The implications of recent developments in the theory of exchange rate targets and currency bands for the credibility of exchange rate policies for member countries of a monetary union such as the European Monetary System are well known, but this recent field of research is also significant for exchange rate management in small countries that do not formally belong to such a union.

In Discussion Paper No. 493, Research Fellow Lars Svensson develops some simple tests of the credibility of an exchange rate target zone and applies them to Swedish data for the period January 1987-August 1990. He first constructs `rate-of-return bands' for foreign-currency bonds of different maturities. These are bounded by the foreign interest rates derived from the minimum and maximum home-currency rates of return that result as the exchange rate tends to the strong and weak edges of its band respectively.

For the exchange rate band to remain intact, the actual home-currency rate of return must lie within this rate-of-return band. Assuming international capital mobility, a target zone's credibility can be tested by determining whether domestic interest rates fall within the rate-of-return band. If they fell outside the band, there would be completely safe arbitrage, so the target zone could not be credible. Under the further assumption of uncovered interest rate parity, expected future exchange rate depreciation and the interest rate differential must be equal, so expected future exchange rates can be computed from interest rate differentials for different maturities, and a target zone can be shown to be credible if these rates fall inside the band. Svensson then uses these tests to show that throughout the period studied the Swedish target zone lacked credibility within a two-year horizon or longer, and occasionally also within a one-year horizon. The expected depreciation within a five-year horizon varied between 6.5% and 18%.

In Discussion Paper No 494, Svensson considers the `foreign exchange risk premium', i.e. the excess expected rate of return on domestic over foreign currency bonds, where both rates include expected currency depreciation. If this risk premium is zero, uncovered interest rate parity holds, i.e. interest rate differentials compensate for expected rates of currency depreciation, so it is easy to compute expected future exchange rates and to test target zones' credibility from current data. Moreover, uncovered interest rate parity also implies that there is no scope for monetary autonomy in target zones, since interest rate differentials cannot deviate from expected rates of currency depreciation.

Previous theoretical and empirical research has generally concluded that foreign exchange risk premiums are fairly small under floating exchange rates, which suggests that they should be even smaller in exchange rate target zones. They are typically characterized, however, by large, discrete realignments and devaluations at irregular intervals, and specific devaluation risks may give rise to sizeable foreign exchange risk premiums.

Svensson derives both real and nominal foreign exchange risk premiums in a target zone model with devaluation risk. In each case the risk premium is the sum of two separate risk premiums: the first, arising from stochastic exchange rate movements within the band, is very small, as expected; while the second, arising from stochastic devaluations or realignments, is considerably larger, but nevertheless small in relation to the expected rates of devaluation or realignment. Even with very risk-averse investors, the foreign exchange risk premium barely exceeds one- fifth of the observed interest rate differentials. Svensson concludes that the foreign exchange risk premiums in target zones are fairly small and clearly dominated by the expected rates of currency depreciation resulting from movements inside the band and anticipated shifts of the bands. Uncovered interest rate parity seems to be an acceptable approximation for target zones.

In Discussion Paper No. 495, Svensson presents a theoretical derivation of the term structure of interest rate differentials for a target zone exchange rate regime and tests some of its implications on Swedish data. First, for a fully credible target zone, with the exchange rate measured in units of domestic per unit of foreign currency, interest rate differentials are decreasing in the exchange rate. When the exchange rate is low at the strong edge of the band it cannot appreciate further: it can stay where it is or depreciate towards the middle of its band. On average it will depreciate, and the interest rate differential must be positive to compensate. When the exchange rate is high, the reverse applies. Further, interest rate differentials are less sensitive to exchange rates as the maturity of the bond lengthens, since the annualized interest rate differential is determined by the total expected exchange rate depreciation divided by the number of years to maturity.

Svensson then considers a target zone with some risk of realignment or devaluation. A risk of devaluation increases both the expected depreciation of the exchange rate and the interest rate differential. If the risk of devaluation is fairly independent of the exchange rate's location in the band, the interest rate differentials will remain a decreasing function of the exchange rate, while if the risk increases as the exchange rate depreciates within the band, the interest rate differential may be an increasing function of the exchange rate.

Svensson tests these theoretical findings on Swedish target zone data for February 1986-October 1990 and finds that interest rate differentials were indeed decreasing in the exchange rate (which indicates that devaluation risk was fairly independent of the exchange rate's position in the band) and also less sensitive to the exchange rate for longer maturities. The average expected rate of devaluation lay between 1.2% and 1.9% per annum, which corresponds to a probability of devaluation (for a devaluation of 10%) of 12-19% per annum (or an expected period between devaluations of 5-8 years), .

The Simplest Test of Target Zone Credibility
The Foreign Exchange Risk Premium in a Target Zone with Devaluation Risk
The Term Structure of Interest Rate Differentials in a Target Zone: Theory and Swedish Data
Lars E O Svensson

Discussion Papers Nos. 493-5, January 1991 (IM)