Media Content, Accounting Quality, and Liquidity Volatility

Abstract

We examine how the linguistic content of news items affects the volatility of a firm's liquidity, and we consider whether accounting quality moderates the media content-liquidity volatility relation. Regarding the unconditional relation between media content and liquidity volatility, one view is media content could reduce liquidity volatility by providing additional information about fundamental values; another view is it could increase liquidity volatility by increasing investor uncertainty, particularly for negative news. Using data from Thomson Reuters News Analytics, we find evidence supporting the view that media content, positive and negative, has incremental information. Regarding the moderating role of accounting quality, pre-existing accounting information of higher quality could enhance investors' reactions to media content by providing a more precise baseline, or it could reduce investors' reactions to the news if investors anchor on higher quality financial statements. Our findings are consistent with more credible accounting information serving an anchor role, and suggest that investors condition their reaction to media content based on the quality of a firm's pre-existing accounting information.

Acknowledgements

We thank Henk Berkman, Mike Bradbury, Asher Curtis, David Emanuel, Jyh-Bang Jou, David Lont, Dimitri Margaritis, Ben Marshall, Philip O'Connor, and seminar participants at Auckland University of Technology, La Trobe University, Massey University, Monash University, National Central University, University of Auckland, 2012 NZ Quantitative Accounting Research Symposium, 2013 American Accounting Association Annual Meeting, and 2013 European Accounting Association Annual Congress for their useful comments. We thank Thomson Reuters for providing access to the TRNA database.

Notes

1 Bushee et al. (2010) find that the number of unexpected press sources covering a firm is negatively related to the abnormal bid-ask spread and positively related to abnormal market depth, consistent with more extensive press coverage reducing information asymmetry and increasing the level of liquidity. However, their other measures of the extent of press coverage (abnormal article word count, abnormal number of Factiva topic codes in an article, and abnormal number of articles with author by-lines or Factiva analysis codes) are not negatively related to information asymmetry.

2 We focus on firm-specific liquidity volatility because we want to examine the effects of the content of firm-specific news items. Liquidity risk is commonly defined in terms co-movement with the market, although there is not a singular definition of liquidity risk. For example, Acharya and Pedersen (2005) identify three different types of systematic risk related to liquidity, that is, the co-variations between firm returns and market liquidity, firm liquidity and market liquidity, and firm liquidity and market returns. Further, Pastor and Stambaugh (2003) capture liquidity risk by estimating a liquidity beta using the Fama and French three-factor model with a market liquidity factor included. Sadka (2006) uses a price-impact measure based on Glosten and Harris (1988) as his proxy for liquidity risk. However, we note that two of the measures we use – the standard deviation and skewness of the Amihud measure – are used by Lang and Maffett (2011) as measures of liquidity risk.

3 Most studies on the business press examine media coverage (e.g. Tetlock, 2007; Fang & Peress, 2009).

4 There is a significant body of research focusing on liquidity risk (e.g. Acharya & Pedersen, 2005; Pastor & Stambaugh, 2003) and liquidity volatility (e.g., Chordia, Subrahmanyam, & Anshuman, 2001).

5 At the same time, the two views suggest endogeneity, an issue we consider in our robustness tests.

6 The high rate of misclassification arises because these programs rely on word lists developed in other academic disciplines. Words that have a negative connotation in a general context are often not negative in a financial context.

7 Callen et al. (2013) examine the effects of pre-existing and newly arrived information on stock price delay while we examine how they affect liquidity volatility. However, the underlying link is information uncertainty. In their case, uncertainty about fundamental values leads to noisy stock prices that later converge. In our case, uncertainty about fundamental values affects the ability of traders to receive funding.

8 The first principle component explains 60% of the total variance, the second and third principle components explain 30% and 9% of the total variance, respectively.

9 TRNA also provides a relevance score for each item. Relevance is scored from 0 (low) to 1 (high). For example, a news item that mentions several firms may not be as relevant for all the firms if one firm is the main focus. However, in our main tests, similar to Tetlock et al. (2008), we use all news stories available because the information in any news story could potentially affect liquidity volatility. Our results hold when we use other cut-offs for the relevance score.

10 Our sample is considerably larger than Tetlock et al. (2008) who focus on the S&P 500 and Kothari et al. (2009) who use a sample of 889 firms from four industries.

11 Given the high correlation between LiqVol between and Ami reported in Table 2, we re-estimate model 3 using Ami as the dependent variable to see if a test using the liquidity level is the same as a test using liquidity volatility (we drop Ami as a control variable for this test). We find that PosMedia is negatively and significantly related to Ami (t-stat. = −7.922, p < .01); however, contrary to Table 3, NegMedia is not significantly related to Ami (t-stat. = −0.992). This suggests that negative news content, in particular, can reduce dispersion in liquidity without changing the level. Thus, a test of liquidity volatility and the liquidity level are not identical.

12 We also estimate Equation (4) for our full sample including contemporaneous positive and negative news to ensure that they are not driving our main results. Controlling for contemporaneous news, we find the coefficients for PosMedia and NegMedia are still negative and highly significant (t-stat. = −7.927, p < .001, and t-stat. = −8.566, p < .001, respectively).

13 Since firm size, leverage, and book-to-market are the set of control variables in Kothari et al.'s (2009) analyses we estimate a model that includes contemporaneous news and these three control variables. The results for PosMedia_t0 and NegMedia_t0 are qualitatively the same as column (5) (t-stat. = −3.185, p < .001, and t-stat. = 8.355, p < .001, respectively), indicating that the different set of control variables do not affect the results for the media variables.

14 Our results also suggest that liquidity volatility has meaning over and above the effect of general returns volatility since we control for return volatility in Table 3.

15 In our sample, the mean for Size (AQ) is 14.621 (−0.039) for firm-months with press coverage and 12.891 (−0.05) for firm-months with no press coverage. Both differences are significant at the 0.01 level.

16 Bushee et al. (2010) also control for analyst following, a measure of the firm's information environment. We re-estimate our analyses in Tables 3 and 5 with the log of one plus the number of analysts covering the firm as an additional control variable. The coefficients for PosMedia, NegMedia, and their interactions with AQ are slightly smaller but remain statistically significant at the same levels. In addition, analyst following is significantly and negatively related to LiqVol, indicating that a larger analyst following is associated with lower information asymmetry and lower liquidity volatility.

17 The business press has its own unique incentives (e.g., see Dyck & Zingales, 2003; Solomon & Soltes, 2011).

18 TRNA includes a code for the source of each news item. Based on Bushee et al. (2010), Solomon and Soltes (2011), Wikipedia (http://en.wikipedia.org/wiki/News_agency), and web searchers, we identify the following as firm-initiated news sources: Asian Corporate Newswire, Business Wire, Cision, Filing Services Canada Newswire, Globe Newswire, Hugin, Marketwire, Prime Newswire, and PR Newswire.

19 Elliott (2011) identifies 9 August 2007 as the start of the crisis because on that date BNP Paribas announced it was curtailing its involvement in three hedge funds specializing in US mortgage debt. Elliott (2011) writes that at that ‘moment it became clear that there were tens of trillions of dollars' worth of dodgy derivatives swilling round which were worth a lot less than the bankers had previously imagined’.

20 Also see Wysocki (2009) and Dechow, Ge, and Schrand (2010) for concerns about the Dechow-Dichev (2002) model.

21 We acknowledge that Barth et al.'s (2013) transparency measure is a measure of the firm's overall earnings quality, and therefore also comingles the performance component and accounting error component of accruals. However, unlike Dechow and Dichev (2002), Barth et al.'s (2013) objective is not to model the error component of accruals. Instead, their measure uses the market's response to infer earnings quality. Thus, their measure reflects the market's assessments of the performance component and the unintentional and intentional accounting error components embedded in a firm's accruals. On the other hand, as Dechow et al. (2010) point out, market-based measures can only provide a conditional measure of earnings quality since earnings quality can be affected by other factors such as monitoring, accounting standards, and other information available to investors.