Top executives on social media and information in the capital market: Evidence from China

Social media platforms are becoming increasingly important channels for information dissemination. This study examines how microblogging by top executives affects the information environment for listed firms in an emerging market. Using manually collected set from Sina Weibo, one of China’s most popular and largest social media platforms, we find that a board chair having a Weibo account is associated with the dissemination of more firm-specific information to the capital market. This result holds up to a battery of robustness tests. We also show that the relationship between board chairs’ Weibo usage and information dissemination is stronger for smaller firms, firms that went public more recently, and firms characterized by less analyst coverage. Findings in this study have important implications for the understanding of the role of social media in the dissemination process of corporate information.


Introduction
Over the last decade, the internet has become the most important source for financial news, with thousands of websites providing information that can be accessed by all types of investors (Drake et al., 2017). More generally, it has been shown that consumers increasingly turn to fellow customers rather than expert advice when making their buying decisions (Chen et al., 2014). The growing importance of social media in information dissemination and the influence it has on buy and sell decisions in all walks of life have also transformed how financial information is disseminated and utilized.
Previous studies have shown that the internet and social media act as channels for dissemination of corporate information, which in turn may affect the capital market (e.g. Hu et al., 2013;Blankespoor et al., 2014;Chen et al., 2014). The Securities and Exchange Commission (SEC) in the United States recognized the growing importance of this new information channel and announced that it would allow firms to disclose news through social media in 2013.
The goal of this study is to examine how social media usage by top executives acts as an information intermediary for the firms they control. To do this, we collect official Sina Weibo accounts of board chairs in Chinese listed firms and analyze how the activation of a microblog account influences the dissemination of information. Sina Weibo is the largest microblogging service in China with 361 million monthly active users as of June 2017. Similar to Twitter, it is a popular platform for quick dissemination of news and opinions, and is thus suitable for examining the research questions in this study.
We hypothesize that microblogging by board chairs in Chinese listed firms complements traditional information sources by helping to improve information dissemination to the market. Moreover, if microblogging by board chairs indeed helps decrease information asymmetries in the financial markets, we believe that the impact of board chairs' activity over social media on firm-specific information dissemination will increase for firms characterized by low transparency. Typical firms with lower transparency include smaller firms, firms that recently went public, and firms with less analyst coverage.
We collect data on board chairs' Weibo accounts and analyze their effect on information pertaining to stock price returns between 1 January 2010 and 31 December 2016. We find that only a small percentage (from 1 percent in 2010 to approximately 2.5 percent in 2016) of the board chairs of all listed firms have signed up for a Weibo account. In our benchmark analysis, we find that there is a significant and negative relationship between board chairs on Weibo and stock return synchronicity. This means that a board chair with a Weibo account is positively related to a better information environment for the firms under his or her control. We confirm this initial baseline result with several robustness tests, including a firm-fixed regression analysis, an instrumental variable two-stage least square analysis, and various matching methods.
Next, we examine how specific firm characteristics influence the relationship between board chair' Weibo usage and information dissemination. We find that the effect of Weibo usage by board chairs is significantly stronger for smaller firms, newly listed firms, and firms with less analyst coverage. In addition, we find that the relationship between board chair' Weibo usage and information dissemination is greater for family firms. For additional robustness, we also carry out an analysis in which we exclude firms in the finance industry and observations from 2015 as that was an unusually turbulent year for the Chinese stock market. Last but not least, we exclude alternative explanations for our findings. To sum up, our empirical findings support the hypothesis that activity over social media by board chairs in Chinese listed firms helps improve the information environment for the firms in question and that this relationship is dependent on several firm characteristics that are associated with transparency.
This study connects to several strands of literature. First, we contribute to a large body of literature that examines the information environment for listed firms. Previous research has analyzed capital market effects of professional reports (e.g., Drake et al., 2014;Dai et al., 2015), press coverage (Fang and Peress, 2009;Kothari et al., 2009;Bushee et al., 2010), and different types of internet intermediaries (Drake et al., 2017).
Our study builds on this research by examining how social media usage by top executives affects the incorporation of firm-specific information in the capital market.
Second, we extend the literature on social media and corporations. Previous studies have shown that firms use social media to reduce information asymmetry and increase liquidity of their stocks (Blankespoor et al., 2014), engage their customers (Lee et al., 2017), attempt to control the negative effects of product recalls (Lee et al., 2015), and for the strategic dissemination of news (Jung et al., 2017). We contribute to this strand of literature by providing evidence for how corporations can utilize social media to improve their information environment. how firm news disseminated via Twitter is related to lower bid-ask spreads and higher market liquidity. We extend this burgeoning literature by shedding light on how social media usage by key personnel in listed firms affects the stock market.
The rest of the study is organized as follows. Section 2 introduces the Chinese microblogging service provided by Sina Weibo and explains why we have chosen this particular social media platform for our analysis. Section 3 presents the data and methodological approach. Section 4 provides the results of the benchmark model as well as alternative model specifications and robustness tests. Section 5 examines different types of firm-specific characteristics that may influence the relationship between board chairs' Sina Weibo usage and the capital market. Section 6 provides further robustness tests that focus on the data sample and the measurement of information dissemination. Finally, Section 7 concludes the study.

Sina Weibo -Microblogging in China
The Chinese government has long enforced a policy of both encouraging the use of the internet while also maintaining vigorous control (Sullivan, 2012). One direct effect of this strategy is that the ecological social media system found in most other countries around the world is not present in China. The global microblogging platform Twitter was established in 2006 and was soon used by Chinese netizens as well. As early as 2007, several Chinese microblogs services (including TaoTao, Jiwai, Fanfou, and Zuosa) similar to Twitter were operational (Qin et al., 2017). However, none of these platforms managed to attract a larger user base. In 2009, microblogging had begun to play an important role as a channel for free-flowing information on current events. However, this was not to last. In July 2009, a series of violent riots broke out in Urumqi, the capital city of the Xinjiang Uyghur Autonomous Region in China. The Chinese government quickly blamed the ethnic riots that resulted in the death of more than 200 people on the uninhibited flow of information and rumors online.
Subsequently, the use of international social media platforms such as Twitter and Facebook was banned. The news portal Sina.com presented a plan for a new microblog social media platform that would control incoming posts by tracking and blocking content that was deemed to be too sensitive (Sullivan, 2012 Today, Sina Weibo, typically called Weibo, is one of China's biggest social media platforms. While it is often called "China's Twitter", Weibo is more versatile and can in many ways be regarded as a combination of Twitter and Facebook. Weibo posts are limited to 140 characters, and photos, videos, images, and gifs can be uploaded as well. There are also significant differences between Twitter and Weibo in terms of user behavior (Gao et al., 2012). Weibo users tend to be more active and disclose more information about themselves (in contrast to Twitter users, Weibo users also post more during the weekends). While Twitter is a popular channel for political news flows and opinions, Weibo users, for natural reasons, tend to avoid issues that relate to political organizations and other institutions.
While some observers have argued that new rules established in 2015 requiring users to register with their real names would result in a strong decline and even death of Weibo, user trends between 2015 and 2017 say otherwise. In the 2017 June-ended quarter, Sina Weibo reported that it had reached 361 million monthly active users, an increase of 28 percent from the same period the previous year. This can be compared to Twitter reporting 319 million monthly active users for the same quarter. A 2016 report from China Internet Network Information Center (CNNIC) shows that the main purpose of using Weibo is to obtain trending news in time. Thus, Weibo constitutes a dominant source of news content which enables its users to acquire, share, and comment on a variety of subjects. The importance of Weibo as a channel for dissemination of new information also means that it is suitable for the analysis of how top executives' use of social media affects the information environment of their firms.

Board Chairs' Sina Weibo Usage
We first identify the board chair for all listed companies during the period 2010-2016. The sample starts in 2010 since Sina Weibo was launched the previous year. The main reason we focus on the board chair rather than the CEO is because the chair often has the most authority in terms of making operation decisions in Chinese companies (Kato and Long, 2006;Feng and Johansson, 2017). 2 For each board chair, we then collect information on whether he or she has opened up an official Weibo account. Table 1 presents the number of all listed firms and the number of board 2 Feng and Johansson (2016) note that the board chair in Chinese companies is often called yi ba shou (number one), while the CEO is er ba shou (number two). chairs with Weibo accounts for each year throughout the sample period. As is evident in the table, only a small fraction of board chairs in listed companies have a Weibo account. In 2010, as few as 18 or less than 1 percent of the total number of board chairs had a Weibo account. The number of board chairs with Weibo accounts has increased slowly but steadily throughout the sample period, with a total of 67 or 2.7 percent of the total number of board chairs having a Weibo account in 2016. Table 1 presents the distribution of listed firms by industry and the number of firms characterized by having a board chair with a Weibo account. Most of the firm observations are within manufacturing and that is also where most of the observations of firms with a board chair using Weibo is found. However, the highest percentage of firms characterized by having a board chair with a Weibo account are found in the hotel and catering, culture, sports and entertainment, and information technology industries.

Panel B in
[

Stock Return Synchronicity
To measure informativeness in the capital market, we use a firm-specific return variation measure. A similar approach has been used in several studies on stock market information (e.g., Durnev et al., 2003;Durnev et al., 2004;Jin and Myers, 2006;Feng et al., 2016). More specifically, we follow Morck et al. (2000) and calculate the stock return synchronicity for each firm. We first obtain the R 2 from an expanded index model 3 : Here, , is the return of stock i in week t. , is the market return in week t, calculated as the weekly tradable market value-weighted returns of all Chinese A-share. Finally, , is the industry return in week t, calculated as the tradable market value-weighted industry index, excluding firm i. R 2 is the coefficient of determination from the estimation of model (1). As R 2 ranges from 0 to 1, we then follow Morck et al. (2000) and use a logistic transformation to obtain a suitable measure of stock return synchronicity: A high value of Synch suggests that less firm-specific information is included in the stock return and that it is instead market-wide information that is driving the price (Roll, 1988;Morck et al., 2000). Correspondingly, a smaller value means that more firm-specific information is incorporated into the stock market.

Firm Data
We obtain firm-level data for all listed firms with A-shares on the Shenzhen and Synchronous Fundamentals, the Spearman correlation between the firm's ROA and its industrial average ROA over the past ten quarters.

Baseline Results
We begin our analysis of the relationship between board chair' Weibo usage and stock return synchronicity with an overview of R 2 and Synch for firms with and without board chairs with a Weibo account. Panel A in Table 2 presents the mean and median, as well as tests for differences between the two groups of firms. The results show that firms characterized by having a board chair with a Weibo account have a smaller R 2 and Synch. The difference is significant for both the mean and median values of the two measures. While preliminary, these findings indicate that board chairs using Weibo is associated with more firm-specific information in their stock prices and thus provide a better information environment for investors.
Although these initial results support our research hypothesis, we need to control for firm-specific variables that may influence stock return synchronicity. To do this, we run a baseline ordinary least square regression and a Fama and MacBeth (1973) panel regression using the following model specification: ℎ is the measure for stock return synchronicity for firm i in year t, which is defined in Section 3.2 and , the main explanatory variable, is a dummy variable which equals one if the board chair of firm i had a Weibo account in year t and zero otherwise. is a vector of the additional control variables introduced in Section 3.3, is a vector of year-and industry-fixed effects. We cluster standard errors by firm and year (Petersen, 2009;Thompson, 2011).
The results for two regressions are presented in Panel B of Table 2. In both estimations, the main explanatory variable Weibo is significantly negative. These results corroborate the initial findings of a significant positive relationship between board chairs' Weibo usage and firm-specific information imputed in the market.
Microblogging board chairs are thus associated with a better information environment for investors.
[ To further handle with the potential endogeneity problem, we run a fixed-effect regression. The fixed-effect model allows us to control for firm-specific effects, and thereby to control for unobservable firm-related factors that may influence stock return synchronicity. It also helps us address the issue of potential reverse causality in the baseline regression. The fixed-effect variable regression model includes the same dependent variables as the benchmark regression, but excludes Control-Ownership and Segments, as these typically change very slowly. The number of reservations in the fixed-effect regression is reduced somewhat since each firm needs at least two observations to be included. The fixed-effect regression results are presented in Table   3. The key explanatory variable Weibo is still negatively significant at the 1 percent level, suggesting that the initial results supporting our hypothesis are robust.
[ The results from the fixed-effect regression at least partially alleviate the concerns for potential endogeneity. However, the model specification in that analysis assumes that the potential unobserved heterogeneity we address is constant over time.
To address the possibility of time-varying omitted variables, we also conduct a two-stage instrumental variable (IV) analysis. In the first stage, we construct a selection model for board chairs to open a Weibo account. In the second stage, we include Predicted Weibo as an independent variable. We construct a measure for board chair personality by identifying his or her personal characteristics in the news. We first flag all news articles we find that contain at least the company name/abbreviation/code and the name of the board chair. Several individuals involved in the project then read the articles in detail to ascertain personal characteristics of the board chair. If at least three different sources claim that the board chair is good at social interaction (shejiao), we define the dummy variable Social as one and zero otherwise. We use the following news sources for the collection of this data: GTA Financial News Database, Genius Finance, INFOBANK, and China Core Newspapers Full-text Database.
Panel A in Table 4 presents the summary statistics of the IV Social. On average, board chairs with a Weibo account are clearly seen as better at social interaction in Chinese media. Board chairs with a Weibo account have higher mean and median scores for Social and simple tests for difference in mean and median of Social are significant at the 1 percent level. Next, we run the two-stage regression analysis. Panel B in Table 4 presents the results, with the first column displaying the first-stage results and the second column showing the second-stage results. In the first-stage regression, the coefficient for Social is positively significant at the 1 percent level, indicating that the IV Social indeed is strongly related to Weibo usage. In the second-stage regression, Predicted Weibo is negatively significant at the 1 percent level. To be sure of the suitability of the estimation, we also carry out tests for exogeneity, relevance and validity of instruments. The Shea partial R 2 values and the F-statistic provide further support for the relevance of our IV in the first stage. The Anderson-Rubin F-statistic rejects the null hypothesis, thereby indicating that the endogenous regressor is relevant.
Finally, the Hansen J-statistic is unable to reject the null hypothesis that the instrument is valid and orthogonal to the residuals. The exclusion of them in the main estimated equation is thus appropriate. To conclude, the two-stage regression again supports the main initial result, namely that microblogging board chairs are positively associated with a better information environment.
[ It is always a challenge to identify a good IV for a two-stage regression analysis.
We acknowledge this and conduct additional robustness tests based on matching samples. Our main aim here is to use matching methods to identify samples of control firms with similar characteristics but without board chairs who have Weibo accounts.
If the matching models are well designed, the difference between the treatment sample and the control sample will be driven by the key explanatory variable Weibo. For comprehensiveness, we use four alternative matching methods commonly found in the literature: firm size and industry; firm size, industry, board chair age, and board chair education; caliper matching; Kernel matching. The first two matching methods are based on firm-and individual-specific characteristics, while the latter two are statistical propensity score matching methods. To look at the strength of our results in Table 5, we also conduct a differencein-difference (DiD) test to analyze the average treatment effects. To do this, we first construct firm samples for "Pre-Weibo usage", with the three-year average stock return synchronicity before the board chair of the firm in question opens up a Weibo account.
We create corresponding subsamples for "Post-Weibo usage". We then match each of these subsamples using the four different matching methods discussed earlier. By doing this, we can analyze the difference between pre-and post-Weibo usage by board chairs and compare this difference for the treatment and control groups. Table 6 presents the results for all four DiD tests. Panel A presents the results for the size-and industry-matched firm groups. Looking at the first two lines, we find that the difference in stock return synchronicity is highly significant for the treatment group, but insignificant for the control group. That is, Weibo significantly reduces stock return synchronicity for treated firms but not for control firms. Next, the first two columns show that the difference in stock return synchronicity for treated and control firms is not significant before Weibo usage, but their difference becomes significant after Weibo usage. That is, the information environment is similar for treated and control firms before Weibo usage, but after a board chair opens a Weibo account, more firm-specific information is incorporated into the stock market.
The remaining three panels exhibit the same results for tests using alternative matching methods. These results show that firms experience a significant improvement in the dissemination of information to the market when their board chair opens up a Weibo account.
[ Overall, however, the results indicate that the relationship between board chairs' Weibo usage and information efficiency is not driven by ownership type.
[ sample period used in this study, 2015 marked a year with very high turbulence. As Figure 1 shows, there was an extremely strong bull-market run leading to what many regarded as the burst of a bubble in June that year. A third of the value of A-shares was lost within a month and market volatility remained high for an extended period.
To take these potential sample issues into account, we run two new sets of regressions, one without financial firms and one in which we exclude all firm observations for 2015. For each of these new samples, we run three separate regressions: a standard OLS regression, a Fama-MacBeth regression, and a two-stage regression.
The results of these estimations are presented in Table 9. As can be seen in the table, the coefficient for our key explanatory variable Weibo is negatively significant at the 1 percent level in all six estimations. We can therefore conclude that our initial results on the relationship between board chairs' Weibo usage and firm-specific information in the stock market also hold up when we take potential data sample issues into consideration. [ Here, , is the market-adjusted monthly returns compounded over a 12-month period that ends the fourth month after the end of firm i's fiscal year; NI is net income deflated by the market value of equity for firm i at the beginning of year t; _ , is the scaled decile rank score. To control for determinants of the relationship between returns and earnings, we include a set of control variables, , : Firm Size, defined as the natural logarithm of the market capitalization; Tobin's Q, the ratio of the sum of market value of equity and book value of liabilities over the book value of total assets; Leverage, the ratio of total liabilities to total assets. Finally, we control for industry and year fixed effects by including dummies for each of them.
The results of the regression on the return-earnings association are presented in Table 10. The coefficient for , is significantly positive, suggesting that earnings are associated with stock returns. Moreover, the coefficient for , * _ , is significantly negative, indicating that the market assigns a lower value to earnings of firms characterized by a higher stock return synchronicity, and that information on corporate earnings is less impounded in stock prices for them. These findings support the argument that our measure of stock return synchronicity captures the extent to which firm-specific information incorporated into stock prices in our sample. [

Conclusion
This study examines how the usage of social media in the form of microblogging by corporate executives in China is affecting the information environment for their firms. We provide strong empirical evidence that Chinese listed firms with a board chair who owns a Weibo account are characterized by a significantly better dissemination of information compared to other listed firms. This finding holds up to a battery of robustness tests, including tests on the measure for firm-specific information dissemination itself. In addition, we find that certain characteristics such as firm size, how recent a firm went public, and analyst following are influential factors behind the relationship between board chairs' social media usage and firm-specific information dissemination. A plausible reason for these findings is the potential effect on information dissemination for smaller firms, firms that were listed more recently, and firms with less analyst coverage is larger since they constitute firms characterized by a lower level of information being disseminated to the market in general.
These findings suggest that social media can act as an effective information channel that complements traditional channels such as public information by the firm,    The dependent variable is Synch, a commonly used stock return synchronicity measure calculated as . R 2 is from regressions of the market model of return of the firm against the stock market index and industry index using weekly data. Weibo is a dummy variable which equals one when the board chair of firm i posted weibo at year t, and zero otherwise. Size is the natural logarithm of market capitalization of firm i at the beginning of year t. Leverage is defined as the book value of all liabilities scaled by total assets, again measured at the beginning of the year t. ROE is the ratio of net profits divided by total equities at the beginning of the year t. Sales Growth is the ratio of sales growth from last year to this year. Segments is the number of segments, including only those with sales that exceed 30% of firm i's total sales at the beginning of year t. Volume is the natural logarithm of trading volume of firm i at year t. Volatility is the standard deviation of the stock return of firm i at year t. Illiquidity is defined as the average ratio of daily absolute returns to the daily trading volume at year t, multiplied by 10 9 . %INST is the ratio of mutual funds' holdings, measured as the aggregate number of shares held by mutual funds, scaled by outstanding shares of firm i in year t. Analyst is the natural logarithm of one plus the number of analysts that cover firm i at year t. Investibility is the investibility measure of firm i at year t. HHI (Herfindahl-Hirschman Index) is an indicator of competition, estimated by using all listed firms' sales from the same industry at the beginning of year t. Family Firm is a dummy variable which equals one if the firm is ultimately controlled by individuals, and zero otherwise. (Control -Ownership) is the difference between the ultimate owner's control rights and ownership. Ownership is defined as the cash flow rights of the ultimate owners. Synchronous fundamentals is defined as the Spearman correlation between the firm's ROA and its industrial average ROA over the past ten quarters. Model (1) presents the results from the OLS in which year and industry dummies are included but not reported. t-statistics are given in parentheses and computed using heteroskedasticity-robust standard errors clustered by firm and year (Petersen, 2009;Thompson, 2011). Column (2) presents Fama and MacBeth(1973) panel results. Industry dummies are included but not reported and t-statistics are computed using heteroskedasticity-robust standard errors clustered by industry. All continuous variables are winsorized at the top and bottom 1%.***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.   . R 2 is from regressions of the market model of return of the firm against the stock market index and industry index using weekly data. Weibo is a dummy variable which equals one when the board chair of firm i posted weibo at year t, and equals zero otherwise. Size is the natural logarithm of market capitalization of firm i at the beginning of year t. Leverage is defined as the book value of all liabilities scaled by total assets, again measured at the beginning of the year t. ROE is the ratio of net profits divided by total equities at the beginning of the year t. Sales Growth is the ratio of sales growth from last year to this year. Volume is the natural logarithm of trading volume of firm i at year t. Volatility is the standard deviation of the stock return of firm i at year t. Illiquidity is defined as the average ratio of daily absolute returns to the daily trading volume at year t, multiplied by 10 9 . %INST is the ratio of mutual funds' holdings, measured as the aggregate number of shares held by mutual funds, scaled by outstanding shares of firm i in year t. Analyst is the natural logarithm of one plus the number of analysts that cover firm i at year t. Investibility is the investibility measure of firm i at year t. HHI (Herfindahl-Hirschman Index) is an indicator of competition, estimated by using all listed firms' sales from the same industry at the beginning of year t. Synchronous fundamentals is defined as the Spearman correlation between the firm's ROA and its industrial average ROA over the past ten quarters. Year dummies are included but not reported and t-statistics are computed using heteroskedasticity-robust standard errors clustered by year. All continuous variables are winsorized at the top and bottom 1%.***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.   (1) and (2), and the corresponding median value is in parenthesis.

Firm Fixed Effect
The last column provides T-statistic ( Wilcoxon-Mann-Whitney test ) for the difference between firms with and without board chair weibo. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively. Weibo

Panel B: Two-stage Regression Results
This panel presents the two-stage regression results of Weibo and stock return synchronicity. Predicted Weibo is the predicted probability of Weibo based on the estimation in the first-stage model. The dependent variables in the second stage model are stock return synchronicity. The lower part of this panel shows the partial F-statistic and the partial R2 from the first stage regression and the values for Anderson-Rubin Fstatistic test and Hansen J-statistic.     (1) and (2) are dimension-by-dimension matching methods. (3) and (4) are propensity score matching methods. "Size/Industry matching" matches each treated firm with control firm which has the nearest market capitalization and is also in the same industry. "Size/Industry and Chair Age/Gender/Education matching" matches each treated firm with a control firm that has the nearest market capitalization, is in the same industry, and whose chair has the nearest age, same gender and educational attainment. The propensity score is estimated using the prediction model in the first stage as seen in Table 3. "Caliper matching" presents the treatment effect using caliper matching with a caliper of 0.05. "Kernel matching" gives the treatment effect using kernel matching. T-statistics are calculated using bootstrapping. ***, **, and * denote significance at the 1%, 5% and 10% level, respectively.

Matching Methods Stock return synchronicity
(1)Size/Industry matching -0.073*** (2)Size/Industry and Chair Age/Gender/Education matching -0.076*** (3)Caliper matching -0.082*** (4)Kernel matching -0.084*** Table 6. Difference-in-Difference Method This table presents average treatment effects, using the difference-in-difference (DiD) method. "Pre-weibo usage" is the three-year average stock return synchronicity before the board chair opened a Weibo account. "Post-weibo usage" is the three-year average stock return synchronicity after the board chair opened a Weibo account. The corresponding median is given in parentheses. The last column reports a T-test (Wilcoxon-Mann-Whitney test) for the difference between "Post-weibo usage" and "Pre-weibo usage". The last row reports a T-test (Wilcoxon-Mann-Whitney test) for the difference between Firms with and without Weibo. All variables are winsorized at the top and bottom 1%. *** ,**, and * denote significance for the difference between two samples at 1%, 5%, and 10%, respectively.

Panel A. Size/Industry matched (N=62)
"Size/Industry matching" matches each treated firm with a control firm that has the nearest market capitalization and is in the same industry.

Panel B. Size/Industry and Chair Age/ Gender/Education matched (N=56)
"Size/Industry and Chair Age/Gender/Education matching" matches each treated firm with a control firm that has the nearest market capitalization, is in the same industry, and whose board chair has the nearest age, same gender and education degree.

Panel C. Caliper matching (N=52)
The propensity score is estimated using the prediction model in the first stage as seen in Table 3. "Caliper matching" presents the treatment effect using caliper matching with a caliper of 0.05.

Panel D. Kernel matching (N=54)
The propensity score is estimated using the prediction model in the first stage as seen in Table 3. "Kernel matching" gives the treatment effect using kernel matching.

Pre-Weibo
Post-Weibo Difference between post-and pre-Weibo  . R 2 is from regressions of the market model of return of the firm against the stock market index and industry index using weekly data. Weibo is a dummy variable which equals one when the board chair of firm i posted weibo at year t, and equals zero otherwise. The focus in this table are Weibo*Smaller Firm (Panel A), Weibo* Younger Firm (Panel B) and Weibo* Fewer Analysts (Panel C), which are the interaction terms of Weibo and different firm characteristics. Size is the natural logarithm of market capitalization of firm i at the beginning of year t. Leverage is defined as the book value of all liabilities scaled by total assets, again measured at the beginning of the year t. ROE is the ratio of net profits divided by total equities at the beginning of the year t. Sales Growth is the ratio of sales growth from last year to this year. Segments is the number of segments, including only those which sales that exceed 30% of firm i's total sales at the beginning of year t. Volume is the natural logarithm of trading volume of firm i at year t. Volatility is the standard deviation of the stock return of firm i at year t. Illiquidity is defined as the average ratio of daily absolute returns to the daily trading volume at year t, multiplied by 10 9 . %INST is the ratio of mutual funds' holdings, measured as the aggregate number of shares held by mutual funds, scaled by outstanding shares of firm i in year t. Analyst is the natural logarithm of one plus the number of analysts that cover firm i at year t. Investibility is the investibility measure of firm i at year t. HHI (Herfindahl-Hirschman Index) is an indicator of competition, estimated by using all listed firms' sales from the same industry at the beginning of year t. Family Firm is a dummy variable which equals one if the firm is ultimately controlled by individuals, and zero otherwise. (Control -Ownership) is the difference between the ultimate owner's control rights and ownership. Ownership is defined as the cash flow rights of the ultimate owners. Synchronous fundamentals is defined as the Spearman correlation between the firm's ROA and its industrial average ROA over the past ten quarters. Model (1) presents the results from the OLS in which year and industry dummies are included but not reported. t-statistics are given in parentheses and computed using heteroskedasticity-robust standard errors clustered by firm and year (Petersen, 2009;Thompson, 2011). Column (2) presents Fama and MacBeth(1973) panel results. Industry dummies are included but not reported and t-statistics are computed using heteroskedasticity-robust standard errors clustered by industry. All continuous variables are winsorized at the top and bottom 1%.***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.

Table 10. Stock Return Synchronicity and Return-earnings Associations
This table present the regression results about the effect of stock return synchronicity on return-earnings associations. The dependent variable is MAR, defined as marketadjusted monthly returns compounded over the 12-month period ending the fourth month after the end of a firm's fiscal year; NI is net income deflated by the market value of equity at the beginning of the fiscal year; Syn_decile is the scaled decile rank score. Control variables include MCap, measured by the natural log of market capitalization; Leverage, measured as the ratio of total liabilities over total assets; and Tobin's Q, measured as the ratio of the sum of market value of equities and book value of liabilities over book value of total assets. Year and Industry dummies are included but not reported, and t-statistics are given in parentheses and computed using heteroskedasticity-robust standard errors clustered by firm and year (Petersen, 2009;Thompson, 2011). All continuous variables are winsorized at the top and bottom 1%. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively.

Variable Definition Syn
A commonly used stock return synchronicity measure, calculated as log ( /(1 − )). R 2 is from regressions of the market model of return of the firm against the stock market index and industry index using weekly data.

Weibo
A dummy variable which equals one if the firm's board chair opens a Weibo account, and zero otherwise.

Size
The natural logarithm of market capitalization of firm i at the beginning of year t. Leverage The book value of all liabilities scaled by total assets at the beginning of the year t. ROE Net profits divided by total equity at the beginning of the year t.

Sales Growth
Sales growth during the last year.

Segments
The number of segments, including only those with sales that exceed 30% of firm i's total sales at the beginning of year t. Volume The natural logarithm of trading volume of firm i at year t. Volatility The standard deviation of the stock return of firm i at year t. Illiquidity The average ratio of daily absolute returns to the daily trading volume at year t, multiplied by 10 9 .

%INST
Mutual funds' holdings, measured as the aggregate number of shares held by mutual funds divided by outstanding shares of firm i in year t. Analyst The natural logarithm of one plus the number of analysts that cover firm i at year t. Investibility The investibility measure of firm i at year t, i.e. the ratio of shares which can be traded in the secondary market.

HHI
Abbreviation for Herfindahl-Hirschman Index, an indicator of competition, estimated by using all listed firms' sales from the same industry at the beginning of year t.

Family Firm
A dummy variable which equals one if the firm is ultimately controlled by individuals and zero otherwise.

Control -Ownership
A proxy for the ultimate owner's control in excess of ownership rights, defined as the difference between the ultimate owner's control rights and ownership (similar to that of La Porta et al., 1999).

Ownership
The cash flow rights owned by ultimate owner (similar to that of La Porta et al., 1999).

Synchronous fundamentals
The Spearman correlation between the firm's ROA and its industrial average ROA over the past ten quarters. (1) (2)