Weekly Wrap and the Week Ahead

A slow and steady week in VIX futures brought down all points along the curve with some additional emphasis on the front two months. Overall the term structure remains somewhat compressed, with just 4.7 points separating 1st and 7th month. 

There was a slight divergence in correlation between XIV and SPY today as we closed at new SPX highs - a signal for some caution in both XIV and SPX longs. From the intraday SPY arbitrage model:

Treasuries and high yield corporate bonds diverged  from SPY during the week as well, widening the gap on the daily SPY arbitrage model, with SPY trading at a $5.00 premium to the model:

And yet there's no reason VIX can't head lower given that actual volatility over the past 3 months (HV63) is just 10.8.

Trading Volatility - Terms of Use

This is a short note to inform you that there are new Terms of Use for the Trading Volatility website.

Please familiarize yourself with these terms, as your use of the website constitutes your acknowledgement and agreement to the contents.

Update to VIX Futures Data Charts

I've made some adjustments to the charts on the VIX Futures Data page.  Changes are mostly cosmetic but I've also added a compare feature which allows you to view the values of all series at a given point by hovering over with your mouse cursor, as shown below.

Secrets of VXX: The Value of the Roll Yield

Most people who trade VXX realize that it faces constant headwinds caused by a term structure that is in contango roughly 90% of the time. However, something you may not know is just how big that headwind is at any given time. To help with this I have posted the weekly roll yields for short-term and med-term VIX futures securities on the VIX Futures Data page.

For more on where these numbers come from, continue reading below.

VXX tracks the performance of the first and second month VIX futures. However the actual dollar weight of each month held by VXX changes each day as iPath continuously rolls their portfolio in order to target a constant maturity weighted average futures maturity of 1 month. You can view the current status of these weights on the iPath site.

Summarizing the roll process as defined in iPath's 200+ page VXX prospectus:
1) At the beginning of the roll period all the weight is allocated to the first month contract.
2) On each subsequent business day a fraction of the first month VIX futures holding is sold and an equal notional amount of the second month VIX futures is bought. The quantity bought/sold depends on the number of business days in the roll period (the number of days varies but averages out to about 21).
3) The following roll period starts after all weight from the front month has been sold and the old second month VIX futures contract becomes the new first month VIX futures contract.

This process has a measurable effect on the price of VXX depending on the term structure.
When the term structure is in contango:
 - At the end of each day VXX must sell a quantity of first month shares and replace them with a quantity of second month shares at a higher value. The "buy high, sell low" scenario is a negative roll yield for the fund (assuming spot VIX is lower).

When the term structure is in backwardation:
- At the end of each day VXX must sell a quantity of first month shares and replace them with a quantity of second month shares at a lower value. The "buy low, sell high" scenario is a positive roll yield for the fund (assuming spot VIX higher).


The actual value of the roll yield depends on the differnce between and 1st and 2nd months VIX futures and spot VIX, and the number of days in the roll period (more days, smaller roll yield). 

The same process applies to inverse VIX futures products. In order to obtain the value of the roll yield for the inverse products (XIV/SVXY), just change a negative sign above to a positive, or vice versa.

In my trading, I like to be able to take advantage of this roll yield by holding securities that have a large positive roll yield, so our ideal timeframe for trades is on the order of weeks/months. Therefore, instead of looking at a daily roll yield we prefer to track the weekly roll yield, obtained from multiplying the daily roll yield by 5 (an approximation).

You can take a similar approach to approximate the roll yield for ZIV (inverse med-term VIX futures) using 4th and 7th month futures. The results of these calculations is what I have listed on the VIX Futures Data page.

To give the roll yield values some context, below I've listed the average, min, max, and standard deviation of the weekly roll yields for XIV (past 9 years) and ZIV (past 5 years):

XIV:
 avg:    +1.2%
 low:   -11.7%
 high:   +7.1%
 stdev: +1.7%

ZIV:
  avg:    +0.3%
  low:    -4.8%
  high:   +5.2%
  stdev: +0.5%




FAQs
Q: When does the roll period start?
A: The roll period starts on the Tuesday prior to the monthly CBOE VIX Futures Settlement Date (the Wednesday falling 30 calendar days before the S&P 500 option expiration for the following month.)

The Week Ahead In VIX Futures

Very little changed on the week for VIX futures, with most months only gaining 1-2%. Spot VIX however came off its low of 11.05 last week and gained 22%. Week-over-week term structure (does not include March VIX futures which expired this week):

The term structure may look normal but if you consider the steepness of the curve you'll see it flattened a bit this week, down to just 4.5 points separating 1st and 7th month (from 5.8 points last week).  Month 1 to month 2 flattened as well, down to 1.25 points from 2.1 last week. This results in a relatively small positive roll yield in XIV, which by itself, isn't likely to get you very far.

Forward implied volatility remains reasonably priced relative to 1- and 3-month historical volatility, with a ~25% premium. From the VIX Futures Data page:

Of course most of next week's action will likely be influenced by the outcome of the Cyprus bailout negotiations. If they can reach a favorable outcome, VIX is likely to fall back toward 1- and 3- month historical volatility, currently near 11. There are many opinions on what will happen but I found this piece from the former Vice Chairman of Moody's to be particularly interesting.

Lastly, a chart of the intraday SPY arbitrage model from today. Model components broke from tracking the S&P's +0.8% move today, with treasuries and volatility remaining flat and high yield credit being sold.

Looking at the New VIX Futures Months After March Expiration

Today is the last day of trading for March 2013 VIX futures.  As discussed would happen in my week ahead post, March futures closed within about 3% of VIX.

Closing term structure:

Looking forward to tomorrow April currently sits at 15.4 and May at 16.25 resulting in a smaller contango spread of -0.85, which applies to XIV, VXX, and UVXY. With VIX at 14.39, April VIX futures are just 7% higher.

For those looking at trading ZIV, the contango spread between month 4 and month 7 will start narrower at -1.5 tomorrow.

Implied volatility has popped up a bit off of realized volatility, but with a volatility risk premium of 26% VIX pretty well priced once more. Here is the current chart from the VIX Futures Data page:

There was not much room to play in the intraday SPY arbitrage model today as the model stuck closely to the SPY.

And in case you missed it I posted about a new tool today to measure Twitter sentiment on VXX which was positive earlier in the day but is unsurprisingly heading toward 50/50 as VXX ended the day flat.

Using Twitter to Measure Sentiment on $VXX

I was chatting with a friend yesterday who is working on a project to scour the Twitter stream universe for data. I immediately started wondering what it might look like to use Twitter to measure conversations on trading of VIX-related products, so I started experimenting a bit.

The result of this work is now located on the Twitter Sentiment page. The graphs there provide a count of the public tweets about $VXX and $UVXY based on whether they have a positive or negative sentiment associated with them. 

Here is a screenshot:

With the downward bias of VXX I would normally expect sentiment on the this to be negative. But since VXX is experiencing a rare moment in the green today I suppose it makes sense that Twitter sentiment is positive.

While I was initially polling for tweets every minute I have scaled back to hourly intervals. I'll continue to keep an eye on this to gauge its usefulness.


UPDATE: 

As I continue to look through the data I thought it would be interesting to get a net value of positive tweets minus negative tweets.  After just this first day of data, it's interesting to see how net Twitter sentiment loosely matches direction of VXX during the day:

The Week Ahead In VIX Futures

Last week wrapped up with VIX and front month futures down heavily, to the point where our normally smooth curve is starting to resemble a cliff.  VIX Futures term structure week over week:

While it is not abnormal to see the front part of the curve steepen as front month futures converge with spot VIX as expiration draws closer, forward 30-day volatility (VIX) is now less than actual realized volatility over the past one and three months, suggesting that VIX is underpriced (a negative volatility risk premium).

Historical volatility vs implied volatility chart from Friday:

With March futures currently 11% above spot VIX it should be noted that this convergence can happen by having a rising VIX, not just a declining front month as some may have grown accustomed to.

VIX Pushing Lower Bounds Set By Historical Volatility

Protection through the purchase of options became even cheaper today as 30-day forward implied volatility (VIX) hit 11.46 -- just 2% above actual realized volatility for the past month (HV21) of 11.24. HV21 is now also roughly equal to the actual volatility over the past three months (HV63), which as I discussed previously, can be used as a reasonable lower bounds for VIX.

Below is a view of today's HV vs IV chart from the VIX futures data page to illustrate where VIX is in relation to historical volatility (three month view):

Next is a view of VIX vs HV63 from June 2004-Present which shows how HV63 can be used as a rough approximation for the lower bounds of VIX in the immediate future under most circumstances:

And while it is possible to say that VIX is running out of room to fall given current levels of historical volatility, VIX futures are another story entirely. As the VIX futures expiration date approaches (March 20th), March VIX futures and spot VIX are likely to converge to within a couple percent of each other (discussed here). At 9% above VIX there is quite a ways for March VIX futures, and consequently VXX, to fall -- assuming spot VIX remains near its current levels.

Looking at April VIX futures, which will become front month futures used in the calculation of VXX and XIV after March expiration, you can see that they now trade at 14.7, more than 27% higher than spot VIX right now. This will provide a new catalyst for a lower VXX and higher XIV as long as VIX remains flat.

Market Wrap - 3/13

In what was otherwise a rather unremarkable sideways day in the markets, I found the movement in the underlying components of the SPY arbitrage model to be quite interesting.

Looking at the model compared to the S&P 500 (SPY) there were two arbitrage opportunities during the day and another rapid convergence of the two into the close.

Taking a look at the components level you can see the Treasury bond component (TBF) dive after strong demand for the 1:00pm 10-year auction drove up prices. As a result the model became deeply discounted relative to the SPY for most of the afternoon until a last minute rally in high yield corporate bonds (HYG) to very neatly close the gap between the S&P 500 and the model.

While the volatility component (XIV) closely tracked the S&P 500 all day, yesterday was a different story as it dragged heavily.

All in all this seems to represents some tension in the market as investors start to take increased caution and diversify a bit after what has essentially been 9 weeks of going nowhere but up for stocks.

I'd also like to point out the movement over the past 6 weeks of stocks vs the dollar index (UUP), which typically move inversely to each other. Another reason for caution in my opinion.

New SPY Arbitrage Tool Available

I'm happy to announce the availability of a new trading tool on the Trading Volatility site, the SPY Arbitrage Model. 

The model is updated in real-time throughout the day, plotting the S&P 500 (SPY) against a model of the implied value for the S&P 500 as derived from a collection of related assets (volatility (XIV), interest rates (TBF), and credit (HYG)). With this model a trader can place relatively low risk arbitrage trades to take advantage of prices as they diverge and recouple during the day.


Let's take a look at Friday's data: 

To get the most our of this trade you want to look for points at which the difference between SPY and model are large -- generally +/-$0.20 is a good rule of thumb depending on your trade sizes and transaction costs.  On the areas I highlighted above, the second area I highlighted (towards the end of the day) provides a better opportunity than the first, as the difference between SPY and model is $0.30. The trade approach is then to sell the SPY while simultaneously buying the components in the model. Once the SPY and model converge, positions should then be closed.

To maintain near market-neutral on the trade the weighting of each component needs to be balanced according to the beta for each component, at least as a starting point (this is a topic I'll cover in another post). Some adjustments can be made based on where each component is trading. Again, consider data from Friday:

When the arbitrage opportunity opened up in the last hour of trading, two components of the model were lagging (XIV and HYG) the SPY, while TBF was essentially in line.  In this case an alternate trade would have been a short position in the SPY, but only take longs in XIV and HYG since it is less likely that TBF would make an upward move while SPY moves lower. This helps to maximize gains and reduce transaction costs.

Important note: If your positions are small relative to your transaction costs or if you are subject to wash sales rules, this may not be a good strategy to employ. I recommend crunching the numbers for your specific situation to determine if this strategy is right for you.

March VIX Futures Expiration Date and a Suppressed VIX

It's one of those unusual months for VIX where there are 35 days between monthly options expiration day instead of 28 as March options expire on the 15th and April options expire on the 19th. This means that March VIX futures don't expire until the 20th this month, since VIX futures always expire 30 days before the next month's options expire. If you don't like counting you can always find a calendar for VIX futures expiration here.

This extended month also causes some issues with the value of VIX during the next week when calculating its constant 30-day volatility.

VIX is calculated using weighted values of near-term (VIN) and next-term (VIF) options of the S&P 500, with near-term options having at least one week to expiration (values can be found on the CBOE's page). Once options expiration is less than a week away VIX rolls to use the next term's values as the new near-term value, weighting the new VIN with a value of 1 and the new VIF with a value of 0. As the month progresses the value of these weights shift toward VIF until the next expiration when VIF has the weight of 1 and VIN has the weight of 0.

For example, if you look at the value of March and April values on the  CBOE's page you will see that current values are 10.47 and 12.71, respectively.  Looking at VIX today you will notice that it is essentially equal to the April VIX value (this should not be confused with April VIX futures).

Since options expire next Friday, VIX must use April and May as VIN and VIF values. However, since both April and May expiration are more than 30 days away, the near-term weight is greater than 1 and the next-term weight will be negative.

Specifically, on Monday April will have a weight of 1.25 and May will have a weight of -0.25.  From there the roll continues as usual with weights shifting a little bit each day (on Tuesday the weights are 1.2 and -0.2, then 1.15 and -0.15 on Wednesday, etc.).

As an example using current values of April and May of 12.71 and 13.76, come Monday VIX will be calculated by (1.25)*(12.71) + (-0.25)*(13.76) = 15.89 - 3.44 = 12.45. You'll notice that the VIX here is actually less than both VIN and VIF, which seems abnormal but is not unusual during these weeks. Sometimes this results in a temporarily suppressed VIX, especially when the next-term month is priced substantially higher than the near-term month.

For more information on how the VIX is calculated see the CBOE's VIX white paper.

Tracking Historical Volatility Vs Forward Implied Volatility

Traders of options and volatility often compare the actual volatility of the market using various lookback periods (10 day, 1 month, 3 month, 6 month, etc) to forward volatility, as expressed by the VIX and VIX futures, to help determine whether options are relatively cheap or expensive.

Therefore I've added a new chart to the bottom of the VIX Futures Data page to track 1 and 3 month historical volatility (HV21 and HV63) of the S&P 500 against 30-day forward implied volatility (VIX) and front month VIX futures (M1).

In the graph I've chosen the actual volatility of the S&P 500 over the past 30 days (HV21) and tracked it against the 30 day implied (forward) volatility, a common measure of the volatility risk premium.

However, I find HV21 to be a bit too noisy so I will often look at HV63 vs VIX instead since the two are more tightly coupled with a smaller average difference and smaller standard deviation. I find it helpful to use HV63 as a rough approximation for a lower bounds of VIX in the immediate future under most conditions.  The exception here is that HV63 will lag quite a bit and be higher than VIX after after large volatility spikes.

As for establishing an upper bounds, HV does not take into account future event risk and therefore is not useful in predicting how high volatility might go. This is demonstrated in the graph in late December as the VIX rose on concerns about the Fiscal Cliff while HV stayed quite low. After the event risk had passed when the deal was made on Jan 1, forward implied volatility fell back in line with HV.

I've included front month futures (M1). Since I focus on trading VIX Futures ETFs I care much more about this value than the actual VIX. Specifically I look at the premium of M1 over VIX to help determine if VIX futures are relatively cheap or expensive. Of course I also look at all months of VIX futures but I already have a graph dedicated to tracking those prices.

The actual values of these metrics can be found in the Quotes box on the VIX Futures Data page.

VIX Futures Weekly Wrap

The big news in the VIX world this week was Monday's 34% gain. And while the S&P 500 retraced the losses to close out the week barely green (+0.17%), VIX remained elevated finishing +8.4% on the week.

Unlike last week, VIX Futures followed the move in spot VIX in a more standard fashion with some additional emphasis on the first three months, compressing the term structure by 1.4 points:

Despite this move in VIX during the week, the volatility risk premium actually fell to 23% as actual volatility over the past month (HV21) rose to 12.55 from 9.89 last Friday. This suggests VIX is pretty well priced given recent market action.

Weekly VIX ETF scoreboard:
VXX:       +8.2%
UVXY:  +12.8%
XIV:       -11.2%
ZIV:         -3.4%
SPY:     +0.17%