More Thoughts on Hedge Fund Replication

So Morgan Stanley jumped into the cloning business recently with announcement that it will be launching “Altera”, in order to compete with products available from other banks such as JP Morgan, Goldman, Merrill, and Deutsche Bank.

This replication competition is getting silly. I have only seen the “guts” of a few of these products, so I cannot comment on all of them, but most seem to use the factor model approach for replication. In a nutshell, use a hedge fund composite index as your dependent variable and some combination of multi-asset broad-based indices as your explanatory variables.

The more I mull on these replication techniques, the more I have a lot of issues buying into the ideaology of these products. For sure, hedge fund replication is at worst an intriguing academic task and, at best, an INCREDIBLY lucrative experiment. But what are we trying to accomplish?

Personally, I want to understand trading styles and strategies that generate these returns – NOT just attempt to replicate the returns by somewhat haphazardly guessing allocation exposures macro-type indices.

For a moment, we should go back to 2004. The two most powerful papers I have read on hedge fund replication are oft-quoted, yet under utilized.

- Bridgewater Associate’s Hedge Funds Selling Beta as Alpha

- JP Morgan’s Have hedge funds eroded market opportunites

These papers break down performance using rule based strategies, not by combining macro-factors: fixed income arb as a combination of illiquid less liquid benchmarks, merger arb as a function of taking positions in the largest deals, managed futures as a simple moving average crossover rule, et cetera.

While my guess is that correlation coefficients of these replication types are not as high as in-sample results seen from recent replication studies, out of sample performance will be much, much better.

The other issue I have is that composite hedge fund indices are composed by combining sub-sector performance. So while we might be able to explain sub-sector performance using some of the techniques in the two papers mentioned above, by the time we jumble fixed-income arb with long/short equity, and merger arbitrage with managed futures – any discernable advantage we had to model sub-sector returns is useless because the return series of the composite index is too noisy to model using these factors. So what do the banks’ products use? They use broad market indices representing geographic equity returns, bond returns, and currency returns. Stupid.

So where should research go? I think generating rule-based trading strategies to model sub-sector performance is the ideal goal for hedge fund replication. Once the sub-sectors have been modeled accordingly – combine the sub-sector replication strategies into a composite replication strategy. Now there is one downside - strategies will change and rules will need to be added and updated over time. Some might argue that changing rules out-of-sample over time defeats the whole purpose of having a replication strategy – the replication process should be constant. Perhaps, but do you really think the hedge fund community is static in the strategies they employ? No.

My goal is to understand the processes that drive hedge fund results – not just replicate the returns. Trust me on this, when factor models blow up out of sample or the cloning results lag the on-going changes in hedge fund behavior, understanding the strategies that drive the return process will be a huge winner.