Course Name: Quantitative Portfolio Management, Section No: 13, Unit No: 1, Unit type: Video
In this video (at 1:15), for the factor selection, it says “…you need to add those factors that have a correlation coefficient very close to zero or negative…”. However, this does not seem to make sense in terms of the linear regression theory. We would like to have factors that are orthogonal to each other (their correlation is close to 0). I understand that for stock selection it makes sense, but for factor selection (assuming it is done after stock selection), it does not make sense.
Am I missing something here?
Hi,
Here, we are looking at factor selection from the perspective of Portfolio Theory and Risk Management.
You are correct that in a regression model, high correlation between factors makes the coefficients unstable.
However, in Factor Investing, we are trying to smooth the ride.
Orthogonal Factors (Correlation ≈ 0): These provide independent sources of return. When one factor is flat, the other might be up. This is the baseline for a healthy multi-factor model.
Negatively Correlated Factors: Here, when one factor goes through a decline, the other tends to gain, which can reduce the volatility of the total portfolio.
Hope this helps.
1 Like
Thanks for the answer, will think about it.