Academic Research Digest

Research evidence excerpt

Academic Research Digest

28 July 2026 Citi Research

rolling regressions, to isolate bond-market-specific pricing from equity-driven

variation. To test the payoff curvature hypothesis, the regressions include

interactions of the disagreement measure with indicators for high idiosyncratic

skewness (equity convexity proxy) and high expected default frequency or non-

investment-grade status (bond concavity proxies). To test the limits-to-arbitrage

channel, the regressions interact disagreement with illiquidity indicators

constructed from the Amihud (2002) measure and turnover for stocks and the Bao,

Pan and Wang (2011) gamma measure for bonds. Long-horizon tests extend the

forecast horizon from one month to twelve months to trace the persistence and

decay of the pricing effect.

The primary measure of investor disagreement is the cross-sectional standard

deviation of analyst EPS forecasts for the upcoming fiscal year scaled by the

absolute value of the consensus mean forecast, sourced from I/B/E/S. Stock return

data come from CRSP and accounting data from Compustat, covering NYSE,

AMEX and NASDAQ common shares. Corporate bond transaction data come from

the enhanced TRACE database and bond characteristics from Mergent FISD. Stock

order flow data come from TAQ, with retail trades identified using the Boehmer et

al. (2021) methodology and institutional trades defined as dollar volume exceeding

USD 50,000. The sample spans July 2002 to December 2024, a period chosen

because TRACE transaction-level data become available from 2002.

Results and Conclusions

The baseline regressions confirm that analyst forecast dispersion negatively

predicts one-month ahead stock returns and positively predicts one-month ahead

corporate bond returns.