Sharpe ratio of 3.32. Alphas unexplained by 14 factors. And a map of how information flows across stocks.

A new paper builds a stochastic discount factor that doesn't just use a stock's own signals — it models how every stock's characteristics predict every other stock's returns.

The framework jointly estimates:

• Λ: which signals matter (investment, value, and profitability dominate),

• Ψ: a spillover matrix encoding cross-asset predictive relationships.

Out-of-sample results (1973–2023):

• Sharpe of 2.21 on spread portfolios, 3.32 on bivariate sorts,

• Monthly alpha of ~0.25% (t > 11) vs. FF5, q-factors, mispricing, and behavioral models,

• Robust across high/low sentiment and VIX regimes.

The network insight: large, low-turnover firms are net transmitters of predictive information. Small stocks absorb it. Momentum and reversal? Near-zero weight in the SDF.