The Easy VIX: VIX Futures Patterns And A Risk-Responsive Portfolio

by: Michael Gettings

In my last article, I explained in some detail how risk-mitigation signals, derived from the VIX futures curve, could double equity returns over the last eleven years.

Today, I’ll supplement the long-term analysis with a quantitative view of how best to deploy cash during periods when equities are deemed too risky.

I’ll compare the use of IEF (10-year Treasury ETF) versus TLT (Long-term Treasury ETF) during those periods of high equity risk.

I’ll also provide far more detail as to the workings of the model than I have previously.

Finally, I reported a sell signal at market close of May 24th. I’ll provide an update on how that signal was resolved with the close of June 4th.

In my last article, linked here, I explained how using a risk-avoidance algorithm derived from the VIX futures curve produced a 15.7% return for a basket of broadly diversified ETFs compared to a 7.3% return for a buy-and-hold strategy. I also noted that the risk avoidance aspect was critical; the same algorithm produced a worst‑case, 12-month loss of only 6.4% versus a 44.7% worst 12-month loss for the buy-and-hold strategy over that eleven-year period. The lower risk and higher return have implications for portfolio structure.

From my perspective, since I'm mostly retired and living on investments, I would normally be more heavily skewed toward fixed income. But having built some confidence in the algorithm's ability to identify high-risk times, I've gradually increased equity allocations as if I were a decade or two younger. During normal periods, I run about 55% stocks and ETFs, 10% preferred, and the remainder split between bonds and bank loan ETFs.

Of the 55 percentage points in stocks and ETFs, I keep most of it in broad-based, very liquid ETFs that can be sold quickly on a signal; the rest stays in a few growth stocks. For now, I maintain 10% to 15% in those growth positions, even in high-risk periods. Given the usual inverse correlation with bonds, that provides a good balance, but I wrestle with it. Why not apply the same algorithm to liquidate the growth stocks and let the bonds run? In fact, why not move the entire portfolio toward fixed income whenever a period of high equity risk is identified?

To inform my own thinking on the question, I've analyzed the use of IEF (10-year Treasury ETF) and also TLT (Long-term Treasury ETF) during those periods of high equity risk. The long-term analysis in my last article assumed that when the ETF basket was sold, the proceeds remained in cash until the next buy signal. Since the model goes to cash about 25% of the time, there should be an exploitable opportunity.

I assumed a 2.5% dividend for IEF and 3.0% for TLT. I felt that those proxies were adequate, and researching actual dividend histories would offer little in the way of further insights. What I found is this. IEF makes a much better choice for those high-equity risk periods. I started the analysis thinking that the longer maturity/duration of TLT would produce greater returns under a risk-off environment, but it was not true. As it turned out, while it was true in the best of cases, the downside results for TLT were far worse than those for IEF. Further, those downside periods brought the average return down substantially. Assuming the "parking" periods (high equity risk is cumbersome) averaged 25% of the year, the average incremental return for IEF was .85%, while TLT added only .22%. Here are the results:

Comparison of IEF vs. TLT "Parking" Returns During Periods of High Equity Risk

Source: Michael Gettings Data: Fidelity

And, here is a graphic of the ranked returns for those parking periods only:

Graph of Ranked Treasury Returns During Periods of High Equity Risk

Source: Michael Gettings Data: Fidelity

Notice in the graphic how TLT's negative returns at the extreme left are bad; they are comparable to the good results at the right side, whereas IEF is very muted to the downside. I suspect that IEF simply benefits more than TLT from the flight to safety in market declines.

That covers the cash deployment question; now, let's look at the current conditions as indicated by the algorithm. The model had called a sell signal on May 24th and exhibited elevated volatility for the week and a half following. By June 3rd, the ETF basket was down 3.5% from the May 25th opening when the sell signal was executed. Then, with the significant market rally on June 4th, the ETF basket rebounded 2.4%, and we gave up 2.4 percentage points of the 3.4 percentage points gained during the interval; early in the day, on June 4th, the model called "Buy". Buying at the close on June 4th would have produced a 1.0% advantage compared to holding through the interval; buying at the open on June 5th would have produced a .4% advantage.

Since June 4th, the ETF basket has rallied by 2.4%, and that has been fully captured.

There are four factors in the model, and they are triggered based on criteria which follow an artificial intelligence ("AI") algorithm. An aggregate score of 2 or more indicates a sell signal; less than 2, a buy-back or hold signal. The factors are listed below. In all cases, contango registers as a positive shape, and movement toward contango registers as a positive slope.

  1. Primary Slope: measures the rate of change in the shape of the VIX futures curve (contango vs. backwardated) - if the Primary Slope is more negative than the criteria, accrue 1 point
  2. Confirming Slope: measures the rate of change in the shape of the VIX futures curve over a longer look back - if the Confirming Slope is more negative than the criteria, accrue 1 point
  3. Safe Shape (sufficiently contango) - if the shape is contango and above the safe criteria, accrue a score of minus 1
  4. "Sell" Shape (sufficiently backwardated) - if the shape is backwardated and less than the criteria, accrue 1 point

Current metrics look like this: The Primary Slope is healthy, while the Confirming Slope remains a concern. The shape itself is in neutral territory - neither safe nor indicating a sell contribution to the overall metrics.

Easy VIX Dashboard

Source: Michael Gettings Data Source:

I'll probably repeat this caution in every article: The algorithm is not designed to call each twist and turn in the market (if that were even possible). It is designed to identify high-risk environments for equities, and then by avoiding them, produce higher returns over reasonable time frames. Sometimes, it will forego small gains in return for safety, but more significantly, it tends to avoid large losses. On balance, probabilities skew substantially to the positive. As stated above, the performance since 2008 produced more than double the returns of an indexed portfolio. Year to date through June 10th, the managed portfolio returned 18.9% versus 14.7% for the indexed portfolio - a 4.2 percentage point advantage over 5-1/3 months.

Here is a graph of the rolling yearly results comparing the raw indexed returns on the left, with algorithm-managed returns on the right. The underlying ETF basket consists of SPY, DIA, QQQ, and IWM, and high-risk periods are invested in IEF. Notice how few loss periods exist on the right graphic compared to the left. The drawdown protection is the key to superior returns.

Side by Side Rolling-Yearly Returns, Indexed vs. Algorithm

Source: Michael Gettings Data: Fidelity

I've explained far more of the model details in this article because I've concluded that more transparency is helpful in building a following, while the AI algorithm will serve as a proprietary firewall. An enterprising analyst could recreate some of this work based on the information provided. My hope is that you'll just follow me, and I'll provide periodic updates. I'll also continue to push the envelope on insights and analytics.

In my next article, I'll apply the same algorithm to a basket of technology stocks. Over the years, they have produced higher returns with accompanying risk, but if the risk can be mitigated with this methodology, why not have a look?

Disclosure: I am/we are long SPY. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Additional disclosure: I trade all tickers mentioned using the subject algorithm.