The effects of risk factor disclosure on analysts’ earnings forecasts: evidence from Chinese IPOs

ABSTRACT

Using Chinese IPO firms, we examine the impact of risk information disclosure on the properties of financial analysts’ earnings forecasts. We construct a disclosure index by manually reading the disclosures of risk factors in IPO prospectuses. We find that high-quality risk factor disclosure is associated with lower earnings forecast bias, less dispersion, and more timely forecasts. Separating total risk factors into five different types, we show that analysts have better ability to process financial risk and operational risk disclosure compared to the other three types (technology-related risk, market risk, and macroeconomic risk). Cross-sectional tests reveal that the positive effect of risk disclosure is mainly concentrated in firms with greater information transparency (e.g., larger firms, firms with higher profitability, and firms with lower performance volatility). We further rule out other potential explanations (e.g., the effect of analyst sentiment) and conduct additional robustness tests. Our study has broad implications for firms in emerging economies.

KEYWORDS:

• Qualitative information disclosure
• risk factors
• analyst forecasts
• forecast bias
• forecast dispersion

Acknowledgments

We thank Ke Wang, Stephen Penman, Isabel Yanyan Wang for their helpful comments. We also thank the participants of the 2018 AAA International Accounting Section Midyear Meeting, and the International Academic Forum Nankai-Birmingham (UK) at Nankai University. Yi Yao acknowledges financial support from National Natural Science Foundation of China Projects (NSFC No. 71672090), Nankai University 100 Young Academic Leaders Program, and the Five First-Batch Social Science Talent Program in Tianjin. Part of the research was conducted when Yi Yao was a visiting scholar at Columbia Business School; she appreciates helpful discussions with the accounting faculty at Columbia University and New York University. Any errors that remain are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Filzen (2015) finds that firms presenting updates to their risk factor disclosures have significantly lower abnormal returns around the filing date of the 10-Q relative to firms without updates. Similarly, Chiu, Guan, and Kim (2018) suggest that risk factor disclosures in 10-K/10-Q filings reduce the information risk premium in credit default swap spreads.

2. We focus on financial analysts because as sophisticated investors, they are more likely to incorporate risk information into their decision-making and earnings forecasts (Demirakos, Strong, and Walker 2004; Hope, Hu, and Lu 2016).

3. While some studies find evidence that risk factor disclosure is informative (e.g., Gupta and Israelsen 2014; Hanley and Hoberg 2010; Heinle and Smith 2017), other studies (Kravet and Muslu 2013; Balakrishnan and Bartov 2011) suggest that risk factor disclosure only increases investors’ risk perceptions without resolving uncertainty. We further discuss the literature in Section 2.

4. One might argue that without the benefit of reducing future earnings volatility, what is managers’ motivation of disclosing an abnormal amount of risk information? We provide one potential explanation from the perspective of capital market relationships (i.e., manager-analyst relationship), please see more details in Section 5.1.

5. Early studies find that the disclosure of risk factors associated with financial assets and liabilities is informative to investors (Rajgopal 1999; Jorion 2002; Linsmeier et al. 2002). More recently, using textual analysis to measure risk disclosure by word counting or by computing the proportion of risk-related keywords, Campbell et al. (2014) and Kravet and Muslu (2013) conduct large sample studies to examine the informativeness of firms’ risk factor disclosures.

6. ‘2015 Chinese securities research industry report’ was published by New Fortune Magazine. (See http://finance.jrj.com.cn/2015/11/28134520151663.shtml).

7. While computer-based textual analysis gained its popularity recently, early studies of voluntary disclosure typically construct disclosure indexes by reading annual reports or IPO prospectuses and adding the number of disclosed items (e.g., Beatty and Welch 1996; Botosan 1997). Under this approach, the higher the disclosure index, the better the level of disclosure. Recent studies try to better, and more objectively, quantify the level of disclosure. For example, Arnold, Fishe, and North (2006) and Deumes (2008) count the number of words in the whole IPO prospectus, risk factor section, or several specific sections (e.g., abstract, capital allocation, and MD&A) and calculate the percentage of risk-related words to proxy for the level of risk disclosure. Lajili and Zéghal (2005) use key word recognition to study risk disclosure in Canadian public companies.

8. Prior studies generally focus on the amount of disclosure (e.g., the existence and the frequency of management earnings forecasts (Healy and Palepu 2001). Li’s (2008) is the first study to examine the readability of firms’ annual reports by using textual analysis, followed by many other studies. While textual analysis has gained popularity in recent literature, its application in Chinese language is still immature, which is another reason why we manually read the prospectuses.

9. Inconsistent with previous literature, Miihkinen (2012) finds that poorly performing firms outperform more profitable firms in risk disclosure.

10. We do not include Timelag when the dependent variable is Timeliness. As they are both heavily influenced by analyst’s forecasting date, more timely Timeliness will result in longer Timelag.

11. In particular, the most prominent change in IPO regulations issued by the Shanghai and Shenzhen Stock Exchange in 2014 is that the maximum IPO underpricing on the first trading day is 44%. Thus, expanding our sample after 2014 will make our main test on the underpricing variable no longer valid and incomparable with the previous period. We also examine the relationship between risk information and post-IPO stock return volatility by extending the sample from 2014 to 2016, and we find that our results are unchanged.

12. Following prior literature (Clement and Tse 2005), if an analyst issued several earnings forecasts on the same company in a year, we only keep the first forecast. This is because follow-up earnings forecast revisions are more likely to be affected by information outside of IPO prospectuses.

13. As analyst characteristics should not affect the firm-level dependent variable, forecast dispersion, we only consider the impact of analyst ability on forecast bias and timeliness, so as Panel B in Table 7.

14. In this part, considering analyst-level sentiment has no effect on the forecast dispersion among different analysts, we do not examine the sentiment effect when the dependent variable is dispersion (Dispersion) in a firm-level model.

15. In China, since all retail and institutional investors, not just those clients of the lead underwriters, have significant access to IPO allocations, the information asymmetry in the premarket for IPOs is more severe, and analysts’ research is in strong demand.

16. The Fog index is defined by Li (2008) as (words per sentence + percent of complex words)×0.4.

Additional information

Funding

This work was supported by the National Science Foundation of China Projects (71672090), Nankai University 100 Young Academic Leaders Program.