Can holding debt make more sense than holding assets?
For debt, read ‘bonds’, because that is all they are. For assets read ‘equities’, because that is what you own when you hold ‘a piece of the equity’. These two claims on other people are the basic building blocks of asset allocation for long-term savers. Clear distinctions about how each behaves are essential for using them sensibly.
There is a lot of nonsense thinking about both types of claim. We shall no doubt see plenty more of it in 2007, first as the debate starts about asset allocation for the new National Pensions Savings Scheme (NPSS) and second as a consequence of reported deficits of occupational pension schemes. Sensible asset allocation matters a lot for individual investors at every level of income and wealth. If you hold any bond investments and you’re not sure how to answer the question I posed, read on.
NPSS: what the Government is proposing The Government’s recent white paper on the new low-cost, soft-compulsion pension scheme has argued that the proposed 6% contribution rate will only achieve its target outcome (50% of final salary from both the basic state e pension and this top-up) if the underlying assets include a high proportion of equity exposure. It suggests the numbers work out at about a 60% equity exposure to 40% bond exposure.
Both equities and bonds have uncertain real future values, once inflation is taken into account. Contributing at a rate that eliminates both market risk and inflation risk is a luxury only the wealthy can afford. The difference between being 50% and 100% certain of achieving a target real outcome is usually between being affordable and not being affordable. It’s that simple.
The NPSS makes no false claims about adequacy of the contribution rate to offer full protection against both risks. Full protection is implicit in the basic state pension, not the semi-voluntary top-up scheme that the NPSS is meant to be. There is clearly room for political debate about whether the present Government’s idea of an adequate level of state pension is realistic but that is a separate debate from the assets to hold in the top-up scheme.
No place for bonds Early criticism of any equity exposure in the NPSS has emerged from the usual sources. My view is that it is the bond exposure that is wrong. The best advice for NPSS savers is likely to be 100% equities until about 50, gradually replaced by index linked gilts. Non-indexed bonds have no place in a funding strategy that is focused on the purchasing power of distant income streams.
As private wealth managers, we see our job as being to know about all threats to capital. To do this we must be a historian of all financial markets, anywhere. What history teaches us is that few threats come bigger than government. Private capital must use the institutional structures of the time and must play be the rules that governments have made. But capital itself is agnostic and non-aligned. In an anarchical framework, equity is the strongest claim and the one that ultimately all other monetary holdings depend on. Even government debt and governments’ future ability to raise taxes depend entirely on the existence of wealth-creating businesses (whether state-owned or private).
In this view of risks, the quality of the claim is more important than the volatility of the asset. Diversification is a sound principle for reducing path volatility but not at the cost of a reduction in the quality of the claim. It is better to reduce risk by diluting the exposure to equities with risk free exposure. When measuring risk in terms of purchasing power, only inflation-indexed securities can provide robust dilution rather than weak diversification.
Conventional wisdom At face sight, the white paper’s proposed 60:40 balanced mix of equities and bonds is not exactly challenging. Almost every packaged investment that has been offered as a self-sufficient investment, from managed funds to with-profits policies, has ‘balanced’ its exposures between bonds and equities. In different developed economies the last half of the century saw the more or less rapid displacement of bond investments by equity exposure. Indeed, the 60:40 equity:bond split has become the default rule of thumb in the USA. Where other assets, such as property – are held, these are often substituted for equities, to retain the underlying logic that asset-backed investments deal with the inflation risk and bonds deal with the nominal capital risk.
The UK moved earlier and went further in embracing equity and property. This is because owning debt with no inflation protection wiped out about 90% of the purchasing power of this nominal claim on other people’s money in the persistent UK inflation of the 1960s through the 1980s. Living off an income which is ‘fixed’ in money terms, rather than in real, inflation-adjusted terms, is a high-risk strategy.
UK occupational pension funds were amongst the leaders of the equity cult. Because the pension promises they make are to accrue pension rights based on final salary (inflating faster than prices) and by law must include compensating pensions (once in payment) for at least part of the erosion of real incomes by price inflation, scheme trustees were more comfortable backing the promise with asset-backed equities than nominal debt instruments. Bonds were allowed to drop to about 20% of the asset allocation. This was roughly equivalent to matching all long-dated promises with equities, as the debt holdings were effectively assigned to meeting the early years’ promises.
Far from being captured by theories, this behaviour is informed by experience and intuition. It is consistent with the trustee, like the capital, being agnostic and non-aligned.
Plus ça change Three things have occurred in the UK, and most other developed economies, in the last 10-15 years to challenge the cult of the equity. However, none of them changes the fundamental nature of either bond or equity investing.
The first change is that inflation has declined and, with it, future expectations about the risk of inflation eroding real incomes or wealth. We have got complacent.
The second is that governments responded to investors’ mistrust of nominal debt by issuing inflation-linked bonds. For the first time in the history of paper money, investors had a form of debt holding that was risk-free in terms of real as well as nominal capital value. The risk associated with the purchasing power of the income stream was also significantly reduced (and with ‘stripped’ index linked gilts can be eliminated). Inflation-linked bonds also provide a new market-derived ‘real interest rate’.
The third is that the accounting rules governing how the assets and liabilities of a pension scheme are calculated and reported in the income statements and balance sheets of the sponsoring companies have been changed. There is not quite one international standard for dealing with the impact but the effect has been the same in many countries, with the UK amongst the most affected. I have explained in earlier articles that the effect is to move the clock back towards matching promises with bonds not equities, on the basis that the promise is itself bond-like. The basis of valuing projected future pension payments is to calculate a present value using a discount rate taken from the bond market – high-quality corporate bonds, to be precise. This latter change is consistent with fair-value accounting principles only because we do not account in real terms. An accounting diktat based on imperfect accounting conventions cannot turn a risky asset in real terms into a risk free asset in real terms.
If both companies and individuals planned and managed their assets and liabilities counting only in real terms, they would reach the same conclusions that served us well for the last 50 years, not the ones that let us down in the preceding 50 years.
Bonds and inflation One test of this is to model the real-return behaviour of bond markets. This requires some assumptions about the inflation process itself.
The inflation process is notoriously tricky to model. It can look easy: near-term future inflation is highly correlated with recent past inflation. Over longer period, however, there may be large cumulative shifts in the rate of change in prices. This is often dealt with in models (including the one the Treasury uses) by having randomly-arising changes in the inflation regime between low and high. From our perspective, defined as long-term, purchasing-power outcomes, the possibility of large changes in inflation that, once they occur, tend to persist, creates a huge uncertainty about the ultimate change in the price level over, say, a period of 20 or 40 years.
Diversification with bonds If we want to move beyond modelling future real returns, to create an optimal portfolio in which both types of risky asset are combined, or ‘balanced’, we need to decide on some values for the mean real return, the standard deviation of real return and the covariance (or ‘co-movement’). Sorry about the maths, but these are at least basic to the investment business. Because of the way we are defined the problem, we are deciding on values valid for long-period outcomes, not just short-period paths.
This makes all the difference in the world. If we allow for the true risk inherent in the inflation process, we will need to assume a standard deviation that is not far different from equities, as it has to allow for the possibility of negative real returns as well as strongly positive real returns. The mean value is likely to be much lower than equities – even the accountants recognise that there is an implicit risk premium between equities and bonds derived from the difference in their ‘path’ or short-term volatility. A realistic mean real bond return is a number between about 2% and 3% pa. Historically, equity markets around the world have achieved mean real returns (derived from regression analysis) of about 5% to 7% pa. We can argue about the exact values, but these orders of magnitude are now widely accepted by academics and investment consultants.
If bonds have the same real-return risk as equities and a much lower expected return, what can they possibly contribute to the typical investor measures of usefulness? Only one: diversification benefits. These materialise as a function of a low assumed covariance. Unless they behave very differently (as in one high when the other low), why would you accept the lower return that adding bonds brings, when risks are equivalent?
Over both short and long periods, the covariance of real bond and real equity returns is highly unstable. There have been periods in which they have moved similarly and also periods of great divergence. These are probably explained by underlying economic fundamentals, particularly the inflation process. So the circle is complete: if investors cannot predict changes in the price level, they cannot predict the real-return covariance between bonds and equities.
If, to allow for this uncertainty about the covariance, you optimise the combination of bonds and equities assuming a high covariance, the answer will come back with no bonds. To squeeze even a 40% bond allocation with these means and standard deviations, you will need a very low covariance assumption. This is an out and out gamble and a pointless one.
Dilution versus diversification If you want to narrow the range of outcomes reliably, replacing equities with index linked gilts (or even short-term cash deposits) is a much more predictable form of risk control than adding another risky asset.
The optimal degree of dilution is a function of the risk aversion of the investor. But whatever the difference in risk aversion between any two investors, there will also be a common sensitivity to the time horizon.
It requires a very low risk aversion to want to hold equities right up to the moment cash is required
It requires a very high risk aversion to want to hold few equities when a long way from the moment cash is required.
This is important information for NPSS savers. They already have a part of their pension income effectively matched by a risk free asset: the basic state pension. Even if offered the alternative of an inflation-indexed holding, young investors ought not to hold it, as the worse-case expected real returns from equities are likely to exceed the fixed risk free rate from an index linked gilt.
Have we correctly defined long-period real equity risk? This last statement is pretty critical. If at any time horizon the range of real equity outcomes is expanding faster than the risk free rate, there is no point at which all of the range of possible equity real returns is above the risk free rate. If there were, it would be apparently irrational to accept any dilution of return, whatever your risk aversion.
Analysing these assumptions, it becomes rapidly clear that the cross-over point depends on your assumptions about ‘mean reversion’ of real equity returns. The idea that over long periods real equity returns will show some mean trend that is fundamentally justified, even if there are large and long-lasting deviations from the trend, is not undisputed. It is implicit in the analysis of long data series for major equity markets from which the 5% to 7% range is taken. However, data in the same form of time series as real equity return indices can give an appearance of a meaningful trend that actually requires more thorough statistical tests.
It is remarkable that, even with such tests having established this is more than a mathematical feature, there is still disagreement about the mean reversion assumption. It would help if there was a robust theory to explain it. Considering the importance of this assumption to rational asset allocation choices, it is surprising how little work has been done on the relationship between real equity returns and ‘average’ investor risk tolerances and ‘utility functions’, as these would help explain both the mean values and the limits on the deviations from them.
The Ralfe argument It is helpful to understand that critics of holding equities to match long-term liabilities may not accept mean reversion. I dealt with one in a previous article. John Ralfe was the Boots finance director who famously switched their pension fund equities into bonds before the last bear market. He is in the ‘fair value’ accounting school but not, it seems, the inflation accounting school (although about two years later somebody spotted the error and most of the bonds were exchanged for index linked gilts).
John argues about mean reversion not from first principles but rather from the rather skimpy evidence of option prices. Options are mainly traded on short terms where it is sensible to assume that price changes are random. If there were substantial actual trades on long terms, option buyers and sellers would also need to confront the issue of mean reversion. Until there is and they do, the fact that some notional long-term contracts assume equity risk expands with time at the same rate as a random series does not make it the right interpretation of equity risk.
Path risk and outcome risk I hesitate to blame John and other fair value accountants for the damage they have done to pension funds, as it may be rational for pension schemes to optimise their asset allocation on the new accounting rules, which affect the company during the tenure of the decision-makers. However, I am wholly critical of his recent writings, as an economist at RBC Capital Markets, which aim to undermine confidence in equities as the bedrock of the NPSS.
What is missing from the debate to date is a clear distinction between path and outcome risk. As noted in this article, you will enjoy a very different range of outcomes if you optimise on short-term nominal volatility rather than long-period real outcomes. The silly thing is that in a soft-compulsion scheme with a modest rate of contributions, path risk should have little or no influence on their attitude to risk or their willingness not to opt out.
Proof that education about the mathematics of saving is woefully lacking is that even wealthy investors fail to realise that what they need during the accumulation period is low markets, not high markets. High markets at this stage (as opposed to when they are close to cashing in) merely serve to reduce the expected return on the contributions they then make.
Mind you, that is another valuable insight that depends on accepting that real equity returns are mean reverting rather than random. Maybe that is the real first focus for education. You’ll find plenty to support it on this website.
For non-indexed bonds to avoid picking up this uncertainty in their own real return outcomes for long holding periods requires the market to anticipate these changes fairly accurately. What the models tell us is that the modellers think this foresight is impossible. We agree. You may feel confident that the inflation of the last 50 years will not be repeated, as markets have now been predicting for about ten years, but it just another bet you are willing to make.