Trend Following and Commitment

I've done a couple posts on trend following in retirement portfolios over the last month. NewFound Research has a new Trend Equity index out so I thought I'd take that for a spin. Their index is what they call a "specification-neutral implementation of the trend equity style" which means that that it is a multi-model, mix parameterization and tranched-rebalance approach and therefore, I guess, lends itself to reducing inter-model comparative idiosyncrasies. Plus it goes back to 1928. We'll use it.

The question is something like this: "If I were to allocate to something that behaves like the index, what happens to convexity and returns if I hold it for different horizons."


The "return premium" is the trend anomaly that we touched on in the last post when we asked if it might be persistent if we first agree that it is trade-able to which we answered "yes" or "probably." The trend return premium in that post was what I likened to a negative, uncertain premium on a put option paid (received in this case) in order to get the convex payoff (Newfound calls it crisis beta rather than crisis alpha which seems about right). Either way, it seems like a pretty good deal if one can get it and keep it: someone offering to probably pay me for my option to probably have a less negative payoff during (at least) long, slow crises. Hard to find those kind of offers. Here, we'll assume persistence and just look at holding something like the index, if we can find it, for 12, 24, and 48 months to see what happens and what it looks like.

The Trend data comes from the Newfound Research Trend Equity Index. From the link:
- diversified across an ensemble of different trend-following models
- diversified formation horizons
- diversified rebalance periods
- each model is applied across 120 different formation periods, spanning 120-to-240 trading days
- the ensemble of 360 different allocations are averaged

S&P data comes from Prof. Shiller.

If I did it right, the basic idea is to get a rolling, chained, annualized geometric return representing 12, 24, and 48 month horizons (for both) and then to scatter the relationship between S&P (x) and Trend Equity Index (y) and see what it looks like.  The trend lines fit to the scatter are a simple second order polynomial line courtesy of excel.  Bystanders can shoot at me if I got any of this wrong. I'm not 100% sure or committed on this yet.

The scatter...

The return profile by way of a probability mass on the Trend index for the different horizons...

and some summary stats...

Conclusions

I don't have any hard conclusions yet since this is a first pass but stepping back and squinting I might say something like:

• A longer hold looks like it gooses the convexity a bit but in a diminishing way. The convexity was less extreme than I was expecting but present,
  
• The return profile, I guess unsurprisingly, shrinks its distribution with longer horizon. Both the right and left tail come in in ways I have not measured but in accumulation I might worry about the right side while in decumulation that movement inward on the left is pure gold,
  
• Stepping way way back I might even say that if I had the courage, my allocation might not be partial, it might be total.  But that seems a bit bold for now given the hard times I've seen in managed futures since 2009.