I've been at the market pretty much my entire adult life, have read most of the literature and honestly don't know.
Back in December 2006 some folks at Credit Suisse took a stab at answering the question, here's their conclusion via the Wharton library:
• Contrarians are all in agreement (?) that a low VIX signals caution while a high VIX heralds rallies. They are right.
• Backtesting shows the ratio of the VIX to its 186-day MA has about 10% predictive power over the SPX 5-month return.
• The signal is worse at predicting short-term moves, while a low reading really means buy puts rather than "sell" VIX predictions have worked well over the last 3 yrs. Today the VIX is 20% below its MA, exercise extreme caution!
Not So Contrarian Anymore
Popular trading wisdom holds that low stock volatility is a bearish signal while high vol indicates a rally. This used to be a contrarian idea: high volatility meant "fear", meaning the weak hands have already been shaken out, while on low vol complacency markets may falter. These days, however, the VIX indicator is pretty mainstream, as likely to be cited on CNBC or by taxi drivers as by contrarian extremists.
Of course the mainstream also once thought that the trend is your friend in tulip bulbs, and more recent crowds of investors loved Pets.com at $11. Clearly its best not always to listen to people, whether mainstream, contrarian or whatever. Historical data makes a more reliable source, though care must
always be taken not to torture it into false confessions. To find out whether implied volatility might really have the signalling power its credited with, we tested the correlation of the VIX relative level to subsequent moves in the S&P 500.
Relative level here is defined as the ratio of the VIX to its own N-day moving average, minus one. So, for example, if the VIX was 15.0, and its N-day moving average was 10.0, its relative level would be 15 / 10 - 1 = -50%. The S&P 500 return was taken from the VIX close to K days out.
Surprisingly (at least to cynical efficient marketeers like ourselves), our results suggest that the VIX has been a strong indicator of direction since its inception in 1993. And in this case one can be unusually confident in the results not because some particular combination of N and K worked, which might result from data-mining but rather because they all did! When it comes to the VIX as a directional indicator, the tulip-bulbers are quite right. But don't scour eBay for old Pets.com share certificates quite yet; some caveats apply.
Interpreting the Results
The VIX is Best at Predicting Medium-Term Moves
The highest degree of linear predictive power from 1990-2006 occured using a 186 trading day (9 calendar month) moving average for the "new" VIX, predicting 114 trading days (5 months) out on the S&P 500. However, the results were nearly as strong for any K and N within a month or two of these values (Fig 2)...MORE (6 Page PDF)