It is generally agreed that, to compensate for their greater short-term volatility and risk of loss, shares should provide a return differential over a risk-free asset like government bonds over the long term, says Shane Oliver, Sydney-based head of investment strategy and chief economist at AMP Capital Investors.

He notes that this return differential is referred to as the equity risk premium (ERP) and is generally thought to be around 5% to 7% annually as this is what it has been for much of the post-war period. But thanks to the recent bear market in shares, he says, US shares have underperformed US bonds by 7% annually over the last decade.

This leads Oliver to ask the questions: Does this mean shares are a dud investment? Is the equity risk premium concept is meaningless?

The answer to both, Oliver says, is no.

In his latest investment strategy report, he explains why:

As the equity risk premium concept relates to the long-term, the bear market of the last two years proves nothing about its merits. There is much confusion surrounding the equity risk premium and much of this flows from the fact that it can refer to three different things: the historically realised return gap between equities and bonds, the gap required by investors to attract them to invest in equities, and the prospective or likely long-term gap based on current valuations. The key issue is not what equities have done relative to bonds in the past but what their potential is going forward. Experience tells us that these two concepts can move in opposite directions.

Many analysts tend to focus on historical data as a guide to what sort of return premium investors require for shares over bonds and to what it will be in future. However, while a useful starting point, this approach has limitations.  

First, the historically realised return differential between equities and bonds varies significantly over time.

Second, to the extent that valuation changes (rising bond yields and rising price to earnings multiples) boosted the realised equity risk premium over the post war period, this would have not been expected by investors and hence the measured ERP over that period is not a good guide as to what they would have required to invest in shares.

Third, in any case it is difficult to justify why investors would have demanded such a large premium (5% to7% annually, for example, over the 50 years from 1950). This would imply an implausibly high degree of risk aversion.

Fourth, historical equity data for countries like the US suffers from a survival bias.

Thus, Oliver notes that the historical record does not provide a definitive guide as to the risk premium that shares should or will offer over bonds. Just as the recent experience of a negative excess return understates the return from shares over bonds, longer term perceptions of a 5% to 7% annual return excess exaggerate it, he says, adding that a range of factors suggests the required ERP is somewhere around 3% to 4% annually. More so, current estimates suggest the prospective ERP is now above this level, particularly for Asian and Australian shares, he says.

"The prospective return differential offered by stocks over bonds today is far more attractive than it was at the height of the last bull market in 2007," Oliver says. "While stocks are vulnerable to a correction after their sharp gains from March, their attractive risk premium compared to bonds suggests that any short-term pullback in shares will simply be a correction in a still rising trend."