Capital market integration in the Pacific Basin region: an ... - CiteSeerX

5 downloads 0 Views 195KB Size Report
Kate Phylaktis. U. City Uni¨ersity ..... The use of cointegration technique, developed initially by Granger 1981 to. 9Such an .... Multi¨ariate Granger causality tests.
Journal of International Money and Finance 18 Ž1999. 267᎐287

Capital market integration in the Pacific Basin region: an impulse response analysis Kate PhylaktisU City Uni¨ ersity Business School, Dept of Banking and Finance, Frobisher Crescent, Barbican Centre, London EC2Y 8HB, UK

Abstract We examine the extent of capital market integration in a group of Pacific Basin countries following the deregulation of their markets, and explore whether the financial influence of Japan in the region has overtaken that of the US. Looking at long-run comovements of real interest rates through the use of cointegration, and using impulse response analysis to examine the speed of adjustment of real interest rates to long-run equilibrium following a shock in one of the markets, which is another indicator of the degree of capital market integration, we find that these countries are closely linked with world financial markets and more so with Japan than with US. 䊚 1999 Elsevier Science Ltd. All rights reserved. JEL classifications: F21; F31; F32; F36 Keywords: Capital Market Integration; Impulse Response Analysis; Real Interest Rate Parity; Pacific-Basin Capital Markets

1. Introduction The objective of the paper is to examine the extent of capital market integration in the Pacific Basin countries following the deregulation of their domestic financial markets and foreign exchange markets in recent years. This deregulation process has varied across the countries both in terms of intensity and timing. For example, Hong Kong liberalised interest rates and international capital flows as early as U

Tel.: q171 477 8735; fax: 171 477 8881; e-mail: [email protected]

0261-5606r99r$ - see front matter 䊚 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 1 - 5 6 0 6 Ž 9 8 . 0 0 0 4 6 - 1

268

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

1973, while Taiwan and Korea took steps towards liberalisation as late as 1987 and 1988, respectively.1 This general deregulation process has been accompanied by signs that Japan has been increasing its financial influence in Asia and possibly overshadowing that of the US. For example, Yuan Ž1986. in his examination of capital flows amongst the Pacific Basin economies finds that 65% of total net flows Žprivate and public. in 1972 and 55% in 1980᎐1982 came from Japan, while the corresponding figures for the US were 31% and 41%, respectively. More recently, data on securities show that Japan purchased net US$725 million worth of foreign securities from the Newly Industrialised countries in Asia on a cumulative basis during the period from 1988 to 1991, while the US sold foreign securities on a net basis in those countries.2 Tavlas and Ozeki Ž1992. examined the role of yen in the region and found that Asian Central Banks in the course of the 1980s have increased their holdings of yen from 13.9% of their foreign exchange reserve portfolios to 17.5%. The yen is also being used more widely to invoice trade and finance in Asia. The percentage of southeast Asian imports denominated in yen increased from 2% in 1983 to 19.4% in 1990. The countries that incurred large international debt in the 1970s and early 1980s subsequently shifted the composition away from dollar-denominated debt towards yen-denominated debt.3 In this paper, we examine these developments by looking at the effects that they may have had on the degree of capital market integration between the financial markets of Pacific Basin countries and the world financial markets, such as the US and Japan. We also examine whether the degree of capital market integration with Japan is greater than that with the US. We look at the covariability of real interest rates, which examines whether prices of financial assets in different countries move in conjunction, but not necessarily have the same level. Different interest-rate levels may prevail because of different levels of transaction costs and taxes, including the implicit tax of foreign exchange controls and the political risk premium relating to future controls.4 We also concentrate on a different indicator of capital market integration than customarily used in the literature. We look at the speed of adjustment of interest rates to re-establish long-run equilibrium following a shock. The various capital market imperfections which drive a wedge between market rates of return also reduce the sensitivity of capital supply that would eliminate differentials beyond 1

Other countries in our sample, such as Singapore, deregulated domestic financial markets and foreign exchange markets in 1978, while Malaysia although began liberalising exchange controls in 1973, completed the process in 1984. A free market interest rate regime was adopted in 1978. Japan began the reform of its domestic financial market in mid 1970s, while its foreign exchange market was liberalised at the end of 1980. 2 See Chinn and Frankel Ž1994.. 3 For example, in Malaysia in 1993 the yen denominated long-term debt was 37.5% compared to 25.1% in dollars Žsee The World Bank Debt Tables, 1994᎐95.. 4 See Phylaktis and Wood Ž1984. and Phylaktis Ž1988. for an explanation of the effects of capital controls on the international parity conditions.

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

269

the wedge. We examine how long it will take to eliminate a differential which is greater than that defined by the wedge. The greater the degree of capital mobility the faster the adjustment to long-run equalisation of real interest rates. Even in the case when capital controls are ineffective, i.e. do not interpose a wedge between domestic and foreign real rates of return, the mere existence of controls may give rise to a gradual adjustment to long-run equilibrium as investors may be facing increasing marginal costs of acquiring foreign assets or liabilities. This can be seen by considering popular channels through which controls can be evaded and capital can be transferred abroad, such as physical movements of cash and leads and lags. Physical movements of cash might involve increasing costs if the potential fine is a proportion of the attempted transfer that is increasing with the size of the transfer. Leads and lags could also involve increasing marginal costs because importers and exporters can operate only with a limited flow Žtheir regular trade transactions ., and lengthening the delay with which they pay their foreign bills becomes increasingly more difficult to justify to the authorities enforcing the controls.5 This indicator of capital market integration provides useful information to policy makers. The existence of a wedge between interest rates gives a country the opportunity to persue a stance of monetary policy which entails a different interest rate from the world interest rate Žunder fixed exchange rates.. If this stance is, however, changed entailing for example a bigger differential than the existing wedge Žthe tightness of capital controls remaining the same., the length of time it takes for the initial wedge to be re-established, indicates the breathing space that the monetary authorities have. Many studies on capital market integration in the Pacific Basin region have concentrated on integration between Japan and the US Žsee e.g. Otani and Tiwari, 1981; Ito, 1988; and Bosner-Neal and Roley, 1994.. Recently, there has been a lot of interest in other Pacific Basin countries. For example, Bhoocha-Oom and Stansell Ž1990. look at interest rates Žadjusted and unadjusted for exchange rates changes. between Hong Kong and Singapore vs. the US. Faruqee Ž1992. examines the uncovered interest rate differential between Singapore, Korea and Thailand vs. the Japanese LIBOR ᎏ taken to represent the world rate of interest.6 Dooley and Mathieson Ž1994. look at seven Pacific Basin countries vs. the US using an analytical framework for interest rate determination, where the prevailing interest rate represents a weighted average of open ŽUS interest rate adjusted for the change in the exchange rate. and closed economy rates that would have existed otherwise.7 Reisen and Yeches Ž1993. using the same framework examine Korea and Taiwan in greater detail. The results of these studies support the view that there is substantial integration between domestic and international financial mar5

Similar assumptions of increasing marginal costs are also made in Khan and Haque Ž1985. and Gros Ž1987.. 6 The change in the exchange rate is assumed to be zero. 7 This is based on work done by Edwards and Khan Ž1985. for the case of Singapore and Colombia; and Haque and Montiel Ž1991. for 15 developing countries.

270

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

kets in Japan, Hong Kong, Singapore and Malaysia, while the views are divided for Korea and Thailand. In Taiwan capital market integration with world financial markets has been found to be limited.8 In the current paper, we present evidence which shows that there is substantial capital market integration even in Korea and especially in Taiwan. That inference is based on the examination of real interest rates across countries. That encompasses both financial integration as well as goods market integration as can be seen from the derivation of real interest rate parity ŽRIP. below. Assuming that the Fisher Hypothesis holds for both countries, i and j in our case, we have Et rtj,k s Et i tj,k y Et ␲ tj,k ,

Ž1.

Et rti,k s Et i ti ,k y Et ␲ ti,k ,

Ž2.

where Et is the expectations operator conditional on information given at time t, i t ,k is the nominal k period interest rate, rt ,k is ex ante real interest rate and ␲ t ,k is the inflation rate from time t to t q k. When one subtracts these equations and invokes rational expectations, since ex ante real interest rates are not observed, one can decompose the resulting relationship as follows: rti,k y rti,k s Ž i tj,k y i ti ,k y fd t ,tqk . q Ž fd t ,tqk y Et ⌬ stqk . q Ž Et ⌬ stqk y ␲ tj,k q ␲ ti,k . q⑀ tqk ,

Et Ž ⑀ tqk < ⌽t . s 0,

Ž3.

where fd t ,t q k is the forward discount on the domestic currency, s is the log spot exchange rate and ⑀ t q k is the inflation prediction error. The first term on the right hand side is the covered interest differential, the second represents the exchange risk premium and the third term the Purchasing Power Parity ŽPPP.. Thus, when one examines RIP, one tests the joint hypothesis of covered interest parity ŽCIP., PPP and the stationarity of the exchange risk premium. In the words of Frankel Ž1993. only ‘CIP is an unalloyed criterion for capital mobility in the sense of the degree of financial market integration across national boundaries’. In our study, however, forward markets were not well developed in all the countries and for the whole sample period. We therefore proceeded to test for financial market integration by examining real interest rates. We were encouraged to do so as earlier work has shown that PPP holds for these Pacific Basin countries over the same sample period Žsee e.g. Phylaktis and Kasimmatis, 1994a., thus implying that favourable results to RIP will be indicative of financial market integration. The paper contributes to the literature in the following ways. First, we use a different methodology to what has been used in the earlier studies on capital market integration in the Pacific Basin countries. We use cointegration methodology which examines whether real interest rates have a long-run relationship, but 8

Recently Chinn and Maloney Ž1996. using a different method of measuring capital mobility based on a portfolio balance model, found evidence of greater degree of openness in Taiwan since early 1989.

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

271

allows for movements in rates within a band defined by transaction costs, taxes and capital controls. Second, we use the speed of adjustment of interest rates following a shock in one of the markets as an additional indicator of capital market integration, by subjecting the system to an impulse response analysis. Finally, we use a multivariate framework which takes into account the interactions of real interest rates between the US and Japan in the examination of capital market integration between each one of these world markets and the Pacific Basin countries. The paper is structured as follows. Section 2 explains methodological issues. It discusses the real interest parity within the cointegration methodology. It explains the application of impulse response analysis in a cointegrated vector autoregressive system ŽVAR. developed by Lutkepohl and Reimers Ž1992. and its relation to capital market integration. It finally examines the application of multivariate Granger Causality tests suggested by Dolado and Lutkepohl Ž1996.. Section 3 discusses the data and presents the empirical results. Finally, Section 4 summarises the main findings and policy implications.

2. Methodological issues 2.1. Bi¨ ariate and multi¨ ariate cointegration Real interest rate parity as a way of measuring the degree of integration between two different financial markets has usually taken the form of estimating the following regression rtj s ␣ 0 q ␣ 1 rti ,

Ž4.

where rtj and rti are the real interest rates in countries j and i, respectively, and testing for the joint hypothesis that ␣ 0 s 0 and ␣ 1 s 1.9 There are two weaknesses with the above test. First, it does not allow for any capital market imperfections, such as transaction costs. Such costs even when they are small can lead to estimates of ␣ 0 and ␣ 1 taking different values from the expected ones of zero and one. Secondly, the usual regression results assume that individual real rates are stationary, which is not always the case.10 If the series are non-stationary then the empirical estimates of the parameters ␣ 0 and ␣ 1 will be consistent but their estimated standard errors will not be consistent Žsee Stock, 1987.. The use of cointegration technique, developed initially by Granger Ž1981. to 9

Such an approach has been followed in earlier studies on real interest rate linkages applied to the US, Canada and some European countries Žsee Mishkin, 1984a,b; Mark, 1985; Cumby and Mishkin, 1986;and Merrick and Saunders, 1986.. The results are on the whole unfavourable to real interest rate equalisation. 10 See Mishkin Ž1995. for evidence on the non-stationary behaviour of real interest rates.

272

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

explore the long-run relationship between two series, overcomes these problems. According to cointegration if two series, such as real interest rate in two different markets, are non-stationary, but there exists some linear combination of them which is a stationary process, then the two rates are said to be cointegrated with a cointegrating parameter, ␣ 1 Žsee Eq. Ž4...11 In the presence of transaction costs and non-synchronous trading, ␣ 1 will be expected to be different from one. In the case where costs are not proportional to interest rates, market efficiency implies that ␣ 0 will be different from zero. Thus, even in the absence of capital controls the joint hypothesis that ␣ 0 s 0 and ␣ 1 s 1 can be rejected because of transaction costs and that will not imply that profitable arbitrage opportunities exist. In a multivariate system the issue of cointegration can be given a different perspective. Stock and Watson Ž1988. have developed a test for the existence of common trends in a set of non-stationary variables. If a set of real rates have a single common trend, that will mean that any single rate is representative of the group of rates examined indicating complete integration of the capital markets. In other words, although each univariate series might contain a stochastic trend, in a vector process these stochastic trends might be common to several of the variables. When the series contain the same stochastic trend then they are said to be cointegrated. In implementing the Stock and Watson Ž1988. test Žas well as the tests for cointegration in the bivariate case., we use the likelihood ratio test due to Johansen Ž1988. and Johansen and Juselius Ž1990.. Let Yt ' (rP B C , rUS , r J ) where rP B C is the real interest rate in the Pacific Basin country, rUS the real interest rate in the US and r J the real interest rate in Japan; n the number of variables in the system, three in this case. If Yt is cointegrated, it can be generated by a vector error correction model ŽVECM.: ky1

⌬Yt s ␮ q

Ý Gi ⌬Ytyi q Gk Yty1 q ⑀ t ,

Ž5.

is1

where ␮ is a 3 = 1 vector of drift, G’s are 3 = 3 matrices of parameters, and ⑀ t is a 3 = 1 white noise vector. The Johansen trace test statistic of the null hypothesis that there are at most r cointegrating vectors 0 F r F n, and thus Ž n y r . common stochastic trends is n

trace s yT

Ý

ˆi . , In Ž 1 y ␭

Ž6.

isrq1

ˆi ’s are the n y r smallest squared canonical correlations of Yt y 1 with where ␭ respect to ⌬Yt corrected for lagged differences and T is the sample size actually used for estimation. In the context of the current paper, if we cannot reject the hypothesis that there 11 Such an approach has been applied to European countries by Goodwin and Grennes Ž1994. with more favourable results to real interest rate parity.

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

273

are two cointegrating vectors, that will imply that there is one common stochastic trend. This will suggest that any single rate will be representative of the rates included in the group, which implies a considerable degree of integration amongst the three markets. 2.2. Impulse response analysis We next subject our cointegrated vector autoregressive system ŽVAR. to impulse response analysis in order to examine how rapidly the real interest rate movements in one market are transmitted to the other markets. As it has been explained the speed of adjustment is an indicator of the degree of capital market integration. Lutkepohl and Reimers Ž1992. show that innovation accounting can be used to obtain information concerning the interactions among the variables. They develop the asymptotic distribution of the impulse responses and forecast error variance components of a Gaussian VAR process with cointegrated variables.12 They suggest that one estimates the undifferenced VAR of VECM of Eq. Ž5.. Yt s ␮ q A1Yty1 q . . . qA p Ytyk q ⑀ t ,

Ž7.

where A i are 3 = 3 coefficient matrices in our case, by multivariate least squares as the resulting estimator has an asymptotic normal distribution where a covariance matrix may be estimated by the usual formula for stationary processes. Eq. Ž7. can be transformed to a vector moving average representation given below which can be used to examine the interactions between the three real interest rates, Yt s ␮ q



Ý ␾i ⑀ tyi .

Ž8.

is0

The coefficients of ␾ i can be used to generate the effects of ⑀ P B C , ⑀ US and ⑀ J shocks on the entire time paths of rP B C , rUS and r J . The shocks can take the form of one standard error of each real interest rate. The accumulated effects of the impulses in ⑀ P B C , ⑀ US and ⑀ J can be obtained by the appropriate summation of the coefficients of the impulse response function. Since the VAR is underidentified Choleski decomposition is often used to orthogonalise the innovations. The results of this approach are not, however, invariant to the ordering of the variables in the VAR. In a recent paper Koop et al. Ž1996. have proposed an alternative approach, the generalised impulse response analysis which does not have that shortcoming. This is achieved by examining the shock in one of the variables, and integrating the effects of other shocks using an assumed or historically observed distribution of the errors. The issue of interest to us is to see how many months it takes for the impulse 12

The asymptotic theory is derived from Johansen Ž1988; Johansen, 1991. maximum likelihood estimation procedure for such systems.

274

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

responses to decay following a shock. The impulse responses should converge to zero because the system is stationary. In the multivariate system, however, of the US, the Pacific Basin country and Japan, our interest is not when all three responses decay, but when the responses of the US and of the Pacific Basin country decay on the one hand, and, when the responses of Japan and of the Pacific Basin country decay on the other hand. If the speed of convergence of the first pair is faster than that of the second, then it can be said that the Pacific Basin country is better integrated with US than with Japan. In examining the speed of adjustment in this multivariate framework we take into account the interaction between the real interest rates in the US and Japan. 2.3. Multi¨ ariate Granger causality tests Apart from the examination of the long-run co-movements of real interest rates, we explore the short-run dynamics by performing multivariate Granger-causality tests for cointegrating systems, by applying the methodology suggested by Dolado and Lutkepohl Ž1996.. Dolado and Lutkepohl Ž1996. propose a method which leads to Wald tests with standard asymptotic ␹ 2-distributions and which avoids possible pretest biases associated with the usual procedure of estimating a first order differenced VAR if variables are known to be IŽ1. with no cointegration, and an error correction model if they are known to be cointegrated. As they point out the testing procedure on the estimation of unit roots, cointegration rank and cointegrating vectors has unknown properties, leaving open the possibility of severe distortions in the inference procedure. Their method is performed directly on the least squares estimators of the coefficients of the VAR process specified in levels of the variables.13 The procedure is based on the argument that the non-standard asymptotic properties of the Wald test on the coefficients of cointegrated VAR systems are due to the singularity of the asymptotic distribution of the least square estimators. Their suggested method gets rid of the singularity by fitting a VAR process whose order exceeds the true order. They show that this device leads to a non-singular distribution of the relevant coefficients. The method involves the following steps. First, one finds the lag structure of the VAR by testing a VARŽk. against a VARŽk q 1., k G 1 using the standard Wald test. Secondly, if the true data generating process is a VARŽk., a VARŽk q 1. is fitted and standard Wald tests are applied on the first k VAR coefficient matrix. In the context of our paper, the above method implies fitting the VAR depicted in Eq. Ž7. for each Pacific Basin country. The expanded version of the VAR is

13

It should be noted that although the variables are allowed to be potentially cointegrated, it is not assumed that the cointegration structure of the system under investigation is known. Therefore preliminary unit root tests are not necessary and, the testing procedure is robust to the integration and cointegration properties of the process.

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

275

given below A10 rP B C rUS s A 20 q rJ A 30

A11 Ž L . A 21Ž L .

A12 Ž L . A 22 Ž L .

A13 Ž L . A 23 Ž L .

A 31Ž L .

A 32 Ž L .

A 33 Ž L .

rP B C ,ty1 ⑀P B C rUS ,ty1 q ⑀ US , r J ,ty1 ⑀J

Ž9.

where A i0 are the parameters representing intercept terms and A i j the polynomials in the lag operator L. We select the lag structure using the Wald test, and then re-estimate the VAR adding one extra lag. Since each equation has the same lag length, we estimate the three equations using OLS as the estimates are consistent and asymptotically efficient Že.g. Enders, 1995.. We perform Granger causality tests to examine whether the past values of the US or Japanese rates have been driving the Pacific Basin country rate, by using the Wald tests once again on the initial VAR coefficients. For example, to test whether rUS Granger causes rP B C , we test the restriction that A12 Ž L. s 0. 3. Empirical results 3.1. Data Six Pacific Basin countries were selected for the empirical analysis: Singapore, Malaysia, Hong Kong, Korea, Taiwan and Japan. The sample period varies for each country according to the availability of data. For Singapore the sample period is August 1973᎐September 1993; for Malaysia, January 1982᎐September 1993; for Taiwan and Korea, February 1972᎐September 1993; for Hong Kong, January 1976᎐September 1993; and for Japan, January 1974᎐September 1993. We have used end of month data apart from the case of Korea, where we have used end of quarter. The money market interest rates used were as follows: 90-day Treasury Bill rate for the US; the three month Gensaki rate for Japan;14 the three month regulated deposit rate for Hong Kong; and the three month interbank rate for Singapore and Malaysia. For Taiwan and Korea, we have used short-term curb rates. The curb market is an unofficial, largely unregulated financial market involving small borrowers and lenders.15 We have used these rates because the domestic financial markets were highly regulated even during the 1980s.16,17 14

Gensaki transactions consist of the resale or repurchase of bonds at a fixed price after a fixed period. They are short-term capital transactions using bonds as collateral. Prior to February 1977 we have used the 60-day Gensaki rate. 15 The size of the markets remains substantial but has fallen over the years. For example, in the mid 1970s the aggregate size of the curb market in Taiwan was as large as that of all financial institutions put together. In 1986, according to flow-of-funds accounts for private business enterprises, the ratio of curb market to total bank borrowing was 48% Žsee Fry, 1990.. The curb rate in Taiwan is the average loan rate against post-dated checks in Kaoshiung city; and in Korea it is a monthly rate which is provided quarterly.

276

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

In order to examine the effects of deregulation we divided the sample into two sub-periods. The first period ends in December 1980 and represents the period of financial regulation. The second sub-period covers January 1981᎐October 1993 and represents the post-liberalisation period. As it has already been mentioned substantial deregulation took place in many of the countries under consideration in the late 1970s and early 1980s. The exact date of the division of the sample period was chosen because of the drastic shift in foreign exchange control policy that took place in Japan. The Foreign Exchange and Foreign Exchange Trade Law was passed which freed most capital flows. In addition, restrictions on non-residents’ Gensaki transactions were completely eliminated. This relaxation of foreign exchange controls makes possible the examination of capital market integration between Japan and the Pacific Basin countries during the 1980s.18 Following Cumby and Mishkin Ž1986., the ex post real interest rate on a j period financial instrument held until maturity was defined from the Fisher Condition as rt ,k s i t ,k y ␲ t ,k ,

Ž 10.

where rt ,k is the real return at time t earned from holding the asset for k periods; i t ,k is the nominal k period interest rate; and ␲ t ,k is the rate of inflation from t to t q k.19 3.2. Bi¨ ariate and multi¨ ariate cointegration results Before testing for cointegration we tested for unit roots in the real interest rates for the two sub-periods, whenever that was possible. The results are not presented but can be made available by the author. We used the Augmented Dickey Fuller test with and without trend as recommended by Engle and Granger Ž1987. and found that the null hypothesis of a unit root for the first difference can be rejected for all real interest rates. On the other hand, the null hypothesis of a unit root in levels was accepted in all cases.20 Thus, like most financial series, these real interest rates are IŽ1., which means that first differencing is required to achieve stationarity. 16

Taiwan completely liberalised interest rates in 1989, while Korea started to lift regulations on interest rates in 1989 and completed the process in 1993. 17 Data on prices refer to consumer price index and were taken from the International Financial Statistics published by the International Monetary Fund, except for Taiwan where prices were taken from Monthly Statistics of the Republic of China. Data on interest rates for Malaysia, Hong Kong and Singapore were provided by Nomura Bank; for Korea by the International Monetary Fund which were obtained from the Korean Development Institute; for Taiwan they were taken from the Financial Statistics Monthly, Taiwan District, Republic of China; and for the US and Japan from Datastream. 18 It should be noted that the US had no foreign exchange controls during the whole sample period, so that the current exercise on the degree of capital market integration reflects developments in the Pacific Basin countries. 19 All returns and inflation are continuously compounded. 20 Lags were added in order to induce whiteness of the residuals. In the case of pre-liberalisation sub-period for Singapore and post-liberalisation sub-period for Taiwan the null hypothesis for unit root could not be rejected at high lags only.

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

277

We proceeded to test for cointegration. We present first the bivariate cointegration results between the real interest rate of each Pacific Basin country vs. US and vs. Japan. We use the Johansen trace statistic which has been corrected for small sample bias Žsee Reimers, 1992..21 Thus, we use ŽT y nk . in Eq. Ž6. instead of T. The lag length is chosen by applying the Schwarz information criterion ŽSIC. on the undifferenced VAR developed by Schwarz Ž1978.. Reimers Ž1992. finds that the SIC does well in selecting the lag length. The results vs. US are shown in Panel A of Table 1. The hypothesis of zero cointegrating vectors ŽH 0 : r s 0., against the alternative of one or more cointegrating vectors, is easily rejected in every case except in the first sub-period in Japan, at the 5% level of significance. On the other hand, the hypothesis of at most one cointegrating vector ŽH 0 : r F 1. is not rejected. The results vs. Japan are reported in Panel B of Table 1. Once again the results are favourable to real interest rate parity. When the statistic is corrected for sample size, the hypothesis of at most one cointegrating vector ŽH 0 : r F 1. is not rejected, whilst the hypothesis of zero cointegrating vectors ŽH 0 : r s 0. is rejected in every case at the 5% level of significance. We have proceeded to the multivariate cointegration tests for real interest rates in each Pacific Basin country, the US and Japan. Table 2 reports the results of calculating the Johansen Maximum likelihood-ratio test statistics to define the dimensionality of the common stochastic trend process. Using a 5% significance level, we cannot reject the null hypothesis that there are zero cointegrating vectors in the first sub-period. This is not surprising as we have found in the bivariate case that the real interest rates of Japan and the US are not cointegrated. On the other hand, the results for the second sub-period show the hypothesis that there are two cointegrating vectors in the full three-dimensional system cannot be rejected in every case, apart from the group which includes Korea. This implies that there is one common stochastic trend and that any single rate is representative of the three rates, which indicates a considerable degree of integration amongst these markets. In the case of Korea, the hypothesis that only one cointegrating vector exists cannot be rejected, so that there are two common stochastic trends and a lower degree of capital market integration. We performed the same exercise for the second sub-period but included all the countries in one group. Namely, we included the US, Japan, Malaysia, Singapore, Taiwan and Hong Kong. We left Korea out because the frequency of the data is quarterly. The results, also presented in Table 2, show that we cannot reject the hypothesis that five stochastic trends are present, or that there is one single stochastic common trend, confirming the existence of integration amongst the capital markets of the Pacific Basin countries, the US and Japan. In general the results presented so far lend substantial support to real interest rate parity within the framework of cointegration. The results, however, do not

21

The trace test appears to be more robust to non-normality of errors compared to the maximal eigenvalue Žsee Cheung and Lai Ž1993. for Monte Carlo results on this issue.

278

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

Table 1 Bivariate cointegration tests for real interest rates Johansen test statistics H0 : r s 0

H0 : r F 1

Panel A: comparisons with US Singapore August 1973᎐December 1980 January 1981᎐September 1993

26.443 48.430

7.623 3.830

Japan January 1974᎐December 1980 January 1981᎐September 1993

18.594 20.987

5.596 5.022

Korea January 1972᎐April 1980 January 1981᎐March 1993

22.691 24.811

7.094 3.948

Malaysia Febuary 1982᎐September 1993

30.685

6.835

Hong Kong January 1976᎐September 1993 January 1981᎐September 1993

51.839 24.247

7.089 4.669

Taiwan January 1972᎐December 1980 January 1981᎐September 1993

21.175 28.270

4.885 3.279

Panel B: Comparisons with Japan Singapore August 1973᎐December 1980 January 1981᎐September 1993

29.747 20.504

7.546 5.307

Korea January 1972᎐March 1980 January 1981᎐March 1993

20.871 21.716

6.726 6.996

Malaysia Febuary 1982᎐September 1993

27.920

5.709

Hong Kong January 1976᎐September 1993 January 1981᎐September 1993

23.418 21.272

7.163 5.307

Taiwan January 1972᎐December 1980 January 1981᎐September 1993

25.026 27.198

7.472 7.171

Notes. If r denotes the number of significant vectors, then the Johansen trace statistics test the hypotheses of at most one and zero cointegrating vectors, respectively. The statistics include a finite sample correction Žsee Reimers, 1992.. The 5% critical value for H0 : r F 1 is 9.243 and H0 : r s 0 is 19.964 ŽOsterwald-Lenum, 1992..

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

279

Table 2 Multivariate cointegration tests of real interest rates Johansen test statistics H0 : r s 0 Panel A. Group: Pacific Basin country, US and Japan First sub-period Singapore 32.03 Korea 27.48 Taiwan 26.57

H1 : r F 1

9.27 14.87 11.25

Second sub-period Singapore 53.94a 21.96a Korea 36.77a 15.65 Malaysia 41.80a 21.39a Hong Kong 54.39a 24.45a a Taiwan 61.04 25.27a Group B: US, Japan, Malaysia, Singapore, Hong Kong and Taiwan H0 : H0 : H0 : H0 : H0 : H0 :

rs0 rF1 rF2 rF3 rF4 rF5

182.341a 126.233a 84.945a 49.515a 24.617a 6.691

H0 : r F 2

3.55 4.17 3.71

3.41 3.43 3.00 6.63 9.13 95% Quantile 102.139 76.069 53.116 34.910 19.964 9.243

Notes. For the precise sample periods for Panel A see Table 1. The sample period for Panel B is Febuary 1982᎐September 1993. If r denotes the numbers of significant cointegration vectors, then the Johansen statistics test the hypotheses of at most two, one, and zero cointegrating vectors using the trace. The statistics include a finite sample correction Žsee Reimers, 1992.. The 5% critical value for H0 : r F 2 is 9.24, for H0 : r F 1 is 19.96 and H0 : r s 0 is 34.91 ŽOsterwald-Lenum, 1992.. a Denotes significance at the 5%.

answer the issues relating to the change in the degree of capital market integration in post-liberalisation times, and whether Japan has come to dominate the region. 3.3. Impulse response analysis Answers to the above questions are provided by using innovation accounting, and in particular impulse response analysis, to examine how long it takes for the real interest rate parity to be re-established following a one standard deviation shock in one of the rates. The greater the speed of adjustment the greater the capital market integration. The lag structure of the VAR has been selected using SIC Žthe same criterion used in the cointegration exercise.. It is well known that SIC is a strongly consistent lag order selection criterion suited for the analysis of finite-lag order VAR models. Furthermore, various studies have concluded that the SIC performs best in small samples Žsee e.g. Lutkepohl, 1991., which is the case in our study. Before proceeding to the impulse response analysis we examined the non-diagonal

280

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

error variance matrix for each VAR and found, using the log-likelihood ratio statistic, that in all the cases we could not reject the null hypothesis that the shocks in different real interest rates are contemporaneously uncorrelated Žsee Appendix A.. In such a situation, Pesaran and Shin Ž1998. show that the orthogonalised impulse responses and the generalised impulse responses, which are invariant to the ordering of the variables in the VAR, coincide. Thus, we performed the exercise using Choleski decomposition to orthogonalise the innovations, knowing that the results will not be sensitive to the ordering of the variables. We explore the interactions of rP B C , rUS and r J due to a shock originating in the US i.e. an impulse in ⑀ US in Eq. Ž8.. We assume that the ordering of variables is rUS , rP B C and r J . Thus, the real interest rate in the US and the Pacific Basin country is permitted to have an instantaneous effect on the rate in Japan, but the latter can only have a lagged impact on the other rates. The results of the first subperiod are given in Table 3. Since we could not reject the null hypothesis that there are zero cointegrating vectors in the multivariate cointegration tests for that period, the analysis uses a bivariate framework. The first column shows the number of months it takes the real interest rate in each Pacific Basin country and that of the US to converge to their long-run equilibrium following an impulse in the US real interest rate; the second column shows the number of months it takes the real interest rate of each Pacific Basin country to converge with that of Japan following an impulse in the Japanese real interest rate.22 It can be seen that Singapore was equally integrated with the US and Japan, the interest rates taking 26 months to converge. In contrast, Taiwan and Korea were both better integrated with the US, the convergence taking 22 and 21 months, respectively. In the case of Taiwan the convergence with the Japanese interest rates was very slow, taking approximately 55 months. The results of the second sub-period are reported in Table 4. The speed of adjustment is faster in the second sub-period compared to the first, vs. both the US and Japan, in the case of Singapore and Taiwan. In Taiwan, the improvement in Table 3 Impulse response analysis for bivariate cointegrating systems ŽPacific Basin country and the US following an impulse in the US real interest rate; Pacific Basin country and Japan following an impulse in the Japanese real interest rate.: number of months for the real interest rates to converge

Singapore Korea Taiwan

Impulse in the US real interest rates Žmonths.

Impulse in the Japanese real interest Žmonths.

26 21 Ž7 quarters. 22

26 24 Ž8 quarters. 55

Notes. The analysis refers to the first sub-period, details of which are given in Table 1. The data in the case of Korea are quarterly and the speed is transferred into months by multiplying by 3. 22

The speed of adjustment refers to the impulse responses decaying up to 0.1.

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

281

Table 4 Impulse response analysis for cointegrating systems: ŽPacific-Basin country, US and Japan.: number of months for the real interest rate of each Pacific Basin country to converge with Ži. US, and with Žii. Japan, following an impulse in the US real interest rate Singapore Žmonths.

Korea Žmonths.

Malaysia Žmonths.

Hong Kong Žmonths.

Taiwan Žmonths.

US

24

19

27

18

Japan

18

30 Ž10 quarters. 24 Ž8 quarters.

13

21

17

Notes. The order of variables is the real interest rate of US, Pacific Basin country and Japan. The analysis refers to the second sub-period, details of which are given in Table 1. The data in the case of Korea are quarterly and the speed is transferred into months by multiplying by 3.

the speed of adjustment is incredible, especially vs. Japan Ž17 months reduced from 55 months.. Korea seems to be less integrated with the US in the second subperiod. One feature stands out clearly. All the countries in the sample are better integrated with Japan than with the US in the second sub-period. This finding is very interesting, as it confirms the other indicators which show that Japan’s influence is dominating the region. Furthermore, Malaysia is the most integrated country with Japan, followed by Taiwan and Singapore. 3.4. Multi¨ ariate Granger causality tests In this last set of tests we look at the short-run dynamics of the system of real interest rates by performing multivariate Granger-causality tests. Our objective in this exercise is to find out whether it is the real interest rate of the US or of Japan which drives the interest rates in the Pacific Basin countries. Applying the methodology suggested by Dolado and Lutkepohl Ž1996. and outlined in Section 2.3, we test whether past values of the US real interest rates Granger cause rates in the Pacific Basin country and vice versa i.e. with reference to Eq. Ž9. we test the restriction A12 Ž L. s 0 using the Wald test. We also test for the reverse causality from the Pacific Basin country to the US, i.e. whether A 21Ž L. s 0. We repeat the exercise for Granger causality between Japan and the Pacific Basin countries. We test whether A13 Ž L. s 0 for causality running from Japan to the Pacific Basin country; and A 31Ž L. s 0 for reverse causality. The results between the US and the Pacific Basin country are shown in Panel A of Table 5 and cover the second sub-period. At the 5% level of significance we find US real interest rates to Granger cause rates in four of the five countries, the exception being Taiwan. On the other hand, there is no causality from the Pacific Basin country to the US. These results show clearly that US real interest rates drive the interest rates in Singapore, Korea, Malaysia and Hong Kong.

282

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

Table 5 Granger causality tests using a multivariate approach between each Pacific Basin country, US and Japan Market

A12 Ž L. s 0

Panel A: Wald tests of causality between US and each Pacific Basin country Singapore 32.586a Ž0.007 = 10y2 . Korea 18.603a Ž0.004 = 10y1 . Malaysia 17.939a Ž0.003. Hong Kong 25.209a Ž0.007 = 10y1 . Taiwan 9.333 Ž0.407. Market A13 Ž L. s 0 Panel B: Wald tests of causality between Japan and each Pacific Basin country Singapore 16.034a Ž0.041. Korea 19.595a Ž0.006 = 10y1 . Malaysia 20.245a Ž0.001. Hong Kong 14.829a Ž0.038. Taiwan 17.517a Ž0.041.

A21 Ž L. s 0 12.475 Ž0.131. 1.267 Ž0.866. 9.746 Ž0.082. 12.356 Ž0.091. 10.101 Ž0.343. A31 Ž L. s 0 11.219 Ž0.189. 6.444 Ž0.168. 11.592a Ž0.040. 18.575a Ž0.009. 6.319 Ž0.708.

Notes. The tests cover the second sub-period details of which are given Table 1. For an explanation of the restrictions see Eq. Ž5.. Figures in parentheses are P-values. a denotes significance at the 5% level.

In Panel B of Table 5, we present the results of the Granger causality tests between Japan and each Pacific Basin country. It can be seen that real interest rates in Japan Granger cause interest rates in all the countries including Taiwan. At the same time, rates in Malaysia and Hong Kong affect Japanese rates. Taking together the results of both Panels, one can make the following points with regard to the interactions between the real rates in the Pacific Basin region, the US and Japan. First, the results show that both the US and Japanese rates drive interest rates in Hong Kong, Singapore, Malaysia and Korea. Secondly, the results show the importance of Japan in the case of Taiwan, as US rates do not Granger cause rates in Taiwan. Finally, the results show that there are close links between Japan and Malaysia, and between Japan and Hong Kong, as the direction of causality is a two way one in each case.

4. Summary and conclusions In this paper, we have examined the extent of capital market integration in a

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

283

group of Pacific Basin countries. Our main concern was to see whether there has been an increase in the degree of capital market integration with world markets, such as the US and Japan, following the deregulation of the region’s financial markets. Furthermore, whether the degree of capital market integration with Japan has been greater compared to that of the US, i.e. whether a Yen block has been created in the area. We examined these issues by looking at the equalisation of real interest rates using cointegration methodology which tests for a long-run relationship between interest rates within a band defined by transaction costs, taxes and capital controls. We have examined another indicator of integration which relates to the speed of adjustment of real rates to re-establish long-run equilibrium following a shock in one of the rates using innovation accounting analysis. Our final exercise concentrated on exploring the short-run dynamics through multivariate Granger-causality tests. In general, the use of multivariate approach in this paper is superior to the bivariate framework used in other studies, as the process takes into account the interaction of interest rates between the US and Japan in examining capital market integration between each one of these world markets and the Pacific Basin countries. The following conclusions have been derived from our analysis: First, the results from the cointegration exercise Žboth bivariate and multivariate. provide supportive evidence for real interest rate parity and capital market integration. Second, the results using the speed of adjustment of real interest rates following a shock as an indicator of capital market integration show that the degree of integration has increased in post-liberalisation period in Singapore and especially in Taiwan. Korea, the other country for which data exist for the pre-liberalisation period, seems to be less integrated with the US in the second sub-period. An interesting result which emerges from the analysis is the greater capital market integration with Japan than with the US in the second sub-period. It confirms the other indicators which show that Japan’s influence is dominating the region. The increased effects of Japan through time could be due not only to the financial liberalisation in the Pacific Basin countries but in Japan itself. This result, which is based on multivariate approach, is in contrast to the bivariate analysis examined in Phylaktis Ž1997., which used a different method to estimate the speed of adjustment within the cointegration methodology and showed that capital market integration was greater with the US than with Japan. Thus, taking into account the interaction of interest rates between the US and Japan in the examination of capital market integration in the Pacific basin countries revealed the increasing financial influence of Japan. Finally, the results of the multivariate Granger-causality tests show that both Japanese and US interest rates have been driving the rates in Singapore, Hong Kong, Malaysia and Korea, while only Japanese interest rates have been driving rates in Taiwan. In Malaysia and Hong Kong, however, there is also reverse causation from both of these countries to Japan. Thus, the results of this exercise

284

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

point also to the fact that some of the countries in the Pacific Basin region have developed close links with Japan and in some of them, such as Taiwan, shocks originating in the US are transmitted via Japan. In this paper, we have established that there are extensive real interest rate linkages with world financial markets even in countries like Taiwan and to a lesser degree Korea, where there are still extensive foreign exchange controls. Important factors which might have helped investors to move funds across borders are the misinvoicing of exports and imports and leads and lags.23 The open character of these economies in terms of exports and imports, together with the low protectionism in current account transactions, have created many such opportunities for evading controls. Mathieson and Rojas-Suarez Ž1993. have also found in their study, on the experience of industrialised and developing countries with capital controls, that controls were most effective when they were combined with trade restrictions. Furthermore, the existence of illicit trade in drugs and other goods in some of the Pacific Basin countries has provided additional channels through which funds could be moved across borders.24 The substantial degree of capital market integration in the Pacific Basin Region found in the paper, both in terms of the comovement of real interest rates and in terms of the speed of adjustment of interest rates following a shock, indicates that the effectiveness of monetary policy to affect these economies might be limited in the long run.25

Acknowledgements Comments from two anonymous referees are gratefully acknowledged. The author wishes to thank Hashem Pesaran and the participants of the 1997 Econometric Society European Meeting in Toulouse, France, for helpful suggestions. The paper was written while the author was a visiting consultant at the Research Department of the International Monetary Fund.

Appendix A: Examination of the non-diagonal error ¨ ariance matrix for each VAR We test the hypothesis that the shocks in the real interest rates rP B C , rUS and r J , are not contemporaneously correlated by using the log-likelihood ratio statistic 23

Kamin Ž1993. finds under-invoicing and over-invoicing of trade transactions has often been a key vehicle for large scale capital flows in developing countries with extensive capital controls. 24 The existence of a substantial black market for dollars as manifested by the high and variable black market premium Žsee Phylaktis and Kasimmatis, 1994b. corroborates the development of such channels. 25 This is subject to the condition that the interest rates used in the analysis are representative of the economy-wide interest rates facing most firms and consumers Žsee Chinn and Dooley, 1995..

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

285

LR Ž H0rH1 . s y2 Ž LLU y LL R . where LLU and LL R are the maximised values of the log-likelihood function under H 1Žthe unrestricted model. and under H 0 Žthe restricted model., respectively. LR is asymptotically distributed as a ␹ 2 variate with three degrees of freedom. The 95 and 99% critical values are 7.815 and 11.345, respectively. In the first sub-period where there is a bivariate VAR of Ži. rB P C ,rUS and Žii. rP B C ,r J , the ␹ 2 variate has two degrees of freedom. The 95 and 99% critical values are 5.991 and 9.210, respectively. First Period: Bivariate VAR models of rP B C and rUS ; and rP B C and r J .

LR

Singapore US 3.442

Singapore Japan 0.060

LR

Korea US 0.368

Korea Japan 5.262

LR

Korea US 0.060

Taiwan Japan 9.200

Second Period: Trivariate VAR models of rP B C ,rUS and r J .

LR

Singapore US Japan 7.868

Korea US Japan 6.365

Malaysia US Japan 5.341

Hong Kong US Japan 2.815

Taiwan US Japan 10.419

References Bhoocha-Oom, A., Stansell, S.R., 1990. A Study of International Financial Market Integration: An Examination of the US, Hong Kong and Singapore Markets. J. Bus. Fin. Account. 17, 193᎐212. Bosner-Neal, C., Roley, V.V., 1994. Are Japanese Interest Rates Too Stable? J. Int. Mon. Fin. 13, 291᎐318. Cheung, Y.W., Lai, K.S., 1993. Finite-sample sizes of Johansen’s Likelihood Ratio for Cointegration. Oxford Bull. Econ. Stat. 55, 313᎐328. Chinn, M.D., Dooley, M.P., 1995. Financial Repression and Capital Mobility: why Capital Flows and Covered Interest Rate Differentials fail to measure Capital Market Integration? Natl. Bureau Econ. Res., Working Paper, 5347. Chinn, M.D., Frankel, J.A., 1994. Financial Links around the Pacific Rim: 1982᎐1992. In: Glick, R., Hutchison, M.M. ŽEds.., Exchange Rate Policy and Interdependence: Perspectives from the PacificBasin. Cambridge, New York, pp. 17᎐47. Chinn, M.D., Maloney, W.F., 1996. Financial and Capital Account Liberalisation in the Pacific Basin: Korea and Taiwan during the 1980s. University of California Santa Cruz, Working Paper 352.

286

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

Cumby, R.E., Mishkin, M.S., 1986. The International Linkage of Real Interest Rates: The European-US Connection. J. Int. Mon. Fin. 5, 5᎐23. Dolado, J.J., Lutkepohl, H., 1996. Making Wald tests for cointegrated VAR systems. Econometr. Rev. 15, 369᎐386. Dooley, M.P., Mathieson, D.J., 1994. Exchange Rate Policy, International Capital Mobility, and Monetary Policy Instruments. In: Glick, R., Hutchison, M.M. ŽEds.., Exchange Rate Policy and Interdependence: Perspectives from the Pacific-Basin. Cambridge, New York, pp. 68᎐95. Edwards, S., Khan, M., 1985. Interest rate determination in developing countries: a conceptual framework. Staff Pap. Int. Monet. Fund 32, 377᎐403. Enders, W., 1995. Applied Econometric Time Series. Wiley, US. Engle, R.F., Granger, C.W.J., 1987. Cointegration and error correction: representation, estimation and testing. Econometrica 55, 251᎐277. Faruqee, H., 1992. Dynamic capital mobility in Pacific-Basin developing countries. Staff Pap. Int. Monet. Fund 39, 706᎐717. Frankel, A.J., 1993. Measuring international capital mobility: a review. Am. Econ. Rev. 82, 197᎐202. Fry, M., 1990. Nine Financial Sector Issues in Eleven Asian Developing Countries’, University of Birmingham, International Finance Group Working Papers. Goodwin, B.K., Grennes, T., 1994. Real interest rate equalisation and the integration of international financial markets. J. Int. Mon. Fin. 13, 107᎐124. Granger, C.W.J., 1981. Some properties of time series data and their use in econometric model specifications. J. Econometr. 28, 121᎐136. Gros, D., 1987. The effectiveness of capital controls: implications for monetary autonomy in the presence of incomplete market separation. Staff Pap. Int. Monet. Fund 34, 621᎐642. Haque, N., Montiel, P., 1991. Capital mobility in developing countries: some empirical tests. World Dev. 19, 1391᎐1398. Ito, T., 1988. Use of Žtime-domain. vector autoregressions to test uncovered interest parity. Rev. Econ. Stat. 70, 296᎐305. Johansen, S., 1988. Statistical analysis of cointegration vectors. J. Econ. Dynam. Control 12, 231᎐254. Johansen, S., 1991. Estimation and hypothesis testing of cointegration vectors in gaussian vector autoregressive models. Econometrica 59, 1551᎐1580. Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on cointegration with applications to the demand for money. Oxford Bull. Econ. Stat. 52, 169᎐210. Kamin, S.B., 1993. Devaluation, exchange controls, and black markets for foreign exchange. J. Dev. Econ. 40, 151᎐169. Khan, M.S., Haque, N., 1985. Foreign borrowing and capital flight: a formal analysis. Staff Pap. Int. Monet. Fund 32, 606᎐628. Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in non-linear multivariate models. J. Econometr. 74, 119᎐147. Lutkepohl, H., 1991. Introduction to Multiple Time Series Analysis. Springer-Verlag, New York. Lutkepohl, H., Reimers, H.E., 1992. Impulse response analysis of cointegrated systems. J. Dynam. Control 16, 53᎐78. Mark, N.C., 1985. Some evidence on the international inequality of real interest rates. J. Int. Mon. Fin. 4, 189᎐208. Mathieson, D.J., Rojas-Suarez, L., 1993. Liberalisation of the Capital Account: Experiences and Issues. Occasional Paper, International Monetary Fund, No. 103. Merrick, J.J., Saunders, A., 1986. International expected real interest rates: new tests of the parity hypothesis and US fiscal-policy effects. J. Monet. Econ. 18, 313᎐322. Mishkin, F.S., 1984a. Are real interest rates equal across countries? An empirical investigation of international parity conditions. J. Fin. 39, 1345᎐1357. Mishkin, F.S., 1984b. The real interest rate: A multi-country empirical study. Can. J. Econ. 17, 283᎐311. Mishkin, F., 1995. Non-stationarity of regressors and tests on real-interest-rate behaviour. J. Bus. Econ. Stat. 13, 47᎐51.

K. Phylaktis r Journal of International Money and Finance 18 (1999) 267᎐287

287

Osterwald-Lenum, M., 1992. A note with quantiles of the asymptotic distribution of the maximum likelihood cointegration test statistics. Oxford Bull. Econ. Stat. 54, 461᎐471. Otani, I., Tiwari, S., 1981. Capital controls and interest rate parity. Staff Pap. Int. Monet. Fund 28, 793᎐815. Pesaran, M.H., Shin, Y., 1998. Generalised Impulse Response Analysis in Linear Multivariate Models. Mimeo, Department of Applied Economics, University of Cambridge, Economic Letters, forthcoming. Phylaktis, K., 1988. Capital controls: the case of Argentina. J. Int. Mon. Fin. 7, 303᎐320. Phylaktis, K., 1997. Capital Market Integration in the Pacific Basin Countries: An Analysis of Real Interest Rate Linkages. Pacific-Basin Finance Journal 5, 195᎐213. Phylaktis, K., Kasimmatis, Y., 1994a. Does the real exchange rate follow a random walk? The Pacific-Basin perspective. J. Int. Mon. Fin. 13, 476᎐495. Phylaktis, K., Kasimmatis, Y., 1994b. Black and official exchange rates in the Pacific-Basin countries: an analysis of their long-run dynamics. Appl. Econ. 24, 399᎐407. Phylaktis, K., Wood, G.E., 1984. An Analytical and Taxonomic Framework for the Study of Exchange Controls. In: Black, J., Dorrance, G.S. ŽEds.., Problems of International Finance. Macmillan, London, pp. 149᎐166. Reimers, H.E., 1992. Comparisons of tests for multivariate cointegration. Stat. Pap. 33, 335᎐346. Reisen, H., Yeches, H., 1993. Time-varying estimates on the openness of the capital account in Korea and Taiwan. J. Dev. Econ. 41, 285᎐305. Schwarz, G., 1978. Estimating the dimension of a model. Ann. Stat. 6, 461᎐464. Stock, J.H., 1987. Asymptotic properties of least squares estimators of cointegrating vectors. Econometrica 55, 1035᎐1056. Stock, J.H., Watson, M.W., 1988. Testing for common trends. J. Am. Stat. Assoc. 83, 179᎐197. Tavlas, G., Ozeki, Y., 1992. The Internationalisation of Currencies: An Appraisal of the Japanese Yen. Occasional Paper, International Monetary Fund, No. 90. Yuan, T., 1986. Capital Flows among Pacific-Basin Countries. In: Tan, A.H.H., Kapur, B. ŽEds.., Pacific Growth and Financial Interdependence. Allen and Unwin.