The Experts

More on parsing the Australian equity risk premium

As I mentioned this week's AFR column, Professor Richard Heaney and Stephen Kidd are in the process of publishing a detailed academic study on the measurement of the 'equity risk premium puzzle' that arrives at broadly similar conclusions to those that I discussed. In a conversation with one of the authors yesterday he proposed this case-study (and the enclosed chart):

  "I’ve used this simple example to illustrate how the commonly employed equity risk premium method can give counter-intuitive results. Over the period from 1980 to 2008 [ie, right after the GFC], 10 yr Commonwealth Government Securities outperformed the All Ords, in total returns. The All Ords returned 10.4 per cent, whereas bonds returned 11.4 per cent. The correctly measured equity risk premium for the period is therefore about negative 1 per cent (or negative 0.94 per cent). If I measure the risk premium using the ‘common approach’ I get a risk premium of positive 3.7 per cent! But then how can you say the ERP is positive, when I would have more money if I had invested in bonds over this period?"

Heaney and Kidd's findings correlate with my own, albeit that Heaney and Kidd are much more thoughtful in their exposition and analysis of the issue. Here's a preview:  

It is difficult to overestimate the importance of the calculation of the market risk premium given regulator and practitioner reliance on the standard CAPM (Gray and Hall, 2006, 2008). There have been a number of papers appearing in the Australian finance literature since the development of the CAPM in the 1960s that provide estimates of the equity market risk premium (Brailsford et al., 2008, 2012; Officer, 1989). These generally use data spanning long time periods, often beginning in the late 1800s, but there are also studies relying on relatively short time periods (Bishop et al., 2011).

Inevitably, a quoted bond yield is used as the proxy for the risk free rate and this yield to maturity is matched against a holding period return calculated for a reasonably broad share price index like the Australian S+P/ASX200 accumulation (total returns) index. We note the lack of symmetry with this approach. While holding period rates of return are used in estimation of the equity market return, quoted yield to maturity on long lived bonds is used for the risk free rate of return.

This can lead to considerable error in average market risk premium estimates, particularly where interest rates are generally rising or falling over the estimation period. We find that this rather subtle change, from quoted yield to maturity to bond holding period rate of return, halves the average market premium estimate for the period from 4 per cent to 2 per cent per annum.