Using ROC techniques to measure the effectiveness of foreign exchange market interventions

ABSTRACT

Using as an example data on official Japanese interventions in the yen/dollar market, I show that techniques developed to analyse relative operating characteristic (ROC) curves can be useful to study the forecasting value and effectiveness of interventions.

KEYWORDS:

• ROC techniques
• exchange rate
• interventions
• Japan

JEL CLASSIFICATION:

• E52
• F31

Notes

¹ The Japanese interventions have been extensively studied in earlier literature. On the effectiveness of Japanese interventions, see Ito (2002). On interventions and exchange-rate volatility, see Frenkel, Pierdzioch, and Stadtmann (2005a). On central bank reaction functions, see Frenkel, Pierdzioch, and Stadtmann (2005b). On large versus small-scale interventions, see Frenkel, Pierdzioch, and Stadtmann (2004). On coordinated interventions, see Frenkel, Pierdzioch, and Stadtmann (2003). For a survey of central bank interventions, see Sarno and Taylor (2001).

² Data on interventions and the exchange rate (noon buying rates in New York City) can be found at http://research.stlouisfed.org/. I use the next available data to fill any missing data on the exchange rate.

³ For similar definitions, see Humpage (1999) and Fatum and Hutchison (2006).

⁴ This definition of the change of the exchange rate accounts for a potential endogeneity of interventions within a cluster (Fatum and Hutchison 2006). For simplicity, the last intervention of an intervention cluster determines whether it is a ‘purchase cluster’ or a ‘sale cluster’.

⁵ Given that sales of foreign currency only occurred in the first half of the sample period, results for the forecasting value of interventions are similar and are not reported.