Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
Empirical evidence shows that the relationship between firm characteristics and stock returns is non-linear, with a stronger correlation at the extreme deciles of the characteristic values. In this paper, we propose a novel portfolio optimization method that models the portfolio weights as a non-linear function of firm characteristics. Our approach allows the weights to vary non-linearly across percentiles of the cross-sectional distribution of each characteristic. We apply our method to the universe of firms listed in the NYSE, AMEX, and NASDAQ and find that non-linear effects in size, value, and momentum anomalies are important for constructing portfolios that have lower risk and higher risk-adjusted returns. Our results suggest that a flexible relation between portfolio weights and firm characteristics can better capture the empirical patterns observed in the data.