They sorted shares according to their monthly volatility in the previous two years, and found that the least volatile 10 per cent of shares subsequently outperformed the most volatile between 1990 and 2011. This was true for every one of the 33 major stock markets they looked at. And except for the three-year periods to 1999 (the tech bubble) and to 2006, it was true for every three-year sub-period of those 21 years.
Striking as this finding is, it merely corroborates and extends other evidence. David Blitz and Pim van Vliet of Robeco Asset Management have shown that low-volatility stocks substantially outperform in the US, Europe and Japan. And back in 1972, Michael Jensen, Fisher Black and Myron Scholes, three of the founders of modern financial economics, showed that low-beta shares in the US did better than they should have between 1931 and 1965.
The tendency for defensive stocks to outperform is therefore as robust a finding as we're likely to get in finance; it exists across countries and across time.
This rejects both the first rule of finance - high risk means high expected returns - and common sense ('you gotta speculate to accumulate').
But why is common sense wrong?
One possibility is that investors get carried away by exciting growth stories and so pay too much for glamour stocks while ignoring duller ones, which subsequently rise as investors wise up to their attractions.
However, Messrs Baker and Haugen favour another explanation - that principal-agent problems within institutional investment companies encourage over-ownership of volatile shares and under-ownership of safe ones, causing the latter to rise as they find it easier to attract the marginal buyer.
One of these problems is that equity analysts working for fund managers have an incentive to recommend volatile stocks, partly because these have more newsflow and so attract attention, and partly because they offer the opportunity to show off their stock-picking skills. A buy recommendation on a small speculative stock will arouse more interest than one on, say, National Grid.
Also, fund managers themselves prefer volatile shares, as they offer the chance of big outperformance and large bonuses and ego gratification. Yes, the chance is small - but managers who are overconfident about their stock-picking skills will overestimate that chance. Buying defensives, by contrast, is a 'get rich slow' scheme. And such strategies are not popular with institutional investors.
You might wonder why it matters what the explanation is for defensives' outperformance. Simple. If defensives do well merely because of a cognitive bias, their outperformance might stop sometime, once investors wise up to their mistake. If, however, defensives do well because of fundamental agency problems within institutional investors, then their good performance is more likely to continue.