Harvard academics reveal blueprint for avoiding stock crashes

[NEW YORK] Good news for everyone waiting for a bubble to burst. Predicting sectorwide crashes in the equity market isn't impossible.

That's the conclusion of Harvard University researchers who studied boom-and-bust cycles in publicly traded US industries since 1926. In a new paper, Robin Greenwood, Andrei Shleifer and Yang You found that while not every violent advance in shares ends in horror, those that do share common traits. Among them: rising volatility and greater share issuance.

As is often the case with academic inquiries into financial assets, the study is a rebuttal to the efficient markets hypothesis, whose Nobel Prize-winning inventor Eugene Fama questioned whether anyone could identify bubbles that ended up wiping out returns.

While the Harvard authors found that much of Fama's view is correct, especially that sharp run-ups in shares aren't by themselves progenitors of market meltdowns, they say tools exist for bailing from the rallies most likely to end violently. Keeping out of crashes can add 10 percentage points to an investor's long-term return, they said.

"When investors look at a very large price runup in an industry and they are concerned if there is a potential bubble, they should look at some of the other non-price features and behaviour as a guide to what might happen," Mr Greenwood said by phone .

"People who have studied bubbles have been solely focused on the price action. We wanted to have a serious attempt to bring in some other information." Fama didn't reply to an email and phone call seeking comments.

Alas, for investors looking for guidance on whether to sell now, the paper holds few clues. The reason is simple: nothing going on in the market today would've qualified for investigation under the authors' base case for bubbles. Maybe it will one day, but the academic bar for studying excesses in the stock market is high, and even industries that seem to be overheating in the Trump bump have a ways to go before entering the history books.

Mr Greenwood and his colleagues defined a bubble as when share prices at least doubled within two years and found that among industries that rallied so much, returns were generally no better or worse than average afterward. Two things explain this. Surging stocks often keep surging, and when they do crash, they sometimes don't fall enough to wipe out previous gains.

At the same time, if you can sidestep a crash, you should. And a methodology for doing so forms the meat of the study. The authors found that roughly half the rallies they examined ended in routs of 40 per cent or more, the avoidance of which would obviously have been good. They posited various telltale signs of when you're in the midst of that kind of runup - with time to get out and avoid the carnage. They are:

"It is still the case that we cannot call the peak of the bubble, and some of the portfolios we examine keep going up," they wrote. "Nonetheless, investment strategies that condition on high past returns in combination with these characteristics exceed the returns to a passive buy-and-hold all industries strategy by 10 percentage points or more on a two-year or longer basis."

The analysis comes with caveats. One, only 40 times in 90 years did industries rise this fast, a fact that might limit statistical significance. Another is that the inquiry centered on industries that had already reached the 100 per cent threshold - it was clear in hindsight that a major rally had occurred. In real time, while an investor might sense a bubble is happening, the study's findings wouldn't be strictly applicable until it got to 100 per cent.

Notably, it found that not all red-hot markets are the same, and the odds of a bad outcome do seem to be a function of the velocity of the rally. Among the 40 bubble episodes, odds of a crash rose to 80 per cent when share prices surged 150 per cent. That compares with 53 per cent when prices double and a random probability of 14 per cent.

While the researchers focused their study on industries, they found there is "substantial" correlation between crashes in sectors of the market and broader underperformance. For instance, the bubbles in coal and steel stocks blew up during the 2007-2009 bear market.

Another more recent example in investors' minds may be biotech shares. The group more than doubled in two years through July 2015 before plunging as much as 40 per cent over the next year. During that period, the S&P 500 Index suffered two separate 10 per cent selloffs.

Other takeaways from the study: Don't bet against bubbles just because they're bubbles. Even in the 21 episodes where a crash does occur eventually, the peak was six months after the first 100 per cent gain was reached. The average return between the bubble signal and the final peak is 30 per cent.

Factors such as stock turnover and sales growth don't appear to have much predictive power on crashes.

BLOOMBERG