A new formula for predicting stock returns beats the old accounting method — but only in theory

Researchers built a model to estimate what investors should demand as returns for holding individual stocks, using cross-sectional data. The new model outperforms the traditional accounting-based approach when tested against actual cross-sectional returns, but the accounting method still wins when you look at how returns actually move over time.

For decades, finance has relied on the accounting-based cost of equity to price risk. If this new model actually works in practice, it means investors and companies have been systematically miscalculating what their capital should cost. The problem is the gap between theory and reality: the model looks good on the benchmark that matters most to academics (cross-sectional prediction), but fails on the benchmark that matters most to practitioners (time-series performance). Until someone figures out why the model wins one test and loses the other, practitioners have no reason to switch.