亞洲央行干預外匯市場的有效性及對美國存託憑證價差的影響 - 政大學術集成

全文

(1)國立政治大學財務管理研究所博士論文. 政治大. 憑證價差的影響. 學. ‧ 國. 立亞洲央行干預外匯市場的有效性及對美國存託 ‧. n. er. io. sit. y. Nat 研究生：張美菁. 撰. al. v i n Ch i U e h n c g 指導教授：張元晨博士. 中華民國一〇四年六月.

(2) 謝辭博士班的求學生涯充滿多采多姿的回憶，有豐富的課程、具挑戰性的學科考試、以及千錘百鍊的研究，這些的歷練豐富我的人生。感謝指導老師. 張元晨老師在我遇到任何研究上的困難時，總能適時地幫. 忙與協助！從您的身上，學生學習到為人的謙卑、處事的圓融，讓我見識到學者的風範，謝謝您的指導！學生的博士論文獲得國科會 102 年度博士論文獎，特別感謝老師您費心地指導寫作！也感謝您介紹 Sandy Suardi 老師一起合作，讓我認識到一位積極樂觀對學術研究充滿著熱誠的老師，學生獲益匪淺！謝謝 Sandy 老師協助研究！. 政治大們的寶貴意見使我的論文更加完整。立感謝口試委員. 林修葳老師、張士傑老師、陳思寬老師、盧秋玲老師，您. ‧ 國. 張元晨老師、顏錫銘老師、周冠男老師、盧敬植老. 學. 感謝博士班授課老師. 師、湛可南老師、郭維裕老師、溫偉任老師、江永裕老師、許振明老師！學生從您們的授課中汲取知識而厚實學問。. ‧. 感謝同學林楚彬、吳周燕的相伴！很懷念那段一起上課、煮東煮西、準備. y. Nat. sit. 學科考試及逛夜市的生活，謝謝你們！. n. al. er. io. 感謝許玉美學姊、翁胤哲學長、吳偉劭學長、劉淑華學姊、謝依婷學姊、. i n U. v. 詹淑惠學姊、富德、苡文、維中、晉吉、之寧、采彤助教曾經對我的幫忙！. Ch. engchi. 感謝周行ㄧ校長、羅時芳老師在我遭遇經濟拮据時，即時伸出援手幫我渡過難關，您們的恩德，學生銘記在心。博士班就讀期間遭遇父親驟逝讓我萬分悲痛，學習堅強面對人生的無常，把悲傷的心情轉化為前進的動力，這是最難學習的一門人生課程。謝謝爸爸在我過往人生中竭盡所能為我默默付出！謝謝媽媽傾盡心力悉心照顧我及我的女兒！謝謝婆婆、哥哥、嫂嫂、姊姊、妹妹、大姑、二姑當我的後盾!謝謝老公一路的陪伴與勉勵！辛苦你了！謝謝兒女帶給我歡樂！最後，謹以此文獻給幫助、關心我的人，謝謝你們！張美菁於政治大學博士班研究室 2015/7/2 i.

(3) 中文摘要【第一篇論文中文摘要】本文使用路透社央行干預匯市的新聞報導，探討哪些因素可以提高亞洲央行成功干預匯市的機率，研究期間為 2005 年 1 月至 2011 年 4 月。此研究期間涵蓋全球金融風暴和美國採行量化寬鬆政策，因此，亞洲貨幣在逐步對美元升值後發生大幅度的貶值。研究結果顯示印尼、馬來西亞、菲律賓、新加坡、台灣及泰國的央行採取逆風而行的策略是有效的干預方式，而且多個國家在同日干預匯市及第一日的干預會有較高成功的機率。. 政治大. 關鍵字：央行干預、金融風暴、亞洲貨幣、逆風而行. 立. ‧ 國. 學. 【第二篇論文中文摘要】. 本文透過不同的研究方法針對亞洲國家央行干預匯率市場的有效性再次驗. ‧. 證，研究期間為 2005 年 1 月至 2011 年 4 月，實證結果顯示亞洲國家在次貸風暴期間面臨美元升值的壓力，央行會採取賣美元的方式來干預匯市，但是這種. Nat. sit. y. 干預匯市的方式卻僅能減緩美元升值的趨勢，其中以印尼盾、新加坡元、新台. er. io. 幣紛紛對美元貶值較為明顯，而在次貸風暴發生之後，美國實施量化寬鬆政策. al. 造成亞洲國家卻面臨美元貶值的壓力，各國央行改採買美元的方式來干預匯市，. n. v i n Ch 但是此種干預匯市的方式也只造成美元緩慢貶值的趨勢，其中印尼盾、馬來西 engchi U 亞令吉、新加坡元、韓圜、泰銖分別對美元升值的趨勢較為明顯，此現象反應亞洲央行干預匯市是採取逆風而行的策略，雖然能降低匯率的波動，但無法改變匯率的升貶值趨勢。. 關鍵字：央行干預、次貸風暴、逆風而行. ii.

(4) 【第三篇論文中文摘要】本研究是在探討印度、印尼、南韓、馬來西亞、新加坡、泰國及台灣央行透過干預匯率市場，對其國家的公司在美國發行存託憑證折溢價的影響，研究期間為2005年1月至2011年4月。研究結果顯示央行對匯市干預造成的變動，確實會影響到該國公司在美國發行的存託憑證產生折價的情形。另外，亞洲央行使用買美元干預匯市的作法會對該國公司在美國發行的存託憑證產生溢價，而央行透過賣美元干預匯市的作法會對該國公司在美國發行的存託憑證產生折價的影響，但是由於樣本資料的限制，其效果在統計上並不顯著。由公司層面的分析可以看出央行透過賣美元來干預匯市對其國家的公司在美國發行的存託憑證會有明顯產生折價的影響。. 治政大關鍵字：匯率市場、存託憑證、折價、溢價立 ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. iii. i n U. v.

(5) 英文摘要. 【第一篇論文英文摘要】 Using Reuters’ news reports on central bank interventions, we investigate the factors that increase the odds of intervention success by Asian central banks in the foreign exchange market from January 2005 to April 2011. This period coincides with the global credit crisis and quantitative easing policy, which have engendered a sharp depreciation followed by a gradual appreciation of Asian currencies. The results show that leaning-against-the-wind intervention strategies are effective in Indonesia, Malaysia, Philippines, Singapore, Taiwan, and Thailand. We also find. 政治大. that joint and first day interventions are associated with higher odds of effective. 立. bank. interventions;. Credit. Asian. currencies;. ‧. Leaning-against-the-wind. crisis;. Nat. sit. 【第二篇論文英文摘要】. y. Central. 學. Keywords:. ‧ 國. intervention.. er. io. This paper examines the effectiveness of central bank interventions in the. al. v i n C h in the foreignUexchange markets by selling that the central banks in Asia intervene engchi n. foreign exchange market from January 2005 to April 2011 in Asia. The results show. U.S. dollars to prevent severe depreciation of local currencies during the global credit crisis. However, central bankers can only slow down the trend of depreciation of local currencies against U.S. dollar. The currencies apparently depreciate against. U.S. dollar in Indonesia, Singapore, and Taiwan. After the global credit crisis, Asian countries confront appreciations of local currencies due to the US quantitative easing policy. The central banks in Asia intervene by purchasing U.S. dollars in the foreign exchange market. Nevertheless, intervention strategies slowly reduce U.S. dollar depreciations. The foreign exchange rate apparently appreciate against U.S. dollar in India, Malaysia, Singapore, South Korea, and Thailand. Results show that Asian central banks adopt leaning-against-the-wind intervention strategies during iv.

(6) the sample period. Their interventions in the foreign exchange market can only reduce fluctuations in the foreign exchange rate, but fail to reverse the trend of Asian exchange rates.. Keywords: Central bank interventions; Credit crisis; Leaning-against-the-wind. 【第三篇論文英文摘要】 This paper examines whether Asian central bank interventions in the foreign exchange market affect the discount or premium of American Depositary Receipt (ADR) of Asian companies from January 2005 to April 2011. The sample consists of. 治政大 discounts of companies in results show that central bank interventions increase ADR 立 Asian countries. In addition, interventions by purchasing U.S. dollars result in higher. companies from Indian, Indonesia, South Korea, Malaysia, Singapore. Empirical. ‧ 國. 學. ADR premiums, and the strategies of selling U.S. dollars affect ADR discounts. Though some of the empirical results are not statistically significant due to limited. ‧. sample size, results based on individual firms show that selling USD interventions by. sit. y. Nat. Asian central banks have a significant impact on their ADR discounts.. al. n. premium. er. io. Keywords: Foreign exchange market; American Depositary Receipt; discount;. Ch. engchi. v. i n U. v.

(7) 目錄前言................................................................................................................................ 1. 【第一篇論文】 Foreign Exchange Intervention in Asian Countries: What Determine the Odds of Success during the Credit Crisis? .......................................................................... 3 1. Introduction ................................................................................................................ 3 2. Literature Review....................................................................................................... 6 2.1 Intervention channels ............................................................................... 6 2.2 Developed and developing markets ......................................................... 7. 政治大. 3. Data .......................................................................................................................... 10. 立. 4. Methodology and Empirical Results ........................................................................ 19. ‧ 國. 學. 4.1 Methodology .......................................................................................... 19 4.1.1 Defining intervention success ..................................................... 19. ‧. 4.1.2 Testing the forecast value of intervention ................................... 20 4.1.3 Determining the odds of successful intervention ........................ 21. y. Nat. sit. 4.1.4 Intervention reaction function ..................................................... 23. er. io. 4.2 Empirical Results ................................................................................... 25. al. v i n 4.2.2 Results forC the intervention reaction h e n g c h i U function............................. 31 n. 4.2.1 Results for the forecast value of intervention ............................. 25. 4.2.3 Factors influencing the odds of successful intervention ............. 35 5. Robustness check ..................................................................................................... 46 6. Conclusions .............................................................................................................. 51 Appendix ...................................................................................................................... 53. Table A1 Sample News Reports ............................................................... 53 References .................................................................................................................... 55. vi.

(8) 【第二篇論文】亞洲央行干預外匯市場有效性的再驗證 ................................................................. 60 1. 緒論......................................................................................................................... 60 2. 文獻回顧................................................................................................................. 62 3. 資料與研究方法..................................................................................................... 66 3.1 資料....................................................................................................... 66 3.2 研究方法............................................................................................... 70 4. 實證結果................................................................................................................. 70 4.1 邏輯斯迴歸模型估計亞洲國家(買美元干預匯市的情況) ................ 70 4.2 未採用配對方法的結果(買美元干預匯市的情況) ............................ 84. 治政大 4.4 亞洲各國邏輯斯迴歸模型估計(賣美元干預匯市的情況) ................ 89 立 4.5 未採用配對方法的實證結果(賣美元干預匯市的情形) ................ 100 4.3 最鄰近配對法(買美元干預匯市的情況) ............................................ 87. ‧ 國. 學. 4.6 採用最鄰近配對法的結果（賣美元干預匯市的情況）................. 103 5. 結論....................................................................................................................... 105. ‧. Nat. sit. 【第三篇論文】. y. 參考文獻.................................................................................................................... 107. er. io. 亞洲央行干預外匯市場對美國存託憑證價差的影響 ........................................... 111. al. 1. 緒論....................................................................................................................... 111. n. v i n Ch 2. 文獻回顧............................................................................................................... 112 engchi U 3. 資料與研究方法................................................................................................... 114 3.1 資料..................................................................................................... 114 3.2 研究方法............................................................................................. 115 4. 實證結果............................................................................................................... 115 4.1 買美元的新聞報導............................................................................. 119 4.2 賣美元的新聞報導............................................................................. 131 5. 結論....................................................................................................................... 141 參考文獻.................................................................................................................... 142. vii.

(9) 表目錄【第一篇論文】 Table 1 Central Bank Intervention News Distribution ............................................. 13 Table 2 Asian Central Bank Intervention and Alternative Success Criteria (Firm News Category) ................................................................................. 27 Table 3 Results of Asian Countries’ Intervention Reaction Functions Using Firm News Category ............................................................................................ 32 Table 4 The Effects of Individual Variables on the Odds Ratio Using Firm News Category ...................................................................................................... 37 Table 5 Joint Significance of Variables on the Odds Ratio Using Firm News. 政治大. Category ...................................................................................................... 42. 立. Table 6 Asian Central Bank Intervention and Alternative Success Criteria (Firm. 【第二篇論文】. ‧. ‧ 國. 學. and Suspected News Category) ................................................................... 47. 表一、美國進口統計表.............................................................................................. 61. Nat. sit. y. 表二、匯率變化的敘述性統計.................................................................................. 68. er. io. 表三、邏輯斯迴歸模型估計亞洲各國(買美元干預匯市) ....................................... 74. al. 表四、參數間穩定性的 Wald test（買美元干預匯市） .......................................... 79. n. v i n Ch 表五、亞洲各國未採用配對的結果(買美元干預匯市) ........................................... 85 engchi U 表六、亞洲各國最鄰近配對法之統計結果(買美元干預匯市) ............................... 88 表七、亞洲各國邏輯斯迴歸模型估計(賣美元干預匯市) ....................................... 91 表八、亞洲各國參數間穩定性的 Wald test（賣美元干預匯市） .......................... 96 表九、亞洲各國未採用配對的結果(賣美元干預匯市) ......................................... 101 表十、亞洲各國最鄰近配對法之統計結果(賣美元干預匯市) ............................. 104. viii.

(10) 【第三篇論文】表一、亞洲樣本國家各變數的敘述性統計表........................................................ 117 表二、亞洲樣本國家的迴歸結果（買美元）........................................................ 121 表三、各公司的迴歸結果（買美元）.................................................................... 126 表四、亞洲樣本國家的迴歸結果（賣美元）........................................................ 132 表五、各公司的迴歸結果（賣美元）.................................................................... 136. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. ix. i n U. v.

(11) 圖目錄【第一篇論文】 Figure 1 Plots of Asian Currencies viz-a-viz US Dollar and Intervention Dummy .... 15. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. x. i n U. v.

(12) 前言本篇論文是在探討亞洲央行干預外匯市場的有效性及對美國存託憑證價差的影響，論文是由三篇研究組合而成，分別說明如下：第一篇論文為「Foreign Exchange Intervention in Asian Countries: What Determine the Odds of Success during the Credit Crisis?」，本篇研究是依據 Humpage（1999）的架構，探討哪些因素可以提高亞洲央行成功干預匯市的機率，央行干預匯市的資料整理自路透社的新聞報導，研究期間為 2005 年 1 月至 2011 年 4 月，此研究期間涵. 政治大化皆會採取逆風而行的干預方式，而此策略在印尼、馬來西亞、菲律賓、新加坡、立. 蓋全球金融風暴和美國採行量化寬鬆政策，因此，亞洲央行面臨匯率市場巨盪的變. 台灣及泰國是有效的干預方式，且多個國家在同日干預匯市及第一日的干預會有較. ‧ 國. 學. 高成功的機率。. ‧. 第二篇論文「亞洲央行干預外匯市場有效性的再驗證」，採用與第一篇論文相同. y. Nat. 的資料，但使用不同的研究方法再次驗證亞洲央行干預匯市的有效性。本研究是根. er. io. sit. 據 Fatum and Hutchison（2010）的架構，實證結果顯示亞洲國家在次貸風暴期間面臨其貨幣兌美元升值的壓力時，央行會採取賣美元的方式來干預匯率市場，但是這. al. n. v i n 種干預匯市的方式卻僅能減緩美元升值的趨勢，而在次貸風暴發生之後，美國實施 Ch engchi U 量化寬鬆政策造成亞洲國家卻面臨美元貶值的壓力，各國央行改採買美元的方式來. 干預匯市，但是此種干預匯市的方式也只造成美元緩慢貶值的趨勢，此篇研究呈現亞洲央行干預匯市是採取逆風而行（Leaning-against-the-wind）的策略，雖然能降低匯率的波動，但無法改變匯率的升貶值趨勢。第三篇論文「亞洲央行干預外匯市場對美國存託憑證價差的影響」，本研究是在探討央行透過干預匯率市場，對該國公司在美國發行存託憑證折溢價的影響。研究結果顯示央行對匯市干預造成的變動，確實會影響到其國家的公司在美國發行的存託憑證產生折價的情形。另外，亞洲央行使用買美元干預匯市的方法下，匯率的變. 1.

(13) 化會使該國公司在美國發行的存託憑證產生溢價，而央行透過賣美元干預匯市的作法，會使該國公司在美國發行的存託憑證產生折價，但是其效果皆不顯著。亞洲央行為提昇其國家產品在全球貿易競爭激烈的環境下增加競爭力，會在匯率市場適時的採取買美元的干預方式，公司營運因此面臨到匯率風險，公司面臨其國家央行頻繁干預匯市下的避險決策為何？是否會與央行不干預匯市情況下有差異是未來可以研究的有趣議題。. 立. 政治大. ‧. ‧ 國. 學. n. er. io. sit. y. Nat. al. Ch. engchi. 2. i n U. v.

(14) Foreign Exchange Intervention in Asian Countries: What Determine the Odds of Success during the Credit Crisis? 1.. Introduction The voluminous literature on foreign exchange interventions has traditionally. focused on advanced economies (Dominguez and Frankel, 1990, Leahy, 1995, Dominguez, 1998, 2006, Dominguez et al, 2013, Chang and Taylor, 1998, Humpage, 1999, 2003, Baillie et al, 2000, Bordo et al, 2012a, 2012b, Ito, 2005, Ito and Yabu, 2007, Fatum and Hutchison, 2002, 2003, 2006, Kearns and Rigobon, 2005, Fatum, 2008, and. 政治大. Neely, 2011). Nevertheless, in recent years the empirical stylized facts of foreign. 立. exchange interventions have changed in several important ways. Major central banks in. ‧ 國. 學. advanced economies, particularly those in the G7, have intervened less frequently post-1996.1 On the other hand, many central banks in emerging markets, especially in. ‧. Asia, have experienced large balance sheet expansions through the accumulation of. Nat. sit. y. foreign reserves and accompanied by more frequent foreign exchange interventions.. n. al. er. io. Cook and Yetman (2012) document that the asset expansions in Asian central banks’. v. balance sheets have been caused by a build-up of foreign exchange reserves.2 Although. Ch. engchi. i n U. we have witnessed a recent rising trend of central bank engagement in competitive exchange rate management in emerging markets, evidence of intervention efficacy remains sparse. This paper is the first to demonstrate the use of Reuters’ news reports as a proxy for Asian central bank interventions, providing a contribution to empirically determine 1. 2. Neely (2011) highlights the mid-1990s as the start of the post-intervention era. Authorities of developed countries doubted of the efficacy of foreign exchange intervention operations and stopped intervening or intervened less frequently in the foreign exchange market. For example, the Bank of Canada stopped intervening in 1988, while the European Central Bank has intervened very rarely since 1999. Foreign exchange reserves made up more than 80% of central bank assets for eight Asian countries, comprising China, Hong Kong, India, Malaysia, Philippines, Singapore, South Korea, and Thailand. While Indonesia may be the only exception, its foreign reserves have constituted more than half of its central bank assets since 2006.. 3.

(15) whether such interventions are effective using data from January 2005 to April 2011. This sample period is chosen, because there are extensive intervention reports on currency markets by Asian central banks that coincide with the global credit crisis and U.S. quantitative easing. International funds have rapidly flowed in and out of Asian countries, giving rise to large fluctuations of Asian currencies during that period. There is a large body of work that has focused on rationalizing the causes and consequences of foreign currency reserve build-up in emerging Asian economies especially since 1998 (Bird and Rajan 2003, Aizenman and Marion, 2003). These studies. 政治大 that this sustained reserve立 build-up has been motivated by a desire to keep currencies have concluded that Asia holds more than enough reserves as a financial safeguard, and. ‧ 國. 學. from appreciating significantly. In addition, during the global financial crisis, part of the large influx of international funds into Asia has been driven by the U.S. Federal Reserve’s. ‧. quantitative easing policy to stimulate the U.S. economy after the global credit crisis,. sit. y. Nat. which resulted in a long-term downward trend of the US dollar viz-á-viz Asian currencies.. n. al. er. io. In the face of appreciation pressure on their currencies, several Asian central banks have. i n U. v. frequently intervened in the foreign exchange market to stabilize their exchange rates. Ch. engchi. with a view of promoting and maintaining their export competitiveness. This paper uses the analytical framework of Humpage (1999) who shows that the U.S. intervention against German Deutsche Mark and Japanese Yen possesses a forecast value under a weak leaning-against-the-wind criterion. This framework bodes well with the coordination channel of intervention. Under this channel, intervention might be important in coordinating the expectations of market participants. Sarno and Taylor (2001), Taylor (2005), and Reitz and Taylor (2008) emphasizes the importance of this coordination channel in communicating the authorities’ belief that the exchange rate is deviating substantially from its long-run value. In light of the appreciating Asian 4.

(16) currencies against the US dollar recently, many authorities believe that this undue appreciation pressure does not reflect the true external value of their currencies, thus potentially hurting their exports. Following Fatum and Hutchison (2002), we use news reports on Asian central banks’ interventions to analyze the effectiveness of exchange rate management by central banks in India, Indonesia, Malaysia, Philippines, Singapore, South Korea, Taiwan, and Thailand. The method of Humpage (1999) also permits the success of intervention operations, as proxied via these news reports to be judged within a probabilistic framework. This. 政治大 operation of the coordination channel, because it relies on traders recognizing the 立. probabilistic framework squares well with the element of market psychology in the. authorities’ intervention operations as a coordinating signal implied by such intervention. ‧ 國. 學. news. Evidence, supporting the coordination channel, leads to the probability of. sit. y. Nat. intervention.. ‧. intervention success to be represented as an increasing function of the degree of. io. er. Our results can be summarized as follows. First, there is a strong correlation between Reuters’ intervention reports and the changes of exchange rates for Indonesia,. al. n. v i n C hTaiwan, and Thailand. Malaysia, Philippines, Singapore, This correlation, however, is engchi U weaker in India and South Korea. Second, the average intervention reports in a given month, which proxy for intervention frequency, are found to affect the correlation between intervention and exchange rate movements. Third, there is evidence of. correlation asymmetry in Asian countries - the correlations of purchasing USD intervention are greater than those of selling USD for India, Malaysia, Philippines, Singapore, South Korea, Taiwan, and Thailand. Such a pattern is not observed for Indonesia. Fourth, when controlling for factors that explain the effectiveness of intervention by Asian central banks, we find that leaning-against-the-wind strategies 5.

(17) increase the odds of intervention by central banks in India, Indonesia, Malaysia, Philippines, Singapore, South Korea, Taiwan, and Thailand. The effect of intervention is found to be more significant for joint interventions and first-day interventions. The remainder of the paper is organized as follows. Section 2 provides a literature review focusing on intervention channels and those studies conducted on both developed and developing countries. Section 3 describes the data. Section 4 discusses the methodology and provides the empirical results. Section 5 presents robustness checks. Section 6 concludes and discusses the implications of our results.. 2. Literature Review. 立. 政治大. ‧ 國. 學. 2.1 Intervention channels. Official intervention in the foreign exchange market occurs when central bank. ‧. authorities buy or sell the US dollar, normally against their own currencies, in order to. y. Nat. io. sit. affect the level of their local exchange rate. There are three channels through which. n. al. er. official intervention might affect the foreign exchange market indirectly: the portfolio. i n U. v. balance channel, the signaling channel, and the coordination channel.. Ch. engchi. The portfolio balance channel alludes to sterilized intervention changing the relative supplies of bonds denominated in different currencies. Given that bonds from different countries are imperfect substitutes and that traders’ demand for bonds from a particular country is determined by their given rate of return, then the relative returns on certain bonds depend on the relative quantities of those bonds to other currency-denominated bonds. Researchers, however, are skeptical that the portfolio balance channel can explain the effectiveness of interventions, because interventions are usually way too small to significantly change the relative quantities of bonds (Humpage, 1999). 6.

(18) The signaling channel suggests that official intervention signals information about future monetary policy. The literature on intervention has not been favourable toward this channel. The effect of intervention on monetary policy has been perverse in a sample of countries examined by Lewis (1995), and Kaminsky and Lewis (1996). In the best possible scenario, Fatum and Hutchison (1999) fail to find that intervention exerts any effect on federal funds future rates. More importantly, the plausibility of the signaling story is in doubt, given that both monetary policy and exchange rate policy are generally managed by two different institutions in an economy. For example, the U.S. Treasury is. 政治大 communicate anticipated monetary policy decided by the FOMC. 立. responsible for the value of the USD and it is unlikely that its intervention signal would. ‧ 國. 學. The coordination channel, which is the most plausible channel of the three, suggests that intervention might be important in coordinating the expectations of market. ‧. participants. This channel has received increasingly greater importance in the literature,. sit. y. Nat. because intervention communicates the authorities’ belief about the exchange rate being. n. al. er. io. misaligned from its long-run value. During such periods, individual traders are subject to. i n U. v. greater risk if they invest capital in hopes of a return to the long-run equilibrium, given. Ch. engchi. that markets have a tendency to remain irrational and the exchange rate can persist to deviate from its long-run equilibrium level for an extended time period. Be that as it may, foreign exchange intervention can facilitate the convergence in the expectations of market participants and lead traders so as to align the exchange rate closer to its long-run equilibrium. It is this channel that we draw upon in this study for assessing the secondary effects of intervention.. 2.2 Developed and developing markets Over the years, there have been radical changes in the conduct of foreign exchange 7.

(19) intervention. The U.S. monetary authorities used foreign exchange intervention to counter disorderly market conditions in 1980, although interventions became less frequent by the late-1990. Empirical works on the effectiveness of intervention have produced very mixed results and generally rely on data from developed countries, specifically the G7 countries (Dominguez and Frankel, 1990, Dominguez, 1998, Humpage, 1999, Fatum and Hutchison, 2002, 2003, 2006, Kearns and Rigobon, 2005). Dominguez and Frankel (1990) show that central bank interventions in Germany and the U.S. are effective in influencing exchange rate volatilities. Dominguez (1998) also presents that interventions by the U.S.,. 政治大 period. Humpage (1999) notes that intervention possesses some forecast value when the 立. Germany, and Japan tend to increase exchange rate volatilities during the 1977-1994. U.S. Treasury undertakes a weak leaning-against-the-wind policy.. ‧ 國. 學. The study of Fatum and Hutchison (2002) is the closest to ours in the sense that they. ‧. utilize news reports from the Wall Street Journal to investigate the effects of intervention. sit. y. Nat. news on the Euro. Using four different categories of intervention news of the European. n. al. er. io. Central Bank (ECB), they find that official statements denying ECB intervention or news. i n U. v. that question the efficacy of intervention do have a significant and negative impact on the. Ch. engchi. value of the Euro. On the other hand, firm reports of intervention, as well as rumors and speculation of intervention in support of the Euro lead to short-term Euro appreciation. Fatum and Hutchison (2006) rely on a non-parametric sign test and a matched-sample test to document strong evidence that intervention by the Bank of Japan systemically affects its short-run exchange rate. Finally, Kearns and Rigobon (2005) show that interventions by the Reserve Bank of Australia and the Bank of Japan have significant effects on the Australian dollar and Japanese Yen. In contrast to the literature of foreign exchange intervention in developed economies, tests of the effectiveness of interventions in emerging markets are limited (Tapia and 8.

(20) Tokman, 2004, Guimaraes and Karacadag, 2004, Disyatat and Galati, 2007, Bernanke, 2010, Chen, 2011, Domaç and Mendoza, 2004, Hua and Gau, 2006, Rincon and Toro, 2011, Menkhoff, 2013). Much of the literature on foreign exchange intervention in emerging markets has featured markets in the western hemisphere. Tapia and Tokman (2004) find that interventions by the Central Bank of Chile in the foreign exchange market affect both the level and volatility of the Chilean Peso during the Asian and Argentinian crises. Guimaraes and Karacadag (2004) use a modified generalized autoregressive conditional heteroskedasticity (GARCH) model to show that sales of USD. 政治大 effect on the Mexican Peso, although purchases do not exert any effect. Domaç and 立 intervention increases volatility both in the short run and long run and have a significant. Mendoza (2004) show that central banks in Mexico and Turkey can effectively influence. ‧ 國. 學. the directions of their currencies. Rincon and Toro (2011) examine the efficacy of. ‧. intervention when combined with capital controls in Colombia during the period 1993 to. sit. y. Nat. 2010, presenting that their simultaneous use tend to support intervention effectiveness by. io. n. al. er. augmenting the exchange rate trend without increasing volatility.. i n U. v. Intervention studies in Asia are more limited to the case of India. Pattanaik and. Ch. engchi. Sahoo (2001) examine the exchange rate policy in India for the period 1995-2003 under the objectives of stabilizing the exchange rate and keeping market conditions orderly. However, the assessment of intervention effectiveness is marred by the fact that these two goals are also achieved via monetary policy and capital controls. The intervention reaction function reveals that India’s central bank reacts to volatility in the foreign exchange market but not to misalignments. Goral and Arora (2010), using intervention data from the Reserve Bank of India, show that interventions are effective in reducing volatilities of the Indian Rupee. As is clear from the limited number of studies on Asian foreign exchange intervention, our paper fills a very important gap in the literature by 9.

(21) using Reuters’ news reports to study the efficacy of intervention in eight economies.. 3. Data This paper focuses on eight economies: India, Indonesia, Malaysia, Philippines, Singapore, South Korea, Taiwan, and Thailand. They are emerging markets in Asia. Our sample period covers January 2005 to April 2011. Daily exchange rate data are obtained from Bloomberg. The data are the composite rates by all banks offering traded prices to Bloomberg. We retrieved Taiwan’s exchange rates (TWD) data using the opening and closing rates from Taipei Forex, Inc. Bloomberg also has the opening and closing. 政治大. exchange rates during the Hong Kong’s trading hours of 8:00 a.m. and 4:59 p.m.,. 立. respectively for Indonesia, Malaysia, Philippines, Singapore, South Korea, and Thailand.. ‧ 國. 學. In addition, we obtain India’s quotes for Mumbai’s trading hours from 9:00 a.m. to 4:59 p.m. We see that exchange rates follow a similar general pattern for Asian countries in. ‧. that they generally appreciated relative to the US dollar, at times rapidly, until March. y. Nat. io. sit. 2008, depreciated very sharply until March 2009, and then appreciated gradually after the. er. implementation of quantitative easing by the U.S. Federal Reserve.. al. n. v i n C hto extract information We use Reuters’ news reports about central bank interventions engchi U. in the foreign exchange market. These news reports consist of the date and time of a reported intervention. The use of news reports as proxies for central bank interventions is featured in past studies (Peiers, 1997, Chang and Taylor, 1998, Fatum and Hutchison, 2002, Dominguez, 2003, Sapp, 2002, Fischer, 2006). The accuracy of intervention reports has been analyzed by Fischer (2006) for the Swiss National Bank. When comparing actual intervention data with intervention reports from Reuters, he shows that intervention reports in the intraday interval often lag behind actual interventions, but interventions reported by Reuters at the daily interval are fairly 10.

(22) accurate with no false reporting. The high percentage of accuracy of actual intervention reports from Reuters indicates that information conveyed in these reports is accurate and provides a high degree of data integrity assurance. This paper uses Reuters’ news reports to proxy for intervention activities in the eight Asian economies. Our classifications of news reports are similar to that of Fatum and Hutchison (2002). Intervention news reports are classified as firm, suspected, supported, and neutral. A report is classified as a firm report when the news clearly state that the central bank intervened in the currency markets. This is synonymous with category D of Fatum and. 政治大 on official intervention in 立the foreign exchange market. This is synonymous with. Hutchison (2002). A report is classified as a suspected report when the news cast doubt. ‧ 國. 學. category A of Fatum and Hutchison (2002). A report is classified as a supported report when central bank or government officials provide statements that show support for. ‧. intervention in the currency market (including suggestions that intervention is a. sit. y. Nat. possibility). This is equivalent to category B of Fatum and Hutchison (2002). A report is. al. er. io. classified as a neutral report when central bank or government officials express neutral. v. n. opinion on intervention activities. Firm reports are more certain on central banks’. Ch. engchi. i n U. intervention activities and suspected reports are possible interventions with doubt. These two categories of reports specifically distinguish intervention in the buying and selling of US dollars. The supported and neutral categories are news reports of central bank or government officials’ verbal interventions on the foreign exchange market, and they do not highlight the difference in the intervention. Table A1 in the Appendix provides examples of the news reports and their categories. Table 1 presents the distribution of intervention reports of the eight Asian countries 11.

(23) categorized into firm, suspected, supported, and neutral reports. Panel (a) shows the number of intervention reports in two sub-sample periods. The first sample period (2005-2007) has a fewer reports for firm and suspected news than that in the second sample period (2008-2011). The second sample period coincides with the subprime crisis, which led to the introduction of the quantitative easing policy adopted by the US Federal Reserve. It is not surprising, therefore, that the majority of intervention reports happened in this period. Panel (b) shows the number of days for which news on central bank interventions are reported on the Asian countries. There are more intervention reports for. 政治大 heavily in the foreign exchange market during the second sample period and the number 立 Taiwan, South Korea, and India. We see that central banks in these countries intervened. of days with intervention reports mainly concentrate in the firm report category, except. ‧ 國. 學. for those of India and Singapore, which have more suspected intervention reports.. ‧. n. er. io. sit. y. Nat. al. Ch. engchi. 12. i n U. v.

(24) Table 1 Central Bank Intervention News Distribution Panel (a) Number of news counts Firm Country. Suspected. Neutral. 2005- 2008- Total 2005- 2008- Total 2005- 2008- Total 2005- 2008- Total 2007 2011. 2007 2011. 2007 2011. 2007 2011. 48. 81. 129. 66. 89. 155. 4. 4. 8. 1. 2. 3. Indonesia. 3. 120. 123. 0. 26. 26. 3. 21. 24. 0. 11. 11. Malaysia. 2. 29. 31. 2. 6. 8. 1. 2. 3. 1. 0. 1. Philippines. 10. 90. 100. 3. 26. 29. 0. 7. 7. 0. 2. 2. Singapore. 1. 17. 18. South Korea. 8. 186. 194. Taiwan. 11. 216. 227. 政治大. Thailand. 4. 124. 128. 立. 4. 17. 21. 1. 0. 1. 0. 1. 1. 3. 56. 59. 13. 48. 61. 0. 14. 14. 2. 21. 23. 0. 17. 17. 0. 7. 7. 6. 28. 34. 0. 30. 30. 0. 7. 7. 學. Malaysia. 79. 99. 7. 3. 95. 23. 18. 9. 24. 7. 3. 1. 7. 2. v1. 1. 54. 10. 16. 7. 24. 7. 82. 28. n. Singapore. Neutral. io. Philippines. Supported. Nat. Indonesia. Suspected. y. India. Firm. 19. al. 16. Ch. South Korea. 133. Taiwan. 201. e n g5420c h i. Thailand. 95. 33. er. Country. sit. Panel (b) Number of days with news reports. ‧. ‧ 國. India. Supported. i n U. Note: A report is classified as a firm report when the news clearly state that the central bank intervened in currency markets. A report is classified as a suspected report when the news cast doubt on official intervention in the foreign exchange market. A report is classified as a supported report when central bank or government officials provide statements that show support for intervention in the currency market (including suggestions that intervention is a possibility). A report is classified as a neutral report when central bank or government officials express a neutral opinion on intervention activities.. 13.

(25) Figures 1(a) to (h) plot the value of the US dollar against the Asian currencies of all eight countries. The plot is superimposed against the intervention dummy, which indicates purchases (+1) and sales (-1) of the US dollar. The figures show that the central banks of India, Indonesia, Philippines, South Korea, Taiwan, and Thailand intervened constantly to keep their exchange rates from falling against the US dollar from 2007 to 2008 and after 2009. The various countries also responded differently to the rise in the value of the US dollar from 2008 to 2009. Some countries allowed their local currency to appreciate more against the US dollar, while others intervened massively to hold their. 政治大 some countries monitor the region’s exchange rates and attempt to keep the relative value 立 exchange rate to fluctuate within a limited range. This is consistent with the notions that. of their currencies in line with the value of selected regional currencies.. ‧ 國. 學. It is clear from these figures that most central banks in Asian countries tend to. ‧. purchase (sell) US dollars when the US dollar depreciates (appreciates), which indicates. sit. y. Nat. that these countries are attempting to counter large changes in the value of their domestic. n. al. er. io. currencies in the sample period. Another noteworthy observation is that in several of. i n U. v. these Asian countries there are many episodes of intervention occurring on the same day,. Ch. engchi. implying persistence by the authorities in coordinating with market participants to influence the level of the exchange rate.. 14.

(26) Figure 1. Plots of Asian Currencies viz-a-viz US Dollar and Intervention Dummy. (USD/ INR) 52. 1. 50 48 46 0 44 42 40 38. -1. 2005. 2006. 2007. 2008. 2010. 2011. 政治大. Intervene(RHS). 立. 2009 SPM(LHS). Figure 1(a): India. ‧ 國. 學. Intervene is a dummy variable equal to 1 when the central bank of India buys US dollars from the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against the Indian Rupee.. ‧ y. sit. al. n. 11560 11060 10560. 1. er. io. 12060. Nat. (USD/ IDR). Ch. engchi. i n U. v. 0. 10060 9560 9060 8560. -1. 2005. 2006. 2007. 2008. Intervene(RHS). 2009. 2010. 2011. SPM(LHS). Figure 1(b): Indonesia Intervene is a dummy variable equal to 1 when the central bank of Indonesia buys US dollars from the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against the Indonesian Rupiah.. 15.

(27) (USD/ MYR) 1 3.9 3.7 3.5. 0. 3.3 3.1 2.9. -1. 2005. 2006. 2007. 2008. 2009. Intervene(RHS). 立. 2010. 2011. SPM(LHS). 治政 Figure 1(c): Malaysia 大. Intervene is a dummy variable equal to 1 when the central bank of Malaysia buys US dollars from. ‧ 國. 學. the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against the Malaysian Ringgit.. y. sit. io. 54. al. n. 52 50 48. 1. er. 56. ‧. 58. Nat. (USD/ PHP). Ch. engchi. i n U. v. 0. 46 44 42 40. -1. 2005. 2006. 2007. 2008. Intervene(RHS). Figure 1(d):. 2009. 2010. 2011. SPM(LHS). Philippines. Intervene is a dummy variable equal to 1 when the central bank of Philippines buys US dollars from the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against the Philippine Peso.. 16.

(28) (USD/ SGD) 1.8. 1. 1.7 1.6 1.5. 0. 1.4 1.3 1.2. -1. 2005. 2006. 2007. 2008. Intervene(RHS). 立. 2009. 2010. 2011. SPM(LHS). 治政 Figure 1(e): Singapore 大. Intervene is a dummy variable equal to 1 when the central bank of Singapore buys US. ‧ 國. 學. dollars from the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against Singapore dollars.. ‧ sit. al. n. 1400 1300. 1. er. io. 1500. y. Nat. (USD/ KRW). Ch. engchi. i n U. v. 0. 1200 1100 1000 900. -1. 2005. 2006. 2007. 2008. Intervene(RHS). 2009. 2010. 2011. SPM(LHS). Figure 1(f): South Korea Intervene is a dummy variable equal to 1 when the central bank of South Korea buys US dollars from the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against the Korean Won.. 17.

(29) (USD/ TWD) 36. 1. 35 34 33 32. 0. 31 30 29 28. -1. 2005. 2006. 2007. 2008. 2009. Intervene(RHS). 立. 2010. 2011. SPM(LHS). 治政 Figure 1(g): Taiwan 大. Intervene is a dummy variable equal to 1 when the central bank of Taiwan buys US dollars. ‧ 國. 學. from the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against NTD.. ‧ y. sit. io. al. n. 41 39 37. 1. er. 43. Nat. (USD/ THB). Ch. engchi. i n U. v. 0. 35 33 31 29. -1. 2005. 2006. 2007. 2008. Intervene(RHS). 2009. 2010. 2011. SPM(LHS). Figure 1(h): Thailand Intervene is a dummy variable equal to 1 when the central bank of Thailand buys US dollars from the exchange market on certain dates, equal to -1 when the bank sells US dollars, or equal to 0 otherwise. SPM is the daily closing price of USD against the Thai Baht.. 18.

(30) 4. Methodology and Empirical Results 4.1 Methodology 4.1.1 Defining intervention success Following Humpage (1999), we specify three success criteria, which include a broad criterion and two component criteria. We treat each criterion as an official prediction of exchange rate movements and test whether intervention has forecast value using a nonparametric test proposed by Merton (1981) and Henriksson and Merton (1981).. 政治大 𝑖𝑓 𝐼 > 0 𝑎𝑛𝑑 ∆𝑆 > 0, 𝑜𝑟 ∆𝑆 > ∆𝑆𝐴𝑀 , 𝑜𝑟 1 {立 ={ 𝑖𝑓 𝐼 < 0 𝑎𝑛𝑑 ∆𝑆 < 0, 𝑜𝑟 ∆𝑆 < ∆𝑆𝐴𝑀 ;. Criterion 1 in the following equation represents the broad measure of success: 𝑡. 𝑡. 𝑡. 𝑡. 𝑡. 𝑡. 𝑡. 𝑡. 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.. (1). 學. ‧ 國. 𝑊𝑡. ‧. 𝑤𝑡 equals one which implies success if intervention purchases (sales) of US dollars are associated either with dollar appreciations (depreciations) at time t, or with smaller dollar. y. Nat. er. io. sit. depreciations (appreciations) when comparing ∆𝑆𝑡 with ∆𝑆𝐴𝑀𝑡 . I t designates official intervention at time t with positive (negative) values being purchases (sales) of US dollar. al. n. v i n C 𝑡h, where SAM and SPM intervention; ∆𝑆𝑡 = 𝑆𝑃𝑀𝑡 − 𝑆𝐴𝑀 are, respectively, morning and engchi U. afternoon exchange rates expressed as foreign currency units per dollar, and ∆𝑆𝐴𝑀𝑡 = 𝑆𝐴𝑀𝑡 − 𝑆𝐴𝑀𝑡−1. The morning-to-closing analysis is used to capture short-lived market. response to intervention. To examine the nature of the response, we break the general success criterion (Equation (1)) into its two component parts. The first component of the general criterion (criterion 2) counts a success if 𝑤1𝑡 equals one, i.e. intervention by Asian countries is successful if a dollar appreciation (depreciation) accompanies an intervention purchase (sale) of US dollars. 19.

(31) 𝑤1𝑡 = {. 𝑖𝑓 𝐼𝑡 > 0 𝑎𝑛𝑑 ∆𝑆𝑡 > 0, 𝑜𝑟 𝑖𝑓 𝐼𝑡 < 0 𝑎𝑛𝑑 ∆𝑆𝑡 < 0; 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.. 1{. (2). The second component of the general criterion (criterion 3), which is consistent with “leaning-against-the-wind” strategy, counts a success if 𝑤2𝑡 equals one, i.e. an intervention purchase (sale) of US dollars accompanies a smaller dollar depreciation (appreciation). 𝑖𝑓 𝐼𝑡 > 0 𝑎𝑛𝑑 ∆𝑆𝑡 > ∆𝑆𝐴𝑀𝑡 , 𝑔𝑖𝑣𝑒𝑛 ∆𝑆𝑡 𝑎𝑛𝑑 ∆𝑆𝐴𝑀𝑡 < 0, 𝑜𝑟 1{ 𝑤2𝑡 = { 𝑖𝑓 𝐼𝑡 < 0 𝑎𝑛𝑑 ∆𝑆𝑡 < ∆𝑆𝐴𝑀𝑡 , 𝑔𝑖𝑣𝑒𝑛 ∆𝑆𝑡 𝑎𝑛𝑑 ∆𝑆𝐴𝑀𝑡 > 0; 0 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.. (3). 政治大 Merton (1981), Henriksson 立 and Merton (1981) develop a nonparametric test to. 4.1.2 Testing the forecast value of intervention. ‧ 國. 學. evaluate investment managers’ forecasting skills. Humpage (1999) apply it to central bank intervention by treating each success criterion as an official forecast of near-term. ‧. dollar movements. Following Humpage (1999), we use this method to assess the. sit. y. Nat. forecasting ability of central bank interventions in Asian countries. Evidence of positive. al. n. superior information.. er. io. forecasting skills would imply that monetary authorities of Asian countries possess. Ch. engchi. i n U. v. For example, consider Asian central banks purchase of US dollars and the general success criterion 1, and define: = 𝑝𝑟𝑜𝑏 [𝐼𝑡 < 0|∆𝑆𝑡 < 0, 𝑜𝑟 ∆𝑆𝑡 < ∆𝑆𝐴𝑀𝑡 ]. (4). 1 − 𝑝1 = 𝑝𝑟𝑜𝑏 [𝐼𝑡 ≥ 0|∆𝑆𝑡 < 0, 𝑜𝑟 ∆𝑆𝑡 < ∆𝑆𝐴𝑀𝑡 ]. (5). = 𝑝𝑟𝑜𝑏 [𝐼𝑡 ≥ 0|∆𝑆𝑡 > 0, 𝑜𝑟 ∆𝑆𝑡 > ∆𝑆𝐴𝑀𝑡 ]. (6). 1 − 𝑝2 = 𝑝𝑟𝑜𝑏 [𝐼𝑡 < 0|∆𝑆𝑡 > 0, 𝑜𝑟 ∆𝑆𝑡 > ∆𝑆𝐴𝑀𝑡 ]. (7). 𝑝1. 𝑝2. In equation (4), p1 is the probability that intervention sale of US dollar by an Asian central bank on day t conditional on the exchange rate conforming to the success criterion given by equation (1). In equation (6), p2 is the probability that the central bank does not 20.

(32) sale dollar on day t conditional on the foreign exchange rate not conforming to the success criterion of equation (1). Equations (5) and (7) are the complementary equations for equations (4) and (6). The null hypothesis that intervention provides no value is equivalent to p1=1- p2 or p1+p2=1. Market participants would not revise their prior estimates of the distribution of exchange rate changes as a result of intervention. Intervention has positive (negative) forecast value if pl+p2>1 (pl+p2<1). In the case when p1+p2>1, market participants can profit from trading in the direction with which Asian central banks intervene in the. 政治大 profitably bet against Asian立 central banks.. foreign exchange market. On the other hand, if p1+p2<1, market participants could. ‧ 國. 學. The number of effective US dollar purchases (sales) by Asian central banks under the null hypothesis has a hypergeometric distribution (Henriksson and Merton, 1981,. ‧. Humpage, 1999), which is independent of the conditional probability or the underlying. y. Nat. io. sit. distribution of the change in exchange rates. In Table 2, if one minus the cumulative. n. al. er. density function (CDF) for the hypergeometric distribution is less than 0.01, we reject the. i n U. v. null in favor of a positive forecast value. If 1-CDF is greater than 0.99, we reject the null. Ch. in favor of a negative forecast value.. engchi. 4.1.3 Determining the odds of successful intervention Using the logistic density function, we estimate the conditional probabilities of successful interventions by Asian central banks. The following logit function measures the odds of success as 𝑃̂ 𝐿̂𝑖 = ln (1−𝑃𝑖̂ ) = 𝑋𝑖 𝑏̂ + 𝑢𝑖 𝑖. (8). , where𝑃̂𝑖 is the probability of a successful intervention, which is estimated by the relative frequency of successes in n trials. 𝑋𝑖 is a (1×K) vector of explanatory variables, which 21.

(33) may affect the probability of successful intervention by Asian central banks. 𝑏̂ is a (K×1) vector of estimated parameters. The first explanatory variable used in this paper is the aggregate number of news on day t (NEWS). Ideally one would measure the success of intervention using the intervention amount, which accords with the portfolio-adjustment theory of exchange rate determination (Humpage, 1999). Due to the lack of intervention amounts data for Asian countries, we construct a proxy for intervention amounts using the number of aggregate intervention news on a specific day. However, given intervention news reports, NEWS,. 政治大 biased. We show in section 4.1(d) how the problem of endogeneity can be circumvented 立. and exchange rate movements are endogenous, the estimated coefficients would be. by using Heckman’s (1979) two-step procedure.. ‧ 國. 學. The second explanatory variable is the number of different sample countries. ‧. intervening on the same day (COUN). Past studies have shown that coordinated. sit. y. Nat. interventions have a bigger impact on exchange rates than do unilateral interventions. io. er. (Dominguez and Frankel, 1990, 1993). Humpage (1999) finds that coordinated interventions with Germany and Japan increase the odds of successful intervention. al. n. v i n C the against the USD. In our sample, number of coordinated interventions is in h ehighest ngchi U South Korea (96 days), and the lowest is in Singapore (11 days).. The third explanatory variable we used is a dummy variable that equals one when intervention day t is the first intervention day in the past five business days (FIRS). This is in accordance with Humpage’s (1999) observation that intervention tends to take place over a number of consecutive days, followed by a number of days of inaction. More often than not, the first intervention of a chain of interventions has a more dominant effect on exchange rates than subsequent ones, implying that the first intervention signals more information to the market than latter interventions. 22.

(34) In addition, we include a fourth variable explanatory variable, SUPP, which is a dummy variable that equals one when official statements supporting interventions appeared on day t. We argue that such statements are likely to carry weight in influencing market expectations and will, therefore, increase the likelihood of intervention success. Finally, the last explanatory variable is a dummy variable equals one when official intervention reports are neutral (NEUT). This variable is used to test whether neutral reports provide any information regarding the odds of intervention success.. Because. some countries may not have enough official statements about supporting interventions or. 政治大 included when estimating the logit regressions. 立. neutral news reports to warrant the estimation coefficients, these variables are not. To account for the impact of coincidental reports of other events on the chances for. ‧ 國. 學. intervention success, we include dummy variables for the announcements of changes of. ‧. repurchase rate (RR), interest rate (IR), interbank overnight rate (IOR), gross domestic. sit. y. Nat. product (GDP), consumer price index (CPI), balance of trade (BOT), and current account. io. er. balance (CAB) of each country in our sample. Only results for reported macroeconomic announcements are presented for brevity and results with all variables are available upon. n. al. request.. Ch. engchi. i n U. v. 4.1.4 Intervention reaction function Intervention reaction functions typically include an exchange rate target, and a measure of volatility, which reflects market disorder (Edison, 1993). We follow this pattern in modeling the decision to intervene, but include a separate equation for the intensity of intervention proxied by the numbers of intervention news reports. Because Asian central banks may intervene out of concern for both the level and the volatility of their exchange rates, we estimate the intervention reaction function using 23.

(35) both ten-day moving average (MOVAG) and a ten-day rolling standard deviation (SIG) of each country. Equation (9) describes the intervention reaction functions for Asian central banks: 𝑍𝑡∗ = Vt a + 𝑢𝑡. (9). where Vt a = â0 + MOVAGt × â1 + SIGt × â2 . To construct a reaction function with unbiased parameter estimates, we estimate a probit model for the decision to intervene and calculate the inverse Mills ratio. We then. 政治大 Mills ratio to control for sample selection bias (Heckman, 1979). 立. estimate a separate equation for the number of intervention reports, using the inverse. ‧ 國. 學. 𝑦𝑡 is the number of intervention reports in Asian countries. We estimate the number of news report in equation (10):. ‧. 𝑦𝑡 = 𝑊𝑡 𝑔 + εt. Nat. of 𝑦𝑡 as. sit. and 𝜀𝑡 follow a bivariate normal distribution with mean zero and variances 2. and. 2. al. n. gives by. Assuming. er. 𝑡. io. both. y. where 𝑊𝑡 𝑔 = 𝑔̂0 + ∆𝑆𝐴𝑀𝑡 × 𝑔̂1. Here, 𝑦𝑡 is observable only when 𝑍𝑡∗ > 0.. (10). i n U. v. , and their correlation, 𝜌, we can obtain the conditional expectation. Ch. engchi. (𝑦𝑡 |𝑦𝑡 𝑖𝑠 𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑) = (𝑦𝑡 |𝑍𝑡∗ > 0) = 𝑊𝑡 𝑔 +. (𝜀𝑡 |𝑦𝑡 > − 𝑡 a) [ ( 𝑡 a)⁄ ( 𝑡 a)]. = 𝑊𝑡 𝑔 + where. ( 𝑡 a) is the normal density function, and. (11). ( 𝑡 a) is the cumulative normal. density function. The term in the square bracket on the right-hand side of the final expression in equation (11) is the Mills ratio.. 24.

(36) 4.2 Empirical Results 4.2.1 Results for the forecast value of intervention Table 2 presents the results of successful intervention and its predicted value for the firm intervention news category. The results show that most Asian central banks have adopted a leaning-against-the-wind strategy and intervened more often by purchasing US dollars to halt the rise of their own currencies when the US dollar depreciated. The effectiveness of central bank intervention varies among Asian countries in Panels A to H of Table 2. By definition, p1 is the probability that a central bank in an. 政治大 success criteria. We find立 that the number is small compared to p2, which is the Asian country sale US dollar on day t conditional on the exchange rate conforming to the. ‧ 國. 學. probability that the central bank does not sale US dollar on day t conditional on the foreign exchange rate not conforming to the success criteria. Even when interventions. ‧. were completely ineffective and randomly undertaken, the martingale nature of exchange. sit. y. Nat. rate changes would insure that it often appeared successful. The “virtual successes”,. n. al. er. io. which counts the number of times the exchange rate mimicked a success criterion without. i n U. v. central bank intervention, indicates that exchange rates conformed to the general success. Ch. engchi. criterion on sixty percent of the sample days. For six of the sixteen general success categories in Table 2, the frequency of actual successes exceeds that of virtual successes. We can reject the null hypothesis at the 95% confidence level in favor of a positive forecast value (p1 + p2 > 1) for the purchases of US dollar intervention in Indonesia and Taiwan, as well as the sales of US dollar intervention in Taiwan using the broad success criterion 1. This suggests that interventions have positive values as recent exchange rate changes will reverse after interventions by these two countries.. In terms of criterion 2,. we show that the null hypothesis can be rejected at the 95% confidence level in favor of a positive forecast value for purchases of US dollar intervention in India, Indonesia, 25.

(37) Malaysia, South Korea, Taiwan, and Thailand, and for sales of US dollar intervention in Indonesia, Philippines, and Taiwan. We can also reject the null in favor of a negative forecast value (pl + p2 < 1) according to the broad success criterion 2 for the sales of US dollar intervention in Malaysia. Results based on criterion 3 indicate that interventions are more effective in reducing local currency changes against the US dollar. We can reject the null hypothesis at the 95% confidence level in favor of a negative forecast value (pl + p2 < 1) in four of the eight countries for purchases of US dollar intervention and six of the eight countries for. 政治大 intervention lead to a smaller 立 local currency appreciation (depreciation) against the US. sales of US dollar intervention. These results imply that purchases (sales) of US dollar. ‧ 國. 學. dollar, which is consistent with the notion that Asian countries have adopted a leaning-against-the-wind intervention policy. The negative values indicate that recent. ‧. exchange rates changes will moderate, but will not reverse.. n. er. io. sit. y. Nat. al. Ch. engchi. 26. i n U. v.

(38) Table 2 Asian Central Bank Intervention and Alternative Success Criteria (Firm News Category) Figures in columns 2, 3, and 4 are the total number of actual intervention, the number of days of successful exchange rate intervention based on a success criterion, and the percentage of successful intervention, respectively. Figures in columns 5, 6, and 7 are total business days, the number of days of successful exchange rate interventions based on a success criterion, and the percentage of virtual successful intervention, respectively. p1 is the probability that a central bank purchases (or sells) US dollars on day t conditional on the exchange rate conforming to the criterion. p2 is the probability that a central bank does not purchase (or sell) US dollars on day t conditional on the foreign exchange rate not conforming to the criterion. p1+p2 denotes the prediction value for intervention. The final column is the value of the test statistic given by one minus the cumulative density function (CDF). ***, **, and * denote significance at the 99%, 95% and 90% confidence levels, respectively.. 政治大 Panel A:. 立. India. Intervention. Sales. #. %. Total. Successes. #. #. %. p1+p2. 1-CDF. 29. 60.4 1,560 1,009 64.7 0.029 0.966. 0.994 0.686. 31. 21. 67.7 1,560. 936 60.0 0.022 0.984. 1.006 0.141. 48. 20. 41.7 1,560. 809 51.9 0.025 0.963. a 31l. 8. 25.8 1,560. 729 46.7 0.011 0.972. sit. y. Nat. n. Leaning-against-the-wind. Ch. er. io. Sales. p2. 48. Appreciate / Depreciate Purchases. p1. ‧. Purchases. #. Successes. 學. General. ‧ 國. Total. Virtual. engchi U. v ni. 0.987 0.901 * 0.983 0.987 **. Purchases. 48. 8. 16.7 1,560. 194 12.4 0.041 0.971. 1.012 0.132. Sales. 31. 13. 41.9 1,560. 191 12.2 0.068 0.987. 1.055 0.000 ***. Panel B:. Indonesia. Intervention Total #. Successes #. %. Virtual Total #. Successes #. p1. p2. p1+p2. 1-CDF. %. General Purchases. 25. 10. 40.0 1,636. 975 59.6 0.010 0.977. 0.988 0.963 **. Sales. 70. 35. 50.0 1,636. 974 59.5 0.036 0.947. 0.983 0.937 *. 25. 5. 20.0 1,636. 756 46.2 0.007 0.977. 0.984 0.994 ***. Appreciate / Depreciate Purchases. 27.

(39) Sales. 70. 25. 35.7 1,636. 794 48.5 0.031 0.947. 0.978 0.981 **. Purchases. 25. 5. 20.0 1,636. 182 11.1 0.027 0.986. 1.014 0.051 *. Sales. 70. 9. 12.9 1,636. 142. 1.023 0.075 *. Leaning-against-the-wind. Panel C:. Malaysia. Intervention Total #. Successes #. %. 8.7 0.063 0.959. Virtual Total. Successes. #. #. p1. p2. p1+p2. 1-CDF. %. General Purchases. 21. 12. 57.1 1,636. 928 56.7 0.013 0.987. 1.000 0.401. Sales. 3. 1. 33.3 1,636. 935 57.2 0.001 0.997. 0.998 0.607. Appreciate / Depreciate Purchases. 19.0 1,636. 685 41.9 0.006 0.982. 0.988 0.976 ** 1.004 0.000 ***. 立. 4. 3 100.0 1,636. 746 45.6 0.004 1.000. 21. 8. 38.1 1,636. 208 12.7 0.038 0.991. 3. 1. 33.3 1,636. 163 10.0 0.006 0.999. 21. Sales. 政治大. 3. io. n. a# l. General Purchases. 42. Sales. 40. Purchases Sales. Successes #. %. Ch 26. 1.005 0.028 **. Philippines. Intervention Total. 1.029 0.001 ***. y. Nat. Panel D:. ‧. Sales. Virtual Total #. Successes #. %. i n U. i e61.9 n g1,635 c h1,006. p1. er. Purchases. sit. ‧ 國. 學. Leaning-against-the-wind. p2. p1+p2. 1-CDF. v. 61.5 0.026 0.975. 1.000 0.421. 22. 55.0 1,635 1,035 63.3 0.021 0.970. 0.991 0.826. 42. 14. 33.3 1,635. 733 44.8 0.019 0.969. 0.988 0.914 *. 40. 13. 32.5 1,635. 860 52.6 0.015 0.965. 0.980 0.992 ***. Purchases. 42. 12. 28.6 1,635. 242 14.8 0.050 0.978. 1.028 0.005 ***. Sales. 40. 9. 22.5 1,635. 166 10.2 0.054 0.979. 1.033 0.005 ***. Appreciate / Depreciate. Leaning-against-the-wind. 28.

(40) Panel E:. Singapore. Intervention Total #. Successes #. %. Virtual Total. Successes. #. #. p1. p2. p1+p2. 1-CDF. %. General Purchases. 14. 10. 71.4 1,634 1,011 61.9 0.010 0.994. 1.003 0.155. Sales. 2. 1. 50.0 1,634 1,070 65.5 0.001 0.998. 0.999 0.429. Purchases. 14. 6. 42.9 1,634. 718 43.9 0.008 0.991. 1.000 0.422. Sales. 2. 1. 50.0 1,634. 900 55.1 0.001 0.999. 1.000 0.303. 14. 4. Appreciate / Depreciate. Leaning-against-the-wind. Total #. Sales. 1.012 0.000 ***. Panel F:. South Korea. Successes #. %. Virtual Total #. Successes #. p1. %. p2. 80. 47. 58.8 1,631 1,009 61.9 0.047 0.947. 53. 31. 58.5 1,631. 997 61.1 0.031 0.965. 28.8 1,631. 782 47.9 0.029 0.933. al. n. Purchases. 166 10.2 0.012 1.000. 80 53. p1+p2. 1-CDF. 0.994 0.683 0.996 0.605. er. io Appreciate / Depreciate. 2 100.0 1,634. ‧. Sales. 1.007 0.076 *. Intervention. Nat. Purchases. 281 17.2 0.014 0.993. 學. ‧ 國. 2. 28.6 1,634. y. 立. Sales. General. 政治大. sit. Purchases. Ch 23 25. i n U. e47.2 n g1,631 c h i 830. v. 50.9 0.030 0.965. 0.962 1.000 *** 0.995 0.659. Leaning-against-the-wind Purchases. 80. 24. 30.0 1,631. 217 13.3 0.111 0.960. 1.071 0.000 ***. Sales. 53. 6. 11.3 1,631. 158. 1.006 0.248. 29. 9.7 0.038 0.968.

(41) Panel G:. Taiwan. Intervention Total #. Successes #. Virtual Total. %. Successes. #. #. p1. p2. p1+p2. 1-CDF. %. General Purchases. 169. 81. 47.9 1,582. 871 55.1 0.093 0.876. 0.969 0.970 **. Sales. 32. 17. 53.1 1,582 1,110 70.2 0.015 0.968. 0.984 0.970 **. Purchases. 169. 49. 29.0 1,582. 632 39.9 0.078 0.874. 0.951 0.999 ***. Sales. 32. 9. 28.1 1,582. 927 58.6 0.010 0.965. 0.975 1.000 ***. 169. 30. 17.8 1,582. 228 14.4 0.132 0.897. 1.029 0.080 *. 立. 8. 25.0 1,582. 171 10.8 0.047 0.983. 1.030 0.005 ***. Appreciate / Depreciate. Leaning-against-the-wind. Total #. Sales. %. #. Successes #. p1. p2. %. 64.1 1,634 1,031 63.1 0.048 0.954. 17. 11. 64.7 1,634. 985 60.3 0.011 0.991. 25.6 1,634. i n 793 48.5 U. 78 17. 1-CDF. 1.002 0.271. 50. al. p1+p2. 1.002 0.382. 78. n. Purchases. #. Total. er. io Appreciate / Depreciate. Successes. Virtual. ‧. Sales. Intervention. Nat. Purchases. Thailand. 學. General. Panel H:. y. 32. ‧ 國. Sales. 政治大. sit. Purchases. C20h. e n g1,634 c h i 797 6 35.3. v. 0.025 0.931. 48.8 0.008 0.987. 0.956 1.000 *** 0.994 0.808. Leaning-against-the-wind Purchases. 78. 28. 35.9 1,634. 213 13.0 0.131 0.965. 1.096 0.000 ***. Sales. 17. 5. 29.4 1,634. 173 10.6 0.029 0.992. 1.021 0.006 ***. 30.

(42) For criterion 3, the number of actual successes always exceeds the virtual success. This suggests that, for the most part, the frequency of success under criterion 3 is greater than that of a random event. We can reject the null hypothesis of no forecast value for purchases and sales of US dollars intervention by Asian central banks with a high level of confidence under criterion 3. Results in this section are based on firm intervention reports. The next two sections of the paper demonstrate that various aspects of how Asian central bank undertake their interventions and factors which can increase the probability of their successes.. 政治大. 4.2.2 Results for the intervention reaction function. 立. Table 3 presents the estimates of the reaction functions for our sample countries. It. ‧ 國. 學. can be seen from the negative and statistically significant coefficient of MOVAG that central banks in India, Malaysia, Philippines, Singapore, Taiwan, and Thailand tend to. ‧. intervene in the markets when their own currencies appreciate against the US dollar. In. Nat. sit. y. addition, exchange rates in India, Philippines, and Taiwan are more volatile as shown in. n. al. er. io. the ten-day rolling standard deviation, which also increases the probability of. v. interventions. It is interesting to note that only the coefficient of ∆SAMt for Malaysia,. Ch. engchi. i n U. Singapore, and Taiwan is statistically significant at the conventional significance level, which suggests that there are more intervention reports for these countries when their currencies have higher exchange rate changes (∆SAMt ).. 31.

(43) Table 3. Results of Asian Countries’ Intervention Reaction Functions Using Firm News Category. We estimate the intervention function using Heckman’ s (1979) two-step approach: 𝑍𝑡∗ = â 0 + MOVAGt × â1 + SIGt × â2 +. (9). 𝑡. 𝑦𝑡 = 𝑔̂0 + ∆𝑆𝐴𝑀𝑡 × 𝑔̂1 + 𝜀𝑡. (10). Equation (9) is a probit model for the decision to intervene. MOVAG is the ten-day moving average of the 9:00 a.m. exchange rates. SIG is the ten-day rolling standard deviation of the 9:00 a.m. exchange rates in equation (10) that the number of intervention news. ∆𝑆𝐴𝑀𝑡 is the opening exchange rate of day t minus the opening exchange rate of day t-1. The sample period is from January 2005 to April 2011. *,**, and *** denotes significance at the 90%, 95%, and 99% confidence levels, respectively.. 政治大 Panel A:. 立. CONS. io. Log likelihood. 6.55 ***. 4.98. 5.53 ***. y. 0.54. Nat. ΔSAM. 1.99. sit. Equation (10). 2.70 -387.88. n. al. Panel B:. Ch. -7.50 ***. ‧. CONS. -0.16. i n U. Indonesia. 1.33 5.45 ***. er. SIG. t-statistic. 學. MOVAG. Coefficient. ‧ 國. Equation (9). India. v. Equation (9). e nCoefficient gchi. MOVAG. 1.89e-11. 0.00. -2.40e-10. -0.00. SIG CONS. -1.59. t-statistic. -31.20 ***. Equation (10) ΔSAM. -2.45e-11. CONS. -0.34. Log likelihood. -0.00 -4.37 ***. -353.75 Panel C:. Malaysia. Coefficient. t-statistic. Equation (9) MOVAG. -3.39. 32. -5.25 ***.

(44) SIG. -20.31. CONS. -1.37. 9.26. 4.59 ***. Equation (10) ΔSAM. -17.62. CONS. 1.69. Log likelihood. -1.71 * 4.14 ***. -111.78 Panel D:. Philippines. Coefficient. t-statistic. Equation (9) MOVAG. -0.10. -5.48 ***. SIG. 1.63. 3.97 ***. CONS. 2.56. 3.07 ***. Equation (10) ΔSAM. -0.06. 立. CONS. CONS Log likelihood. Singapore. -3.95 ***. -32.93. y. t-statistic. -0.80. 4.34. sit. Coefficient. -4.57. 2.73 ***. er. al. n. ΔSAM. -350.87 Panel E:. io. Equation (10). 2.37 **. ‧. CONS. 1.16. Nat. SIG. -0.28. 學. MOVAG. ‧ 國. Log likelihood. Equation (9). 政治大. Ch. -38.46 1.19 e n g c-81.12 hi. Panel F:. i n U. v. -2.03 ** 2.70 ***. South Korea. Coefficient. t-statistic. Equation (9) MOVAG. 4.40e-10. 0.00. SIG. 2.18e-09. 0.00. CONS. -1.42. -20.23 ***. Equation (10) ΔSAM. 2.51e-09. CONS. -0.79. Log likelihood. -509.57. 33. 0.00 -1.88 *.

(45) Panel G:. Taiwan. Coefficient. t-statistic. Equation (9) MOVAG. -0.11. -3.08 ***. SIG. 2.36. 4.22 ***. CONS. 2.17. 1.87 *. ΔSAM. -0.59. -2.15 **. CONS. 1.25. Equation (10). Log likelihood. 7.61 ***. -683.55 Panel H:. Thailand. Coefficient Equation (9) MOVAG. 立. SIG. -2.45 **. 0.46 1.66 -413.92. y. Nat. io. n. al. Ch. engchi. 34. 5.08 ***. 0.72. sit. Log likelihood. -1.13. ‧. CONS. -6.83 ***. er. ΔSAM. -0.16. 3.88. ‧ 國. Equation (10). 政治大. 學. CONS. t-statistic. i n U. v. 4.23 ***.

(46) 4.2.3 Factors influencing the odds of successful intervention Having estimated the intervention reaction function, we obtain NEWSHAT to proxy for the number of intervention reports and which adjusts for sample selection bias. Table 4 presents the odds of successful interventions, according to the broad criterion 1. It can be seen that the odds of successful interventions according to this criterion increase significantly with the number of firm intervention reports in India, Indonesia, Malaysia, Singapore, South Korea, Taiwan, and Thailand. However, the coefficient estimates of NEWSHAT for both Indonesia and South Korea are virtually zero to make any economic. 政治大 intervention from having no firm intervention report to having one firm intervention 立. significance. Based on the estimated coefficient, the increase in the odds of successful. report is about 45% for India, 24% for Malaysia, 46% for Singapore, 49% for Taiwan,. ‧ 國. 學. and 47% for Thailand, all things remaining unchanged. The probability of having no firm. ‧. intervention report is 0.5, ceteris paribus. Consequently, having a firm intervention report. sit. y. Nat. significantly improves the likelihood of intervention success in five of the eight Asian. io. er. countries. In the Philippines case, the negative coefficient of -38.039 suggests that the odds of successful interventions decrease by about 50% when the number of firm. al. n. v i n intervention reports increases C from zero to one. Philippines is the only country that hengchi U. exhibits the opposite and perverse effect from having a firm intervention report on the probability of intervention success. The odds of successful interventions increase significantly with the number of coordinated interventions (i.e. COUN) and with the first intervention day in the past five business days (i.e. FIRS) for all the countries. When no country is jointly intervening on the same day, the probability of successful intervention is 0.5. When another country jointly intervenes on the same day, the probability of successful intervention goes up by 31% for South Korea (the highest increase in the odds) and 23% for Taiwan (the lowest 35.