Are stocks the best bet for the long haul? In a paper to be published in the Journal of Finance, two researchers question the long-held investing assertion.

“There have been periods in which stocks underperformed [Treasury] bonds and bills over 30 years, [and] 40-year periods in which stocks barely [outperformed] bonds,” Lubos Pastor of the University of Chicago Booth School of Business, one of the study’s authors, tells MarketWatch’s Howard Gold.

Gold says Pastor and fellow study author Robert F. Stambaugh of Wharton make their case through statistics. “They say that the moderate volatility of stocks over long periods of time is a result of classic reversion to the mean. So, the long-term standard deviation of stock returns (a measure of variability) is 17% per year,” he writes. “But they claim that when they use different statistical techniques that allow them to factor in uncertainty about average future returns, the resulting ‘forward looking’ measures of volatility produce a much higher annual standard deviation of 21%. That makes stocks nearly 25% more risky each year over any 30-year period.”

Pastor also argues that the 200 years of stock market data that Jeremy Siegel uses to make his case in Stocks for the Long Run is insufficient. That’s because, while it’s two centuries worth of data, it covers a period in which the U.S. moved from a frontier market to an emerging market to a superpower. Now, we don’t know what will come next.

“The past 200 years have been very kind to the U.S., [but] there’s some probability you might lose a war. There’s some probability you might have a financial meltdown. There’s no guarantee we’re going to bounce back” from them, Pastor says.

But Pastor isn’t exactly predicting disaster for the stock market — far from it. He says that he “wouldn’t be surprised if the 7% historical average real return were, say, 4% expected return plus 3% unexpected return, where the 3% represents good luck (a positive surprise).” His best guess for future after-inflation returns are that they will be about 2 or 3 percentage points lower than they’ve been historically, which would still put them around 4% to 5%.

To read the full study from Pastor and Stambaugh, click here.