Combining SPY-COMP and the Volatility Curve Models

In today’ post, I’ll formally outline a model I’ve alluded to several times in the past, the VOL-COMP model. The model is a combination of the SPY-COMP model (which is the basis for all the models in the Economic Pulse Newsletter) and the VOL Curve Original model with several enhancements. While the combination of SPY-COMP and VOL Curve can be used with many assets, in this post and as part of QuantPulse I’m introducing a specific implementation of the model, the VOL-COMP Global TAA model, which uses global equities as the risk-on asset classes and uses short and long term government bond durations for the risk-off assets.

Basics

I’ve discussed and posted about the basics of this model (in this Q&A post for example) in the past. I’ve presented the VOL-COMP model as a combination of the SPY-COMP model and the VOL-Curve model. The combination is a simple logical AND function. In other words, for the model to be risk-on both the SPY-COMP model and the Vol Curve model need to be risk-on. Otherwise, i.e. if either is risk-off, then the VOL-COMP model is risk-off.  I’ve been able to test this model back to 1993 using SPY and TLT as the risk-on and risk-off ETFs respectively. I’ve replicated the relative performance table below. The second column, VOL-COMP OG is the model I’m building on since it has the best performance over the full time period and the lowest number of trades.

The Model

The VOL-COMP Global TAA model takes the VOL-COMP OG model, adds global stocks (VEU), expands the US universe to all stocks (VTI), and uses relative momentum (average of 3/6/12 month performance) between the two to choose the risk-on ETF. Basically, the DM-COMP model from Econ Pulse combined with the VOL Curve model. In addition, for the risk-off component, to make the model more robust to interest rate changes, the model uses relative momentum (average of 3/6/12 month performance) between long term US government bonds (TLT) and and short-term US government bonds (BIL).

Another nice feature of the model is that it uses both time frames of the sub models that it is made up of. The overall risk-on risk-off decisions are monitored on a daily basis, like the Vol Curve model, but the relative momentum decisions, between VTI VEU and between TLT BIL are only made on a monthly basis. This keeps the number of trades and turnover to a more reasonable level.

That’s about it.

Results

The Table below shows the results for the model back to 1999, the earliest I can take it back with the global ETFs. The table shows two options, one with VEA as the international ETF, and one with VWO as the international ETF. The inception date for VEU was in March 2007 so it can’t be used to back test before then. The back test was done on Portfolio123. The DM-COMP model is from the Econ Pulse Newsletter and is a global version of SPY-COMP.

Source: Portfolio123

Very good performance and drawdowns for the VOL-COMP model. Better than either VOL Curve Original or DM-COMP on their own and also significantly better than the benchmarks. Sharpe ratios for the full period from 1999 were 1.07 for the VTI-VEA model and 1.14 for the VTI-VWO model.

We can also model with VEU and the real vol curve data since in inception back to 2009 for full year data. Below are the results.

Source: Tiingo, CBOE data, my model calculations

Implementation and Tracking

I’ve implemented the VOL-COMP model like I have the VOL Curve model and track them online in a Google Sheet. Current status is shown below. This is what is available to subscribers along with an email when the model changes status.

Summary and Launch

The VOL-COMP Global TAA Model provides the best of the Vol Curve model and the SPY-COMP model. Performance and risk are better than either of the individual models on their own. The model is now available as part of Quant Pulse. Econ Pulse Subscribers can add the Quant Pulse Service for an additional $25/month.