Fahiz Baba-Yara

Following portfolios beyond their formation date reveals return patterns that benchmark factor models struggle to explain.

Neither the persistent nor the transitory component is uniformly dominant.

A common component captures about two-thirds of the predictability, while the remainder is specific to each asset class.

This is a share of return predictability, not a share of realized return variation.

Three latent factors summarize the factor zoo in this empirical application.

The model accounts for uncertainty about the factor structure rather than treating the selected structure as known.

We link the decline to benchmark-tracking capital that broadens risk sharing and changes which investors set prices.

Risk depends on how factors move together, not just on their individual variances.

Firms whose employees read more macroeconomic news beforehand are more exposed to subsequent changes in economic conditions.

An earlier draft circulated as “Are Uncertain Firms Riskier?”

A strong performance metric does not by itself establish that one model spans another.

I build economic restrictions into the network’s architecture, then examine forecasts across firms, time, portfolios, and horizons.