Why bonds are not good diversifiers: a primer
In a market correction, such as we are experiencing currently, bonds can appear a safe haven. But over longer periods it is bonds that have least to contribute to efficient, diversified portfolios. I touched on this in December. It struck a number of professionals as well as non-professionals as little short of heresy so I decided a ‘chapter’s worth’ on bonds would be a useful ‘primer’.
The historical portfolio role of bonds was to dilute outcome risk, by introducing some certainty of return. It was not to smooth return paths. In the presence of inflation, they do neither well. In this basic explanation we use two No Monkey Business ‘distinctions’: dilution and diversification as ways of managing risk; and path risk versus outcome risk. A resulting insight is that fixed-income bonds without inflation protection have no place in most investors’ portfolios.
The diversification principle The overwhelming argument for diversification is that a diversified portfolio has lower volatility, or a smoother return path, than an undiversified portfolio, without necessarily giving up any expected return.
The theory starts with portfolios of securities in the same market, which is why it is plausible to assume no reduction in mean expected return. Not to diversify a portfolio of securities is to give up the closest thing to a ‘free lunch’ – in a word, dumb.
It is not necessary for the stocks to have very low correlations for this ‘free lunch’ to be enjoyed. When extended to asset allocation between different types of asset (such as different equity markets, hedge funds, property and commodities), the theory relies on low correlations, or different short-period price movements, because the ‘normal’ expected systematic returns are not the same for each type of asset held.
Rather than being a free lunch, a trade-off has to be made between reducing risk and giving up some expected return. The nature of a trade off is that you pay a price for something. There is also a price at which you are indifferent to not having it versus having it. In decisions where a trade off has to be made, the diversification argument is much more complex. It also depends critically on the investor’s personal risk tolerance, or the price (in lost return) at which he or she is indifferent between a certain and uncertain outcome – between taking or rejecting a bet.
The popular asset classes The typical building blocks of a portfolio diversified across asset types can be seen from insurance-company ‘managed funds’ as well as private-client portfolio ‘benchmarks’ adopted by the industry:
equities (in turn diversified between companies in the UK and foreign share markets)
property (direct investment, not property shares which are just more equity)
bonds (‘fixed income’ securities issued by governments or companies)
Few individuals, other than the very wealthy, have diversified further by adding illiquid, unregulated funds and strategies in private equity and commodities. Private client holdings of hedge funds tend to be via funds of funds and add little diversification benefit.
Bonds represent a much more important part of most diversified portfolios than property. Holdings of corporate bonds have also increasingly displaced gilts. Both observations are generally true whether individuals make their own investments, have portfolios managed by stockbrokers, use an IFA to allocate money between the asset classes or invest, by whatever means, in diversified managed funds. With-profits investments, which used to hold the same mix as managed funds, are no longer well diversified, most being forced by reserving rules to hold predominantly bonds.
Risks and returns of the popular asset classes The expected returns over long holding periods for equities and property are broadly of the same order. Both are ‘real assets’. The value of the asset in the market place on any day is given by the rights to a stream of future income. Whether profits or rents, this income will be denominated in ‘current’ pounds and will reflect general retail prices, as well as general growth, in real terms, of the economy.
The expected return of a bond is not of the same order. It is lower. Theoretically, its lower expected return is a function of the fact that the payment of its due interest and the repayment of principal are relatively secure – obviously more so for gilts than for corporate bonds. The income stream is more predictable. But also more of the present value of the stream is accounted for by early years than late years, whereas the market value of most company shares reflects income due to be earned in the distant future. There is much more uncertainty about the ‘correct’ present value.
This idea of the ‘duration’ of an investment has a significant impact on its ‘volatility’: the amount by which its market price fluctuates. Volatility is an important source of investment risk. In No Monkey Business we define this as ‘path risk’.
Individual investments with lower volatility should be expected to earn less than ones with higher volatility, for which a higher expected return is the incentive that motivates people to take more risk. The technical term for this incentive is a ‘risk premium’. The ‘equity risk premium’ describes the difference in expected or actual returns between gilts and equities.
It follows from this explanation that bonds have return attributes that make them different from equities and property when using them to build a portfolio. They add diversification, to the extent they do not have identical paths, but they also reduce expected returns.
There is a standard mathematical process for combining assets with both different risks and different returns, called ‘optimisation’. It solves for the combination with the highest return (however defined) per unit of risk, or risk-adjusted return. There is an array of portfolio combinations that theoretically have the highest risk-adjusted return, known as the ‘efficient frontier’. Though they may all have equivalent risk-adjusted returns, and each is optimal, they have different absolute levels of risk. Selecting the right combination for an individual, or an individual goal, requires a knowledge of their own risk tolerance.
The asset allocation in each efficient portfolio is decided on the basis of the risk and return of each individual building block and the correlations, or co-movement, between them.
Correlations This key measure of how assets fluctuate jointly is a powerful determinant of the efficient portfolio. High assumed correlations will make the resulting weights (or exposure to each asset) look like they are a function of expected return differences, because there is little reduction in risk to trade off as a benefit when holding those assets with lower returns. Low assumed correlations will therefore tend to increase diversification, making the portfolio look more equal-weighted and less obviously explained by the expected returns.
There are two fundamental problems with building diversified portfolios this way:
Correlations between asset classes themselves fluctuate greatly over time
The levels of correlation vary as between short, medium and longer holding periods
The FSA recently referred to these problems in a paper anticipating possible financial risks in 2007. It listed as one of its concerns investors’ and their agents’ over-reliance on diversification across asset classes to keep downside risk within their true tolerance ranges.
The FSA’s warning reflects both of the problems above. It is prompted partly by the fact that correlations can be observed to have converged in recent years. It is also willing to speculate that the cause of this, strong growth in global ‘liquidity’, will lead to highly-correlated falls in asset prices when liquidity turns down. Being cyclical, it surely will – and it may already have started.
The second problem is related to the first, via what we understand to be the FSA’s insight. As such, it is more important for bond holdings than equities and property. Correlations between bonds and equities are much higher at about five years than at shorter periods or longer periods. The long-period correlations are largely a function of inflation, which we will come on to. The higher intermediate-period correlations are to do with liquidity cycles and business cycles. Interest rates mainly drive bond yields and bond prices and interest rates are highly sensitive to the balance between supply and demand for funds or liquidity. Central banks and governments influence the supply of liquidity by setting interest rates but in doing so they tend to add to the cyclicality of economic activity.
We can also anticipate that the risk premium for credit risk, as between corporate bonds and gilts, will also vary in a cyclical rather than counter-cyclical manner. In a liquidity squeeze, defaults rise and recovery rates on distressed securities fall. The yield premium required to make investors indifferent to accepting these risks, instead of holding a lower-yielding but safer gilt, is currently at unprecedented low levels in the UK and many other countries. Because yields have been generally low, investors have in this respect implicitly increased their risk tolerance, probably without realising it. The fluctuations in this credit risk premium add to the intermediate and cyclical correlation with equities.
‘Money illusion’ and the outcome risk of bonds For all investors with long horizons and objectives for their money that are measured in terms of what the money will buy in the future, the relevant measure of uncertainty of outcome is ‘real terms’, encompassing both market and inflation uncertainty.
The difference we have already noted between equities and property as real assets and bonds as ‘monetary contracts’ is one that makes its impact felt over long holding periods. This is because long periods expose the investor to the greatest uncertainty about cumulative changes in the future price level, as represented by some index of retail prices.
The inflation process is poorly understood by economists and therefore poorly predictable by investors and their agents. The recent period of stability in the rate of inflation does not diminish this uncertainty. However, it has clearly had that effect for most investors. There is now virtually no risk premium for inflation uncertainty.
If we could be sure that markets could anticipate future changes in the inflation rate reasonably well, we would not need to worry too much about outcome risk, in real terms, when holding bonds. The capital repayment element of the current market value for a long-duration bond would be (as noted earlier) a relatively small part of the current market value and prices would alter to reflect new expectations of future inflation. The reinvestment of income would therefore be at yields that preserved purchasing power. (Even this would not be of any use if interest was being spent rather then reinvested.)
The assumption that markets can anticipate future inflation is not borne out by history. Based on historical evidence, the uncertainty about real returns when holding nominal bonds for periods as long as 20 years is virtually indistinguishable from that holding equities. Over periods no longer than 10 years, the actual real gilt yield looking forward to known subsequent inflation has varied over the past 80 years between 10% pa (in the early 1930s and again in 1990) to -5% at the start of the 1980s. If we assume that a stable real yield, using hindsight, in a tight band of say 0% to 2% denotes sustained accurate inflation forecasting, there is only one spell of 30 years out of 70 that mostly meets this condition (1943 to 1973).
It remains to be seen whether the last ten years of unusual ‘stability’ will also meet the condition. The gyrations in nominal yields between about 7% and 4% suggest it will not be that accurate. The impression given has been that there has been increasing willingness to trust the monetary authorities’ ability to stabilise and restrain the force of inflation and therefore of investor’s ability to forecast changes in the cumulative price level. Individuals and their agents may have taken this optimistic view of inflation predictability on board. More likely, both are myopic about the true forecasting errors and so not alert to the dangers of money illusion.
If the expected real bond return is much lower than equities (by the amount of the expected risk premium) and the risk (for a long holding period, in real terms) is as high as or not much less than equities, it is not rational to hold bonds.
There are no bonds in a real-return optimisation Using the same mathematical optimisation process referred to earlier, limited to bonds and equities, but this time inputting long-term real returns and outcome risks, we find that bonds do not win an allocation. It requires unrealistically low correlation assumptions in a real-return optimisation to overcome both lower expected returns and similar outcome risk to the competing real assets.
If using a formal optimisation approach, bonds do not need to be excluded a priori, but it will make little difference if you do. If using some other rule of thumb, or intuitive guesswork, to build a diversified portfolio, knowing this result from formal models means you might as well treat bonds as an asset you can safely ignore. It will certainly simplify your decisions if you do.
Dilution as a better means of risk control We come to the second key distinction. Diversification can only be a very approximate way of controlling risk because of the uncertain, time-varying and time-dependent nature of the correlations between the building blocks. This particularly penalises bonds. But historically the role of bonds was not diversification but risk dilution.
This was because a gilt (with no credit risk) was treated as a risk-free asset. Adding a risk free asset to a set of risky assets reduces risk in a much more predictable way than adding some lower-risk but still risky assets.
As we have seen, if matching a gilt duration to the investor’s own time horizon it is only inflation risk that makes the bond a risky asset. In the past, when prices fluctuated due to harvests and wars, but with no overall trend, it was sensible to treat a government bond as risk free. Nowadays, understanding how retail prices can trend cumulatively and so introduce high uncertainty about long-holding period real returns, only index linked gilts have this risk free characteristic. This is because the government guarantees to uplift both the interest payments and the capital sum repayable at maturity by the general price inflation experienced over its life. There is therefore no real outcome risk (though path risk for longer-dated index linked gilts remains).
The risky asset portfolio, itself diversified, has a band of possible outcomes at future horizons. This band of uncertainty grows somewhat larger with time even if the assets themselves are ‘mean reverting’. The mid-point of this expanding ‘funnel of doubt’ is the slope or angle of the mean expected return. Adding index linked gilts will have the entirely predictable effect of narrowing the funnel and also reducing its slope. Adding conventional bonds would also reduce the slope but will not appreciably narrow the funnel at most horizons (intermediate: because of higher correlation; longer: because of inflation uncertainty).
In a portfolio in which risk is managed largely by diluting risky asset exposure, the degree of dilution can be arrived at by applying the investor’s own degree of risk tolerance. A more risk-averse investor will hold less of the optimal risky portfolio and more of the risk free asset. In terms of the ‘indifference’ approach referred to, risk tolerance determines how much exposure to a risk premium an investor should accept on the basis of indifference between the risk free rate of return and the alternative rate of return from risky assets.
Illustrating the dilution principle In the chart we show how we might sensibly approximate the location of equities, bonds, cash and ILGs in space defined by each of path risk (monthly volatility) and outcome risk (for about ten years). The areas occupied by each asset allow for the time-varying nature of actual risk observations – hence they are depicted as fuzzy or approximate rather than precise and accurate values.
The green area assumes the existence of some dynamic strategy of managing the mix of equities (as risky) and ILGs (as risk free), as a function of a personal risk tolerance and responsive to changing asset prices. This is what we regard as an optimal approach, superior to a static mix. If the nature of the portfolio goal involves counting down to a known requirement for the cash, such as the drawdown of capital to meet spending in retirement, the locations alter as the horizon shortens. With about five years to go, cash, ILGs and short-dated conventional bonds (but not long bonds) will converge in the space at bottom left where path volatility is a reasonable approximation of real outcome uncertainty. At this distance from the consumption of the capital, no equities are likely to be held.
Lifestyle options and the annuity fallacy The principle described above, when counting down rather than rolling forward, is the same principle as in the ‘lifestyle option’ that some personal pension products include. This builds in a crude, linear switch of equity exposure to bonds as the expected retirement date approaches. The intention is to minimise the impact on the eventual pension of the joint risk residing in the future value of the equities and the unknown gilt yield at the point of retirement. Because the plan assumes the purchase of a level annuity without inflation protection, the switch is to bonds of a matching duration, not cash. The principle is sound but it assumes the investor will optimally select a level annuity.
It follows from our observations about inflation risk, particularly where bond interest is not being reinvested, that we regard a level annuity without inflation protection as a very inefficient way to manage inflation risk in retirement. With increasing longevity, an annuity has to last about 30 years. Inflation will erode the purchasing power of the income but at an uncertain rate. If we could be sure of the rate, and also that it roughly matched some intentional profile of declining spending in real terms, a level annuity would be a reasonably sound way to plan for retirement spending. But in practice it is both arbitrary and uncertain. If a level annuity is not the right pension solution, a lifestyle option which switches to gilts will not reduce risk in a useful way.
Conclusions UK private investors have been seduced by both lower and more stable inflation to treat fixed-income bonds as diversification building blocks that do not reduce portfolio returns dramatically but do reduce portfolio risk. They – or their advisers and managers – have also been willing to substitute corporate bonds for gilts. These offer higher yields to reflect the risk of default. This risk premium has been whittled down to very low levels.
The FSA is right to warn that diversification benefits are probably being relied on excessively to keep client portfolios inside the actual tolerance ranges of their owners. They are right because different equity markets, and property relative to equities, are likely to have much higher correlations than required to make a big difference to the range of probable portfolio movements or outcomes.
The only reliable way to manage risk so as to ensure portfolio paths and outcomes are consistent with true tolerances is by narrowing the ranges by reducing exposure to risky assets as a whole. This is done simply and cheaply by substituting risky assets by an appropriate risk free asset.
As usual, consumer ignorance of this approach goes hand in hand with an industry conflict of interest. The industry loses by advising dilution rather than diversification. Its revenue stream is typically a percentage of asset values. Dilution will either reduce the fee rate it can charge or commission it can earn on the lower-risk assets or, if it loses control of the diluting assets (such as if cash is used), it loses any fee or commission on that amount. The diversification supposedly provided of a managed fund or a lifestyle option has the merit for an agent of sustaining the asset base and the fee stream for as long as possible.