Voluntary information disclosure on social media

Download PDF
4 downloads 21188 Views 1MB Size Report
Comment
Mar 5, 2015 - Companies are increasingly using social media to communicate with ... In April 2013, the SEC stated that companies can use social media to.
Decision Support Systems 73 (2015) 28–36

Contents lists available at ScienceDirect

Decision Support Systems journal homepage: www.elsevier.com/locate/dss

Voluntary information disclosure on social media Juheng Zhang ⁎ Department of Operations and Information Systems, Manning School of Business, University of Massachusetts Lowell, One University Ave., Lowell, MA 01854, USA

a r t i c l e

i n f o

Article history: Received 16 July 2014 Received in revised form 25 February 2015 Accepted 26 February 2015 Available online 5 March 2015 Keywords: Information asymmetry Information disclosure Social media Technology adoption

a b s t r a c t Companies are increasingly using social media to communicate with consumers, and the content of those media affects consumer decision making. We investigate the adoption of new communication media (social media, mobile apps, email alerts, etc.) by companies to disclose company news to consumers and investors. Using a cluster analysis, we group companies either into a cluster with high adoption of new media or into one with low adoption, based on their use of new media for disclosure. We find that a company's voluntary information disclosure on social media is positively related to its adoption level of new media. Our findings suggest that the engagement of information disclosure on new media increases a company's influence and reach. © 2015 Elsevier B.V. All rights reserved.

1. Introduction and motivation In July 2012, Netflix CEO Reed Hastings and Facebook were put on the front line of corporate disclosure. Hastings posted the news on Facebook that the number of monthly online viewing hours on Netflix in June 2012 had exceeded one billion for the first time. Netflix's stock price at the time of this post was $70.45; it went up to $81.72 by the close of the following day. This was the first time that Hastings had disclosed Netflix's monthly viewing hours on his Facebook wall. The Security and Exchange Commission (SEC) investigated whether the posting violated the regulation of fair disclosure (Regulation FD), since the performance information was not released in official filings, press release, or other traditional disclosure outlets. Eventually, the SEC chose not to charge Hastings with any wrongdoings. However, it confirmed that Regulation FD applies to social media and other emerging communication outlets [20]. In April 2013, the SEC stated that companies can use social media to announce material information to investors [29], but companies should first alert investors as to which media they plan to use. The media must also be publicly accessible and nonexclusive. Earlier, in 2008, SEC had ruled that companies can use their websites to disseminate key information. The more recent ruling means that social media are as acceptable as company websites in disseminating information. Companies are increasingly making use of social media to blog corporate news and reach out to consumers [23]. Among Fortune 500 companies, 79% show fresh content on corporate blogs, 77% have an active Twitter account, and 69% maintain a YouTube channel. They produce a variety of social media content, such as blogs and videos, ⁎ Tel.: +1 978 934 3261. E-mail address: [email protected].

http://dx.doi.org/10.1016/j.dss.2015.02.018 0167-9236/© 2015 Elsevier B.V. All rights reserved.

and use multiple social-network sites to communicate with consumers. In addition, companies are also seeking social media to disclose sensitive information [29]. For example, eBay, PepsiCo, and Dell use Twitter to announce their earnings and profit numbers. Surveys show social media to be the most powerful outlet of information [12]; it is through social media that people get updated with the latest corporate news, market trends, investment information, and so on. Even though more companies are embracing social media, some are cautious about using new media to disclose financial information. Many are worried about the uncertainty of social media and are reluctant to adopt new media for information disclosure. In this study, we examine the factors affecting both a company's overall information disclosure and its financial information disclosure on social media. We select publically traded companies and collect the data of their new media adoption. Using a cluster analysis, we group those companies into two clusters based on their adoption of new communication media. We investigate whether voluntary information disclosure on social media is affected by the level of new media adoption or other factors such as information environment, information asymmetry, or firm profitability. To the best of our knowledge, this is the first study of information disclosure and social media adoption. We summarize the contributions of this paper below: (1) We investigate information disclosure on social media from the perspective of companies rather than consumers. Information Systems (IS) literature focuses on User Generated Content (UGC) and examines the value of social media content provided by users. However, here we consider how companies disclose information on social media, i.e., Enterprise Generated Content (EGC), which differs from the existing social media literature in the IS field.

J. Zhang / Decision Support Systems 73 (2015) 28–36

(2) We explore the cutting-edge communication media for information disclosure. Researchers of voluntary information disclosure widely study traditional media, such as official SEC reports or government filings. New technologies are increasingly made available for the use of information release; for instance, RSS (Rich Site Summary), email alerts, mobile apps, social media platforms, etc. However, until now, no one had yet investigated new communication media for voluntary information disclosure. (3) We cluster companies into two groups based on their adoption of new media (social media, mobile apps, RSS, email alerts, etc.) for disclosure. Using the North American Industry Classification System (NAICS), we randomly selected companies from the information industry and the manufacturing industry. We find that companies in the information industry do not lead in the adoption of new media for disclosure or in the amount of news released through social media. Our analysis clusters companies into either the group with a high adoption level or one with a low adoption level, regardless of their NAICS industry classification. We find that the companies in the high adoption group embrace a larger audience on social media and also lead in the intensity of information disclosure on social media. (4) We show that the level of new media adoption—as well as such other factors as the information environment, information asymmetry, and firm profitability—determines voluntary information disclosure on social media. We also find that a company's adoption level affects its intensity of information disclosure on social media. In addition, after a new medium becomes widely adopted and more popular for disclosure, voluntary information disclosure on that medium then follows the pattern of voluntary information disclosure in traditional filings and reports. Voluntary information disclosure on the medium can be explained by the determinants of voluntary information disclosure on traditional media (i.e., SEC filings or reports). In the next section, we review IS studies on social media content and corporate finance. We focus on the issue of voluntary information disclosure in the finance field and review the traditional determinants of voluntary information disclosure. In Section 3, we describe data collection and new communication media for information dissemination, and provide basic statistics from collected data. In Section 4, we perform a cluster analysis to group the selected companies. In Section 5, we compare the intensity of information disclosure on social media between the two clusters and investigate the determinants of voluntary information disclosure on social media. We discuss the managerial implications of our study in Section 6. We present our conclusion and note future work in Section 7. 2. Literature review Researchers in the IS field who study social media mainly work on the information value of UGC [e.g., 4,8,21,22,32], information quality

29

of UGC [e.g., 4], or information sharing on social networks [e.g., 30,32]. Most related to our work are the studies [22,33] that examine the relationship between social media content and firm stock performance. Yu et al. [33] focus on sentiment analysis and show that different social media (blogs, forums, Twitter) have different impacts on firm financial performance, and the overall media have a stronger relationship to that performance than traditional media (television, newspapers, magazines). Luo et al. [22] study whether social media can be used to predict firm equity value. They suggest that social media have stronger and faster leading prediction power of firm equity value than traditional online behavioral metrics, including Google Search and web traffic. They find that social media are a significant leading indicator of firm equity value. The study examines the value of social media in finance and justifies the investment in social media and new technology initiatives. DiStaso and Bortree [7] study the value of social media in the improvement of company transparency. The IS studies of social media differ from our study in several ways. First, they examine such UGC as product ratings or web logs written by individual consumers, while our study investigates EGC, the information that companies release through social media. Second, these social media studies analyze the value of UGC as an influence factor on product sales or a firm's equity value. Seldom have they addressed the motivations of companies to disclose information on social media. Our work examines which factors affect voluntary information disclosure on social media. Another branch of related studies is voluntary information disclosure in the finance field. These studies identify such determinants of voluntary information disclosure as firm size [e.g. 5,6,9,10,24–26], information asymmetry [17–19], firm profitability [24,25,28], debt structure [6,25,26], or information environment [27]. Large firms generally have a better information environment and are more transparent. As for the proxy of firm size, some studies [e.g., 26] use the logarithm of assets, while others may use the number of employees [5] or sales turnover [25]. In this work, we follow the existing study [26] and use the logarithm of assets as the proxy of firm size. Regarding information asymmetry, one study [1] shows that when a company has more intangible assets, it is considered to have higher uncertainty in the future; the degree of information asymmetry is associated with the intangible assets. Therefore, we use intangible assets as the proxy of information asymmetry. Firm performance is highly correlated with EBITDA (Earnings Before Interest, Taxes, Depreciation and Amortization) or EBIT (Earnings Before Interest, Taxes) as a percent of sales or total assets [24,25]. EBITDA is a more conservative earning measurement than EBIT, so we use EBITDA for firm profitability. The debt structure is commonly measured by the debt in relation to assets [e.g., 6]. We use debt as the proxy of debt structure. The existing literature of voluntary information disclosure mainly focuses on traditional media such as reports or government filings. Only a few studies investigate voluntary information disclosure on new technologies, and those studies are limited to company websites [e.g., 2,9,10,24]. Little attention has been devoted to voluntary information disclosure using new communication paradigms including social media, RSS, mobile apps, or other media.

Table 1 Summary of technology adoption. Industry NAICS #51 sector NAICS #31–33 Overall

1 0 −1 1 0 −1 1 0 −1

Twitter

YouTube

Facebook

LinkedIn

Pinterest

Blogs

Mobile apps

RSS

Web casting

Conf. calls

Email alerts

9% 75% 16% 38% 20% 42% 24% 46% 30%

2% 69% 29% 3% 60% 37% 3% 63% 34%

0% 95% 5% 7% 52% 42% 3% 72% 24%

17% 63% 20% 7% 91% 2% 12% 77% 11%

0% 10% 90% 2% 5% 93% 1% 8% 91%

5% 22% 73% 5% 67% 27% 5% 43% 51%

7% 33% 60% 24% 51% 25% 15% 42% 43%

77% 0% 23% 84% 4% 13% 80% 2% 18%

95% 0% 5% 96% 2% 2% 96% 1% 3%

95% 0% 5% 96% 4% 0% 96% 2% 3%

57% 0% 43% 100% 0% 0% 77% 0% 23%

30

J. Zhang / Decision Support Systems 73 (2015) 28–36

Table 2 Summary of intensity of information disclosure. Twitter Total tweets Max 18,318 Min 0 Mean 2053 Std. dev. 2976 Mean-#51 2958.1 Mean-#31/33 1171.5

Table 4 Summary of web traffic.

YouTube

Corporate websites

Finance tweets

Total videos

Finance videos

122 0 7.6 18.7 4.2 11.0

16,223 13 0 0 216.0 0.2 1526.0 1.4 363.9 0.1 73.5 0.3

Total news

Finance news

1880 15 186.9 209.9 176.1 204.2

496 0 57.6 58.9 49.9 67.8

This work differentiates itself from existing studies of voluntary information disclosure in two aspects. We focus on new communication media; existing works study traditional government filings or corporate website updates. In addition, we explore whether a company's technology adoption level affects its intensity of information disclosure on social media. Technology adoption or investment has not been studied in the literature of voluntary information disclosure. Technology adoption for disclosure shows the engagement of a company in information disclosure; it can be considered as a factor of voluntary information disclosure. 3. Data and description statistics 3.1. Data collection We chose companies in two industries based on NAICS, which is largely used by government and business to classify industries with its six-digit code. The first two digits are the most general business sector. We used business sector #51 (the information industry) and #31–33 (the manufacturing industry) to study the adoption of new media for information disclosure. The information industry includes software companies and other media communication companies; the manufacturing industry comprises companies that manufacture products, parts, or materials. (This survey [11] shows that manufacturing companies employ YouTube to disseminate ideas and topics relating to their products, or use LinkedIn as their social channel.) We randomly selected 120 companies in these two industries—60 in each industry—and downloaded their financial information for fiscal year 2013 from the Compustat database. The key financial information we used includes total assets, EBITDA, intangible asset, and liability, as suggested in the literature of voluntary information disclosure [2,5,6,9,10,13,15,19,24,25,28]. For communication media, we chose web blogs, mobile apps, RSS, email alerts, company websites, webcasting, conference calls, and social media platforms such as Twitter, Facebook, YouTube, and LinkedIn. The chosen social media platforms are especially popular: Twitter has one billion registered users and Facebook has close to that number. LinkedIn is a professional social network that companies often use to release news. Other communication media, such as mobile apps, have recently become popular and some companies provide them for investors to access their company news.

Table 3 Summary of medium reach. Twitter

Max Min Mean Std. dev. Mean-#51 Mean-#31/33

YouTube

Facebook

Followers

Days

Subscribers

Days

Likes

Days

214,000 0 10,055 28,873 6123.7 13,358.4

2019 0 789.1 587.1 692.4 430.5

2,723,191 0 29,490 261,859 54,771.3 8321.0

2884 0 1694 774.8 1232.2 718.1

82,621,473 0 1,018,517 7,995,408 347,280.6 1,583,282.4

2699 21 1410 569.1 1229.6 783.6

Max Min Mean Std. dev. Mean-#51 Mean-#31/33

Global rank

Bounce rate

Daily page views

Daily time

12,945,293 1396 1,024,659 2,076,150 420,944.72 1,593,916.07

0.7 0.12 0.46 0.12 0.44 0.48

11 1 2.81 1.35 3.1 2.52

17.4 1.01 2.91 2.18 3.45 2.34

For each company, as described below, we collected two main types of data: (1) a company's adoption of a communication medium for information disclosure, and (2) the intensity of its information disclosure on the medium. We coded the adoption of each communication medium in three values: −1, 0, 1. The coded value is “1” if a company adopted the communication tool and uses it for financial disclosure; “0” if the company adopted the tool for releasing information but not for financial news; and “−1” if the company has not adopted the tool. For the intensity of information disclosure, we collected data on the quantity of news (financial news and news of all kinds) released on each medium, if the medium was adopted for the disclosure of financial information. Such data measure the intensity of information disclosure, which we collected for three media only—Twitter, YouTube, and company websites—and used such keywords as “profit,” “review,” or “$” to separate financial messages from general ones. In addition, we collected basic information from these social media platforms, including a company's reach on each one, since audience size can be a proxy for the reach or impact of information disclosure through a medium. We downloaded the number of each company's Twitter followers, YouTube subscribers, and Facebook “likes,” as well as the date on which each adopted a medium (the date it joined Facebook; the date of its first Facebook post; and so on). We downloaded such website-traffic data as website global rank, bounce rate, and daily page views from Alexa.com. “Website global rank” indicates website popularity, where a higher rank means greater popularity (smaller numbers indicate higher ranks). Using data on how many visitors view only one page on a website, “bounce rate” measures how often users bounce from one website to another; the smaller the value, the fewer users who switch to other websites after a short visit. “Daily page views” shows the average number of pages on a website viewed by each unique visitor; and “daily time” is the average amount of time that each visitor spends on a website. These metrics measure how engaged visitors are. 3.2. Descriptive statistics For the data of media adoption, we include the basic summary in Table 1. Table 1 lists the percentage of companies that have adopted the chosen communication media for information disclosure. These media are all included in the table except for corporate websites that are adopted entirely for financial information disclosure. As shown in Table 1, companies have widely adopted traditional tools (including corporate websites, webcasting, and conference calls) to disclose financial information. They are also using new media, such as RSS, email alerts, mobile apps, and Twitter, for information disclosure. For example, as for Twitter, 24% of companies have adopted it for financial information disclosure; 46% have adopted it for disclosing general information but not financial information; and 30% have not adopted it for any kind of disclosure. Table 1 also includes the adoption percentage for the information industry and the manufacturing industry separately. It shows that the information industry (#51) leads the manufacturing industry on new media adoption for financial information disclosure only as regards LinkedIn, but it dominates manufacturing on disseminating general information on Twitter, YouTube, Facebook, and Pinterest.

J. Zhang / Decision Support Systems 73 (2015) 28–36

31

Table 5 Clusters of new-media adoption. Cluster means Cluster

Twitter

YouTube

Facebook

LinkedIn

Pinterest

Blogs

Mobile apps.

RSS

Web casting

Conf. calls

Email alerts

1 2

0.10 −0.50

−0.19 −0.63

−0.10 −0.53

0.13 −0.37

−0.93 −0.87

−0.35 −0.77

−0.10 −0.77

0.9 −0.2

0.99 0.73

0.99 0.80

0.90 −0.47

Twitter

YouTube

Facebook

LinkedIn

Pinterest

Blogs

Mobile apps.

RSS

Web casting

Conf. calls

Email alerts

0.48 0.49

0.40 0.57

0.41 0.49

0.26 0.43

0.11 0.69

0.11 0.61

Cluster standard deviations Cluster 1 2

0.69 0.68

For intensity of information disclosure, including by industry, we give a basic summary of data in Table 2. It shows that the information industry releases more general messages than the manufacturing industry on Twitter and YouTube but not on corporate websites, and fewer finance-specific messages. We give a descriptive summary of social media reach in Table 3, which includes the social media platforms Twitter, YouTube, and Facebook. From that table we see that the information industry adopted Twitter, Facebook, and YouTube earlier than the manufacturing industry did but has a smaller audience than the manufacturing industry on Twitter and Facebook. In Table 4 we summarize web-traffic data downloaded from Alexa.com. This table includes a company website's global rank; bounce rate, daily page views by visitor, and daily time. As expected, this data show that information-industry websites are more popular than those of manufacturing companies; further, visitors spend more time on information-industry websites. 4. Cluster analysis on new-media adoption The data description and discussion in the previous section show that the companies in the information industry are not leading in the adoption of cutting-edge media for information disclosure. These results are rather mixed. NAICS code can be helpful for separating business sectors but is not informative in terms of new technology adoption.

0.57 0.57

0.71 0.43

0.40 1.00

0.43 0.90

As noted, we cluster the companies into two groups—one with a high level of new-media adoption and one with a low level. We use the data of new-media adoption for clustering analysis; however, to avoid skewing group membership, we omitted the data for intensity of information disclosure. We use the k-means method [31] for clustering analysis. This method finds cluster centers and assigns each data record to the nearest center. It minimizes the total of squared distance from cluster centers: 2 X  X  x −μ 2 ; where μ is the center of the group k: Min i k k k¼1 xi ∈C k

Here, the number of clusters, k, is preset to 2. A data point, xi, is clustered into a group based on its distance from each cluster centroid, μ k. That is, xi ∈ C1 if ‖xi − μ1‖2 b ‖xi − μ2‖2. Otherwise, xi ∈ C2. The results of the cluster analysis are included in Table 5. With the k-means method, each cluster is represented by a single mean vector. As shown in Table 5, the high adoption group (cluster 1) includes the companies that have largely adopted new communication media. Overall, the adoption rates of cluster 1 are higher than those of cluster 2, except for Pinterest's. Pinterest was launched in 2010 and is a new medium, one in which users can share and manage pictures (also called “pins”) in collections known as “pinboards.” It is barely used by the selected companies for disclosure. We compare the membership of a company based on our cluster analysis with its NAICS code (business sector), and we demonstrate our results in Fig. 1. Each data point stands for a company's adoption of new media and is plotted based on its distance from its cluster center. The sampled companies in the information industry are plotted separately from those in the manufacturing industry. As shown in Fig. 1, the application software companies with NAICS subsector code 511210 are classified into the high adoption group; they dominate others in the adoption of new media for information disclosure. Other companies in the information industry and manufacturing industry are roughly scattered into either the high adoption group or the low adoption group. We summarize the frequency of industry subsectors appearing in each cluster. The detailed information of cluster membership for NAICS code is given in Table 6 and the results are consistent as shown Table 6 Frequency of information subsectors in cluster. NAICS Code

Common keywords

Information industry #51 Cluster 1 511210 Applications software, computer, packaged 519130 Advertising periodical publishers, exclusively on Internet 51Others

Fig. 1. Cluster analysis with NAICS classifications.

Manufacturing industry #31–33 31Food, beverage, tobacco, textile, leather production 32Wood product, paper, chemical, plastics production 33Machinery, computer, electronics, furniture production

Frequency

100% 91% 90%

33% 69% 57%

Cluster 2 0% 9% 10%

67% 44% 43%

32

J. Zhang / Decision Support Systems 73 (2015) 28–36

Table 7 Comparison of clusters in the reach of social media. Twitter followers

Facebook likes

YouTube subscribers

Cluster

1

2

1

2

1

2

Max Min Mean Std. dev. 95% CL mean Diff (1–2) pooled mean Folded F method F Value PrN|t| Kruskal–Wallis test Chi-square Pr N Chi-square

214,000 0 13,198.9 33,100.1 28,717.6 11,840.8

15,700 0 1358.1 3881.8 3091.5

82,621,473 0 1,360,834.0 9,234,979.0 8,012,251.0 1,357,293.0

93,000 0 3540.5 16,941.2 13,492.1

2,723,191 0 38,188.5 299,952.0 260,238.0 36,694.9

40,818 0 1493.6 7444.9 5929.2

72.71 b.0001

297,154.00 b.0001

1623.24 b.0001

19.64 b.0001

21.47 b.0001

13.42 0.0002

in Fig. 1. All software companies are clustered into the high adoption group. Other companies are mixed in the adoption of new media.

transparent in information disclosure than companies that do not actively use such tools. We state our hypotheses below. (i) The high adoption cluster has a larger reach on social media than the low adoption cluster. E(mj |C1) ≥ E(mj |C2), where j ∈ {Facebook, Twitter, YouTube}, mj is medium reach. (ii) The high adoption cluster releases more news through social media than the low adoption cluster. E(intensityj |C1) ≥ E(intensityj |C2), where j ∈ {Twitter, YouTube}, intensityj is the amount of information released through the medium j.

5. Voluntary information disclosure on social media We conduct our cluster analysis in Section 4, above, on the data for the adoption and use of a number of new media, including mobile apps, RSS, Twitter, etc. In this section, we focus on social media platforms and further investigate the impact of media adoption on the intensity of information disclosure. We base our information disclosure analysis on the intensity data (e.g., the amount of news released through a medium). We use the pooled t test to compare the means of two clusters. We also conduct the Folded F method [3] to test if the variances of two groups are equal. The Folded F method does not require the specification of which variance is expected to be larger, but it is sensitive to non-normality. The Kruskal– Wallis test [16] is a non-parametric method of testing whether two groups originate from the same distribution. The test does not assume the normal distribution of the residuals. 5.1. Social media use for two clusters We compare two clusters regarding information disclosure on social media. The companies that are clustered into the high adoption group actively invest in and adopt new communication tools. When companies actively use new tools for disclosure, we expect to see that they embrace larger-sized audiences on social media and that they are more

We use the number of Twitter followers, Facebook likes, and YouTube subscribers as the reach of Twitter, Facebook, and YouTube, correspondingly. We examine the difference between the two clusters of the average numbers of followers on each social media platform. The results are included in Table 7. We see that the high adoption group (cluster 1) embraces a larger audience. Table 7 shows that the difference of the reach of social media between two clusters is significant at 99% confidence level, as suggested by both the Folded F method and the Kruskal–Wallis test. The first proposed belief is supported. We further examine the intensity of information disclosure on social media, measured by the number of messages being disseminated through

Table 8 Comparison of clusters in intensity of information disclosure. Website news

Twitter tweets

Total

Finance

YouTube videos

Total

Finance

Total

Finance

Cluster

1

2

1

2

1

2

1

2

1

2

1

2

Max Min Mean Std. dev. 95% CL mean Diff (1–2)

1880 16 215.6 232.4 201.7 108.1

387 15 107.5 92.5 73.6

496 6 59.8 66.8 58.0 8.3

101 0 51.5 27.4 21.8

18,318 0 2573.8 3213.7 2788.2 1962.6

5069 0 611.1 1437.6 1144.9

73 0 8.5 17.3 15.0 3.6

122 0 5.0 22.4 17.9

16,223 0 283.8 1777.5 1542.1 255.3

310 0 28.5 82.1 65.4

13 0 0.3 1.6 1.4 0.29

0 0 0 0

Pooled method t value Pr N |t|

2.47 0.02

0.66 0.51

3.22 0.002

0.89 0.38

0.78 0.43

Folded F method F value Pr N |t|

6.32 b.0001

5.95 b.0001

5.00 b.0001

1.68 0.07

469.18 b.0001

Kruskal–Wallis test Chi-square Pr N Chi-square

13.66 0.0002

0.70 0.40

20.50 b.0001

3.22 0.07

16.98 b.0001

0.96 0.34

Infty b.0001 1.10 0.29

J. Zhang / Decision Support Systems 73 (2015) 28–36

33

Fig. 2. Information disclosure on social media.

such media. We consider both the intensity of finance-specific news and that of overall news. These results are given in Table 8. We observe that the high adoption group (cluster 1) releases more general messages through Twitter and YouTube than the low adoption group (cluster 2). The difference between the two clusters is significant at 90% confidence level, based on the Folded F method and the Kruskal– Wallis test. In terms of financial news, we find no significant difference between the two clusters. We use radar charts to illustrate the intensity of information disclosure on these three media (or dimensions): corporate websites, Twitter, and YouTube. Radar charts are also called “star charts” or “web charts,” and are generally used to display the multivariate data. Each star (shown as a triangle here, in three-dimensional radar charts) represents a data point. The value of a data point on a dimension is represented by the relative line length from the origin. In Fig. 2, the left radar chart is the disclosure of overall information and the right chart is for financespecific disclosure. The star with solid lines represents the high adoption group; the star with dashed lines represents the low adoption group. As shown in the left chart of Fig. 2, the star of the high adoption group has a larger line length on all dimensions than does the one for the low adoption group, and the same is true for financial information disclosure in the right chart. This suggests that the magnitude of information released on any of these three media is larger for the high adoption group than for the low adoption one. Among the three media, Twitter is dominantly used for the disclosure of general news by both groups, but not for financial news. Company websites are more often used for finance-specific messages. 5.2. Voluntary information disclosure Firm size [e.g., 5,6,9,10,24–26], information asymmetry [e.g., 17–19], firm profitability [24,25,28], debt structure [6,25,26], and information environment [27] are the main determinants of voluntary information disclosure, as suggested in the literature of voluntary information

disclosure [e.g., 5,9,10,13,14,19,25,28]. Following the literature, we take the logarithm of asset as the measure of firm size, EBITDA for firm profitability, intangible assets for information asymmetry, and liability for debtleverage ratio. That is, the regression model of information disclosure as stated below: Disclosurei ¼ β0 þ β1  logðAsseti Þ þ β2  EBITDAi þ β3  Intangiblei þ β4  Liabilityi þ εi We first examine these traditional determinants without considering the impact of new-media adoption. The analysis of variance for disclosure intensity, along with the reach of new media, is provided in Table 9. As indicated in Table 9, the explanatory power of those determinants is limited on social media platforms, Twitter, and YouTube. The total number of tweets and of YouTube videos cannot be explained by those determinants. The adjusted R-squares are negative: −0.0002 for tweets and −0.03 for YouTube videos. A negative adjusted R-square means the explanatory factors of the model are not meaningful predictors to the explained variable. The results in Table 9 suggest that the determinants of voluntary information disclosure in traditional media are of limited use in explaining information disclosure on social media. These determinants do not have much explanatory power in the regression model of disclosure on social media. Voluntary information disclosure on social media may be affected by other factors. We now investigate whether information disclosure on social media is affected by the media adoption level. The adoption level of new media shows a company's engagement with, and willingness to invest in, new technologies. In Table 10 we provide analysis of variance results of voluntary information disclosure, controlled for the media adoption level. When the IT adoption is controlled for, the explanatory power of those determinants increases overall. For cluster 1, the adjusted R-square for total website news is increased from 0.48 to 0.56; website financial news goes up from 0.11 to 0.19; total tweets, up from −0.0002 to 0.27;

Table 9 Analysis of variance: voluntary information disclosure. Website

R-sq. Adj. R-sq F value Intercept LogAT EBITDA Liability Intangible

Twitter

YouTube

Audience

Total news

Finance news

Total tweets

Finance tweets

Total videos

Finance videos

Twitter followers

Facebook likes

YouTube subscribers

0.50 0.48 27.33 −122.52 0.14 43.98 0.0005 −0.12 b.0001 0.03 b.0001 −0.009 0.29

0.14 0.11 4.52 −15.34 0.61 10.65 0.02 −0.02 0.05 0.003 0.01 −0.001 0.85

0.04 −0.0002 0.99 848.87 0.60 167.22 0.49 −0.66 0.19 0.07 0.30 0.24 0.17

0.15 0.11 4.61 −18.58 0.05 3.73 0.01 −0.01 0.08 0.0006 0.12 0.001 0.15

0.004 −0.03 0.11 708.46 0.40 −76.30 0.55 0.07 0.80 0.004 0.91 −0.02 0.81

0.25 0.23 9.13 0.49 0.47 −0.07 0.47 0.0002 0.32 0.00006 0.02 −0.0001 0.04

0.58 0.57 37.95 16,013 0.12 −2064.95 0.18 −4.71 0.14 1.66 b.0001 4.82 b.0001

0.63 0.61 45.01 9,172,136 0.0008 −1,538,218 0.0002 1890.58 0.03 −305.63 0.01 2025.82 b.0001

0.01 0.03 0.22 130,860 0.36 −16,331 0.44 9.05 0.84 0.17 0.98 3.872 0.79

34

J. Zhang / Decision Support Systems 73 (2015) 28–36

Table 10 Analysis of variance: information disclosure with adoption level controlled. Website

Cluster 1 R-sq. Adj. R-sq F value Intercept LogAT EBITDA Liability Intangible

Cluster 2 R Adj. R-Sq F value Intercept LogAT EBITDA Liability Intangible

Twitter

Total news

Finance news

Total tweets

0.59 0.56 22.04 −203.13 0.09 49.42 0.0043 −0.19 b.0001 0.03 b.0001 0.0044 0.66

0.24 0.19 4.97 0.74 b.0001 −0.04 0.07 0.000097 0.03 −0.000009 0.09 0.0000005 0.97

0.32 0.27 6.88 244.69 0.87 −65.47 0.76 −0.45 0.29 0.04 0.43 0.27 0.04

0.86 0.83 28.79 −12.03 0.76 12.74 0.08 0.04 0.03 0.01 0.08 −0.05 0.04

0.43 0.31 3.64 0.82 0.01 −0.02 0.74 0.000005 0.08 0.00 0.82 0.0003 0.10

0.43 0.31 3.63 −1498.16 0.23 351.36 0.12 0.08 0.87 0.11 0.30 −1.43 0.07

and financial tweets, from 0.11 to 0.25. For cluster 2, the adjusted Rsquare for website news increases from 0.48 to 0.83, financial news, from 0.11 to 0.31, total tweets, from −0.0002 to 0.31, and total finance tweets, from 0.11 to 0.23. Readers may argue that the increased Rsquares in this analysis may be limited, but this issue is observed in other studies of voluntary information disclosure on the Internet. For instance, a study [24] shows that the largest adjusted R-square is 0.6 for the regression model of voluntary information disclosure on corporate websites, using different data transformations and datasets. Here, the adjusted R-Square is 0.83 for cluster 2 in the regression model of corporate website news, which is not lower than the model used in the study [24]. The adjusted R-square is 0.56 for cluster 1, which is comparable to the results in the study [24]. From Table 10, we see that the explanatory power of finance-specific disclosure is more significant for website news than for social media; for instance, adjusted R-squares are 0.83 and 0.31 for website news and Twitter tweets, respectively, in cluster 2. A possible explanation is that

YouTube Finance tweets

Total videos

Finance videos

0.30 0.25 6.56 −37.62 0.0015 6.71 b.0001 −0.01 0.08 0.0005 0.15 0.00097 0.32

0.03 0.025 0.6 959.1 0.48 −183.4 0.35 0.11 0.78 0.003 0.94 −0.02 0.87

0.28 0.24 6.06 0.64 0.56 −0.06 0.72 0.0004 0.16 0.00005 0.21 −0.0002 0.02

0.37 0.23 2.78 0.27 0.99 −0.34 0.93 0.004 0.65 −0.0014 0.42 −0.01 0.44

0.49 0.38 4.57 −86.1 0.2 18.1 0.14 0.08 0.003 −0.016 0.008 −0.05 0.19

website technology is more mature and thus more widely adopted compared to such other media as Twitter, Facebook, or YouTube, which are emerging technologies. Some companies are still cautious about moving to the social media world. When a communication technology is widely adopted, its information disclosure can be explained by the determinants suggested in the literature of voluntary information disclosure in traditional filings. In addition, we find that the number of YouTube subscribers is closely related to the number of videos posted on YouTube. This suggests that the former is not affected by those determinants but by the number of videos. The results in Table 11 suggest that YouTube differs from Twitter and Facebook. On Twitter and Facebook, the number of followers or likes can be determined by firm size, profitability, intangible asset, and

Table 11 Analysis of variance: YouTube subscribers. Pr N F

Source

DF

Sum of squares

Mean square

F value

Model Error Corrected total Root MSE Dependent mean Coeff var

1 111 112 29,880 28,447

7.31E+12 9.91E+10 7.41E+12 R-square Adj R-sq

7.31E+12 8.93E+08

8187.64 b.0001

0.987 0.987

Parameter estimate −7720.72 167.45

Standard error 2839.11 1.85

105.03755

Parameter estimates Variable

DF

Intercept YouTube Channel

1 1

t value

Pr N |t|

−2.72 90.49

0.0076 b.0001 Fig. 3. Financial information disclosure.

J. Zhang / Decision Support Systems 73 (2015) 28–36

liability. People are more likely to follow a company on Twitter or like it on Facebook if the company is profitable or large. The number of YouTube subscribers is mainly dependent on the number of videos uploaded onto the site; the more videos posted on YouTube, the more views and subscribers a company has. In Fig. 3, we graphically demonstrate the increase of adjusted Rsquare for financial information disclosure when the adoption level is controlled for. We plot the change of adjusted R-square of the regression model separately for the high adoption group and the low adoption one. In each group, we include Twitter and corporate websites as the media for financial information disclosure. The figure shows the comparison of a cutting-edge medium (Twitter) and a relatively mature medium (websites). Bar height represents the amount of improvement in the explanatory power of the regression model. As shown in Fig. 3, the adjusted R-squares are improved by including the adoption level. In the high adoption level group, the improvement vis-à-vis Twitter is greater than for corporate websites. One can observe the reverse pattern in the low adoption group. The possible reason for this observation is that when companies are highly active in media adoption for disclosure, they also tend to use new media for disclosure more intensively. Following a company's wide adoption and intensive use of a medium for disclosure, its voluntary information disclosure in that medium is driven by the same factors underlying voluntary information disclosure in traditional media—such as firm performance, information environment, information asymmetry level, and debt structure. At that point (of wide adoption and intense use), its disclosure on social media follows the pattern of voluntary information disclosure in traditional SEC filings and government reports. When companies are less active in new-media adoption, corporate websites are a mature technology relative to Twitter; voluntary information disclosure on these companies' corporate websites is more likely to show the pattern of disclosure in traditional filings. In short, when a new technology becomes more mature and well adopted, the factors for voluntary information disclosure through the medium converge to those in traditional filings. 6. Managerial implications It is important for companies to use social media in order to reach target audiences and influence consumer purchasing decisions. We find that the reach of a company via social media is related to its technology adoption level. Active companies, in adopting new communication media, embrace larger audiences and have greater impacts than less active companies. Thus, the adoption of new technologies creates value for companies. Our findings justify the IT investment in new communication technologies of information disclosure. Firm managers make disclosure decisions in traditional media based on their companies' characteristics and their benefits or costs of disclosure. It is uncertain what incents firm managers to voluntarily disclose information on social media. As discussed above, we find that voluntary disclosure via the new medium follows the pattern of voluntary disclosure using traditional media; moreover, such disclosure on new media is explained by the determinants at play in the use of traditional media (SEC filings and government reports), such as firm performance, information environment, information asymmetry level, or debt structure. Our findings suggest that firm managers can use social media to mitigate managerial agency problems. 7. Conclusion and future research We explored voluntary information disclosure on social media, focusing on new media (social media, mobile apps, RSS, email alerts, etc.) for technology adoption. We find that a company's new-media adoption level affects the intensity of its disclosure of financial information on social media. Our findings also suggest that when the technology adoption level is controlled, voluntary information disclosure on social media can be

35

explained by the determinants suggested in the literature on voluntary information disclosure in traditional filings. Our study indicates that company investment in new technologies and communication media is justified. We find that the companies with high adoption levels attract more interested users or investors through social media than do companies with low levels. When a company adopts new technologies and invests resources on new communication media, it embraces larger audiences on social media. Thus, it can influence more people and have a bigger impact on investor decisions and stockholder returns. The limitations of this study may be related to its sample size, which can be addressed in future research. However, 120 selected companies in 2 industries disclosing thousands of news items over 10 different media should provide a reasonable degree of generality regarding the results. We selected companies from 2 industries, based on NAICS: the information industry and the manufacturing industry. Sensitive industries with stricter disclosure regimes, such as the financial industry, are not included. In a future study, we may consider a greater number of companies, in more industries, to strengthen the stability of our analysis.

References [1] D. Aboody, B. Lev, Information asymmetry, R&D, and insider gains, The Journal of Finance 55 (6) (2000) 2747–2766. [2] H. Ashbaugh, K.M. Johnstone, T.D. Warfield, Corporate reporting on the internet, Accounting Horizons 13 (3) (1999) 241–257. [3] G. Casella, R.L. Berger, Statistical Inference, Cengage Learning, Stamford, CT, 2002. [4] J. Chen, H. Xu, A.B. Whinston, Moderated online communities and quality of usergenerated content, Journal of Management Information Systems 28 (2) (2011) 237–268. [5] B.M. Craven, C.L. Marston, Financial reporting on the internet by leading UK companies, The European Accounting Review 8 (2) (2010) 321–333. [6] F. Depoers, A cost-benefit study of voluntary disclosure: some empirical evidence from French-listed companies, The European Accounting Review 9 (2) (2000) 245–263. [7] M.W. DiStaso, D.S. Bortree, Multi-method analysis of transparency in social media practices: survey, interviews and content analysis, Public Relations Review 38 (3) (2012) 511–514. [8] W. Duan, B. Gu, A.B. Whinston, Do online reviews matter?: an empirical investigation of panel data, Decision Support Systems 45 (4) (2008) 1007–1016. [9] M. Ettredge, V.J. Richardson, S. Scholz, The presentation of financial information at corporate web sites, International Journal of Accounting Information Systems 2 (3) (2001) 149–168. [10] M. Ettredge, V.J. Richardson, S. Scholz, Dissemination of information for investors at corporate web sites, Journal of Accounting and Public Policy 21 (4) (2003) 357–369. [11] L. Green, How do different types of manufacturing businesses use social media? Retrieved February 21, 2014 http://sbninteractive.com/2014/02/21/how-dodifferent-types-of-manufacturing-businesses-use-social-media/. [12] R. Greenslade, More digital disruption ahead for mainstream news groups, says surveyRetrieved June 11, 2014 http://www.theguardian.com/media/greenslade/ 2014/jun/12/digital-media-social-media. [13] P.M. Healy, K.G. Palepu, Information asymmetry, corporate disclosure, and the capital markets: a review of the empirical disclosure literature, Journal of Accounting and Economics 31 (1) (2001) 405–440. [14] S.S. Ho, K. Shun Wong, A study of the relationship between corporate governance structures and the extent of voluntary disclosure, Journal of International Accounting, Auditing and Taxation 10 (2) (2001) 139–156. [15] J.W. Kim, Y. Shi, Voluntary disclosure and the cost of equity capital: evidence from management earnings forecasts, Journal of Accounting and Public Policy 30 (4) (2011) 348–366. [16] W.H. Kruskal, W.A. Wallis, Use of ranks in one-criterion variance analysis, Journal of the American Statistical Association 47 (260) (1952) 583–621. [17] R. Lambert, C. Leuz, R.E. Verrecchia, Accounting information, disclosure, and the cost of capital, Journal of Accounting Research 45 (2) (2007) 385–420. [18] R.A. Lambert, C. Leuz, R.E. Verrecchia, Information asymmetry, information precision, and the cost of capital, Review of Finance 16 (1) (2012) 1–29. [19] M.H. Lang, R.J. Lundholm, Voluntary disclosure and equity offerings: reducing information asymmetry or hyping the stock? Contemporary Accounting Research 17 (4) (2000) 623–662. [20] N. Lipschutz, Netflix CEO and the new disclosureRetrieved April 12, 2013, http:// blogs.wsj.com/marketbeat/2013/04/12/netflix-ceo-and-the-new-disclosure/. [21] X. Luo, J. Zhang, How do consumer buzz and traffic in social media marketing predict the value of the firm? Journal of Management Information Systems 30 (2) (2013) 213–238. [22] X. Luo, J. Zhang, W. Duan, Social media and firm equity value, Information Systems Research 24 (1) (2013) 146–163. [23] B. Marsteller, Twitter followers of Fortune Global 100 companies are more connected and influential than average Twitter usersRetrieved March 25, 2014, http://www. burson-marsteller.com/press-release/twitter-followers-of-fortune-global-100companies-are-more-connected-and-influential-than-average-twitter-users/.

36

J. Zhang / Decision Support Systems 73 (2015) 28–36

[24] C. Marston, A. Polei, Corporate reporting on the internet by German companies, International Journal of Accounting Information Systems 5 (3) (2004) 285–311. [25] G.K. Meek, C.B. Roberts, S.J. Gray, Factors influencing voluntary annual report disclosures by US, UK and continental European multinational corporations, Journal of International Business Studies 26 (3) (1995) 555–572. [26] K.J. Merkley, More than numbers: R&D-related disclosure and firm performance(Ph.D. diss.) Business Administration, The University of Michigan, 2011. [27] G.F. Peters, A.M. Romi, Discretionary compliance with mandatory environmental disclosures: evidence from SEC filings, Journal of Accounting and Public Policy 32 (4) (2013) 213–236. [28] B. Raffournier, The determinants of voluntary financial disclosure by Swiss-listed companies, The European Accounting Review 4 (2) (1995) 261–280. [29] SEC, SEC says social media OK for company announcements if investors are alerted, Retrieved April 2, 2013 http://www.sec.gov/News/PressRelease/Detail/PressRelease/ 1365171513574#.U5ppdPldXwg. [30] S. Stieglitz, L. Dang-Xuan, Emotions and information diffusion in social media: sentiment of microblogs and sharing behavior, Journal of Management Information Systems 29 (4) (2013) 217–248.

[31] I.H. Witten, E. Frank, M.A. Hall, Data mining: practical machine learning tools and techniques, Elsevier Science, Amsterdam, The Netherlands, 2011. [32] Y. Xu, C. Zhang, L. Xue, Measuring product susceptibility in online product review social network, Decision Support Systems (2015) (January 10, 2013 available online, in press). [33] Y. Yu, W. Duan, Q. Cao, The impact of social and conventional media on firm equity value: a sentiment analysis approach, Decision Support Systems 55 (4) (2013) 919–926. Juheng Zhang is an assistant professor in the Department of Operations and Information Systems from the Manning School of Business at University of Massachusetts Lowell. She earned a Ph.D. in Business Administration from the University of Florida. Her research focuses on data analytics and examines information disclosure and manipulation on decision making. Juheng Zhang has published in Information Systems Research, Decision Support Systems, and other academic journals.