Heterogeneous behaviours and the effectiveness of central bank intervention in the yen/dollar exchange market

Abstract

A nonlinear heterogeneous agent model is applied to the yen/dollar exchange rate market to discuss the channels of an effective central bank intervention. The existence of two hypothetical channels proposed by Hung (1997) and Taylor (2004, 2005) are tested and confirmed. Evidence shows heterogeneous agents are active in the yen/dollar market where the stabilizing force from the fundamentalists declines in large misalignments. Central bank intervention is effective in arousing the trend-reversing sentiment among chartists to prevent market from explosion. The intervention is also effective in strengthening fundamentalists’ confidence that the market will move toward its theoretical equilibrium. The intervention has significant effects on fundamentalists’ confidence, regardless of whether the forecasting method relies on Purchasing Power Parity (PPP) only or on a PPP plus Uncovered Interest rate Parity (UIP) condition. The interest rate differential can affect the exchange rate changes through influencing demand orders of the short-run fundamentalists.

Keywords:

• nonlinearity
• heterogeneous agents
• central bank intervention
• foreign exchange market
• uncovered interest rate parity

JEL Classification::

• E58
• F31
• G15

Notes

¹ Sarno and Taylor (2002) provide a comprehensive review on the topic, in which various aspects of the effectiveness of intervention are discussed.

² Hung (1997) states the demand orders of noise trader are influenced by beliefs or sentiments that are not fully consistent with economic fundamentals. Besides, most noise traders rely on a feedback rule, so they generate momentums in price.

³ The simulation results in Taylor et al. (2001) show the mean-reversion speed (i.e. the half-life of a shock) of yen/dollar exchange rate ranges between 14–42 months, depending on the magnitude of shocks.

⁴ One possible reason is the fundamentalists may not exactly know the fundamental value of a given currency. Even if they do know the true fundamental equilibrium value of that currency, the unpredictable future changes in price expose them to high risks. As the markets are ruled by irrational expectations of non-fundamentalists, large and persistent misalignments prevail. Individual trader who sticks his analysis on fundamental factors would experience large losses, exhausting his liquidities (Reitz and Taylor, 2008), or his credibility with the managers (Shleifer and Vishny, 1997). Besides, most agents are subject to periodical reviews on their performance, making it even harder to wait for a correction in the current price. Confidence in the fundamental analysis fades with the enlargement in exchange rate misalignments.

⁵ The BFGS algorithm provides robust covariance matrices and consistent estimates under the absence of conditional normality (Bollerslev and Wooldridge, 1992).

⁶ The NRNL statistics is obtained by re-estimating the auxiliary regression of Equation 10 with standardized residuals (instead of exchange rate returns) used in the left-hand side. Based on a LM-type test, the null hypothesis with no remaining nonlinearities (i.e. H ₀: λ_i  = ϕ_i  = 0 for i = 2, 3, 4, 5 and j = 2, 3, 4) is tested against the alternative of additional nonlinear structure.