Lessons in Replication
Replication is an exercise in truly understanding the underlying funds. The Index must be constructed in a certain way, and the constituents are those that abide by their mandate among other constraints. The assets used and the number of assets involved in the procedure should be chosen thoughtfully. Our goal is to obtain a portfolio that is able to capture that performance conditional on the reduction of tracking error.
The reason why we look to replication are things we’ve alluded to earlier. The primary benefits are an allocation to the space while minimizing single manage risk and outperformance through fee disintermediation.
In our view, managed futures is the sole category worth having an allocation towards for the majority of investors.
A bottom up approach is less ideal than a top down approach. It’s another form of a single manager product, the risk we’re trying to avoid.
The underlying index has a rotating list of funds with different underlying strategies and portfolio construction, making the bottom up approach unlikely to accurately reduce future tracking error. Slightly different model specifications can result in vast return dispersion, and can result in a vast hit to long-term peformance. This paper by Quantica offers more insight into the changing nature of the SG CTA Index.
A top down approach is able to capture not only the performance of the underlying index, but is also able to adapt to the changing fund constituents and strategy improvements. When a fund like AQR adds a new strategy dubbed "economic trend,” that is reflected in the top-down approach.
The components of the SG CTA Index utilize strategies ranging from trend-following, carry, mean reversion, relative value, global macro, and even equity long-short, although trend-following remains the most significant piece.
The top down approach must be robust and adaptive to moves in the index, changes to the underlying funds, and changes within the existing funds. A well designed approach is able to look at new data points, adapt accordingly, and be robust to different specifications that don’t result in significant changes in tracking error.
Is there single manager risk in replication?
Done right, we believe that answer is, No! A well designed model should be more apt to minimize single manager risk than any of the underlying constituents! Even if provided with the wrong weights, a well designed model will be able to self-correct and adapt to new data.
Even though we minimize single manager risk, there are still going to be periods of deviations from the benchmark as tracking error can result from *very-rare* idiosyncratic market movements in the Index. Often, this has led to a reduction in performance for the Index relative to our tracker, but there are other times when it is a bit less beneficial.
Therefore, it might even be prudent to combine different replication managers together in an effort to reduce tracking error further, which we have a unique internal take on.
