Podcast Review: TTU #148 (part 2)
Top Traders Unplugged, host: Niels Kaastrup-Larsen, guests: Yoav Git, Rob Carver, Graham Robertson
Link to episode:
CTAs - The Good, The Bad, and The Misunderstood ft. Rob Carver, Graham Robertson and Yoav Git
Two ex-AHL guys, Yoav and Rob along with a current AHL partner Graham. You know this one will be interesting. Also, Yoav and Rob had the same role at AHL as Head of Fixed Income at different times.
You can actually read Parts 1 and 2 of my discussion independently, but if you wanted to read Part 1 first, here’s the link.
Podcast tried to answer the following question:
Why are CTAs not widely represented in portfolios?
This is an age-old question that every CTA tries to answer. You will typically see the same analysis in various forms from every CTA: you add a little bit of CTA exposure to a 60/40 portfolio and voila! — you have better Sharpe, less drastic drawdowns. Most importantly, CTAs allow investors to stay in invested the (stock) market during a downturn, since they get a nice hedging benefit from their CTA exposure. Think 2008 and 2022. Also, when other trends persist such as the USD trend of 2014, you get an amazing year. Other year, you will need to sit miserably and wish you invested in equities.
Yoav makes an excellent point on the podcast, which helped me understand why trend following is not very loved by the institutional investors: CTAs try to compete in the big league since they delivered amazing returns in the 80s, 90s and early 2000s, yet today CTAs are roughly a 0.5 to 0.7 Sharpe strategy. This is nothing to sneer at, but it’s also not an amazing Sharpe when you consider a portfolio of other strategies that deliver 1.5+ Sharpe. Yoav says that adding a 0.7 Sharpe strategy to a book of bunch of 1.5 Sharpe strategies does not increase the overall Sharpe. This is a great point, and I am going to argue that life isn’t always that simple.
Let’s use a concrete toy numerical example to understand the nuances of his argument. First, let’s assume I have two 1.5 Sharpe uncorrelated non-CTA strategies, each targeting 10% annual volatility. We will call this Portfolio A. If I were to allocate 50% of my capital to each, I get:
Note that these two strategies deliver 15% annualized return on their own capital. In order to understand what the portfolio level return would be, we need to consider two alternatives:
Portfolio A, No gearing: These strategies are at their limit. Adding more risk capital to them is either not feasible from a regulatory or credit risk perspective (think very high margin use to equity), or slippage increases to an arbitrarily high value eroding Sharpe. In this case, we achieve the same annual return as the individual strategies at the portfolio level, but at a higher Sharpe
\(r_{portf,A,NoGearing} = sharpe_{portf} * \sigma_{portf} = 2.12 * 7\% = 15\%\)Portfolio A, With gearing: Strategies can be levered up. We can gear the individual strategies (give them additional capital) in order to get the portfolio risk back up to 10%. Gearing amount is simply the ratio of decrease in portfolio volatility, i.e. \sqrt(2). In this case the geared portfolio return becomes:
\(r_{portf,A,Gearing} = G \cdot sharpe_{portf} \times \sigma_{portf} = \sqrt{2} \times 2.12 \times 7\% = 21.2\%\)
Now let’s add the CTA strategy to this mix, at 0.7 Sharpe with very high capacity, running at 10% vol. Assume all three strategies are uncorrelated. I will not run a parameter sweep (left as an assignment to the reader), so we will again assume equal risk allocation to these three strategies.
Adding the 0.7 sharpe uncorrelated strategy barely increases the sharpe. Since adding each strategy has a cost, it is understandable that allocators might be hesitant to add the strategy to the mix, or multi-strats being hesitant to add personnel to add a CTA strategy to their arsenal.
Portfolio B, No Gearing: If all three strategies are at their limit, the portfolio would be worse off by adding the CTA strategy. At this point we should note that this is an unlikely scenario, given the very high capacity of CTA strategies.
\(r_{portf,B,NoGearing} = sharpe_{portf} * \sigma_{portf} = 2.14 * 5.77\% = 12.33\%\)Portfolio B, With Gearing: If none of the strategies are at their limit, we can gear the strategies by a factor of the diversification factor sqrt(3) = 1.73. Note that Portfolios A and B have roughly the same return, as one would expect since they have the same sharpe and we are gearing them up to the same volatility level.
\(r_{portf,B,Gearing} = G \cdot sharpe_{portf} \times \sigma_{portf} = \sqrt{3} \times 2.14 \times 5.77\% = 21.4\%\)Portfolio B, Partial Gearing: This is where things start to get interesting in favor of CTAs. What if we could not gear up the first two strategies with higher Sharpe, but we could gear up the CTA strategy? This would be a very realistic comparison to Portfolio A without gearing, since we know CTAs have very high capacity.
In this case, we want to gear up the CTA strategy (but not the other two) such that the return of the total portfolio is greater than the return of portfolio A, with no gearing. This is simple to calculate, we would need to gear up the CTA strategy by the ratio of Sharpe ratios, in other words 1.5 / 0.7 = 2.14x in order to get the same expected dollar return as the higher Sharpe strategies.\(r_{portf,B,PartialGearing} \triangleq 15\%\)\(\sigma^2_{portf,B,PartialGearing} = \frac{1}{3^2} \left(1 + 1 + \frac{1.5}{0.7} \right) \times (10\%)^2\)\(\sigma_{portf,B,PartialGearing} \approx 8.56 \%\)\(sharpe_{portf,B,PartialGearing} = \frac{15\%}{8.56\%} = 1.75\)
Partial gearing Portfolio B obviously reduces the Sharpe ratio but with one important benefit: higher expected returns. Allocators who are trying to increase expected returns for their clients should consider adding more CTAs to their portfolios and perhaps require higher risk targets from them. In an environment with low capacity and high Sharpe strategies, CTAs fill a very important niche: high capacity at very low correlation to a 60/40 portfolio. CTAs provide volatility and returns that other strategies cannot provide. After all, you can eat returns, but not Sharpe.
Note that we haven’t even talked about the positive skewness of CTAs. Positive higher moments make the case for CTAs even stronger, since they increase total returns at the same volatility and Sharpe level