«Internal and External Attributions by Managers in Earnings Conference Calls by Zhenhua Chen Business Administration Duke University Date:_ Approved: ...»
4. Firm size. Baginski et al. (2004) document that larger firms are more likely to provide external attributions with management earnings forecasts. I include the natural logarithm of market value (LNMVE) as a control variable and expect its coefficient to be negative.
5. Length of conference calls. As shown in Table 2 Panel F, the statistical distribution of IvsE is affected by the length (LENGTH) of conference calls. I therefore include LENGTH in the regression analysis and expect a negative coefficient.
Panel A of Table 5 shows the Pearson correlation matrix of CEO attributions and firm characteristics. The correlation between IvsE and ROA is positive and statistically significant, indicating that better performance leads to more internal attributions relative to external attributions. This suggests that self-serving bias seems to be the more dominant force in management attributions.1 IvsE is also positively correlated with assets and the logarithm of market value of equity, which is not in line with the result in Baginski et al. (2004) that larger firms tend to make external attributions. The correlation between IvsE and DRND is -0.15 and significant, consistent with the notion that firms with higher proprietary costs are less likely to make internal attributions. IvsE is positively correlated with DSI, suggesting that firms that have special items are more likely to make internal attributions, consistent with the results in Baginski et al. (2004).
IvsE is weakly related to contemporaneous market reaction and not correlated with future stock returns.
It is not clear whether the correlation is economically significant. In the later part of the paper I examine the incrementally explanatory power of firm performance relative to the control variables to see if the association between performance and managers’ attributions is economically significant.
Table 5: The Association between Firm Performance and Managers’ Attributions Panel A: Pearson Correlation Matrix between IvsE and Firm Characteristics
Panel A presents the Pearson correlation matrix of IvsE and firm characteristics. Panel B reports ordinary least squares regression estimation of equation (1). T-statistics are shown in parentheses with standard errors clustered at both firm and quarter level. ***/**/* means significance at 0.01, 0.05 and 0.10 level, respectively.
Panel B of Table 5 presents the results of the OLS regression analysis. As can be seen in column 1, the coefficient on ROA is 0.233 (t = 3.897). This indicates that managers’ attributions are positively associated with firm performance, suggesting that self-serving attribution bias dominates leadership in earnings conference calls.
The coefficient on DRND is negative (coefficient = -0.056) and statistically significant, consistent with the notion that firms with higher proprietary costs are less likely to provide internal attributions since it may reduce the firms’ competitive advantage. The coefficient on DSI is positive, consistent with Baginski et al. (2004) that firms with special items are more likely to make internal attributions. The coefficient on LNMVE is positive, which is inconsistent with Baginski et al. (2004).
The model has reasonable explanatory power with an R2 of 10.5%. Untabulated results show that when I stepwise introduce the dependent variable, the explanatory power of ROA is trivial (0.8%). The majority of the explanatory power comes from industry fixed effects (5.6%). Combined with the finding reported in Panel A of the correlation between IvsE and ROA (9.0%), these results suggest that earnings are not the first order determinant of internal and external attributions.
Bloomfield (2008) suggests that spontaneous conversations (such as Q&A sessions of conference calls) are more suitable in testing psychology theory than prescripted texts (such as the presentation sections of conference calls). If this claim is valid, I expect to see stronger results using the attribution measure based on CEO responses in the Q&A sessions. However, Larcker and Zakolyukina (2011) do not find significant differences in predicting financial misreporting when they compare the presentation sessions and Q&A sessions. To test whether CEO attributions differ from presentation to Q&A sessions, I measure IvsE separately for the presentation and Q&A sections of the conference calls, and re-estimate equation (1) using IvsE from presentation and Q&A sections separately.
Column 2 presents the regression results based on management attributions during presentation section of the conference call. The results are largely in line with using IvsE based on the entire conference calls, suggesting that CEOs tend to refer to internal factors for good performance and external factors for poor performance even in the presentation section. Column 3 shows the regression results based on Q&A sections of the conference calls. The coefficient on ROA is larger for results for the Q&A sections.
However, untabulated results show that the difference is not statistically significant, consistent with the recent finding in Larker and Zakolyukina (2012) that managers do not have different speaking patterns during the presentation and Q&A sections.
7. Market Reaction and Analyst Forecast Revisions to Managers’ Attributions In this section I examine how investors and financial analysts respond to management attributions by examining the contemporaneous stock market returns and analyst forecast revisions. Investors’ and analysts’ future earnings forecasts are affected by their diagnosis of performance drivers in the previous periods (Koonce et al. 2011). If managers’ attributions help investors evaluate the performance drivers, investors may adjust their future earnings forecasts and react accordingly.
To test for market reactions, I regress the three-day cumulative abnormal returns measured around the conference call date CAR(-1,1) on the unexpected component of IvsE1. To measure the expected component of IvsE, I use seasonal prior quarter value of IvsE. I control for quantitative accounting news contained in the earnings conference call by including earnings surprises (UEt). I also control for news in linguistic tone documented in prior research (Davis et al. 2008; Demers and Vega 2008; Engelberg 2008;
Loughran and McDonald 2011). Kinney et al. (2002) find that market reaction to earnings surprise is not linear but S-shaped, and thus I include LargePosUE and LargeNegUE as explanatory variables to control for the non-linearity of the return-earnings relation. In addition, I control for other firm characteristics that may affect market reaction during Using unadjusted IvsE as the dependent variable yields similar empirical implications.
the conference call window: firm size (LNMVE), growth option (BM), and momentum (MOM).
I use the following model to test market reaction:
Table 6 shows that the market reacts negatively to more internal attributions provided by managers. This is consistent with the results in a recent study that finds a negative market reaction to CEO’s self-reference words during CNBC interviews (Kim 2011). This negative reaction suggests investors interpret managers’ self-serving attributions as a negative signal. Whether this stock market inference is rational is investigated in more detail in section 8. The market also react to news in earnings (UE) and news in linguistic tone (POSWORDS and NEGWORDS), consistent with prior studies.
This table reports ordinary least squares regression estimation of the association between attributions and contemporaneous stock market reaction CAR(-1,1). All the variables are defined in Appendix A. T-statistics are shown in parentheses with standard errors clustered at both firm and quarter level. ***/**/* means significance at 0.01, 0.05 and 0.10 level, respectively.
To examine analyst reactions, I use analyst forecast revisions (FREV) as the dependent variable instead of CAR(-1,1) in equation (2). I also modify equation (2) to include the contemporaneous market reaction CAR(-1,1) as an additional independent variable to control for other contemporaneous news that is not captured by other explanatory variables. I expect the coefficient on CAR(-1,1) to be positive.
I measure FREV as the one-quarter-ahead forecast revision, which is calculated as the difference between the median one-quarter-ahead forecast issued after and before the current quarter earnings announcement, scaled by stock price two days before the earnings conference call. I calculate the median of analyst forecasts before (after) the earnings announcement using the last (first) forecast of all individual analysts who issue forecasts during the 90-day period before (after) the current quarter earnings announcement.
analysts do not incorporate the information in managers’ attributions in their future earnings forecasts. Consistent with prior studies, I find that the coefficients on earnings surprises (UE) and the contemporaneous returns (CAR(-1,1)) are positive and significant.
The insignificant coefficient on IvsE is possibly explained by the following three possibilities. First, analysts do not incorporate managers’ attributions when forecasting earnings. Second, analysts do incorporate CEO attributions into their long term earnings forecasts, but not into their short term earnings forecasts. To disentangle these possibilities requires an examination of how CEO attributions associate with firms’ future performance, which I examine in section 8. Third, the information contained in CEO attributes is fully captured by the contemporaneous stock returns and hence the effect is controlled away. However, this explanation is unlikely because when contemporaneous stock returns are removed from the model, I still do not observe a significant coefficient on the CEO attribution signal.
This table reports ordinary least squares regression estimation of the association between attributions and one-quarter-ahead analyst forecast revisions (FREV). All the variables are defined in Appendix A. T-statistics are shown in parentheses with standard errors clustered at both firm and quarter level. ***/**/* means significance at 0.01, 0.05 and 0.10 level, respectively.
8. Managers’ Attributions and Earnings Persistence Results in section 7 show that investors react negatively to managers’ internal attributions. In this section I examine whether this negative reaction is consistent with deteriorating future performance accompanied with managers’ attributions. Motivated by theory and prior research, I investigate whether deteriorating earnings persistence is a plausible explanation for the negative market response to unexpected internal attributions.
I examine earnings persistence because it is related to whether managers’ claims about performance drivers are truthful information or reflect their conscious or unconscious biases. 1 If managers’ attributions are truthful, we would expect to observe a positive association between internal attributions and earnings persistence. On the other hand, if managers’ attributions represent conscious bias for impression management purpose2, we would expect to observe a negative association between internal attributions and earnings persistence. If the attribution bias is due to cognitive Attribution theory suggests that an event is likely to reoccur or change according to whether the cause is going to remain or change (Weiner et al. 1971). Weiner et al. (1971) suggest that if an event is caused by a stable factor (for example, managerial ability) rather than an unstable factor (for example, efforts or luck), then the event is more likely to be persistent. In an annual report setting, Bettman and Weitz (1983) find that the locus of causality (i.e., internal vs. external attributions) and stability are closely related. They find that most internal attributions provided by managers are stable. In this paper, I assume that earnings innovations from internal and external factors have different persistence: internal innovations are more persistent and external shocks are more transitory.
2 CEOs have incentive to portray firm performance as better than it is. Prior research suggests that CEOs use corporate narratives to manipulate the impression of analysts and investors so as to be perceived favorably (Staw et al. 1983; Westphal and Graebner 2010).
factor such as overconfidence, then we may not observe an association between attributions and earnings persistence.
If a manager makes self-serving attributions to portray favorable firm performance as more persistent than it truly is, a rational market will unravel this and react negatively. Furthermore, one should observe less persistent earnings in the future.
To examine the relation between managers’ attributions and earnings persistence, I regress ROA in future periods (ROAt+1 and ROAt+4) on the interactive term of current period ROA and IvsE. The interaction term captures the relation between CEO attributions and earnings persistence. I also include other variables in equation (1) as additional explanatory variables. To control for the accrual effect documented by Sloan (1996), I include the absolute value of accruals (ABSACC) and its interactive term with
ROA in the earnings persistence tests. The final empirical specification is as follows:
Table 8 presents the results of OLS regression of equation (3). The coefficients on ROA*IvsE are mostly negative and statistically significant, in line with ROA being less persistent in the presence of internal attributions. The negative association between managerial attributions and earnings persistence suggests that CEO attributions are for impression management purpose rather than providing incremental information about managerial ability. The results also suggest that the negative market reaction reported earlier is rational.
To gauge the economic magnitude, I examine how changes of attributions affect changes of earnings persistence. A one standard deviation change in IvsE yields a change in one-quarter-ahead ROA of 0.029 (-0.102 * 0.28), which is comparable to the standard deviation of ROA (0.04). This suggests that the negative relation between managers’ attributions and earnings persistence is also economically significant.