With hedge fund performance starting to look up, a reader recently suggested we check up on the trials and tribulations of “hedge fund replicators” – those who aim to clone the returns of hedge funds via passive exposure to highly liquid and ubiquitous investments. The most well understood method of doing is to use a factor model based on a trailing regression of hedge fund industry returns. One of the most prominent players in this space is probably Merrill Lynch, purveyors of the “Merrill Lynch Factor Model” (factor model website).
The firm describes the index this way in its marketing sheet:
“The Merrill Lynch Factor Model (the “Model”) is designed to provide a high correlation to “hedge fund beta,” the portion of hedge fund returns which can be explained by exposure to certain market measures, and not individual manager skill. To accomplish this, the Model seeks to emulate the performance of the HFRI Fund Weighted Composite Index (the “HFRI Composite Index”) using the publicly reported prices of a basket of liquid, well-known market measures. The HFRI Composite Index is a non-investable index which tracks the performance of over 2,000 hedge funds on a monthly basis.”
So how has it done at “emulating” the HFRI? According to Merrill’s data, pretty well. Although its marketing sheet hasn’t been updated since the end of 2007, raw data provided on the product’s website shows the Factor Model and the HFRI Composite have a rolling 24 month r-squared of about 0.9 over the past several years.
But why look at monthly returns when you can examine daily returns? The ML Factor Model publishes daily returns that you can compare against the daily results of the HFRX (the daily-valued cousin of the HFRI).
As you can see from the chart below created by us from data on the product’s website, the trailing 30 day r-squared of the ML Factor Index and the HFRX has been much more volatile than that of the monthly data (over the past 18 months at least).
Thankfully, the falling correlation between the ML Factor Model and the HFRX is due to the fact that the ML Factor Model has been dramatically outperforming the index this year. The following chart shows the performance of the HFRX (via HFR), the ML Factor Model (via the MLFM website) and the S&P 500 (total return, via Yahoo!) on an index of 100 (100=Jan 1, 2008). As you can see, the ML Factor Model seemed to enjoy the downside protection of the HFRX last fall, and is now also enjoying the tailwind of the S&P’s recent pop as well.
Unfortunately, beating the HFRX isn’t really part of the mandate for these models. Still, the Merrill Lynch Factor Model seems to be nearing its pre-crash levels – and that’s a lot more than can be said about either the HFRX or the S&P 500. How did it achieve this? Check out the website to see the actual historical weightings used in the model over the past year.