Paul Duncombe looks behind the marketing stories

As trustees around the world increase their plans' holdings of international investments, many are faced with the challenge of developing and implementing a currency policy. This is not easy - there is no one single policy that suits all. Each plan has its own particular objectives and idiosyncrasies, which demand a unique solution.

In all cases, the process should broadly stick to the following path -=

p Understand the impact of currency on your plan.

p Assess how it might impact the ability to meet investment objectives and liabilities.

p Decide on a suitable strategic hedge ratio.

p Decide whether to use an active currency manager.

Selecting an overlay manager is one decision, which many trustees find very difficult. The market for overlay business is very concentrated compared to other asset classes - the top 10 firms account for over 80% of overlay contracts awarded. Relatively few mandates are awarded each year and the competition to win them is intensive. Understandably, trustees have less experience of currency than other asset classes and therefore find it more difficult to assess the subtle differences between the various styles. The managers themselves do not make it any easier. Most of the top firms have spent years refining their marketing stories and, as one trustee said, they all sound very plausible". There is a lot of packaging of processes, which attempts to increase the differentiation from others.

For the moment, let's focus on the pure currency overlay market, which I am defining as the management of pre-existing currency risk. This definition excludes hedge funds and other programmes where currency risk is created specifically to generate profits. In other words, a plan has currency risk from its international in-vestments and uses overlay management to reduce that risk.

Risk is a word that can be defined in different ways. It is most commonly used to mean volatility, or the standard deviation of a series of returns. However, most plan sponsors are not primarily concerned with the volatility of currency returns - their objective is to avoid losses from adverse currency moves.

This means that currency overlay managers interpret the expression 'managing risk' to mean, 'avoiding losses'. Therefore the objective of most programmes is to reduce exposure to foreign currency selectively - the plan wants to keep the upside risk when foreign currencies are rising and avoid the downside risk when they are falling. This means that the performance of the perfect overlay looks like a zero-cost option with 100% participation in gains from upward currency moves and 0% participation (or 100% protection) in losses from downward moves (see below).

Despite the different philosophical styles of overlay managers, all, in their own way, are seeking to achieve as closely as possible the pay-off of this zero-cost option.

There are inefficiencies in the foreign exchange market which active managers can identify and capitalise on. The vast majority of managers use one of two sources. The first source is the non-random pattern of currency returns, which leads to clearly observable trends. The distribution of these returns is 'fat-tailed', with a larger than expected number of observations of high magnitude. The volatility of the returns is also somewhat non-random, with volatility having a tendency to be mean-reverting. This allows managers to use historical price data and statistical analysis to predict directional movements as well as volatility. This type of modelling is not founded on any economic theory, but is by nature empirical and the structure of the models will be driven purely by the statistical properties of the historical data. Skeptics of technically based approaches have been predicting its demise for years, citing increasing efficiency in the markets. However, the reality is quite the opposite. There have been more rather than less distinct trends in the last three years, and the market shows no sign of losing these characteristics. One of the key traps to avoid in the use of technical approaches is data-mining. It is possible to construct large numbers of simulated strategies that look wonderful historically, but then completely fail when let loose on real money. The art here is to avoid over-designing models and to identify consistent, predictable patterns of behaviour. Remember the rule used by engineers - 'the more moving parts the more that can go wrong!'

The second source of 'inefficiency' is the use of fundamental or macro-economic data to predict currency movements. This type of approach has the advantage that models can be based on economic theory. Theories such as Purchasing Power Parity, Portfolio Balance model, Monetary model, Uncovered Interest Arbitrage model and Balance of Payments model all postulate causality between exchange rate movements and macro-econom-ic factors. The use of these factors in currency models does create other challenges - data is only available monthly and therefore such models tend to have investment horizons that are longer term in nature.

Data quality is also an issue. The methods used to calculate some data series, for example unemployment and money supply, are revised frequently so that the value of this type of data has to be close to worthless. Programmes using 'harder' variables such as inflation rates and market derived data (bond yields) will be more ro-bust. Nevertheless, the long-term causality between these factors and exchange rate movements means that many managers recognise the value of including fundamental modelling as a part of their processes.

Broadly speaking, there are two styles of overlay management. Both styles attempt to capture one or more of the inefficiencies described above in forecasting models. The first style has its origins in dynamic hedging, sometimes known as option replication, in which the level of hedging is increased as the exchange rate falls and decreased when it rises. The pay-off of these strategies is asymmetric, similar to an option, but depends on how volatile prices are during the life of the strategy and also the degree of trending in prices. These strategies perform best when volatility is low and prices trend. Many managers using this type of strategy now generate their own forecasts of volatility rather than relying on the price of implied volatility from the market. Studies show that volatility has a tendency to be mean reverting and so ARCH/GARCH techniques can be used to give better estimates of volatility than using simple historical values. Predicting volatility means that the magnitude of the price change in the next period can be forecast, and therefore the size of the trade appropriate for a given level of risk. A forecast of market direction is still required to determine whether the hedge should be increased or decreased. Additionally, some managers use more than one volatility-forecasting model in their process, with each one focusing on a different time horizon. Regardless of the number of variables being forecast in these types of program, all are based on technical analysis, where the sole source of information is the historical price series of the exchange rate.

The second style of management forecasts the direction and/or magnitude of returns to determine the level of hedging required. The philosophy of this style relies on both fundamental and technical approaches to model exchange rate returns. In these programmes technical analysis is used to identify trends, and also the degree to which trends have run away excessively, leaving currencies 'overbought' or 'oversold'. Trends are most often identified by using moving averages, and have been used successfully for many years. Measures of price acceleration are very useful as acceleration has mean-reverting properties- if it reaches a high level it always declines subsequently. Most of these types of indicator are directional and the level of hedging can be adjusted to reflect the strength or degree of confidence in the trend. The second approach within this style models fundamental data. Most of the econometric theories described in the preceding paragraphs are used by one or more managers. They can be very useful in identifying currencies which are grossly over or under-valued, although the manager needs to have a mechanism in the process to avoid buying a cheap currency which is still declining. This is because the signals from fundamental approaches tend to be leading indicators. This is the opposite of technical approaches, which are lagging indicators, but the complementary nature of the two approaches means that most managers in this group use some combination of them.

Styles differences mean that portfolio allocation techniques also vary from manager to manager. Once the front end of the investment process has generated a signal/forecast/decision, that output must be converted into an active position. The size of the position taken will be the most significant factor in determining the degree of active risk assumed. This degree of risk assumed might also be influenced by recent performance, which can lead to a dilemma. Consider the situation where a client has specified, for example, that the portfolio should not underperform by more than 2%. Because of a period of volatile, directionless markets, the manager has lost money and is 1.5% behind the benchmark. Now, the manager only takes small active positions because he does not want to risk falling below his -2% target. The market recovers from its directionless phase and the program makes money. But the manager only adds a small amount of value and an opportunity to improve the performance has been partially lost. So there is the dilemma. All programmes are going to have periods of underperformance, and very often the worst action to take after such a period is to reduce the program. In order to maximise the potential of a program the best course is usually to stick with it even if this means that the risk of underperforming the benchmark has to be occasionally higher. The important issue to remember is that guaranteeing a floor to performance does not come without a cost.

It is sometimes tempting to undermine the approach used by a competitor, citing academic articles as 'evidence' of the folly of following adopting such an approach. This should be treated with a degree of scepticism. Academic articles can be very useful and have often opened up whole new areas of knowledge (e.g. modern portfolio theory). However, these tend to be general techniques rather than specific strategies. All of this may not make the decision of choosing an overlay manager any easier. The intention of this article has been to help potential users of currency overlay see through the marketing stories so that they can ask really pertinent questions about the strengths and weaknesses of the different programmes. It is important to remember that a potential client should really dig deeply into the investment process and understands its workings. Some plan sponsors feel a degree of discomfort with quantitative processes, the so-called 'black boxes' of the investment world.

An experienced manager should be able to demonstrate that the 'black box' is really a 'glass box'. The investment process should be sufficiently transparent that the client can see how it works and how the inputs are an-alysed and translated into hedging decisions for the portfolio. This should leave the client with a much greater feeling of comfort. Obviously other considerations such as strength and depth in the investment team, performance and organisational stability will also play a part in the decision making process. Having weighed all these factors up, the decision, as ever, is in the hands of the buyer.

Paul Duncombe is investment director at State Street Global Advisers in London"