Does fisher hypothesis hold for the East Asian Economies? An application of panel unit root tests.

Ling, Tai-hu ; Liew, Venus Khim-sen ; Wafa, Syed Azizi Wafa Syed Khalid 等

INTRODUCTION

Irving Fisher hypothesized that there should be a long-run relationship in the adjustment of nominal interest rate corresponding to changes in expected inflation. If the Fisher hypothesis holds, then short-term interest rates will be an efficient predictor of future inflation (Granville and Mallick, 2004). More importantly, the monetary authority will then be able to influence long-term interest yields in order to enhance macroeconomic stabilization. Owing to its importance, the hypothesis has been subjected to rigorous research (Evans and Lewis, 1995; Daniels et al., 1996; Payne and Ewing, 1997; Lee et al., 1998; Koustas and Serletis, 1999; Cooray, 2002; Fahmy and Kandil, 2003; Granville and Mallick, 2004, just to name a few). One commonly adopted method to scrutinize the hypothesis is to examine the stationarity of the real interest rates. In this respect, if the hypothesis holds, then the real interest rate should be stationary. Empirical findings obtained from this approach are abundant but inconclusive thus far; see Cooray (2003) and Johnson (2006) who provide excellent overviews of the theoretical and empirical issues on the Fisher effect.

One well-accepted explanation of the contrasting evidence is the low power of conventional unit root tests with the relatively short span of data employed (Rapach and Wohar, 2002; Baharumshah et al., 2005). It is expected that with a longer span of data, the power of test could be improved, thereby yielding more reliable results. However, long data sets are normally unavailable. (1) An alternative solution to circumvent the problem is to perform panel analysis, which has higher power. (2) In this regard, most of the East Asian economies have a history of about half a century since independence.

Moreover, by pooling the data, the analysis can consider cross-country financial markets interactions, which need to be appropriately dealt with in this era of increasing international markets globalization and integration. Wu and Chen (1998, 2001) and Holmes (2002), for instance, demonstrated that by exploiting cross-country variations of the data in the estimation, panel analysis can yield higher test power than conventional unit root tests. Owing to its usefulness, recent studies have adopted panel analysis to investigate the stationarity of nominal interest rates (for instance, Wu and Chen, 2001) and real interest parity (Holmes, 2002; Baharumshah et al., 2005), just to mention a few. However, to the best of our awareness, panel analysis is yet to be applied in the context of the Fisher hypothesis. This study tests the long-run validity of the Fisher hypothesis using panel unit root tests. Specifically, this note aims to examine whether the Fisher hypothesis holds for the East Asian economies. East Asia is a fascinating economic region which has undergone rapid economic transformation and experienced spectacular growth over the past four decades. This study includes ASEAN-5 (Malaysia, Singapore, Taiwan, Thailand and the Philippines), China, Hong Kong and Taiwan, Japan and South Korea, which have strong trade and economic relationship. The intra-regional trade shares of these economies amounted to over one-half of their total trade in 2005 (United Nations, 2008). The combined merchandise exports of these East Asian economies amounted to over three trillion USD, accounting for one-quarter of the world exports in 2005-2006. The extraordinary growth of these economies in the recent decades and the important roles they play in the international trade have put East Asia under the spotlight of economic research (see for instance, Sarel, 1996). Among others, Baharumshah et al. (2005) recently documented evidence of the real interest rate parity by examining the stationarity of real interest rate differentials of East Asian economies. Ling (2008) argues that the existence or non-existence of the real interest rate parity in these economies can be affected by the soundness of the Fisher hypothesis. If the hypothesis does not hold, then the resultant real interest rate differentials will not reflect the actual international financial linkages. Thus, it is important to verify the validity of the Fisher hypothesis in the case of these East Asian economies. To accomplish this task, the stationarity of 10 East Asian economies' real interest rates are examined using few commonly adopted panel unit root tests developed by Maddala and Wu (1999), Choi (2001), and Im et al. (2003).

The remainder of this note is structured as follows: The next section describes the data and methodology employed in this study. This is followed by results and interpretation. The final section concludes this study.

DATA AND METHODOLOGY

This study analyses the stationarity of real interest rates of 10 East Asian economies, namely China, Hong Kong, India, Indonesia, Japan, Malaysia, Singapore, South Korea, Taiwan, Thailand, and the Philippines. The sample data, which are obtained from the International Financial Statistics, Asian Development Bank and Central Banks, span from the first quarter of 1987 to the third quarter of 2006 (1987:Q1 to 2006:Q3). Various short-term interest rates are considered, depending on data availability: deposit rate (China), money market rate (India, Indonesia, South Korea, Taiwan and Thailand), and 3-Months Treasury bill rate (Japan, Malaysia and the Philippines). Following Atkins and Coe (2002), the inflation rate, [[pi].sub.t], is defined as the percentage change of the quarterly consumer price index multiplied by four. The expected inflation is then obtained by estimating an autoregressive model for inflation rate as shown below:

[[pi].sup.e.sub.t] = [[alpha].sub.0] + [k=3.summation over (t-i)] [[alpha].sub.i][[pi].sub.t-i] + [[eta].sub.t] (1)

where [[pi].sup.e.sub.t] is the expected inflation and [[pi].sub.t-i] is the inflation rate calculated from the CPI. Expected inflation rate is defined by [[??].sub.t] = [[pi].sup.e.sub.t].

The real interest rate for each economy, in turn, is obtained by subtracting the expected inflation rate from the nominal interest rate. For the Fisher hypothesis to hold, the resultant ex ante real interest rate should be stationary. To test for stationarity, several panel unit root tests due to Im et al. (2003), Maddala and Wu (1999) and Choi (2001) are adopted in this study. For comparison purpose, the conventional univariate augmented Dickey-Fuller (ADF) and its improved version known as Generalized Least Squares augmented Dickey-Fuller (ADF-GLS, due to Elliott et al., 1996; see also Ng and Perron, 2001) unit root tests are included in this study.

Im et al. (2003) panel unit root test

Im et al. (2003) proposed a t-bar statistic, which is based on the average of the individual cross-sectional ADF t-statistics, to examine the unit root hypothesis for panels. (3) In particular, the test is performed by combining individual unit root tests to derive their panel counterpart. Im et al. (2003) based their panel unit root test on a separate ADF test for each cross section (in our case, country) in the panel. Then the average of the t-statistics of individual ADF statistics is adjusted to obtain the unit root test statistic for the panel, namely the t-bar statistic. For a sample of N groups observed over T time periods, the panel unit root regression of the conventional ADF test is written as

[DELTA][y.sub.it] = [[alpha].sub.i] + [[beta].sub.i][y.sub.it] + [[p.sub.t].summation over (j=1)] [[gamma].sub.ij] [DELTA][y.sub.it-j] + [e.sub.it], i = 1, ..., N, t = 1, ..., T (2)

where [y.sub.it] is the real interest rate, [DELTA][y.sub.it] = [y.sub.it]-[y.sub.it-1], [[alpha].sub.i], [[beta].sub.I] and [[gamma].sub.ij] are the parameters to be estimated, and [e.sub.it] stands for disturbance terms.

The null hypothesis of the Im et al. (2003) test is characterized as:

[H.sub.0]: [[beta].sub.i]: 0 for all i (3)

against the alternatives that all series are stationary processes

[H.sub.1]: [[beta].sub.i] < 0, i = 1, 2, ..., [N.sub.1]; [N.sub.2] + 2, ..., B (4)

This alternative hypothesis allows for [[beta].sub.i] to differ across groups and is more general than the uniform alternative hypothesis, namely [[beta].sub.i] = [beta] < 0 for all i.

To test the hypothesis, Im et al. (2003) proposed a standardized t-bar statistic given by

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)

where

[bar.[t.sub.NT]] = [N.sup.-1] [N.summation over (i=1)] [t.sub.i], r([p.sub.i], [[beta].sub.i]) (6)

and [t.sub.i]([p.sub.i], [[beta].sub.i]) is the individual t-statistic for testing [[beta].sub.i] = 0 for all i. E[[t.sub.i], r([p.sub.i], 0)| [[beta].sub.i] = 0] and Var[[t.sub.i], T([p.sub.i], 0)[absolute value of [[beta].sub.i] = 0] can be found in Table 2 of Im et al. (2003). Since E[[t.sub.i], r([p.sub.i], 0)|[[beta].sub.i] = 0] and Var[[t.sub.i], T([p.sub.i], 0)| [[beta].sub.i] = 0] vary as the lag length in the ADF regression varies, in practice, we are restricted implicitly to using the same lag length in all the individual ADF regressions. Under the null hypothesis, the standardized t-bar statistic [[psi].sub.t] is asymptotically distributed as a standard normal distribution, [psi] ~ N(0, 1).

There are several advantages of using the Im et al. (2003) panel unit root test as compared to previously developed panel-based unit root tests (Quah, 1992; 1994; Levin and Lin, 1993). First, it takes into account heterogeneity across countries in two aspects, comprising of individual-specific effects and different patterns of residual serial correlations. Second, the proposed t-bar statistics allow for residual serial correlation and heterogeneity of the dynamics and error variances across time series data. Therefore, the Im et al. (2003) panel unit root test is adopted here to examine the validity of the Fisher hypothesis for the East Asian economies, allowing for heterogeneity and contemporaneous serial correlations due to financial markets interactions among these economies.

Maddala and Wu (1999) panel unit root test

Maddala and Wu (1999) proposed a Fisher test statistic solely based on joining the p-value of the test statistic from the individual unit root tests. The test is non-parametric and is based on Fisher (1932). Similar to Im et al. (1997), this test allows for different first-order autoregressive coefficients and has the same null and alternative hypothesis in the estimation procedure. The Fisher test statistic, P([chi]) is written as follows:

P([chi]) = -2 [N.sub.summation over (j=1)] log([pi].sub.j]) (7)

where [[pi].sub.j] is the p-value of the test statistic for j. The Fisher test statistic p([chi square]) is a chi-squared distribution with 2N degree of freedom.

As pointed out by Maddala and Wu (1999), Fisher test has more accurate size and better power compared to Levin and Lin (1993). Moreover, this test provides flexibility in choosing different lag lengths in each series of ADF regressions. Thus, the usefulness of the test is that it may reduce the bias due to the lag selection (see Banerjee, 1999).

Choi (2001) panel unit root test

Choi (2001) extends the Fisher test statistics of Maddala and Wu (1999) by demonstrating that

Z = 1/[square root of N] [N.summation over (j=1)] [[phi].sup.-1] ([[pi].sub.j) [right arrow] N(0, 1) (8)

where [[phi].sup.-1] is the opposite of the standard collective distribution function. [right arrow] N (0, 1) refers to asymptotically distributed as standard normal distribution.

There are several features that distinguish the Choi (2001) test from above-mentioned panel unit root tests. First, this test is devised for finite N as well as for infinite N, where N denotes the number of groups. Second, it is assumed that each series has different types of non-stochastic and stochastic elements. Third, there is flexibility in the length of time series whereby each series can appear in different number of time series. Fourth, this test also deals with problems where some groups have a unit root and the others do not. Thus, the Choi (2001) test can be used under more general assumptions than the panel unit root test of Im et al. (2003) and Levin and Lin (1993). (4) Moreover, as mentioned by the author, the Choi (2001) test is superior to that of Maddala and Wu (1999) in terms of finite sample size and power.

In sum, the fundamental element that differentiates the above three tests is that the Fisher test (Maddala and Wu, 1999 and Choi, 2001) is calculated from a combination of the significance levels of the different tests, whereas the Im et al. (2003) statistic is computed from a group of test statistics.

Therefore, the Fisher test has the flexibility of using heterogeneous lag lengths and capability of easing restrictive assumptions assumed by Im et al. (2003). More importantly, these three statistics will be computed in this study based on conventional adopted ADF, as well as ADF-GLS estimation procedures. (5)

RESULTS AND INTERPRETATION

As a preliminary analysis, the ordinary ADF and ADF-GLS univariate unit root tests are deployed to check the stationarity of the real interest rates for the sample period (2001:Q1 to 2006:Q3) and the results are summarized in Table 1. It is evident in Table 1 that the null hypothesis of non-stationary series can be rejected for China, Malaysia and Singapore by the ADF test. (6) This is because the probability value of the t-statistics for the three countries is less than 0.10, implying that the real interest rates concerned are stationary at 10 % significance level or better. The implication of this finding is that there is a long-run relationship between nominal interest rate and inflation rate in these countries. Hence, the Fisher hypothesis is valid for these countries. Applying the same principle, the results suggest that Fisher hypothesis does not hold for other countries.

In contrast, the ADF-GLS test is able to detect more cases supportive of the Fisher effect. In particular, the null of non-stationary series can be rejected at 5% level or better for China, Hong Kong, Indonesia, Japan, Malaysia, Singapore and Thailand, implying long-run validity of Fisher hypothesis for these countries. Since the results from ADF and ADF-GLS are inconsistent, one has to rely on a more robust test for decision. In this matter, the fact that the ADF-GLS test provides evidence in favor of the Fisher effect for most countries in the sample but the ADF test does not is in accordance with previous discussion in the literature that the ADF-GLS test has more power than the ADF test in detecting stationarity (Ng and Perron, 2001; Rapach and Wohar, 2002). As such, relying on results obtained from the ADF-GLS test, it

(3) Unlike another panel unit root test advocated by Levin et al. (2002) who imposed the restrictive assumption of homogeneity, Im et al. (2003) allow for heterogenity across groups and serial correlation errors across groups. Therefore, it achieves more accurate size and higher power relative to the Levin et al. (2002) test. By using a Monte Carlo simulation, Im et al. (2003) showed better finite sample performances of [[psi].sub.t] in relation to the Levin et al. (2002) test. may be noted for this moment that Fisher hypothesis hold for all countries under consideration with the exception of Korea, the Philippines and Taiwan.

It was mentioned earlier that conventional unit root test such as the ADF test has a low power when a relatively short span of data is employed. Therefore, it is possible that a longer period could improve the results. (7) For the purpose of comparison, we report the results of examining the longer sample period of data covering the period from 1987:Q1 to 2006:Q3 in Table 2. Based on the ADF test, the null hypothesis of a unit root is rejected for most of the countries with an exception of Hong Kong, Indonesia, Japan and Thailand. In other words, using a longer set of data, the ADF test is able to discover more evidence favoring the Fisher hypothesis. In addition to the evidence found earlier (China, Singapore and Malaysia), this time evidence is also found for South Korea, the Philippines and Taiwan. On the other hand, the results obtained from the ADF-GLS test suggests that the non-stationary real interest rate can be rejected for all of the countries at 10% or even better significance level. Overall the results from Tables 1 and 2 suggest that the Fisher hypothesis holds better for a longer set of data. This finding is in accordance with the view that univariate unit root tests can perform better if they are applied to a longer set of time series data (Rapach and Wohar, 2002; Baharumshah et al., 2005).

While the ADF-GLS test is more reliable than the ADF test, it does not consider financial market interactions across countries, which exist due to strong trade and investment relationships among these economies. To circumvent the weakness of these univariate unit root tests, panel unit root tests are employed. For this purpose, the test statistics of Im et al. (2003), Maddala and Wu (1999) and Choi (2001) are computed from both the ADF and ADF-GLS tests. The results for the periods 2001:Q1 to 2006:Q3 and 1987:Q1 to 2006:Q3 are presented in Tables 3 and 4, respectively. It is observed from these tables that the null hypothesis of non-stationary series can be rejected at 1% significance level regardless of the type of unit roots employed. Thus, it can be concluded that based on panel analysis which allows for the consideration of cross-country variations, all the East Asian real interest rates are stationary. Holmes (2002) points out that panel unit root tests work better than univariate unit root tests in the case of real interest parity. As such, this study concludes that, as a whole, the Fisher hypothesis holds for all the 10 countries under investigation based on panel testing procedures. Recall that in the case of univariate unit root tests, we need to lengthen the sample data to reveal more evidence in favor of the Fisher hypothesis. In sharp contrast, empirical findings of the Fisher hypothesis are obtained from panel unit root tests even when a shorter sample period was used in our study and the use of a longer sample provides consistent results. Thus, our findings are in line with those who found that panel unit root tests are an improvement over univariate unit root tests for finite data.

CONCLUSION

In general, a long-run relationship between nominal interest rates and inflation rates for all the East Asian economies under investigation has been identified by the panel but not the univariate unit root tests. The finding should come as no surprise as basically, these economies share quite similar monetary policies over the past few decades.

The key implications of this finding are: first, the validation of the Fisher hypothesis in these economies will encourage borrowers to make productive investments that promote economic growth and develop better banking system (Pill and Pradhan, 1997). Second, the stationarity finding for real interest rates provides convincing foundation for the applications of various capital asset pricing models in this region (Johnson, 2006).

Third, and perhaps more importantly, monetary policy can be used as an effective tool to influence long-term interest rates in these East Asian economies (Granville and Mallick, 2004). However, considering the fact that supportive evidence of the Fisher hypothesis is only obtained when cross-country interdependence in real interest rates is incorporated in the estimation, it is expected that monetary policy will work better with regional collaboration. This would require the coordination of policy-makers from ministries of finance; the central banks and the financial market regulators of these economies to develop a shared vision in their macroeconomic goals (see Sheng and Teng, 2007). Such collaboration is especially important in combating the recent global financial crisis and economic downturn. In this respect, the authorities across East Asia economies had used an array of similar policies (such as liquidity support, deposit guarantees, and foreign exchange intervention and swap arrangements) to support their banking systems and ensure financial stability in response to the global financial turmoil. According to the Asian Development Bank (2009), these policies indeed have successfully restored public confidence in the region's financial systems, and as a result, these economies managed to make a remarkable recovery (Lipsky, 2009). Nonetheless, to date, it is still early to safely conclude that crisis is over and as such, the leaders of China, Japan and South Korea recently emphasized that it is necessary to reinforce regional collaboration to face the world economic crisis (AsiaNews, 2009). Perhaps, instead of competitive interest rate reduction to boost exported-oriented industries during crisis (Ito, 2009), a more closely coordinated regional exchange rate mechanism and the establishment of an East Asian regional financial facility as proposed by the East Asian Study Group (2002) should now be seriously considered and pursued by the economies in this region to enhance financial and economic stability.

Acknowledgements

The authors sincerely thank the two anonymous referees and the Editor, Paul Wachtel, for their insightful comments on the earlier version of this study. The authors are solely responsible for remaining errors.

REFERENCES

AsiaNews. 2009: China, Japan, South Korea united against world economic crisis. 13 December 2009. Available at: http://www.asianews.it/index.php?l=en&art=13999#, accessed 31 December 2009.

Asian Development Bank. 2009: Asia Economic Monitor. June, 1-101. Available at: http://aric.adb.org/pdf/aem/jul09/Jul_AEM_complete.pdf, accessed 31 December 2009.

Atkins, FJ and Coe, PJ. 2002: An ARDL bounds test of the long-run Fisher effect in the United States and Canada. Journal of Macroeconomics 24(2): 255-266.

Baharumshah, AZ, Haw, CT and Fountas, S. 2005: A panel study on real interest rate parity in East Asian countries: Pre and post-liberalization era. Global Finance Journal 16(1): 69-85.

Bank for International Settlements. 2006: Monetary policies in Asia: Approaches and implementation. Proceedings of a BIS/HKIMR Conference in Hong Kong SAR, 21-22 November 2005.

Banerjee, A. 1999: Panel data unit roots and cointegration: An overview. Oxford Bulletin of Economics and Statistics 61(0): 607-629.

Choi, I. 2001: Unit root tests for panel data. Journal of International Money and Finance 20(2): 249-272.

Cooray, A. 2002: Interest rates and inflationary expectations: Evidence on the Fisher effect in Sri Lamka. South Asia Economic Journal 3(2): 201-216.

Cooray, A. 2003: The Fisher effect: A survey. Singapore Economic Review 48(2): 135-150.

Daniels, JP, Nourzad, F and Toutkoushian, RK. 1996: Testing the Fisher effect as a long-run equilibrium relation. Applied Financial Economics 6(2): 115-120.

East Asia Study Group. 2002: Final Report of the East Asia Study Group. Available at: http://www.aseansec.org/viewpdf.asp?file=/pdf/easg.pdf, accessed 31 August 2009.

Elliott, G, Rothenberg, TJ and Stock, JH. 1996: Efficient tests for an autoregressive unit root. Econometrica 64(4): 813-836.

Evans, MDD and Lewis, KK. 1995: Do expected shifts in inflation effect estimate of the long-run Fisher relation? Journal of Finance 50(1): 255-253.

Fahmy, YAF and Kandil, M. 2003: The Fisher effect: New evidence and implications. International Review of Economics and Finance 12 (4): 451-465.

Fisher, RA. 1932: Statistical Methods for Research Workers. Oliver and Boyd: Edinburgh.

Granville, B and Mallick, S. 2004: Fisher hypothesis: UK evidence over a century. Applied Economics Letters 11(2): 87-90.

Holmes, MJ. 2002: Does long-run real interest parity hold among EU countries? Some new panel data evidence. The Quarterly Review of Economics and Finance 42 (4): 733-746.

Im, KS, Pesaran, MH and Shin, Y. 1997: Testing for unit roots in heterogeneous panel. Department of Applied Economics, University of Cambridge, Mimeo.

Im, KS, Pesaran, MH and Shin, Y. 2003: Testing for unit roots in heterogeneous panels. Journal of Econometrics 115(1): 53-74.

Ito, T. 2009: Global financial crisis and Asian currencies. Global Crisis Debate 3 March 2009. Available at: http://www.voxeu.org/index.php?q = node/3163, accessed 21 December 2009.

Johnson, PA. 2006: Is it really the Fisher effect? Applied Economics Letters 13(4): 201-203.

Koustas, Z and Serletis, A. 1999: On the Fisher effect. Journal of Monetary Economics 44(1): 105-130.

Lee, JK, Clark, C and Ahn, SK. 1998: Long and short-run Fisher effects: New evidence and new results. Applied Economics 30(1): 113-124.

Levin, A and Lin, CE 1993: Unit root tests in panel data: Asymptotic and finite-sample properties. Unpublished manuscript, University of California, San Diego.

Levin, A, Lin, CF and Chu., C. 2002: Unit root tests in panel data: Asymptotic and finite sample properties. Journal of Econometrics 108 (1): 1-24.

Ling, TH. 2008: Fisher hypothesis and real interest rate equalization: Empirical evidence from Asian economies. Unpublished Master Thesis, Labuan School of International Business and Finance, Universiti Malaysia Sabah.

Lipsky, J. 2009: Asia, the financial crisis, and global economic governance. Speech at the Federal Reserve Bank of San Francisco Conference, Santa Barbara, California, 20 October 2009. Available at: http://www.imf.org/external/np/speeches/2009/102009.htm, accessed 21 December 2009.

Maddala, GS and Wu, S. 1999: A comparative study of unit root tests with panel data and a new simple test. Oxford Bulletin of Economics and Statistics 61(0): 631-652.

Ng, S and Perron, P. 2001: Lag length selection and the construction of unit root tests with good size and power. Econometrica 69(6): 1519-1554.

Payne, JE and Ewing, BT. 1997: Evidence from lesser developed countries on the Fisher hypothesis: A eointegration analysis. Applied Economics Letters 4(11): 683-687.

Phil, H and Pradhan, M. 1997: Financial liberalization in Africa and Asia. Finance and Development 34(2): 7-10.

Quah, D. 1992: International patterns of growth: I. Persistence in cross-country disparities. Unpublished manuscript, London School of Economics.

Quah, D. 1994: Exploiting cross-section variations for unit root inference in dynamic data. Economics Letters 44(1-2): 9-19.

Rapach, DE and Wohar, ME. 2002: Testing the monetary model of exchange rate determination: New evidence from a century of data. Journal of International Economics 58(2): 359-385.

Rose, A. 1988: Is the real interest rate stable? Journal of Finance 43: 1095-1112.

Sarel, M. 1996: Growth in East Asia: What We Can and What We Cannot Infer. International Monetary Fund Publication: Washington, DC.

Sheng, A and Teng, KK. 2007: East Asian capital markets integration: Steps beyond ABMI EABER Working Paper Series, Paper No. 11, East Asian Bureau of Economic Research.

United Nations. 2008: Foreign direct investment and financing for development: Trends and selected issues. TD/B/COM.2/80, Geneva. Available at: http://www.unctad.org/en/docs/c2d80_en.pdf, accessed 31 August 2009.

Wu, JL and Chen, SL. 1998: A reexamination of real interest rate parity. Canadian Journal of Economics 31 (4): 837-851.

Wu, JL and Chen, SL. 2001: Mean reversion of interest rates in the Eurocurrency market. Oxford Bulletin of Economics and Statistics 63 (4): 459-473.

(1) One exception is the recent work of Granville and Mallick (2004) who is able to provide evidence supportive of Fisher hypothesis by employing a century data covering from 1900 to 2000. In contrast, Rose (1988) is unable to find result in favor of the hypothesis using shorter span of data from 1892 to 1970 for the US.

(2) Im et al. (1997) demonstrated a substantial increase in power in panel unit root test, which allows for cross-sectional variation, even for fairly short time series.

(4) The work of Levin and Lin (1993) is published as Levin et al. (2002).

(5) Choi (2001) demonstrates the use of ADF-GLS test in his proposed Fisher test. By applying the Fisher test in the study of purchasing power parity (PPP), he demonstrated the proposed test is more powerful than ADF-GLS and t-bar test of Im et al. (2003).

(6) This finding may reflect the fact that these three countries share quite similar monetary policies in the sense that they had the experience of fixing their respective currencies against US dollar for the past few decades, as well as same goals to maintain low inflation and a stable exchange rate (see Bank for International Settlements, 2006). However, it is too early to base our conclusion on the finding of ADF test, in which the shortcomings of this test had been discussed earlier.

(7) Analysis of the longer sample has been included at the suggestion of one of the reviewers.

TAI-HU LING [1], VENUS KHIM-SEN LIEW [2] & SYED AZIZI WAFA SYED KHALID WAFA [3]

[1] Labuan School of International Business and Finance, Universiti Malaysia Sabah, Jatan Sungai Pagar, Labuan 87000, Malaysia.

[2] Department of Economics, Faculty of Economics and Business, Universiti Malaysia Sarawak, 94300 Kota Samarahan, Sarawak, Malaysia.

[3] School of Economics and Business, Universiti Malaysia Sabah, 8899 Kota Kinabalu, Malaysia.

Table 1: Unit root test results for
individual country (2001:01-2006:03)

Series            ADF                      ADF-GLS

         Lag     t-Stat     Prob.   Lag     t-Stat     Prob.

CH        1    -3.314 **    0.018    5    -3.087 ***   0.003
HK        1    -2.012       0.281    6    -2.091 **    0.041
ID        0    -2.152       0.226    0    -2.166 **    0.033
JP        2    -1.104       0.710    6    -2.595 **    0.012
KR        0    -1.760       0.397    2    -1.341       0.184
MS        0    -2.628 *     0.092    1    -2.055 **    0.043
PH        0    -1.905       0.328    4    -1.152       0.253
SG        0    -3.616 ***   0.008    4    -2.065 **    0.043
TH        0    -2.149       0.227    0    -2.161 **    0.034
TW        3    -0.723       0.834    7    -0.717       0.476

Note: In all cases, intercept has been included in the
estimation. ***, ** and * denote the rejection of the null
hypothesis of unit root at 1, 5 and 10% significance levels,
respectively. CH, HK, ID, JP, KR, MS, PH, SG, TH and TW denote
China, Hong Kong, Indonesia, Japan, South Korea, Malaysia, the
Philippines, Singapore, Thailand and Taiwan.

Table 2: Unit root tests for individual country (1987:41-2006:43)

Series         ADF                        ADF-GLS

         Lag   t-Stat       Prob.   Lag   t-Stat       Prob.

CH        5    -3.312 *     0.073    1    -3.600 ***   0.001
HK        1    -1.973       0.606    1    -1.806 *     0.075
ID        0    -2.155       0.507    0    -2.182 **    0.032
JP        4    -2.434       0.360    4    -2.364 ***   0.021
KR        1    -3.477 **    0.049    0    -2.389 **    0.019
MS        0    -3.225 *     0.087    0    -2.820 ***   0.006
PH        0    -3.999 **    0.013    0    -2.591 **    0.012
SG        0    -3.851 **    0.019    0    -3.832 ***   0.000
TH        0    -2.744       0.223    0    -2.534 **    0.013
TW        0    -6.910 ***   0.000    4    -4.070 ***   0.000

Note: In all ** * cases, intercept has been included in the
estimation. ***, and denote the rejection of the null hypothesis
of unit root at 1, 5 and 10% significance levels, respectively.
CH, HK, ID, JP, KR, MS, PH, SG, TH and TW denote China, Hong
Kong, Indonesia, Japan, South Korea, Malaysia, the Philippines,
Singapore, Thailand and Taiwan.

Table 3: Panel unit root tests of 10 countries (2001-2006)

Panel unit root test          Computed from

                            ADF         ADF-GLS

Im et al. (2003)        -1.9940 ***    -1.7213 **
Maddala and Wu (1999)   36.2467 ***    60.7641 ***
Choi (2001)             -7.1312 ***   -15.4874 ***

Panel unit root test            Critical value

                          1%        5%        10%

Im et al. (2003)         -1.960    -1.645    -1.282
Maddala and Wu (1999)    40.289    33.924    30.813
Choi (2001)              -1.960    -1.645    -1.282

Note: ***, ** and * denote the rejection of the null hypothesis
of unit root at 1, 5 and 10% significance levels, respectively.

Table 4: Panel unit root tests of 10 countries (1987-2006)

Panel unit root test           Computed from

                            ADF          ADF-GLS

Im et at. (2003)         -5.1537 ***    -2.7575 ***
Maddala and Wu (1999)    72.4849 ***   104.9268 ***
Choi (2001)             -14.8187 ***   -24.7780 ***

Panel unit root test          Critical value

                          1%        5%        10%

Im et at. (2003)         -1.960    -1.645    -1.282
Maddala and Wu (1999)    40.289    33.924    30.813
Choi (2001)              -1.960    -1.645    -1.282

Note: ***, ** and * denote the rejection of the null hypothesis
of unit root at 1, 5 and 10% significance  levels respectively.