A correlation-based portfolio choice algorithm

Analyzing the correlation matrix of listed stocks, we identify "singletons" that table minimal cross-sectional correlations. Portfolios comprising 100-500 singletons all have lower betas and standard deviations and, correspondingly, higher average Sharpe and Treynor ratios than the Center for Research in Security Prices (CRSP) universe over the sample time period 1950-2017. Portfolios of singletons chosen from subsets of the CRSP universe, including small-value, low-variability, and momentum stocks, similarly realize lower portfolio standard deviations and higher risk-adjusted returns. These well-diversified portfolios suggest that the positive abnormal returns to low-beta portfolios are driven by their component stocks having low average cross-sectional correlation. One of the authors invested $20,000 of his own money in the algorithm-chosen 240 stock singleton portfolio over a 4-year period (2015-2018) and beat the market year-by-year on a risk-adjusted basis just as our results predicted.