MANAGEMENT OF FOREIGN CURRENCY RESERVES

Performance Measurement

Benchmarks

It is widely accepted that central banks should manage their foreign currency portfolios against benchmark portfolios that are designed to reflect the risk appetite embodied in their investment mandates. Subject to this constraint, a well defined benchmark should also ensure that the return on the foreign currency portfolio is maximised and that returns in excess of the benchmark return are readily attributable to the individuals that made any discretionary portfolio management decisions.

Generally, a good benchmark should be:

• Representative – the benchmark should reflect the central bank's broad objectives in regards to reserves management as well as the range of eligible instruments specified in the investment mandate;
• Replicable – all instruments represented in the benchmark should be eligible instruments and should be readily tradable;
• Transparent – the benchmark should be clearly defined and easily understood; and
• Computable – portfolio managers should be able to review the benchmark portfolios and assess their relative exposures on a continuous basis.

A three-month instrument (such as a Treasury Bill) is often used as a reserves management benchmark as it represents the funding (opportunity) cost of managing the reserves portfolio. Superficially, it appears to satisfy the required characteristics of a good benchmark – it is replicable, it is transparent and it is computable. It is, however, unlikely to be representative of the range of instruments available to the central bank and, because it is relatively easy to achieve, it is not likely to represent the highest portfolio return that the central bank could reasonably expect to achieve.

Many central banks rely on indices generated by global investment houses.3 These indices are attractive benchmark options – they are transparent, they are computable, they represent a significant proportion of a bank's investment mandate and (importantly for central banks that face resource constraints) they are updated daily and are usually available at no cost. One limitation with them is, however, that while they are superficially replicable their coverage is so wide that it is often not possible for a central bank to purchase all the securities that they represent. Further, notwithstanding that these indices can be aggregated together and therefore can represent a significant proportion of a central bank's investment mandate, they are often narrowly defined in terms of instrument types (i.e. US Treasury Notes) and, as such, are unlikely to incorporate the range of investment alternatives available to a central bank.

Another alternative is for central banks to customise their own benchmarks. Although these benchmarks can be readily designed to meet all the characteristics identified above, they may expose the central bank to criticism about their appropriateness and effectiveness. A benchmark that is customised by an external consultant overcomes some of this criticism but this is not a costless solution.

Benchmark Composition

Irrespective of the decision about whether to use an internally or externally generated benchmark, a central bank has to determine the optimal composition of its benchmarks – in terms of their duration, their composition and their currency composition.

Portfolio Duration

A decision about the duration of the benchmark portfolio is essentially a decision about managing interest rate risk – the risk that the value of the instruments will decline as interest rates increase. This decision involves three competing factors – a central bank's investment horizon, the expected return on the assets and the expected variability of these returns. Taking a simple example of an investor with a one-year investment horizon and assuming normal (upward sloping) yield curves, average portfolio returns should increase as the duration of the portfolio increases (Graph 1). The risk in this case is, however, that the variability of expected returns around this average also increases as the duration of the portfolio increases. Taking the example in Graph 1, there would be a 1 in 40-year probability of a negative return if the duration of the portfolio were 2½ years (the point where the 95 per cent confidence interval transects the x axis). The impact of increasing the initial investment horizon restriction from one to two years is illustrated in Graph 2. Here the variability of returns is reduced in such a way that the duration of the portfolio and, as a result, the return on the foreign currency portfolio can be higher for a given portfolio mandate (risk tolerance).

Portfolio Composition

A decision about the duration of a portfolio benchmark simply identifies the interest rate risk that a central bank is prepared to accept given its broad investment objectives. Even with the narrowest of investment mandates, however, there are an infinite range of combinations of instrument types and terms to maturity that could be chosen to represent this duration and these are likely to have quite different risk and return profiles. For example, a benchmark of 2½ years can be achieved by holding only 3-year securities or alternatively by holding a weighted combination of 3-month and 10-year securities. While the returns on these portfolios will be identical for a parallel shift in the yield curve, it will not be the case if the slope of the curve changes. Similarly, if the investment universe were to also include products with different credit risk profiles, benchmark returns would vary depending on the extent to which these instruments were included even if the duration of the instruments held were identical.

Another important consideration is whether the benchmark is replicable and liquid. A benchmark that includes a multitude of ‘off-the-run' securities meets neither of these objectives – it is unlikely that a portfolio manager would be able to hold all the instruments and, by definition, if they were to hold them then they may have difficulty selling them at a fair price at short notice. Similarly, a diversified portfolio of corporate bonds may earn a higher return over the long run, but may require an investment in so many securities that it may only be possible to liquidate the portfolio at short notice at a substantial discount.

To some extent, however, these replication and liquidity issues can be overstated. Portfolio management techniques enable market-capitalisation indices to be readily replicated using a considerably smaller sample of securities than is contained in the indices. Indeed, in many cases, the tracking error or deviations from benchmark can be relatively small (and still be consistent with a central bank's broad investment mandate). Similarly, with the development of deep and liquid secondary markets and the plethora of hedging and funding instruments that have been developed over the past two decades (such as futures and repurchase agreements), the liquidity requirements of a fixed-income portfolio can be readily achieved.

Currency Composition

When selecting the currency composition of the foreign currency portfolio, consideration needs to be given to the ultimate objective of holding the reserves. If reserves are being held for intervention purposes then the portfolio may need to be relatively overweight the intervention currency(s) of choice. On the other hand, if foreign currency reserves are being held as a hedge against official or non-official foreign currency liabilities or are funded with foreign currency loans, it may be appropriate to match their currency composition. Finally, if foreign currency reserves are used to fund imports or cover current account deficits, consideration should be given to a currency composition that reflects the numeraire of trade flows.

Consideration also needs to be given to the degree to which prospective currencies are correlated and whether, for operational reasons, larger portfolio allocations would be more efficient. For example, the currency allocation to Australian and New Zealand dollar assets is likely to be relatively small in most foreign currency portfolios. Assuming that the high correlation between these currencies were to persist, it may be practical to merge the benchmark allocation and only have an exposure to one or the other currency. This doesn't preclude investments in either market but it does make the benchmark easier to maintain and easier to replicate.

Benchmark Optimisation

Once the duration, composition and currency composition of the benchmark portfolios has been determined, the next task is to assign weights to each portfolio to determine the amount of funds to be allocated to each one. This process involves optimising returns subject to a number of policy-related operational constraints (such as minimum portfolio sizes). Using mean-variance analytical tools, an efficient frontier can be determined which indicates the highest return achievable for each level of risk (return variability).

The choice of where to operate along the efficient frontier is somewhat arbitrary. At one extreme, a portfolio can be chosen such that the diversification benefits are maximised. This point (A in Graph 3) represents the least risky point on the efficient frontier but, by definition, also represents the lowest expected return. An alternative strategy is to select the point where the marginal compensation for an additional unit of risk is equal to the marginal return (B).

One limitation with this approach is that it is based on historical price movements and, as such, there is no guarantee that these returns will be repeated in the future. What this methodology does, however, is provide an objective means by which a central bank can gauge the portfolio composition that would have generated the highest returns for a particular risk profile in the past.

Reporting Issues 

Numeraire

When reporting foreign currency portfolio returns, a central bank can choose from a variety of numeraires (or units of account). Ideally, a numeraire should represent a risk-free measure for the value of the foreign currency portfolio such that the purchasing power of the reserves is unaffected by currency movements. Three types of numeraire are usually considered:

• Most private sector fund managers report portfolio returns in the domestic currency of their clients. Although this numeraire has greatest relevance when reporting portfolio returns to a domestic audience, it is likely to represent a significant moral hazard problem for central banks that actively intervene in their own currency.
• Using a non-domestic currency (such as the US dollar or euro) as the numeraire avoids the moral hazard problem and may be attractive if it is the currency of choice for intervening in foreign currency markets. The downside to such a numeraire is that it is likely to distort portfolio management decisions. For example, if the US dollar were used as the numeraire and it is one of the investment currencies it is quite likely that the ‘optimal' currency composition of the portfolio would only contain US dollar assets. This is because a portfolio that only consists of assets expressed in the numeraire would involve no currency risk and would therefore tend to have the lowest risk profile.
• Reporting returns in a weighted currency basket, such as Special Drawing Rights (SDRs), eliminates the moral hazard problem and trading distortions associated with the previous two measures. It does, however, suffer because it has very little relevance to a domestic audience.

Ultimately, the choice of numeraire is very much dependent on its actual use. In Australia's case, the optimal portfolio composition is estimated using portfolio returns expressed in SDRs. For operational purposes, returns are reported in US dollar terms on a daily basis and the portfolio's actual and relative performance is reported in the Annual Report in Australian dollar terms.

Financial vs Accounting Returns

Best practice suggests that foreign currency portfolios should be marked-to-market each day using publicly available bid prices (i.e. the prices that would be achieved in the event that the reserve assets were liquidated). Importantly, performance data should reflect the interest earned as well as all realised and unrealised capital gains and losses on all financial exposures (including all contracted and settled transactions as well as all on- and off-balance sheet exposures). Depending on a country's accounting standards, this measure of portfolio performance may, however, differ from returns published in the Annual Report. This may especially be the case where the accounting standards do not recognise income on off-balance sheet and contracted exposures or which exclude unrealised capital gains and losses. As with the decision about the numeraire, it may be that a central bank will choose to report returns on a different basis depending on its ultimate purpose.

Footnote

3. Most of these indices are capitalisation-weighted, i.e. they are return indices based on the total supply of securities on issue with a particular characteristic. For example, a 3 to 5-year US Treasury Index would include all US Treasury securities which have between three and five years to maturity. (back to text)