Appendix 2

The risk/return characteristics of ECU-denominated bonds were such that the ECU figured prominently in global portfolios. Somewhat surprisingly, the share of ECU assets even increased for investors moving along the efficiency frontier, that is, investors who did not wish to minimize risk. This outcome suggests that the ECU had more to offer beyond hedging properties and, for that matter, afforded an attractive remuneration for the risks it involved. What were those risks?

We identified three types of risk that are analytically distinct. In all but the notional European habitat, the ECU was a foreign currency and as such contained a "foreign currency" risk. We show that this component was the largest risk factor. Of course, in a European habitat, the "foreign currency" risk disappeared and the ECU became a very low-risk investment.

The second source of risk derived from the open-ended construction of the ECU. ECU interest rates reflected expected changes in its currency weights and such expectations could be, and usually were, off the mark. Therefore, there was a type of risk which did not exist for a basket of currencies replicating the ECU. This risk was quite small for short maturities but could rise up to 30 basis points (bp) for longer maturities. This type of risk existed in all habitats including the European one. It could only be eliminated by either fixing the weights rather than the currency units in the ECU, or by doing away with the basket construction altogether.

Both "foreign currency" and "basket" risks are habitat-specific. An ultimate source of risk derived from basket revisions and liquidity problems; the premium for this type of risk was reflected in the spread between the theoretical and market interest rates of the ECU. Unlike the two previous ones, this particular type of risk was similar across all habitats. Its size was small relative to the "foreign currency" risk, but large relative to the "basket" risk. This type of risk could have been substantially reduced if ECU weights had no longer been revised at five-year intervals, as argued by Girard and Steinherr (1989a).

For remunerated financial assets that are traded internationally, it is obvious that one cannot define risk in an absolute sense. Modern portfolio theory suggests that risk can be defined in a meaningful way only by reference to a relevant portfolio. Any asset has two risk components: one representing the risk of the whole market (market risk) and one that is asset-specific and defined in relationship to this market (the “beta” value of the asset). Analysis of the risk of any particular foreign currency asset could be carried out through the same approach. It would require construction of an international portfolio which would serve as a benchmark for the purpose of assessing the degree of risk associated with the assets available in a particular currency. This approach raises a number of conceptual difficulties. The most significant appears to revolve around how to arrive at an appropriate definition of such an international benchmark portfolio which, moreover, must be habitat-specific.

This is why we adopted a different approach. We took national habitats for benchmarks and were, therefore, only able to assess the ECU-specific risk in relation to the currency of a specific national habitat. However, because we also retain an ECU habitat, the non-national habitat approach is not altogether overlooked. Such an approach, if more limited, seems warranted not only on account of the greater simplicity it affords but also for the task at hand: that is, the attractiveness of the ECU as an alternative to national currency assets in European habitats.

In habitat h we define:
ÐT h = ÐEh + S (A1)

where ÐT h is the difference between the ex post return on an ECU-denominated asset converted into currency h on the one hand and, on the other hand, returns on assets denominated in currency h; ÐEh is the difference between the ex post theoretical return on the ECU and effective returns on assets in habitat h; S is the spread between the effective and theoretical ECU returns.

The spread S is independent of the choice of habitat, but PT h (defined in equation A1) is clearly habitat-specific. However, the contribution of the spread to the total variability (risk) of an ECU asset varies across habitats, as long as the covariance (Cov) between PE and S is different from zero:

V(PT h ) = V(PEh) + 2 Cov (PEh, S) (A2)

The variable PT h assumes different values across time periods, which is precisely the risk of holding ECU assets. For a given time period, the volatility of PT h (measured by its standard deviation, V) is a measure of the risk in holding ECU-denominated assets in habitat h. The average value of PT h is a measure of the average risk premium an investor in habitat h will require if he is to hold an ECU asset.

Equation (A1) can be used to compute average risk premiums for a choice of relevant time periods. This equation shows that the total ECU risk can be decomposed into two elements, namely PEh and S.

The risk premium associated with the variable PEh is usefully called a premium for “intrinsic” risk. This risk itself originates in two sources: one is the individual risk premium attached to every single currency in the basket from the viewpoint of an asset holder in habitat h. This can be called “foreign exchange risk.” Another, much less apparent source of risk derives from the way ECU interest rates are determined. For a detailed description, see Girard and Steinherr (1989b).

The theoretical ECU interest rate for a very short-term instrument (say, overnight) will simply be the average of appropriate national interest rates weighted by the actual shares of currencies in the ECU. However, for longer maturities, consistency requires that covered interest rate parity not be overlooked. This means that interest rate spreads are compensated by corresponding forward exchange discounts. Therefore, on this basis, one cannot assume that currency shares in the ECU remain constant. Rather they must be projected using forward exchange discounts equal to interest rate differentials.

While interest rate parities yield consistency, experience shows that forward exchange rates are “blue” (i.e., best-linear-unbiased) estimators, though their record is very poor. This means the margin for error is large, implying higher risk. Such risk would not exist were the components of the ECU held directly and separately. For this reason we call it “basket risk.” Interest rates on each component currency were determined in the relevant domestic markets and given current weights; the return on an ECU-equivalent basket is determined without bringing changing weights into the picture.

The second component in equation (A1) is the spread. Spreads are influenced by several factors, such as measurement errors, illiquidity in some component currencies, shocks too specific to domestic markets to spill over to Euromarkets (this is important since the theoretical ECU interest rate ought to be computed on the basis of tax-free euro-instruments, but for reasons of data availability is, in fact, computed on the basis of domestic market rates), and, finally, any degree of uncertainty about future basket revisions. Thus, the spread is a measure of a premium which includes several distinct factors, of which only two are associated with risk: the risk of illiquidity which characterizes ECU markets to various degrees over time and uncertainty about future basket revisions. For empirical results see Girard and Steinherr (1989b).

Download this Discussion Paper [ PDF 274.5KB| 38 pages ].

Comment(s)

There are [0] comment(s) for this entry. Post a comment.