Bias and Predictability in Analysts' Beliefs
A new study finds financial analysts’ forecasts are not only biased but follow a U-shaped pattern. Extreme optimism or pessimism predicts higher short-term returns, while moderate bias leads to lower returns. Using machine-learning benchmarks and a vast US dataset, the research shows analysts mostly stick to consensus, rarely acting independently. The asset-pricing model links consensus anchoring and skewed stock shocks to these patterns. The study suggests analysts’ underreaction to new information can distort market prices.
What it examines
This paper studies how financial analysts' forecast biases relate to future stock returns by comparing their predictions to machine learning benchmarks. It analyzes price targets, earnings, and long-term growth forecasts, uncovering patterns of bias and consensus anchoring, and develops a model to explain these effects in asset pricing.
What it concludes
The research finds that forecast bias predicts returns differently over time and is shaped by analysts' tendency to follow consensus. These insights can improve investment strategies, risk management, and financial modeling. Future work could use text analysis to study how analysts' reasoning and explanations influence market expectations and consensus formation.
Evidence objects
Financial analysts forecast biases show a U-shaped pattern: extreme optimism or pessimism predicts higher short-term returns, but in the long run, these same biases lead to lower returns.
key_findings bullet 1 · key_findings · validation V0
The study introduces innovative methods, measuring bias against machine-learning benchmarks and classifying analyst behavior as 'conformism' or 'contrarianism,' revealing analysts overwhelmingly anchor predictions to consensus and rarely challenge the crowd.
key_findings bullet 2 · key_findings · validation V0
An asset-pricing model combining consensus anchoring and positively skewed stock shocks explains analysts underreaction to new information, but further research is needed on why some analysts break consensus and international applicability.
key_findings bullet 3 · key_findings · validation V0
This paper uniquely compares analyst forecasts (PTG, EPS, LTG) to machine-learning benchmarks, revealing novel nonlinear (U-shaped) bias-return relationships and systematic anchoring on consensus. Its joint analysis across forecast types and modeling of anchoring behavior advances both academic understanding and practical trading, offering compelling, original insights beyond prior literature.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- … To directly test whether bias and returns are non-monotonically related, I examine whether the return predictability of bias in PTG and LTG varies with the forecast horizon …
Source row: 314 · abstract type: snippet