Social Network Analysis and Mining (2024) 14:131

https://doi.org/10.1007/s13278-024-01289-8

ORIGINAL ARTICLE

The “Non-Musk” effect at X (Twitter)

Dmitry Zinoviev1 · Arkapravo Sarkar2 · Pelin Bicen2

Received: 9 January 2024 / Revised: 8 June 2024 / Accepted: 23 June 2024

© The Author(s), under exclusive licence to Springer-Verlag GmbH Austria, part of Springer Nature 2024

Abstract
Elon Musk, a notable entrepreneur, often influences Wall Street with his controversial social media presence. Drawing on
Social Identity and Social Influence Theories, this study explores the “Musk Effect,” specifically how Musk’s acquisition
of Twitter impacts other entrepreneurs’ social media activities. Analyzing 4 million tweets from 36,828 entrepreneurs
across seven English-speaking countries (US, Australia, New Zealand, UK, Canada, South Africa, and Ireland) over 56
weeks, the study captures the period from the rumors of Musk’s Twitter purchase to its completion. Findings reveal that
only about 2% of these entrepreneurs significantly altered their tweeting behavior, suggesting that the direct influence of
Musk’s actions on entrepreneurs’ social media activity may be limited, despite his high-profile status. This research is
among the first to investigate how Musk’s Twitter takeover affects entrepreneurs’ actual Twitter usage, offering insights
into the Musk Effect beyond its Wall Street implications. By identifying specific entrepreneur characteristics most influ-
enced by the Musk Effect, this study underscores the nuanced nature of social influence in the digital age. It reveals that
while high-profile figures like Musk can generate widespread interest, their direct impact on the digital behavior of peers
may be limited. Understanding this dynamic is essential for accurately assessing the influence of prominent entrepreneurs
on industry trends and behaviors in a rapidly evolving digital marketplace. The findings suggest that applying Social
Identity and Social Influence Theories in digital contexts requires a more nuanced approach, considering the unique char-
acteristics of online environments and the multiple factors shaping social identities and behaviors.

Keywords Leadership · Elon Musk · X (Twitter) · Complex network analysis · Topic modeling · Entrepreneurship

1 Introduction and theoretical background —Steve Johnson, Chief Investment Officer, Forager Funds
Management, Sydney in the context of Elon Musk’s $44 bil-
lion Twitter acquisition deal in 2022
It’s been a heck of a rollercoaster, that’s for sure, and Elon Musk, the Chief Executive Officer of Tesla and
we’re certainly happy from a financial perspective SpaceX, is a prominent figure recognized for his unconven-
that (the deal) has closed. tional personality and controversial claims (Flake 2022). His
actions and statements have frequently incited substantial
fluctuations in the stock market (Shead 2021), the crypto-
currency market (Ante 2023), and notably within the tech-
nology sector (Luri et al. 2023). Previous studies, such as
Pelin Bicen Niburski and Niburski (2023), have highlighted the power
[email protected]
of influential figures like Elon Musk in shaping public opin-
Dmitry Zinoviev ion and market trends through social media. However, these
[email protected]
studies primarily focused on the general public.
Arkapravo Sarkar Despite the extensive research on Musk’s influence on
[email protected]
the general public, the impact of his conduct and actions
1
Department of Computer Science, College of Arts and on social media, specifically on the behavior of entrepre-
Sciences, Suffolk University, 73 Tremont, Boston, neurs, remains largely unexplored in academic literature.
MA 02108, USA Our study specifically focuses on this gap by investigat-
2
Sawyer Business School, Marketing Suffolk University, 73 ing the so-called “Musk Effect” on the Twitter activity of
Tremont, Boston, MA 02108, USA

13
131 Page 2 of 17 Social Network Analysis and Mining (2024) 14:131

individuals who identify themselves as entrepreneurs, par- sentiments, might align their Twitter behavior in response
ticularly in the wake of Musk’s acquisition of Twitter. to Musk’s activities, showcasing the social influence at play.
Given Elon Musk’s prominent influence on public opin- Our belief is consistent with the Social Influence Theory
ion through his social media activities, this study aims (Kelman 1953; Li 2013) which is a psychological theory
to address the specific impact of his actions on the entre- that explores how an individual’s thoughts, feelings, and
preneurial community. We seek to answer the following behaviors are influenced by the presence and actions of oth-
research questions: How does Elon Musk’s acquisition of ers which includes influence, particularly from authority fig-
Twitter influence the social media activity and behaviors of ures, peers, and celebrities. It examines the ways in which
self-identified entrepreneurs on the platform? Would politi- people’s attitudes and behaviors can be shaped by the real,
cal affiliation of self-identified entrepreneurs moderate the imagined, or implied presence of others in a social context.
influence of Elon Musk’s acquisition of Twitter on their One of the key aspects of Social Influence Theory includes
social media activity and behaviors? the process of changing someone’s attitudes, beliefs, or
These questions are crucial as they explore the extent to behaviors through the use of communication, which Musk
which Musk’s actions and statements, particularly his Twit- often does on Twitter. In one study (Rudeloff and Damms
ter activities, affect the tweeting behavior of entrepreneurs. 2022), authors explored how the personal communication
By focusing on entrepreneurs rather than the general activities of influential entrepreneurs, such as Elon Musk,
public, our study aims to fill a significant gap in the litera- on social media lead to the emergence of parasocial interac-
ture regarding the “Musk Effect” on social media utiliza- tions (PSIs) which refer to the one-sided relationships that
tion among this specific group. Understanding this dynamic individuals develop with media personalities, celebrities, or
can provide valuable insights into how high-profile figures influencers. These interactions occur when audiences feel
influence not only market trends but also the behaviors and a sense of intimacy and connection with a media figure
strategies of other influential individuals within the entre- despite the lack of reciprocation.
preneurial ecosystem. High-profile individuals can significantly sway follower
Contemporary capitalist societies trust heroic individuals opinions and behavior through identification, where individ-
to address major societal issues (Bardon et al. 2021). Heroes uals look to others they can identify with, to determine their
are known for sacrifice, risk-taking, or superior ethics. The own actions. As Luri et al. (2023) synthesize work from var-
“delivering hero” archetype, who saves those in distress, is ious fields to map out how entrepreneurs, like Musk, engage
prevalent in media (Murray 2017). Entrepreneurs leverage consumers, media, and the state in their visions of social
this by promoting their products as saviors (Sørensen 2008). and market progress and critically examines the impact of
For instance, promises to combat global warming or tran- heroic discourses of technological utopias and the ideologi-
scend human limits through space ventures position entre- cal support for such discourses. As such, Elon Musk could
preneurs like Bezos and Musk as heroes if their visions are be considered a reference group by himself for entrepre-
embraced. neurs on Twitter. His actions, particularly his acquisition
Elon Musk as an entrepreneur is often cited as he propa- of Twitter and subsequent behaviors on the platform serve
gates narratives of a utopian future, and often called by the as a benchmark for other entrepreneurs. They may adjust
public and media as an entrepreneur, who has wrapped him- their tweeting behavior, content, and engagement strategies
self in big ideas about the future, is viewed as a market hero in alignment or contrast to Musk’s, influenced by his per-
(Bardon et al. 2021). His public image has been defined by ceived success and authority in the technology and business
his unwavering focus on space exploration, climate solu- sectors. Notably, Musk has gone on record to say that his
tions, human-computer integration, and avoiding AI extinc- acquisition of Twitter was motivated by his inherent desire
tion. Musk’s ventures, such as SpaceX, Tesla, SolarCity, to protect freedom of speech on Twitter. As detailed by Isaa-
Neuralink, and The Boring Company, are cited as examples cson (2023) in Elon Musk’s biography, Musk’s pursuit of
of businesses that are seen as philanthropic efforts to “save Twitter was underpinned by his commitment to uphold prin-
the world,” illustrating the convergence of market profits ciples of free speech, viewing the platform as a battleground
and social good (Shamir 2008). for First Amendment rights in the digital age. This ambition
When it comes to social media usage and Twitter activ- reflects a broader vision beyond mere corporate acquisition,
ity, we expect that Elon Musk, due to his heroic status and suggesting Musk’s engagement with Twitter was more ideo-
significant following, acts as a pivotal figure whose regular logically driven. A recent study by Broockman et al. (2019)
interactions and frequent statements on Twitter might influ- has shown that a significant majority of entrepreneurs sub-
ence the behaviors of other users, particularly aspiring entre- scribe to market oriented neoliberal ideologies and in turn
preneurs. In our rationale, we believe that entrepreneurs, may also subscribe to Musk’s ideals of safeguarding public
seeking to align with successful figures or gauge market discourse liberties on Twitter. As such, we anticipate that

13
Social Network Analysis and Mining (2024) 14:131 Page 3 of 17 131

entrepreneurs’ responses to Musk’s acquisition of Twitter liberal-conservative scale provides a means of assessing
will be influenced by their political and ideological group entrepreneurs’ ideological alignment with Musk’s stance,
memberships, which leads us to our second research ques- strengthening the application of the Social Identity Theory
tion: Would the political affiliation of self-identified entre- to our research context.
preneurs moderate the impact of Elon Musk’s Twitter In the period surrounding his acquisition of Twitter,
acquisition on their social media activity and behaviors? Musk increased his own tweeting activity significantly.
Social Identity Theory, developed by Tajfel and Turner Prior to the acquisition, Musk engaged with the platform
(1986), is a social psychological theory that explains how just over 15 times a day, on average. Following his purchase
people’s sense of who they are and how they behave is of Twitter, he was seen to be tweeting 23.9 times a day1, or
based on their group membership. The theory proposes that almost once every hour, on average. This increase in tweet-
people classify themselves and others into various social ing activity is expected to influence other entrepreneurs as
categories, such as gender, age, religion, politics, sports, or well, prompting them to intensify their interactions on Twit-
educational affiliation. Social Identity Theory has been used ter - a phenomenon we term the “Musk Effect.”
to understand various social phenomena, such as intergroup In this context, the timeline of events pertaining to Elon
relations, stereotyping, and collective action. It highlights Musk’s recent and highly publicized acquisition of Twitter
the importance of group membership in shaping an indi- warrants attention (Wile 2022). On November 29, 2021,
vidual’s self-concept and behavior within social contexts Jack Dorsey resigned as the Chief Executive Officer of Twit-
group (Tajfel and Turner 1986). Its key findings, perhaps ter (Conger and Hirsch 2021). Subsequently, on January 31,
the most famous of which is the conformity to group norms 2022, Musk initiated the purchase of Twitter stock (Paris
that generates collective action have been widely replicated et al. 2022). On April 14th, 2022, Musk extended an unso-
(Brown 1995). Huddy (2001) suggests that individuals shar- licited and nonbinding proposition to acquire the company,
ing a strong ideological identity are likely to respond in a instigating considerable speculation and debate across social
similar manner to specific situations. media platforms (Paris et al. 2022). However, on July 8th,
In the context of Elon Musk’s acquisition of Twitter, 2022, Musk declared his intention to terminate the $44 bil-
entrepreneurs who subscribe to market-oriented neoliberal lion proposed deal, inciting one more round of turbulence in
ideologies and share Musk’s views on safeguarding public the financial markets (Feiner 2022). Ultimately, on October
discourse liberties may perceive themselves as part of an 27, 2022, Elon Musk and Twitter finalized the agreement,
“in-group” aligned with Musk’s ideals. This sense of shared actualizing Musk’s vision of a more free and decentralized
identity may lead to a convergence in their social media social media platform (Conger and Hirsch 2021).
activity and behaviors post-acquisition. Given this backdrop, our research quantifies the pro-
To capture the ideological alignment of entrepreneurs posed “Musk Effect,” on the Twitter activities of entrepre-
with Musk’s stance, we employ the liberal-conservative neurs from seven English-speaking nations (United States,
scale, a widely used measure in political science research Australia, New Zealand, United Kingdom, Canada, South
(e.g., Jost et al. 2009). Entrepreneurs who identify as more Africa, and Ireland) over a span of 56 weeks. For the pur-
conservative on the liberal-conservative scale may be more poses of this investigation, we compiled a dataset comprising
likely to express support for Musk’s actions, engage with 12.2 million authentic tweets from over 47,000 self-identi-
his content, and adapt their own tweeting behavior to reflect fied entrepreneurs (Bruns and Burgess 2012). A comprehen-
their ideological alignment with Musk. sive analysis was conducted on all the tweets, encompassing
Conversely, entrepreneurs who identify as more liberal account details such as username, account age, number of
on the scale may perceive themselves as part of an “out- lists, posts, followers, following, and any hashtags and men-
group” and exhibit different behavioral patterns on Twit- tions associated with each tweet (Tufekci 2014). Employing
ter, such as expressing criticism of Musk’s acquisition or trajectory cluster analysis, we classified the entire cohort of
disengaging from the platform altogether. By examining entrepreneurs into eight distinct, non-overlapping groups
the moderating role of political affiliation, as measured by or clusters (Andrienko et al. 2009). We extracted prevalent
the liberal-conservative scale, through the lens of Social conversation topics from these entrepreneurs’ tweets utiliz-
Identity Theory, we aim to uncover how the formation of ing a hashtag-based analysis (Bruns and Burgess 2012).
ideologically-driven in-groups and out-groups among entre- Furthermore, by examining the “following” information of
preneurs shapes their responses to Musk’s acquisition of the entrepreneurs, i.e., whom each entrepreneur follows on
Twitter. This approach allows us to better understand the
complex interplay between individual identity, group mem- 1
Elon Musk has used Twitter once per hour on average since buy-
bership, and social influence in the context of a significant ing the company (https://www.businessinsider.com/elon-musk-twitter-
event within the entrepreneurial ecosystem. The use of the use-increase-since-buying-company-2022-11; Accessed on 25 April,
2024).

13
 131 Page 4 of 17 Social Network Analysis and Mining (2024) 14:131

Twitter, we were able to estimate and graphically represent “manager,” “managing director,” “chief executive,” “direc-
the political leaning of these entrepreneurs (Conover et al. tor,” and “VP” (Hanik et al. 2023). We collected each user’s
2011). Lastly, we employed multivariate logistic regression identifier, name, username, and essential public statistics
to identify characteristics that could be ascribed to each of such as account age and the number of lists, posts, follow-
the entrepreneur clusters that emerged from our analysis ers, and following. Furthermore, we recorded whether their
(Hosmer Jr et al. 2013). accounts had URLs and were verified. For every tweet in
In the following sections of this manuscript, vertical the dataset, we have collected its unique identifier as well
dashed lines depicted in the figures correspond to the fol- as the unique identifier of its author. Furthermore, we have
lowing significant events: recorded the date of posting and the original text, including
all mentions (“@”) and hashtags (“#”).
● January 31, 2022: The commencement of Elon Musk’s In our study, we utilized a series of simple heuristics
acquisition of Twitter stock, as represented in certain based on gendered names, usernames, pronouns, family
figures. roles, and occupations to identify the users’ genders, where
● April 14, 2022: The date on which Elon Musk extended feasible. Subsequently, we employed gendered accounts to
an offer to Twitter for the procurement of the company. train a logistic regression algorithm for predicting the gen-
● July 8, 2022: The announcement by Elon Musk of his der of remaining users. Our model exhibited a precision of
intention to discontinue the proposed acquisition. 99.3%, thus rendering us confident in our acceptance of the
● October 27, 2022: The finalization of the agreement be- predicted genders (Hanik et al. 2023).
tween Elon Musk and Twitter. Due to the limited availability of tweets (at most 3,000
most recent tweets per account, including retweets, imposed
The remainder of this manuscript is organized as follows: by the Twitter API), we could only obtain partial timelines
The subsequent section delineates the procedures involved for some highly active users, with most posts made in the
in our data acquisition methodology and elucidates the cri- last weeks of the observation period. This disparity created
teria employed for final sample selection. The third section a distorted perception of increasing posting frequency over
offers a comprehensive examination of our data, inclusive of time. Since we aimed to employ weekly posting frequency
an in-depth description of the user classification methodol- to gauge Twitter’s response to Musk’s acquisition, such bias
ogy and the estimation of political leaning. The penultimate was untenable.
section engages in a discussion of the regression analysis, To examine and mitigate this bias, we plotted the cumu-
accompanied by the associated findings for each cluster of lative distribution of the dates of the first recorded tweet.
entrepreneurs. The final section provides a conclusion to the For example, accounts with complete and uninterrupted
paper, summarizing the key findings and their implications. timelines originating on or before March 25, 2022, well
before any overt acquisition-related Musk activity, consti-
tuted 78% of all harvested accounts. Consequently, if we
2 Data Acquisition sacrificed 22% of the dataset, we could assure timeline uni-
formity. We could expand the timeline at the expense of the
Our dataset comprises 12,176,276 tweets posted by 47,189 number of users or include more active users at the cost of
unique users for 81 weeks, commencing October 1, 2022, a shorter timeline. As previously mentioned, we ultimately
and concluding on April 21, 2023. The users were randomly opted for a subset of 78% of users whose timelines began on
selected from the live Twitter stream based on two crite- or before March 25, 2022. The chosen observation window
ria. Firstly, they must originate from an officially English- encompassed the entire Twitter acquisition saga, including
speaking country (the USA, Great Britain, Canada, Ireland, its prelude.
New Zealand, or Australia) or South Africa. Secondly, their The truncated dataset contains 4,034,007 (39.5%) tweets
Twitter profile descriptions must contain at least one of the posted by 36,828 (78%) unique users throughout 56 (69%)
following keywords, either verbatim or as a hashtag: “entre- weeks. To summarize the composition of the dataset,
preneur,” “owner,” “founder,” “co-founder,” “business Table 1 provides the number of unique users per country of
owner,” “CEO,” “co-owner,” “representative,” “consul- tweet origin.
tant,” “assistant,” “analyst,” “officer,” “partner,” “devel-
oper,” “head,” “coordinator,” “specialist,” “executive,”

Table 1 Number of unique users per country in the truncated dataset

Country Count Country Count Country Count Country Count
AU 3,732 CA 5,453 IE 2,866 NZ 812
UK 4,561 US 13,918 ZA 5,486

13
Social Network Analysis and Mining (2024) 14:131 Page 5 of 17 131

3 Data Analysis 3.1 Analyzing Tweet and Hashtag frequencies

In this section, we describe the analytical tests and meth- Figure 1 presents the mean tweeting frequency, measured as
odologies that we have applied in our study to dissect the the number of tweets per week per person for each country
Twitter dataset concerning entrepreneurial reactions to of origin, and the mean tweeting frequency for the entire
Elon Musk’s involvement with the platform. Initially, the dataset. The noticeable and consistent decline in tweeting
analysis focuses on the frequency of tweets and hashtags to frequency across all countries on the right side of the chart is
gauge the temporal dynamics of user engagement, particu- attributed to the Christmas/New Year holiday season, with
larly during pivotal events surrounding Musk’s acquisition the decline being less pronounced in South Africa. Addition-
of Twitter. This is followed by a sophisticated clustering ally, the frequencies display an almost continuous decline
approach that segments users based on their tweeting pat- between April and early July, which will be explored further
terns, revealing distinct groups that reacted differently to the in the following section upon introducing Fig. 3.
corporate developments. This study examines the user perception of Elon Musk’s
Further analytical depth is achieved through an exami- online persona on Twitter. Users have the option to refer
nation of the conversation topics prevalent among entre- to Mr. Musk through either the mention “@elonmusk” or
preneurs on Twitter. By analyzing networked interactions the hashtags “#elonmusk,” “#elon,” and “#musk.” The lat-
of hashtags and tokens, we uncover the thematic structures ter option is considered indirect, as it does not immediately
that dominate the discourse and assess their alignment with notify Mr. Musk of the tweet’s contents; it is primarily
Musk’s public persona and actions. Finally, we assess the intended to contribute to the trending activity on the site’s
political leanings of these entrepreneurs, utilizing a net- homepage and attract the attention of other subscribers
work analysis to understand how political affiliations may (Bruns and Moe 2014).
influence their responses to Musk’s leadership and policy On the other hand, a mention option is a direct form of
initiatives on Twitter. These combined efforts provide a communication that immediately alerts Mr. Musk or his des-
comprehensive view of the entrepreneurial landscape on ignated account handler of the mention. As such, it may be
Twitter, illustrating both the direct and subtle influences of viewed as a more personal approach to garner Mr. Musk’s
Musk’s high-profile activities. attention.
Figure 2 illustrates that the entrepreneurs exhibited a high
frequency of mentions and hashtags twice. The first instance
occurred in late April to early May, shortly after Mr. Musk’s
offer to purchase Twitter. The second instance occurred
Fig. 1 Tweeting frequency, per
country, over time. The thick red
line shows the mean tweeting
frequency

13
 131 Page 6 of 17 Social Network Analysis and Mining (2024) 14:131

Fig. 2 Mentions of Elon Musk

(@elonmusk) and the namesake
hashtags (#elonmusk, #musk, and
#elon) over time

from late October to early November, following the finaliza- personally significant to the users we observed, leading to
tion of the acquisition. Based on Fig. 2, the initial surge in more direct and personal responses.
activity was brief, lasting only one week, and featured both It should be noted that the legal dispute in July 2022,
mentions and hashtags being used equally excessively. In in which Elon Musk declared his intention to terminate the
contrast, the second surge lasted approximately two months. acquisition, did not elicit a significant response in terms of
It was primarily driven by direct mentions, which could be the hashtag and mention frequencies. Specifically, there was
because the news of the completed acquisition was more

Fig. 3 Tweeting frequency per

group over time. The groups are
arranged in descending order of
size. The point values represent
the mean tweeting frequencies
during the preceding observation
period

13
Social Network Analysis and Mining (2024) 14:131 Page 7 of 17 131

no observable change in the frequency of mentions, and the size and activity level, suggesting that most Twitter users
increase in hashtag usage was minimal. are passive observers or re-tweeters.
Groups 5 and 8, where the latter is the smallest, consist-
3.2 Classifying the users ing only of 0.6% of the dataset, exhibited reduced activ-
ity between the announcement of the intention to acquire
Our goal is to identify subgroups of users that exerted dif- Twitter and the attempted withdrawal from the deal. These
ferent reactions to the acquisition process. We pursued the groups are primarily responsible for the linear decline
trajectory clustering approach proposed by Andrienko et observed on the left side of Fig. 1.
al. (2009) to achieve this. Although the aggregate tweeting Lastly, the users in group number 7 demonstrate a sig-
activity over the observation period was remarkably stable, nificant increase in activity in anticipation of and imme-
both on average and by country (excluding the dip during diately after the closure of the acquisition deal, despite
the holiday season and the gradual decline in the first half of initially being aligned with (and potentially members of)
2022), we believe that there might be hidden patterns in the group number 4. This group, although the second smallest
tweeting behavior of subgroups of users that are relevant to (1.9%), is the most vocal responder to the corporate leader-
our investigation. Therefore, by applying trajectory cluster- ship change.
ing, we aim to identify groups of users with similar tweeting
patterns and analyze their reactions. 3.3 Extracting token-based conversation topics
We computed a tweeting trajectory for each observed
user by quantifying their tweeting frequencies at, a week and Grouping the users revealed a structural response of the
two weeks before, a week and two weeks after, and between entrepreneurs on Twitter to the takeover. On the contrary,
the notable events (depicted as the “dashed lines”)2. Thus, topic analysis via tokens addresses their semantic response,
each trajectory was represented as a 15-dimensional vector if any.
of real numbers. The averaging of the intervals between the We utilized complex network analysis on a network of
acquisition events aimed to eliminate any sensitive response hashtags to identify the topics being discussed (Zinoviev
to unrelated episodes, such as national elections. 2018). The network consists of nodes representing the
We further applied the k-means clustering algorithm (Jin most frequently used hashtags (#ai, #leadership, #busi-
and Han 2010) to divide the vectors into eight distinct and ness, #twitch, #bitcoin, #marketing, #crypto, #motivation,
non-overlapping groups—clusters. Each vector in a cluster #technology, #realestate, etc.) Aside from the hashtags, we
is closer (based on the Euclidean distance) to the vectors in included hashtag lookalikes: lemmatized words that are
the same cluster than those in any other cluster. The cen- spelled exactly as the chosen hashtags (e.g., “ai,” “leader-
troid, the arithmetic mean of all vectors in a cluster, is rep- ship,” and “business”). Below, we refer to the hashtags and
resentative of that cluster. We call all users whose tweeting lookalikes as tokens. 67.6% of tweets contained at least one
frequency trajectories belong to the same cluster, a group. token of interest, and 43.2% contained two or more.
Figure 3 shows the centroids of the eight groups obtained Two nodes are connected with an edge if the correspond-
through this process, arranged in descending order of group ing tokens are used in the same tweet at least once, and the
size. weight of the edge is proportional to the number of simulta-
We selected the number of groups to be eight to ensure neous occurrences, normalized to the unit range. The result-
that the centroids of each group were not too similar and to ing network has 1,640 distinct tokens and 111,591 edges of
separate users with abnormal behavior into distinct groups. various weights. We removed edges between tokens that
Selecting any other number of groups would have violated co-occurred 30 times or fewer. The ten strongest edges in
one of these requirements. the network are (#follow, #people), (#time, #work), (#game,
Five of the eight groups consist of users with a stable #play), (#people, #work), (#birthday, #happy), (#people,
tweeting history who were not significantly affected by the #time), (#team, #work), (#health, #mental), (#life, #people),
company acquisition events. The centroids of these groups and (#game, #time), each sorted alphabetically.
differ only in activity levels but are otherwise close to We employed the Louvain community detection method
straight horizontal lines in ascending order: 1, 2, 3, 4, and 6. (Blondel et al. 2008) with the resolution r = 0.8 (for a bet-
Interestingly, there is an inverse correlation between group ter understanding of the fine community structure) to iden-
tify groups of tightly connected network nodes representing
token-based conversation topics. This method produced 15
2
Specifically, we used the following dates: 2022-03-31, 2022-04-07, topics, from which we selected the nine most productive
2022-04-14, 2022-04-21, 2022-04-28,2022-06-25, ’022-07-01, 2022- topics for further analysis, as shown in Table 2.
07-08, 2022-07-15, 2022-07-22, 2022-10-13, 2022-10-20, 2022-10-
27, 2022-11-03,2022-11-10, 2023-01-29, and 2023-04-21.

13
 131 Page 8 of 17 Social Network Analysis and Mining (2024) 14:131

Table 2 The most frequent tokens in the most productive conversation attributed to the FIFA World Cup 2022 and the winter holi-
topics
day season, and the plunge of the Finance topic is likely to
Tokens Conversation Topics
be the October effect. We imply that there was no observable
Entertainment twitch; podcast; tiktok; gaming;
gamedev; music conversation topic on the entrepreneurs’ Twitter accounts
Entrepreneurship business; marketing; technology; that reflected or was affected by the takeover.
cybersecurity; innovation; tech
Finance bitcoin; crypto; nft; btc; invest- 3.4 Estimating political Leaning
ing; blockchain
Lifestyle photography; fashion; food;
The research question remaining to be investigated concerns
travel; family; christmas
Healthcare health; healthcare; community; the potential influence of political leaning on entrepreneurs’
charity; veterans; covid reactions to the change of ownership. A subnetwork of
Geopolitics canada; ukraine; pakistan; usa; entrepreneurs as Twitter followers was extracted to exam-
women; nightcafe ine this. It is assumed that if a user follows another user,
Spirituality motivation; inspiration; peace; they share some views with the followed user (Burgess and
cio; life; jesus
Baym 2022) and may lean politically in the corresponding
Education education; art; edtech; books;
author; teachers direction.
Professional leadership; jobs; job; inclusion; However, constructing a one-dimensional political spec-
diversity; work trum scale compatible with the liberal-conservative scale in
Note: Both topics and tokens within topics are listed in decreasing the USA is a challenging task due to the international nature
order of frequency of the dataset. Creating such a scale necessitates a profound
and multifaceted understanding of local politics, which we
Cluster labeling is a problem commonly associated with as the authors lack. Therefore, we have opted to use the
topic extraction via complex network analysis. Instead of existing North American liberal-conservative scale as a
labeling clusters with their most frequently used tokens, universal yardstick (Jost et al. 2009), acknowledging the
we approached ChatGPT4 and requested to summarize all potential for accusations of Americentrism while fully rec-
tokens in each cluster with a single word (the left column ognizing that the scale may not fully capture local nuances.
in Table 2). For each user in the dataset, we collected a list of up to
Most topics depicted in Fig. 4 exhibit notably stable 1,000 users they follow (“followees”). The choice of the
dynamics except for three topics. Specifically, the surge limit was affected by the recent changes in Twitter’s API
in the Entertainment and Lifestyle topics can be readily policies that severely restrict the followee lists’ download

Fig. 4 Token-based conversation

topic frequencies over time

13
Social Network Analysis and Mining (2024) 14:131 Page 9 of 17 131

Fig. 5 The network of Twitter

accounts based on co-following,
with major network communities
detected: ?liberal? (baracko-
bama, joebiden), ?conservative?
(elonmusk, realdonaldtrump),
?entertainment? (mrbeast, oprah,
vancityreynolds, ellendegeneres),
?business? (billgates, wsj),
?South African? (trevorhoah,
rihanna), etc

speed. A bipartite, directed, unweighted network was con-

structed with entrepreneurs as one set of nodes and their 4 Regression analysis and findings
followees as the other set (we refer to them as “opinion
leaders,” for lack of a better term). If multiple entrepreneurs We employed a logistic regression model to identify the
followed the same two opinion leaders, a new edge was factors that explain the entrepreneurs’ association with the
created connecting the leaders, thereby building a new pro- groups. Due to missing gender descriptions, only 33,668
jected network of co-followed (Neal 2014). The resulting users were included in the regression analysis. The results
projected network has 2,840 nodes and 662,452 edges. It is of this analysis are presented in Tables 4 and 5.
relatively sparse (density d ≈ 8.2%) and has high modular- The first insight from Table 4 is that except Group 1,
ity (m ≈ 0.575) and a crisp community structure (Newman all other groups have relatively newer Twitter accounts,
2006). which shows that they have not been on Twitter that long.
Upon applying the Louvain community detection algo- We remind the readers that Group 1 is the biggest of all the
rithm (Blondel et al. 2008), we identified 14 groups of groups and constitutes about 56.5% of all the entrepreneurs
opinion leaders, including highly politically polarized com- in our study. With that in mind, it will be interesting to not
munities (Conover et al. 2011). The findings are summa- only examine how some of the smaller groups are similar
rized in Table 3 and visualized in Fig. 5. to and different from this majority group in terms of their
Of particular interest are clusters #1 (“liberal”) and #3 characteristics but also observe what makes these smaller
(“conservative”). As a result, we can calculate the number groups similar to and different from each other. Interest-
of opinion leaders on lists #1 and #3 that each entrepreneur ingly, we find that Groups 2 and 3 have similar traits in sev-
in the dataset follows and interpret the difference between eral dimensions:
the two as an indicator of their political leaning.
Figure 6 displays a histogram of the political leaning for ● Entrepreneurs who belong to either of these groups have
all included entrepreneurs, revealing a symmetrical distri- strong followers but do not necessarily follow a high
bution with the mean and median leaning close to 0. The number of accounts themselves (as captured by the “Fol-
shape of the distribution suggests that, within the context of lowing” count from their respective Twitter profiles);
the proposed political leaning estimation framework, Twit- ● They have a relatively high frequency of tweets;
ter as a whole is politically neutral, at least for the sample ● They provide fairly long bio descriptions, including
of entrepreneurs. their personal website URL, that shows that they want
to be known and want people to land on their webpage

13
 131 Page 10 of 17 Social Network Analysis and Mining (2024) 14:131

Table 3 The top opinion leaders in each cluster too, have relatively newer Twitter accounts, long bio
Sl # of Label Top Members descriptions including personal URLs, strong followers but
# Followers
not necessarily following many accounts themselves, and
1 709,802 Liberal barackobama, joebiden,
potus, kamalaharris, aoc, a high frequency of tweets. What separates Group 6 from
vp, nasa, nytimes, michel- Groups 2 and 3 is that they are primarily located in the UK
leobama, bbcbreaking and New Zealand. Also, this group has significantly more
2 414,959 ZA + Soccer trevornoah, rihanna, drake, male entrepreneurs.
kevinhart4real, fabrizio-
romano, cyrilramaphosa,
Next, looking at Group 8, we see that this group also
kimkardashian, cristiano, has a strong male presence with a high number of tweets
premierleague, iamcardib with a short description and without URLs, and consists of
3 329,174 Conservative elonmusk, realdonaldtrump, a disproportionately lower number of people from the UK.
spacex, potus45, tesla, joero-
Its members have relatively newer accounts on Twitter and
gan, jack, snowden, garyvee,
jordanbpeterson strong followers but do not necessarily follow a high num-
4 229,897 Entertainment mrbeast, oprah, youtube, ber of accounts themselves. These features indicate that
kingjames, therock, twitch, those entrepreneurs provide opinions to their followers but
espn, sportscenter, snoop- do not necessarily want to be known or interested in people
dogg, verified
knowing who they are or what they do. This trait of entre-
5 130,348 Entertainment vancityreynolds, ellendege-
neres, markhamill, jim- preneurs in Group 8 is similar to those in Groups 2, 3, and
myfallon, robertdowneyjr, 6, who are more into being known/promoting themselves
tomhanks, netflix, chrissytei- and their businesses. What mainly separates these groups
gen, taylorswift13, ladygaga (Groups 8, 2, 3, and 6) is mostly where they are located. We
6 90,631 Business billgates, wsj, theeconomist,
call them self-promoters.
forbes, whitehouse45, richard-
branson, business, techcrunch, Continuing to examine our regression results further, we
harvardbiz, simonsinek see that Groups 4, 5, and 7 stand out in a different spec-
7 7,295 Formula-1 f1, danielricciardo, landonor- trum. While being similar to each other and some of the
ris, redbullracing, mclarenf1, earlier groups in terms of having relatively newer Twitter
max33verstappen, scude-
riaferrari, georgerussell63,
accounts with long bio descriptions, including their personal
charles leclerc, skysportsf1 URL and high tweet frequency, entrepreneurs who are part
8 3,752 SkyNews skynews, garylineker, sky- of this group neither follow many other Twitter accounts
newsbreak, lord sugar themselves nor necessarily have many followers (i.e., the
9 2,882 Top Gear jeremyclarkson, mrjamesmay, estimated coefficients were statistically insignificant). We
richardhammond
decided to label them as aspirants—entrepreneurs on their
10 1,824 MS Windows microsoft, windows
path to becoming self-promoters one day but do not have the
11 1,677 PGA tigerwoods, pgatour
12 1,512 Twitter Media twittercreators, vinecreators
critical mass of followers qualifying them as self-promoters.
13 872 Country blakeshelton, Also, we notice that entrepreneurs from Groups 4, 5, and 7
carrieunderwood mostly seem to originate from the US. Finally, let us exam-
14 805 Advertising adweek, adage ine Group 1—our most populous group of entrepreneurs.
Note: Both clusters and leaders within clusters are listed in the Besides accommodating an overwhelming majority
decreasing order of the number of followers. (56.5%) of our entrepreneurs in our study, this group has a
set of starkly distinct features that help entrepreneurs in this
for more information about themselves and what they group stand out from all the other groups.
do; and, For example, the entrepreneurs belonging to Group 1
● They are not from South Africa. have relatively older Twitter accounts, a relatively shorter
bio description with no URL (indicating that these entre-
Based on the above set of common characteristics, we preneurs are possibly not keen on self/business promotional
believe that entrepreneurs belonging to Groups 2 and 3 activities), and do not have many of followers. However,
may best be considered as opinion providers who use Twit- they follow many accounts and have the distinctive trait
ter for the interest of self/business promotion. In other of not tweeting much. In other words, entrepreneurs in our
words, these entrepreneurs’ Twitter presence seems largely most populous group are on Twitter mostly to silently read
business-driven. and observe. This group also consists of a significantly
Entrepreneurs in Group 6 have many traits similar to large proportion of entrepreneurs who do not have a veri-
those in Groups 2 and 3, except for where they are. They, fied account, i.e., they do not have the blue check associated

13
Social Network Analysis and Mining (2024) 14:131 Page 11 of 17 131

Fig. 6 Estimated political leaning

of ~45,500 users (“liberals” on
the left, “conservatives” on the
right)

with their Twitter accounts3. That means these people are groups of groups) based on the broad similarity of their
not interested in being publicly known by promoting the tweeting trajectories (see Fig. 3).
authenticity of their Twitter accounts. We call them silent Entrepreneurs in Group 7 who belong to the aspirants’
observers or lurkers. They consume content but do not group, easily stand out and seem to have the most distinct
actively participate by posting or replying to tweets. This change in their tweet trajectory in response to Elon Musk’s
finding is consistent with research studies that show that a official acquisition of Twitter (November 2022). We find
small percentage of Twitter users contribute a large percent- that, just before the acquisition deal was closed, their tweet
age of the content (Hughes and Wojcik 2019). frequency increased significantly, continued increasing for
Overall, entrepreneurs in Groups 2, 3, 6, and 8, whom a little but post-acquisition, had a flat trajectory for a very
we broadly label as self-promoters, are regularly on Twitter short while after closing, and then declined sharply after
to be known and use the platform for self/business inter- that. In sharp contrast to the overall muted sentiments from
ests. Twitter is mostly a marketing/business tool for them to entrepreneurs belonging to most other groups in response
connect with their target audience and find new audiences. to the official acquisition event, this group showed the most
Collectively, they comprise about 31.6% of the sample we visible reaction to Musk’s acquisition of Twitter. Inciden-
studied. We remind the reader that we label the entrepre- tally, the members of this group tend to lean right.
neurs in Groups 4, 5, and 7 as aspirants and believe they Knowing they are aspirants and want to self-promote
have the same intention as the self-promoters but are not and tweet with high frequency, their strong reaction (posi-
quite there yet. They make up about 11.9% of the sample tive or negative) makes sense. Groups 5 and 8 (also aspi-
that we studied. Interestingly, entrepreneurs belonging to rants), both with a desire to have strong followers, had high
only Group 7 seem to own a relatively higher proportion of tweet frequency when Musk announced his intention to buy
blue-check verified Twitter accounts. About 43.5% of the Twitter (May 2022). However, their tweeting frequency
entrepreneurs, we studied use the Twitter platform for self- declined sharply after that and went almost flat when Musk
promotion or want to be linked/known for who they are and announced his intention to withdraw from the deal (July
what they do, while a larger number (56.5%) of entrepre- 2022). Both had an instantaneous reaction to the final offi-
neurs use Twitter just to read and observe silently. cial acquisition, which eventually became flat. The tweeting
Next, in Table 5, we report the summary statistics of trajectory is similar over our entire observation period for
our entrepreneurs by further grouping them (i.e., forming all the remaining groups (1, 2, 3, 4, and 6). Though these
groups vary significantly from each other in terms of their
3
The blue verified badge on Twitter lets people know that an account frequencies of tweeting, their overall reaction to Musk
of public interest is authentic.

13
 131 Page 12 of 17 Social Network Analysis and Mining (2024) 14:131

Table 4 Group sizes and logistic regression coefficients, with effect sizes (Cohens’s d)
Groups 1 2 3 4 5 6 7 8
(Size, %) 56.5 19 8.9 6.7 3.3 2.8 1.9 0.6
Male? -0.02 0 0.08⁺ -0.1* 0 0.13⁺ 0.18* 0.24
(-0.01) (0.02) (-0.01) (0.06) (0.02) (-0.02) (-0.08) (-0.12)
Leaning -0.39 -0.25 0 -0.15 1.33 0.98 2.38* 3.15
(-0.02) (0.02) (0.04) (0.03) (-0.05) (-0.02) (-0.09) (-0.16)
From CA? -0.12* 0 0.08 0.08 0.18 0.2 0.29⁺ -0.31
(0.04) (-0.01) (-0.04) (-0.05) (-0.04) (-0.04) (-0.04) (0.02)
From IE? 0.06 0 0.06 -0.29* -0.29⁺ 0.22 -0.14 -0.4
(-0.04) (-0.01) (0.00) (0.09) (0.11) (0.03) (0.11) (0.08)
From NZ? -0.11 0 -0.09 -0.24 0.21 0.38 0.53⁺ -0.45
(0.00) (-0.02) (0.01) (0.03) (0.00) (-0.02) (-0.05) (0.04)
From UK? -0.08 0 0.13 -0.1 -0.28⁺ 0.34* 0.17 -1.05**
(0.03) (-0.04) (-0.05) (0.04) (0.10) (-0.09) (0.00) (0.18)
From US? -0.17*** 0.03 0.03 0.19* 0.22⁺ 0.23⁺ 0.3⁺ -0.38
(0.09) (-0.02) (-0.01) (-0.15) (-0.17) (-0.06) (-0.10) (-0.01)
From ZA? 0.25*** -0.22*** -0.17* -0.19⁺ -0.1 -0.1 0.11 0.07
(-0.14) (0.08) (0.10) (0.13) (0.06) (0.15) (0.05) (-0.20)
Account age 0.09*** -0.03*** -0.04*** -0.04*** -0.09*** -0.11*** -0.1*** -0.2***
(0.13) (-0.04) (-0.11) (-0.20) (0.02) (-0.12) (-0.05) (0.21)
Descr. Length -0.58*** 0.29*** 0.5*** 0.29** 0.32* 0.76*** 0.57** -0.24
(0.24) (-0.12) (-0.20) (-0.14) (-0.13) (-0.28) (-0.18) (0.00)
Has URL? -0.47*** 0.17*** 0.31*** 0.41*** 0.39*** 0.51*** 0.36*** 0.3⁺
(0.32) (-0.13) (-0.24) (-0.26) (-0.21) (-0.37) (-0.26) (-0.18)
log(#followers) -0.28*** 0.19*** 0.21*** -0.13* 0.05 0.22** 0.21* 0.4*
(0.51) (-0.21) (-0.36) (-0.34) (-0.32) (-0.58) (-0.48) (-0.51)
log(#following) 0.38*** -0.09* -0.18*** -0.31*** -0.04 -0.44*** -0.3*** -0.51**
(0.25) (-0.12) (-0.18) (-0.17) (-0.21) (-0.23) (-0.22) (-0.18)
log(#listed) -0.03 -0.03 0.02 0.06 -0.16* 0.16* -0.04 0.01
(0.45) (-0.17) (-0.34) (-0.38) (-0.22) (-0.57) (-0.38) (-0.22)
log(#tweets) -1.25*** 0.31*** 0.62*** 1.3*** 1.13*** 1.5*** 1.3*** 1.6***
(0.70) (-0.23) (-0.42) (-0.65) (-0.53) (-0.77) (-0.35) (-0.67)
Verified? -0.23* -0.18⁺ 0.13 0.15 0.25 -0.12 0.43* -0.97
(0.15) (-0.03) (-0.14) (-0.12) (-0.11) (-0.18) (-0.25) (0.03)
+ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001

Table 5 Average statistics by groups with similar tweeting trajectories of their political affiliation (left-wing or right-wing). When
Groups A = 1,2,3,4,6 B = 5,8 C=7 we look at their behavioral tendencies, we see that some are
Male? 0.66 0.66 1 closer to one another: Groups 2, 3, and 6 as the self-promot-
Leaning (×1000) -2 0.79 1.81 ers; Groups 4, 5, 7, and 8 as the aspirants; and Group 1 as
Account age 9 8.94 9.28 silent observers/lurkers. Irrespective of their diverse inten-
Descr. Length 1 0.74 0.76 tions and motives to be on Twitter and their varying fre-
Has URL? 1 0.67 0.7 quency of platform usage, they all have similar reactions to
log(#followers) 2.66 2.92 3.01 Elon Musk’s acquisition of Twitter, i.e., no reaction— which
log(#following) 2.73 2.85 2.86 we thus call a “Non-Musk Effect.” We believe the instanta-
log(#listed) 0.81 0.97 1.09
neous reaction we observe in only a few groups (Groups
log(#tweets) 3.33 3.71 3.79
5, 7, and 8) also conforms with the behavior pattern of the
Verified? 0.02 0.03 0.05
other groups, as all their tweeting trajectories went flat soon.
In the Twitterverse, entrepreneurs quickly returned to their
buying Twitter is quite the same - and can be considered world or never left it in most cases, even as Musk’s antics
very muted at best. vis-à-vis the Twitter acquisition deal had caught the popular
In our study, we observe and analyze the tweeting behav- press and technology world in a frenzy!
ior of almost 37,000 entrepreneurs who seem to sort them
into eight distinct groups based on their motives, intentions,
and usage of the Twitter platform. Surprisingly, we do not
find any significant difference between the groups in terms

13
Social Network Analysis and Mining (2024) 14:131 Page 13 of 17 131

5 Discussion and conclusion they do not necessarily translate into significant behavioral
changes among his entrepreneurial peers on social media.
Our investigation into the “Musk Effect” amid his Twitter This highlights a critical distinction between the inspira-
acquisition merges two pivotal discussions: the ideologi- tional power of the “delivering hero” and the practical, day-
cal influence of neoliberalism on entrepreneurship and the to-day actions of other entrepreneurs. This further suggests
social dynamics of utopian market hero formation. Starting a disconnection between Musk’s high-profile persona and
from the 1960s, neoliberal ideology has celebrated mar- his actual influence on entrepreneurial behavior on social
ket-oriented solutions and entrepreneurial heroism, view- media. The “utopian market hero” concept further illumi-
ing business leaders as agents of societal change who use nates this disconnection. It describes how entrepreneurs like
market-driven strategies to tackle global challenges (Luri et Musk create narratives that resonate by envisioning futuristic
al. 2023; Shamir 2008; Yngfalk and Yngfalk 2020). Elon solutions to current problems. These narratives, which often
Musk, a quintessential figure in this narrative, embodies the project idyllic visions like a world with free speech where
“utopian hero”—a persona who not only proposes innova- people can freely express themselves without censorship,
tive solutions, like space-based technologies, renewable captivate public imagination (Anderson and Musk 2017).
energy, and solving the commuting problem with his Boring The ability to inspire through these visions, rather than their
Inc., but also frames these solutions as essential to societal immediate practicality, is what elevates entrepreneurs to
progress (Luri et al. 2023; Sørensen 2008). hero status. This process involves both top-down visionary
Based on “delivering hero” archetype and social influ- influence and bottom-up collective longing, which together
ence theory, we expected to find a significant ‘Musk Effect’ form a feedback loop in the public imagination, constrained,
on the Twitter activity of self-identified entrepreneurs, par- however, by market structures and ideologies that limit the
ticularly after Musk’s acquisition of Twitter. However, our choice of heroes to those sanctioned by the market (Li et al.
study did not yield evidence supporting a significant influ- 2023 Maclaran and Brown 2005; Mosco 2005).
ence. Our empirical analysis of 12.2 million tweets from The reality, as revealed by our tweet analysis, indicates
over 47,000 entrepreneurs collected over the term of 56 that while Musk’s narratives may significantly shape public
weeks from seven English speaking countries (US, Austra- perceptions and desires (Niburski and Niburski 2023; Rude-
lia, New Zealand, UK, Canada, South Africa, and Ireland) loff and Damms 2022), they do not necessarily translate into
offers a nuanced perspective. It shows that despite Musk’s active engagement or behavioral change among his entre-
significant ideological stance—particularly his commitment preneurial peers, nor create in-group or out-group reaction
to free speech as a motivation for acquiring Twitter—there differences to Musk acquisition of Twitter. In simple terms,
was minimal reaction from the entrepreneurial community one reason may be that bottom up collective longing was not
to his acquisition of Twitter. Results revealed that only less resonating with Musk’s top-down visionary influence, and
than 2% of entrepreneur react to Musk’s acquisition news, feedback loop was not as strong as one may expect among
which is signaled by these accounts’ tweet frequency. That entrepreneurs (Luri et al. 2023). Our findings showed that
means the media reaction to Twitter’s acquisition by Musk discussions on Twitter remained largely focused on other
was not necessarily aligned with main street’s reaction. Peo- interests, and even areas previously influenced by Musk,
ple, in our case, entrepreneurs, did not care that much about like cryptocurrency, showed declines in engagement.
Musk’s ownership of Twitter. Most entrepreneurs, around While the “delivering hero” archetype, as described by
56.5%, use the platform to read and observe. 43.5% of Luri et al. (2023), suggests that entrepreneurs like Elon Musk
them are there to promote their business for self-promotion. can become heroes in the public eye by promising to solve
Moreover, the group that reacted mildly to Musk’s acquisi- pressing societal problems (e.g., free speech), our findings
tion is Group 7, which is in the self-promoters cluster. indicate that admiration for such figures does not necessarily
The “delivering hero” archetype, as described by Luri et translate into direct changes in social media behavior. The
al. (2023), presents entrepreneurs as utopian market heroes influence of these “utopian heroes” may be more symbolic,
who are celebrated for proposing futuristic solutions to cur- shaping broader industry narratives and aspirations rather
rent societal challenges. These entrepreneurs, such as Elon than immediate actions. Entrepreneurs become heroes when
Musk, are often seen as visionary leaders who drive inno- their narratives of utopian futures resonate with consumers’
vation and progress. Musk’s ventures, including SpaceX, desires for a better world. However, the lack of significant
Tesla, SolarCity, and Neuralink, illustrate his role in pro- changes in entrepreneurs’ Twitter activity following Musk’s
moting narratives of a utopian future where technological acquisition suggests that while they may admire his vision-
advancements address global issues. ary leadership, they prioritize their own strategic goals and
Our null findings suggest that while Musk’s visionary market realities over aligning their social media behavior
narratives and heroic status captivate public imagination, with his actions.

13
131 Page 14 of 17 Social Network Analysis and Mining (2024) 14:131

This disconnect underscores the complexity of the “deliv- and contextual factors that mediate the impact of high-pro-
ering hero” archetype in practice. As Luri et al. (2023 p.107) file figures.
note, “achieving widespread acceptance of and desire for In addition to the complexity of the “delivering hero”
their personal visions of a future utopia is what turns this archetype in shaping entrepreneurial behavior, the lack of
type of entrepreneur into a hero.” Our findings suggest that support for the moderating role of political affiliation in our
this acceptance and desire may not always lead to direct study also prompts a reconsideration of Social Identity The-
emulation, particularly when it comes to entrepreneurs’ ory in the context of digital influence. According to Social
social media strategies. Identity Theory, individuals who strongly identify with a
Instead, the influence of high-profile figures like Musk particular social group are more likely to conform to the
may operate on a more symbolic level, shaping the broader norms and behaviors associated with that group (Tajfel and
ideological context in which entrepreneurs operate. As Turner 1986). In our case, we hypothesized that entrepre-
the Luri et al. (2023) argue, the intersection of three ide- neurs who share Musk’s political ideology would be more
ologies—the market, technology, and individual heroes as likely to align their social media behavior with his actions
saviors—provides a fertile ground for entrepreneurs to be following the acquisition of Twitter.
framed as heroes. While this ideological context may inspire However, our findings suggest that even among entrepre-
and influence entrepreneurs, it does not necessarily dictate neurs who may identify with Musk’s political views, there
their day-to-day actions on social media platforms. was no significant change in their Twitter activity or behav-
Social Influence Theory posits that individuals’ beliefs, ior. One possible explanation is that in digital networks, the
attitudes, and behaviors are shaped by their social interac- formation and influence of social identities may be more
tions, particularly with influential figures and peers. Tradi- fluid and context-dependent than in traditional offline set-
tionally, this theory suggests that high-profile individuals can tings. Entrepreneurs may identify with Musk’s political ide-
directly influence the actions and decisions of those within ology in certain contexts, but this identification may not be
their social sphere. However, our findings challenge the strong enough to override their own strategic priorities and
assumption that such influence is both direct and uniform, market considerations when it comes to their social media
particularly within the context of social media networks. behavior. Furthermore, the complex nature of digital influ-
Again, our study revealed minimal changes in tweeting ence, as discussed earlier, may also play a role in attenuat-
behavior among entrepreneurs following Musk’s acquisi- ing the effects of social identity on behavioral conformity.
tion of Twitter. This outcome indicates that the dynamics of In online environments, the influence of high-profile figures
social influence in digital networks may differ significantly may be mediated by a range of factors, such as the specific
from those in more traditional, face-to-face settings. While platform dynamics, the presence of competing influencers,
Musk’s actions and statements undoubtedly generate con- and the individual’s own network structure. These findings
siderable media attention and public discourse, this height- suggest that the application of Social Identity Theory in
ened visibility does not necessarily translate into substantial digital contexts requires a more nuanced approach, one that
behavioral shifts among the entrepreneurial community. considers the unique characteristics of online environments
This discrepancy suggests that digital influence operates and the multiple factors that can shape the formation and
in a more complex and diffuse manner. In digital networks, influence of social identities. Future research could explore
the influence of high-profile figures like Musk may not these factors in more depth, examining how social identi-
manifest as straightforward behavioral mimicry. Instead, ties intersect with other variables to shape entrepreneurial
their impact might be more about shaping the broader narra- behavior in digital networks.
tive and discourse within entrepreneurial communities. For Overall, findings challenge the expected “Musk Effect,”
instance, Musk’s tweets and public statements might inspire revealing a complex landscape of digital communication
discussions, reflections, and indirect influences that shape where entrepreneurial actions are influenced by a blend of
the overall entrepreneurial ecosystem rather than causing admiration for visionary narratives and the practical realities
direct, immediate changes in individual behavior. These of business operations. Our study highlights the importance
insights prompt a reevaluation of Social Influence Theory of understanding both the ideological narratives that shape
when applied to digital contexts and/or considering other entrepreneurial identities and the actual behaviors prompted
mediating factors when the context is social media. The tra- by these narratives. It underscores the need for a nuanced
ditional view of social influence as a direct cause-and-effect view of social influence in the digital age, recognizing that
relationship may need to be adjusted to account for the sub- public personas like Musk may inspire widespread interest
tler, more varied ways influence is exerted in online envi- but have a limited direct impact on the digital behaviors of
ronments. In digital networks, the pathways of influence are their peers. This nuanced understanding is crucial for accu-
likely more intricate, involving a range of intermediaries rately assessing the role of high-profile entrepreneurs in

13
Social Network Analysis and Mining (2024) 14:131 Page 15 of 17 131

shaping industry trends and behaviors in a rapidly evolving the limitations of our approach. One potential limitation is
digital marketplace. the use of the North American liberal-conservative scale
as a universal yardstick for assessing political ideology.
Although this scale is widely recognized and allows for
6 Limitations and Future Research comparability across contexts, it may not fully capture the
implications nuances and complexities of political beliefs and attitudes in
different local settings. Political ideology is a multidimen-
Our study focused on individuals who self-identified as sional construct that can vary significantly across countries
entrepreneurs on Twitter, analyzing a substantial dataset and cultures. By relying on a single, US-centric scale, we
comprising 4 million tweets from 36,828 entrepreneurs may be overlooking important variations in how entrepre-
across seven English-speaking countries (US, Australia, neurs align themselves politically in different contexts.
New Zealand, UK, Canada, South Africa, and Ireland) over Additionally, the liberal-conservative dichotomy may not
a period of 56 weeks. The primary objective was to capture adequately represent the full spectrum of political views,
broad trends and behaviors within the entrepreneurial com- particularly in countries with multi-party systems or where
munity in response to Elon Musk’s acquisition of Twitter. political identities are shaped by factors beyond the tradi-
While this broad approach allowed us to identify gen- tional left-right divide.
eral patterns and the overall impact of Musk’s social media Another limitation of our study is the reliance on network-
activities on a diverse group of entrepreneurs, it also pre- based measures of political ideology. While our approach of
sented some limitations. Notably, our data collection did not constructing a bipartite network of entrepreneurs and their
include detailed industry-specific information. Categorizing co-followed opinion leaders provides a useful proxy for
entrepreneurs by specific industries could have provided political alignment, it may not fully capture the complexity
more granular insights, particularly as certain subgroups, of entrepreneurs’ political beliefs and attitudes. The opinion
such as technology entrepreneurs, might be more directly leaders that entrepreneurs follow on Twitter may represent
influenced by Musk’s actions and statements. The absence only a partial or superficial reflection of their true political
of such categorization may have diluted potential significant identities.
effects among these subgroups, overshadowing the nuanced To address these limitations and extend our findings,
behaviors of technology entrepreneurs who are more aligned future research could explore several promising directions.
with Musk’s industry and interests. To address this limita- First, studies could employ more context-specific measures
tion and enhance the depth of future research, we recom- of political ideology that are tailored to the local political
mend future studies to obtain more granular data that allows landscape of the entrepreneurs being studied. This could
for industry-specific analyses. Such detailed data would involve using multi-dimensional scales that capture various
enable a deeper understanding of how specific groups of aspects of political beliefs, such as economic, social, and
entrepreneurs respond to influential figures like Elon Musk, cultural dimensions, or developing entirely new measures
providing more targeted and actionable insights. Despite that are grounded in the specific political realities of differ-
this limitation, our findings still offer valuable insights into ent countries or regions. Second, future research could com-
the general influence of high-profile figures on entrepre- bine network-based approaches with other data sources to
neurial behavior. By analyzing a large and diverse sample, gain a more comprehensive understanding of entrepreneurs’
we captured overarching trends and provided a comprehen- political ideology. For example, studies could analyze the
sive view of the “Musk Effect” across different industries. content of entrepreneurs’ social media posts or public state-
This broad perspective is essential for understanding the ments to infer their political views more directly. Third,
generalizability of Musk’s influence and lays the foundation future studies could explore how the relationship between
for more targeted future studies. entrepreneurs’ political ideology and their social media
Additionally, longitudinal studies could provide deeper behavior varies across different contexts and over time. This
insights into how the influence of high-profile figures like could involve conducting comparative analyses of entrepre-
Musk evolves over time and affects long-term entrepreneur- neurs in different countries or regions or tracking changes
ial strategies. Qualitative research methods, such as inter- in entrepreneurs’ political alignment and social media
views and case studies, could also offer deeper insights into activity longitudinally. Such research could provide valu-
the motivations and perceptions of entrepreneurs regarding able insights into how the dynamics of political ideology
influential figures like Musk. and social influence play out in different settings and how
While our study provides valuable insights into the they evolve in response to changing political and economic
relationship between entrepreneurs’ political ideology and conditions. By addressing these methodological limitations
their social media behavior, it is important to acknowledge and pursuing these future research directions, scholars can

13
 131 Page 16 of 17 Social Network Analysis and Mining (2024) 14:131

Ante L (2023) How Elon musk’s Twitter activity moves cryptocur-

build on our findings and develop a more nuanced and com- rency markets. Technol Forecast Soc Change 186:122112
prehensive understanding of how entrepreneurs’ political Bardon T, Brown AD, Puyou FR (2021) Citius, Altius, Fortius: man-
ideology shapes their social media behavior and broader agers quest for heroic leader identities. Organ 30. https://doi.
influence in the entrepreneurial ecosystem. org/10.1177/13505084211030644
Blondel VD, Guillaume JL, Lambiotte R, Lefebvre E (2008) Fast
unfolding of communities in large networks. J Stat Mech Theory
Author contributions Dr. DZ was instrumental in the empirical aspect Exp 2008
of the research, taking charge of the collection and management of Broockman DE, Ferenstein G, Malhotra N (2019) Predispositions
the longitudinal data that forms the backbone of our analysis. This in- and the political behavior of American economic elites: evidence
volved designing the data collection methodology, ensuring the data’s from technology entrepreneurs. Am J Polit Sci 63:212–233
integrity and reliability, and employing sophisticated analytical tech- Brown R (1995) Prejudice: its social psychology. Blackwell, Oxford
niques to interpret the data. Dr. DZ’s expertise was pivotal in the quan- Bruns A, Moe H (2014) Structural layers of communication on Twitter.
titative analysis of the study, facilitating the extraction of actionable in- In: Weller K, Bruns A, Burgess J, Mahrt M, Puschmann C (eds)
sights from complex datasets.Dr. AS provided substantial intellectual Twitter and society [Digital formations, volume 89]. Peter Lang,
input in situating the study within the broader research landscape. Dr. New York, pp 15–28
AS spearheaded the literature review process, identifying key theo- Bruns A, Burgess J (2012) Researching news discussion on twitter:
retical frameworks and previous empirical studies that align with and new methodologies. Journal Stud 13(5–6):801–814. https://doi.
support our research objectives. This comprehensive literature review org/10.1080/1461670X.2012.664428
set the stage for our study’s contributions, grounding it firmly in the Burgess J, Baym NK (2022) Twitter: a biography. NYU, New York
existing academic discourse. Dr. PB contributed significantly to the Conger K, Hirsch L (2021) Twitter’s Jack Dorsey Steps Down From
manuscript by articulating the paper’s positioning in relation to the C.E.O. Role. The New York Times. https://www.nytimes.
current knowledge base and theoretical understanding. Dr. PB meticu- com/2022/10/27/technology/elon-musk-twitter-deal-complete.
lously crafted the interpretations of our findings, drawing connections html. Accessed 8 March 2024
between our data and established theories, and engaging with the lit- Conover M, Ratkiewicz J, Francisco M, Gonçalves B, Menczer F,
erature to highlight the study’s novel contributions. Furthermore, Dr. Flammini A (2011) Political polarization on Twitter. In: Proceed-
PB led the discussion sections, synthesizing our research findings with ings of the international AAAI conference on web and social
broader implications and suggesting directions for future research by media. Vol 5, pp 89–96
highlighting the limitations of our study. Together, Dr. AS and Dr. PB’s Feiner L (2022) Elon Musk notifies Twitter he is terminating deal.
collaborative efforts in framing the paper’s intellectual narrative were CNBC. https://www.cnbc.com/2022/07/08/elon-musk-notifies-
crucial in clarifying the study’s significance and ensuring a coherent twitter-he-is-terminating-deal.html. Accessed 8 March 2024
and persuasive argument throughout the manuscript.All authors have Flake E (2022) Elon Musk’s battle with Twitter shows that even the
participated in drafting the manuscript and revising it critically for im- brashest, loudest leaders can face a reckoning. Business Insider.
portant intellectual content. https://www.businessinsider.com/elon-musk-scandal-contro-
versy-leadership-tesla-spacex-ceo-2022-8. Accessed 4 December
Data availability The dataset utilized in this study consists of a mas- 2023
sive collection of tweets obtained through Twitter’s Academic account Hanik K, Bendell B, Sullivan D, Zinoviev D (2023) How Gender
API following all of Twitter’s data use policies. The data includes pub- Shapes Role-Based Entrepreneurial Self-Presentation on Twitter.
lic tweets and related metadata. Due to privacy, ethical considerations, Working Paper
and Twitter’s terms of service, the raw tweet data cannot be shared Hosmer DW Jr, Lemeshow S, Sturdivant RX (2013) Applied logistic
directly. However, tweet IDs can be made available upon reasonable regression. Wiley, Hoboken
request to the corresponding author, subject to Twitter’s policies. Huddy L (2001) From social to political identity: a critical examination
Users can use these tweet IDs to download the tweet data for their use of social identity theory. Political Psychol 22:127–157
directly from Twitter, in accordance with Twitter’s terms of service. Hughes A, Wojcik S (2019) Key takeaways from our new study of
Please note that some tweets may no longer be available if they have how Americans use Twitter. https://policycommons.net/arti-
been deleted or the user’s account has been deactivated or made pri- facts/616845/key-takeaways-from-our-new-study-of-how-ameri-
vate after the data was collected. cans-use-twitter/1597555/. Accessed 11 Dec 2023
Huyhn TLD (2023) When Elon Muck changes his tone, does Bitcoin
Declarations adjust its tune? Comput Econ 62:639–661
Isaacson W (2023) Elon Musk. Fayard, London
Jin X, Han JW (2010) K-means clustering. In: Sammut C, Webb GI
Competing interests The authors declare no competing interests.
(eds) Encyclopedia of machine learning. Springer, Boston, pp
563–564
Jost JT, Krochik M, Gaucher D, Hennes EP (2009) Can a psychologi-
References cal theory of ideological differences explain contextual variability
in the contents of political attitudes? Psychol Inq 20:183–188
Kelman HC (1953) Attitude change as a function of response restric-
Anderson C, Musk E (2017) Elon Musk the future were building - and
tion. Hum Relat 6:185–214
boring. Ted Conferences. https://www.ted.com/talks/elon_musk_
Li CY (2013) Persuasive messages on information system acceptance:
the_future_we_re_building_and_boring? Accessed 14 February
a theoretical extension of elaboration likelihood model and social
2024
influence theory. Comput Hum Behav 29:264–275
Andrienko G, Andrienko N, Rinzivillo S, Nanni M, Pedreschi D, Gian-
Luri I, Kaliyamurthy AK, Farmer M (2023) Sometime in the future -
notti F (2009) Interactive visual clustering of large collections of
the technology entrepreneur as utopian market hero. Mark Theory
trajectories. In: 2009 IEEE Symposium on Visual Analytics Sci-
23:99–118
ence and Technology. IEEE, pp 3–10
Maclaran P, Brown S (2005) The center cannot hold: consuming the
utopian marketplace. J Consum Res 32:311–323

13
Social Network Analysis and Mining (2024) 14:131 Page 17 of 17 131

Mosco V (2005) The Digital Sublime: myth, power, and Cyberspace. Tajfel H, Turner JC (1986) The Social Identity Theory of Intergroup
MIT Press, Cambridge, MA Behavior. In: Worchel S, Austin WG (eds) Psychology of Inter-
Murray M (2017) The work of art in the age of the superhero. Sci Fic- group Relation. Hall, Chicago, pp 7–24
tion Film Telev 10(1):27–51 Tufekci Z (2014) Big questions for social media big data: Representa-
Neal Z (2014) The backbone of bipartite projections: inferring rela- tiveness, validity and other methodological pitfalls. In: Proceed-
tionships from co-authorship, co-sponsorship, co-attendance and ings of the international AAAI conference on web and social
other co-behaviors. Soc Networks 39:84–97 media. Vol. 8, pp 505–514
Newman ME (2006) Modularity and community structure in net- Wile R (2022) A timeline of Elon Musk’s takeover of Twitter. NBC
works. Proc Natl Acad Sci USA 103:8577–8582 News. https://www.nbcnews.com/business/business-news/
Niburski K, Niburski O (2023) Effects of Elon Musk’s tweet about twitter-elon-musk-timeline-what-happened-so-far-rcna57532.
psychotropic drugs on the internet, media, and purchasing: Accessed 4 Dec 2023
Observational Study. J Consum Health Internet 27:12 Yngfalk AF, Yngfalk C (2020) Modifying markets: consumerism
Paris M, Hull D, Adler M (2022) Elon Musk’s takeover of Twit- and institutional work in nonprofit marketing. Mark Theory
ter: Everything you need to know. Yahoo! Finance. https:// 20:343–362
finance.yahoo.com/news/elon-musk-takeover-twitter-every- Zinoviev D (2018) Complex network analysis in Python: recognize-
thing-193231867.html. Accessed 4 Dec 2023 construct-visualize-analyze interpret. Complex Netw Anal
Rudeloff C, Damms J (2022) Entrepreneurs as influencers: the impact Python, 1–200
of parasocial interactions on communication outcomes. J Res
Mark Entrep. https://doi.org/10.1108/JRME-04-2022-0052 Publisher’s Note Springer Nature remains neutral with regard to juris-
Shamir R (2008) The age of responsibilization: on market-embedded dictional claims in published maps and institutional affiliations.
morality. Econ Soc 37:1–19
Shead S (2021) Elon Musk’s tweets are moving markets — and some Springer Nature or its licensor (e.g. a society or other partner) holds
investors are worried. CNBC. https://www.cnbc.com/2021/01/29/ exclusive rights to this article under a publishing agreement with the
elon-musks-tweets-are-moving-markets.html. Accessed 4 Dec author(s) or other rightsholder(s); author self-archiving of the accepted
2023 manuscript version of this article is solely governed by the terms of
Sørensen BM (2008) Behold, I am making all things new: the entrepre- such publishing agreement and applicable law.
neur as savior in the age of creativity. Scand J Manag 24:85–93

13