How do investor relations firms create value for their clients? Evidence from financial restatements

Transcription

1 How do investor relations firms create value for their clients? Evidence from financial restatements Jun-Koo Kang Nanyang Business School Nanyang Technological University Singapore, Lingwei Li* Research School of Accounting The Australian National University Canberra, ACT 2601 Huai Zhang Nanyang Business School Nanyang Technological University Singapore, ABSTRACT We examine whether investor relations (IR) firms create value by improving their clients communications. We find that IR clients experience higher restatement announcement returns than propensity-score matched control firms. Inconsistent with IR firms hyping for their clients, we do not observe return reversals in the post-restatement periods. The positive effect of IR is greater for clients with higher information asymmetry. Further investigations show that the valuation impact goes through both the expected cash flows and the discount rate. Overall, our findings suggest that IR professionals increase value for their clients by helping them communicate effectively during periods of crises, when communication is vital. Keywords: Investor relations firms; Financial restatements; Market reactions. JEL Classification: G14; G29; M41. *Corresponding author. 1

2 Sloane & Company are experts at assisting clients when unforeseen events threaten to impact their business or damage their reputation. We have become a go-to firm when these crises and special situations occur by listening to our clients, understanding the situation, determining the risks to their business and delivering candid advice to management teams, boards, executives and organizations when they need it most. 1 Sloane & Company, one of the largest investor relations firms 1. Introduction Firms investor relations (IR) programs have received academic attention. A central question is whether and how IR activities create value. Bushee and Miller (2012), Kirk and Vincent (2014) and Karolyi and Liao (2015) find that IR activities increase firms visibility and firms with higher IR activities exhibit higher valuation ratios. However, conceptually, high visibility does not readily translate into high market value. While Merton (1987) is supportive of this causal relationship, Da, Engelberg, and Gao (2011) show that firms with higher visibility (proxied by the number of Google search) have lower future returns, because investors overprice prominent stocks. The high valuation ratio could also be due to IR professionals hyping the stock. Solomon (2012) finds that IR firms paint a rosy picture for their clients and investors are disappointed when hard information arrives. Together, these two papers suggest that the visibility-based high valuation might be transitory and unsustainable. In an effort to better understand IR activities, Brown, Call, Clement, and Sharp (2017) survey IR officers and show that they play a very important role in corporate communications. Motivated by Brown, Call, Clement, and Sharp (2017), we ask whether professional IR firms create value by doing their job, i.e., improving the effectiveness of 1 Sloane & Company. 2

3 their clients corporate communications. Specifically, we empirically assess the valuecreation role of IR firms in the setting of financial restatements. This setting offers several advantages. First, as discussed in detail in Section 2, communication is vital during corporate crises. Since financial restatements can be considered as a crisis in corporate credibility, the effectiveness/ineffectiveness of IR firms is heightened during financial restatement periods. Therefore, financial restatements offer a powerful setting to analyze our research question. Second, this setting helps to address the endogeneity concern. This concern arises, because the decision to hire IR firms is likely determined by firm characteristics and these characteristics may have valuation implications. The shortwindow return around financial restatements however is unlikely affected by these characteristics, since they shall be priced by investors prior to the financial restatements. Furthermore, financial restatements are of intense academic interest. A long line of literature examines determinants of financial restatements, the market response to such restatements and subsequent developments afterwards (Palmrose, Richardson, and Scholz, 2004; Farber, 2005; Desai, Hogan, and Wilkins, 2006; Hennes, Leone, and Miller, 2008; Chakravarthy, dehaan, and Rajgopal, 2014). Since a substantial number of firms are clients of IR firms (more than 1,000 CRSP firms each year in our sample), what role IR firms play in this important corporate event is likely an issue of interest to academics, investors and managers. Using a sample of restatements obtained from Audit Analytics in the period of , we find that the restatement announcement return is substantially higher for IR clients who restate their earnings (treatment firms) than for restating firms matched based 3

4 on the propensity of being an IR client (control firms). 2 The three-day window return centered on the restatement announcement date is higher by 1.2% for treatment firms than for control firms, after controlling for restatement severity, firm characteristics, year dummies and industry dummies. These results are consistent with that IR firms create value during corporate crises by improving corporate communications. If indeed IR firms create value through proper corporate communications, we hypothesize that the valuation-creation role of IR firms shall be more pronounced for firms with high information asymmetry where communications are of even higher importance. Following prior literature (Coller and Yohn, 1997; Yohn, 1998; Aboody and Lev, 2000; Zhang, 2006), we identify small firms, firms with high bid-ask spreads, firms with high R&D intensity and firms with high earning volatility as firms with high information asymmetry. Our subsample analyses yield results consistent with this hypothesis, confirming that the impact of IR professionals is related to firms corporate communications. We next track the return subsequent to the financial restatements. Solomon (2012) offers evidence that IR professionals tend to hype their clients corporate news. If the higher financial restatement return is due to IR hyping, we expect that investors of IR-firm clients will be disappointed in the future, which manifest in lower subsequent returns. However, if IR firms do create value for their clients by appropriately formulating and disseminating their clients messages, we expect to observe no return reversals for IR firms 2 We use all the restatements in our main analyses for the following reasons. First, investors could have concerns on the firm's accounting quality and the valuation consequence could be negative even if the restatement has a positive impact on the financial statements. Second, this helps to have a general idea on the impact of IR firms in the setting of financial restatements and, in general, restatement announcement returns are negative. 4

5 clients. Specifically, we examine the 60-day/120-day buy-and-hold market-adjusted after the restatement announcement window. The returns are not statistically different for treatment firms and for control firms. This finding is inconsistent with IR professionals simply hyping for their clients. We examine the possible mechanisms though which IR firms increase value. To this end, we analyze the sentiment imbedded in the restatement press releases. We find that IR clients issue significantly more upbeat press releases when announcing the financial restatements. Their message seems convincing, because the media sentiment is more optimistic around the restatement announcements. These results are consistent with the notion that IR firms help their clients to carefully formulate their message and avoid negative media vibe. We continue to examine the possible mechanisms though which IR firms help increase value. In theory, value can be created by either increasing the expected cash flows or reducing the risk. We study both aspects. We find that, relative to non-ir clients who restate their earnings, IR clients experience an uptick in future operating cash flows. In addition, they have lower risk, measured through standard deviation of stock returns and idiosyncratic volatility. Overall, our results are consistent with that IR professionals help to create value by increasing the expected cash flows and lowering the risk. To check the robustness of the main results, we conduct several additional checks. First, we limit our sample to the samples of intentional restatements and adverse restatements, and our results continue to hold. Second, we use alternative measures of cumulative abnormal returns. The new measures of cumulative abnormal returns are calculated by estimating the Fama-French three-factor model and the four factor model. 5

6 The results are robust to these alternative measures of cumulative abnormal returns. Third, given the criticism on propensity score matching (DeFond, Erkens, and Zhang, 2016), we test the sensitivity of our conclusion using a different approach. Specifically, we repeat the analyses of the market reactions to restatement announcements using the whole sample instead of limiting to the sample of matched firms. The positive association between IR firm use and the restatement announcement returns continues to hold. Lastly, we use a different way to identify control firms. Following Bushee and Miller (2012), we match on industry, year, exchange listing, time listed, and size. Our conclusions do not change. In sum, we find robust evidence that IR firms limit the valuation damage of restatement announcements. The contribution of our study lies in the following aspects. First, this study extends the literature on investor relations. Bushee and Miller (2012) find that external IR firms help to increase small firms visibility. Kirk and Vincent (2014) show that IR activities increase visibility even for big and established firms. Karolyi and Liao (2015) show similar findings internationally. All three papers report that IR professionals/activities elevate firms valuation ratios. Solomon (2012) however paints a somewhat different picture. He shows that IR firms spin positive stories. When hard information arrives, investors are disappointed and IR clients experience lower returns. His results counter the notion that IR professionals create value. Brown, Call, Clement, and Sharp (2017) survey IR officers and show that their activities center on corporate communications. We show that IR firms limit their clients valuation damage in the event of financial restatements by devising effective communication strategies rather than hyping. Our results not only enrich our understanding 6

7 of the IR firms valuation creation role, but also shed light on the IR professionals effectiveness in corporate communications, a question yet to be addressed. Second, this study adds to the literature on financial restatements. Prior studies examine how financial restatements affect firm value and the management s career prospects (e.g., Palmrose, Richardson, and Scholz, 2004; Desai, Hogan, and Wilkins, 2006; Hennes, Leone, and Miller, 2008). Farber (2005) and Chakravarthy, dehaan, and Rajgopal (2014) find that fraud firms take actions to improve their corporate governance and restore investors trust after the restatements, and investors react positively to these actions. Our findings suggest that hiring an outside IR firm could be a useful way to limit the negative valuation consequences of financial restatements. Third, our study provides practical implications for managers and investors. To the extent that IR firms help their clients to communicate appropriately during the crises, firms may consider hiring external IR firms to reduce the negative consequences of the crises. Investors may also take into account the role played by IR firms when making investment decisions. The rest of this paper is organized as follows. Section 2 develops hypotheses. Section 3 presents research designs. Section 4 covers sample formation and descriptive statistics. Section 5 reports empirical results for the hypotheses. Section 6 covers the possible mechanisms and Section 7 covers robustness checks. Section 8 concludes. 2. Hypotheses Development Communication is vital in corporate crises. Effective communication helps firms to quickly recover from damages stemming from unexpected events while poor 7

8 communication deepens crises. For instance, after two Domino s employees posted an offending video on YouTube, the company did not issue a formal press release to handle the situation within the first 24 hours. The lack of response to the crisis caused detrimental consequences to Domino s brand and its customers trust (Flandez, 2009). 3 Financial restatements are a type of corporate crises, since they raise serious concerns on the firm s credibility. Prior studies show that restatements destroy value and lower the management s career prospects (e.g., Palmrose, Richardson, and Scholz, 2004; Desai, Hogan, and Wilkins, 2006; Hennes, Leone, and Miller, 2008). Similar to other crises, communication plays a very important role in restatements. Implementing the appropriate communication tactics helps restating firms to take control of the situation and repair trust and reputation quickly. For example, after Royal Dutch Shell s financial restatement in 2004, its management quickly admitted the company s wrongdoing and communicated effectively with investors about the extent of the restatement and the plans to prevent future occurrence. Their proper communications limited the damage from the embarrassing event. (Gertsen, van Riel, and Berens, 2006) 4 IR professionals specialize in corporate communications. According to the National Investor Relations Institute, investor relations is a strategic management responsibility that integrates finance, communication, marketing and securities law compliance to enable the most effective two-way communication between a company, the financial community, and other constituencies, which ultimately contributes to a company's securities achieving fair valuation. 5 IR firms are not shy about their role in corporate crises. Sloane & 3 Domino s did not hire an outside IR firm at the time of the crisis. 4 Royal Dutch Shell had an outside IR firm at the time of the restatement announcement. 5 Please refer to 8

9 Company, one of the largest IR firms, states on its website that Sloane & Company are experts at assisting clients when unforeseen events threaten to impact their business or damage their reputation. We have become a go-to firm when these crises and special situations occur by listening to our clients, understanding the situation, determining the risks to their business and delivering candid advice to management teams, boards, executives and organizations when they need it most. 6 If IR firms help their clients to communicate appropriately and effectively upon restatement announcements, we expect that restatement announcement returns are higher for firms hiring IR professionals. The valuation impact can come from both the cash flow effect and the risk effect. Better communication reduces information asymmetry, averting investors making worst-case scenario cash flow estimates (Verrecchia, 1983). Moreover, it reduce investors uncertainty related to the restating firms' future prospects and effectively lowers the risk assessment. Our main hypothesis is stated as follows. H1: The restatement announcement returns are higher for IR clients than for non- IR clients. Our main hypothesis is based on the notion that the effect of IR firms is through communications. If the effect is indeed through communications, we hypothesize that the impact of IR professionals is more pronounced for firms with higher information asymmetry, for whom communications are especially important. This is our second hypothesis. H2: The impact of IR on restatement announcement returns is more pronounced for firms with high information asymmetry. 6 Please refer to 9

10 Solomon (2012) shows that IR firms spin their clients news in a positive way to inflate short-term stock prices. When hard information, such as earnings, arrives, the stock market reacts more negatively. If the valuation creation role is due to hyping rather than effective communications, we expect the stock returns after the restatement announcements to be lower for IR clients than for other restating firms. If the hyping story is not true, we expect that IR client firms do not experience return reversals. Since we do not have a clear prediction for the stock returns after restatement announcements, the third hypothesis is stated in the null form. H3: The stock returns after restatement announcements are the same for IR clients and non-ir clients. 3. Research Design 3.1. Propensity Score Matching We investigate the effect of IR firms on market reactions to restatement announcements. The decision of hiring IR firms may be affected by many firm characteristics and these characteristics may have valuation implications, which raises the endogeneity concern. Our restatement setting helps to address this concern since the shortwindow return around financial restatements is unlikely affected by these characteristics. To further mitigate the concern that our results are driven by these firm characteristics rather than the presence of IR firms, we adopt the propensity score matching approach based on ex ante characteristics to identify control firms. The propensity score is estimated by using the following logistic model: IR =α 0 + α 1 SIZE + α 2 LEVERAGE + α 3 BM + α 4 ROA + α 5 LOSS + α 6 LNAGE + α 7 RET + α 8 TURNOVER + α 9 SPREAD + α 10 STDRET + α 11 INSTOWN + α 12 LNANL + 10

11 α 13 LNMEDIA + α 14 MAJORCHG + α 15 ISSUANCE + α 16 MERGER + Year Dummies + Industry Dummies + ɛ (1) IR is a dummy variable which equals one if the firm hires an IR firm at the time of the restatement announcement, and zero otherwise. The data on whether firms hire an IR firm at the time of the restatement announcement are obtain from O Dwyer s Directory of Public Relations Firms and from Factiva and Google search. We require the control firms to be in the same year as the treatment firms and use the closest propensity score to obtain optimal matches, with a caliper of 0.05 and without replacement. Following prior studies on investor relations (e.g., Bushee and Miller, 2012; Solomon, 2012; Kirk and Vincent, 2014), we include a set of independent variables that could be related to firms decisions of hiring IR firms. We use the log of total assets (SIZE) to control for size effects. We include prior leverage (LEVERAGE), defined as book value of debts over total assets, and the book-to-market ratio (BM), measured as the book value of equity divided by market value of equity. Two proxies for firm performance are included: income before extraordinary items divided by total assets (ROA) and a dummy variable which equals one if the income before extraordinary items is negative (LOSS). We also include the log of the number of years since first listed on CRSP (LNAGE). To capture recent changes in firm performance, we include the 12-month buy-and-hold marketadjusted returns (RET). TURNOVER, the average daily number of shares traded divided by shares outstanding, and SPREAD, the average daily quoted spread, are included to proxy for liquidity. We also include the standard deviation of daily returns (STDRET) to capture valuation uncertainty. To measure firms prior visibility, we include the average percentage of institutional ownership (INSTOWN), the log of 1 plus the number of unique analysts issuing earnings forecasts (LNANL), and the log of 1 plus the number of Dow Jones news 11

12 articles (LNMEDIA). Finally, we include variables for recent major changes (MAJORCHG), equity issuance (ISSUANCE), and merger or acquisition activity (MERGER). MAJORCHG is the sum of three indicator variables for changes in exchange listing, changes in the CEO or CFO, and changes in auditors. ISSUANCE is a dummy variable which equals one if the firm issued equity and MERGER is a dummy variable which equals one if the firm was involved in a merger or acquisition. Year fixed effects and industry fixed effects are also included in the model. Following Malmendier and Tate (2005) and Cai and Sevilir (2012), we define industries according to Fama-French 12 industries classification Research Design for H1 For the analyses of the impact of hiring IR firms on market reactions to restatement announcements, we use the following model: CAR(-1,1)/CAR(-7,7) = α 0 + α 1 IR + α 2 FRAUD + α 3 CORE_RESTATE + α 4 NI_RESTATE + α 5 NUM_RESTATE + α 6 YR_RESTATE+ α 7 SIZE + α 8 LEVERAGE + α 9 BM + α 10 ROA + α 11 LOSS + α 12 LNAGE + α 13 RET + α 14 TURNOVER + α 15 SPREAD + α 16 STDRET + α 17 INSTOWN + α 18 LNANL + α 19 LNMEDIA + α 20 MAJORCHG + α 21 ISSUANCE + α 22 MERGER + Year Dummies + Industry Dummies + ɛ (2) CAR(-1,1) and CAR(-7,7) are the cumulative abnormal returns around the restatement announcement dates over the window (-1, 1) and (-7, 7), respectively, where day 0 is the restatement announcement date. Daily abnormal returns (ARs) are calculated using a market model with a 200 trading day estimation period beginning 210 days before and ending 11 days before the restatement date. The CRSP value-weighted return is used as a proxy for the market return. The daily ARs are cumulated to obtain the CAR from one day (seven days) before the restatement announcement date to one day (seven days) after 12

13 the restatement announcement date. The main variable of interest is IR. If IR firms help their clients to reduce the negative market reactions to restatement announcements, we expect the coefficient on IR to be positive and significant. Following Palmrose, Richardson, and Scholz (2004), we include a set of variables related to the severity of the restatements. FRAUD is dummy variable which equals one if the restatement is due to fraud. CORE_RESTATE is a dummy variable which equals one if the restatement involves revenue, cost of sales or operating expense accounts. NI_RESTATE is the restated net income less originally reported net income over restated period scaled by book value of assets reported at the fiscal year end prior to the restatement announcement. NUM_RESTATE is the number of issues contained in the restatement and YR_RESTATE is the sum of years restated. In addition, we control for firm characteristics (SIZE, LEVERAGE, BM, ROA, LOSS, LNAGE, RET, TURNOVER, SPREAD, STDRET, INSTOWN, LNANL, LNMEDIA, MAJORCHG, ISSUANCE, and MERGER). SIZE, LEVERAGE, BM, ROA, and LOSS are measured at the fiscal year end prior the restatement announcement. RET is measured in the 12 months prior to the restatement announcement. LNAGE, TURNOVER, SPREAD, STDRET, INSTOWN, LNANL, LNMEDIA, MAJORCHG, ISSUANCE, and MERGER are measured in the year prior to the restatement announcement. All these firm characteristics are the same as those included in the propensity score matching model. Variable definitions and data sources are provided in Appendix. Year fixed effects and industry fixed effects are included in the model Research Design for H2 13

14 Prior studies suggest that small firms, firms with high bid-ask spreads, firms with high R&D intensity and firms with high earning volatility are subject to high information asymmetry (Coller and Yohn, 1997; Yohn, 1998; Aboody and Lev, 2000; Zhang, 2006). Therefore, we use firm size, bid-ask spread, R&D intensity and earnings volatility to proxy for information asymmetry. To test whether the positive impact of hiring IR firms on market reactions to restatement announcements is more pronounced for firms with high information asymmetry, we estimate the following four regressions: CAR(-1,1)/CAR(-7,7) =α 0 + α 1 IR + α 2 L_SIZE + α 3 IR*L_SIZE + Controls + Year Dummies + Industry Dummies + ɛ (3) CAR(-1,1)/CAR(-7,7) = α 0 + α 1 IR + α 2 H_SPREAD + α 3 IR*H_SPREAD + Controls + Year Dummies + Industry Dummies + ɛ (4) CAR(-1,1)/CAR(-7,7) =α 0 + α 1 IR + α 2 H_RDEX + α 3 IR*H_RDEX + Controls + Year Dummies + Industry Dummies + ɛ (5) CAR(-1,1)/CAR(-7,7) = α 0 + α 1 IR + α 2 H_STDEARN + α 3 IR*H_STDEARN + Controls + Year Dummies + Industry Dummies + ɛ (6) L_SIZE is an indicator that equals one if total assets at the fiscal year end prior to the restatement announcement is not in the highest quartile, and zero otherwise. H_SPREAD is an indicator that equals one if the average daily quoted spread, (ask - bid) / midpoint, in the year prior to the restatement announcement is not in the lowest quartile, and zero otherwise. H_RDEX is an indicator that equals one if research and development (R&D) expenditures are greater than zero, and zero otherwise. H_STDEARN is an indicator that equals one if the standard deviation of annual earnings before extraordinary items scaled by total assets during the past five years is not in the lowest quartile, and zero otherwise. We are interested in the interaction between IR and L_SIZE (H_SPREAD, H_RDEX and H_STDEARN). If the positive impact of IR firms is more pronounced for 14

15 firms with small size, high bid-ask spread, high R&D expenditure and high earnings volatility, we expect that the coefficient on the interaction between IR and L_SIZE (H_SPREAD, H_RDEX and H_STDEARN) is significantly positive. We use the same set of control variables as in equation (2), except that we include L_SIZE rather than SIZE in equation (3) and include H_SPREAD rather than SPREAD in equation (4). Year fixed effects and industry fixed effects are included in the model Research Design for H3 We use the following model to examine whether firms hiring an external IR firm experience lower stock returns after restatement announcements: POSTRET_60D/POSTRET_120D = α 0 + α 1 IR + Controls + Year Dummies + Industry Dummies + ɛ (7) The dependent variables are buy-and-hold market-adjusted returns over the window (8, 67) and (8, 127) after the restatement announcement date (day 0), respectively. If IR firms hype their clients news upon restatement announcements, we expect the coefficient on IR to be negative and significant. If the hyping story is not true, we expect the coefficient on IR to be insignificant. The control variables are the same as those in Equation (2). Year fixed effects and industry fixed effects are included in the model. 4. Sample Formation and Descriptive Statistics 4.1. Sample Formation We start with 12,757 observations of financial restatements from 2002 to The data on financial restatements are obtained from the Audit Analytics database. We then require the dependent variables, i.e., CAR(-1,1), CAR(-7,7), POSTRET_60D and 15

16 POSTRET_120D to be non-missing. In addition, we require all control variables, i.e., FRAUD, CORE_RESTATE, NI_RESTATE, NUM_RESTATE, YR_RESTATE, SIZE, LEVERAGE, BM, ROA, LOSS, LNAGE, RET, TURNOVER, SPREAD, STDRET, INSTOWN, LNANL, LNMEDIA, MAJORCHG, ISSUANCE and MERGER, to have nonmissing values. These leave us with 4,545 observations. Variable definitions and data sources are provided in Appendix. We obtain the client names of IR firms from O Dwyer s Directory of Public Relations Firms. This directory lists IR firms and their long-term (six months or more) clients each year. In order to obtain client identifiers, we match client names with several possible CRSP names using the spelling distance function. We then manually check whether the matching is correct. A Google search is done if the matching is ambiguous. Following Solomon (2012), we exclude clients identified as subsidiaries or regional branches of CRSP firms. If firms hire IR firms both in the year before and in the year of the restatement announcement, then IR is equal to 1. If firms do not hire IR firms either in the year before or in the year of the restatement announcements, then IR is equal to 0. For firms that only hire IR firms in the year before or in the year of the restatement announcements, it is not very clear whether they hire IR firms at the time of the restatement announcements. To solve this issue, we hand collect the data through Factiva and Google search. IR is equal to 1 when the news show that the firm hires an IR firm at the time of the restatement announcement. We delete firms that do not have such evidence. The full sample therefore consists of 4,182 observations, of which 560 belong to the treatment group. 16

17 In order to obtain the propensity score matched sample, we require the control firms to be in the same year as the treatment firms and have the closest propensity score. We regress IR on SIZE, LEVERAGE, BM, ROA, LOSS, LNAGE, RET, TURNOVER, SPREAD, STDRET, INSTOWN, LNANL, LNMEDIA, MAJORCHG, ISSUANCE, MERGER, year dummies and industry dummies to obtain the propensity scores. We do not allow replacement in the matching procedure and require a caliper of The propensity score matched sample consists of 992 observations, including 496 observations in the treatment group and 496 observations in the control group Descriptive Statistics Table 1 describes the descriptive statistics of the control variables. Panel A reports the characteristics of firms hiring IR firms at the time of the restatement announcements (IR=1) and other firms (IR=0) separately based on the full sample. There are 560 observations in the treatment group and 3,622 observations in the control group. Tests for differences in means and medians are reported in the last two columns. Tests for differences in means are based on t-statistics for continuous variables and z-statistics for binary variables, and tests for differences in medians are based on the Wilcoxon rank sum test. All the continuous variables are winsorized at the top and bottom one percentile. In terms of the severity of restatements, the treatment firms, on average, are more likely to have core accounts affected, a greater number of issues contained and a longer period restated. As for firm characteristics, in general, treatment firms have a larger size and a lower book-to-market ratio. Moreover, treatment firms have better performance (i.e., higher ROA and higher stock returns), longer firm age and less return volatility. They are 17

18 also more liquid (i.e., higher turnover and lower bid-ask spreads) and more visible (i.e., higher institutional ownership, greater analyst coverage, and greater media coverage). Finally, treatment firms tend to be involved in equity issuance and mergers and acquisitions. These findings are generally consistent with Solomon (2012). Panel B of Table 1 reports comparison between treatment firms and propensity score matched control firms. Both the treatment group and the control group have 496 observations. As for the restatement severity, the differences between treatment group and control group are insignificant, suggesting that treatment firms and propensity score matched control firms have similar restatement characteristics. In terms of firm characteristics, we find that the differences between the two groups of firms, except for the median value of BM, are insignificant. This finding indicates that treatment firms and propensity score matched control firms are similar in almost all the dimensions considered in the propensity score model, and adds to our confidence that the matching is valid. Table 2 reports descriptive statistics of restatement announcement returns based on the propensity score matched sample. The mean and median values of daily abnormal returns (ARs) and cumulative abnormal returns (CARs) around restatement announcements are negative both when IR is equal to one and when IR is equal to zero. However, the mean and median values of CAR(-1,1), CAR(-5,5) and CAR(-7,7) are statistically higher for the treatment group than for the control group. This finding suggests that IR firms help to alleviate the negative restatement announcement returns. 5. Empirical Results 5.1. Empirical Results for H1 18

19 To test whether firms hiring an external IR firm realize higher restatement announcement returns, we estimate the regression specified in Equation (2). We use CAR(- 1,1) and CAR(-7,7) as the dependent variable respectively. Inferences are based on standard errors adjusted by heteroskedasticity and clustered by firm to control for time-series dependence in residuals. Table 3 reports results regarding restatement announcement returns. We control for restatement severity 7, firm characteristics, year dummies and industry dummies in the regressions. When the dependent variable is CAR(-1,1), the coefficient on IR is and significant at the 5% level. It suggests that IR firm use is related to an increase of 1.2% in the 3-day cumulative abnormal returns surrounding the restatement announcement. The cumulative abnormal returns are measured by estimating the market model. The second regression uses CAR(-7,7) as the dependent variable. The coefficient on IR indicates that having an IR firm is associated with an increase of 1.9% in the 15-day restatement announcement returns. Results from both regressions show that hiring an IR firm is positively associated with restatement announcement returns, which suggests that IR firms help their clients to manage the crises and take control of the situation. 8 7 We also examine the impact of IR firms on restatement announcement returns by partitioning the sample on restatement severity. The results do not vary with restatement severity. 8 An alternative explanation for the finding is that IR firms prep the public with some hints of upcoming restatements and effectively leak the information prior to the restatement announcements. To examine this alternative explanation, we use the 60-day and 120-day buy-and-hold market-adjusted return before the restatement announcement window (PRERET_60D and PRERET_120D) as the dependent variables and repeat the analyses. If the less negative market response is due to the information leakage, we expect the stock returns prior to the restatement announcements to be more negative for firms being an IR client. If the information leakage story is untrue, we do not expect that hiring IR firms is associated with the stock return before the restatement announcement. The results show that IR is not associated with PRERET_60D and PRERET_120D, suggesting that the main results are not driven by information leakage before the restatement announcement. 19

20 5.2. Empirical Results for H2 The results of the subsample analyses regarding firms information asymmetry, measured by firm size, bid-ask spread, R&D expenditure and earnings volatility, are reported in Table 4. H2 predicts that the positive impact of having IR firms on restatement announcement returns is more pronounced for firms with high information asymmetry. This prediction is supported by the empirical results. Specifically, in Regressions (1) and (2), the coefficients on the interaction between the IR dummy (IR) and smaller firm size (L_SIZE) are positive and significant at 1% level and 5% level, respectively. The coefficients on the interaction between IR and higher bid-ask spread (H_SPREAD) are significantly positive in both Regressions (3) and (4). When we partition the sample based on R&D intensity (H_RDEX) in Regressions (5) and (6), the coefficients on the interaction term are positive and significant. Moreover, in Regressions (7) and (8), the coefficients on the interaction between IR and higher earnings volatility (H_STDEARN) are significantly positive. In sum, the results suggest that the effect of IR firms on restatement announcement returns is more pronounced for firms with higher information asymmetry, i.e., smaller firm size, higher bid-ask spread, higher R&D intensity and higher earnings volatility Empirical Results for H3 To examine whether IR client firms experience return reversals after the restatement announcements, we estimate the model specified in Equation (7). We require the dependent variables, i.e., POSTRET_60D and POSTRET_120D, to be non-missing. The results are reported in Table 5. In the first regression, the dependent variable is the buy- 20

21 and-hold market-adjusted returns over the window (8, 67) after the restatement announcement date (day 0). In the second regression, the dependent variable is the buyand-hold market-adjusted returns over the window (8, 127) after the restatement announcement date (day 0). The coefficients on IR are insignificant in both regressions. Therefore, there is no association between the presence of IR firms and the buy-and-hold market-adjusted returns after the restatement announcements, indicating that IR firms help their clients to communicate in an unbiased manner instead of hyping their clients news upon restatement announcements. 6. Possible Mechanisms 6.1. Press Release Sentiment and Media News Sentiment To confirm our story that IR firms help to limit the damage from financial restatements by providing useful communications, we examine the press release sentiment and the media news sentiment around the restatement announcements. We obtain the sentiment score from RavenPack, which is calculated based on the tone of the press release or the media news. A higher value means more positive sentiment. The results are reported in Table 6. In Regressions (1) and (2), the dependent variables are the change in the press release sentiment. Specifically, CHG_PRSEN3/CHG_PRSEN15 is the difference in the average press release sentiment score between the window (-1, 1)/(-7, 7) surrounding the restatement announcement date and the window (-4, -2)/ (-24, -8) before the restatement announcement date. The coefficients on IR are positive and significant at 10% level in both regressions, indicating that the press releases around the restatement announcements are more positive for IR-client firms after controlling for restatement severity, firm 21

22 characteristics, year dummies and industry dummies. In Regressions (3) and (4), the dependent variables are the change in the media news sentiment. CHG_MSEN3/CHG_MSEN15 is the difference in the average media news sentiment score between the window (-1, 1)/(-7, 7) surrounding the restatement announcement date and the window (-4, -2)/(-24, -8) before the restatement announcement date. We find that the coefficients on IR are positive and significant at 5% level in both regressions, suggesting that IR firms help to generate more positive media news. To sum up, our results suggest that IR firms positively affect their clients valuation through press releases and media news Cash Flow and Risk: Difference-in-differences Tests As we discussed earlier, IR firms could help their clients to increase future cash flows and decrease the risk. The findings on press release sentiment and media news sentiment cannot provide direct evidence on whether IR firms increase their clients value through the increase in future cash flows or the decrease in the risk or both. The reason is that both the increase in future cash flows and the decrease in uncertainty could be positively associated with press release sentiment and media news sentiment. To have a better idea on the mechanisms through which IR firms increase their clients value, we examine restating firms cash flows and risk in the tree years before and three years after the restatements, using difference-in-differences regressions. Specifically, we include the POST dummy and the interaction between IR and POST in the regressions. POST is an indicator that equals one for the post-restatement year, and zero otherwise. The interaction between IR and POST captures the effect of IR firms on their clients cash flows and risk. 22

23 The results are reported in Table 7. In regression (1), the dependent variable is OCF, which is operating cash flow to total assets. The coefficient on IR*POST is positive and significant at 5% level, indicating that IR firms help to increase their clients cash flows. In regressions (2) and (3), we use two measures of risk as the dependent variables. STDRET is the standard deviation of the firm s daily returns over the year. IDIOSRISK is the idiosyncratic risk of the firm s stock returns, measured as the standard deviation of the error term of the market model, which is estimated using daily returns over the year. The coefficients on the interaction between IR and POST are significantly negative in both regressions, which suggests that IR-client firms experience relatively lower risk after the restatements. Overall, the findings supports the notion that IR firms help increase their clients value through increasing the expected cash flows and lowering the risk. 7. Robustness Checks 7.1. Using the Samples of Intentional Restatements and Adverse Restatements We examine the market reactions to restatement announcements using the samples of intentional restatements and adverse restatements. We start with all the intentional (adverse) restatements and use the same propensity score matching approach to identify control firms. Panel A of Table 8 reports results. In regressions (1) and (2), we use the propensity score matched sample of intentional restatements that are not solely driven by errors. The sample consists of 944 intentional restatement firms (472 treatment firms that have an IR firm and 472 control firms that do not have an IR firm). The coefficients on IR are positive and significant in both regressions. In regressions (3) and (4), we use the propensity score matched sample of adverse restatements that have negative impact on the 23

24 financial statements. The sample consists of 804 adverse restatement firms (402 treatment firms that have an IR firm and 402 control firms that do not have an IR firm). We find positive and significant coefficients on IR in both regressions. Therefore, our results are robust if we limit to the samples of intentional restatements and adverse restatements Using CARs Estimated with Alternative Models To check the robustness of the main results, we use alternative measures of cumulative abnormal returns. CAR(-1,1)_FF and CAR(-7,7)_FF are cumulative abnormal returns over the window (-1, 1) and (-7, 7) surrounding the restatement announcement date (day 0) respectively, measured by estimating the Fama-French three-factor model over the window (-210, -11). CAR(-1,1)_FFM and CAR(-7,7)_FFM are calculated by estimating the Fama-French three factor plus the momentum factor model. Panel B of Table 8 reports results using these alternative measures of restatement announcement returns. The coefficients on IR are positive and significant in all the regressions. Therefore, the results are robust to the alternative measures of cumulative abnormal returns Using the Full Sample and Alternative Matched Sample We repeat the analyses of the market reactions to restatement announcements by using the full sample and an alternative matched sample. Panel C of Table 8 reports the results. Regressions (1) and (2) report results based on the full sample. The coefficients on IR are positive and significant in both regressions. Therefore, the positive association between having an IR firm and market reactions to restatement announcements continues to hold after using this sample. 24

25 Following Bushee and Miller (2012), we construct a different matched sample by matching on industry, year, exchange listing, time listed and size. The control firms are required to be in the same Fama-French 48 industry, in the same year, and traded on the same exchange as the treatment firms in the year before the restatement announcement. We then search for the closest match in firm size within the set of firms whose time listed on the exchange was within four years of the treatment firm. The sample consists of 702 observations. Regressions (3) and (4) report the results after using this alternative matched sample. The coefficients on IR are positive, significant at 5% level. Therefore, the positive association between IR firm use and restatement announcement returns is robust to this alternative matched sample, suggesting that IR firms help their clients to manage the crises and communicate effectively. 8. Conclusions IR firms are important information intermediaries that help their clients to communicate, engage and build relationships with investors and other stakeholders. On average, more than 1,000 CRSP firms hire an external IR firm each year. In this study, we examine whether hiring an IR firm increases shareholder value through crisis communications in a setting of financial restatements. We find that firms hiring an IR firm realize less negative market reactions to restatement announcements than propensity score matched control firms, consistent with that IR firms increase value for their clients by communicating with investors and stakeholders effectively during periods of crises. We future show that the positive impact of hiring an IR firm on restatement announcement returns is more pronounced for smaller firms, firms with higher bid-ask 25

26 spread, firms with higher R&D intensity and firms with higher earnings volatility. The results suggest that the role of IR firms in crisis communications depends on firms information environment and IR firms impact tends to be stronger for firms with higher information asymmetry. Moreover, IR clients do not experience return reversals after the restatement announcement. This findings is inconsistent with IR hyping. Rather, it shows that the value created is real and sustainable. Our further analyses show that the sentiment in press releases and media news around the restatement announcements are more positive for IR clients, supporting the notion that the positive valuation impact of IR firms is indeed through communications. What s more, a difference-in-different test shows that IR clients experience higher cash flows and lower risks after financial restatements, than non-ir clients. Finally, we conduct several robustness checks. We find that the positive association between hiring IR firms and restatement announcement returns is robust to the samples of intentional and adverse restatements, alternative measures of cumulative abnormal returns, and alternative matched samples. Overall, the results suggest that crisis communications is a mechanism through which IR firms increase shareholder value. Specifically, IR firms assist their clients in managing the crises and communicating appropriately at the time of restatement announcements, thus alleviating the negative valuation consequences of financial restatements. Our study not only contributes to academic literature but also offers practical implications for investors, managers and the public. 26

27 References Aboody, D., and Lev, B., Information asymmetry, R&D, and insider gains. The Journal of Finance, 55(6), Brown, L.D., Call, A.C., Clement, M.B. and Sharp, N.Y., Managing the Narrative: Investor Relations Officers and Corporate Disclosure. Working paper, Temple University, Arizona State University, University of Texas at Austin and Texas A&M University. Bushee, B. J., and Miller, G. S., Investor relations, firm visibility, and investor following. The Accounting Review, 87(3), Cai, Y., and Sevilir, M., Board connections and M&A transactions. Journal of Financial Economics, 103(2), Chakravarthy, J., dehaan, E., and Rajgopal, S., Reputation repair after a serious restatement. The Accounting Review, 89(4), Coller, M., and Yohn, T. L., Management forecasts and information asymmetry: An examination of bid-ask spreads. Journal of Accounting Research, 35(2), Da, Z., Engelberg, J. and Gao, P., In search of attention. The Journal of Finance, 66(5), DeFond, M. L., Erkens, D. H., and Zhang, J., Does PSM really eliminate the Big N audit quality effect?. Working paper, University of Southern California, Georgetown University and The University of Texas at Dallas. Desai, H., Hogan, C. E., and Wilkins, M. S., The reputational penalty for aggressive accounting: Earnings restatements and management turnover. The Accounting Review, 81(1), Farber, D. B., Restoring trust after fraud: Does corporate governance matter?. The Accounting Review, 80(2), Flandez, R., Domino s response offers lessons in crisis management. The Wall Street Journal, April 20. Available at /2009/04/20/dominos-response-offers-lessons-in-crisis-management/. Gertsen, F. H., van Riel, C. B., and Berens, G., Avoiding reputation damage in financial restatements. Long Range Planning, 39(4), Hennes, K. M., Leone, A. J., and Miller, B. P., The importance of distinguishing errors from irregularities in restatement research: The case of restatements and CEO/CFO turnover. The Accounting Review, 83(6),

28 Karolyi, G. A., and Liao, R. C., The economic consequences of investor relations: A global perspective. Working paper, Cornell University and Rutgers University. Kirk, M. P., and Vincent, J. D., Professional investor relations within the firm. The Accounting Review, 89(4), Malmendier, U., and Tate, G., CEO overconfidence and corporate investment. The Journal of Finance, 60(6), Merton, R.C., A simple model of capital market equilibrium with incomplete information. The Journal of Finance, 42(3), Palmrose, Z. V., Richardson, V. J., and Scholz, S., Determinants of market reactions to restatement announcements. Journal of Accounting and Economics, 37(1), Solomon, D. H., Selective publicity and stock prices. The Journal of Finance, 67(2), Verrecchia, R. E., Discretionary disclosure. Journal of Accounting and Economics, 5, Yohn, T. L., Information asymmetry around earnings announcements. Review of Quantitative Finance and Accounting, 11(2), Zhang, X., Information uncertainty and stock returns. The Journal of Finance, 61(1),