One aspect about using objective modeling is that the modeler will usually continue to tweak the original model long after it appears to show positive results, regardless of the circumstances. With that said, it should be no surprise that I continue to play with VIX data, seeing if there is a better systematic and simple way to use the VIX to judge the future performance of the market. At the very least, a better way to use the VIX data to raise the likelihood that the market will rise or fall in future periods.
Over the last few weeks, I have been reading a number of articles, reports, and commentary concerning the VIX and specifically the skew. Generally speaking, a skewed distribution is one where the outcomes are unevenly scattered around the average. I could describe skew further, but I think it is far more illustrative to show skew graphically.
In most of case, the skew is used in regards to trading options and volatility. For instance, see here, here, here, or here. But my thought is/was, can the VIX be used to effectively trade equities. The answer appears to be yes.
The model I used takes the 6-month skew of the VIX, standardized, and relates that to the daily forward 3-month returns of the S&P 500. I use data back to January 2000. The model segments the daily returns in to two segments. One segment consisting of daily forward 3-month returns following a standardized VIX skew that is between 0 and -2 and another when the standardized VIX skew is greater than 0 or less than -2. Unlike previous models I constructed using the VIX, there is no period in over 3000 daily periods that I do not receive a 'signal' using this methodology.
The average the daily 3-month returns for market since January 2000 works out to be 40 basis points or 133 basis points for the median 3-month return. Additionally, out of 3,200+ observations, there are over 1,800 days where 3-month return was positive, resulting in a batting average of 56.5%. This is in comparison to the returns and the batting average of the two skew segments. The 3-month average, median, and batting average for periods following a standardized skew between 0 and -2 is 152 basis points, 224 basis points, and 61%, respectively. Ahead of the simple buy-and-hold average. As for the second segment (calculated using returns on days when the standardized VIX skew is either less than -2 or greater than 0) results in an average return of -58 basis points, a median return of 54 basis points, and a batting average of 53%. In my mind, these are interesting results and suggests that the standardized VIX skew can be used to help judge entry and exit points for equities.
Now, I have toyed with the notion of completely revamping the VIX-Trading model portfolio and instead of using some form of standardized VIX, instead use the VIX skew in its place. The one drawback I found using the VIX skew to make equity trading calls is that the signals are relatively short-lived. In fact, the power of any signal diminishes after three months, as both the returns and batting averages begin to converge on the population's statistics after the time period begins to widen more than three months. Once the calculation periods goes to six months, the power of the signal appears fully diminished.
It is my intention to use the standardized VIX skew data to assist in my decision making (for the VIX-Trading Portfolio or otherwise) and I will report on this measures in future posts.Just to note, the standardized VIX skew is -0.85.
Over the last few weeks, I have been reading a number of articles, reports, and commentary concerning the VIX and specifically the skew. Generally speaking, a skewed distribution is one where the outcomes are unevenly scattered around the average. I could describe skew further, but I think it is far more illustrative to show skew graphically.
In most of case, the skew is used in regards to trading options and volatility. For instance, see here, here, here, or here. But my thought is/was, can the VIX be used to effectively trade equities. The answer appears to be yes.
The model I used takes the 6-month skew of the VIX, standardized, and relates that to the daily forward 3-month returns of the S&P 500. I use data back to January 2000. The model segments the daily returns in to two segments. One segment consisting of daily forward 3-month returns following a standardized VIX skew that is between 0 and -2 and another when the standardized VIX skew is greater than 0 or less than -2. Unlike previous models I constructed using the VIX, there is no period in over 3000 daily periods that I do not receive a 'signal' using this methodology.
The average the daily 3-month returns for market since January 2000 works out to be 40 basis points or 133 basis points for the median 3-month return. Additionally, out of 3,200+ observations, there are over 1,800 days where 3-month return was positive, resulting in a batting average of 56.5%. This is in comparison to the returns and the batting average of the two skew segments. The 3-month average, median, and batting average for periods following a standardized skew between 0 and -2 is 152 basis points, 224 basis points, and 61%, respectively. Ahead of the simple buy-and-hold average. As for the second segment (calculated using returns on days when the standardized VIX skew is either less than -2 or greater than 0) results in an average return of -58 basis points, a median return of 54 basis points, and a batting average of 53%. In my mind, these are interesting results and suggests that the standardized VIX skew can be used to help judge entry and exit points for equities.
Now, I have toyed with the notion of completely revamping the VIX-Trading model portfolio and instead of using some form of standardized VIX, instead use the VIX skew in its place. The one drawback I found using the VIX skew to make equity trading calls is that the signals are relatively short-lived. In fact, the power of any signal diminishes after three months, as both the returns and batting averages begin to converge on the population's statistics after the time period begins to widen more than three months. Once the calculation periods goes to six months, the power of the signal appears fully diminished.
It is my intention to use the standardized VIX skew data to assist in my decision making (for the VIX-Trading Portfolio or otherwise) and I will report on this measures in future posts.Just to note, the standardized VIX skew is -0.85.
No comments:
Post a Comment