Home > News > 2002 > December 2002 > The equity risk premium
The equity risk premium
Introduction
This note discusses the out performance of equity assets over bond assets (known as the equity risk premium or ERP).
Historically, equities have shown much higher volatility of capital value over the short term compared to bonds, but higher long term total investment returns (as a combination of both capital growth and income). A pension fund will typically invest part of its fund in bonds, to secure assets to meet short term liabilities (such as payments to current pensioners), and part in equities to try and generate high long term returns and maximise the chances of meeting the longer term liabilities (such as payments to younger members who are not expected to retire for some time). With these longer term liabilities, the short term volatility of equities is less of a concern.
Bonds represent the "low risk" asset because they share the most similar financial characteristics to pension scheme liabilities. The capital value of bonds varies as interest rates change (so they are more volatile in absolute terms than cash, for example) but the value of pension scheme liabilities moves in the same way (so the value of bonds relative to the liability value is more stable than cash).
Within the broad definition of bonds, arguably gilts (of appropriate term) represent the best match as they minimise currency and credit risks. Index-linked gilts provide the best match for inflation-linked liabilities, although fixed income gilts provide greater income to meet regular pension outgo. However, it is not possible to match pension scheme liabilities exactly, even with gilts - for example, there can be no precise match for increases in liabilities resulting from salary growth or improved life expectancy.
The key questions for pension scheme trustees, when deciding how much exposure they should have to equities and bonds are: Over what future periods do we expect equities to outperform bonds? and What long-term outperformance can one expect?
Historical analysis
When discussing actual ERP figures, it is important to know which equities and which bonds have been compared. We have focused on Government bonds below, in order to strip out the effect of credit risk (which can be a substantial element of the yield on corporate bond issues, but which does not come without a price, as recent high profile down-gradings of credit ratings have illustrated).
Although history is not always a good guide to the future, it seems a sensible place to start. For example, over the last 50 years (1952 to 2001) the UK stock market outperformed UK fixed- income gilts by an average of over 5% pa.
However, it would perhaps not be prudent to look at only the average equity outperformance and we believe it is more appropriate to consider that performance that has historically been exceeded 9 times out of 10, or else 3 times out of 4 (the 90th and 75th percentiles respectively). This then offers a risk-adjusted measure of the historical outperformance.
For example, over this period we can observe that the UK stock market outperformed fixed- income gilts in more than 9 out of 10 periods of seven years or more, and more than 3 out of 4 periods of three years or more. Another conclusion that emerges from this analysis is that, beyond the short-term (say periods of more than 10 years) the equity risk premium has, broadly, exceeded 2% pa 9 times out of 10, and it has exceeded 3% pa 3 times out of 4.
The graph below summarises the results of this historical analysis:
New research
These figures are consistent with those put forward by many other historical studies. For example, a recent study by BGI concluded that the expected ERP over the next 10 years is 3.75% pa. However, some commentators and economists currently advocate a more conservative view than this, on the grounds that future ERPs may be lower, and less certain, than those in the past.
For example, a recent study by Elroy Dimson, Paul Marsh and Mike Staunton of the London Business School challenges several aspects of the current thinking concerning the ERP. In particular, it highlights a number of weaknesses in the methods previously used to calculate the ERP based on historical data:
- Survivorship bias - equity indices compiled retrospectively tend to be based on only those companies which have survived and prospered - the effect of potentially investing in ultimately unsuccessful companies is not included.
- Choice of market - most studies have concentrated on the US and UK markets, which may not have provided results that were representative of a sufficiently wide range of possible economic scenarios.
- Time period - the majority of research focuses on the second half of the 20th century, when equity performance was high by historical standards.
This new research deals with these issues - first by including all relevant companies in the index construction (including those that subsequently failed), second by using data from 16 countries including France, Germany, Japan, Canada, South Africa and Australia as well as the US and UK and, finally, by basing results on the 102-year period from 1900-2001.
Defining the ERP as the geometric mean of the excess annual return of local equities compared to Government bonds over the whole period, the ERPs obtained in this manner are found to vary from 1.6% pa in Denmark to a maximum of 7.1% pa in France. The UK figure of 4.5% pa fell very close to that for the World index (4.6% pa), and the US fared a little better with 5.6% pa.
However, these results were derived from a historical (or "ex post") perspective, and therefore it may be necessary to make an adjustment to produce realistic estimates of likely future (or "ex ante") ERPs.
To achieve this, the research attempted to remove the effect of certain upwards re-ratings of stock prices which occurred over the last century. For each year, the real dividend growth of a company's shares was compared with investors' expectations for future dividend growth, and the differences compounded to give an estimate of the impact of unanticipated cash flows. This could then be subtracted from the historical ERP. Similarly, an allowance was made for the gain from decreases in the required risk premium.
As a result, the (forward-looking) ex ante ERP for the UK was estimated to be only 2.3% pa, approximately half the (historical) ex post figure of 4.5% pa, while the World index ERP was correspondingly reduced from 4.6% to 2.9% pa.
The research concluded that equities are not guaranteed to outperform bonds over the medium term, which could mean anything up to 20 years. Whilst it is true that in the US and the UK, ERPs are positive for all 20-year periods, negative premia over this length of time can be found in other countries. In fact, for most European countries, including Ireland, the Netherlands, Switzerland, Denmark, Belgium, Spain, Italy and Sweden, at least 10% of 20-year periods show a negative ERP. As a salutary warning of the risks inherent in equity investment, the research suggests that, based on the experience of Dutch stocks, the shortest period of time required to guarantee a positive ERP may be as high as 40 years.
Conclusion
After considering this research, one might re-assess the future period to be reasonably confident of equity outperformance as a longer term of up to 20 years. Furthermore, the probability of future long-term equity outperformance of +2% pa or +3% pa might be estimated to be lower than the historical analysis above suggests.
However, this latest research may prove to be too pessimistic. For example, we would accept that there have been positive "surprises" that have helped equity markets over the last 100 years, but it is pessimistic to assume that there will be no further surprises to come - who can predict future developments in, say, technology or energy efficiency?
In our view one should still be reasonably confident of the performance of equities over periods of 10 years or more.
Barnett Waddingham, December 2002.