The effect of manager-specific optimism on the tone of earnings conference calls

Studies on dispositional optimism often measure this construct using the Life Orientation Test (or LOT). Longitudinal studies have noted high correlations in LOT scores across time, consistent with optimism being a personality trait (Scheier and Carver 1993). In addition, studies of identical twins raised separately and together suggest a strong hereditary component to optimism/pessimism (Plomin et al. 1992).

The emotional Stroop task involves having subjects identify the ink color of a list of words that vary in emotional significance, ignoring the word’s meaning. Ignoring the meaning of words with high emotional significance is more difficult, leading to response latency in identifying the ink color (referred to as “interference”). This study examined the relation between interference associated with positive and negative words and subjects’ scores on the LOT test (discussed in footnote 1).

Because the presentation portion of the call generally consists of prepared remarks, which may or may not be read by the individual manager, it is possible that the CEO influences the remarks made by the CFO during the presentation portion of the call and vice versa. If so, we should use all comments made during the presentation rather than just the comments made by our manager of interest. We tested the sensitivity of our results to using the tone of the entire presentation transcript along with the tone of the specific comments made by the manager during the Q&A. Results are inferentially similar.

We start our sample period from 2002 because the CQ FD Disclosure database only provides earnings conference call transcripts for conference calls that occurred since 2001.

We use the variable “titlean” in Execucomp to identify the CFO of the firm. The following key words are chosen: “Chief Financial Officer,” “CFO,” “Vice President in Finance,” “VP Finance,” etc. Among all CEO and CFO job changes on Excecucomp from 2002 through 2009, 77 of them are at the CEO position in both firms they have worked for, 120 of them are at the CFO position in both firms, nine of them moved from the CFO position to the CEO position at the second firm, and three of them moved from the CEO position to the CFO position at the second firm.

Execucomp does not identify the exact month that a CEO or a CFO joins a company; therefore we read each conference call transcript to ensure that the name of the manager is mentioned in the transcript and this manager participates in the conference call. If not, we consider that firm-quarter observation as a “filler quarter.”

By construction, TONE_D, TONE_H, and TONE_LM are weighted averages of the net positive words used by managers in the presentation and Q&A portions of the call; that is, the sum of the net positive words used in the presentation and Q&A portions divided by the total number of words used in both portions of the call. In untabulated analysis, we find that the average net positive words during the presentation (Q&A) portion of the call is 1.67 % (1.34 %), 2.22 % (1.43 %), and 0.76 % (0.34 %) using TONE_D, TONE_H, and TONE_LM. Thus the average positive tone expressed by managers during the presentation portion of the call is generally higher than the average positive tone expressed during the Q&A.

Consistent with this conjecture, when we exclude MBE from the regression, the coefficient on SURP is positively associated with our tone variables and generally significant.

In untabulated analysis, we examined the relation between the tone of the entire presentation comments and current and future performance. In this analysis, the coefficients on one- and two-quarter-ahead ROA are significantly positive for all three tone measures, and the coefficients on three-quarter-ahead ROA are significant for the Henry and LM tone measures.

Similar to Dyreng et al. (2010), we find greater statistical significance for the manager fixed effects when we cluster standard errors at the firm level. However, as noted by Dyreng et al. (2010), testing the joint significance of the manager fixed effects is not advisable when clustering standard errors by firm. The rank of the variance/covariance matrix when clustering standard errors is equal to the minimum of either the number of clusters minus one or the number of independent variables that vary within clusters. Therefore testing a large set of restrictions associated with the manager effects requires inverting a matrix that is near singular, resulting in overly large F-statistics.

We also examine the correlations among manager fixed effects based on the three different tone measures. The average Spearman correlation is 0.62, suggesting that manager-specific optimism measured using the three dictionaries are similar, although not perfect substitutes.

Bamber et al. (2010) does find some evidence on the impact of military and educational background on management forecast characteristics, although results vary across the different forecast properties. However, because this study examines a variety of forecast properties—including frequency, accuracy, bias, precision, and direction of news—the underlying cognitive characteristics associated with their measures of style are likely varied (e.g., optimism is not likely the cognitive characteristic driving manager-specific forecast accuracy). The variable most closely related to our tone measure is forecast bias. They find one manager-specific characteristic associated with forecast bias: having a finance or accounting background. Thus the evidence of Bamber et al. (2010) is equally limited.

It is difficult, in an archival setting, to distinguish between the effects of optimism, overconfidence, and risk-aversion. Any of these three cognitive characteristics may lead to the consistent use of more optimistic language in conference calls. The psychology literature defines overconfidence as one’s “assuredness” about one’s own judgments (Reber 1995). Thus, at the construct level, overconfidence is “nondirectional,” i.e., one could be overconfident about the possibility of negative outcomes for the firm. However, one might argue that managers generally have positive expectations of the firm and therefore, empirically, overconfident managers will use more positive words. We believe it is beyond the scope of this paper to distinguish between these three underlying cognitive characteristics.

For example, it is not clear how investment banking experiences are associated with managers’ cognitive characteristics. On one hand, investment bankers may be more optimistic due to the sales nature of their job. On the other, these managers (especially those who have worked as financial analysts in investment banks) are likely highly aware of the negative consequences of disappointing investors (e.g., reduced institutional ownership or legal costs).

There are 212 CEOs and 325 CFOs in the sample, with 11 managers who worked as both a CEO and a CFO (thus 212 CEOs + 325 CFOs − 11 duplicates = 526 distinct managers).

We note, however, that we can confirm prior findings that the market reacts to tone using our sample of conference calls. Specifically, we regress two-day (t, t + 1) value-weighted market-adjusted returns centered around the conference call date on TONE_i ^ALL and all control variables from Eq. (1). Consistent with prior research, we find a positive and statistically significant coefficient on TONE_i ^ALL (using all three measures of TONE), suggesting that the market reacts to the portion of tone that is unrelated to current and future realized performance and firms’ strategic incentives. We also examine whether tone is negatively associated with future stock returns (suggesting that investors correct their initial pricing error). Specifically, we regress value-weighted market-adjusted returns (t + 2, t + 121) on our tone measures and control variables. We find that, across all three tone measures, the coefficient on manager’s tone is significantly negative, consistent with unwarranted optimistic tone misleading investors at the time of conference call dates. These results are consistent with Huang et al. (2014) that provide evidence of price reversals associated with the tone in earnings announcement press releases.