Rational speculation, central bank intervention and the negative forward discount bias

The negative forward discount bias in the foreign currency was rationalized and tested in the context of rational speculation model. This study reexamines the issue within the context of a two-agent rational speculation model and finds empirically that the negative forward discount bias is still a puzzle. In an extension of the rational speculation model, non-speculative demand for foreign currency that is influenced by the difference in interest rates in two countries can be shown to influence the extent of the forward discount bias. This effect is generally difficult to test. However, if the intervention policy of central banks changes over time, as they did in the 1980s, then there are testable predictions about how the intervention will affect the forward discount bias. This study extends the theoretical rational speculation model to include central banks intervention. An increase in the foreign interest rate tends to appreciate the foreign currency. A central bank that leans against the wind by selling the foreign currency to slow down its appreciation will add to the forward discount bias and may even make it negative. A central bank that leans with the wind by buying foreign currency to increase its rate of appreciation will tend to make the forward discount bias smaller. The hypothesis is tested using the thirteen-country currency data against the dollar for the period of 1980.01 through 1994.12. Since, the sample period includes two well-publicized cooperative attempts to influence dollar exchange rate among G-5 and G-7 central banks, the model predicts that “the β 1 coefficient” increases in the sub-period of 1985.01–1987.01 and decreases (and can be negative) in the sub-period of 1987.02–1989.12. The hypothesis was strongly supported by the empirical data. In most of the currencies, regression estimates of “the β1 coefficient” increased in the Plaza period, and decreased and were mostly negative in the Louvre period. Empirical data do not support the hypothesis that these changes in the β1 coefficients result from structural shifts in the process of the interest-rate differentials. Therefore this study also supports the hypothesis that intervention can have an effect through a portfolio-balance channel