The VXN is the implied vol on the QQQQs and VIX is on the S&P500Charts below plot the ratio VXN/VIX: very interesting downtrend since 2001 -- any thoughts on why?Link to full chart

Hmmm. A few of thoughts:1. 2001 was an especially volatile time for the cube with the aftermath of the dotcom wreckage2. The Nasdaq 100 has rotated in more non-technology issues since 20013. I think that the S&P has gained more technology exposure (but I could be wrong)I doubt that explains all of this downtrend. Has the beta between QQQQ and SPY decreased over this time frame? To my eye, it looks like the beta of QQQ was high in the earlier years, than it is now.

Tech used to be the hot stock - it's the opposite now -people are very cautious, almost fearful of tech stock - they just don't trust it (or the analysts) after getting burnt in 2000. Behavioural phynance 101. Hence lower vol.Cheers,TJ.

Thanks for both your comments.In thinking about it a little more myself, you can think of the VXN and VIX as composed of two pieces: the'fundamental piece', due to your best predictions of the future volatility of the underlying indices, and the'psychological piece', which is the illusive volatility risk premium.If the vol. risk premium were zero, to justify a current ratio VXN/VIX so close to one, I suspect you almostneed to predict a QQQQ beta of less than one. (Traden4Alpha's remark on beta suggested this.). This would seem -very surprising-, given that most 3-5 year betas for that that I seetend to be 1.4-1.7. Can the composition of QQQQ have changed so much? Or ... maybe the market looks at the historical betas very short-term, say over the last month. I wish I had the energy to develop a rolling 20-trading-day beta using, say hourly observations during the regular NYSE session. This would be about 120 obs. per beta, so not too bad. If anybody can do that and post a chart of how this QQQQ rolling historical beta has behaved over the last few years, it might be very enlightening on this issue.Then, apart from that, even if your best future beta prediction is a more normal 1.4 or so, maybe the changing vol. risk premiums on the two can explain it. Usually this is estimated to be negative (meaning people pay up forvol. exposure leading to VIX > real-world SPX vol.). But, I don't know why this premium couldn't switch sign for the VXN (VXN < real-world QQQQ vol.), a somewhat similar take as TJ's comment, except on vol. If anybody has access to historical and current variance (or vol.) swap prices, and can post a similar ratio chart (say for 6 month swaps),this would be very interesting to look at, too. One swap price would be the fair strikefor SPX vol. The other could be the fair strike for whatever NASDAQ-related index vol. is the most favored for swaps (I don't know.)At the end of the day, after all the analysis, who thinks there is (was?) a good trade here betting that this ratio will rebound strongly from the last chart point (1.06)? (As I type this, the ratio is now at about 1.10)

I agree with your breakdown of the volatility into fundamental and psychological pieces, but wonder if it tells the whole story. What is interesting is that volatility is a measure of the dispersion of expected outcomes, but there are three distinct phenomena that can create an expectation of a spread of outcomes. The first is the true macroeconomic uncertainty -- will the economy tank or boom, carrying the value of the components of the index in one direction or another? The second reflects internal diversification of the index -- to the extent that the components of the index are more or less mutually correlated, expected index volatility (for a constant level of volatility at the component level) will be higher o lower respectively. Third is the diversity of the participants -- ceteris paribus, if all participants in the options market agree on the distribution of future outcomes, there will be lower implied volatility than if participants disagree on the future fundamentals. Your notion of the "fundamental" is incorporated in the 1st and 2nd of my categories and your notion of "psychology" is probably covered in my 1st and 3rd categories.The N100 has changed quite a bit in the past 6 years -- with almost 90 addition/deletion pairs since Jan 2001! The list ( http://www.nasdaq.com/indexshares/historical_data.stm ) doesn't tell an obvious story, although it looks like the N100 has fewer communications companies now than in 2001. There's this article ( http://query.nytimes.com/gst/fullpage.h ... A9679C8B63) in the NYT from late 2001 suggesting that the N100 will become less volatile and more like the S&P 500. The article suggests that even if the N100 retained the same numbers of tech-stock components, the massive drop in tech-stock P/E ratios means/meant a substantive drop in the weight given to technology in the N100.The radically different makeup of QQQQ and SPY also means they should vary differently at different parts of the business cycle or credit cycle. QQQQ has very little exposure to financial, consumer goods, or raw materials stocks. It looks like a solid 6%-10% of the SPY is in petroleum and commodity metals, so that's got to explain some of the differences.

Excellent point -- a lot of the drivers of volatility recently are interest rates changes and oil prices. To the exent that QQQQ lacks financials and energy sector stocks, extra volatility in those sectors may help explain some of the recent volatility convergence with SPX, depending on the correlations with the tech sector and others. On theother hand, this means that SPX continues to be the more diversified index, QQQQ continues to be more weighted inthe volatile tech sector, and VXN/VIX = 1.0, for example, just looks fundamentally wrong to me. (The Times link doesn't work for me.)

Sorry the NYT URL is DOA. Try the Google Cache ( uery.nytimes.com/gst/fullpage.html%3Fres%3D9C05E6DD113EF932A15753C1A9679C8B63+nasdaq+component+weights+lehman+2001&hl=en&ct=clnk&cd=1&gl=us&lr=lang_en">http://64.233.167.104/search?q=cache:an ... lr=lang_en ) or search (http://www.google.com/search?hl=en&c2co ... tnG=Search)Regardless of the rationale for a declining VXN/VIX ratio, a value of 1.0 does seem far too low. Not only is the QQQQ less diversified than the SPY, but many of the N100 components are tech companies whose fortunes are tied to the capex budgets of companies which should make the N100 much more sensitive to the business cycle than is the SPY (which contains more consumer goods companies). I've not traded anything associated with the VIX or VXN. My only concern would be whether one can lock-in a long-term pair trade so that if the ratio remained anomalously low for several months (or a couple of years), the trade would still work. It can be hard to craft a long-term position with instruments that have short-term maturities/expirations because it seems too easy to lose the expected profits in the rollover.

QuoteOriginally posted by: AlanThanks for both your comments.If the vol. risk premium were zero, to justify a current ratio VXN/VIX so close to one, I suspect you almostneed to predict a QQQQ beta of less than one. (Traden4Alpha's remark on beta suggested this.). This would seem -very surprising-, given that most 3-5 year betas for that that I seetend to be 1.4-1.7. Can the composition of QQQQ have changed so much? Or ... maybe the market looks at the historical betas very short-term, say over the last month. I wish I had the energy to develop a rolling 20-trading-day beta using, say hourly observations during the regular NYSE session. This would be about 120 obs. per beta, so not too bad. If anybody can do that and post a chart of how this QQQQ rolling historical beta has behaved over the last few years, it might be very enlightening on this issue.I quote myself.Is the recent QQQQ beta less than 1? Yes!Here is the rolling 20day beta: (noisy estimator with daily obs -- don't have hourlys)Link to full chart

Interesting graph. Did you use OLS or TLS for the beta calculations? I'd probably use TLS (or an errors-in-variables approach) because OLS assumes that all residual variance resides in the dependent variable (which is reasonably true when computing beta for a single stock versus a very diversified index). But for regressing data from two indices, TLS would be more accurate (and probably show slightly higher betas). But the real paradox is in comparing your two curves. In late 2006, for example, the VXN/VIX spikes at 1.6, but the beta is 1.8! These two numbers are extremely hard to reconcile because they say something about the relative amounts of idiosyncratic variance in the SPY vs. QQQQ. In particular, SPY must have a greater amount of non-market (or mis-priced) variance than QQQQ. In general, I'd expect the VXN/VIX ratio to roughly equal to the beta of QQQQ divided by the correlation coefficient between QQQQ and SPY. That is, VXN/VIX > Beta. Data windowing and weighting effects could also underly this paradox -- if the time-to-expiration on the options used by VIX & VXN is longer than 20 days, then the effective beta encoded in VXN/VIX would be lower than that suggested by the 20-day rolling beta.P.S. I'm impressed by your data set!!! Where did you get the Jul 17th QQQQ price used in your graph for your post on the Jul 16th. Is there some data service I need to subscribe to in order to get tomorrows' prices????

Good eye -- I wondered if anyone would notice my typo on the dates. (really through yesterday's close).I don't compute these beta's; just getting them from ivolatility.com. Their documentationsays beta is just the OLS value, using the log P(t)/P(t-1). Nor do I compute the VXN/VIX ratio; justgetting it from QCharts/ESignal. So, there is an element of faith in the data and the computations.Your points about reconciling the two charts are good ones. Of course, beta is backwards looking and VXN/VIX isa forward looking expectation. But the market's predicted beta should be consistent. We don'treally know what that number is. The 20 day rolling beta is probably a very noisy estimatorof even my earlier desired 20 day beta based on hourly returns. A little better might be the 90 day rolling beta, whichlooks a little more consistent with the VXN/VIX ratio: Rolling 90 day QQQQ beta

Yes, the backwards vs. forwards issue does make the beta and VXN/VIX ratio not strictly comparable ( with the exception of the consistency issue that you mention). Yet, I find it very interesting that the 90-day beta and the VXN/VIX ratio seem to cross-correlate remarkable well when, in fact, one might expect the forward-looking ratio to lead the backward-looking beta.I still can help but wonder if one can glean some useful estimates of VXN vs. VIX price distortions by looking at beta and R^2 of QQQQ vs. SPY in the context of a model of how the variances of these two indices are related.

VXN/VIX closed at what is probably a record low of around 0.94Full chartIt's was an ugly day in the market, with the SPX down about 2.5% at the close The QQQQ is indeed showing its recent "low beta" stripes, down less than 1% at the close. Plus, I think the steeper skew of the SPX options has to be contributing to this (historically low?) ratio.

Today's market action makes one think about the current curcuit breaker rules: