Predictive VIX

As you may have noticed there was a bit of a lag last week between posts. While I make no promises to update the blog daily, I will attempt to provide something new at least three times per week. Last week I slipped because I got distracted by some work I was doing on the VIX and the S&P, or more accurately the SPY which is a reasonable proxy for the index itself.


Anyone who has traded options for an extended period has sooner or later developed a theory regarding the predictive value of the VIX. Numerous papers have also been written on the subject evaluating returns on the S&P 500 given certain conditions in the VIX that may indicate a peak or trough in the market. After all, the VIX tends to move in an opposite direction from the market, rising during uncertain periods as the market moves lower and falling during periods of calm as the market moves higher, so there should be some value. What is important to remember is that the future is not only dependant on present conditions but also on how the market got to its present state, as any technician will tell you. So the question might not be what does the VIX tell us about the future of the market, but rather when does it tell us about the future of the market.


One point that has been raised in the past surrounds unusual action in the VIX, in particular the impact of future returns when both the VIX and the SPY are higher on a given day. I followed this line of reasoning but examined only extreme moves higher, greater than 5%, in the VIX while the SPY was higher by any amount. Looking back to March 1, 1995, giving 15 full years of data, there were only 38 instances when the VIX was higher by 5% or more and the SPY was higher by any amount. While I expected these events to be turning points for the market, it turns out that these days tended to confirm trends.


If, when these up/up days occurred, the SPY was higher than it was 60 trading days earlier, then one would enter the SPY on the long side; if lower, then enter on the short side. In either case, the position would be held for 120 trading days (or roughly 6 calendar months) barring a contrary event. In the case of a confirming up/up day during the 120 day period, the holding period would be extended 120 days from the most recent event, again barring a contrary signal. Of the 38 up/up days that occurred during the period in question, 2 occurred within 60 days of the beginning of the data set and could therefore not be acted upon (although I plan on extending the starting point in the future) and 23 occurred during a holding period, thus extending the time in the market. The result was 13 tradable events with holding periods averaging 217 days on the long side and 131 days on the short side. Additionally, the average time out of the market between positions was 117 days. The results are included in the table below. Three points should be noted: 1) no interest for holding cash during the days out of the market was included in the calculation of cumulative return, but that would only serve to increase the final value, 2) no dividends received or paid (don't forget that you have to pay dividends when you are short) were included, in other words this is pure price action, and 3) this compares to the cumulative return on a long position in the SPY from March 1, 1995 through February 26, 2010 of 227.3922%, again without dividends included.

Action	Price on	Date	Price	Period Return	Cumulative Return
CLOSE	CLOSE	1/19/2010	115.06	17.8169%	405.3056
LONG	OPEN	7/28/2009	97.66
CLOSE	CLOSE	3/20/2009	76.71	36.9317%	344.0131
SHORT	OPEN	9/15/2008	121.63
CLOSE	CLOSE	2/28/2007	140.93	7.5719%	251.2297
LONG	OPEN	3/16/2006	131.01
CLOSE	OPEN	3/16/2006	131.01	-8.6499%	233.5458
SHORT	OPEN	11/1/2005	120.58
CLOSE	CLOSE	3/28/2005	117.31	3.0391%	255.6600
LONG	OPEN	10/5/2004	113.85
CLOSE	CLOSE	2/17/2004	116.17	19.5780%	248.1194
LONG	OPEN	6/3/2003	97.15
CLOSE	CLOSE	4/29/2003	92.11	19.1876%	207.4959
SHORT	OPEN	4/2/2002	113.98
CLOSE	CLOSE	4/23/2001	122.24	11.3432%	174.0919
SHORT	OPEN	10/27/2000	137.88
CLOSE	CLOSE	9/8/2000	149.81	6.5808%	156.3561
LONG	OPEN	11/16/1999	140.56
CLOSE	CLOSE	12/23/1998	123.22	12.4065%	146.7019
LONG	OPEN	3/23/1998	109.62
CLOSE	CLOSE	1/12/1998	94	26.7872%	130.5102
LONG	OPEN	11/19/1996	74.14
CLOSE	OPEN	11/19/1996	74.14	-12.7605%	102.9365
SHORT	OPEN	9/4/1996	65.75
CLOSE	CLOSE	7/29/1996	63.02	17.9929%	117.9929
LONG	OPEN	6/1/1995	53.41
			Initial Investment:		100

Since one cannot know that the up/up event has occurred until the close of business on that day, both long and short positions were established on the opening price of the day following the event. As an example, the first long position established on 6/1/1995 was the result of an up/up day on 5/31/1995. Also I have highlighted the two incidents of contrary indications that occurred within the initial 120 day holding period which, in both cases, this caused the closing of a short and opening of a long position on the given day's opening price. If a position was held to the termination of a 120 day or extended holding period, the closing price was used. Graphically, it looks like this:

Clearly not all highs and lows were caught and there were incidents, in some cases of both high volatility that led to in short term draw downs during the holding periods and entering on the same side of the market over various intervals. None the less, the results are interesting. I am in the process of examining the time leading up to the investment periods in more depth to discover what motivations might have existed to inspire extreme moves in the VIX despite positive days on the SPY, so there is likely more to come, but it might be interspersed with strategy discussions.


As a final note, the strategy currently suggests a cash position as we wait for the next up/up event.