister,” “Hi, teacher,” or “Hi, hot KOL”), personal remarks (e.g., “beautiful,” “handsome,” or “smart-looking”), and greetings (e.g., “great to see you”). They were positive and matched the conceptual meaning of charisma—that is, compelling attractiveness or charm that inspires devotion in others. The more a video receives these remarks, the more likely the influencer had used charisma strategy in the video. Audience members who felt negative (or neutral) toward an influencer would be unlikely to have left a live text in this category.

Influencer–Audience Bond (Bonding Strategy)

To assess an influencer’s effort to bond with the audience, we deployed texts that reflected the way the audience responded to the influencer. The live texts in this category included acknowledgment of what the audience had learned (e.g., “got it,” “now I know”) and the way they received the message (e.g., “loud and clear,” “well received”). They reflect the parasocial bond between the audience and the influencer, as the former “replies” to the influencer. The more a video elicits these indications of a parasocial bond, the more likely an influencer had deployed bonding strategy in the video.

Control Variables

We controlled for video length (in seconds) and product heterogeneity. The latter is based on the unit price of the promoted product in the video (in RMB).

Analysis and Results

We first checked for multicollinearity and endogeneity issues in our data. The variance inflation factors (VIFs) among the IVs in this regression model were around 1.0, indicating that multicollinearity was not a problem in this study. To inspect the issue of endogeneity, a test-stage least squares (TSLS) was conducted using the EndoS SPSS macro (Daryanto 2020) with an instrumental variable of the video launch time, which is considered a strategic decision related to IVs but not to DVs (i.e., recorded brand sales). The weak instrument test gave a Cragg–Donald statistic of less than 10, showing evidence that the instrumental variable was weak, and ordinary least squares (OLS) regression was sufficient for our model. In addition, the IVs were coded by research assistants using content analysis, while the DVs were collated from Douchacha. Thus, there was no common method bias in the study.

The skewness and kurtosis values for all variables were also determined. Because of substantial variations, we performed logarithmic transformations for audience likes, audience shares, brand sales, and comments to ensure their normality. To test the hypotheses, we ran a series of OLS regression analyses. We report the results for audience likes and shares in Table 4 and brand sales in Table 5.

Table 4. Hypothesis test: Number of audience likes and shares as dependent variables (DVs).

Variables	DV: Audience Likes		DV: Audience Shares
	Microinfluencers	Macroinfluencers	Microinfluencers	Macroinfluencers
Control variables
Product price	−.142^† (−.190)	.079 (1.254)	−.097 (−1.314)	.112^† (1.69)
Video length	.043 (.586)	.273** (4.306)	.020 (.267)	.191** (2.87)
Independent variables
Influencer charisma (charisma strategy)	−.122^† (−1.647)	.144*^# (2.069)	−.147^† (−1.962)	.143*^# (1.96)
Message content (content strategy)	.143* (1.931)	−.092^# (−1.325)	.039 (.526)	−.002 (−0.30)
Influencer–audience bond (bonding strategy)	.060 (.831)	.058 (.906)	.132^† (1.805)	.152* (2.24)
R^2	.064	.139	.051	.097
F	2.501*	7.17***	1.97*	4.53***
df	5, 187	5, 222	5, 184	5, 215

Note. Standardized regression coefficients are reported (t values in parentheses).

^#Comparison of effect size across micro- and macroinfluencers is significant.

***p < .001; **p < .01; *p < .05; ^†p < .10 (two tailed).

Table 5. Hypothesis test: Brand sales as dependent variables.

Variables	Dependent Variables: Brand Sales
	Microinfluencers	Macroinfluencers
Control variables
Product price	−.191** (−2.912)	−.051 (−0.91)
Video length	−.026 (−.408)	.00 (1.57)
Independent variables
Audience shares^#	.317** (3.720)	.658** (9.16)
Audience likes^#	.165* (1.928)	−.14* (−1.92)
R^2	.254	.370
F	15.69***	32.74***
Degree of freedom	2, 184	2, 223

Note. Standardized regression coefficients are reported (t values in parentheses).

^# Comparison of effect size across micro- and macroinfluencers is significant.

***p < .001; **p < .01; *p < .05; ^†p < .10 (two tailed).

Effects of Charisma, Content, and Bonding Strategies on Audience Likes

As shown in Table 4, we found that influencer charisma significantly drives audience likes for macroinfluencers (β = .144, p < .05; column 2), supporting hypothesis 1(a), but not for microinfluencers, thus supporting hypothesis 1(b). However, message content significantly drives audience likes for microinfluencers (β = .143, p < .05; column 1), supporting hypothesis 2(a), but not for macroinfluencers, thus supporting hypothesis 2(b). The difference in effect size for the two paths (charisma: β = .095, t = 2.61, p < .01; message content: β = −.137, t = 2.30, p < .05) between micro- and macroinfluencers was also significant (F (9, 408) = 79.21, p < .001), as hypothesized in hypotheses 1(b) and 2(b). Bonding strategy was insignificant for both micro- and macroinfluencers; hence, hypothesis 3 was not supported.

Effects of Influencer Charisma and Message Content on Audience Shares

We ran separate analyses for audience shares, as we posited that they differ from audience likes. While audience shares capture the propensity to share the video with others, audience likes capture the personal affect toward the video and/or influencer.

For microinfluencers, the results (column 4, Table 4) indicated that message content (β = .039) is insignificant, so hypothesis 5(b) was not supported; but influencer–audience bond (β = .132, p < .05) exerts a significant positive impact on audience shares, supporting hypothesis 6. Interestingly, charisma strategy (β = −.147, p < .1) effect was negative with marginal significance, thus supporting hypothesis 4(b).

As for macroinfluencers (column 5), influencer charisma (β = .143, p < .05) and influencer–audience bond (β = .152, p < .05) were found to significantly promote audience shares, thus supporting hypotheses 4(a) and 6. Furthermore, the results showed that the effect of influencer charisma on audience shares was stronger for macroinfluencers than for microinfluencers (F (9, 408) = 38.71, p < .001; β = .107, t = 2.41, p < .05), demonstrating that macroinfluencers are more effective in utilizing their charisma to enhance audience shares than microinfluencers. In addition, as suggested in hypothesis 5(a), content strategy for macroinfluencers was insignificant (β = −.002, p > .1).

Effects of Audience Likes and Audience Shares on Brand Sales

For microinfluencers, we found that both audience shares (β = .317, p < .01; column 2 in Table 5) and audience likes (β = .165, p < .05) had significant positive impacts on brand sales, confirming hypothesis 7(a). As suggested in hypothesis 7(b), for macroinfluencers, audience shares were significant (β = .658, p < .01), but audience likes were not (column 3). In other words, audience likes cannot serve as an effective predictor of brand sales in relation to macroinfluencers. Differences in effect size (β = −.128, p < .05 for shares; β = .393, p < .05 for likes) were found to be significant across micro- and macroinfluencers (F (5, 412) = 95.70, p < .001). Interestingly, for microinfluencers, both audience likes and audience shares registered significant main effects on brand sales. Table 6 summarizes the results of the hypothesis testing. We discuss the findings in the next section.

Table 6. Summary of hypothesis testing results.

Hypothesis	Relationship	Result
1(a)	Charisma strategy would be positively associated with audience likes for macroinfluencers	Supported
1(b)	Microinfluencers would be less likely to deploy charisma strategy to enhance their audience likes	Supported
2(a)	Content strategy would be positively associated with audience likes for microinfluencers	Supported
2(b)	Macroinfluencers would be less likely to deploy content strategy to enhance audience likes	Supported
3	Bonding strategy would be positively associated with audience likes for both macro- and microinfluencers	Not supported
4(a)	Charisma strategy would be positively associated with audience shares for macroinfluencers	Supported
4(b)	Microinfluencers would be less likely to deploy charisma strategy to enhance audience shares	Supported
5(a)	Macroinfluencers would be less likely to deploy content strategy to enhance audience shares	Supported
5(b)	Content strategy would be positively associated with audience shares for microinfluencers	Not supported
6	Bonding strategy would be positively associated with audience shares for both macro- and microinfluencers	Supported
7(a)	Audience likes and shares would be positively associated with brand sales for microinfluencers	Supported
7(b)	Audience shares would be positively associated with brand sales for macroinfluencers	Supported

Discussion and Summary

This study examined how microinfluencers thrive, an understudied topic in the influencer marketing literature. This is a salient and timely topic, as the marketplace is increasingly becoming overwhelmed by brand proliferation, with products’ performance and benefits unclear to consumers. This development has urged consumers to seek independent advice from like-minded but knowledgeable peers, such as consumer communities (Hung and Li 2007). As influencers are perceived as authentic and trustworthy (Hung 2021), the information provided by them is highly valued. This suggests that influencer marketing has earned a valuable position in the advertising and branding ecosystem and will continue to thrive.

The growth of microinfluencers is fascinating because their emergence is counterintuitive and defies existing theories in the conventional brand endorsement literature. Most microinfluencers do not have a high level of charisma, they are not expert spokespersons, and often their messages are not professionally delivered. In short, microinfluencers do not have the theoretical support to thrive. Yet, surprisingly, there are at least three reasons for their success. First, celebrity endorsement has often been perceived as “paid and scripted,” thus opening the door for lay spokespersons who can make genuine remarks. Second, through social media, lay spokespersons can interact directly and personally with their followers. This deep follower engagement enables microinfluencers to cultivate their follower base in a down-to-earth and authentic manner. Third, the low cost of uploading shared content on social media means that the career entry cost for microinfluencers is not high. Indeed, surveys have revealed that around one-third of U.S. university students want to be influencers. In China, the percentage is over 50%, and in Malaysia, it is as high as 72%. Thus, we will likely see a growing number of influencers in the future.

The study focused on three persuasion strategies—namely, charisma, content, and bonding—that microinfluencers may deploy. Using a major influencer database, our study gathered objective performance data on influencer TikTok videos and derived objective ways to measure these strategies. The study made use of three output parameters (likes, shares, and brand sales) and three strategy constructs for two groups of influencers. Using macroinfluencers as a comparison group, the study determined the strategies that underpin the performance of microinfluencers.

The results were consistent with our postulations and confirmed 10 of the 12 hypotheses. For microinfluencers, charisma strategy does not significantly affect audience likes or shares. This is in contrast to content strategy, which is significant in affecting audience likes. Bonding was found to be significant for audience shares and only marginally so for likes (p < .1). These results strongly support microinfluencers’ reliance on content strategy. Meanwhile, for macroinfluencers, charisma strategy is significant in affecting audience likes and shares, but content strategy is insignificant for both. Bonding is positive in affecting shares only. These results support macroinfluencers’ reliance on charisma.

These findings are consistent with the career challenges that microinfluencers face. Compared to macroinfluencers, microinfluencers have less strategic freedom when it comes to enhancing their performance. Given the intense competition in the profession, it is advisable that microinfluencers focus on content strategy to gain likes and shares. Our findings showed that charisma strategy exerts a negative (though not significant) effect on likes and shares, underscoring the potential pitfalls of charisma strategy for microinfluencers.

Interestingly, bonding strategy does not significantly affect likes for either micro- or macroinfluencers but positively affects shares for both influencer groups. This finding confirms the need to analyze likes and shares separately. Because parasocial bonds may trigger more shares than likes, this area calls for more research efforts.

Theoretical Implications

This research enriches the emerging influencer literature by assessing the persuasion techniques deployed by influencers. Prior studies predominantly focused on analyzing highly successful (macro-)influencers as a basis for developing theories related to effective influencer marketing strategies. However, influencers are not all the same (Hung, Tse, and Chan 2022). Each influencer type “comes with a unique skill set and therefore a unique strategic marketing purpose” (Campbell and Farrell 2020, 476). By conceptually delineating and empirically verifying an influencer framework for microinfluencers, this study unraveled the salience of content strategy and built on Park et al.’s (2021) findings to demonstrate the relative effectiveness of brand endorsements promoted by microinfluencers compared to those promoted by macroinfluencers.

In addition, we identified that charisma, content, and bonding strategies are employed by influencers, although micro- and macroinfluencers use them to varying degrees. This study substantiates Lou and Yuan’s (2019) social influencer value model, confirming the value of the message in addition to source factors in driving influencer effects. This multiroute of influencer promotion emphasis on source, message, and bonding extends McGuire’s (2001) communication–persuasion matrix and builds on what Hung, Tse, and Chan (2022) uncovered in the context of influencer marketing. Going forward, this study may form the basis of a comprehensive framework for future research to better understand the equivocal findings regarding the persuasiveness of different influencer types.

Moreover, the extant endorsement literature has often cited the positive impact of influencers with a large number of followers (Schwarz 2020). The present study challenged this perspective by highlighting the different behaviors of followers in liking and sharing sponsored posts made by macro- and microinfluencers, which in turn drive brand sales. The results echo findings in the information diffusion literature (Li, Lai, and Lin 2017; Rudat, Buder, and Hesse 2014), indicating that audience liking and sharing, while interconnected, are influenced by distinct underlying processes.

Finally, our investigation of content characteristics revealed some of the unique aspects of influencer marketing in today’s era, using TikTok as our study context. The brevity of video-based brand promotion through influencers demands concise, authentic, and creative content to engage audiences. TikTok’s algorithm is effective at surfacing content to users, even content from relatively unknown influencers. This helps to level the playing field for influencers of varying degrees of fame and allows researchers to make cross-influencer comparisons. As the field of influencer marketing research grows, it is imperative for future studies to incorporate brand sales, leveraging TikTok’s e-commerce integration, as a metric for assessing the effectiveness of influencer marketing. This study thus established a tangible link between theoretical insights and actionable strategies, providing a road map for future research in this dynamic field.

Managerial Implications

Microinfluencers have now become an integral part of influencer marketing, and they are a channel used by sporting goods brands (e.g., Nike, Adidas), retailers (e.g., Sephora, Target), service brands (e.g., Starbucks, Dunkin’), and fashion brands (e.g., Daniel Wellington, Sperry) alike. The findings of this study offer several insights into this practice. First, both micro- and macroinfluencers may enhance firm performance, though the two influencer types should concentrate on their respective strengths: content strategy for microinfluencers and charisma strategy for macroinfluencers.

Second, the findings may benefit firms in identifying the types of influencers they should employ in this competitive marketplace. Firms seeking to reach different niche markets should pay attention to microinfluencers whose message content is more appealing and whose charisma is not a relevant criterion. In leveraging large follower bases, charisma is a key attribute of macroinfluencers. Meanwhile, firms should encourage both micro- and macroinfluencers to engage and interact with followers in brand-sponsored content, utilizing the primary benefits of bonding strategy.

Furthermore, firms should relinquish control over content creation to influencers to reap the benefits of influencer marketing. While there is risk and reluctance for firms to surrender full control of how to disseminate the brand’s message, influencer marketing works best when the content created by influencers is authentic, personal, and engaging.

Fourth, firms must also consider the potential differences between audience likes and shares as indicators of influencers’ performance. Videos promoting a brand and/or new product may be liked because the content is creative and useful; however, firms should be aware that followers may like a video simply because of the influencers themselves, which in some cases may not be indicative of purchase intentions. However, sharing or reposting reflects a deeper form of commitment than just liking; it implies consumers’ volitional act of propagating the content to their own networks (Leung, Gu, and Palmatier 2022), which serves as a better predictor of brand sales.

Finally, influencer/talent agencies should look to identify, develop, and enhance laypersons as influencers. This study offers some insights into the particular strategies different influencers should adopt as well as the qualities these agencies should look for in potential influencers.

Future Research and Limitations

Several areas require future research. First, it is unclear whether the three strategies (content, charisma, and bonding) are mutually exclusive. Conventional wisdom acknowledges that there are limiting conditions underlying these strategies. In particular, macroinfluencers are bound by their large numbers of followers, which limits the level of personalized engagement they can extend. Likewise, microinfluencers are bound by their down-to-earth appeal, which keeps them from deploying a charisma strategy. Yet a tactic discussed in this study—the personalized advice that microinfluencers use to address follower inquiries—may embody both content and bonding components. Furthermore, technological advances (e.g., artificial intelligence [AI]) may open up engagement options for macroinfluencers to overcome their limiting conditions. Although the data in this study did not enable an examination of these issues, future research may explore the exclusivity of strategies.

Second, it is unclear whether the list of strategies investigated is exhaustive. In an interesting paper, Hung (2014) showed that entertainment is a viable communication strategy for celebrity brand endorsements. As entertainment is likely a key strategy to gain audience interest in gaming platforms (Taylor 2018), the question of exhaustiveness posited by the current model needs to be examined as researchers expand the platforms they investigate.

Another area that deserves attention is the potential moderators of the proposed model. Our database offers limited details, yet we posit that there are potential variables of interest. For example, the audience’s cognitive, affective, and emotional arousal can be interesting moderators. These factors may enhance the effectiveness of the persuasion strategies examined in this article and also point to other interesting effects (e.g., signaling or bandwagon effects) that can be explored in future works. In addition, other potential moderators, such as influencer empathy and message authenticity, were not examined in this study and need further investigation.

Finally, we analyzed the content found in live texts in this study to understand how the audience reacts to influencer videos. In some regard, we benefited from social listening, a new research tool in Big Data context. Thus, textual analysis can be an insightful extension of the current use of social listening as a research tool.

This study has several limitations. Two key data sources were not included in the study. The first is tips awarded by the audience to the influencer. Tipping is visible on screen in TikTok and is regarded as highly motivating to the influencer and likely would be instrumental to audience likes, shares, and brand sales. The second is the amount of commissions the influencer receives from product sales. As an industry practice, commissions would be higher for macroinfluencers than microinfluencers. Without these data, we were not able to fully assess the performance impact of the three communication strategies.

Our empirical context was Douyin, the Chinese version of TikTok. It remains a subject for future research to validate whether the findings of our study can be generalizable to other formats and media platforms. The increasingly diverse social media tools and platforms available for influencer marketing may require different content measures. For instance, Instagram stories, which appear for only 24 hours, can be more impactful than regular posts by riding on the psychology of millennials, specifically the fear of missing out (FOMO). Moving beyond text analysis, further research could benefit from analyzing other content features, such as creativity, entertainment, and the aesthetic value of visual elements. As research efforts continue to expand, our knowledge of influencer marketing, an exciting topic in the advertising and branding ecosystem, will be enriched.

Disclosure Statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Terri H. Chan

Terri H. Chan (PhD, University of Hong Kong) is an assistant professor, Chinese University of Hong Kong.

Kineta Hung

Kineta Hung (PhD, York University) is a professor, Hong Kong Baptist University.

David K. Tse

David K. Tse (PhD, University of California Berkeley) is a research fellow, Institute of Behavioural and Decision Science, University of Hong Kong.