Lest we forget: Using Out-Of-Sample Errors in Portfolio Optimization

Socially responsible investment portfolios: Does the optimization process matter?

“…Bloomfield et al (1977) found that naive portfolio allocation methods are superior to more sophisticated methods, and that 1/N performs well; while Jorion (1991) demonstrated that 1/N is superior to the sophisticated techniques of Markowitz, Bayes-Stein and minimum variance. More recently, Barroso (2015), Duchin and Levy (2009), Jagannathan and Ma (2003), Kan et al (2016) and Pflug et al (2012) show that 1/N is superior to Windcliff and Boyle (2004) find that 1/N is preferable to both Markowitz and Bayes-Stein, and Fischer and Gallmeyer (forthcoming) demonstrate the superiority of 1/N to minimum variance. Kirby and Ostdiek (2012) point out that the stability of naive diversification is one of the main causes behind its strong performance, and that the reward-to-risk approach can yield stronger results, even in the presence of high transaction costs.…”

Horses for courses: Mean-variance for asset allocation and 1/N for stock selection

“…A number of studies have applied MV and 1/N to either asset allocation or stock selection problems and obtained results consistent with each of the parts of our hypothesis. For example, Allaj (2019), Durand et al (2011, Han (2016), andShigeta (2016) have found that MV has a bigger Sharpe ratio than 1/N for asset classes; while Barroso and Saxena (2019), Board and Sutcliffe (1994), DeMiguel et al (2009a), Dickson (2016, Hwang et al (2018), Jagannathan andMa (2003), andLi (2016) have found that 1/N has a bigger Sharpe ratio than MV for individual equities.…”

Socially Responsible Investment Portfolios: Does the Optimization Process Matter?

“…Bloomfield et al (1977) found that naive portfolio allocation methods are superior to more sophisticated methods, and that 1/N performs well; while Jorion (1991) demonstrated that 1/N is superior to the sophisticated techniques of Markowitz, Bayes-Stein and minimum variance. More recently, Barroso (2015), Duchin and Levy (2009), Jagannathan and Ma (2003), Kan et al (2016) and Pflug et al (2012) show that 1/N is superior to Markowitz; Windcliff and Boyle (2004) find that 1/N is preferable to both Markowitz and Bayes-Stein, and Fischer and Gallmeyer (forthcoming) demonstrate the superiority of 1/N to minimum variance. Kirby and Ostdiek (2012) point out that the stability of naive diversification is one of the main causes behind its strong performance, and that the reward-to-risk approach can yield stronger results, even in the presence of high transaction costs.…”