Skip to content

Research Synthesis: The VIX as a Moderator of Small vs Large Cap Performance and the “Generals Lead the Troops”

January 17, 2010

Readers are encouraged to check out the following background reading:

On the VIX as a moderator of short-term relative performance between large and small caps:

On the use of the S&P500 and Russell2000 relative strength relationship to predict S&P500 returns:

Sometimes interesting findings results from combining two different ideas that are both good—sort of a chocolate and peanut butter kind of thing. In this case, the relationship between the S&P500 and the Russell2000 is interesting because it captures the lead/lag autocorrelation effect documented in the academic literature. In addition it also  captures the difference in risk premium to holding large diversified companies that are more heavily tied to economic factors versus smaller and more concentrated companies that have: 1) less sensitivity to economic factors most of the time, but more stock-specific risk 2) a high sensitivity under times of extreme stress–generally due to higher credit risk. Nonetheless, the explanantion and theory is a subject for another day (or another blog), my goal is to show that there is a synergistic effect by combining both. In the MarketSci post (see link above), a strategy was introduced that showed that the S&P500 tended to outperform when the 12-month or 1-year return was higher than the Russell2000 index. This phenomenon was labelled as the “Generals Leading the Troops.” This strategy did better   with less risk and less exposure than a standard buy and hold strategy. In the link above from CXO Advisory they review a paper by Dean Leistikow and Susana Yu who tested the significance of the relative VIX level as a signal for short-term switching between the Russell2000 (small-cap) and the S&P500 (large-cap) stock indexes. They found that the S&P500 outperformed the Russell2000 when the VIX was high, and underperformed when the VIX was low. These high/low levels were identified as being within 10-20% above or below the 75-day moving average of the VIX.

So with that background, lets get down to our little experiment to see how these two separate findings work together. In terms of methodology: I used the S&P500 and Russell2000 index prices from Yahoo Finance to increase the sample size of the dataset. I defined relative outperformance for the relative strength strategy as the 252 day return of the S&P500 being greater than(less than) the 252 day return of the Russell2000. For the VIX, I used the 75-day sma mentioned in the article, and decided to split the difference of 10-20% as being either 15% above or below the sma. Lets look at the results for each strategy in isolation, and in combination.

S&P500 Average Daily Performance Under Different Conditions (5000 bars)  
(non-compounded–ie fixed bet)      
    avg daily ret  
  avg daily return /avg daily sd w% (next day)
Baseline (Buy and Hold the S&P500) 0.03% 0.17% 53.1%
S&P500 (SP) is leading Russell (RUT) 0.09% 0.71% 55.0%
Russell (RUT) is leading S&P500 (SP) -0.01% -0.10% 51.5%
VIX/75day sma>1.15  (Vsma75>1.15) 0.10% 0.88% 54.8%
VIX/75day sma<.85 (Vsma75<.85) -0.02% -0.28% 51.8%
SP> RUT  and (Vsma75>1.15) 0.22% 2.98% 58.0%
SP> RUT  and (Vsma75<.85) 0.10% 2.60% 54.60%
RUT> SP  and (Vsma75>1.15) -0.01% -0.16% 51.40%
RUT>SP  and (Vsma75<.85) -0.11% -2.10% 49.66%

Conclusions: The results for combining the two studies showed  higher absolute and risk-adjusted performance  than either study individually over most categories. It appears that the dual screen of relative strength and relative VIX are more powerful together than in isolation. On the positive side, when the S&P was leading the Russell and the VIX was high, the S&P had by far its best performance in absolute and risk-adjusted terms. On the opposite side, when the Russell was leading the S&P and the VIX was low, the S&P put in its worst performance in absolute and risk-adjusted terms. As a third note, the S&P performed well when the VIX was low and the S&P was outperforming the Russell. While not shown for brevity’s sake, it is clear that the S&P underperforms buy and hold in the “normal” VIX environment (1.15>VIX/sma>.85) regardless of whether it is leading or lagging. Creating long and long/short strategies based on these regimes is a subject for a future post.

5 Comments leave one →
  1. dha3936420 permalink
    January 19, 2010 8:39 am

    DV, I know you don’t typically do economic indicators . .. . . But, what do you think of research suggesting: if (28-day T Bills – urban CPI < 0) then small caps else large caps (actual formula in the book includes 12 months data)

    • david varadi permalink*
      January 20, 2010 1:04 am

      hi , indeed there is a slight difference in rate sensitivity, and something i am in fact aware of. likely this is due to the credit factor–where smaller firms get hurt more by higher financing rates since their borrowing costs are higher than large diversified firms.
      great comment!

  2. Bill permalink
    January 20, 2010 1:57 am


    I was doing some research with DVO indicators on my own. I did get some very good results.

    My question to you is if there is any need for more optimizing the settings.
    I am doing PERCENTRANK on 250 days should i go for a smaller time period. How do you make it adaptive. Adaptive at some point of time looks like curve fitting.

    Also i am ausing d7/(.3*b7+.3*c7+.4*d7) , where b,c,d are High,Low,close.

    each days data is then again smoothed = .5*today +.3*yesterday +.15*2 day before +.05*3days before.

    Now the percent rank is on this final number.

    I can post my excel if you would be interested to see it.

  3. Herb Geissler permalink
    January 20, 2010 2:50 pm

    Looks interesting, but this analysis is telling only part of the story. While the avg daily return may be much higher when both conditions are met, how many days per year does that occur (on average). The VIX is very spiky and does not spend much time 15% above its MA75. Similarly, I think the SP moving annual return is greater than the RUT annual return less than half of the time. If both conditions must occur, the % occurrence may be a small portion of the year. Making a quarter of a percentage point per day for only a few weeks per year does not appear attractive

    • david varadi permalink*
      January 20, 2010 3:02 pm

      herb, all good comments–suprisingly the absolute performance numbers were still high–however I will be doing a follow up article on this that coincidentally will address your concerns. as a hint, i will be normalizing the VIX regime and the Relative Strength into percentile units.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: