Do successive performance matter for tone management? Evidence from China

ABSTRACT

This paper investigates whether managers deliberately adjust the tone of their non-financial disclosure depending on their firm performance over multiple periods. Using 13,995 firm-year MD&As (Management Discussion and Analysis) in annual reports of companies listed on the Shanghai and Shenzhen Stock Exchanges from 2009 to 2019, we first observe a bias towards good news disclosure by managers in their MD&A section. More importantly, we find that this asymmetric tone management is more pronounced when a firm’s performance declines consecutively. In cross-sectional tests, we provide evidence that the relationship between asymmetric tone management and successive performance is weaker for firms with state ownership and those with a high marketability index. We suggest that in non-U.S. markets, such as China, asymmetric tone management is more likely to reflect the opportunistic behaviour of managers. Overall, our study sheds new light on tone management in emerging markets and highlights the importance of a multi-period relation in understanding managers’ tone management behaviour.

KEYWORDS:

• management discussion and analysis (MD&A)
• tone management
• non-U.S. settings
• reputation management
• multi-period relation

JEL CLASSIFICATION:

• G10
• M40

Disclosure statement

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

Data availability statement

Data used in this paper are available from public sources identified in the paper.

Notes

1. In this paper, we interchangeably use tone management, asymmetric tone management, and tone asymmetry to indicate disclosure behaviour by managers associated with unusual positive tone in the MD&A section.

2. In this paper, we use ‘multi-period’ or ‘multiple periods’ to indicate two or three consecutive fiscal years.

3. Python program is used for text mining and sentiment analysis.

4. Return on asset (ROA) and return on equity (ROE) is used as a proxy for performance. See Section III for details.

5. Disclosure tone is estimated following the methodology of Loughran and McDonald (2011). See Section III for details.

6. It is important to note the limitations inherent in our approach to sentiment analysis using Chinese sentiment dictionaries. While we have endeavoured to align our methodology with that of existing studies (e.g. Bian et al., 2021; Shan, 2019), the adaptation and integration of multiple sentiment dictionaries may introduce variations in tone assessment due to differences in the scope, definitions, and classifications of terms across different dictionaries. We caution readers to consider these potential limitations when interpreting our findings.

7. W. R. Chen (2008) develops a behavioural forward-looking search model using different weights between the actual prior-year firm performance and the expected prior-year firm performance from zero to one in increments of 0.1. The results based on 0.6 give the highest model log-likelihood. Our inferences are unchanged to different weights between the actual prior-year firm performance and the expected prior-year firm performance.

8. We check the robustness of our main findings using alternative tone asymmetry model in D’Augusta and DeAngelis (2020b) and find a similar inference (Section 4.2).

9. We expect no difference in the asymmetric tone disclosure between rational and irrational managers in cases of superior performance (i.e. [Case 1] – [Case 4] in Figure 1). In the case of a continuous increase in performance ([Case 1], [Case 2] in Figure 1), we expect both rational and irrational managers to deliver good news to the capital market. The extent to which good news is emphasised might be stronger than in the case of unfavourable performance observed in the previous year ([Case 3], [Case 4] in Figure 1). In short, we expect γ₁^PGood and γ₁^PBad to be positive by an equal degree between rational and irrational managers. Given current bad news, on the contrary, we anticipate that managers will under-disclose ([Case 5] – [Case 8] in Figure 1). When the current period has fresh bad news ([Case 5], [Case 6] in Figure 1), we expect γ₂^PGood to be negative; again, there is no theoretical distinction between rational and irrational managers. However, our second hypothesis gives rise to a distinctive prediction of the γ₂^PBad estimate ([Case 7], [Case 8] in Figure 1). A significantly negative estimate of γ₂^PBad indicates that negative tone is suppressed or avoided when performance deteriorates, which means that the asymmetric tone results from managers’ irrational decision-making ([Case 7] in Figure 1). If γ₂^PBad is positive, then we can interpret the asymmetric tone in the opposite direction due to rational decision-making ([Case 8] in Figure 1).

10. All listed firms’ annual report is uploaded on the Juchao Information Network (http://www.cninfo.com/cn).

11. We note that our findings are robust to a three-year performance model, alternative benchmarks for assessing bad news or good news, and alternative models for estimating the relation between firm performance and tone management (D’Augusta & DeAngelis, 2020b). See Section 4.2 for details.

12. The marketisation index comes from the book China Marketisation Index published by the Beijing National Economic Research Institute (https://cmi.ssap.com.cn/teamIntroduction), one of the first non-governmental, non-profit research institutes on private economic theory and policy in China.

13. In Table 2, we find that managers are most sensitive to changes in industry level ROE. Following this line, the results in Columns (7) and (8), where the benchmark for good and bad news is measured by expected industry level ROE, can be interpreted as even managers with less motivation for opportunistic behaviour are likely to engage in asymmetric tone management following successive declines in industry level ROE.

14. One drawback of this measure is that we cannot include firms without analyst coverage.

15. For variable definitions, refer to Appendix.

Additional information

Funding

Iny Hwang appreciates financial support from the Institute of Management Research at Seoul National University.