Peer-based performance comparison and tone management

ABSTRACT

While performance comparison against peers is documented as a key decision-making factor in strategic management, its potential role as a strategic determinant of disclosure tone has so far been overlooked in the literature on narrative disclosures. In this study, we examine whether, and how, peer-based performance comparison (i.e. the comparison of earnings/cash flows of a firm with its peers) affects tone management. Using a measure of tone management based on 10-K filings over 1993–2013, we show that firms with performance falling below the peer-based benchmark engage in greater downward tone management in their earnings-related narrative disclosures, and the extent of this over-pessimism increases the further performance falls below the benchmark. We further document that the observed over-pessimism not only prompts security analysts to revise their earnings forecasts downward, but it is also associated with a higher probability of meeting or beating these forecasts (i.e. expectation management). This is the first study that links tone management with expectation management, in the context of peer-based performance comparison.

KEYWORDS:

• Tone management
• peer-based performance comparison
• narrative disclosures
• performance feedback
• expectation management

JEL CLASSIFICATION:

• G3
• M41
• M51

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

¹ The component of tone obtained after controlling for these economic variables is discretionary (or abnormal) tone, which is a tool that managers may strategically employ to inform or mislead investors about future firm performance (Huang, Teoh, and Zhang 2014).

² These scenarios are as follows: (i) meet or beat vs. miss consensus analyst forecast, (ii) high vs. low earnings, (iii) profit vs loss, and (iv) positive vs. negative earnings change.

³ This conjecture is plausible knowing that the decision about narrative disclosure tone is subject to managerial discretion, and implementation of greater downward tone management subsequent to their performance shortfall to manage analysts’ expectations is thus an easier and less costly tool than taking real or tangible actions, such as altering R&D investments (O’brien and David 2014; Deb et al. 2019) to adjust performance.

⁴ Narrative disclosures constitute an ‘unregulated window’ because there are minimal explicit rules regarding the content and format of these disclosures, as words are much more elastic than numbers and thereby more difficult to regulate (Merkl-Davies and Brennan 2007; Henry 2008).

⁵ Earning, earnings, EPS, income, loss, losses, profit, profits, sales, revenue, revenues, expense, expenses, EBT, EBIT, EBITDA.

⁶ The mean (median) number of firms within a peer group with available data to calculate EARNINGS equals 39 (26).

⁷ For simplicity, reported findings are based on the benchmark derived from EARNINGS, while untabulated results based on CFO are not materially different.

⁸ This argument also holds credence in the case of quantitative disclosures, as managers have been found to execute their discretion in financial reporting to affect information asymmetry between them and investors with regard to their firms (Subramanyam 1996).

⁹ A caveat here is that we do not obtain significant results when we split our main sample into subsamples of firms with earnings below (BELOW = 1) or above (BELOW = 0) the peer-based benchmark.

¹⁰ The analysis of the marginal effects of ABTONE_EARN in the case of outperforming firms reveals that a one unit of change in PEERCOMP × ABTONE_EARN increases the likelihood of meeting or beating analysts’ forecasts by .0075% point. This relatively low increase in the likelihood (compared to underperforming firms) feeds our doubt about considering upward tone management in the case of outperforming firms as a tool to increase the likelihood of meeting or beating analysts’ forecasts.

¹¹ The univariate correlation between analyst forecast error (AFE) and BELOW is −.0049 (p-value equal to .3281). Hence, we did not use AFE as instrument in the first-stage regression model, despite that (untabulated) results with AFE being included in the set of instruments are not materially different. Also, untabulated results using only LOSS as instrument in the first-stage regression model are qualitatively similar.

¹² With respect to last year earnings as performance expectation, PERF equals reported earnings in year t minus earnings in year t-1 scaled by beginning total assets. Obviously, this variable is the same as HISTCOMP and that is why HISTCOMP is omitted in both columns (1) and (2). With respect to analysts’ forecasts, PERF equals reported EPS minus analyst consensus, scaled by beginning stock price.

¹³ The weighting scheme was modified to additionally adjust for document length [similar to Loughran and Mcdonald (2011)] and account for the variation in length over time, since the 10-K has become significantly lengthier over time and it is more likely for a word appearing in 1993 to have a different impact than a word appearing in 2010 (my sample period is from 1993–2010).

¹⁴ To better gauge the economic significance of ABTONE_EARN, we use annual decile ranks for ABTONE_EARN in all the equations with ABTONE_EARN as independent variable.