Mian and Sankaraguruswamy (2012, hereafter MS) provide evidence that market-wide investor sentiment is related to the stock market's response to earnings windfalls around the earnings announcement date. Investor sentiment across the market is expected to influence the market's response to good and bad news in different ways. Thus, our research extends Tetlock (2007) by examining the effects of media sentiment at the firm level and complements MS and Seybert and Yang (2012) by examining whether the market-wide effects of investor sentiment are limited, amplified, or unaffected by the firm. -specific media mood.
As such, investors' response to media sentiment may be conditioned by prevailing investor sentiment. We also estimate our main model separately for subsamples based on extreme market-wide investor sentiment. We find that positive company-specific media sentiment increases (decreases) the market reaction to good (bad) unexpected earnings, while negative company-specific media sentiment increases (decreases) the reaction to bad (good) unexpected earnings, after controlling for market-wide investor sentiment.
Furthermore, we find that there is no evidence of an interaction between media sentiment and investor sentiment. For example, we find that media sentiment is significantly related to mispricing both when market-wide investor sentiment is extremely low and extremely high. We extend this literature by examining the joint effects of firm-specific media sentiment and market-wide investor sentiment on stock price sensitivity to earnings news.
MS finds that the stock market's reaction to unexpected earnings around the earnings announcement date is influenced by market-wide investor sentiment.
Data and Method
When Sent_Media is the dependent variable, we use the residual of equation (1) as a measure of firm-specific press-initiated media sentiment. When Sent_Firm is the dependent variable, we use the remainder of equation (1) as a measure of the. Size is the log value of total assets (Compustat item ATQ) in the previous quarter.
ROA is net income before extraordinary items (IBQ) divided by total assets (ATQ) in the previous quarter. BM is the book-to-market ratio of the previous quarter, which is calculated as the book value of common shares (CEQQ) divided by the market value of common shares defined as the closing price of the common shares at the end of the quarter (PRCCQ) times the common shares outstanding (CSHPRQ). DivYield is the dividend yield in the previous quarter, defined as dividend per share (DVPSXQ) divided by the closing price of common stock at the end of the quarter (PRCCQ).
MktSent is the monthly Baker-Wurgler Sentiment Index at the beginning of the current month. MediaSent is press-initiated firm-specific sentiment, calculated as the residual of equation (1) where Sent_Media is the dependent variable.
Results 1 Main findings
More importantly, we find that our first interest rate coefficient, UEUpxMediaSent, is positive and significant after controlling for market-wide sentiment in the model, which is consistent with investors overreacting to positive earnings surprises when company-specific media sentiment is high. Our second yield coefficient, UEDownxMediaSent, is negative and significant, indicating that investors overreact to negative earnings surprises when company-specific media sentiment is low. In this case UEDownxMktSent, which had a negative and significant coefficient in the reduced model. columns 1 and 2), is not significant when company-specific media sentiment is included.
We examine the returns over a 60-day period from t+2 in case there is a short-term continuation of the effect of firm-specific media sentiment in the post-earnings announcement period. Third, we examine whether media sentiment is correlated with contemporaneous financial analyst earnings forecast error. We examine this issue by comparing the analyst forecast errors for deciles of media sentiment.
We scale the residual by the raw media sentiment, so a smaller (larger) scaled residual indicates that the fundamentals explain more (less) of the raw media sentiment. As a result, the interaction of windfall earnings and media sentiment should have a larger impact on earnings announcement CAR when the scaled residual is large, if the information history holds. If media sentiment reflects this other information, UEDownxMediaSent or UEUpxMediaSent will lose significance once we control for analyst forecast revisions.
If media sentiment affects investor behavior in a similar way to market sentiment, we expect indirect differences in ERC sensitivity to media sentiment based on firm value uncertainty. For hard-to-value firms, investors always react to positive (negative) earnings windfalls when positive (negative) media sentiment is high. Also, we find evidence that market-wide investor sentiment affects the sensitivity of stock prices to earnings windfalls when media sentiment is controlled for, but only for hard-to-value firms.
In other words, the relationship between earnings windfall (either positive or negative) and media sentiment is unaffected by prevailing market sentiment. When market sentiment is very high (columns 3 and 4), UEUpxMediaSent and UEDownxMediaSent remain statistically significant, confirming the finding from Panel A that the role of media sentiment is independent of market sentiment. These two results are consistent with the MS results for overall market sentiment and suggest that both media sentiment and market sentiment are important when market sentiment is high.
Overall, the results in Table 9 support the view that the effect of media sentiment is independent of market sentiment. Importantly, these results show that media sentiment and investor sentiment are two distinct drivers of sentiment.
Conclusion
BM is the book-to-market ratio of the previous quarter, which is calculated as the book value of common shares (CEQQ) divided by the market value of common shares defined as the closing price of the common shares at the end of the quarter (PRCCQ) times the outstanding ordinary shares. Lev is the leverage ratio in the previous quarter, defined as the sum of long-term debt (DLTTQ) and the debt in current debt (DLCQ) divided by total assets (ATQ). ChgNOA is the change in net operating assets in the last quarter, defined as the change in net operating assets scaled by total assets (ATQ), where net operating assets are calculated as operating assets (total assets (ATQ) minus the sum of cash and investments (CHEQ)) minus operating liabilities (total liabilities (LTQ) minus total long-term debt (DLTTQ) and the debt in short-term debt (DLCQ)).
Variable definitions: CAR is the cumulative abnormal return for the period -1, +1 surrounding the earnings announcement. UE is the current quarterly profit minus the same quarterly profit from the previous year, scaled by the stock price at the end of the current quarter. OwnSent is a firm-initiated sentiment, calculated as the residual from equation (1) where Sent_Firm is the dependent variable.
MktPE is the total PE of the stock market in the previous month minus the average PE of the market over the previous 12 months. The table provides the results for the second-stage regression model, i.e., equation (2), where the dependent variable is CAR, which is the cumulative abnormal return for the -1, +1 window around the earnings announcement. Variable definitions: In columns 3 and 4, UE is the current quarter's earnings minus the same quarter's earnings in the previous year, scaled by the stock price at the end of the current quarter.
In columns 5 and 6, UE is the difference between actual earnings and the consensus analyst forecast scaled by the period's opening share price. Panel A provides the results for equation (3) where the dependent variable is Earnings, which is the 1-8 quarter forward earnings per share before extraordinary items. Panel B provides the results for equation (3) where the dependent variable is CFO, which is the 1-8 quarter forward operating cash flow per share. t-statistics are adjusted for clustering by month.
UE is profit in the current quarter minus profit in the same quarter of the previous year, calculated based on the share price at the end of the current quarter. UEDown) is the product of UE and an indicator variable equal to 1 if UE is positive (negative) and 0 otherwise. Variable definitions: UE is the current quarter's profit minus the profit in the same quarter in the previous year, calculated based on the stock price at the end of the current quarter.
The table provides the results for the second-stage regression model, i.e., equation (2), where the dependent variable is CAR, which is the cumulative abnormal return for the -1, +1 window around the earnings announcement, for the subsamples based in uncertainty or difficulty in evaluating the firm. R The table provides results for second-stage regression models where the dependent variable is CAR, which is the cumulative abnormal return for the -1, +1 window around the earnings announcement, for two analyzes examining the interaction of media and market sentiment. - the broad sentiment of investors.
