The authors introduce Stability-Adjusted Portfolios: a methodology for incorporating estimation error in covariances into the portfolio formation process.
The authors compute covariances from all independent subsamples of a chosen size and measure composite errors in these subsamples. These composite errors comprise small-sample error, independent-sample error, and interval error. They then add these errors to a base-case covariance matrix and, assuming normality, generate stability-adjusted return distributions for all subsamples. They then combine these distributions into a stability-adjusted return distribution, which is non-normal.
The authors then use full-scale optimization (that works with non-normal distributions) and utility functions to derive optimal portfolios. These portfolios tend to be less volatile.
KRITZMAN, M., & TURKINGTON, D. (2016). Stability-Adjusted Portfolios. Journal of Portfolio Management, 42(5), 113–122.
No comments:
Post a Comment