At a conference last week, Professor Harry Kat of hedge fund replication fame presented a list of his specific rebuttals to 10 criticisms. He’s since included the audience’s feedback in a new article on distributional replication released this weekend.
Both factor-model and distributional hedge fund replication have attracted a lot of attention over the past year – and both have also been criticized.
Kat’s recent presentation in Geneva was aimed squarely at critics of his distributional approach. After presenting his list of 10 criticisms, some in the audience volunteered a few others. Since last week, Kat has been busy incorporating these additional unjustified criticisms to the existing list. What resulted was a new paper available here.
In descending order, here are Kat’s rebuttals to the top 10 criticisms of his distributional replication approach:
1. Due to the dynamic nature of FundCreator-based trading strategies, transaction costs will dramatically erode returns.
Summary of Kat’s response: Not true, our model explicitly accounts for transaction costs, futures are highly liquid instruments and trades needn’t be made immediately.
2. When used for hedge fund replication, the returns generated by FundCreator do not match the returns of the target fund or index on a month-to-month basis.
Summary of Kat’s response: True, but unjustified. The whole point of our approach is to match return distributions, not month-to-month results.
3. Average returns depend on the reserve asset that is chosen by the user.
Summary of Kat’s response: True, but that is a strength, not a weakness. Our approach doesn’t create returns on its own. It’s a risk management tool, not an investment product. Users make their own decisions about the underlying reserve asset and thereby are able to incorporate any views and/or skills they may have.
4. You cannot expect investors to wait for several years to see whether their returns indeed have the desired properties.
Summary of Kat’s response: This is no different from what appears perfectly acceptable in the fund management industry. Sure, over short periods, results may differ from the target. That is normal sampling error. In the longer run, everything will converge.
5. FundCreator only targets the risk profile, but not the expected return.
Summary of Kat’s response: Not true. One parameter must always be left open for the market to determine. But this needn’t be the expected return. For example, we can fix the expected return and volatility and leave the skew to the market instead.
6. Although much more complicated, FundCreator is not very different from mean-variance optimization.
Summary of Kat’s response: True in cases where returns are normally-distributed. But contrary to mean-variance our approach addresses higher moments such as skew and explicitly allows for dynamic trading over the investment horizon.
7. The dependence with other assets than the reference portfolio is not targeted.
Summary of Kat’s response: True, but who cares about that. Our approach produces a portfolio with a given dependence structure to the investor’s existing portfolio. That’s all that really matters.
8. The model of monthly returns that is used to determine the desired payoff function could be incompatible with the Black-Scholes-type model that is used to price it.
Summary of Kat’s response: True, but an academic point with little or no implication on the success of the model.
9. Due to the non-normal return distribution of the assets traded, the use of Black-Scholes type hedge ratios is inappropriate.
Summary of Kat’s response: Not true. The portfolios traded by the model are well diversified and normally-distributed – making normality an excellent approximation.
10. You will never be able to replicate the best hedge funds.
True. Special skills cannot be replicated. That’s what makes them special. One has to distinguish between pre-fee and after-fee alpha though. Synthetic funds derive their return from market risk premiums. By construction therefore, synthetic funds produce no pre-fee alpha