
ãé©åãªæšèšã®ããã®ããŒã¿åŠçðããJapanese Foreign Exchange Interventions, 1971-2018:Estimating a Reaction Function Using the Best ProxyãïŒå è¡ç 究解説 No.9ð2023/10/23
IntroductionïŒåæ¥è«æã¯æ©ãã«ä»äžãããð
ç§ãããããåæ¥è«æã®å·çã«
åããããææããã£ãŠåããŸããð
äœäºãã¢ãŠããããåæã®ã€ã³ãããã
倧äºã§ãããšãnoteã§æ¯æ¥çºä¿¡ããŠããŸãã
ããã¯ãã©ã®ãããªå
容ã§
ãã£ãŠãåœãŠã¯ãŸããŸãð
è«æãäžæŠã«èªãã§ã
èšæ¶ã«æ®ã£ãŠããªãã£ãã
倧åãªèгç¹ãå¿ããŠããŸã£ãŠãããããã
åæ¥è«æã®é²æã¯æ»ã£ãŠããŸããšæããŸã
ã ããããããã®ãnoteãããã«æŽ»çšããŠ
åæ¥è«æã1ïŒ
ã§ã
宿ã«åããŠé²ããŠãããããšæããŸã
ç§ã®åè«å·çãžã®è»è·¡ã
ã©ãããæèªãã ããð
ä»åã®åèæç®ð¥
ä»åãèªã¿é²ããŠããè«æã¯
ãã¡ãã®URLã«ãªããŸãð
ãJapanese Foreign Exchange Interventions, 1971-2018: Estimating a Reaction Function Using the Best Proxyã
Takatoshi Ito(a), Tomoyoshi Yabu(b)
(a) School of International and Public Affairs, Columbia University, and GRIPS, Tokyo
(b) Department of Business and Commerce, Keio University
èªã¿çµããå è¡ç ç©¶ð
ãæ¥æ¬ã®çºæ¿ä»å
¥ã®åæããäŒè€éæã»è
çµæžç ç©¶ãVol.54 No.2 Apr. 2003
ãEffects of the Bank of Japanâs intervention on yen/dollar exchange rate volatilityã21 November 2004
Toshiaki Watanabe (a), Kimie Harada(b)
ãThe Effects of Japanese Foreign Exchange Intervention: GARCH Estimation and Change Point Detectionã
Eric Hillebrand Gunther Schnabl Discussion
Paper No.6 October 2003
Japanese Foreign Exchange Interventions, 1971-2018: Estimating a Reaction Function Using the Best Proxy
December 12, 2019
Takatoshi Ito(a), Tomoyoshi Yabu(b)
(a) School of International and Public Affairs, Columbia University, and GRIPS, Tokyo
(b) Department of Business and Commerce, Keio University
ååã®ã埩ç¿ãð
3.3. Important adjustments in the test
Let us denote interventions conducted during the last two business days of the month by Int_t^End.
These amounts are recorded in the fiscal accounts that are used in the monthly proxy of month t+1.
Similarly, Int_t+1^End is recorded in the monthly proxy of month t.
Thus, the proxy of month t should be compared against settlement-based intervention amounts, calculated as:
Int_t^End ã«ãã£ãŠæã®æåŸã® 2 å¶æ¥æ¥ã®éã«è¡ãããä»å
¥é¡ã瀺ãããŸã
ãããã®éé¡ã¯ãt+1æã®ææ¬¡ä»£ç倿°(Proxy)ã§äœ¿çšãããäŒèšå£åº§ã«èšé²ãããŸãð
åæ§ã«ãInt_t+1^Endã¯ãt æã®ææ¬¡ä»£ç倿°ã«èšé²ãããŸã
ãããã£ãŠãt æã®ä»£çé¡ã¯ã次ã®ããã«èšç®ãããååŒããŒã¹ã®ä»å
¥éé¡(settlement-based intervention amounts)ãšæ¯èŒãããå¿
èŠãããã®ã§ã
In order to see the point of this discussion, the estimated errors of the best proxy ( Error= Int_t - Proxy_t) are examined.
We argue that this is mostly explained by the end-of-month interventions and the t+2 difference between transaction and settlement. We define the t+2 difference as EoMDifft (abbreviation of âend of month differenceâ):
ãã®è°è«ã®èŠç¹ãçè§£ããããã«ãæè¯ã®ä»£ç倿°ã®æšå®èª€å·® (Error= Int_t - Proxy_t) ã調ã¹ãŸã
ããã¯ãææ«ã®ä»å
¥ãšååŒããŒã¹éã®æ±ºæžæ¥æ°ã®ã©ã°ã§ãã t+2æ¥åŸãšããå·®ã«ãã£ãŠäž»ã«èª¬æããããšããäž»ã«ãªãããŸã
ãããŠæ±ºæž t+2 ã®å·®ã EoMDifft (ãææ«ã®å·®ãã®ç¥èª) ãšããŠå®çŸ©ããŠããŸã
Fig. 2 shows the time series of Errort and EoMDifft where Errort is again computed using âAdjusted MOF FEEFâ as a proxy.
As expected, the âErrorsâ appear to be explained mostly as a result of end-of-month interventions and the t+2 settlement.
To be statistically accurate, a regression is conducted:

å³2ã¯ãError_tãšEoMDiff_tã®æç³»åã瀺ããŠããŸã
Error_tã¯ãèª¿æŽæžã¿ MOF FEEFãã代ç倿°(Prixy)ãšããŠäœ¿çšãããå床èšç®ãããŠããŸã
äºæ³ã©ãããããšã©ãŒãã¯äž»ã«t+2 決æžã«ãŒã«ã«åºã¥ãææ«ã®ä»å
¥ã®çµæãšããŠèª¬æãããããã§ã
çµ±èšçã«æ£ç¢ºã«ããããã«ã以äžã®ååž°ãå®è¡ããããšã«ããŸã
Here HAC standard errors are in brackets.
Neither H0: β0= 0 nor H0: β1= 1 is rejected; the deviation of the proxy from transaction-based interventions is well explained by the end-of-month interventions.
In fact, 95.6% of the deviation is due to end-of-month interventions, as shown by R^2 = 0.956.
ããã§ãHACæšæºèª€å·®ã®å€ã¯ãæ¬åŒ§å
ã«ç€ºãããŠããŸã
æšèšçµæãããåž°ç¡ä»®èª¬ãH0: β0= 0ãããH0: β1= 1ããæ£åŽãããŸããã§ãã

決æžããŒã¹ã®ä»å
¥ããã®ä»£ç倿°ã®åå·®ã¯ãææ«ä»å
¥ã«ãã£ãŠååã«èª¬æãããŸã
å®éã®ãšããããã®ã¢ãã«ã®æ±ºå®ä¿æ°R^2ã¯ã0.956ã§ç€ºãããããã«ããã®ååž°åŒã§èª¬æããã被説æå€æ°ã®å€åã®ãã¡ã95.6%ã¯ææ«ä»å
¥ïŒèª¬æå€æ°ïŒã«ããå€åã§ãããšããããšãèšããã®ã§ã
We further examine how close settlement-based intervention data are to various proposed proxies for the period after the disclosure, April 1991.
Table 3 shows the regression results of Eq. (1) with the dependent variable being settlement-based interventions ( Int_t^Settlement). As shown, the explanatory power of proxies is much improved, compared to the one in Table 2.
For the regression using âAdjusted MOF FEEF,â the R2 improves from 0.969 in Table 2 to 0.998 in Table 3.3 This is a confirmation that the proxy is an accurate measure, explaining 99.8% of the movements of settlement-based interventions.
ããã«ã決æžããŒã¹ã®ä»å ¥(settlement-based intervention)ããŒã¿ããå ¬åŒããŒã¿é瀺åŸã®1991幎4æã®ããŸããŸãªä»£æ¿å€æ°ã®æ¡ã«ã©ãã ãè¿ããã調æ»ããŸã

ãŸã衚3ã¯ãååž°åŒ(1)ã®æšèšçµæã瀺ããŠããŸã
ãŸãååž°åŒ(1)ã®åŸå±å€æ°ã¯æ±ºæžããŒã¹ã®ä»å
¥ ( Int_t^Settlement ) ãšãªããŸã
衚3ã§ç€ºãããŠããããã«ã代ç倿°(prixies)ã®èª¬æåã¯ã衚2ã®ãã®ãšæ¯èŒããŠå€§å¹
ã«åäžããŠããããšãããããŸã

ãŸãã調æŽããã MOF FEEFãã䜿çšããååž°ã§ã¯ãã¢ãã«ã®æ±ºå®ä¿æ°(R^2)ã¯è¡š2ã®0.969ãã衚3.3ã®0.998ã«æ¹åããŸããð
ããã¯ä»£ç倿°(Proxies)ãæ£ç¢ºãªæž¬å®å€ã§ãããæ±ºæžããŒã¹ã®ä»å
¥ã®åãã®99.8%ã説æããŠããããšãè£ä»ããŠãããšè§£éã§ããã®ã§ã
The above detailed examination in search of the best proxy for interventions is thus concluded with both encouraging and discouraging results.
We are encouraged to have found that âAdjusted MOF FEEFâ is the best proxy with 99.8% accuracy.
However, our enthusiasm must be tempered slightly given that this accuracy is measured against settlement-based interventions. Hence, considering that daily data are not available prior to April 1991, the available monthly interventions data for months (t) covers the period from the last two days of month (t-1) to the third-to-last business day of month (t).
ãããã£ãŠãäžèšã§è¿°ã¹ãŠããä»å
¥ã«æé©ãªä»£æ¿å€æ°ãæ¢ãããã®è©³çްãªèª¿æ»ã¯ã倧å€ç ç©¶ã®å±ã¿ã«ãªããããªçµæãšã¯ããŸãèœèãããªçµæã®äž¡æ¹ã§çµè«ã¥ããããããšãããããŸã ã調æŽããã MOF FEEFãã99.8%ã®ç²ŸåºŠãæã€æè¯ã®ä»£ç倿°ã§ããããšã倿ããããšã¯å¬ããçµæãšèšããã®ã§ã¯ãªãã§ãããã
ãããããã®æ£ç¢ºæ§ã決æžããŒã¹ã®ä»å
¥ã«å¯ŸããŠæž¬å®ãããããšãèãããšãç§ãã¡ã®ç±æã¯å°ãå¶åŸ¡ãããå¿
èŠããããŸã
ãããã£ãŠã1991幎4æããåã¯ã調æ»ã«å¯Ÿããæ¥æ¬¡ããŒã¿ãå
¥æã§ããªãããšãèæ
®ãããšã第tæã«ãããå
¥æå¯èœãªææ¬¡ä»å
¥ããŒã¿ã¯ãt-1æã®æåŸã®2æ¥ããæã®æåŸãã 3çªç®ã®å¶æ¥æ¥ãŸã§ã®æéãã«ããŒããŠãããšããããšã«ãªãã®ã§ã
This may pose a problem in analyzing the effectiveness of interventions or the reaction function of the monetary authorities when intervention actions are concentrated at the end of the month.
However, an adjustment can be applied when such circumstances are suspected based on the volatility of the exchange rate; for example, one can use the exchange rate from the third to-last business day as the monthly exchange rate.
This is what we will do in section 5 when estimating the reaction function of the monetary authorities.
ãã®ããšã¯ãææ«ã«ä»å
¥ãéäžããå Žåã«ãããŠãä»å
¥æ¿çã®æå¹æ§ãéèåœå±ã®åå¿é¢æ°ãåæããäžã§åé¡ãšãªãå¯èœæ§ããããŸã
ãã ããçºæ¿ã¬ãŒãã®å€åã«åºã¥ããŠãã®ãããªç¶æ³ãçãããå Žåã«ã¯ã調æŽãé©çšãããå ŽåããããŸã
ããšãã°ã3 å¶æ¥æ¥ããæçµå¶æ¥æ¥ãŸã§ã®çºæ¿ã¬ãŒããææ¬¡çºæ¿ã¬ãŒããšããŠäœ¿çšå¯èœã§ã
ããã¯ãåŸè¿°ããã»ã¯ã·ã§ã³5ã«ãããŠéèåœå±ã®åå¿é¢æ°ãæšå®ãããšãã«è¡ã察åŠã«ãªããŸãð
æ¬æ¥ã®è§£èª¬ã¯ããããŸã§ãšããŸã
ãã®ãããªæŽå²ãå
è¡ç ç©¶ããã£ããçè§£ããäžã§ãåæ¥è«æå·çã«åãçµãã§ããããã§ã
ä»åãç§ãåæ¥è«æå·çã«ãããŠåãäžãã
24幎ã¶ãã®ãå宿¯æ£ãä»å
¥ã¯æ¬åœã«ã¬ã¢ãªçµæžæ¿ç
ãšããããšãåèªèã§ãããããªæ°ãããŸãð
ç§ã®ç ç©¶ããŒãã«ã€ããŠð
ç§ã¯ãçºæ¿ä»å
¥ã®å®èšŒåæããããŒãã«
åæ¥è«æãå·çããããšèããŠããŸãð
æ¥æ¬çµæžãèãããšãã«ãçºæ¿ã¬ãŒãã«ãã£ãŠ
貿æååŒãçµåžžåæ¯ãå€åããã
æ ªã蚌åžã嵿š©ãšãã£ãéèè³ç£ã®åççã
å€åããããšæ¥æ¬çµæžãšçºæ¿ã¬ãŒããšã¯
åã£ãŠãåããªãçžãããã®ã§ãð
ïŒåðŽã ãã«ïœ¥ïœ¥ïœ¥ïŒ
çµæžã·ã§ãã¯ã«ãã£ãŠ
çºæ¿ã¬ãŒããå€åãããš
ãã®åœ±é¿ã¯ç§ãã¡ã®ç掻ã«å€§ãã圱é¿ããŸã
ã ãããããçºæ¿ã¬ãŒãã®å®å®æ§ã
æ
ä¿ãããããªçºæ¿ä»å
¥ã¯ãã¯ãçµæžæ¿çã«
ãããŠãéåžžã«éèŠãªæçŸ©ãæã£ãŠãããš
æšæž¬ããŠããŸã
決ããŠåŠéšçãæ¥œããŠå·çã§ãã
ç°¡åãªããŒããéžæããŠããããã§ã¯ç¡ããšä¿¡ããŠããŸã
ãã ããã®åæ¥è«æãããåãããšã
ç§ã®åŠççæŽ»ã®é倧æãšãªãããšã¯äºå®ãªã®ã§
æåŸãŸã§ã³ãã³ããšåãçµãã§åããŸãð¥
æ¬æ¥ã®è§£èª¬ã¯ã以äžãšããŸãð
ä»åŸãçµæžåŠçè«éãªãã³ã«
瀟äŒèª²é¡ã«å¯ŸããçµæžåŠçèŠç¹ã«ãã説æãªã©
ææçŸ©ãªå
容ãçºä¿¡ã§ããããã«
åªããŠãŸãããŸãã®ã§
ä»åŸãšãå®ãããé¡ãããŸãð¥º
ãã¬ãžã³ã®ã玹ä»ð
ãã¡ãã®ãã¬ãžã³ã«ãŠ
åæ¥è«æå·çãžã®è»è·¡
ãšãã»ã³ã·ã£ã«çµæžåŠçè«éããªãã³ã«
ãåœéçµæžåŠðãã®åºç€çè«ããŸãšããŠããŸã
ä»åŸãããã«ã³ã³ãã³ããæ¡å
ã§ããããã«åªããŠåããŸãã®ã§äœåãããããé¡ãç³ãäžããŸãð
æåŸãŸã§ãæèªããã ãèª ã«æé£ãããããŸããïŒ
ãããŸã§ãç§ã®èŠè§£ãæã£ãããšã
ãŸãšããããŠããã ããŠãŸãã
ãã®ç¹ã«é¢ããŸããŠããäºæ¿ãã ããð
ãã®æçš¿ãã¿ãŠãã ãã£ãæ¹ã
ã»ãã®å°ããªäºã§ãåŠã³ããã£ãïŒ
èãæ¹ã®åŒãåºããå¢ããïŒ
èªæžããåŠã¹ãããšãå€ãïŒ
ãªã©ãªã©ããã©ã¹ã®åç©«ããã£ãã®ã§ããã°
倧å€å¬ããæããŸãããæçš¿äœæã®å¥å©ã«å°œããŸãïŒïŒ
ãæ°è»œã«ã³ã¡ã³ããããããã¹ããð
ãããŠããå·®ãæ¯ããªããã°
ãã©ããŒïŒã·ã§ã¢ããé¡ããããã§ãð
ä»åŸãšãäœåãããããé¡ãããããŸãïŒ