Click here to download all the XIV, front month VIX futures and contango and backwardation data since 2004. Note that the data was updated on August 21 2013 02:03:23 (California time).
The estimated XIV price is modeled based on VXX data. Market VXX data when available (since 2009) and estimated VXX data before. VXX data was estimated based on VIX futures. Like this I obtained the price of the XIV since the VIX futures started trading (2004/march) until now. The price model is also used to forecast future XIV data based on VXX prices.
Historical extrapolation could be done because with the XIV futures the VXX was calculated and therefore the XIV historical behavior could be deduced. You could also do historical extrapolation using the underlying index. It gives the same result but the underlying index was available after the VIX futures started trading so less history can be extrapolated by that method.
This is the model:
 For any given day n, calculate R(n) as the return that you would have made from shorting VXX from day (n-1) to day n (or from shorting the underlying index if you prefer not to use VXX data).
 Apply XIV’(n+1) = XIV’(n) + XIV’(n) x R(n+1)
Take as initial value the market value of the XIV on the first trading day: XIV’(1)=9.56
 Calculate the daily tracking error, F, by solving:
XIV(n+1) = XIV’(n+1) x F
With border conditions:
XIV’(N)= Market price of the XIV on the last trading day (last close).
The daily tracking error, has a small part due to the daily management fee. The estimated XIV’ model, based only on  gives higher values than the market prices. That difference with respect to the market data is used to calculate the daily tracking error. The tracking error is a headwind, the final model  is 11.9% per year smaller than what it would be if no tracking error nor management fees existed .
Doing the above we obtain a model that gives an error of plus or minus 0.8% in average when compared to market data. Some days more or less, but on average extremely close to it. Both measures of root squares error and absolute deviation give almost the same value.
This is how the model and the data looks like. Note that the graph was made with info up to the 24th July 2012 but the spreadsheet has updated data :
Looking at the graph you can see that in low volatility periods, when the VXX is falling the XIV goes up and vice versa. You can also see that while the VXX has fallen 99% the XIV has multiplied by 3. That’s why several prefer to short the VXX instead of going long XIV. The XIV had a very long and big run up from under 5 til the low 30s. This happened in the low volatility years of 2004-2006, until it peaked in February 2007. The extreme low volatility caused the VXX price to fall dramatically and correspondingly the XIV to increase. When volatility increased in 2007 the XIV fell violently. Notice how after bouncing at 10 at the end of 2007 the XIV falls back much more to under 5 in the 2008/2009 period of extreme high volatility which simultaneously caused the VXX price to more than triple for a while. After the 2008/2009 period it made a big come back, temporarily interrupted by the 2010 fall, to reach a bit under 20 in 2011. Then it fell back again to 5 during the 2011/august volatility spike which in parallel caused the VXX to almost triple. It recovered, quite much, to over 10 recently. With so much ups and downs it is understandable that why many traders like the XIV.
If you want to play conservative the goal is to buy it at the moment when you want to sell short the VXX. So at moments of extreme volatility, when its highly probable that things are so bad that volatility will fall or not increase anymore. While still at lower levels than the previous 2 peaks, the product seems to have the potential of having big spikes if you manage to buy it in periods of very high volatility (end 2008 / begin 2009 or august-sept 2011).
I like the strategy of permanently shorting the VXX. Adding to it at extreme volatile periods. But having seen this historical behavior I think that I can have less risk if I short the VXX and also go long the XIV in periods of extreme volatility. Note that it’s not a hedge. It’s in both cases betting on a volatility fall. But like this half of the amount (the XIV part) has limited loss (can not lose more than 100%). While the other half has unlimited potential loss since if the VXX more than duplicates I could have more than 100% of paper losses. That strategy seems safer than going only short on the VXX since at least like that I reduce the unlimited potential loss to half of the amount. Actually the good thing about dividing that trade is that I could use much less than half for the XIV long because if it explodes, due to volatility falling from high levels, the XIV could multiply itself, while the short VXX can theoretically at most go down to 0. The total amount at risk could be reduced since I would need a smaller amount to go long on the XIV than the amount used to short the VXX to obtain the same profit. That’s a good argument that XIV traders often give. Not only do they limit the theoretical loss to just what they invested but they also invest less. So if you plan to short volatility it is worth considering to short the VXX and at the same time to go long on the XIV. Both have advantages, VXX has in the long run an almost guaranteed fall but XIV limits the paper loss or margin call risk better and requires smaller amounts.
If you want to calculate new XIV values you will have to implement the formulas. It is good that you implement how to calculate the data in order to have a better understanding of the XIV dynamics. The XIV historical prices or the VXX prices are available from yahoo finance and you can use them to keep on updating the model since I may not update it regularly. The VXX data back to 2004 is here.
You may not have time or do not wish to implement the pricing formulas or want future XIV forecasts based on VXX values. I sell for 25 US$, or the equivalent on any currency accepted by paypal, the same spreadsheet with a data model for historical and future forecasted data. It includes:
1) Pricing formulas.
2) Forecast future values based on VXX values, with optional levels of random market noise to see different outcomes.
3) Latest and future data updates if you don’t manage to do them.
4) Front month VIX futures, contango and backwardation data up to 2004.
Besides the historical and modelling data it has the advantage that you can understand how the VXX affects the VIX price. You can also use it to make forecasts of how the VIX will be affected depending on future VXX prices. It includes XIV price forecasts based on VXX values with which you can play with to estimate what the XIV price will be depending on different scenarios.
You can make the payment via paypal to my email firstname.lastname@example.org or via the button below. Via the paypal payment button, besides allowing you to use paypal, you may also chose cards such as visa, master, american express, discover or maestro. You may pay in dollars or the equivalent amount in the supported currency of your choice. Once paid I will be notified by email and I’ll send you the excel spreadsheet with the latest data and the formulas. I’ll gladly give you e-mail support if you need it. You may request an invoice by email if you specify it before or during the payment process.
Alternatively if you are in Europe you can send a transfer, free among EU members, for the equivalent amount to my European bank account.
Note that if you trade both the XIV and the VXX and are interested in the VXX pricing model I sell both for 50 US dollars, or the equivalent on any currency accepted by paypal. That would be cheaper than buying them separately (the XIV model alone for 25 US$ and the VXX model alone for 35 US$). Both spreadsheets work well together, you can use them to, for example, directly predict the price of the XIV only from the price of the 1st and 2nd VIX futures. How ? –> With the 1st and 2nd VIX futures you predict the VXX and with that you predict the XIV.
Originally I made a VXX pricing model which I used to short the VXX. Having done it initially for myself I was surprised to see people adopting it. Some among them who trade with the XIV asked me about a XIV price model or how it would have behaved in the past. I thought about it sometimes a bit but never came to a solution because I prefer shorting the VXX due to the faster decay. I finally did it when someone offered to pay for it.
It’s basically a tool you can use to guide and feel more under control with your trading. Anyone else interested in sharing modelling or trading ideas feel free to contact.
Hope this data is useful and if you find any interesting patterns by analyzing it please do not doubt to let me know !
PD1 -> Why doesn’t the XIV return to the same level when the VXX goes up and down ? If the VXX goes from Price A to B and then back to A then the XIV should also come back to its starting price. Unfortunately that does not happen. No product has been created that can reach close to that level. If it had then the XIV would be more than twice its value by now. It would be possible if a perfect inverse could be obtained, like the name of the product often (mis)leads to believe. But what the product does is to go up or down the opposite percentage of the VXX on a daily basis, but opposite in sign is not the same as an inverse. Many traders think that the VIX should behave like a mathematical inverse and get disappointed and poorer when they realized it does not.
PD2 -> How is Contango/Backwardation calculated and why does it affect the VXX and therefore the XIV price ? That is answered in the VXX blog post where it says PD4.
PD3 -> Why does the calculated data change as more data becomes available ?
That’s due to the way that the model is designed to learn from available market data and extrapolate. All calculated data changes depending on the amount of market data that the model learns from. Not only does the first 2004 calculated value change but all of them do.
A way to have a fixed initial 2004 calculated value is by arbitrarily fixing it. But that would be probably an error since you do not know what it was. At most you can calculate it by learning from the available market values.
Differences in calculated values happen every day. They are normal because the data is calculated all the way backwards to 2004 by taking into account all the available market data since the XIV started trading up until the last available market value. The model fixes the 1st and last available XIV market values and uses all the XIV market values in between to calculate the XIV and project it back to 2004. That’s a mathematical consequence of the model (see XIV model above).
For example from 2009 to 2012 there is one less year of data than from 2009 to 2013. So its normal that the calculated values, including the first calculated value, on both cases differ.
The changes may look big and concerning but they are actually small if you consider that the annualized percentage growth of the XIV in both cases is almost the same.
You can also graph different series of calculated data taken in different periods and you will notice slight differences. No calculated data series is better than another, that depends on the market data quality. Most probably the more market data that you feed the model the better that the calculated data will be, since in such case the model will use more information to learn from.