The Value of Using Predictive Information Optimally

41 Pages Posted: 16 Oct 2020 Last revised: 17 Nov 2020

See all articles by Michael Ashby

Michael Ashby

Faculty of Economics, University of Cambridge; Downing College, Cambridge

Date Written: October 24, 2020


For mean-variance investors, using predictive information unconditionally optimally produces better portfolios than using predictive information conditionally optimally. The latter is more usually done in practice. Empirically, the unconditionally optimal portfolios have higher Sharpe ratios and certainty equivalents than the conditionally optimal portfolios. They also have lower turnover, leverage, losses and draw-downs. Moreover, measures of the whole distribution tend to prefer the unconditionally optimal portfolios, especially once transaction costs are accounted for. With transaction costs, the unconditionally optimal portfolios often second-order stochastically dominate the conditionally optimal portfolios. The unconditionally optimal portfolios are also preferred in terms of Sharpe ratio, certainty equivalent, costs, losses, draw-downs and stochastic dominance to mean-variance optimal portfolios that do not use predictive information. However, whether unconditionally optimal portfolios are preferred to minimum variance or 1/N portfolios depends on the asset universe.

Keywords: conditional efficiency, unconditional efficiency, signal, predictive information, prediction, risk-return trade-off, mean-variance

JEL Classification: G11, G14, G17

Suggested Citation

Ashby, Michael, The Value of Using Predictive Information Optimally (October 24, 2020). Available at SSRN: or

Michael Ashby (Contact Author)

Faculty of Economics, University of Cambridge ( email )

Sidgwick Avenue
Cambridge, CB3 9DD
United Kingdom

Downing College, Cambridge ( email )

Regent St
Cambridge, CB2 1DQ
United Kingdom

Do you have negative results from your research you’d like to share?

Paper statistics

Abstract Views
PlumX Metrics