I am interested in coming up with better risk calculations for alternative asset classes. As these are illiquid, not a lot of historical data is available.

My idea is to use performance of stocks and bonds to estimate, say Private Equity returns, back in time and use these estimates, to improve risk estimates.

What would be the best approach to this task?


My recommendation is to focus on improving estimates based on available data. For example, the returns of these assets classes tend to exhibit very strong serial correlations as a result of smoothing, which dampen observed volatilities. You can add a ton of value to the investment process by de-smoothing these returns and estimate risk thereafter.

I'm inclined to believe that this is more valuable than extending the returns – it's hard to gauge what kind of biases we might introduce. And if adding the extra returns doesn't change the results meaningfully (likely because we're imposing our will based on observable data), there's no value in adding them anyways.

That being said, we do extend these time series in a lot of portfolio construction research, mostly when we're studying diversification benefits (e.g., how these assets interact with others in different economic environments):

  • For private equity, it's common practice to use something as simple as the returns of Russell 2000 scaled by 1.1 to 1.2, plus an in-house alpha assumption.
  • Private real estate can be proxied by house prices, which already go back to the late 1800s.
  • REITs behaved more like equities in the early days and can be proxied by equity benchmarks, or a blend of equities and house prices.
  • Hedge fund is much more challenging. Insofar that your hedge fund portfolio is truly market neutral, the "replication" can be as simple as cash + an alpha assumption. Otherwise, you can regress either the HFRI or the CS indices against equities, bonds, credit spreads, etc. and go from there. If you're really adventurous, you can replicate some of the popular strategies (trend following, carry, etc.) and create a blend of them. Most banks also offer hedge fund replication products and they provide very detailed documentation on how they do it (mostly just simple regressions), which you can reference for ideas.
  • $\begingroup$ I really appreciate your nuanced answer and new angles on the problem. Can you elaborate on what you mean by smoothing and "de-smoothing"? It sounds really interesting. Perhaps even point me to some libraries I can use. I prefer Python, but R will do as well - thanks a bunch! $\endgroup$
    – DBE7
    Aug 10 '18 at 9:35
  • $\begingroup$ You might google "hedge fund smoothing" for example $\endgroup$
    – Alex C
    Aug 10 '18 at 14:57
  • 1
    $\begingroup$ There are quite a few approaches. The most popular one is the Geltner method. See "De-Lagging the NCREIF Index: Transaction Prices and Reverse-Engineering" and "Value Indices of Com- mercial Real Estate: A Comparison of Index Construction Methods" $\endgroup$
    – Helin
    Aug 10 '18 at 19:56

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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