Does equity analyst research lack rigour and objectivity? Evidence from conference call questions and research notes

Abstract

Doubts have been raised about the rigour and objectivity of sell-side analysts’ research due to institutional structures that promote pro-management behaviour. However, research in psychology stresses the importance of controlling for biases in individuals’ inherent cognitive processing behaviour when drawing conclusions about their propensity to undertake careful scientific analysis. Using social cognition theory, we predict that the rigour and objectivity evident in analyst research is more pronounced following unexpected news in general and unexpected bad news in particular. We evaluate this prediction against the null hypothesis that analyst research consistently lacks rigour and objectivity to maintain good relations with management. Using U.S. firm earnings surprises as our conditioning event, we examine the content of analysts’ conference call questions and research notes to assess the properties of their research. We find that analysts’ notes and conference call questions display material levels of rigour and objectivity when earnings news is unexpectedly positive, and that these characteristics are more pronounced in response to unexpectedly poor earnings news. Results are consistent with analysts’ innate cognitive processing response counteracting institutional considerations when attributional search incentives are strong. Exploratory analysis suggests that studying verbal and written outputs provides a more complete picture of analysts’ work.

Keywords:

• capital markets
• textual analysis
• analyst reports
• conference calls

Acknowledgements

We are grateful for very helpful comments from two anonymous reviewers whose suggestions significantly improved the paper. Comments and suggestions were also provided by Kathleen Andries, Nathalie Beckers, Robert Bloomfield, Mark Clatworthy and Edward Lee (the editors), Simon Dekeyser, Daniel Giamouridis, Tom Guilmet, Katlijn Haesebrouck, Simon Hayley, Gilad Livne, Peter Pope, Richard Taffler, Christophe Van Linden, Sofie Verbieren, and seminar participants at the 2011 British Accounting and Finance Association Doctoral Colloquium, Athens University of Economics and Business, Cass Business School, Exeter University, Katholieke Universiteit Leuven, Lancaster University, and University of Southampton.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Related work examines analyst participation in the Q&A portion of the call. Findings suggest that participation is determined both by analysts’ level of private information and managers’ desire to avoid unfavourable questions (Cohen et al. 2013, Mayew et al. 2013).

2. Matsumoto et al. (2011) examine conference call transcripts and find the Q&A segment is more informative when performance is poor. However, because their tests aggregate analysts’ questions with managers’ responses, it is unclear whether this result is due to analysts seeking out more information or management voluntarily disclosing more information. Our evidence speaks directly to analysts’ information acquisition activities.

3. Although studies show that analysts revise key summary outputs in response to earnings surprises (Abarbanell and Bernard 1992, Yezegel 2015), this evidence does not speak directly to our research question for two reasons. First, revisions in summary outputs may be driven by factors other than attributional search behaviour (Altinkilic and Hansen 2009). Second, summary outputs such as earnings forecasts and investment recommendations yield limited and inconsistent insights about analysts’ research activities (Schipper 1991, Bradshaw 2011).

4. Analysts may view unexpectedly favourable earnings with particular scepticism, leading to higher levels of attributional search aimed at determining if reported results are persistent. This effect will work against our prediction that analysts focus more on future threats and weaknesses in response to poor earnings news.

5. A decline in management’s reputation and perceived competence may also threaten firm value through a higher cost of capital resulting from increased information risk (Barton and Mercer 2005). We view this potential discount rate effect as part of the overall rise in uncertainty discussed above.

6. Contrary to the view that written research is more visible, conference calls may have higher permanence because they are available for web viewing and their contents are transcribed. These features could cause managers to censor their comments during the call. However, research demonstrates that managers who refuse to answer questions or who adhere to scripts during the Q&A are perceived negatively (Hollander et al. 2010, Lee 2016), implying that investors value the spontaneity and unexpected content of the Q&A section.

7. We focus on the sign of the surprise because social cognition theory offers no clear prediction concerning the impact on attributional search of the magnitude of the earnings surprise, while evidence on the market reaction to NES suggests complex non-linearities (Kinney et al. 2002). Untabulated tests examined whether analysts’ response to NES is conditional on the magnitude of the surprise but the results provided no support for an interaction effect.

8. Negative prospects are distinct from negative tone studied by Chen et al. (2015). Tone applies to backward- and forward-looking discussions whereas prospects are exclusively forward-looking.

9. Differences in the approaches used to code research notes and conference call questions means that direct comparisons between analysts’ written and verbal responses should be interpreted with caution. We address this issue in a later section by constructing content metrics designed to permit direct comparison across modalities.

10. We do not condition UNCERT^RN on negative tone in our main tests because uncertain language is more unequivocally negative in written format. For example, the sentence ‘will revenue growth achieve target levels?’ in a research note would imply downside uncertainty, whereas the same question posed in a Q&A setting does not automatically imply downside risk. In supplementary tests described below we construct a conditional measure of uncertainty for research notes. Results are not materially different using this metric.

11. An individual analyst may not view an announcement as a negative surprise (a) when the firm achieves the analyst’s individual forecast but misses the consensus, (b) where the street consensus differs from the IBES consensus or (c) where a firm pre-announces disappointing earnings news after the last IBES consensus date.

12. Our sample window post-dates Regulation Fair Disclosure and rules arising from the Global Analyst Research Settlement to reduce the impact of analyst optimism (Hovakimian and Saenyasiri 2010). As a consequence our conclusions are silent on the impact of these regulatory changes on the properties of analysts’ research.

13. GEE and random effects analysis are the two most commonly applied methods for analysing data where repeated measurements for an individual or firm are correlated. GEE provides unbiased estimates of regression coefficients and variances without needing to specify the correct covariance structure (Hardin and Hilbe 2003, Twisk 2004). Choice of specific working correlation structure for GEE estimation is irrelevant for matched pairs data because all non-identity structures produce the same result (Liang and Zeger 1986).

14. Results for research notes display skewness: the median percent of management-related discussions that are negatively toned in the representative report is zero for both surprise partitions.

15. The majority of non-NES are positive schema-discrepant: 93.5% (94%) of the MBE sample are positive based on the mean (median) consensus forecast.

16. Our research design also creates interpretational problems. Ideally we would have compared comments for the same analyst across alternative modalities. Unfortunately, this is not possible because many conference call transcripts published before 2008 do not identify analysts by name.

17. An important caveat associated with this approach is that by constructing a metric that is more comparable across modalities we risk compromising statistical power gained from using format-specific metrics that reflect fundamental differences in content and style. If the reduction in power affects these outputs differentially, then observed variation in relative strength will be driven by statistical biases rather than fundamentals. An alternative way of comparing effects is to compute standardised regression coefficients for regression models in Tables 2–4. Unfortunately, interpreting standardised coefficients for indicator variables such as D_NES is problematic because a one standard deviation change is not meaningful for binary variables.

Additional information

Funding

Financial support was provided by the Economic and Social Research Council (studentship PTA-031-2006-00412 and contracts ES/J012394/1 and ES/K002155/1). A previous version of this paper was circulated under the title ‘Valuation implications of negative earnings surprises: evidence from analyst research reports’.