Interest Rate Spillover Effects Across Asian Pacific Countries VAR

Document Sample
Interest Rate Spillover Effects Across Asian Pacific Countries VAR Powered By Docstoc
					      A Network View of Capital Market Integration and
         Disintegration- An Example by VAR Model

                              HSIAO-TANG HSU
                     Investment Portfolio Management Group
                                 IndyMac Bank
                  12728 Moorpark St, #14, Studio City, CA, 91604
                                      USA



Abstract: Traditional empirical work on market integration based on looking at two
countries at a time. This paper uses interaction between members (spillover effect) to
examine the market integration and disintegration. Asian Currency Crisis in 1997 was
said to happen because high market integration made crisis in one country contagious
(through spillover effect) to other countries and caused the disaster. This paper
chooses 7 Asian countries –Hong Kong, Singapore, Taiwan, Malaysia, Korea,
Thailand and Indonesia –to study the integration and disintegration through interest
rate spillover effect (lead-lad relationship) among them. This paper examines how the
spillover effect worked before the crisis and how this relationship changed after the
crisis. VAR (Vector Autoregressive) model is used to analyze the multivariate
spillover effects. This paper examines capital market integration periods before and
after crisis separately. We found that before Asian currency crisis, the market
integration is strong. After Asian currency crisis, regional integration decreased and
the spillover effects among the members in this region changed. The only exception is
that Taiwan had the same spillover effect as before.

Key-Words: Capital market integration, VAR.
1 Introduction                                rate (i*) and then there is no chance for
                                              arbitrage.     In     financial    market
      Asian countries are facing a            integration point of view, if there exists
process      of    financial      market      interest rate differential in two
liberalization and growing capital            countries, then it represents the degree
flows over the past 20 years. For             of market segmentation and obstacle of
economic development, Asian Pacific           capital mobility. Therefore, we can use
countries like Korea, Hong Kong,              interest rate differential as a
Singapore and Taiwan, etc are                 measurement of international capital
gradually     deregulating     domestic       mobility. In Frankel(1991), he uses
financial     market    and     relaxing      interest rate differentials to measure
restrictions on international capital         capital mobility and market integration
flows from 1980, which makes Asian            in the 1980’s for 25 countries. The
Pacific area one of the fastest               international interest rate is US interest
economic growth area. Therefore, the          rate. He calculated the means and
influence of financial liberalization on      standard errors and his empirical result
Asian countries is particularly worthy        is quite consistent with his own prior
of being studied.                             expectations: the five closed LDC’s
                                              have the highest variability and five
       The changing degree of capital         open Atlantic countries have the lowest
mobility—broadly defined as the               variability. However, there are some
degree of linkage between interaction         exceptions. For example, he found that
of domestic and international interest        though France has very strict financial
rates—can help us to understand the           regulations during that time, it has
influence on market interaction and           smaller interest rate differentials than
integration after financial liberalization.   Japan,      the      Netherlands       and
With deregulation or relaxation in            Switzerland—countries known as
financial restrictions (i.e. financial        highly free of capital controls.
liberalization), capital inflow or            Moreover, he found that no interest
outflow increases (the interest rates         rate differential is zero of any country.
interaction     increases),    and     the    It means that there still exits barriers of
differential between domestic and             capital mobility in these countries.
international interest rates will
gradually get reduced (by Interest Rate           2. Problem Formulation
Parity). According to Interest Rate
Parity (IRP), if we assume rational                This paper will use VAR as the
expectation, domestic interest rate (i)       structure of estimation of spillover
should be equal to international interest     effect. To study cointegration of
interest rates between these countries,           Korea, Thailand and Indonesia. To
we use VAR to analyze the                         construct VAR model,
multivariate time series behavior. We
also substitute Japanese interest rate            Y = XB + U            , where
for US interest rate as benchmark
interest rate because Japanese interest           Y= [Yhk , Ysingapore , Ytaiwan , Ymalaysia ,
rate might have more impact on Asian              Yindonesia ,Y Korea ,Y Thailand ] (T×K),
countries.
                                                  X = [1 , Yt-1 , Yt-2 , ……. ,Yt-p ]
                                                  (T×(1+K*P)),
      As we know that Asian currency
                                                  B = [v , A1 , A2 , …. ,Ap ]’
crisis happened in 1997. According to
                                                  ((1+K*P)×K),
Figure 1 , we can find that there is
                                                  U = [uhk usingapore utaiwan      umalaysia
possible a structure change around                uindonesia u Korea u Thailand ]    (T×K),
1997( index 140). The volatility of HK,
                                                  And T=168, K=7
Singapore, Indonesia and Malaysia
                                                  To get B, we use vec operator:
became unusually high than before.
The only exception is Taiwan. To
                                                  y=vec(Y)=vec(XB)+vec(U)=(I ⊗ X)ve
study this structural change (Asian
                                                  c(B)+vec(U)=(I ⊗ X) β +u
Currency Crisis in 1997), therefore,
                                                  β =(I ⊗ (X’X) −1 X’)y
besides dealing with the whole period
data, perhaps we can split the data in to
                                                  3. Problem Solution
two period: one is from January 1986
to December 1996, and the other
period is from January 1997 to
                                                  PART I Pre-Currency Crisis Period
December 1999. We want to know
                                                  (Jan 1986 ~ Dec 1996)
whether the property of the times
series changes due to this structure
                                                  3.1     Unit Root Test:
change and whether this structure
change has huge influence on the
                                                              First of all, we want to
spillover effect among these countries.
                                                  know whether there are unit roots in
Finally we want to discuss a little bit of
                                                  this period. Asymptotic Critical Values
why Taiwan is kind of an exception.
                                                  under α=0.5 for t, F and KPSS Tests
                                                  are –2.86 ,4.59 and 0.146 respectively.
VAR Model

                                                  3.2     Identification: Lag
     Let Yt , j = it , j , the interest rate in
                                                          Selection
country j at time t, where j=Hong
Kong, Taiwan, Singapore, Malaysia,
                                                              We found that minimum
     AIC , HQ, FPE and SC goes to lag                        confidence interval are –2.83 for t test,
     1. Therefore we choose VAR(1)                           4.59 for F test and 0.146 for KPSS test.
     for pre-period model.
                                                                It tell us that the residuals are I(1).
                                                                Therefore, there exists
3.3          Estimation (Pre                                    cointegration and we can estimate
             period) : 01/1986 –                                the model directly.
             12/1997
                                                             Part II Post-Currency Crisis Period
                                                             (Jan 1997 ~ Dec 1999)
ASIAN PACIFIC COUNTRIES
               INTEREST RATE                                 3.5. Unit Root Test
               SPILLOVER EFFECT
                                                                   Asymptotic Critical Values
     Hong     Singa Taiwa Malay Indon Korea Thail            under α=0.5 for t, F and KPSS Tests
t    Kong     pore    n       sia    esia           and
                                                             are –2.83 ,4.59 and 0.146 respectively.
HK 0.275( 0.519( 0.595( 0.709( 0.581( 0.472( 0.623(
                                                             We found that the data still have unit
     0.03)* 0.01)* 0.03)* 0.05)* 0.12)* 0.05)* 0.02)*
     *        *       *       *      *      *       *        root; therefore, we need to test
Sing 0.798( 0.275( -0.227 -0.454 -1.49( 0.382( 0.384(        cointegration.
apor 0.07)* 0.02)* (0.09) (0.06) 0.14)* 0.02)* 0.01)*
e    *        *       *       **     *      *       *
Tai 0.812( 0.016( 0.204( 0.655( 0.352( 0.286( 0.625(
                                                             3.6. Identification: Lag
wan 0.04)* 0.02)      0.02)* 0.07)* 0.11)* 0.01)* 0.01)*     Selection
     *                *       *      *      *       *
Mal -0.043 0.158( -0.042 0.365( 0.169( 0.035( 0.451(
aysi (0.03) 0.13)     (0.03) 0.02)* 0.08)* 0.12)    0.05)*
                                                                Because all the minimal criteria
a                             *      *              *           goes to two, so we choose VAR(2).
Indo 0.022( -0.13( -0.16( 0.037( 0.412( 0.023( 0.23(0.
nesi 0.01)    0.11)   0.12)   0.02)* 0.13)* 0.25)   08)**
                                                             3.7. Estimation (post
a                                    *
Kor 0.075( 0.19(0. 0.752( 0.612( 0.245( 0.258( 0.542(        period): 01/1997 –12/1999
ea   0.12)    11)     0.05)* 0.02)* 0.02)* 0.04)* 0.11)*
                      *       *      *      *       *
Thai 0.012( -0.034 0.017( 0.521( 0.301( 0.033( 0.274(
                                                             ASIAN PACIFIC COUNTRIES
land 0.02)    (0.02) 0.02)    0.03)* 0.03)* 0.02)   0.03)*             INTEREST RATE
                              *      *              *
                                                                       SPILLOVER EFFECT


3.4          Diagnostic:
             Cointegration Test


         The critical values in 95%
      Hong    Singa Taiwa Malay Indon Korea Thail
                                                               3.8. Diagnostic:
t     Kong    pore    n       sia     esia            and
                                                               Cointegration test
t-1

HK 0.302( 0.036( 0.462( -0.043 0.145( 0.044( -0.028
      0.03)* 0.08)    0.05)* (0.13) 0.23)     0.15)   (0.18)
                                                                    The critical values in 95%
      *               *                                        confidence interval are –2.83 for t test,
Sing 0.177( 0.307( 0.074( 0.592( 0.385( 0.233( 0.75(0.         4.59 for F test and 0.146 for KPSS test.
apor 0.12)    0.15)* 0.04)    0.14)* 0.01)* 0.84)     8)
e             *               *       *
Tai 0.65(0. 0.248( 1.09(0. 0.079( 0.069( 0.45(0. 0.32(0.            According to the result, Taiwan
wan 03)** 0.64)       17)** 0.24)     0.21)   62)     87)      accepted the null hypothesis and is not
Mal -0.058 -0.8(0. 0.049( 0.645( 0.045( 0.025( 0.021(
                                                               cointegrated with other countries. If we
aysi (1.51) 79)       0.21)   0.01)* 0.12)    0.22)   0.18)
a                             *                                look at figure 1 again, we found that all
Indo 0.040( 0.026( -0.001 0.228( 0.287( 0.069( 0.112(          the countries have a peak around 1997
nesi 0.07)    0.04)   (0.01) 0.32)    0..03)* 0.51)   0.28)
                                                               (after index 140) except for Taiwan.
a                                     *
Kor 0.055( 0.049( 0.062( 0.045( 0.027( 0.647( 0.036(           The intuition is that Taiwan is not
ea    0.22)   0.18)   0.38)   0.18)   0.34)   0.08)* 0.31)     affected by Asian Currency crisis to
                                              *
                                                               the same degree with other countries.
Thai 0.15(0. 0.72(0. 0.054( 0.043( 0.057( 0.064( 0.288(
land 16)      86)     0.11)   0.15)   0.45)   0.29)   0.01)*
                                                      *        4.   Conclusion
t-2

HK 0.551( 0.045( 0.725( 0.124( 0.079( 0.25(0. 0.012(                We found out that after currency
      0.11)* 0.19)    0.06)* 0.28)    0.41)   34)     0.56)
                                                                    crisis, the spillover effect
      *               *
Sing 0.03(0. 0.771( 0.014( 0.53(0. 0.85(0. 0.47(0. 0.54(0.          changes.
apor 02)      0.04)* 0.31)    02)** 15)** 51)         87)                 1. In post-period, the model
e             *
                                                                             changes     to    VAR(2)
Tai 0.59(0. 0.044( 0.431( 0.084( 0.146( 0.039( 0.075(
wan 14)** 0.12)       0.12)* 0.14)    0.28)   0.31)   0.35)                  instead of VAR(1).
                      *                                                   2. HK has spillover effect
Mal 0.033( 0.015( 0.012( 0.367( 0.025( 0.51(0. 0.073(
                                                                             to Taiwan.
aysi 0.03)    0.32)   0.13)   0.02)* 0.41)    63)     0.18)
a                             *
                                                                          3. Singapore      only   has
Indo 0.028( 0.044( 0.441( 0.034( 0.589( 0.024( 0.147(                        spillover     effect   to
nesi 0.02)    0.12)   0.51)   0.21)   0.02)* 0.12)    0.25)                  Malaysia and Indonesia.
a                                     *
                                                                          4. Malaysia doesn’t have
Kor -0.025 0.035( 0.024( 0.51(0. 0.019( -0.234 0.25(0.
ea    (0.13) 0.42)    0.36)   84)     0.61)   (0.10) 47)                     any spillover effect to
                                              **                             others.
Thai 0.132( 0.28(0. 0.002( 0.34(0. -0.11( 0.12(0. -0.384
                                                                          5. Thailand doesn’t have
land 0.14)    45)     0.35)   52)     0.22)   58)     (0.02)
                                                      **                     any spillover effect to
                                                                             others.
                                                                          6. Indonesia doesn’t have
                                                                             any spillover effect to
             others.                        Reference
          7. The spillover effect of
             Taiwan stays the same:         Bollersler, Tim, 1992, "ARCH
             Taiwan has significant         Modeling in Finance-- a review of the
             spillover effect to HK.        theory and empirical evidence",
                                            Journal of Econometrics, Vol. 52, pp.
     The reason why these spillover
                                            5-59.
effects changes is probably because
                                            Bollerslev, Tim, 1986, "Generalized
that the FDI and trade among these
countries shrunk after currency crisis.     Autoregressive Conditional
During currency crisis, most of the         Heteroskedasticity", Journal of
countries lost huge amount of foreign       Economics, Vol. 31, pp. 307-327.
exchange      reserves      (Table 15);     Donald J. Mathieson and Liliana
therefore, government couldn’t use
                                            Rojas-Suarez, "Liberalization of the
foreign     exchange       reserve like
                                            Capital Account Experiences and
pre-period to finance its fiscal debt.
The only way was to use the money           Issues", IMF, March 1993,pp. 166-17.
once was used for FDI to liquid             Edwards, Sebastian and Moshin S.
government’s debt. Since the capital        Khan, "Interest Rate Determination in
flows became smaller, the spillover         Developing Countries: A Conceptual
effects in post period became not as        Framework," Staff Papers,IMF, Vol.
significant as that in pre-period.
                                            32 (September 1985), pp. 377-403.
                                            Engle, Robert F., 1982,
        The reason why Taiwan’s
spillover effect stays the same might       "Autoregressive Conditional
be explained by its survival in currency    Heteroskedasticity with Estimates of
crisis. During the whole currency crisis,   the Variance of United Kingdom
Taiwan had the slightest suffering from     Inflation", Econometrica, Vol. 50,
this unexpected shock in these
                                            No.4, pp. 987-1007.
countries. By the end of December
                                            Hamid Farugee, "Dynamic Capital
1997, New Taiwan dollar weakened
                                            Mobility in Pacific Basin Developing
moderately and depreciated about only
10 percent, which the level was only        Countries, estimation and Policy
one-third or fifty percent compared to      Implications," IMF Staff Papers, Vol.
ASEAN-4 countries. We will illustrate       39, No. 3 (September 1992), pp.
it more in detail next.                     707-717.