Interpretable Deep Learning for Stock Returns: A Consensus-Bottleneck Asset Pricing Model

TL;DR

The paper proposes the CB-APM, an interpretable asset pricing model that uses analyst consensus as a structural bottleneck, enhancing prediction and interpretability in stock return forecasts.

Contribution

It introduces a novel model embedding consensus as a structural bottleneck, improving interpretability and predictive accuracy over traditional methods.

Findings

Portfolios sorted on CB-APM forecasts show a strong return gradient.

The learned consensus captures priced variation beyond canonical factor models.

The model maintains robustness across macroeconomic regimes.

Abstract

We introduce the Consensus-Bottleneck Asset Pricing Model (CB-APM), which embeds aggregate analyst consensus as a structural bottleneck, treating professional beliefs as a sufficient statistic for the market's high-dimensional information set. Unlike post-hoc explainability approaches, CB-APM achieves interpretability-by-design: the bottleneck constraint functions as an endogenous regularizer that simultaneously improves out-of-sample predictive accuracy and anchors inference to economically interpretable drivers. Portfolios sorted on CB-APM forecasts exhibit a strong monotonic return gradient, robust across macroeconomic regimes. Pricing diagnostics further reveal that the learned consensus encodes priced variation not spanned by canonical factor models, identifying belief-driven risk heterogeneity that standard linear frameworks systematically miss.