-->What is the VIX really telling us?
They instead keep reincarnating in ever different ways, in hopes that eventually they will capture something truly meaningful in the market's gyrations.

Take the Volatility Index, or VIX, which is a measure of the implied market volatility based on a basket of widely traded options on the S&P 500 (SPX: news, chart, profile). The Chicago Board Options Exchange, the (VIX: news, chart, profile) Volatility Index's creator, refers to it as an"investor fear gauge," and contrarians latched onto it soon after it was created in the early 1990s.

Because high VIX values supposedly represented high levels of investor fear, they were taken to be bullish - on the contrarian grounds that the market likes to climb a wall of worry. By the same token, low VIX levels supposedly represented investor complacency and were considered bearish.

There turned out to be a major problem with this contrarian interpretation of the VIX: It's only half right. While high VIX levels are indeed bullish, low levels are not bearish. Over the last 15 years, in fact, the market has actually performed at an above-average rate following both low and high VIX readings.

In the face of this failure, some contrarians in recent years had modified their use of the VIX, focusing on its relative rather than its absolute value. According to this newer interpretation, it would be bullish whenever the VIX was rising quickly and bearish whenever it was falling rapidly.

But this newer way of utilizing the VIX turned out to have no greater support in the historical record than the original contrarian interpretation.

(For a more in-depth discussion of VIX's history, see my Aug. 21 column in the New York Times.)

Earlier this week, however, I came across yet another way of interpreting the VIX, from Ivan Martchev, one of the trio of editors of the Wall Street Winners newsletter. The newsletter has an excellent long-term track record, according to the Hulbert Financial Digest, and Martchev is a gifted and serious student of the market.

So I sat up and took notice when Martchev argued that:"When the VIX crosses above its [20-day] moving average from depressed levels, it's a sell signal for stocks. When it crosses below the moving average from extended levels, it's considered a buy signal for stocks."

On this interpretation, the VIX flashed a buy signal in early September.

Before I report on how Martchev's interpretation has worked over the longer term, let me point out how the original interpretation of the VIX's significance has now been turned on its head. Instead of higher (relative) VIX levels being bullish, they are now taken to be bearish - and vice versa.

Regardless, my PC's statistical package was unable to find support for Martchev's interpretation of the VIX. As I found with the original contrarian interpretation, his is at best half right: While the S&P 500 performs at an above-average rate following buy signals, it also performs at an above-average rate following sell signals.

The following table gives a flavor of what I found.

Triggering event S&P 500's average return over next month S&P 500's average return over next quarter
VIX crosses above its 20-day moving average 0.95% 2.42%
VIX crosses below its 20-day moving average 0.96% 2.47%
All trading days since 1990 0.78% 2.36%

Notice that the S&P 500 tends to perform at an above-average rate over the next month and quarter after both rising above and falling below its 20-day moving average.

To be sure, Martchev writes that it takes more than the VIX rising above or below its MA to trigger a signal. Another precondition is that it must occur from"depressed levels" (for a sell signal) or"extended levels" (for a buy signal). So the results reported in the table are not the final word.

But I was still unable to find support for Martchev's interpretation in any of a number of other econometric tests that I ran that looked at both the VIX's crossing its 20-day moving average as well as where the VIX's current level stood in relation to its recent past.

In other words, I am not confident that this latest use of the VIX will turn out to be any more successful than prior ones.

In response to all of you who wrote to me following my previous expressions of skepticism about the VIX's worth as a market timing indicator, let me say that I am entirely open to a market timing theory based on the VIX that actually works. But don't just assert that this or that theory works; you need to show that it works in a statistically significant way.

I'll immediately report any market timing theory of the VIX that can jump over this hurdle.

<ul> ~ http://www.marketwatch.com/news/story.asp?guid={50E1AAC0-5F57-46A5-B85A-CD99ADCE</ul>