Post by jeffolie on Nov 14, 2013 7:29:07 GMT -6
Nov. 13, 2013
The anniversary no market timer should ever forget
Commentary: In November 2008, the VIX said ‘buy,’ but the market fell 40%
CHAPEL HILL, N.C. (MarketWatch) — This month is the fifth anniversary of something no market timer should ever forget.
It was five years ago that the CBOE’s Volatility Index /quotes/zigman/2766221/realtime VIX -2.34% rose above its previous all-time high. That triggered buy signals from virtually all VIX-based market-timing systems, since the assumption at that time was that record-high VIX levels only occurred at or close to market bottoms.
Yet the market kept falling, and the VIX kept rising. By the time the VIX finally hit its peak five years ago this coming week, it was nearly double its previous record.
That was bad enough, but it got worse.
The VIX’s peak above 80 on Nov. 20, 2008, was a screaming buy signal. Its closing high that day was 80.86, versus a previous record high prior to the fall of 2008 of 45.74. And yet the market continued plunging even more: Over the next three months, the S&P 500 dropped an additional 10.1% and the Dow Industrials fell 13.7%.
So the anniversary I am noting in this column is not one that market timers will want to celebrate. But nonetheless we forget it at our peril.
There are several crucial lessons to draw from that experience:
•Past extremes do not represent an inviolable barrier. Just because the VIX had previously never risen above 45.74 didn’t mean it couldn’t rise even further. An investor in the fall of 2008 who bought into the market when the VIX first hit 45.74 would have lost nearly 40% of his money by the time the market hit bottom in March 2009.
•To the extent chart patterns do have validity, it’s when they rest on a strong theoretical foundation. And yet, the foundation of most VIX-based models was wanting, even prior to the 2008 debacle. Though the VIX was (and is) thought to be an “investor fear gauge,” it actually measures expected volatility. And, needless to say, there is both upside and downside volatility. Rigorous econometric tests of the pre-2008 data showed that VIX levels were not correlated with the market’s subsequent returns at even minimal levels of statistical significance.
•When testing an indicator, it’s vital to avoid what statisticians refer to as look-ahead bias. We’re guilty of this when we assume we knew something at a crucial point in the past when, in fact, that something would have been unknowable until a later point in time. Consider, for example, what the VIX’s record was prior to hitting 45.74. That previous record was 38.20, so for a long time this was the level thought to represent a buy signal. It was a sleight of hand, after the VIX did rise to 45.74, to assume that market timers would have known that a VIX level of 45.74 was a buy signal. By the same token, market timers today are guilty of look-ahead bias when they backtest their models by assuming that investors should act in certain ways whenever the VIX reaches any level close to 80.86.
The bottom line? You should never bet all or nothing on any indicator. But especially one whose theoretical foundation is as weak as it is for many of the popular market-timing models based on the VIX.
If we remember these lessons, perhaps the trauma of November 2008 will not be entirely in vain.
www.marketwatch.com/story/the-anniversary-no-market-timer-should-ever-forget-2013-11-13?link=mw_home_kiosk
The anniversary no market timer should ever forget
Commentary: In November 2008, the VIX said ‘buy,’ but the market fell 40%
CHAPEL HILL, N.C. (MarketWatch) — This month is the fifth anniversary of something no market timer should ever forget.
It was five years ago that the CBOE’s Volatility Index /quotes/zigman/2766221/realtime VIX -2.34% rose above its previous all-time high. That triggered buy signals from virtually all VIX-based market-timing systems, since the assumption at that time was that record-high VIX levels only occurred at or close to market bottoms.
Yet the market kept falling, and the VIX kept rising. By the time the VIX finally hit its peak five years ago this coming week, it was nearly double its previous record.
That was bad enough, but it got worse.
The VIX’s peak above 80 on Nov. 20, 2008, was a screaming buy signal. Its closing high that day was 80.86, versus a previous record high prior to the fall of 2008 of 45.74. And yet the market continued plunging even more: Over the next three months, the S&P 500 dropped an additional 10.1% and the Dow Industrials fell 13.7%.
So the anniversary I am noting in this column is not one that market timers will want to celebrate. But nonetheless we forget it at our peril.
There are several crucial lessons to draw from that experience:
•Past extremes do not represent an inviolable barrier. Just because the VIX had previously never risen above 45.74 didn’t mean it couldn’t rise even further. An investor in the fall of 2008 who bought into the market when the VIX first hit 45.74 would have lost nearly 40% of his money by the time the market hit bottom in March 2009.
•To the extent chart patterns do have validity, it’s when they rest on a strong theoretical foundation. And yet, the foundation of most VIX-based models was wanting, even prior to the 2008 debacle. Though the VIX was (and is) thought to be an “investor fear gauge,” it actually measures expected volatility. And, needless to say, there is both upside and downside volatility. Rigorous econometric tests of the pre-2008 data showed that VIX levels were not correlated with the market’s subsequent returns at even minimal levels of statistical significance.
•When testing an indicator, it’s vital to avoid what statisticians refer to as look-ahead bias. We’re guilty of this when we assume we knew something at a crucial point in the past when, in fact, that something would have been unknowable until a later point in time. Consider, for example, what the VIX’s record was prior to hitting 45.74. That previous record was 38.20, so for a long time this was the level thought to represent a buy signal. It was a sleight of hand, after the VIX did rise to 45.74, to assume that market timers would have known that a VIX level of 45.74 was a buy signal. By the same token, market timers today are guilty of look-ahead bias when they backtest their models by assuming that investors should act in certain ways whenever the VIX reaches any level close to 80.86.
The bottom line? You should never bet all or nothing on any indicator. But especially one whose theoretical foundation is as weak as it is for many of the popular market-timing models based on the VIX.
If we remember these lessons, perhaps the trauma of November 2008 will not be entirely in vain.
www.marketwatch.com/story/the-anniversary-no-market-timer-should-ever-forget-2013-11-13?link=mw_home_kiosk