search

WLM1ke / LedoitWolf

Ledoit-Wolf covariance matrix estimator of stock returns
35 stars 9 forks source link

README #1

Open dschenck opened 1 year ago

dschenck commented 1 year ago

Hi - what do you mean in the below (found on your README) ?

Ledoit-Wolf Estimator from sklearn use shrinkage to zero correlation equal variance target and implies no systematic risk and equal total risk of all stocks.

WLM1ke commented 1 year ago

In the case of 100% shrinkage, all stocks will have the same variance (total risk) and zero correlation, and therefore beta (systematic risk)

dschenck commented 1 year ago

Yes agreed 😊

My question was more, why should you not use the estimator from sklearn?

WLM1ke commented 1 year ago

In the sklearn approach, all stocks have zero systematic risk and in a portfolio with large number of stocks it is possible to reduce its risk to zero or significantly underestimate portfolio risk

The proposed in "Honey, I Shrunk the Sample Covariance Matrix" approach preserves the systematic risk of the portfolio