Bonds: why the debt bet is a bad bet
No Monkey Business, is highly unusual, though not alone, in rejecting conventional bonds. When we take on new clients, it is the asset allocation choice we most often need to challenge, is easiest to change and makes the biggest difference to risk management. Where does this insight come from and why is it so powerful?
Conventional bonds or fixed income investments are one of the key building blocks of the investment portfolios of private clients, trusts and institutions.
Allocations to bonds are typically the main means by which managers and product packagers customise a diversified or ‘balanced’ portfolio so it matches some definition of client risk tolerance
Bond holdings are used by personal pension managers in the crude ‘lifestyle’ feature of gradual risk reduction in the years leading up to retirement
Bond yields are used in modern accounting as the discount rate for measuring the present value of long-term liabilities such as final-salary pension promises, encouraging their use as a ‘matching’ asset
Bonds are widely used to maximise income where this is a priority for investors.
Each of these conventional uses of bonds is intellectually flawed.
Whether intuitive or well-informed, recognition of this is actually implied by many private investors’ affinity with mortgage debt (they love repaying nominal debt with depreciated currency) and aversion to fixed-income annuities (they hate receiving a pension in depreciated currency). However, when it comes to investment, in the face of such entrenched portfolio practice, they need logical argument to resist the industry’s recommended use of bonds.
The counter arguments rest to some extent on the same theories we embrace when seeking to integrate investment management and goal planning, such as maximising some goal-specific ‘utility’ which is defined in terms of purchasing power. Stripped of technical language, this means a portfolio is organised to deliver specific outcomes the client defines and which are expressed in real or inflation-adjusted terms. Preferences and trade offs are made at the planning stage by selecting between alternative sets of outcomes.
Being based on established theories, the arguments have been discussed and reviewed in a number of academic papers. But they also stand up to simple logical tests and to the lessons of historical evidence. They are not complex and, though near enough complete, what follows is not a long explanation.
Bonds as a bet We have a general preference for distinguishing investments as bets. The industry avoids this language either because they think it creates the wrong impression of them or because they do not want investors to be totally clear. So what is the bond bet?
Bonds pay a fixed income, fixed in money terms, for a fixed period and then return your original capital, also in money terms. What you risk is that the borrower cannot pay the interest or repay the capital. The default probability is evidenced and manageable. A bigger risk, which cannot be diversified, is that the nominal cash flows fail to provide a satisfactory return once inflation is taken into account.
The public market in bonds plays an important collective role in determining the best guess as to what future inflation will be, so that nominal yields always take into account inflation expectations. Borrowers will try to set the interest rate and governments are the borrowers who can best manipulate rates (and boy they try). But bond investors do not have to lend, or lend long, so they ultimately set long-dated rates.
How good or bad the bond investment crowd’s inflation guesses prove to be is what determines the outcome of any long holding-period investment in bonds. The record is poor. Cumulative real returns have turned out to be quite strongly positive or negative as often as they have ‘normal’. For investors concerned about long-term purchasing power, bonds come with very large ‘outcome uncertainty’ or risk. Why they might want to take this gamble when they can easily avoid it is a question every investor should ask themselves.
Bonds in a portfolio of bets Historically, bonds were not added to equities to diversify risks. Rather, as theories about the true risk of diversified portfolios of company shares versus the risk of individual share speculation took hold in the first part of the last century, equities were added to bonds.
Early portfolio theory did not need to think in real terms to advance combinations of bonds and equities as generating superior risk-adjusted returns. In the best-read investment book of all time, The Intelligent Investor, Benjamin Graham explained as early as 1950 the principle of combining two assets that moved over most short periods in opposite directions (negative correlation). It proved not to be the case, other than in brief intervals, but as long as correlations were much lower than 1 there would be a benefit to normally risk-averse investors from holding both.
This assumes that the right measure of risk to weigh against the cost of slightly lower return is the volatility of the return path over short periods. In other words, we will accept lower wealth outcomes if we can take away some of the anxiety provoked by the bumpy ride. As a description of typical investor utility, or preferences for making trade offs, this works fine as long as the bumpiness of the ride is also the same as, or similar to, the uncertainty about the outcome. In a world where everyone focuses on nominal returns from investment, not adjusted for inflation, the two are near enough the same. But in a world where many investors are not fooled by money illusion and focus on the purchasing power of future wealth levels, real outcomes adjusted for inflation matter as much as, or more than, the bumpiness of the path and, because of the nature of the inflation process, the two degrees of risk are not at all the same.
In the chart below, we have approximated the position of the main financial assets used in conventional portfolios on two axes:
outcome uncertainty (the range of possible real outcomes over horizons of 15 or more years) and
path volatility (satisfied by monthly standard deviation of returns)
Reflecting the fuzzy and time-varying nature of the observations, they are shown as areas rather than points. Bonds are shown here as having outcome risk that is not a lot less than equities. The historical evidence of bond real returns is drawn from the UK rather than some other countries whose bonds were devastated by hyperinflation (otherwise they would be even riskier than equities).
In real outcome terms, the most certain asset is an index linked gilt (ILG) matched to the term of the desired outcome (which we elaborate on below when discussing a better approach).
The fact that an ILG secures all inflation risk yet is still quite volatile in the short term (ie is plotted a long way to the right) illustrates perfectly the principle that investors need to be clear about priorities, either for all their money or for particular goals. There is no perfect asset that has a smooth return path and delivers certain real outcomes at distant horizons, so a choice has to be made as to which benefit they most value.
How inflation affects risk How does inflation risk drive a wedge between path volatility and wealth outcome uncertainty? It is because of the nature of the inflation process.
The price level with fiat money (paper currencies issued by governments) is not subject to self-equilibrating forces. There is no ‘natural’ level. The cumulative change in the price level over long periods is therefore a function of changes in the rate of change in the price level, or changes in inflation. This, though not well understood, is generally thought to be best characterised by ‘regime change’. This means it is well-behaved for long periods (next year’s inflation rate will be close to last year’s) but then encounters shocks as a result of which it first becomes unstable and then moves to a new level of rate of change at which it becomes stable again. Subsequent shocks may restore it to earlier or even different stable levels.
These shocks are not predictable. This explanation accounts for the breakdown in the wisdom of crowds as a basis for reasonably accurate forecasts of future inflation.
Investors need not have experienced the shock of living on fixed incomes in the 1970s to accept this humbling inadequacy. Markets then failed for nearly a decade to build in the moderation of inflation into their estimates of future inflation. As if that were not sufficient proof, consider how badly anticipated has been the inflation acceleration around the world in the last few years.
There is no shortage of less-than humbled forecasters willing to bet (if only with other people’s money) on recession leading to slower inflation or even deflation. This ignores the scope for governments, with the active endorsement of voters, to print money to ward off debt deflation.
Correlation uncertainty and the portfolio bet Portfolio theory needs expected returns and standard deviation, equivalent to path volatility, but also needs an estimate of correlations between them. As noted, correlations change dramatically over time. This process is also poorly understood and not that widely researched in academia. However, the historical evidence points to the possibility that it is unexpected inflation, or estimation errors, that may mostly account for the big changes in correlations. It is highly inconvenient for portfolio theory if the errors in estimating the key variables are not independent.
These findings support our case for building portfolios on the basis of real wealth outcomes rather than short-period returns and risks.
We do not need to exclude bonds as a prior. Allowing them to compete in a portfolio optimisation process, whereby an efficient trade-off is sought between risk and return, bonds will not win an allocation if realistic holding-period real returns and risk (largely inflation risk) are used as inputs.
Bond income: the distribution fallacy Bond investment also involves a transfer of potential consumption from the future to the present whenever the income component which represents the market’s guess of future inflation is spent or distributed. It cannot be both spent and added to the nominal capital to support future spending.
In this case, the full amount of the inflation rate, rather than the error in estimating it, is what reduces future wealth outcomes.
The inflation guess is usually the largest component of the yield (historically, say 2-10% versus a ‘real yield’ component of 1-3%).
This ‘consumption time shift’ may be a deliberate preference, it may be unrecognised or even a fudge between the two. A level annuity to meet retirement spending, for instance, is probably seen by most people as a declining income in real terms. In this case, the rate of decline is left to the mercy of the gods rather than managed.
When recognised, it may be the investor’s preference but it may also be the agent’s preference, as it will take a long time for the customer to discover that the solution is the wrong one. The industry has been cute about its use of bonds to maximise current income as well as to minimise the use of cash (on which it is harder to take fees) as a means of diluting risky exposures.
The consumption shift is also a source of potential conflict within trusts. As well as leading to unmanaged risk of decline in real incomes from the trust, holding trust capital in bonds can favour a life tenant at the expense of other beneficiaries’ purchasing power in a way the settlor had not intended. In the nature of trustee liabilities, this should never be a fudge, let alone unrecognised.
The tax inefficiency of bonds Finally, governments collude in your ignorance of these simple truths about bonds by taxing the inflation compensation component in exactly the same way as if they were real. Real yields need to be higher to earn any economic rent from capital after tax. In other words, the bet on the market’s inflation guess has to be right even to stand still in real terms.
Governments have nonetheless made index-linked investments available that are taxed less heavily. Because the capital protection comes as a tax-free uplift to the sum at redemption, only the bond interest suffers tax and this is running at under 2% versus nearly 5%. A difference of 1.2% for a 40% tax payer is a hefty price to pay for the inflation bet. HMRC may be indifferent to which form of inflation protection we choose but we should not be.
Better ways to manage risk Risk control, when purchasing power matters, is better performed differently. There are far better assets with uncertain real payoffs than bonds and these can be customised to personal risk tolerance by diluting them with holdings of assets that have near certain real outcomes.
Using dilution instead of relying on diversification between more risky assets creates a far wider spectrum of possible outcomes with which to satisfy different investor preferences and reduces the number of assumption errors which might taint the projected outcomes.
Index linked gilts, matched to the maturity (or duration) of the goal are the most certain. In a portfolio construction process that maximises some investor utility, the appropriate index linked gilt yield provides the risk free rate and risk aversion describes the ‘indifference’ between that rate and some uncertain (but higher) rate.
In our approach, risk aversion is not derived from questions about investment but by preferences between model outputs in the form of the range of possible outcomes. Outcomes are usually expressed in the terms relevant to the investor and understood by them, such as a sustainable rate of spending in real pounds (eg if drawing down to meet lifetime spending). In this approach, the cost of avoiding risk, both capital market and inflation risk, is explicit, as a reduction in the range of potential outcomes as well as a narrowing of the range.
Rolling over cash (or very short-dated bonds) is less certain but much less risky than long-dated bonds. With cash or short bonds, you get to reinvest the capital coming back to you at the market’s new estimate of inflation. With long bonds, only the interest gets reinvested with the benefit of new inflation estimates, assuming the interest is not being spent. Hence the bet on the estimate is much greater. You might be on the wrong side of a regime change.