This is a general question on how fund managers use factor models to add alpha. I understand how a risk model can tell you what factors a portfolio has exposure to. But can a risk model actually tell you how to position a portfolio to generate alpha?

If anyone has some good links or articles which explain this that would be helpful.

  • $\begingroup$ one way is to try to predict which factor returns are going to revert and which are going to trend over some horizon. Assuming you can do that, then, if you weight your factor exposures accordingly over that same horizon, you should obtain excess return over that same horizon. The question is how to predict the future factor returns. $\endgroup$
    – mark leeds
    Commented Aug 1, 2019 at 10:53
  • $\begingroup$ Oh, to answer your other question: generally speaking, you won't find anything terribly useful in the literature because, if someone has a useful method, they probably wouldn't publish it. $\endgroup$
    – mark leeds
    Commented Aug 1, 2019 at 10:55
  • $\begingroup$ In Factor Investing, as described for example in Andrew Ang's book, some factors are thought to earn factor premiums in the long run. It therefore makes sense to seek exposure to such factors to enhance long term returns above the cap weighted portfolio's return. You could consider this outperformance a form of alpha, but not everyone does. $\endgroup$
    – nbbo2
    Commented Aug 1, 2019 at 12:29
  • $\begingroup$ From my understanding of fundamental factor models, the factor returns are derived from cross-sectional regression. How do you interpret these "factor returns"? They aren't based on what the securities returned historically. Say the long term average monthly return for the value factor is 3% which is derived from the regression. Can you generate alpha by having a positive active exposure to the value factor? But then you kind of assume the factor return of 3% holds up. $\endgroup$ Commented Aug 2, 2019 at 6:46
  • $\begingroup$ hi: the factor model is generally re-estimated every whatever ( let's say month ) so the factor returns change each month. If you bet on them, you're betting that they are either going to revert or trend or stay the same etc. But next month's factor returns won't be the same as this month's factor return. All one can do is control the exposures to the factor returns. I would think of the factor returns as the betas in a cross sectional regression model. $\endgroup$
    – mark leeds
    Commented Aug 2, 2019 at 13:28

1 Answer 1


By definition, they don't (assuming you're considering something in the vein of APT). The primary benefit of equity factor models in terms of portfolio risk is allowing you to decompose risk to understand what you're actually exposed to aside from simply market risk.

  • $\begingroup$ Factor models (i.e: barra) are also used to track indices. Say you want your portfolio to mimic the S&P 500 index without buying all the stocks. You can use an equity factor model to make sure that your portfolio has the same factor exposures that being perfectly invested in the S&P 500 would result in. This should reduce tracking error. $\endgroup$
    – mark leeds
    Commented Aug 2, 2019 at 5:15
  • $\begingroup$ Yup, same thing from two different perspectives...assuming the factors included are exhaustive. I don't know that it's an open and closed case that matching factor exposures will give matching performance outcomes, but it's at least the way I think about it. $\endgroup$
    – Chris
    Commented Aug 3, 2019 at 3:36
  • $\begingroup$ Hi Chris: Not matching but it's better than just buying the big cap stocks and hoping for similar outcomes. It's also a way to model tracking error. $\endgroup$
    – mark leeds
    Commented Aug 4, 2019 at 12:30

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.