Optimal characteristic portfolios

Characteristic-sorted portfolios are the workhorses of modern empirical finance, deployed widely to evaluate anomalies and construct asset pricing models. We propose a new method for their estimation that is simple compute, makes no ex-ante assumption on nature relationship between characteristic returns, does not require ad hoc selections percentile breakpoints or portfolio weighting schemes. Characteristic weights implied directly from data, through maximizing Mean–Variance objective function with mean variance estimated non-parametrically cross-section assets. To illustrate method, we size, value momentum find overwhelming evidence outperformance our methodology compared standard methods constructing characteristic-sorted portfolios.