Electronic Commerce Research and Applications 61 (2023) 101305

Contents lists available at ScienceDirect

Electronic Commerce Research and Applications

journal homepage: www.elsevier.com/locate/elerap

Reputational assets and social media marketing activeness: Empirical

insights from China
Anders C. Johansson a, Zhen Zhu b, *
a
Stockholm School of Economics, Sweden
b
University of Kent, United Kingdom

A R T I C L E I N F O A B S T R A C T

Keywords: We explore the linkages between social media marketing activeness and reputational assets on digital platforms
Social media marketing with a unique sample of over 8,000 customer-to-customer (C2C) sellers registered on both Taobao, China’s
Reputational assets largest C2C online shopping platform, and Sina Weibo, China’s largest microblogging platform. A unique
Electronic commerce
collaborative effort between the two platforms enables us to examine whether C2C sellers are motivated to
China
engage in marketing activities on a separate social media platform. Applying machine learning methods, we first
classify whether C2C sellers conduct social media marketing on their microblogs or not, which allows the
measurement of social media marketing activeness. We then use logistic regression models and find that earned
reputational assets such as the rating scores and the number of followers are significantly associated with social
media marketing activeness on both platforms. However, we identify a conflict of owned reputational assets such
as the shop age and the paid membership between the two platforms, which provides a potential explanation for
the limited success of the cross-platform collaboration.

1. Introduction et al., 2013; Pascucci, et al., 2018), and large corporations versus SMEs
(small and medium-sized enterprises) (Abbasi et al., 2022; Abed, 2020;
Social media marketing has become a standard practice for busi­ Chatterjee and Kumar Kar, 2020; Heavey, et al., 2020; Sinclaire and
nesses in the twenty-first century. It enables people to interact with each Vogus, 2011). The existing literature is primarily focused on various
other in a variety of ways and for firms and marketers to directly reach drivers of customer engagement on social media (de Oliveira et al.,
out to and engage with both businesses and consumers (Appel et al., 2016; Liu et al., 2019a, 2019b; Shi et al., 2016) or factors affecting a
2020). Social media marketing takes advantage of how online re­ firm’s decision of adopting and using social media as a marketing
lationships function, i.e. in a seamless, omnichannel, highly personal­ channel (Chatterjee and Kumar Kar, 2020; Dahnil et al., 2014; Guesa­
ized, and anthropomorphized way (Steinhoff et al., 2019), and has been laga, 2016; Michaelidou et al., 2011; Siamagka et al., 2015). However,
utilized, both in theory and in practice, as a facilitator of individual the usage of social media marketing in the C2C context remains
expression, a decision support tool and a market intelligence source underexplored (Puriwat and Tripopsakul, 2021).
(Lamberton and Stephen, 2016). Social media marketing incurs lower This paper takes a quantitative approach to improve our under­
costs than other traditional offline methods (Du et al., 2015), which is standing of the driving forces behind social media marketing activeness.
perhaps especially appealing to customer-to-customer (C2C) sellers We focus on C2C sellers on Taobao, China’s largest C2C online shopping
(Wang et al, 2019). The effectiveness of social media marketing has been platform. Alibaba, the parent of Taobao, started to invest in Sina Weibo
well documented in the literature (Chang, et al., 2015; Godey, et al., in 2013. After Sina Weibo’s IPO on the New York Stock Exchange in
2016; Kim and Ko, 2012; Kumar, et al., 2016). Meanwhile, the adoption 2014, Alibaba increased its ownership in the company to 31.4%, giving
and usage of social media has been studied in varying contexts such as it a 14.8% voting power. Sina Weibo was seen as constituting a crucial
individuals (Cheung, et al., 2011; Zolkepli and Kamarulzaman, 2015), component in Alibaba’s social commerce strategy (Jia and Han, 2020).
B2C (business-to-customer) versus B2B (business-to-business) (Dwivedi Alibaba thus saw an opportunity to combine its leading e-commerce
et al., 2021; Iankova, et al., 2019; López-López and Giusti, 2020; Moore, platform with the leading microblogging platform Sina Weibo, thereby

* Corresponding author.
E-mail address: [email protected] (Z. Zhu).

https://doi.org/10.1016/j.elerap.2023.101305
Received 6 October 2022; Received in revised form 1 April 2023; Accepted 22 August 2023
Available online 26 August 2023
1567-4223/© 2023 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

competing with emerging and increasingly popular social commerce reputation factors and develops the research hypotheses. Section 5
platforms such as Xiaohongshu and Pinduoduo. In that endeavor, the presents the empirical methodology by introducing the data and vari­
two platforms developed certain features meant to push sellers on ables, descriptive statistics, and then specifying the empirical model.
Taobao to use Sina Weibo to market their products. While Sina Weibo Section 6 presents and discusses the empirical results, after which the
had undoubtedly become a leading social media marketing channel for final section concludes the study.
large brands and celebrities, its popularity among the mass of C2C sellers
remained limited (Jia and Han, 2020). Nevertheless, this collaboration 2. C2C e-commerce and social media in China
between a leading e-commerce platform and social media platform
constitutes a unique opportunity to analyze how e-commerce sellers Since the early 2010s, a number of e-commerce and social media
optimize their marketing on a separate but integrated platform. There­ platforms have emerged in China. C2C e-commerce has long been
fore, this study aims to shed light on how e-commerce C2C sellers use dominated by Alibaba’s Taobao. In the social media space, on the other
integrated social media marketing. We believe a quantitative analysis hand, there are a large number of competitors, each with their own
will improve the understanding of the use of social media marketing for specific characteristics. As the access to international social media
e-commerce and, perhaps more importantly, whether a collaboration platforms such as Facebook, Instagram, and Twitter has remained
between e-commerce and social media platforms results in an effective elusive to Chinese netizens, native platforms have been developed and
way for e-commerce C2C sellers to use social media marketing in their become hugely popular. While many of these platforms were originally
sales process. To do this, we use a data sample that was collected from very similar to their international counterparts, they have evolved in
August to October in 2016 and consists of 8,637 C2C sellers (hereafter ways that make the Chinese social media landscape highly unique. In
referred to as Taobao-Weibo users). Because of a collaborative effort this section, we first introduce Taobao, the leading C2C e-commerce
between the two platforms, we were able to collect information on C2C platform in China. We then provide a brief introduction to Sina Weibo,
sellers on Taobao as well as their activities on Sina Weibo. After col­ the leading microblogging platform that we focus on in our empirical
lecting all the microblogging posts for the C2C seller sample, we then use analysis, as well as the collaboration between the two platforms that was
machine learning and natural language processing methods to deduce if initiated back in 2013.
microblogging posts in the sample incorporate social media marketing
content. 2.1. Taobao
This study makes several important contributions. First, and to the
best of our knowledge, this is the first large-scale study that empirically Taobao (https://www.taobao.com) was founded by Alibaba Group
analyzes how reputation on both e-commerce and social media plat­ in 2003. With 836 million active users (Chen, 2020), it is the largest e-
forms is associated with social media marketing activeness by e-com­ commerce platform for online shopping in China (Gao et al, 2016; Li et
merce C2C sellers. The only work we have identified studying the al, 2008). It provides C2C retailing services to customers in China as well
adoption of social media marketing in the C2C context is by Puriwat and as markets abroad. It also provides business-to-consumer (B2C) services
Tripopsakul (2021). However, whereas their study relies on self- through its sub-platform Taobao Mall or Tmall for short.
reported usage of social media platforms such as Facebook and Line, Like other e-commerce platforms, Taobao has introduced a series of
we collect objective metrics for the same individuals from both e-com­ interactive features to mitigate uncertainty in online exchange re­
merce and social media platforms. As mentioned earlier, a unique lationships (Pavlou et al, 2007). For example, sellers and buyers can rate
collaboration between Taobao and Sina Weibo makes it possible for us to each other after each transaction and the accumulated ratings are
trace Taobao sellers’ microblogging activeness, thereby providing a attached to user profiles and are therefore publicly visible (Wang et al,
unique opportunity to study C2C social media marketing activeness. 2016; Ye et al, 2009). Taobao sellers are also provided an instant
Second, we provide empirical evidence that improves our understanding messenger tool called Aliwangwang to communicate with customers
of why the collaboration between e-commerce and social media plat­ which can be used to build trust and reputation (Cai et al, 2018).
forms never became successful in terms of attracting the mass of C2C
sellers as we identify a conflict between reputational assets on the two 2.2. Sina Weibo
platforms. Third, we shed light on one of the world’s most important
C2C online platforms by examining to what extent sellers on that plat­ Chinese citizens are very active online. According to a report by We
form engage in the marketing of their products on social media. Fourth, Are Social (2019), over 1 billion Chinese citizens were social media users
we contribute to the literature on reputation and social media market­ in 2019. They spent an average of close to two hours per day on social
ing. Previous studies have primarily focused on if and how social media media. The social media landscape in China differs significantly from
marketing affects online reputation. In this study, we define reputational that found in most of the rest of the world. One important reason for this
assets as the observable reputation-related quantitative or qualitative is that the Chinese government encourages the use of the internet while
metrics accumulated on the digital platforms at the user level over time maintaining control of information (Feng and Johansson, 2019). Early
and instead focus on whether the existing reputational assets are factors Chinese adopters of social media started using Twitter soon after its
shaping social media marketing activeness. This has important impli­ inception in 2006. Soon, several similar microblogging platforms were
cations for our understanding of how online C2C actors make decisions developed in China (Qin et al., 2017), but none of these gained large
about their marketing activities. Moreover, for the reputational assets traction during this period. Social media platforms quickly became
considered, we differentiate earned reputation assets from owned repu­ channels where information flowed freely in China. However, after vi­
tation assets for both platforms. Fifth, while the ownership of a social olent riots erupted in the provincial capital Urumqi in Xinjiang in 2009,
media account is often used as a proxy for social media marketing in the the Chinese government banned the use of international social media
literature, we argue that such a proxy is too crude as it fails to capture platforms (Sullivan, 2012).
actions taken by the owner of that account. By applying machine Sina Weibo was subsequently launched in August 2009. Several
learning and natural language processing methods, we can identify other microblogging platforms soon emerged. The main competitor is
whether or not C2C sellers’ microblogging accounts are characterized by Tencent’s platform, but Sina Weibo remains the most popular micro­
social media marketing activities. blogging platform in China. While many observers tend to call Sina
The following section provides a brief introduction to the two Chi­ Weibo China’s Twitter, there are significant differences between the two
nese online platforms Taobao and Sina Weibo. Section 3 highlights the platforms. Sina Weibo allowed graphic content to be posted early on,
concept of social media marketing activeness, as opposed to social media something Twitter incorporated relatively late. One of the main differ­
marketing adoption. Section 4 then reviews the related literature on ences between the two is how the platforms are used. Twitter is often

2
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

seen as a channel used by observers of politics, policies, and social is­ activeness in the C2C context becomes feasible because we can cross link
sues. This is not limited to the US, but a phenomenon seen in many other the C2C sellers using both platforms.
places, including Asian countries such as Indonesia and the Philippines. In a broader sense, as both e-commerce and social media platforms
Weibo users, on the other hand, are often very active on the platform heavily rely on user interactions, the persistence of purchase intention
and tend to post more about themselves (Gao et al., 2012). from buyers’ perspective as well as social media marketing adoption and
In terms of the size of the user base, the report by We Are Social usage from sellers’ perspective can only be maintained with
(2019) shows that WeChat, a highly successful multi-purpose mobile relationship-building and trust (Chiu, et al., 2012; Herjanto and Amin,
app developed by Tencent since its release in 2011, remains the most 2020; Melnik and Alm, 2002). Moreover, previous studies have long
popular social media platform. In a survey, 79% of the respondents said considered reputation systems as the primary approach to establish trust
they used WeChat. However, as many as 60% also reported that they use in the sharing economy (Ter Huurne, et al., 2017; Thierer, et al., 2015).
Sina Weibo, suggesting that it remains an important channel for many in We therefore devote the next section to the discussion of reputation
China. As noted by Gong et al. (2017), the growth of Weibo has resulted factors.
in many businesses exploring it as a channel for their marketing
activities. 4. Reputational assets on digital platforms

2.3. Collaboration Previous studies have roughly explored three different groups of
antecedents of social media marketing adoption, and they are respec­
On August 5, 2013, Sina Weibo in collaboration with Taobao tively personal, organizational, and external factors (Guesalaga, 2016;
released a new module specifically designed for Taobao sellers. The new Pascucci, et al., 2018). Focusing on B2B organizations, Michaelidou
module gives the verified Taobao sellers additional capabilities et al. (2011) highlight a large variation in managerial mindsets towards
(compared with other regular Sina Weibo users) to promote their social media and show that this has direct implications for social media
merchandise. More importantly, it grants the verified Taobao sellers an adoption. Siamagka et al. (2015) find that features related to perceived
identity of “Tao,” which is highlighted in their profiles on the Sina usefulness and organizational innovativeness are positively associated
Weibo platform. Moreover, in January 2014, Sina Weibo established with social media adoption among B2B organizations. In the context of
cooperation with Alipay (Taobao and Alipay are both subsidiaries of the SMEs, Dahnil et al., (2014) identify both internal and external factors
Alibaba Group) to launch a new platform called Weibo Payment. The that could affect the adoption of social media marketing. The internal
new platform makes money transfer much easier for Sina Weibo users. factors include end users’ training and knowledge, organizational re­
The introduction of these services has encouraged more Taobao sellers sources, leaders’ attitude whereas the external factors include business
to create accounts on Sina Weibo and to make full use of it as a mar­ and economic environments. More recently the internal factors have
keting channel for their products. Unlike other social media platforms been further specified at individual and organizational levels in the B2C
such as Twitter or Facebook, the explicit collaboration between Sina and B2B contexts (Guesalaga, 2016; Pascucci, et al., 2018; Dwivedi
Weibo and Taobao has made it possible to systematically ascertain et al., 2021).
whether or not C2C sellers use social media marketing (e.g., via the In the C2C context, the organizational factors become irrelevant.
above-mentioned “Tao” or profile tags such as “Taobao seller”). The When it comes to the post-adoption social media marketing activeness
gathering of Taobao sellers on Sina Weibo thus gives us a unique op­ within the platforms, the initial technical and perceptual barriers should
portunity to analyze C2C sellers’ marketing behavior on social media. be less of a concern. As stated above, the persistence of social media
usage needs to be maintained by relationship and trust building with
3. Social media marketing activeness other users. As a result, we empirically explore the linkages between
social media marketing activeness and reputational assets on digital
What drives social media marketing? Previous studies have identi­ platforms, which are often regarded as the chief means to establish trust
fied various factors that affect social media adoption by different types in the sharing economy (Ter Huurne, et al., 2017; Thierer, et al., 2015).
of businesses (Dahnil et al., 2014; Dwivedi et al., 2021; Michaelidou We focus on reputation as its importance in e-commerce has long been
et al., 2011; Siamagka et al., 2015). However, to the best of our recognized by academics as well as practitioners (Dellarocas, 2003;
knowledge, the extant literature has not yet fully investigated the factors Melnik and Alm, 2002; Pavlou and Dimoka, 2006). Since most partici­
behind social media marketing activeness, e.g., whether firms or indi­ pants on digital platforms are anonymous, reputation signaled by user
vidual users indeed post to social media after they adopt the social media profiles and other relevant metrics is an invaluable means for customers
accounts. Although the post-adoption usage of social media marketing to mitigate uncertainty when conducting online transactions (Kim and
in terms of branded social content has been analyzed and the importance Krishnan, 2015). When developed properly, reputation can help busi­
of frequent updates has been highlighted (Ashley and Tuten, 2015), the nesses amplify their message and reach out to more customers and in
determinants of the active creation of branded social content are still some cases charge a premium for the products and services. On the other
unclear. One exception is the study by Toubia and Stephen (2013), albeit hand, a business can be ruined quickly if its customers stay away
in a noncommercial setting, which conducts a field experiment on because of a bad reputation.
Twitter to identify two types of utilities, intrinsic utility and image- Closely related to the image-related utility that motivates users to
related utility, that motivate users to post to social media. Other more post to social media (Toubia and Stephen, 2013), reputation is poten­
recent efforts of measuring social media marketing activeness include tially a key determinant of social media marketing activeness. We,
the work by Moore, et al. (2015) in the B2B and B2C contexts and the therefore, extend the study by Toubia and Stephen (2013) to the com­
one by Puriwat and Tripopsakul (2021) in the C2C context. However, mercial setting of C2C social media marketing. We measure the moti­
their frequency or intensity measures of social media marketing vating utilities by reputational assets, which are defined as the
activeness are based on self-reported survey results rather than objective observable reputation-related quantitative or qualitative metrics accu­
social media metrics. Therefore, the factors behind social media mar­ mulated on the digital platforms at the user level over time. As Floreddu
keting activeness remain underexplored on a large scale. One practical et al. (2014) point out, “reputation is a crucial intangible asset formed as
challenge in the setting of C2C is that while it is straightforward to a result of past actions and stakeholders’ direct experience”. For an
identify a big company or a famous brand’s social media account, it is far online C2C seller with a social media account, accumulative reputa­
more difficult to systematically identify C2C sellers on various social tional assets are generated from both the shopping and social media
media platforms. Thanks to the above-mentioned collaboration between sites. In this study, we classify these reputational assets into two groups
Taobao and Sina Weibo, a large-scale study of social media marketing depending on whether they are accumulatively earned or owned by C2C

3
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Fig. 1. Reputational Assets and Social Media Marketing Activeness.

sellers. A reputational asset is defined as earned reputation if it is received with social media marketing activeness by C2C sellers
by C2C sellers and is therefore relatively beyond their direct control, for Earned reputation can be built through a variety of channels. Pre­
example, the rating scores received from Taobao or the number of fol­ vious studies also find that third-party recognition can positively influ­
lowers and the verifications received from Sina Weibo. On the other ence C2C e-commerce trust (Jones and Leonard, 2008; Leonard and
hand, a reputational asset is defined as owned reputation if it is provided Jones, 2021; Yoon and Occena, 2015). In addition to recognition on the
by C2C sellers and is therefore relatively under their direct control, for e-commerce platform itself, C2C sellers can build up their reputation on
example, the number of years operating on Taobao or the resulted his­ other channels, including the social media platforms they opt to use for
torical sales and the paid membership on Sina Weibo. The paid mem­ their marketing efforts. We hypothesize that there is a positive rela­
bership could be categorized as paid reputation but is rather grouped tionship between earned reputation, this time on the social media
into owned reputation for simplicity since it is under C2C sellers’ direct platform, and social media marketing activeness. The reason for this is
control. As detailed below, we also consider the non-accumulative similar to why we believe there is a positive and more direct relationship
measure, gender, as another potential determinant of C2C social between C2C platform earned reputation and social media marketing
media marketing and other non-accumulative measures such as region activeness: a potential customer attracted via social media marketing
and price, quantity and category of products as controls. can directly observe the C2C seller’s earned reputation on the social
Previous studies on reputation and social media marketing typically media platform without following a click-through link.
examine how social media activity determines reputation, in particular H2. Earned reputation on a social media platform is positively
from a business-to-business (B2B) or business-to-consumer (B2C) associated with social media marketing activeness by C2C sellers
perspective. For example, Horn et al. (2015) examine the threat of social Earned reputation is thus likely to act as an amplifier for social media
media to corporate reputation by customers, employees, and the cor­ marketing activities. What about owned reputational assets such as
poration itself. Liu et al. (2019a, 2019b) analyze how social media overall sales as seen in the form of a ranking on the C2C platform? One
marketing affects the reputation of luxury fashion brands and shows that could argue that higher sales would lead to C2C being able to allocate
brand exclusivity is not negatively affected by strong social media more capital to social media marketing activities. It has been shown that
exposure in China. While studies such as the ones mentioned here help there is a direct link between sales (or shop age) and reputation, espe­
improve our understanding of how social media activity can influence cially in the case of Taobao (Gao et al., 2016). Moreover, the collabo­
reputation, here we instead focus on how online reputation affects social ration between Taobao and Sina Weibo encourages established C2C
media activity. C2C platforms often provide measures of activity, sales sellers with high overall sales to use social media as a verified “Tao”
volumes, or reviews that help customers navigate between alternative identity requires a minimum threshold of overall sales. This would imply
providers of the products they seek as the reputation feedback system a positive relationship between sales and social media marketing
and web site quality can positively influence C2C e-commerce trust activeness. On the other hand, sales, as proxied by an experience ranking
(Jones and Leonard, 2008; Leonard and Jones, 2021; Yang, et al., 2007). on the C2C platform, may instead suggest a measure of effort exclusively
How do reputational assets accumulated on the C2C platform influence dedicated to the platform and that there is less need for omnichannel
social media marketing activeness? C2C sellers can be more motivated marketing activities. If this holds, we would thus expect a negative
and confident when doing social media marketing if they have secured relationship between sales and social media marketing activeness. These
good ratings received from previous customers on the C2C platform. two opposing forces make it difficult to determine which of these factors
That is, a potential customer attracted via social media marketing will dominates the relationship between sales and social media marketing
eventually find out what ratings the C2C seller has received by following activeness theoretically. Similarly, the relationship between shop age, as
the shopping link embedded. In other words, a cemented reputation another owned reputational asset, and social media marketing active­
received on the C2C platform helps amplify the message that the C2C ness is unclear. We, therefore, leave it to the empirical analysis to
seller wants to spread across other channels, such as social media plat­ determine how owned reputation factors on a C2C platform are asso­
forms. We, therefore, start with the hypothesis that earned reputation is ciated with social media marketing activeness among C2C sellers.
positively associated with social media marketing activeness by C2C H3. Owned reputation on a C2C platform is associated with social
sellers. media marketing activeness by C2C sellers
H1. Earned reputation on a C2C platform is positively associated When it comes to the owned reputational assets on social media

4
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Fig. 2. Geographic Distribution of Taobao-Weibo Users.

Fig. 3. Product Distribution of Taobao-Weibo Users.

platforms such as paid membership and number of follows, their ex­ 2014. We select this period because November 11, the largest online
pected relation to social media marketing activeness by C2C sellers is shopping day in the world, occurs during this month. As a result, Taobao
unambiguously positive. Previous studies have established the linkage sellers should have substantial economic incentives to be active during
between the expected social and economic returns from social media that period. We identified 12,744 sellers who added the links to their
platforms and the willingness to pay for memberships and services (Li, Taobao shops. We then gathered data on them from both platforms
et al., 2014; Vock, et al., 2013). Because the social media marketing during the period August to October in 2016, which allowed us to
activity is eventually deployed on a social media platform, a high level of establish reliable measures for variables such as median price and the
prior owned reputation, which can also be considered as an accumulated number of items for each Taobao shop. After removing the ones with
investment in terms of both time and money spending, would indicate a invalid links or missing information for key variables, our final dataset
strong tendency to use it for marketing activity and investment returns. contains 8,637 Taobao C2C sellers with a Sina Weibo account.
H4. Owned reputation on a social media platform is positively Fig. 1 shows the geographical distribution of our sample of Taobao-
associated with social media marketing activeness by C2C sellers Weibo users. The regions with over 500 Taobao-Weibo users are
Finally, Fig. 1 below summarizes our hypotheses regarding the fac­ Guangdong (1,103), Zhejiang (758), Shanghai (649), and Beijing (580),
tors of social media marketing activeness as well as other control which are commonly known as the most e-commerce-active areas in
variables. China (Wang et al, 2019). On the other hand, the vast majority of West,
Northwest, and Northeast China have Taobao-Weibo users below 50.
5. Methodology Geographically speaking, our sample thus provides a relatively good
representation of Taobao sellers. Finally, although not shown in Fig. 1,
5.1. Data 250 users are identifying their location as “Overseas”. The non-trivial
number of Taobao sellers outside of China shows the rise of ‘purchas­
We initially identified Taobao sellers on Sina Weibo in November ing agents’ for Chinese domestic customers (Wang et al, 2019).

5
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Table 1
Descriptive Statistics.
(1) (2) (3) (4) (5) (6)

CODE VARIABLES # Obs mean sd p10 p50 p90

Weibo Variables
1 weibo marketing (d) 8,637 0.168 0.374 0 0 1
2 weibo use (d) 8,637 0.322 0.467 0 0 1
3 number of follows (o) 8,637 462.2 531.4 47 273 1,196
4 number of followers (e) 8,637 2,938 18,819 38 410 4,501
5 tao (e) 8,637 0.389 0.488 0 0 1
6 verify (e) 8,637 0.11 0.312 0 0 1
7 member (o) 8,637 0.327 1.274 0 0 0
8 female (c) 8,637 0.746 0.435 0 1 1
Taobao Variables 8,637
9 number of items (c) 8,637 114.4 374.3 2 26 273.3
10 median price (c) 8,637 240.9 545.3 18.29 92.59 474.7
11 taobao rank (o) 8,637 8.612 3.038 4 9 12
12 shop age (o) 8,637 5.825 2.299 2.949 5.568 9.13
13 credit as buyer (e) 8,637 619 743.6 40 388 1474
14 rating as seller (e) 8,637 99.11 1.147 97.93 99.44 100
15 description industry comparison (e) 8,637 35.9 38.12 − 0.423 29.51 100
16 description rating (e) 8,637 3.408 2.117 0 4.767 5
17 logistics industry comparison (e) 8,637 36.51 38.21 − 0.0267 30.98 100
18 logistics rating (e) 8,637 3.42 2.126 0 4.8 5
19 service industry comparison (e) 8,637 37.15 38.47 0 31.79 100
20 service rating (e) 8,637 3.426 2.128 0 4.8 5

“d” stands for dependent variables, “c” stands for control variables, “e” stands for earned reputational assets, and “o” stands for owned reputational assets.

Fig. 2 shows the distribution of Taobao-Weibo users by the main ● The final evaluation with the test set is based on standard perfor­
product category on Taobao. The top categories to some extent support mance indicators such as precision, recall, accuracy, and the area
the notion of a certain level of ‘feminism’ in China’s e-commerce in­ under the receiver operating characteristics curve (AUC) score
dustry (Yu and Cui, 2019), as the dominating category is “Clothes and (Lessmann et al, 2008).
bags” with 3,392 users, followed by “Mother and child” with 996 users
and “Beauty care” with 933 users. The binary classifiers considered include logistic regression, decision
tree (CART), random forests, and multinomial naive Bayes (MNB). After
running the classification and test procedures for the various algorithms,
5.2. Variables we find that the MNB classifier performs the best.
The MNB classifier is based on the probability of being in class c
A simple way to measure a Taobao-Weibo user’s activeness is to see if given a document d:
the user posted microblogs or not in the observation period. To do so, we ∏
define a dummy variable, weibo use, equaling 1 if at least one Weibo was P(c|d)∝P(c) P(tk |c)
posted in the observation period (i.e., from August to October in 2016) 1≤k≤nd

or 0 otherwise. However, the postings by individual Taobao-Weibo users P(tk |c) is the probability that feature tk appearing in a document
may not be marketing-oriented. For instance, a user may post on Weibo belonging to class c. Note that document d has nd features.
to share music or other aspects of his or her social life (Wang et al, 2019). The class that a given document d is assigned to is determined by the
As a result, Weibo activeness does not necessarily imply Weibo mar­ maximum a posteriori (MAP) class cmap :
keting activeness. We thus need to take a closer look at the content of ∏
Weibo posts to ensure that our key variable captures marketing efforts. ̂
cmap = argmax P(c|d) ̂
= argmax P(c) ̂ k |c)
P(t
To automate the process of classifying Weibo posts into marketing and c∈C c∈C 1≤k≤nd

non-marketing types, we employ a process in which we train a variety of

P(c)
̂ can be simply estimated as the number of documents belonging
machine learning algorithms and then select the best binary classifier in
to class c divided by the total number of documents in the training set,
terms of test performance.
whereas:
The sample used for comparing binary classifiers contains 5,000
microblogs randomly selected from 774,429 microblogs collected dur­ ̂ k |c) = Nctk + α
P(t
ing the periods of November in 2014 and from July to October in 2016 Nc + αn
(Wang et al, 2019). The 5,000 microblogs are manually labeled as either
Nctk is the fractional counting of TF-IDF feature tk appears in a sample
marketing or non-marketing type. The machine learning task is defined
of class c in the training set and Nc is the fractional counting of all TF-IDF
as predicting whether a given microblog is of marketing type (1 if true or
features for class c in the training set. n is the number of TF-IDF features
0 otherwise) based on its textual content. The procedure can be
for class c. Finally, α ≥ 0 is the smoothing prior that accounts for fea­
described as follows:
tures not present in the training set and prevents zero probabilities in
future computations (Manning et al., 2008).
● The microblogs (which are segmented into words, phrases, etc.) are
In our case, the MNB classifier’s AUC score and precision equal 0.96
tokenized, and stop words are removed.
and 0.91, respectively. We thus used it to classify all the 774,429
● The microblogs are mapped into feature vectors with the term
microblogs (Wang et al, 2019), which include the sample of microblogs
frequency-inverse document frequency (TF-IDF) method.
used in this paper. Once we have run the MNB classifier algorithm on our
● The sample of 5,000 microblogs is randomly divided into a training
sample, we find that out of 170,802 microblogs posted by the 8,637
set of 4,000 microblogs and a test set of 1,000 microblogs.
Taobao-Weibo users, 46,326 microblogs are classified as marketing
● A 10-fold cross-validation is applied to tune the hyperparameters of
type.
the classifiers.

6
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Using the above results, we can define a dummy variable, weibo

20
marketing, that equals 1 for a Taobao-Weibo user if the user posted at

1
least one Weibo classified as marketing type by the MNB classifier
during the observation period, i.e. from August to October in 2016, or

0.62
19

0 otherwise. Out of the 8,637 Taobao-Weibo users in our sample, 1,454

1
or 16.8% of the users posted at least one marketing Weibo during the

0.613
0.998
observation period. We further check the heterogeneity of marketing
18

activeness in terms of geographical distribution. Fig. A1 in the appendix

1
shows the number of actively marketing Taobao-Weibo users (i.e., weibo

0.616
0.926
0.608
marketing equalling 1) by region. The region with the most Taobao-
17

Weibo users in our sample, Guangdong (see Fig. 2), remains the most

1
marketing active one in Fig. A1, with 17.8% of its users (196 out of
1,103) being marketing-active. However, the second most marketing

0.611
0.998
0.616
0.998
16

active region becomes Shanghai (131 active users) whereas the third

1
and fourth places are taken by Zhejiang (109 active users) and Beijing
(107 active users) respectively.

0.904
0.598
0.925
0.597
0.61
15

Similarly, we check the heterogeneity of marketing activeness in

1

terms of product category. Fig. A2 in the appendix presents the number

of actively marketing Taobao-Weibo users by product category, with a
− 0.002

− 0.005

− 0.005
0.169

0.158

0.162
similar ranking to that in Fig. 3. Although only 18.6% of its users are
14

classified as actively marketing Taobao-Weibo users (630 out of 3,392),

the most populous product category “Clothes and bags” (see Fig. 3) still
0.124
0.077
0.074
0.079
0.074
0.072
0.073

dominates in Fig. A2. However, the second most marketing-active

13

category is “Beauty care” (with 222 users being marketing-active)

whereas the second most populous category “Mother and child” only
0.302
0.118
0.025
0.114
0.026
0.114
0.018
0.113

takes third place in Fig. A2 with 116 users being marketing-active.

12

When it comes to the potential driving factors behind social media

marketing activeness, we consider the reputational assets both earned
0.435
0.172
0.026

0.448
0.069
0.449
0.058
0.448

and owned by sellers on both C2C and social media platforms. First, there
0.05
11

are binary measures of earned reputation on both platforms. Sina Weibo,

for example, has both “Tao” and “Verify” identities that provide a signal
− 0.085

− 0.031

− 0.078

of trust as the verification of the identities is associated with some

− 0.08

− 0.08
0.078
0.041

0.023

0.025
0.02

thresholds of credibility and prestige (Wang et al, 2019). Second,

10

continuous earned reputation variables include the number of followers

on Sina Weibo and rating as seller, credit as buyer (as a C2C seller can
0.007
0.216

0.067
0.022
0.036

0.033
0.181
0.034
0.13

0.18

0.18

also be a buyer on Taobao and be rated by sellers), and specific rating

9

scores at both absolute and relative (i.e., industry comparison) levels

regarding product description (i.e., how realistic is the online product
− 0.004
− 0.045
− 0.015

− 0.042

− 0.041

− 0.042
0.067
0.139
0.097
0.032

0.043

0.033

description after a buyer receiving the product in hands?), logistics (i.e.,

how long is the delivery?) and customer services (i.e., how responsive is
8

the seller when a question is raised?). Table A1 in the appendix provides

0.059

0.022
0.146
0.045

0.085
0.064
0.095
0.063
0.095
0.064
0.095

detailed definitions of the variables.

0.05

0.13

On the other hand, the owned reputational assets include historical

7

sales measured by Taobao rank and shop age on Taobao and paid
− 0.072

membership and number of follows on Sina Weibo. We focus on sales

0.191

0.016
0.002
0.105
0.021
0.028

0.022
0.075
0.016
0.073
0.016
0.073
0.01

here as the other variables are self-explanatory. While it is practically

6

difficult to gather information about sales volume in a traditional sense

(e.g., measured by total transaction value or assets) in a virtual online
− 0.022

− 0.029

− 0.004
0.152
0.111

0.302
0.116
0.046

0.191
0.054
0.191
0.052

setting, Taobao does provide Taobao rank as a proxy for sales volume
0.08

0.05

0.19

(Gao et al, 2016). A similar approach was taken by previous work on

5

Amazon (Chevalier and Mayzlin, 2006). Taobao rank from the lowest to
the highest goes from 1 to 20. Each rank corresponds to an accumulated
− 0.006
0.032

0.285
0.028
0.009

0.112
0.018
0.058
0.026
0.006
0.048
0.011
0.048
0.011
0.048
0.21

score based on the customer rating after each transaction. After each
4

transaction, a customer can give a rating of “positive” (a + 1 score),

“neutral” (a 0 score), or “negative” (-1 score). Importantly, if no rating is
− 0.075

− 0.049
− 0.006

− 0.016
− 0.01

− 0.01

provided by the buyer, a rating of “positive” (a + 1 score) will be

0.054
0.097
0.099
0.122

0.008

0.025
0.002

0.025

0.024

0.025
0.02

assumed by default. Most customers do not leave any ratings after

3

transactions. As a result, most of the ratings are positive and the accu­
mulated score can thus be seen as serving as a proxy for the total number
− 0.011
0.112
0.124
0.106
0.127
0.271

0.031

0.014
0.136
0.085

0.069
0.083

0.089
0.12

0.01

0.08

0.07

0.07

of historical transactions. In other words, the sellers with a larger

2

number of sales transactions would typically have higher ranks.

Correlation Matrix.

Therefore, Taobao rank effectively serves as a measure of the historical

0.652
0.127
0.149
0.135
0.146
0.307

0.068
0.006
0.087
0.037
0.103
0.058
0.078
0.122
0.083
0.123
0.086
0.123
0.07

number of sales.
1

We also consider the binary variable, gender, as a potential deter­

Table 2

minant of C2C social media marketing activeness as well as control for

10
11
12
13
14
15
16
17
18
19
20
#

region, product category, median price, and the number of items

1
2
3
4
5
6
7
8
9

7
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Table 3
Logistic Regression on weibo marketing.
(1) (2) (3) (4) (5)
VARIABLES Weibo Taobao Weibo + Taobao Weibo + Taobao Weibo + Taobao

number of items (c) 0.000111 0.000258*** 0.000238*** 0.000124

(7.89e-05) (8.62e-05) (8.89e-05) (9.25e-05)
median price (c) 0.000104 3.95e-05 3.32e-05 5.54e-05
(6.66e-05) (5.73e-05) (5.93e-05) (7.19e-05)
taobao rank (o) 0.0205 − 0.0445*** − 0.0476*** − 0.0470***
(0.0162) (0.0142) (0.0143) (0.0178)
shop age (o) − 0.0370** − 0.00701 − 0.00720 − 0.0193
(0.0178) (0.0155) (0.0156) (0.0189)
credit as buyer (e) 0.000260*** 0.000179*** 0.000177*** 0.000183***
(4.25e-05) (3.92e-05) (3.99e-05) (4.59e-05)
rating as seller (e) 0.175*** 0.0974*** 0.120*** 0.117**
(0.0478) (0.0345) (0.0359) (0.0480)
industry comparison average (e) − 0.000697 − 0.000458 − 0.00101 − 0.00107
(0.00137) (0.00115) (0.00116) (0.00144)
rating average (e) 0.165*** 0.151*** 0.165*** 0.147***
(0.0302) (0.0246) (0.0253) (0.0318)
number of follows (o) 0.000420*** 0.000441*** 0.000469*** 0.000371***
(5.33e-05) (5.40e-05) (5.44e-05) (6.88e-05)
number of followers (e) 1.13e-05*** 1.16e-05*** 1.06e-05*** 1.33e-05***
(2.83e-06) (2.88e-06) (2.82e-06) (3.93e-06)
tao (e) 0.535*** 0.494*** 0.481*** 0.590***
(0.0623) (0.0660) (0.0665) (0.0813)
verify (e) 0.519*** 0.514*** 0.515*** 0.558***
(0.0891) (0.0898) (0.0904) (0.111)
member (o) 0.348*** 0.326*** 0.319*** 0.324***
(0.0199) (0.0204) (0.0205) (0.0248)
female (c) 0.504*** 0.478*** 0.420*** 0.342***
(0.0774) (0.0791) (0.0809) (0.0993)
Constant − 2.725*** − 20.21*** − 12.61*** − 15.23*** − 14.82***
(0.0842) (4.750) (3.413) (3.562) (4.767)
product category fixed effects (c) No Yes No Yes Yes
region fixed effects (c) No Yes No No Yes
Observations 8,637 5,522 8,637 8,637 5,522

Standard errors in parentheses; ***, **, and * denote significance at the 1%, 5% and 10% level, respectively.
“c” stands for control variables, “e” stands for earned reputational assets, and “o” stands for owned reputational assets.

displayed when doing empirical analyses. Again, the data of the Taobao- regressions.
Weibo users were collected for three months to establish reliable mea­
sures for variables such as median price and the number of items. As a 5.4. Model specification
result, except for static measures such as gender, all the measures are
monthly averaged over the observation period from August to October Since the outcome variable weibo marketing is binary, we apply a
in 2016. standard logistic regression model to understand the impact of the
variables collected from both platforms on Weibo activeness. Our main
5.3. Descriptive statistics empirical model is as follows:

Pr(weibomarketingi = 1|Weiboi ,Taobaoi )

Table 1 shows the descriptive statistics of the variables collected
from both platforms. Table A1 in the appendix shows the definitions of =
exp(α0 + β1 Weibo1i + β2 Weibo2i + ⋯ + γ 1 Taobao1i + γ2 Taobao2i + ⋯)
the variables. In Table 1, the variables 1–8 are from Sina Weibo whereas 1 + exp(α0 + β1 Weibo1i + β2 Weibo2i + ⋯ + γ 1 Taobao1i + γ2 Taobao2i + ⋯)
the ones 9–20 are from Taobao. Again, there are 8,637 Taobao-Weibo
users identified in our sample. For the observation period, 32% of where weibomarketingi = 1 if user i posted at least one marketing Weibo
them posted at least one microblog (as measured by the weibo use var­ during the observation period and Weiboi represents variables collected
iable) and 17% of them posted at least one marketing microblog ac­ from the platform of Weibo whereas Taobaoi represent variables
cording to our MNB classifier (as measured by the weibo marketing collected from the platform of Taobao.
variable). Moreover, 39% and 11% of them have obtained the “Tao” and
“Verify” identities respectively, 33% of them have paid memberships, 6. Empirical results
and 75% of them are female according to their profiles. For readers’
convenience, we add the variable types in the parentheses next to the 6.1. Baseline results
variable names, where “d” stands for dependent variables, “c” stands for
control variables, “e” stands for earned reputational assets, and “o” Table 3 shows the results of logistic regressions with different model
stands for owned reputational assets. Table 2 shows the pairwise cor­ specifications. The first and the second columns include the Weibo
relations among the variables. Note that the variables description industry variables and the Taobao variables respectively. The third to fifth col­
comparison, logistics industry comparison, and service industry comparison umns include the variables from both platforms by varying the fixed
are highly correlated with each other and the variables description rating, effects we control for.
logistics rating, and service rating are highly correlated with each other As can be seen in the table, the results across different specifications
(all with the correlation coefficients above 0.9). To avoid the collinearity are quite stable. We find that whether a registered Taobao seller is
problem, the two groups of three are averaged to the variables industry marketing-active in Weibo is associated with factors from both plat­
comparison average and rating average respectively when used in the forms. Looking first at earned reputation, our dependent variable is

8
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Table 4
Logistic Regression on weibo use.
(1) (2) (3) (4) (5)
VARIABLES Weibo Taobao Weibo + Taobao Weibo + Taobao Weibo + Taobao

number of items (c) 1.76e-05 0.000140* 0.000127 1.02e-05

(7.13e-05) (7.95e-05) (8.13e-05) (7.92e-05)
median price (c) − 2.48e-05 − 8.35e-05* − 8.62e-05* − 7.17e-05
(5.76e-05) (4.86e-05) (4.99e-05) (6.18e-05)
taobao rank (o) − 0.0297** − 0.0839*** − 0.0875*** − 0.0870***
(0.0132) (0.0113) (0.0114) (0.0144)
shop age (o) − 0.0257* − 0.0120 − 0.0124 − 0.0118
(0.0146) (0.0125) (0.0126) (0.0153)
credit as buyer (e) 0.000353*** 0.000284*** 0.000281*** 0.000275***
(3.93e-05) (3.47e-05) (3.53e-05) (4.09e-05)
rating as seller (e) 0.174*** 0.117*** 0.135*** 0.131***
(0.0360) (0.0254) (0.0263) (0.0363)
industry comparison average (e) 0.000682 0.00125 0.000791 0.000335
(0.00113) (0.000933) (0.000945) (0.00118)
rating average (e) 0.0886*** 0.0604*** 0.0751*** 0.0748***
(0.0233) (0.0187) (0.0192) (0.0244)
number of follows (o) 0.000334*** 0.000356*** 0.000369*** 0.000296***
(4.62e-05) (4.69e-05) (4.72e-05) (6.01e-05)
number of followers (e) 2.36e-05*** 2.53e-05*** 2.37e-05*** 2.55e-05***
(4.37e-06) (4.50e-06) (4.42e-06) (5.60e-06)
tao (e) 0.301*** 0.395*** 0.386*** 0.513***
(0.0502) (0.0536) (0.0539) (0.0663)
verify (e) 0.429*** 0.440*** 0.446*** 0.398***
(0.0799) (0.0811) (0.0814) (0.101)
member (o) 0.378*** 0.365*** 0.359*** 0.362***
(0.0243) (0.0246) (0.0246) (0.0297)
female (c) 0.677*** 0.598*** 0.546*** 0.526***
(0.0608) (0.0623) (0.0636) (0.0800)
Constant − 1.772*** − 18.66*** − 13.02*** − 15.06*** − 14.79***
(0.0645) (3.575) (2.516) (2.610) (3.605)
product category fixed effects (c) No Yes No Yes Yes
region fixed effects (c) No Yes No No Yes
Observations 8,637 5,538 8,637 8,637 5,538

Standard errors in parentheses; ***, **, and * denote significance at the 1%, 5% and 10% level, respectively.
“c” stands for control variables, “e” stands for earned reputational assets, and “o” stands for owned reputational assets.

Fig. A1. Geographic Distribution of Marketing-Active Taobao-Weibo Users.

9
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Fig. A2. Product Distribution of Marketing-Active Taobao-Weibo Users.

positively associated with the two verified identities tao and verify and use online information for purchasing decisions. Nadeem et al. (2015)
number of followers on the Weibo platform. This result supports hy­ find that the influence of peer recommendations is significantly stronger
pothesis H2, namely that earned reputation on social media is positively on the attitudes of females. Women constitute an important customer
associated with social media marketing activeness. If we instead look at segment in B2C and C2C e-commerce. This is evident when we look at
earned reputation on the Taobao platform, we find a similar pattern. Our how social media marketing is evolving. For example, in the quickly
dependent variable is positively related to both proxies for earned growing subdomain of influencer marketing, as many as 77% of influ­
reputation, credit as buyer and rating as seller. These results provide encers are women who most often reach out to and influence other
support for our hypothesis H1, i.e. that earned reputation on the C2C women (Gesenhues, 2019). This again suggests that women tend to
platform is positively associated with social media marketing activeness. place a high value on information from their peers when making pur­
We also find that owned reputational assets on Sina Weibo, as measured chasing decisions. The geographic focus of this study is China, a country
by number of follows and member, matter positively, which supports our where it is commonly known that women drive online sales and fuel
hypothesis H4. what has been labeled the ‘sheconomy’. Even though the country ex­
Next, we focus on the owned reputation variables on the Taobao side hibits a significant gender imbalance with about 31.6 million more
of the dataset. Our key explanatory variables here are taobao rank, which males, recent official figures show that women account for 55% of online
we argue serves well as a proxy for the number of sales on the Tabao spending (Liu, 2019). A typical example of this phenomenon is the social
platform, and shop age, which measures the number of years between the shopping platform Xiaohongshu (RED), where shoppers go to find
Taobao shop opening date and the data collection date. As we shopping information, tips, and reviews in the form of text, images, and
mentioned in Section 3 of the paper, there are two potential and oppo­ video. In 2019, over 80% of its users were female (Lim, 2020). Looking
site effects that sales may have on social media marketing activeness. at gender as a potential determinant of social media marketing active­
Focusing on the main regression model in Column 5, we see that the ness in our empirical analysis, we find that the coefficient for female C2C
coefficient for taobao rank is negative and significant. This lends support sellers is significant and positive in both Columns 1 and 3–5. This finding
to the argument that more sales on Taobao are associated with C2C ties into discussions on the issue of ‘feminism’ in China’s e-commerce
sellers feeling less need to engage in social media marketing activities. industry found in the literature and on the emergence of the so-called
The alternative measure, shop age, also behaves similarly when no fixed ‘sheconomy’.
effects of region or product category are controlled for in Column 2. But
its significance is crowded out by taobao rank once the fixed effects are
present. Looking at the other control variables from the Taobao part of 6.2. Robustness checks
the dataset, it appears that number of items and median price are by and
large insignificantly related to Weibo activeness. Note that the signifi­ We carry out a robustness check in which we instead use weibo use as
cance of number of items disappears after the region fixed effects are the dependent variable. Here, the dependent variable equals 1 if the Sina
included. Also recall that, to avoid the collinearity problem, rating Weibo user posted on his or her microblog (not necessarily marketing
average is the average across description rating, logistics rating, and service oriented) during the observation period. As a result, weibo use is a more
rating and its coefficient is consistently positive and significant. On the generous measure of social media marketing activeness than weibo
other hand, industry comparison average is the average across description marketing.
industry comparison, logistics industry comparison, and service industry The results of the new set of logistic regressions are presented in
comparison, and its coefficient is insignificant. Table 4. The main findings remain qualitatively unchanged. In partic­
While hitherto a relatively unexplored topic, previous research has ular, the proxies for earned reputation on the two platforms are still
shown that attitudes and behavior on social media differ across gender. highly significant and positively associated with weibo use. Similarly, the
In general, women tend to be motivated by being able to uphold close coefficient for gender, female, is still significant and positive. Finally, the
ties and obtain social information while men tend to be motivated by the proxy for sales, taobao rank, is once more negatively associated with the
opportunity to obtain more general information (Krasnova et al., 2017). dependent variable. These results lend further support to our initial
Differences between males and females can also be found in how they findings and our working hypotheses presented in Section 3.
Another robustness check we have tried is to recompute the variable

10
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Table A1 7. Conclusions
Variable Description.
VARIABLES DESCRIPTION Social media marketing has become a standard tool for businesses of
all sizes. In this study, we contribute to the literature on social media
Weibo Variables
1 weibo marketing (d) A dummy variable that equals 1 if at least one marketing by examining a unique dataset of C2C sellers’ social media
marketing Weibo during the 3-month observation marketing activeness on China’s leading microblogging platform. Based
period. on a sample of over 8,000 C2C sellers registered on both Taobao, China’s
2 weibo use (d) A dummy variable that equals 1 if at least one Weibo largest C2C online shopping platform, and Sina Weibo, China’s largest
during the 3-month observation period.
3 number of follows (o) The number of follows in Sina Weibo (3-month
microblogging platform, this paper empirically explores the factors
average). influencing C2C sellers’ social media marketing use. We find that
4 number of followers (e) The number of followers in Sina Weibo (3-month whether a registered Taobao seller is actively marketing in Weibo is
average). associated with factors on both platforms. In particular, it relates to
5 tao (e) A dummy variable that equals 1 if the user is verified as
reputational assets accumulated on both the C2C e-commerce platform
a Taobao seller in Sina Weibo.
6 verify (e) A dummy variable being 1 if the user is identity- as well as the microblogging platform. Interestingly, we also find that
verified. female users are more likely to be active, which further confirms the
7 member (o) A dummy variable being 1 if the user has paid for a ‘feminism’ in China’s e-commerce industry and the advancement of the
Sina Weibo VIP membership. so-called ‘sheconomy’. Finally, owned reputational assets such as Tao­
8 female (c) A dummy variable being 1 if the user claims as female
in Sina Weibo.
bao rank, a proxy for its turnover, are negatively associated with social
Taobao Variables media marketing activeness. One potential reason for this is that owners
9 number of items (c) The number of items listed in the Taobao shop (3- of highly ranked Taobao shops feel that they are less dependent on social
month average). media marketing efforts.
10 median price (c) The median price of all the listed items (3-month
We contribute to the literature in multiple ways. First of all, we
average).
11 taobao rank (o) Taobao rank (1–20), the higher the value the better identify several new factors that potentially influence social media
rating points accumulated. It is effectively a measure of marketing activeness, including reputational assets classified into
sales. earned and owned reputational assets on the C2C as well as the micro­
12 shop age (o) The number of years between the Taobao shop opening blogging platform and gender. Our finding that earned reputation on
date and the data collection date. The days are
both platforms is positively associated with social media marketing
converted to decimal years (3-month average).
13 credit as buyer (e) The credit level as a buyer, the higher the better ratings activeness is particularly interesting, as the relationship is often dis­
received from sellers (3-month average). cussed in terms of how social media marketing affects reputation, not
14 rating as seller (e) The percentage of good ratings out of all ratings the other way around. Perhaps more importantly, we also shed light on
received from buyers (3-month average).
the limitations of collaborative efforts between C2C e-commerce and
15 description industry The relative difference between the Taobao shop’s
comparison (e) description rating (defined below) and the industry social media platforms as our empirical analysis provide evidence sug­
average (3-month average). Its range goes from − 100 gesting that there is an inherent conflict between reputation on the
to 100. different platforms, something that has a direct effect on C2C sellers’
16 description rating (e) The average rating (1–5) received in terms of willingness to engage in marketing activities on social media platforms.
description matching product delivered, the higher the
This is important as the collaboration between a C2C e-commerce
better (3-month average).
17 logistics industry The relative difference between the Taobao shop’s platform and a social media platform is a potential alternative to ven­
comparison (e) logistics rating (defined below) and the industry dors such as Xiaohongshu and Pinduoduo that provide an integrated
average (3-month average). Its range goes from − 100 social commerce experience on the same platform. Furthermore, we
to 100.
shed new light on the important and quickly growing C2C market in
18 logistics rating (e) The average rating (1–5) received in terms of delivery
speed, the higher the better (3-month average).
China. Finally, our empirical approach allows us to develop a more
19 service industry The relative difference between the Taobao shop’s suitable measure for social media marketing than that found in most
comparison (e) service rating (defined below) and the industry related studies. In most of the existing literature that has analyzed fac­
average (3-month average). Its range goes from − 100 tors that may affect the adoption of social media marketing and its po­
to 100.
tential effects, social media marketing is often measured by whether a
20 service rating (e) The average rating (1–5) received in terms of customer
service, the higher the better (3-month average). business has a social media account or not. However, given the fact that
industry comparison The average of the above description industry a social media account can be easily created, it is more relevant to know
average (e) comparison, logistics industry comparison, and service to what extent the social media account is used for marketing purposes.
industry comparison.
Through the adoption of machine learning and natural language pro­
rating average (e) The average of the above description rating, logistics
rating, and service rating.
cessing techniques, we are able to develop a measure for social media
activeness instead of merely social media adoption, thereby shedding
“d” stands for dependent variables, “c” stands for control variables, “e” stands for new light on the driving forces behind social media marketing by C2C
earned reputational assets, and “o” stands for owned reputational assets.
sellers.
To the best of our knowledge, this is the first study that examines the
weibo marketing by randomly reclassifying some microblogs. To do so, interplay between two leading C2C e-commerce and social media plat­
we randomly select 10% of the microblogs and reclassify them into forms as they experiment with a new collaborative effort aimed at
marketing type if they are non-marketing type according to our MNB helping C2C sellers marketing their products on social media. As
classifier and vice versa. Even though our MNB classifier has achieved mentioned throughout the text, the quickly growing social commerce
impressive performance scores (AUC score and precision equaling 0.96 space in China shows the importance of understanding how e-commerce
and 0.91 respectively), there might still be some room for improvement interacts with social media. A successful integration of C2C e-commerce
with other machine learning models, such as neural-network-based and social media platforms could theoretically provide a powerful
ones. Therefore, this additional check will test how sensitive our re­ alternative to integrated social commerce platforms such as Xiaohong­
sults react to the accuracy of our MNB classifier. The results are pre­ shu and Pinduoduo. Our analysis suggests that the conflict between
sented in Table A2 in the appendix. Again, our findings are robust as the different reputational assets on the two platforms constitutes one
results remain qualitatively unchanged. important factor for the limited success of the collaboration that was
initiated in 2013. Finally, the empirical relationships examined in the

11
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Table A2
Logistic Regression on randomly reclassified weibo marketing.
(1) (2) (3) (4) (5)
VARIABLES Weibo Taobao Weibo + Taobao Weibo + Taobao Weibo + Taobao

number of items (c) 7.82e-05 0.000197** 0.000185** 7.86e-05

(7.35e-05) (8.40e-05) (8.66e-05) (8.09e-05)
median price (c) 9.49e-05 4.59e-06 5.46e-06 6.14e-05
(6.43e-05) (5.52e-05) (5.68e-05) (6.77e-05)
taobao rank (o) 0.0107 − 0.0439*** − 0.0481*** − 0.0409**
(0.0155) (0.0133) (0.0134) (0.0167)
shop age (o) − 0.0346** − 0.00320 − 0.00276 − 0.0226
(0.0170) (0.0146) (0.0147) (0.0177)
credit as buyer (e) 0.000274*** 0.000224*** 0.000220*** 0.000205***
(4.15e-05) (3.69e-05) (3.75e-05) (4.35e-05)
rating as seller (e) 0.202*** 0.0925*** 0.110*** 0.151***
(0.0465) (0.0320) (0.0331) (0.0465)
industry comparison average (e) 0.000653 0.000895 0.000366 0.000421
(0.00132) (0.00109) (0.00111) (0.00137)
rating average (e) 0.124*** 0.104*** 0.117*** 0.107***
(0.0285) (0.0229) (0.0235) (0.0295)
number of follows (o) 0.000304*** 0.000325*** 0.000344*** 0.000309***
(5.16e-05) (5.23e-05) (5.26e-05) (6.55e-05)
number of followers (e) 7.70e-06*** 7.87e-06*** 7.26e-06*** 7.48e-06**
(2.34e-06) (2.39e-06) (2.30e-06) (3.00e-06)
tao (e) 0.415*** 0.397*** 0.384*** 0.495***
(0.0590) (0.0626) (0.0630) (0.0764)
verify (e) 0.392*** 0.390*** 0.391*** 0.447***
(0.0873) (0.0881) (0.0884) (0.107)
member (o) 0.308*** 0.285*** 0.277*** 0.260***
(0.0193) (0.0198) (0.0199) (0.0238)
female (c) 0.587*** 0.534*** 0.487*** 0.435***
(0.0742) (0.0757) (0.0773) (0.0948)
Constant − 2.479*** –22.41*** − 11.79*** − 13.66*** − 17.77***
(0.0794) (4.627) (3.168) (3.286) (4.617)
product category fixed effects (c) No Yes No Yes Yes
region fixed effects (c) No Yes No No Yes
Observations 8,637 5,531 8,637 8,637 5,531

Standard errors in parentheses; ***, **, and * denote significance at the 1%, 5% and 10% level, respectively.
“c” stands for control variables, “e” stands for earned reputational assets, and “o” stands for owned reputational assets.

paper are at most correlational, rather than causal. It is impossible to References

deny the possibilities that unmeasurable internal factors such as Taobao-
Weibo users’ intrinsic willingness to engage customers are behind the Abed, S.S., 2020. Social commerce adoption using TOE framework: An empirical
investigation of Saudi Arabian SMEs. Int. J. Inf. Manage. 53, 102118.
empirical relationships found in this study and that external factors such Ali Abbasi, G., Abdul Rahim, N.F., Wu, H., Iranmanesh, M. and Keong, B.N.C., 2022.
as direct competition from WeChat, Xiaohongshu, Pinduoduo, and other Determinants of SME’s social media marketing adoption: competitive industry as a
platforms represent plausible reasons for the limited success of Taobao moderator. Sage Open, 12(1), p.21582440211067220.
Appel, G., Grewal, L., Hadi, R., Stephen, A.T., 2020. The future of social media
and Sina Weibo’s social commerce collaboration. We therefore leave the marketing. J. Acad. Market. Sci. 48, 79–95.
analysis of causality to future studies once more data and research de­ Ashley, C., Tuten, T., 2015. Creative strategies in social media marketing: An exploratory
signs become available. study of branded social content and consumer engagement. Psychol. Mark. 32 (1),
15–27.
Cai, Y.-J., Wang, Y., Zhang, J., 2018. Enhancing e-platform business by customer service
CRediT authorship contribution statement systems: a multi-methodological case study on Ali Wangwang instant message’s
impacts on TaoBao. Ann. Oper. Res. 291 (1-2), 59–81.
Chang, Y.T., Yu, H., Lu, H.P., 2015. Persuasive messages, popularity cohesion, and
Anders C. Johansson: Conceptualization, Methodology, Writing –
message diffusion in social media marketing. J. Bus. Res. 68 (4), 777–782.
review & editing. Zhen Zhu: Conceptualization, Methodology, Soft­ Chatterjee, S., Kumar Kar, A., 2020. Why do small and medium enterprises use social
ware, Data curation, Writing – review & editing. media marketing and what is the impact: Empirical insights from India. Int. J. Inf.
Manage. 53, 102103.
Chen, W., 2020. Studious shoppers can now find online classes on Alibaba’s Taobao.
Declaration of Competing Interest KrAsia https://kr-asia.com/studious-shoppers-can-now-find-online-classes-on-
alibabas-taobao.
The authors declare that they have no known competing financial Cheung, C.M., Chiu, P.Y., Lee, M.K., 2011. Online social networks: Why do students use
facebook? Comput. Hum. Behav. 27 (4), 1337–1343.
interests or personal relationships that could have appeared to influence Chevalier, J.A., Mayzlin, D., 2006. The effect of word of mouth on sales: Online book
the work reported in this paper. reviews. J. Mark. Res. 43 (3), 345–354.
Chiu, C.M., Hsu, M.H., Lai, H., Chang, C.M., 2012. Re-examining the influence of trust on
online repeat purchase intention: The moderating role of habit and its antecedents.
Data availability Decis. Support Syst. 53 (4), 835–845.
Dahnil, M.I., Marzuki, K.M., Langgat, J., Fabeil, N.F., 2014. Factors influencing SMEs’
Data will be made available on request. adoption of social media marketing. Procedia Soc. Behav. Sci. 148, 119–126.
de Oliveira, M.J., Huertas, M.K.Z., Lin, Z., 2016. Factors driving young users’
engagement with Facebook: Evidence from Brazil. Comput. Hum. Behav. 54, 54–61.
Appendix Dellarocas, C., 2003. The digitization of word of mouth: Promise and challenges of online
feedback mechanisms. Manag. Sci. 49 (10), 1407–1424.
Du, R.Y., Hu, Y., Damangir, S., 2015. Leveraging trends in online searches for product
Figs. A.1 and A.2, Tables A.1 and A.2.
features in market response modeling. J. Mark. 79 (1), 29–43.

12
A.C. Johansson and Z. Zhu Electronic Commerce Research and Applications 61 (2023) 101305

Dwivedi, Y.K., Ismagilova, E., Rana, N.P., Raman, R., 2023. Social media adoption, usage López-López, D., Giusti, G., 2020. Comparing digital strategies and social media usage in
and impact in business-to-business (B2B) context: A state-of-the-art literature review. B2B and B2C industries in Spain. J. Bus. Bus. Mark. 27 (2), 175–186.
Inf. Syst. Front. 25 (3), 971–993. Manning, C.D., Raghavan, P., Schütze, H. (Eds.), 2008. Introduction to Information
Feng, X., Johansson, A.C., 2019. Top executives on social media and information in the Retrieval. Cambridge University Press.
capital market: Evidence from China. Finance 58, 824–857. Melnik, M.I., Alm, J., 2002. Does a seller’s ecommerce reputation matter? Evidence from
Gao, Q., Abel, F., Houben, G.J., Yu, Y., 2012. A comparative study of users’ eBay auctions. J. Ind. Econ. 50 (3), 337–349.
microblogging behavior on Sina Weibo and Twitter. In: Masthoff, J., Mobasher, B., Michaelidou, N., Siamagka, N.T., Christodoulides, G., 2011. Usage, barriers, and
Desmarais, M., Nkambou, R. (Eds.), User Modeling, Adaptation, and Personalization: measurement of social media marketing: An exploratory investigation of small and
20th International Conference, UMAP 2012, Montreal, Canada, July 16–20, 2012 medium B2B brands. Ind. Mark. Manage. 40 (7), 1153–1159.
Proceedings. Springer, Berlin Heidelberg, Heidelberg, pp. 88–101. Moore, J.N., Hopkins, C.D., Raymond, M.A., 2013. Utilization of relationship-oriented
Gao, B., Chan, W.K.V., Chi, L., Deng, X.N., 2016. Size and growth dynamics of online social media in the selling process: A comparison of consumer (B2C) and industrial
stores: A case of China’s Taobao. Com. Electronic Commerce Res. Appl. 17, 161–172. (B2B) salespeople. J. Internet Commer. 12 (1), 48–75.
Gesenhues, 2019. Women make up majority of influencer community, still earn less than Moore, J.N., Raymond, M.A., Hopkins, C.D., 2015. Social selling: A comparison of social
male influencers. Marketing Land, June 6. media usage across process stage, markets, and sales job functions. J. Mark. Theory
Godey, B., Manthiou, A., Pederzoli, D., Rokka, J., Aiello, G., Donvito, R., Singh, R., 2016. Pract. 23 (1), 1–20.
Social media marketing efforts of luxury brands: Influence on brand equity and Pascucci, F., Ancillai, C., Cardinali, S., 2018. Exploring antecedents of social media usage
consumer behavior. J. Bus. Res. 69 (12), 5833–5841. in B2B: a systematic review. Manage. Res. Rev. 41 (6), 629–656.
Gong, S., Zhang, J., Zhao, P., Jiang, X., 2017. Tweeting as a marketing tool: A field Pavlou, P.A., Dimoka, A., 2006. The nature and role of feedback text comments in online
experiment in the TV industry. J. Mark. Res. 54 (6), 833–850. marketplaces: Implications for trust building, price premiums, and seller
Guesalaga, R., 2016. The use of social media in sales: Individual and organizational differentiation. Inf. Syst. Res. 17 (4), 392–414.
antecedents, and the role of customer engagement in social media. Ind. Mark. Pavlou, P.A., Liang, H., Xue, Y., 2007. Understanding and mitigating uncertainty in
Manage. 54, 71–79. online exchange relationships: A principal-agent perspective. MIS Q. 105–136.
Heavey, C., Simsek, Z., Kyprianou, C., Risius, M., 2020. How do strategic leaders engage Puriwat, W., Tripopsakul, S., 2021. Explaining social media adoption for a business
with social media? A theoretical framework for research and practice. Strateg. purpose: an application of the UTAUT model. Sustainability 13 (4), 2082.
Manag. J. 41 (8), 1490–1527. Qin, B., Strömberg, D., Wu, Y., 2017. Why does China allow freer social media? Protests
Herjanto, H., Amin, M., 2020. Repurchase intention: the effect of similarity and client versus surveillance and propaganda. J. Econ. Perspect. 31 (1), 117–140.
knowledge. Int. J. Bank Mark. 38 (6), 1351–1371. Shi, S., Chen, Y., Chow, W.S., 2016. Key values driving continued interaction on brand
Iankova, S., Davies, I., Archer-Brown, C., Marder, B., Yau, A., 2019. A comparison of pages in social media: An examination across genders. Comput. Hum. Behav. 62,
social media marketing between B2B, B2C and mixed business models. Ind. Mark. 578–589.
Manage. 81, 169–179. Siamagka, N.T., Christodoulides, G., Michaelidou, N., Valvi, A., 2015. Determinants of
Jia, L., Han, X., 2020. Tracing Weibo (2009–2019): The commercial dissolution of public social media adoption by B2B organizations. Ind. Mark. Manage. 51, 89–99.
communication and changing politics. Internet Histories 4 (3), 304–332. Sinclaire, J.K., Vogus, C.E., 2011. Adoption of social networking sites: an exploratory
Jones, K., Leonard, L.N., 2008. Trust in consumer-to-consumer electronic commerce. Inf. adaptive structuration perspective for global organizations. Inf. Technol. Manage. 12
Manage. 45 (2), 88–95. (4), 293–314.
Kim, A.J., Ko, E., 2012. Do social media marketing activities enhance customer equity? Steinhoff, L., Arli, D., Weaven, S., Kozlenkova, I.V., 2019. Online relationship marketing.
An empirical study of luxury fashion brand. J. Bus. Res. 65 (10), 1480–1486. Journal of Academy of Marketing Science 47 (3), 369–393.
Kim, Y., Krishnan, R., 2015. On product-level uncertainty and online purchase behavior: Sullivan, J., 2012. A tale of two microblogs in China. Media Cult. Soc. 34 (6), 773–783.
An empirical analysis. Manag. Sci. 61 (10), 2449–2467. Ter Huurne, M., Ronteltap, A., Corten, R., Buskens, V., 2017. Antecedents of trust in the
Kumar, A., Bezawada, R., Rishika, R., Janakiraman, R., Kannan, P.K., 2016. From social sharing economy: A systematic review. J. Consum. Behav. 16 (6), 485–498.
to sale: The effects of firm-generated content in social media on customer behavior. Thierer, A., Koopman, C., Hobson, A., Kuiper, C., 2015. How the internet, the sharing
J. Mark. 80 (1), 7–25. economy, and reputational feedback mechanisms solve the lemons problem.
Lamberton, C., Stephen, A.T., 2016. A thematic exploration of digital, social media, and U. Miami L. Rev. 70, 830.
mobile marketing: Research evolution from 2000 to 2015 and an agenda for future Toubia, O., Stephen, A.T., 2013. Intrinsic vs. image-related utility in social media: Why
inquiry. J. Mark. 80 (6), 146–172. do people contribute content to Twitter? Mark. Sci. 32 (3), 368–392.
Leonard, L.N., Jones, K., 2021. Trust in C2C electronic commerce: Ten years later. Vock, M., Dolen, W.V., Ruyter, K.D., 2013. Understanding willingness to pay for social
J. Comput. Inf. Syst. 61 (3), 240–246. network sites. J. Serv. Res. 16 (3), 311–325.
Lessmann, S., Baesens, B., Mues, C., Pietsch, S., 2008. Benchmarking classification Wang, X., Baesens, B., Zhu, Z., 2019. On the optimal marketing aggressiveness level of
models for software defect prediction: A proposed framework and novel findings. C2C sellers in social media: Evidence from China. Omega 85, 83–93.
IEEE Trans. Softw. Eng. 34 (4), 485–496. Wang, Z., Li, H., Ye, Q., Law, R., 2016. Saliency effects of online reviews embedded in the
Li, D., Li, J., Lin, Z., 2008. Online consumer-to-consumer market in China–a comparative description on sales: Moderating role of reputation. Decis. Support Syst. 87, 50–58.
study of Taobao and eBay. Electron. Commer. Res. Appl. 7 (1), 55–67. We Are Social, 2019. Digital in 2019: China.
Li, G., Liu, H., Li, G., 2014. Payment willingness for VIP subscription in social networking Yang, J., Hu, X., Zhang, H., 2007. Effects of a reputation feedback system on an online
sites. J. Bus. Res. 67 (10), 2179–2184. consumer-to-consumer auction market. Decis. Support Syst. 44 (1), 93–105.
Lim, S., 2020. Emerging platforms: What is Xiaohongshu and how can brands leverage it? Ye, Q., Li, Y., Kiang, M., Wu, W., 2009. The Impact of Seller Reputation on the
The Drum, 16 March. Performance of online sales: evidence from TaoBao buy-it-now (BIN) data. ACM
Liu, S., Perry, P., Gadzinski, G., 2019a. The implications of digital marketing on WeChat SIGMIS Database: the DATABASE for Advances in Information Systems 40 (1),
for luxury fashion brands in China. J. Brand Manage. 26 (4), 395–409. 12–19.
Liu, X., Shin, H., Burns, A.C., 2019b. Examining the impact of luxury brand’s social Yoon, H.S., Occena, L.G., 2015. Influencing factors of trust in consumer-to-consumer
media marketing on customer engagement: Using big data analytics and natural electronic commerce with gender and age. Int. J. Inf. Manage. 35 (3), 352–363.
language processing. J. Bus. Res. Yu, H., Cui, L., 2019. China’s E-Commerce: Empowering Rural Women? China Q. 238,
Liu, D., 2019. China’s $670 billion ‘sheconomy’ is growing like crazy. NBC News online, 1–20.
3 May. Zolkepli, I.A., Kamarulzaman, Y., 2015. Social media adoption: The role of media needs
and innovation characteristics. Comput. Hum. Behav. 43, 189–209.

13