Analysts’ forecast anchoring and discontinuous market reaction: evidence from China

Abstract

Based on a comprehensive dataset of Chinese A-share stocks, we find significant evidence for analysts’ anchoring in China. This anchoring behavior exists only when forecasts are lower than the industry median and correlates to a series of analysts’ and corporate characteristics. The following market reactions to analyst anchoring are discontinuous: only forecasts without analysts’ anchoring achieve higher and positive returns in the next period, while those with anchoring bias have no influence on the following returns. Further analyses based on the earnings response coefficient reveal that analysts’ anchoring inhibits earnings information dissemination, and discontinuous stock performance is mainly from professional institutions. Overall, this paper provides evident results about analyst anchoring and its impact, and the findings are beneficial for investors to efficiently use analysts’ forecast information.

JEL:

• G14
• G12
• G11

KEYWORDS:

• Analysts’ forecasts
• anchoring
• market reaction
• company credit rating
• forecast error
   

Acknowledgments

Financial support from the National Natural Science Foundation of China (72001033, 72141304, 71901160), Key R&D Project of the Ministry of Science and Technology (2022YFC3303304), the Fundamental Research Funds for the Central Universities (DUT23RW105), and China Scholarship Council is greatly acknowledged. All remaining errors are the authors’.

Disclosure statement

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

Notes

1 We thank the anonymous referee’s suggestion.

2 All Chinese firms end their fiscal year on December 31, but the deadline for an annual report is April 30.

3 Stocks with low company credit ratings obtain lower returns and induce low firm visibility (Ashour and Hao 2019; Hao and Li 2021).

4 $\mathrm{CAF}$ represents the changes between analysts’ forecasts and the industry median value, which is different from $\mathrm{FE}$ that measures the difference between forecasted and real earnings per share. The correlation between $\mathrm{CAF}$ and $\mathrm{FE}$ is 0.05 and significant at the 1% level.

5 Based on the definition of analysts’ forecast anchoring, forecasts in quintile 3 have a value similar to the industry median, and those in quintiles 4 and 5 (quintiles 1 and 2) have values higher (lower) than the industry median.

6 We only report the significant levels of the difference between firms with the lowest and highest $\mathrm{CAFs}$ in (5-1) in Panels B and C of this table, considering the comparison of these features to zero to get the t-values is meaningless. Besides, we thank the anonymous referee’s suggestion to add a discussion about the economic significance of our findings.

7 We also study the correlation and relationship between analysts’ characteristics and $\mathrm{CAF}$ based on the Fama–Macbeth cross-sectional regressions. The results are consistent with Table 1, and we do not report them for brief. The results are always available upon request.

8 In Table 1, the value of $\mathrm{CAF}$ in quintile 3 is −0.001 close to zero, which means the forecasts in quintile 3 have a value similar to the industry median according to the definition of $\mathrm{CAF}$.

9 For firms with forecasts below the industry median, the following returns are uninfluenced by the forecast values, which might be because investors recognize overvalued facts during long trading histories and tend to ignore these forecasts in their trading.

10 The different market reactions to stocks forecasted by analysts with and without anchoring behavior might be due to heterogeneous investor trading preferences. During long-term trading, investors might have recognized the analysts’ anchoring behavior, and based on this knowledge, investors tend to allocate more (less) trading for stocks with forecasts above (below) the industry median and lead to the related higher (uninfluenced) stock returns. We test the trading volume of stocks with different $\mathrm{CAF}$ values to analyze the effect of anchoring on investor trading to prove this intuition. Specifically, we separate stocks into five quintiles (two groups) based on the value of $\mathrm{CAF}$ each month and get the average trading volume of each group in the next month. The results are in Table A.4. The significantly positive difference between stocks with positive (highest) and negative (lowest) $\mathrm{CAF}$ indicates that investors tend to trade stocks with accurate analyst forecasts and ignore those with biased forecasts, which is consistent with the discontinuous market reaction in Tables 4 and 5. In addition, we perform a Fama–Macbeth cross-sectional regression after controlling for potentially related factors to further verify analysts’ anchoring on trading volume, and the findings are consistent with the portfolio-level evidence. The results further prove our Hypothesis 2 that stocks with higher $\mathrm{CAF}$ values are favored by investors and have higher trading volume. This trading behavior is highly related to the analyst forecast information and could provide a potential explanation for the following asymmetric market reactions.

11 To be consistent with the dependent and main independent variables, for control variables $\mathrm{Size}$, $B/M$, $\mathrm{Lev}$, $\mathrm{Illiq}$, $\mathrm{Beta}$, and $\mathrm{AG}$, data frequency is year and for other control variables, data frequency is month.

12 We guess this result might be from the fact that the investment quota restriction for QFII/RQFII was released in 2019, and the subsamples before and after this policy are unbalanced (there are only two years after this policy and the influence is limited). Even if the results are contrary to our intuition considering the data limitation, we think it is important to reveal the entrance of more QFII/RQFII in our findings, and we are grateful for the comments of the anonymous referee. To be brief, we do not report the detailed results of this part in this manuscript, and the results are always available upon request.

13 Specifically, $\mathrm{CAF}$ ($\mathrm{CA}{F}_{\mathrm{median}}$) measures the difference between the mean (median) value of analysts’ forecasts and the analyst consensus forecast. Theoretically, the mean value is more susceptible to extremes than the median value, and $\mathrm{CA}{F}_{\mathrm{median}}$ is different from $\mathrm{CAF}$, which is based on an adequate number of samples.

14 In our sample, the average number of analyst forecasts issued per month for each company is 3.27, the max value is 50, and the correlation coefficient between $\mathrm{CAF}$ and $\mathrm{CA}{F}_{\mathrm{median}}$ is 0.996 and significant at the 1% level.

15 Besides, when we consider the significance of the differences in stock feature variables between quintiles with the highest and lowest $\mathrm{CAFs}$, we find that most differences are significant at the statistical and economic levels. For example, the difference in $B/M$ between the two quintiles is −0.06 with an associated t-statistic of −7.39, which is relatively large at the economic level considering the mean value of $B/M$ is 0.61 in the whole sample (reported in Appendix Table A.2).

16 We try to keep the consistency of control variables when we use the same dependent variables to do the regression, however, the control variables for studies on different dependent variables might be slightly different. We thank the anonymous referee’s comment.

17 For example, the Global Analyst Research Settlement in the early 2000s. Global Analyst Research Settlement requires brokerages to set up an independent research department to isolate it from other departments, reduce the pressure and influence of investment banking and other business departments on analysts, and require that analyst compensation depends on the quality and accuracy of research reports.

Additional information

Funding

Financial support from the National Natural Science Foundation of China (72001033, 72141304, 71901160), Key R&D Project of the Ministry of Science and Technology (2022YFC3303304), the Fundamental Research Funds for the Central Universities (DUT23RW105), and China Scholarship Council is greatly acknowledged.

Notes on contributors

Ruixin Fan

Ruixin Fan is a PhD student in College of Management and Economics at Tianjin University in Tianjin, China and her main research interests include asset pricing, investors' behavior, and green finance.

Xiong Xiong

Xiong Xiong is a Professor in College of Management and Economics at Tianjin University and is a director of Laboratory of Computation and Analytics of Complex Management Systems (CACMS) in Tianjin, China and his main research interests include asset pricing, agent-based modeling, computational method, financial risk, and financial management.

Youwei Li

Youwei Li is a Professor in the Hull University Business School in UK and his main research interests include asset pricing, financial econometrics, heterogeneous agent models of financial markets, and quantitative finance.

Ya Gao

Ya Gao is an Associate Professor in School of Economics and Management at Dalian University of Technology in Liaoning, China and her main research interests include asset pricing, investors' behavior, financial market micro-structure, and green finance.