Equity risk: the inescapable impacts of business conditions
When living through a period of unusual economic instability it is useful to try to unlearn most of what we have learnt as a standard framework for savings and investment, and focus instead on the essential power, unpredictability and unfairness of business. In any form of market economy, business, and only business is the source of both national and personal wealth creation and it is also the source of the sometimes massive changes in economic fortunes between generations, even when the overall trend is for increasing affluence. These tides overwhelm the investment outcomes arising from diversification across asset types and strategies, all of which ultimately are derivatives of equity risk. In this Insight we demonstrate the ubiquity of equity risk and ask whether it is possible to quantify it, communicate it to lay investors, and manage it.
Equity risk as a mirror of business Market economies, like religions, are adaptive systems responding to the shifting needs and attitudes of society from one period to another. What they retain is their essential dependence on businesses, great and small, as the driver of the performance of the economy through time. Governments can make transfer payments between individual citizens and can help soften the collective blow of changes in the tides of economic performance. But these will make relatively little difference to the impact of these shifting tides in business conditions on individuals’ financial welfare.
With the benefit of a long historical perspective it is possible to see that market economies have in fact adapted very successfully, hence the fundamental nature of equity risk is much more stable and predictable than we might intuitively expect. Nonetheless, it is clear that the shifting fortunes of business through time mean that different generations may well experience substantially different economic outcomes. Investment returns are an important manifestation of those outcome differences. It may seem unfair and undesirable but it is part of the essential nature of equity risk.
For the purposes of this Insight, it is much more sensible to think of equity risk as the measure of business uncertainty, across these shifting tides, than it is to focus on some much narrower statistical definition of risk such as volatility, or dispersion of short-term returns. Because these tides are themselves best observed in real terms, undistorted by inflation, it is uncertainty of real outcomes that is the measure of equity risk we want to encourage. Contrary to the view the financial services industry seeks mischievously to inculcate in its customers, equity risk does not reduce with time and this is an important piece of conventional thinking to unlearn.
In practice, it is difficult, if not impossible, for individuals to live their lives free from equity risk even when they actively seek to avoid its most obvious form, when holding equity investments. It is ubiquitous and unavoidable, through employment, accrued pension rights, asset holdings, liabilities and contractual investments.
Since equity is the source of wealth creation, we will later ask why, knowing the cost of trying to avoid equity risk, people should even want to, instead of leading their lives in some deliberate and sensible accommodation with risk.
Equity risk overwhelms diversification When investors hold financial assets directly, under advice or via a discretionary manager, they are likely to follow a fundamental portfolio principle of diversification: spreading risk. Our high-level view of equity risk poses no problem for one aspect of diversification, which is the avoidance of unrewarded specific risks, but it does pose a problem for a second aspect, which is spreading the exposures across a number of different types of asset.
Specific risk refers to the idiosyncratic risks of individual assets, such as shares in a particular company, and contrasts with undiversifiable ‘systematic’ risk of a market or asset class as a whole. Academic evidence suggests that the system risks, which tie in closely with our high-level notion of tides in business conditions, are rewarded by ‘risk premiums’ whereas undiversified exposure to specific risk is not rewarded. The investment industry can adopt this distinction very easily by using diversified vehicles that reflect the market system, such as actively-managed or index-tracking collective funds, or by designing portfolios of individual securities within a market that are well diversified.
Entrepreneurs will protest that individual business risks are rewarded and that diversifying away all the exposures to specific risks is the wrong principle. There is no need for dispute here as they are essentially two very different ways of investing. Problems tend to arise when investors do not clearly select between the two approaches or when they fail to identify their true preferences, or belief systems, and invest out of character, such as imagining they are entrepreneurial when in fact they are system players. It is also perfectly possible that the same investor will apply both approaches, perhaps for different purposes and with different sources of capital.
The distinction between assets classes The principle of diversification of capital between a range of different types of assets or strategies is to reduce the statistical measure of risk, volatility or dispersion of short-term returns, on the basis that it should be possible to do so without reducing expected returns. In a world abhorring free lunches, diversification between asset classes can look like the only free lunch available.
The numerical characteristic on which this depends is correlation: the way different asset types move together. If those movements are not perfectly correlated, there is scope to smooth the return path of the combined portfolio.
Our high-level view of equity risk overwhelms the differences between these asset classes, in terms of the underlying risks that characterise them, and reveals them to be essentially different versions of, or derivatives of, equity risk.
To illustrate this, we take as a starting point a diversified portfolio benchmark of the Association of Private Client Investment Managers (APCIMS). To reflect the increasing use of structured products and private equity at the top end of the private wealth management market we have arbitrarily reassigned 5% to each from the equity share of the benchmark.
The three conventional slices of the pie, equity, property and fixed income, are distinguished here from the additions to diversification, hedge funds, private equity and structured products, whose slices have been lifted out.
If we apply the thinking of underlying business risk to each of these portfolio building blocks, we can summarise as follows:
Commercial property is a subset of corporate balance sheets and cash flows, since the rental income streams are entirely dependent on the capacity of businesses to pay and honour their covenants in all other respects.
Fixed income has three risk components, nominal interest rates, inflation and credit quality, each of which is a reflection of the high-level business context. If within a structure of diversification we were to avoid unrewarded bets, consistent with first principles, then we would not want to take on inflation risk but might retain credit risk. It is self-evident that the market’s changing perceptions of the chance of default are a first and foremost a judgement on equity capital surviving bankruptcy.
After the chastening of the credit crunch, high-brow opinions about private equity are shifting towards the more prosaic academic view that returns are a function of public equity conditions with the addition of leverage. Both buying and selling prices are set largely by public equity-market conditions. Proponents may also argue that the private phase allows transformation in the underlying performance of the business that would not be possible in the public market but this flies in the face of the history of change in public companies.
Structured products are not necessarily equity-based but we think the vast majority are. The structure only alters equity risk to the extent of introducing asymmetry in the payoffs, as in any option strategy. On a theoretical basis, rolling over equity options is self-defeating, like all forms of portfolio insurance, as the mean cost is a function of volatility which itself is related to the mean risk premium for bearing equity risk.
Hedge funds encompass a wide range of strategies but equity risk still dominates, particularly when private clients obtain their exposure through funds of funds.
On this analysis, the 45% that is diversified away from pure equity volatility is not avoiding equity risk as we have defined it. Virtually all the portfolio is at the mercy of the shifting tides of business conditions and this is what long-period portfolio outcomes, in real terms, will tend to reflect.
What do we know about equity risk?
As soon as we define this in terms of real returns from markets as systems we know we can use historical return data for published, representative indices for different equity markets to understand and quantify equity risk. A corollary of viewing the market as a system is that we should measure ‘total’ returns, income and capital (so that income distributions are treated as being reinvested), since the consumption of income is an individual choice, often arbitrary, not a system feature.
We show below a series for the UK equity market from 1900 that is based on the annual Barclays Gilt Equity Book, which provides back-filled data prior to the formation firstly of the 30-share industrial index and later the launch of the All Share Index. This is similar to the UK data assembled by Dimson, Marsh and Staunton for their ground-breaking book, The Triumph of the Optimists, which examined a century of returns from equity investing around the world. The UK market data benefits from genuine continuity, whereas many European markets and Japan suffered bouts of hyperinflation that distort the history of real returns. Post-war reconstruction can also effectively mean that the returns were enjoyed by different investors before and after that hiatus. The UK and US equity markets are the two major markets that are unusual in avoiding all such distortions. In all cases, money returns are converted to real returns using a recognised, contemporaneous measure of retail prices in each country.
UK Long Term Real Return Trend
The index itself is calculated by compounding short-period, deflated total returns. It is presented in log scale to ensure that the changing rates of change are always proportional (contrasted with non-log series for financial assets that tend to show exponential growth towards the top right-hand corner of a chart).
The visible features are i) a trend and ii) cycles around the trend. Cycles are different from random deviations from a trend because successive observations are themselves serially-correlated. The trend is a measure of systematic ‘reward’ for bearing equity risk and the cyclical deviations are a measure of either or both of the underlying business cycle and a tendency for market sentiment to follow a herding principle.
Two trends have been fitted to the data: one (grey) for the observations that could have been made without hindsight up to any point of time and the other (yellow) a single trend fitted to the entire series. Differences between them speak of the broad shifts in business conditions we like to focus on. The slope of the whole-period trend for the UK is 6.2% per annum. This represents the ‘typical’ systematic reward for bearing UK equity risk.
However, because the deviations from the trend are clearly very large (after cumulating in a cyclical fashion perhaps over many years), the typical rate is never going to serve as a good forecast of the future rate over some shorter period than another century. These deviations are a key component of equity risk, pointing to a wide degree of uncertainty about future real levels. Uncertainty about where we will end up is not the only measure of equity risk since we may also be unsure about both the validity of the observed trend and the starting position in relation to the trend. These sources of uncertainty are directly related to the amount of data we have to work with, although we will never know how much data is required to be truly representative.
Mean reversion The feature implied when a trend appears valid over a long time series is ‘mean reversion’. Since a purely random series may also show an upwards drift, it is not enough to assume mean reversion is actually at work. But there are statistical tests (summarised in No Monkey Business: what Investors need to know and why) that confirm the likelihood the process is genuinely mean reverting.
From a theoretical point of view, this is entirely consistent with other widely-accepted evidence of mean reversion in economic components of equity returns that are based on some equilibrium theory. Examples are the competition between returns to labour and capital, which tend to make swings in the balance of advantage self-correcting, marginal returns on capital in business and rates of dividend distribution compared with retention for investment. If these fundamental business drivers are mean-reverting it follows that deflated measures of corporate financial performance, such as earnings per share, will be mean-reverting and that the market valuation process based on such observations, such as price-earnings ratios, will then in turn be mean-reverting.
The size and duration of deviations from trend evident in this UK series nonetheless imply that the mean reversion process is quite weak in explaining short-period returns. The profile of these deviations is also itself highly variable, because of the unpredictability of the large shifts in business conditions that prompted this Insight.
In the UK’s history we can single out 1973/4 as a truly exceptional bear market, coming from a peak ratio of trend of 126% in 1972 to an eventual ratio at the bottom of the bear market of 36%. Yet within 12 months of the low, the ratio had recovered to 67% and was back on trend by 1984, less than a decade later.
This V-shaped recovery contrasts dramatically with recent experience in Japan. The same chart for Japan, using data from 1957 (which we treat as undistorted by the post-war recapitalisation of the Japanese equity base), shows a bear market commencing in 1989 at a peak ratio of 187% which reached a ratio of 67% in three years. For the next 10 years, mean reversion failed to occur and the ratio eventually drifted down to 39% (very close to the UK’s bear market low point).
Japan Long Term real Return Trend
Although Japan is widely viewed as enduring an exceptional period of economic adjustment, not shared by other countries, this is simply not the case. This sort of massive adjustment process, impacting dramatically on achievable equity returns, has also occurred in Europe and the USA. Reflecting the poor recent performance, the observed trend for Japanese real returns has been drifting off and at 5.3% pa is substantially below other markets.
Europe ex UK Real Return Trend
Japan’s ‘lost decade’ was mirrored in the period of ‘Eurosclerosis’ that marked the 1980s. Partly an inability to adjust to high inflation, Continental Europe also struggled to adjust its post-war industrial model to much slower domestic growth and increased global competition. During its lost decade, both real profits and dividends declined dramatically before the pendulum of mean reversion in underlying economic performance started to swing back, and with it stock market returns. Its whole-history trend is 6.9% pa so it has clearly made up for this blip.
The US history of real returns (based on the S&P 500 Index) also provides examples of long-duration deviations from trend. Because of the data length (and quality) we can be more confident the observed trend of 6.8% pa is representative.
US Real Return Trend
Contrary to popular opinion, the Depression era was not marked by the failure of equity returns to mean revert. Though unemployment ravaged the nation, aggravated by the dust bowl that was impervious to economic initiatives, industrial output was pushed up within three years of the 1929 Crash and real stock prices followed. It is impossible to separate the initial effects of the New Deal from the impact of world war and many observers believe it was the latter, not the former, that eventually underpinned America’s economic revival.
On the other hand, America’s adjustment to the testing economic conditions of the late 1970s led to a much slower recovery in real share prices, albeit from a less extreme ratio in 1974, than the UK. It took 20 years for the market to get back on trend.
Positive deviations that persisted for a long time are just as important in determining actual returns and in shaping the views of contemporary investors about equity risk. Many current UK investors are likely to have been influenced by the persistence of above-trend real returns from about 1990 until this recent bear market began. Likewise, the overvaluation of US stocks that peaked in the late 1960s (particularly growth stocks that were in the front-line of globalisation) may well have reflected complacency about the persistent above-trend returns from the mid-1950s.
Quantifying equity risk
These data histories provide us with a measure of the realistic scale and duration of the shifting tides that lead to substantial differences in achievable returns from equities and other asset classes. They also provide us with a basis for making mean expected returns a function of starting conditions, provided we believe the process really is mean reverting.
Given the similarities between the return processes (and even the observed trends) in these four market groups, it is sensible to assume that the equity return process, as s system, operates very similarly in different countries, overwhelming cultural differences and even persistent differences in economic power and performance. Mean reversion makes for a flatter world than intuition (or a little economic knowledge) might imply.
It is therefore sensible to assume that:
Anything that has occurred in one place can occur in another
Anything that has occurred in the past can occur on the future
Things will occur on a scale and with a frequency in your own lifetime that ensure that your own outcomes will be subject to the uncertainty of large-scale changes in business conditions
Any complete range of possible outcomes must allow for multi-asset class diversification to be overwhelmed by the dominance of this form of equity risk.
Using a long-term asset model (‘Lambda’) We believe that we can quantify the full range of outcome uncertainty at any future horizon based on our interpretation of current conditions. Even when we believe the data is representative, the uncertainty is still going to be greater than many professional investors and advisers glibly suggest to their clients. However, in markets with less data or less reliable data, notably Europe (ex UK) and Japan, we will expand the range of uncertain future outcomes even further to allow for this possible source of error.
A further adjustment we need to make in quantifying uncertain future outcomes is for currency risk when projecting returns in foreign markets. In a real-return modelling framework, foreign markets are as capable of generating real returns in one currency as another, on the basis that differences in their inflation rates will tend over time to be offset by nominal exchange rate changes (another equilibrium theory known as Purchasing Power Parity). The risk associated with currency is therefore the movements in the exchange rate not explained by actual observed inflation differences. This is easy to measure. Combining the two sources of risk, equity and currency, needs to reflect the fact that they appear to be independent of each other.
Equity risk and time What emerges very clearly from our quantification of equity risk is that it is completely at odds with the popular industry canard that equity risk declines with time, or that uncertainty about returns reduces the longer the forecasting (or assumed holding) period. The only measure of equity risk that declines is the annualised standard deviation or measure of dispersion. Mathematically, this is a function of the square root of time for a random series, and some greater rate of reduction in the case of a mean-reverting time series. As soon as you compound the annualised rates over the actual forecasting period, it becomes clear that equity risk continues to expand with time. This phenomenon was expressed by a leading actuary many years ago as the ‘expanding funnel of doubt’ and it is a phrase we find salutary to repeat in our conversations with clients.
How can it be that the investment industry can so blatantly misrepresent such an important attribute of equity risk in its marketing literature and in its communications with clients? We are at a loss to know whether the explanation is ignorance or cynical mischief-making. The cynical view assumes that the industry has a bias to understating risk so as to encourage investors to buy high-margin products, because the margins typically follow the underlying investment risk. Indeed, there may even be a twisted logic to the tactic since it becomes irrational to invest in products with high charges unless they have some scope for high returns, however small the chance of clearing the cost hurdle and earning those returns. Unfortunately, this is not a conscious rationale investors are invited to engage in.
We show below an example of the expanding funnel of doubt for a foreign equity market with a fairly long data history, an observed trend of 6.6% pa and a starting ratio of 50% – so reasonably close to current conditions in several major markets. Costs are assumed to be minimised by using trackers rather than active funds. The range of possible outcomes is with 99% confidence, which is theoretically as confident as any modeller can be and allows for the very small chance that the entire model assumption about systematic equity returns is either wrong or for some reason breaks down globally.
At 15 years, the probability of earning a zero real rate of return is 1%. The model assumptions about the strength of mean reversion are such that most of the return difference from starting conditions should have been reflected in 15 years.
What better ways to manage equity risk? If diversification between different classes is not a good way to control the outcome uncertainty that can dramatically impact individuals’ financial position, is there a better way?
Clearly there is but though consistent with modern investment theory (and some institutional practice) it is inconsistent with the dominant retail investment business format.
The solution is dilution of equity risk, by holding offsetting positions in risk free assets. For this purpose, it is vital to define the individual’s purpose for the money as this will determine whether the risk free asset is one that needs to include inflation protection. For any goal whose target outcomes are defined in terms of purchasing power (such as retirement spending), the only risk free asset is index linked gilts, since these carry a guarantee of full (and largely untaxed) compensation for future inflation whatever it turns out to be. For very short horizons, it may be practical to bear some inflation risk, however, and treat cash as risk free.
The array of possible alternatives to equity risk looks very different when defined in terms of each of nominal return volatility (which we also call ‘path risk’) and long-horizon real outcome uncertainty, which we have identified as more important for personal wealth.
The y axis plots the uncertainty of real outcomes and so has cash as relatively risky, whereas on the x axis cash has no path volatility and is essentially risk free. The outcome risk arises because the nominal interest rate paid mostly consists of the market’s assumed required compensation for future inflation. If markets were both good at anticipating future inflation and also able (subject to monetary policy goals) to secure the required compensation, there would be little error in real outcomes from holding cash. In practice, both inflation forecasts and monetary policy influences have given rise to large differences in inflation compensation from cash, ranging from very generous to negative real rates. These errors can compound over long periods, hence introducing an expanding funnel of doubt for cash too.
It is conventional fixed interest whose location, and portfolio role, most changes when expressed in terms of cumulative outcome uncertainty. This is another version of the inflation guess but with the important difference that with fixed income there are far fewer opportunities to adjust the guess. At least with cash, the market can make new forecasts in response to new evidence about inflation, possibly as frequently as daily. With fixed income, you have one guess at the point of purchase and then only small opportunities to reset the guess as a function of the yields at which interest is reinvested. If interest is being spent, the accuracy of the forecast will not even benefit from these other opportunities.
The location of real outcome risk reflects the UK’s particularly bad experience of inflation forecasting and its impact on fixed income, with deeply negative real yields in the 1970s followed by over-compensation in the 1980s. Because the inflation process is so hard to understand and model, it is pointless taking on this sort of bet.
It is also entirely unnecessary since index linked gilts, perfectly matched to your personal time horizon (or series of horizons), get the job done perfectly. Though there may be quite high path volatility prior to redemption, particularly for longer-dated index linked gilts, the investor is in practice indifferent to that movement since the cash flows are perfectly matched.
In the schematic below we illustrate the principle of using index linked gilts to narrow the funnel of doubt by diluting the equity exposure. It is clear that dilution also lowers the slope of the funnel.
The cost of avoiding risk The expected real return given up when equity risk is diluted needs to be quantified if individuals are to determine their personal risk preferences. In a quantifiable modelling framework like Lambda, each of the key parameters, i) target outcomes ii) time horizons iii) resources required and iv) level of risk taking, can be expressed numerically and illustrated graphically. Running different iterations can quickly demonstrate the trade-offs required, such as between the additional resources required if taking more risk, assuming an unchanged level of confidence in the outcome.
In the example below, we show a retirement spending goal in which the client’s stream of cash flows for spending are organised by three-year time slices. Each is separately funded to produce outcomes within the agreed bands, represented by the candlesticks. In this example, the resources required with optimised risk taking, to meet a schedule of gradually tapering spending targets, are £2m. The approach to risk is dynamic so that the mix between risky and risk free will change both as market conditions change and as the plan gets shorter. These dynamics are reflected in the amount of time-slice risk, as if for a plan rather than for a portfolio.
On a worst-case basis, the sustainable rate of draw from the portfolio will be £125,000 pa tapering to £57,000 pa. But if the mean expected returns are achieved, the initial rate of real spending will barely need to taper, ending at £105,000 pa in real terms. There is a 50% chance spending can be higher than this, or surplus assigned to other goals.
To avoid all equity risk and fund these cash streams entirely using index linked gilts would require assets of £2.2m to match the worst-case outcome and as much as £3.2m to match the mean expected outcome with risk taking.
Clearly, there is a high cost for not staying at the table in this casino. But the client can also at some stage leave the casino itself, by purchasing an index linked annuity. However, applying all of the available £2m to an annuity now will generate a sustainable real income of just £64,000, nearly half the initial rate of draw and barely above the worst-case tapered draw late in the plan. At a much later stage, on the other hand, maximising spending may be better achieved with an annuity than continued risk taking, particularly if a client assigns a low value to leaving capital at death as a bequest.
Knowing these costs of avoiding even pure, direct equity risk, it is not at all clear that most investors would want to.
It is because this process is entirely customised that the existing industry formats would struggle to adopt it. It is important that the process be highly quantitative, and delivered with slick technology, to offset the cost implications of mass customisation. But it also requires other changes in business approach, not least the fixing of flat rates of fee for a portfolio service, to ensure that the manager is completely indifferent to the risk-taking preferences exhibited by the client.