Skip to main content
Disciplines
Automotive Business IT + Informatics Construction + Real Estate Electrical Engineering + Electronics Energy + Sustainability Insurance + Risk Finance + Banking Management + Leadership Marketing + Sales Mechanical Engineering + Materials
Events
DE
EN
Books
Journals
Start single access now
Request access for companies
EXTENDED SEARCH
Log in
Top
Review of Quantitative Finance and Accounting

Hint

Swipe to navigate through the articles of this issue

16-09-2023 | Original Research

The impact of social media on venture capital financing: evidence from Twitter interactions

Authors: Onur Bayar, Emre Kesici

Published in: Review of Quantitative Finance and Accounting

Log in

Abstract

This paper examines how information acquisition through social media affects venture capital (VC) investments into entrepreneurial startup firms. We collect a unique data set from Twitter API to measure the impact of owned social media (OSM) and earned social media (ESM) of portfolio companies on the structure of VC investments they receive. We find evidence consistent with the hypothesis that startup firms’ social media engagement affects the staging of VC financing, the VC syndicate structure, and the probability of a successful exit. When a portfolio company’s social media accounts are more active and the company has a higher engagement volume with its followers, VC firms reduce the extent of stage financing and are less likely to syndicate with each other in financing such a portfolio company. Overall, our results demonstrate that entrepreneurial firms with higher OSM and ESM engagement volume have fewer VC financing rounds, a smaller number of VCs in their VC syndicates, a lower probability of VC syndication, a higher successful exit probability, and a higher amount total funding across all rounds.

Dont have a licence yet? Then find out more about our products and how to get one now:

Springer Professional "Wirtschaft+Technik"

Online-Abonnement

Mit Springer Professional "Wirtschaft+Technik" erhalten Sie Zugriff auf:

  • über 102.000 Bücher
  • über 537 Zeitschriften

aus folgenden Fachgebieten:

  • Automobil + Motoren
  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Elektrotechnik + Elektronik
  • Energie + Nachhaltigkeit
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Maschinenbau + Werkstoffe
  • Versicherung + Risiko

Jetzt Wissensvorsprung sichern!

inform now

Springer Professional "Wirtschaft"

Online-Abonnement

Mit Springer Professional "Wirtschaft" erhalten Sie Zugriff auf:

  • über 67.000 Bücher
  • über 340 Zeitschriften

aus folgenden Fachgebieten:

  • Bauwesen + Immobilien
  • Business IT + Informatik
  • Finance + Banking
  • Management + Führung
  • Marketing + Vertrieb
  • Versicherung + Risiko




Jetzt Wissensvorsprung sichern!

inform now
Appendix
Available only for authorised users
Footnotes
1
See also the empirical studies by Tumasjan et al. (2021) and Que and Zhang (2021), which focus on the effect of social media on the valuation of venture capital backed firms.
 
2
Colicev et al. (2018) state that firm-generated contents on social media are termed as “owned social media” (OSM). In addition to OSM, users can generate contents (e.g., retweets, likes, mentions) to promote their recommendations, propagates their brand awareness, purchase intent, and customer satisfaction. Such user-generated contents are commonly termed as “earned social media” (ESM). See also Stephen and Galak (2012).
 
3
Tumasjan et al. (2021) argue that Twitter sentiment as a weak signal does not correlate with actual long-term investment success, whereas patents (which are considered a stronger signal of intrinsic company value) do.
 
4
Baker and Wurgler (2006), Baker, Wurgler, and Yuan (2012), and Huang et al. (2015) provide evidence of return predictability using stock market-based investor sentiment measures. Baker and Wurgler (2006) suggest that stocks that are difficult to value and costly to arbitrage are more sensitive to manager sentiment-driven mispricing. Tetlock (2007) quantitatively measures the nature of the media’s interactions with the stock market using daily content from a popular Wall Street Journal column. He finds that high media pessimism predicts downward pressure on market prices followed by a reversion to fundamentals, and unusually high or low pessimism predicts high market trading volume. Loughran and McDonald (2011) develop a negative word list used in a financial context and find a positive relation between firm-level manager sentiment and four-day event period return in the cross-section. Similarly, Jiang et al. (2019) construct a new manager sentiment index and find that manager sentiment negatively predicts cross-sectional stock returns. See also Tumasjan et al. (2021) who analyze the effect of Twitter sentiment on venture valuations.
 
5
We use the Harvard-IV-4 category of the Harvard Psychological Dictionary to construct sentiment measures as perceived by human readers by counting “power” words. Power sentiment is measured using the “strong words” category of Harvard psychosocial dictionary (Stone 2002) indicating power, control, authority, and success (e.g., superior, accomplishment, amazing, award, astound, etc.).
 
6
Wang (2016) analyzes how social networks influence both a start-up’s likelihood of being screened for evaluation and its likelihood of being funded. Bernstein, Korteweg, and Laws (2017) conduct a randomized field experiment on AngelList and find that information about human assets (e.g., founding team) is causally important for the funding of early-stage start-up firms. Gompers et al. (2020) find that VCs see the management team as somewhat more important than business-related characteristics. Similarly, Cao (2019) investigates frictions in the process of VC information acquisition and uses “Product Hunt” website, which ranks popular new products using VCs and developers’ votes. She finds a significant and positive association between product rankings and early-stage funding.
 
7
While the sample of Tumasjan et al. (2021) consists of technology-based VC-backed startups only, our sample consists of all VC-backed startups which received their first round of financing between 2010 and 2018. Different from Tumasjan et al. (2021), who consider the Twitter sentiment of portfolio company tweets preceding all rounds of VC financing, we focus on the social media content prior to the first round of VC financing only to mitigate reverse causality concerns in our empirical analysis. Finally, while Tumasjan et al. (2021) measure the sentiment of portfolio company tweets only, our paper measures the effect of both the social media content of company tweets (owned social media) and the reactions of Twitter users to these tweets (earned social media).
 
8
Cornelli et al. (2006) examine retail investor sentiment in European pre-IPO markets. They find that high pre-IPO markets driven by the overoptimism of retail investors are a good predictor of first-day aftermarket prices.
 
9
Certo (2003) and Williams et al. (2010) argue that startup firms signal their specific characteristics of top management team to mitigate information asymmetry and investor uncertainty during their IPO process and seek to gain legitimacy. See also Vismara (2016) and Tumasjan et al. (2021).
 
10
More recently, Bayar, Chemmanur, and Tian (2019) study a new rationale for VC syndication: peer monitoring among VCs. They show that syndicates mitigate VCs’ moral hazard problem in value addition. Additionally, the second-opinion hypothesis (Lerner 1994; Tian 2012) conveys the argument that VCs collect information from their peers in the syndicate to resolve asymmetric information and decide whether to invest in risky startups.
 
11
Lerner (1994) provides evidence consistent with the view that syndication enables VCs to collect information about potential deals, mitigating asymmetric information between entrepreneurs and VCs when VC firms choose their investments in portfolio companies. Tian (2012) shows that VC syndication can create product market value for startups because it facilitates nurturing portfolio companies’ innovation capacity and increasing post-IPO operating performance.
 
12
Tian (2011) notes that stage financing may be costly for three reasons: (1) every stage incurs negotiation and contracting costs before capital infusion to venture capital, (2) staging could induce entrepreneurs to aim short-term performance rather than long-term success (termed as “window dressing”), (3) staging may also lead to an underinvestment problem.
 
13
Bernstein et al. (2016) empirically analyze the introduction of new airline routes that reduce the travel time between VCs and their portfolio companies to address similar concerns about asymmetric information between VCs and portfolio companies and the cost of monitoring. Their results are consistent with the idea that if the cost of monitoring is smaller, the number of VC financing rounds (stages) should decrease.
 
14
Tetlock (2007), Tetlock, Saar-Tsechanksy, and Macskassy (2008), and Hanley and Hoberg (2010) pioneered the application of psychological dictionaries in measuring the sentiment of financial texts. Loughran and Macdonald (2011) developed capital-market specific dictionaries which have since been applied to large-scale computation of tones and sentiment in financial texts, e.g., Dow Jones newswires (Da, Engelberg, and Gao 2011), New York Times financial articles (Garcia, 2013), 10-K and IPO prospectuses (Jegadeesh and Wu 2013), corporate press releases (Ahern and Sosyura 2014), earnings conference calls (Jiang, Lee, Martin, and Zhou 2019), and all wired news from Factiva (Huang et al. 2020). See also the survey article Loughran and McDonald (2016). Recent studies analyzing downloads of SEC filings include Bernard et al. (2020), Cao, Du, Yang, and Zhang (2020), Chen, Cohen, Gurun, Lou, and Malloy (2020), and Crane, Crotty, and Umar (2020).
 
15
Cao et al. (2020) analyze how corporate disclosure has been reshaped by machine processors, employed by algorithmic traders, robot investment advisors, and quantitative analysts. They find that increasing machine and AI readership, proxied by machine downloads, motivates firms to prepare filings that are friendlier to machine parsing and processing. Furthermore, firms with high expected machine downloads manage textual sentiment and audio emotion by differentially avoiding words that are perceived as negative by computational algorithms as compared to those by human readers, and by exhibiting speech emotion favored by machine learning software processors.
 
16
The use of computational textual analysis has recently been extended to non-text data such as the audio content of conference calls (Mayew and Venkatachalam 2012), and the video content of startup pitch presentations (Hu and Ma, 2021).
 
17
Tian (2011) notes that VC funds are required to be liquidated within ten years from the inception. Hence, we assume that startup firms are written off if they fail to receive investments within ten years after the last round.
 
18
If there are fewer than four public firms in the four-digit SIC industry, we use three-digit SIC industry codes, and if there are fewer than three public firms in the three-digit SIC industry, we use two-digit SIC industry codes.
 
19
The relevant information is accessible from this source: “How Twitter is fighting spam and malicious automation” retrieved from https://​blog.​twitter.​com/​en_​us/​topics/​company/​2018/​how-twitter-is-fighting-spam-and-malicious-automation.
 
21
In a large sample of 10-Ks during 1994 to 2008, Loughran and McDonald (2011) argue that almost three-fourths of the words identified as negative by the Harvard Dictionary are words typically not considered negative in financial contexts. Therefore, they develop an alternative negative word list, along with five other word lists, that better reflect tone in financial text. Based on prior theoretical and empirical research from psychology and linguistics on deception detection, Larcker and Zakolyukina (2012) select the word categories that theoretically should be able to detect deceptive behavior by executives. They use these linguistic features to develop classification models for a large sample of quarterly conference call transcripts.
 
22
The theoretical model of Merton (1987) model has been extended by Van Nieuwerburgh and Veldkamp (2009), who assume that such attention/information acquisition has a cost.
 
23
Bollen, Mao, and Zheng (2011) argue that Twitter mood predicts the stock market. Tirunillai and Trellis (2012) examine whether user-generated content (UGC) is related to stock market performance and conclude that idiosyncratic risk increases significantly with negative information in UGC. Luo et al. (2013) examine the relationship between social media and firm equity value, and they conclude that social media-based metrics (web blogs and consumer ratings) are significant leading indicators of firm equity value.
 
24
Mumi, Obal, and Yang (2019) investigate social media as a firm's signaling strategy through an IPO and find a positive association between social media presence and firms’ IPO valuations. Bianchi, Kind, and Kung (2019) find some market-based evidence that President Trump influences monetary policy expectations of market participants through his tweets that criticize the conduct of monetary policy.
 
Literature
Agrawal AK, Catalini C, Goldfarb A (2011) The geography of crowdfunding (No. w16820). National bureau of economic research
Cao R (2019) Information Frictions in New Venture Finance: Evidence from Product Hunt Rankings. Working Paper
Cao S, Jiang W, Yang B, Zhang AL (2020) How to talk when a machine is listening? Corporate disclosure in the age of AI (No. w27950). National Bureau of Economic Research
Crane AD, Crotty K, Umar T (2020) Public and private information: complements or substitutes. Available at SSRN, 3127825
Statista (2018) Active Twitter users in the US
Metadata
Title
The impact of social media on venture capital financing: evidence from Twitter interactions
Authors
Onur Bayar
Emre Kesici
Publication date
16-09-2023
Publisher
Springer US
Published in
Review of Quantitative Finance and Accounting
Print ISSN: 0924-865X
Electronic ISSN: 1573-7179
DOI
https://doi.org/10.1007/s11156-023-01199-4