Free Historical VXX Data

18Apr12

I use Bloomberg data on the blog to show VXX’s estimated performance prior to the ETP’s launch in 2009, but that’s not an option for most readers.

So to help readers in their own research, I’ve calculated historical VXX data all the way back to 03/2004.

CLICK TO DOWNLOAD HISTORICAL DATA

Data prior to VXX’s launch is estimated based on VIX futures (not Bloomberg), so for the uber nerds, I’ve also included the Excel workbook I used to make those estimates.

CLICK TO DOWNLOAD EXCEL WORKBOOK

You’ll find instructions for updating the workbook on the first tab.

Happy Analyzing,
ms

Note: the II Blog charges a fee for a similar file. I haven’t seen his calculations, but I couldn’t imagine asking readers to pay for something that should be freely available to all, so I created my own version, including formulas and futures data, from scratch. Enjoy.

. . . . .

To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.



19 Responses to “Free Historical VXX Data”

  1. 1 Alex

    Nice work. I have Bloomberg as well. You can also pull up the SPVXSP INDEX ticker which goes back to December 2005. It’s the underlying that VXX tracks. SPVXMP INDEX corresponds to VXZ.

  2. 2 pablofrommars

    You are more than generous on that one!

    Cheers,
    Pablo

  3. 3 brekekex

    Very nice of you, thank you.

  4. 4 Mike

    Thank you for sharing this.

    Question: I noticed that when I input vix futures & vxx data for April 17th, the divisor changed (as expected) and the value of the final VXX estimate for the first day of VIX futures trading (cell L3), March 26, 2004 changed from 1693.12 to 1720.04. Subsequently, when I added the data for today, April 18th, the final value for VXX shifted from 1720.04 to 1695.25.

    Should these types of fluctuations be expected as more data is added to the model? From what I gather, it appears you calculate a divisor on the fly, which relies on the most recent Actual VXX (col D, last cell w/data) equal to the most recent Final VXX (col L, last cell w/data).

    As stated above, I thank you very much for sharing the model, especially free of charge. My intent is to learn.

    -M

    • 5 MarketSci

      Hello Mike – you’re correct, the divisor trues everything up to the last VXX datapoint and will change every time you add a new VXX datapoint.

      Note that the difference between days in % terms never changes (i.e. a 1% difference in the “raw” column will always be a 1% difference in the “final” column), so the data isn’t really changing, just the scaling.

      If you use the workbook just to estimate data up until when VXX launched, it’s a non-issue b/c that first VXX datapoint never changes.

      If you use the workbook to do things like estimate the impact of the term-structure on the daily rebalance between 1st/2nd month futures, it should also be a non-issue (assuming you’re looking at things in % terms).

      In other words, don’t know why it would ever be an issue =)

      michael

  5. 6 Nick Iversen

    You don’t subtract the daily management fee that is charged on VXX. This is fine as we can easily do it ourselves. Also you don’t add the daily interest that SPVXSTR earns – but that’s not too difficult either.

    What I was wondering was why does SPVXSTR earn daily interest? And why does SPVXSP (which XIV is based on) NOT earn the daily interest?

    • 7 MarketSci

      Hello Nick – I mentioned that on the first tab. Could be estimated, just not significant enough for most applications. I also can’t account for tracking error (which is by far the most significant factor of the three).

      Don’t know the answer to your second question off hand, I’d have to take a look see.

      michael

  6. 8 aristotle33

    When will the trading system be available?

    • 9 MarketSci

      Hello Aristotle33 – soon. I wanted the program to go through a rough patch and then recover to show readers what that would look like (I don’t like selling something that’s only had sunny days). Expect something over the next month or so. michael

  7. 10 xls5929

    Michael,

    I finally had a chance to look at your historical data. Good work. I have a few questions and comments.

    1) From the raw VX contract data, there are 33 Month-2 data values that have missing values, How did you fill those values? I filled them by using the same Month-2 to Month-1 ratio as the first subsequent non-missing value.

    2) As your “INSTRUCTIONS” tab stated, the estimate does not account for fees. When I fit my model, I used the following structure where the “c”
    coefficient is an approximation for the fees (I was going to add a term for interest, however I think it biases the backward extrapolation):

    ln(VXX) = a + (b * ln(Futures)) + (c * time) <—– Model B

    where: a, b, and c are "fit" using the data from 01/29/2009 to some final date.

    Your model is:

    VXX = d * Futures <—– Model A

    3) Comparing the backward extrapolations (from 01/29/2009 backwards) for the two models, they are close at 01/29/2009. However, at 03/26/2004, Model B is about 8% higher than Model A. I doubt this will make much of a difference in comparing any future calculations, but I thought I would point it out.

    Thanks again for providing your calculations. If you want to look at my comparison (it's somewhat messy, I've been short on time), I can send a spreadsheet.

    xls5929

  8. 11 David R.

    Thanks a lot for this long-view datasheet.
    Your blog is a true source of inspiration.

  9. 12 Alex

    MarketSci
    Thank a lot for your datasheet. Could you mind to provide the source of the formula which can calculate VXX? By the way it would be very helpful if you don’t mind to explain its principle.

    thanks again.

    Alex

  10. 13 Dai

    Thanks for the calculated data and Excel workbook. Have you done something similar for VXZ?

    • 14 MarketSci

      Hello Dai – unfortunately, no. I use Bloomberg data for my own analysis, so I never took the time to update the workbook beyond what you see here. michael

  11. 15 Jay

    Thanks for the data, MarketSci. On your ‘Estimate VXX’ workbook there are times in 2004 and 2005 where you have 2-month roll periods. Is this what you were referring to as ‘wonky’data during that period?

    • 16 MarketSci

      Hello Jay – forgot where I wrote that, but yes, I’d have to assume that’s what I was referring to. Early on, there wasn’t a contract available every month, so the estimate will be a bit off. michael

  12. 17 Michael Moses

    Would it be possible for you to update the VXX Excel data? It only goes through 4/12. Thanks.


  1. 1 Flirting with Models™ | Reconstructing Volatility-Based ETPs and Some Observations

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com 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