Abstract
This paper formulates a two-stage model to capture the decision process of financial analysts when issuing earnings forecasts. Our model extends the model of Chen and Jiang [(2005). Analysts’ weighting of private and public information. Review of Financial Studies, 19 (1), 319–355], by allowing for a distortion of forecasts independent of whether an analyst has private information. Using quarterly earnings forecasts, we provide empirical evidence on the coexistence of overconfidence and strategic incentives. Financial analysts overweight their private information and at the same time strategically inflate their forecast.
Acknowledgements
We thank Ester Eiling, Daniel Kohlert, Ralph Koijen and Bertrand Melenberg for comments that improved the paper. Furthermore we thank seminar participants from the K.U.Leuven and Tilburg University and conference participants at the 2010 MFA, the 2010 FMA-European and FMA-Annual Meeting, the 2010 EFMA meeting and the Swiss Financial Markets conference.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes
1. We also estimate the model on the full sample of positive and negative forecasts, and results remain unchanged.
2. As analysts are reluctant to issue negative forecasts, stocks with such very negative future prospects are possibly dropped from coverage (see McNichols and O'Brien Citation1997).
3. A full derivation can be found in Appendix 1.
4. As a robustness check we also estimated a model that includes the deviation from the median consensus and we obtain similar results. These estimation results are available upon request.
5. We thank one of the referees for pointing this out.
6. Deleting negative forecasts implies a reduction in the data set of 11% from 362,040 to 322,123 forecasts. A sample of only positive forecasts still implies a symmetric loss function of forecast errors.
7. For completeness, Section 5 reports the estimation results of our model applied to the subsample of negative earnings forecasts. These findings are in line with our main conclusions. Also a full sample analysis, including both positive and negative forecasts was performed with similar outcomes.
8. Adding quarter dummies for a possible fixed time effect or adding industry dummies for a possible fixed industry effect leads to similar results. All conclusions remain the same.
10. Also Green et al. (Citation2009) and Kumar (Citation2010) are able to match approximately 95% of the observations with gender.
11. National Association of Securities Dealers.
12. SRO stands for Self-Regulatory Organization Rulemaking. These new rules target research analyst conflicts of interest and aim to promote greater independence of research analysts. To this end, actual and potential conflicts of interest need to be disclosed to investors. See Securities Exchange Act Release No. 45908, 67 FR 34968 (16 May 2002). For a detailed explanation see http://www.sec.gov/rules/sro/34-48252.htm.