Equity markets around the world have taken a beating over the last two years and most analysts are focused on trying to work out whether there will be a rally this year. Here we take a longer-term perspective and consider the question of what returns we can expect for equities over the long haul. This is not merely an academic exercise. Small changes to rate of return assumptions can have radical effects on funding requirements for pensions.
Our conclusions are unlikely to bring much cheer. We expect returns in the future to be much lower than they have been in the past. Many pensions are funded on the basis of returns that are, in our judgement, unrealistically high. Given the fall in markets over the last year or so, this suggests that many funds risk being unable to meet their commitments. There are, broadly speaking, two ways to estimate long-term equity returns: in terms of what equities can return given reasonable economic assumptions or, alternatively, what they must return to make investors hold them. The latter approach involves the ‘equity risk premium’ (ERP). This has been described as the most important concept in finance – it is also one of the most misused. The ERP is the extra return required from equities over a safe asset, like cash, to compensate for their higher volatility.
Over long periods of time, it is reasonable to suppose that the actual average return will be a good indicator of the ERP, otherwise we are assuming that investors are systematically surprised. This is where the problem lies. Over the last 50 years, equities have outperformed cash by roughly 7% pa in both the UK and US. In the last 20 years, the margin has been even higher. Looking at other countries produces a range of results but the most authoritative study by Elroy Dimson, Paul Marsh and Mike Staunton in a London Business School/ABN-AMRO report which looked at 16 countries, found an average in excess of 6%. Having looked at so many countries over such a long time period you might think that these results form a sound basis for determining the ERP. Not so. Numbers like 6 or 7% are implausibly high. Remember that the ERP reflects compensation for risk.
This leaves us with the problem of trying to decide why the observed data shows such a high premium. The first explanation is that the data are wrong. They reflect survivor and success bias. The survivorship bias is a familiar point: if you look at companies that exist today and construct an index of their shares prices over the last 50 or 100 years, all those companies which existed at the beginning but subsequently went bust will have been ignored. Estimated returns from such an index will therefore be biased upwards. Recent work which recalculates traditional indices to remove survivorship bias has reduced ERP estimates by around one percentage point a year. Survivor bias may also occur in other more subtle ways. The indices that we look at are from countries that have been successful.
A similar explanation is based on the idea that the past 50 years have been especially good. Investors may have feared a disastrous event – war, revolution, hyperinflation, etc, and demanded a premium to compensate for this possibility. One estimate suggests a meagre 0.5% probability of a 50% loss could explain the upward bias in estimates of the ERP. This explanation receives additional support from the work done by Dimpson, Marsh and Staunton that extends the data on historical returns to cover the first 50 years of the twentieth century. Examples of economic depression, war, revolution and hyperinflation abound in this period and, sure enough, the historical ERP is lower. Relative to cash, the median ERP based on sixteen countries was 2% for 1900-49, a fraction of the figure of over 6% for the following 50 years.
Even if we accept that the future is likely to be closer to the 50 ‘good’ years in the second half of the last century than the preceding 50 ‘bad’ years, we still have to accept a lower ERP. Investors have presumably factored the good news into their calculations and boosted stockmarket valuations as a result. Only if the next 50 years showed a similar improvement in economic fortunes would equities maintain their exceptional returns.
We conclude from all this historical analysis that the ERP is below 6%, probably significantly so, and perhaps closer to 2%. This is a wide range but we can narrow it considerably by applying the second approach. At the most basic level, equity returns in the long term must be based on the cash paid out to investors. Cash can be returned to investors via dividends or share buy-backs. If we divide this free cash flow by share prices it becomes a yield. This yield, plus the growth rate of free cash flows equals the long-term return to equities. The free cash flow yield in major equity markets is around 2% to 3% a year.
And how are those free cash flows likely to grow over time? Probably in line with GDP. The corporate profit share of GDP has been close to trendless over the long run, and we have no reason to expect that to change. Similarly, there is little reason to suspect that the proportion of those earnings that firms will be able to make available to shareholders will change either. Some of that free cash flow growth, however, will likely reflect the emergence of new firms and new shares of outstanding firms (just as some of GDP growth owes to population growth).
A better proxy for the concept that we are after may be trend productivity. If productivity grows at, say, 2% per year in the long run, then it is not unreasonable to expect free cash flows of existing shares to grow at a similar real rate: 2% is a reasonable assumption for mature industrial countries in the long run. In Europe, where the population is roughly stable, this would equate to trend GDP growth. The real return to equities is therefore 4-5%.
To get a forecast for the actual nominal return we must add on inflation. For the UK, we assume that the Bank of England Committee continues to hit their inflation target, and that it stays at 2.5%. Most other countries do not have an explicit inflation target. In Euroland, the goal is for inflation to be ‘at most 2%’ which we take to be 1.5%. In the US they use 2.5%, as in the UK. Our assumption for Japan, currently suffering from deflation, is that inflation averages 1% in the long run.
We end up with a nominal long-term annual return for UK equities of 6.5% to 7.5%. Overseas markets would deliver slightly different returns depending on their respective inflation rates. But if we assume that exchange rates move over the long run to offset differences in relative inflation rates, overseas markets would all generate the same return of around 6.5% to 7.5% a year measured in sterling terms.
The equity risk premium is the difference between this and the expected return to safe assets like bonds or cash. Given our assumptions about inflation and productivity, projecting these is relatively simple. There is considerable theoretical support for the idea that real short-term interest rates should be equal to the rate of productivity growth, which we have already assumed to be 2% pa. Adding in inflation gives 4.5% for nominal interest rates. Bonds are a little more volatile so they should return slightly more, say 5%. As a side check we note that this is close to the actual yield on UK government bonds. Actual yields have been remarkably good predictors of subsequent nominal returns on government bonds over reasonably long time frames. If interest rates rise, bond prices fall but this is offset somewhat by higher reinvestment rates. Of course, this does not maintain real returns when inflation rises.
These estimates suggest that 3% is a reasonable estimate for the ERP relative to cash. Even this is based on the upper end of our range for equity returns and the ERP relative to bonds is 0.5% lower. Note also that these estimates make no allowance for the impact of tax or management fees, which would tend to lower the figures still further.
Many pension schemes and investment products around the world are based on the assumption that returns will exceed these estimates. In the 1980s and 1990s, market returns typically exceeded even the most optimistic estimates. More recently the reverse has been true, particularly for those funds that invest heavily in equities. The implication of our analysis is that many pension schemes are unsustainable and if they are to continue they face an unpleasant choice between reducing benefits and/or increasing contributions.
Steven Bell is global chief economist at Deutsche Asset Management in London
This article is based on Joshua Feinman’s ‘What can we expect for asset returns in the long term?’ Deutsche Asset Management (2001)