In any discussion about the risks of investing, the starting point must be the liabilities an investor has to meet. If your investments drop in value, how will this impact on your ability to meet the cost of your
Reflecting this fundamental liability-related risk, risk in institutional investing is traditionally divided into two categories:
• Strategic investment risk. The risk of investments which are in line with a strategic-investment benchmark not meeting the liability requirements.
• Tactical investment risk. The risks which are taken relative to the strategic-investment benchmark by a fund manager investing in a way which
differs from the benchmark.
The first risk is very much the domain of the investor. The second risk is one which a fund manager can understand and manage. The first is generally longer-term while the second can have a shorter-term focus. More time tends to be spent examining tactical risk but, as strategic risk is a much greater part of the overall investment risk, let’s redress the balance and look at strategic risk in more detail.
Assessing the strategic investment risk typically involves looking at both assets and liabilities under a range of possible future economic scenarios. The aim is to construct an investment-performance benchmark designed to reflect the investor’s preferences for risk and return. Inevitably, this involves using what is a new vocabulary for the investor: suitability, diversification, matching and benchmarks.
The pension plan has liabilities, namely pensions to be paid at some point in the future. This liability can be valued in a variety of ways by the actuary, but the meaning of risk for those responsible for the plan depends on the relative values of the assets and the liabilities. By choosing a suitable investment strategy the risks can be managed.
But there are limits to what an investment strategy can achieve. The value of liabilities can change for a number of reasons, some of which depend solely on the intrinsic qualities of the liabilities (for example, the number of people in the plan or the definition of the benefits to which they are entitled). You can’t use investments to manage these particular ‘liability risks’. The plan sponsor and trustees must use other means to control these risks.
Other risks are more interesting from the asset-liability modelling perspective. Higher-than-expected inflation means that pensions and salaries generally increase faster than predicted. Lower-than-expected yields on investments also mean that more money is needed to meet the same payments in the future. Both of these risks can be eliminated by choosing assets with values that will also adjust in line with inflation or yield changes. The value of matching assets and the value of liabilities always move in tandem, so that the risk of a relative change in values disappears.
In the real world, the matching assets may not exist, or there may be other reasons why the client may not need to adopt such a cautious position. Diversification means that assets which don’t match liabilities can still be used in the strategy, as long as they are combined with other assets in a way which dampens the ups and downs relative to the plan’s liabilities. If the scheme is in a healthy position they can afford to take risks and earn rewards along the way. Asset-liability modelling or scenario-testing will demonstrate the level of risk of different investment strategies.
After examining the various options, the benchmark is born. The benchmark is the way that the investor’s preferences for risk are translated into a structure for fund managers and corresponding performance objectives. Setting the benchmark involves choosing which asset classes are preferred for investment and how they should be combined within the portfolio. It may be expressed as an explicit combination of the asset classes (for example, 50% Euro-zone equities; 50% euro bonds) or implicitly by referring to some average strategy adopted by a peer group.
Don’t be deceived if this process sounds straightforward. The investor’s preference for risk is not easy to pin down, especially where the investors are a diverse group with different perspectives. The process of arriving at the investment benchmark is where the investment consultant’s understanding of the liabilities and legislative requirements can add enormous value.
Deciding on a comfortable level of strategic risk is a tricky business. And, as is often the case with difficult
decisions, it is often avoided or put to one side in favour of easier and less important decisions. Picking a fund manager and then monitoring the risks they take is much easier and probably more enjoyable than setting the overall benchmark. However, investors, attracted by the potential rewards, are tackling this decision – the size of pension-fund assets is increasingly important in the context of the sponsoring employer’s finances and so pension costs are coming under the spotlight. Pension provision is now less about designing a scheme to attract and retain the right employees and more about money. And investment risks.
Once the strategic risks have been assessed, the next step is to have one or more fund managers manage assets relative to the strategic benchmark. To the extent that they are required to produce extra performance, they will need to take extra risks – and we are now in a situation where the benchmark is all-important and the liabilities are left to one side.
The focus when measuring tactical risk relative to the benchmark is not on sophisticated mathematical
techniques. The aim is to translate risk theory into a concept that everyone can understand. Whilst there are some very sophisticated sets of investors, most UK pension fund trustees, for example, spend an average of only eight hours a year on investment-related issues.
One risk measure which has become an industry standard is tracking error. Tracking error has a number of pseudonyms: active risk, commitment factor and relative risk. Tracking error attempts to show us the uncertainty of returns around a portfolio’s benchmark. A greater tracking error implies that portfolio returns may fall further away from benchmark returns.
There are two forms of tracking error. Historic tracking error is simply the standard deviation of the relative returns. It is backward-looking and a matter of historical fact. It does not reflect the actual risks taken. For example, a portfolio of 72% Marks & Spencer and 28% Vodafone would have returned 13.8% in 1998. This is exactly in line with the index, therefore the tracking error is zero, but clearly this was not a no-risk portfolio. While other examples would be less extreme, it is never clear whether a manager has experienced their tracking error by design or chance.
Prospective tracking error is the expected standard deviation of the difference in return between the portfolio and an equal investment in the benchmark, as calculated by a quantitative statistical computer model, often the ‘Barra’ model. Note the word ‘expected’ denoting the prediction of future tracking error.
Prospective tracking error can be a useful tool. Tracking error can be used to compare trustees’ expectations of different managers. An analogy used to help trustees thinking about tracking error and its relationship to returns is a car’s engine capacity. A 1.6 litre engine is less powerful than a that of a three-litre car, and similarly a portfolio with a tracking error of 1.6 is less punchy than one with a tracking error of three. Investors sometimes need reminding that they are driving a Mini rather than a
Risk levels for a manager’s portfolios within one house may be wildly inconsistent or could change rapidly. This could prompt investors to reassess their choice of manager on the grounds that they are no longer driving the same car, so to speak.
Based on an assumption that relative returns are normally distributed, the usual interpretation of prospective tracking error is that two out of three annual returns for a portfolio, around its benchmark, will be lower in size than its tracking error.
Last year this rule of thumb was tested by comparing a portfolio’s prospective tracking error and its subsequent performance relative to its benchmark. The results showed that the prospective tracking error calculated by the Barra model tends to underestimate the level of risk in a portfolio. Where tracking error was high, the results were much more extreme. This is of particularly concern given that in this area trustees are especially keen to keep tabs on the risk a manager is taking.
There are more straightforward and perhaps more intuitive ways of assessing tactical risk, and ‘active money’ (or ‘load difference’) is the key
building block. Active money is
usually defined as the sum of the positive differences between the stock weightings in a portfolio and the benchmark. For example, suppose that XYZ Holdings is 2.8% of the portfolio but only 0.3% of the index – the active money is 2.5%. In assessing risk against the benchmark, investors should focus on this holding
rather than a far larger holding which has no active money, say 6.9% in ABC plc which is exactly in line with the index.
As part of our reports to clients we use active money to look at the biggest positions, namely the biggest positive and negative levels of active money. We use the sum of the active money in the top-10 positive positions as a measure of concentration in the portfolio. Another way of using active money is to arrange the stocks in a portfolio in descending order of their active money levels, to produce an ‘active slope’ for the portfolio. For most managers, significant positions
are confined to a small number of stocks.
We also define the measure ‘common money’ to complement active money – it represents what is in both the portfolio and in the benchmark. Because it doesn’t need to be
measured against an index, common money can also be used to compare different fund managers within
the same investment house to see how strongly the ‘house investment process’ is enforced and how much flexibility individual fund managers have to vary from the house process.
Sector and size weightings in a
portfolio are similar to active money positions but supplement the overall portfolio active money with an
indication of where the balance of risk lies. Looking at the sector and size weightings can help investors monitor the level of risk their manager is taking in these positions. As well as the risks a manager is taking relative to the index, investors will also want to know whether the level of their manager’s risk is changing and whether he is still taking the risks they expected when they appointed him.
We are increasingly moving to a situation where a range of risk measures are required to understand a portfolio. For a specialist manager where the level of variation away from the benchmark is likely to be high then traditional risk measures need supplementing by techniques which reflect the manager’s particular style and process. This has led us to develop a completely new technique, Simian, to assess portfolio-construction decisions. We expect this to give a fuller picture of risks and returns – and of managers’ luck and skill.
Sally Bridgeland is head of investment research and a partner and Matthew Levine is a quantitative investment analyst at Bacon & Woodrow, London