Japan is known to have resisted rapid yen appreciation by occasionally conducting foreign exchange interventions since it departed from the Bretton Woods system in August 1971.
The frequency and scale of interventions have been relatively large for a developed country, and Japan has been a popular country for studies in the intervention literature.

The daily data of interventions for the period since April 1, 1991 have been officially made available to the public. However, intervention data before March 1991 have not been officially disclosed.

Some researchers have relied on newspaper reports based on leaks and market rumors in order to analyze particular episodes.

Others have used either changes in foreign reserves or the data titled “Treasury Funds and Others/Foreign Exchange” released by the Bank of Japan as a proxy for intervention.
However, changes in foreign reserves, for example,include receipts of interest income on the foreign reserves, which are not interventions.

Ito and Yabu (2017) identified an accurate proxy for the monthly intervention amount for the period for which official data have not been disclosed.

The accuracy of different possible proxies was tested against “transaction-based” interventions which took place during the period from which data are available.

The best proxy, identified by Ito and Yabu, is more accurate than other popular proxies used in the existing literature, which include change in foreign reserves and the Bank of Japan’s "Treasury Funds and Others/Foreign Exchange."

In this paper, we first show that, in testing for the best proxy, it is important to distinguish “transactionbased” intervention from “settlement-based” interventions.

Due to the market practice in the foreign exchange market, settlement is done two business days later than transactions—a rule known as t+2.

We find that Ito and Yabu’s (2017) best proxy captures the movements of settlement-based interventions very well, explaining 99.8% of settlement-based interventions.

Our first contribution is that we provide reliable data for intervention, from August 1971—the end of the Bretton Woods system—to March 1991. From April 1991, the official data for intervention are available.

Second, we analyze the conditions under which interventions are most likely to occur for the period from August 1971 to March 2018.
The officially disclosed data are used for April 1991 to March 2018 and our proxy is used from August 1971 to March 1991.

Since reaction functions, which represent the monetary authorities’ tolerance of and resistance to exchange rate volatility, are expected to have evolved and transformed over the 47 years under consideration, possible structural breaks in reaction functions will be carefully examined in this paper.

Third, the US and German monetary authorities have disclosed daily intervention data for the period beginning in 1973 and 1976, respectively.
By analyzing intervention timings and amounts in the three countries, insights can be obtained on policy coordination or lack thereof, among the G3, in the past four decades.

Fourth, we present the episode of international coordination represented by the Plaza and Louvre agreements as a case study of notable interventions during the period under consideration.

The rest of the paper is organized as follows.
Section 2 will review the foreign exchange intervention literature. Section 3 will review Ito and Yabu (2017) and show that in testing for the best proxy, it is important to distinguish transaction-based intervention from settlement-based intervention.

Section 4 will provide an overview of the yen/dollar rates and interventions over the past 47 years.
Section 5 will analyze the conditions under which the Japanese monetary authorities are likely to intervene in the foreign exchange market.
Section 6 will present a case study on notable interventions during the period.