Causality Relationships between the Twin Deficits in the Regional Economy

Transcription

1 Version: 04/06/2009 Causality Relationships between the Twin Deficits in the Regional Economy Jui-Chuan Chang * and Zao-Zhou Hsu + Department of Economics National Chi Nan University ABSTRACT Utilizing the simpler Granger non-causality procedure developed by Toda and Yamamoto (1995), this paper attempts to investigate the causality relationship between the budget deficit and the current account deficit for five North European countries, Four Asian Tigers and the United States. In order to circumvent the distortion of the causality inferences which could be due to omission of relevant variables, this study conducts the modified Wald test procedure in a multi-variate framework, apart from previous researches in a bi-variate model. The empirical results not only provide some alternative views of theoretical prediction, but also show the structural imbalances of the economies studied. Keywords: Twin deficits; causality test JEL classification: F41, C22 * Corresponding Author: Assistant Professor, Department of Economics, National Chi Nan University, 1 University Rd., Pu-li, Nantou 54561, Taiwan; dellachang@ncnu.edu.tw; Phone: ext. 4632; Fax: Graduate, Department of Economics, National Chi Nan University; s @ncnu.edu.tw.

2 1 Introduction During the early 1980s, the term twin deficits was initially invented to describe the co-movement between the budget deficit and the current account deficit in the United States (Abell, 1990; Gordon, 1986; Laney, 1984; McKinnon, 1980, 1990; Miller and Russek, 1989). 1 Since the switch of the century in the United States the recurrence of huge fiscal and trade deficits have revived public interests in the edge relations between the twin deficits (Bordo, 2006; Coughlin et al., 2006; Mann, 2002; Obstfeld and Rogoff, 2004, 2005; Sinai, 2006; Salvatore, 2006). As shown in the literature, however, the causal directions between budget and current deficits are not exclusive to the United States yet. In the 1990s, European countries, such as Sweden and Germany, also faced similar situations where the rise in budget deficits was accompanied by the real appreciation of their national currencies that adversely affected the current accounts (Ibrahim and Kumah, 1996). Due to the differences in the structure of the economy, furthermore, developing economies are more likely to experience the same expansion of budget and current account deficits (Laney, 1984; Khalid and Kuan, 1999; Anoruo and Ramchander, 1998; Edwards, 2001; Megarbane, 2002; Kouassi et al., 2004). As such the macroeconomic dynamics governing the two deficits may be different from the developed economy. Thus, this paper mainly attempts to investigate the causality relationship between the twin deficits using the expanded data from five North European countries, Four Asian Tigers and the United States. This paper plans to extend the existing literature on causality between the twin deficits in three significant aspects. First of all, this paper explicitly takes into account the lately causality testing procedure proposed by 1 The so-called twin deficits hypothesis that emerged in the 1980s is characterized by an unusual shift in current account and budget deficits. Another feature is the strong appreciation of the dollar. 1

3 Toda and Yamamoto (1995). This method has no regards that the VARs may be stationary integrated of any order, or cointegrated of any order; consequently, one can test linear or nonlinear restrictions on the coefficients by estimating a VAR in levels and applying the modified Wald statistics. Besides, this approach is much attractive not just because it is computationally simpler than the traditional F-test for Granger causality, but also because the finite sample performance of this Monte Carlo based testing procedure has proven to be superior to that of Toda and Phillips (1993, 1994). Second, this study employs a multi-variate rather than a bi-variate framework, in order to keep away from the distortion of the causality inferences which could be due to omission of relevant variables. For example, the role of the dollar in causing the trade deficit is a key part of the widely accepted doctrine that links trade deficit to budget deficit in the United States. To our best knowledge, since Abell (1990) discusses the importance of the mediating variables in the twin deficits nexus, there are few studies conducting in the multi-variate frameworks, except Anoruo and Ramchander (1998) and Kim and Kim (2006). Third, this research provides broader evidence on such a debating relationship using the longer time spanned data from five North European countries, Four Asian Tigers and the United States, respectively. This will enhance the robustness of our empirical findings by potentially revealing both specific and general information on the vastly structural imbalances of the economies studied. Once the causality relationships could be confirmed, policymaker might effectively make the twin deficits under control and keep economic growth sustainable. The structure of this paper is organized as follows. Section 2 reviews four possible causality relationships between the twin deficits in the extant literature. Section 3 discusses the empirical method of Toda and Yamamoto (1995). Section 4 gives the data description and presents the empirical results. Last section draws the 2

4 conclusions. 2 Literature Review There might be four possible causation linkages between budget deficit (BD) and current account deficit (CAD), including BD CAD, BD / CAD, CAD BD and BD CAD. A first theoretical explanation of the relation between BD and CAD is the Mundell-Fleming framework. The Mundell-Fleming approach argues that an increase in BD induces an upward pressure on interest rates that, in turn, will trigger capital inflows and an appreciation of exchange rates, ultimately leading to an increase in CAD. A second theoretical explanation of the linkage between the twin deficits is the Keynesian absorption theory, which suggests that an increase in BD would induce domestic absorption and hence import expansion, causing an increase or a worsening of the CAD. Both the Mundell-Fleming model and Keynesian absorption theory support the uni-directional relationship of BD causing CAD (BD CAD). Another explanation of the relation between BD and CAD is based on the presumption that the twin deficits are not related. Adherents of the Ricardian equivalence hypothesis dispense entirely with the income-expenditure approach and rely instead on the inter-temporal approach. Since a government s means of finance do not alter private agents inter-temporal budget constraints, the real interest rate, the quantity of investment, or the current account balance. They claim that budget deficits do not cause any interest and exchange rate changes (Garcia and Ramajo, 2004), which thus have no effect on current account imbalances. Hence, under the Ricardian equivalence hypothesis, budget and current account deficits are causally independent; that is, BD / CAD. Third, a uni-directional causality that may run from current account to budget 3

5 deficits may also exist. This outcome occurs when the deterioration in current account leads to a slower pace of growth and hence an increase in the budget deficit. A country experiencing a financial or solvency crisis resulting from chronic, excessive current account deficits may face a situation in which large injections of public funds are required to rehabilitate troubled financial sectors, to improve the corporate governance system, and to attenuate a recession. In Korea, for example, fiscal deficits were allowed to increase considerably for the purposes of supporting economic activity and strengthening the social safety net after the financial crisis of The causal relationship in this instance proceeds from the current account deficit to the budget deficit (CAD BD). This reverse causation is termed current account targeting by Summers (1988). He argued that external adjustment may be sought via fiscal policy. This causal pattern may be more relevant for developing countries that have accumulated large foreign debts. Recently, Alkswani and Al-Towaijari (1999) provide empirical evidence on reverse causation between the two deficits for Saudi Arabia. While Anoruo and Ramchander (1998) discover that trade deficit causes fiscal deficit in some Asian countries. They argue that governments in developing countries might engage in fiscal stimulus to lessen the deleterious economic and financial consequences of large trade imbalances. The economic slowdowns brought about by huge current account deficits not only increase government spending but also reduce tax revenues. Besides, bi-directional causality might exist between budget and current account deficits; that is, BD CAD. While budget deficits may cause current account deficits, the existence of significant feedback may cause causality between the two variables to run in both directions. The Latin American currency crisis in the 1980s may have reflected just such a vicious circle. Feldstein and Horioka (1980) also find that savings 4

6 and investments are highly correlated, causing BD and CAD to move together. Hence, the validity of the twin deficits hypothesis could be linked to the degree of international capital mobility. According to the above-mentioned literature, each of possible causal linkages between budget deficit and current account deficit corresponds to one of theoretical predictions. To be clarified, Figure 1 demonstrates that the four possible causality relationships between the twin deficits are associated with their corresponding theory. 3 Empirical Method Many economists have applied the Granger causality to test causal relationships among variables since 1969, but the procedure of Granger causality analysis is very sensitive to model specifications, such as the chosen lag length and stationary properties. If a VAR system is cointegrated, the Granger causality test may be conducted in the environment of vector error correction model (VECM); otherwise, the analyses may be conducted as a standard first-difference VAR model. In other words, when the variables are cointegrated, the corresponding error correction representations must be included in the system (Granger, 1988). By doing so, one can avoid misspecification and omission of the important constraints; meanwhile, the traditional F-test for Granger non-causality is not valid as well. That is because the test statistic does not follow its own distribution when the variables are integrated or cointegrated. On the other hand, if the variables are not integrated of the same order or are not cointegrated, the VECM cannot be applied either. In addition, possibly severe pre-test biases in ECM may exist, especially for finite samples. In this sense that the workhorses of testing the non-causality in the very of VECM are cumbersome and sensitive to the values of nuisance parameters in finite 5

7 samples; therefore, the virtues of simplicity and ease of application have been largely lost (Rambaldi and Doran, 1996, p. 3). Besides, the formulation does have its drawbacks in that it is implicitly dependent upon a pre-test of integration and cointegration (Lütkepohl and Reimers, 1992). One way to circumvent this problem is to posit a VAR in which variables appear purely in their level form. Toda and Yamamoto (1995) have proposed the modified wald (MWALD) for testing Granger non-causality that allows causal inference to be conducted in the level VARs that may contain integrated of an arbitrary order and/or (non-) cointegrated of an arbitrary order processes. That is, this procedure imposes (non-) linear restrictions on the parameters of VAR models without having to pre-test for unit roots and cointegrating ranks. Thus, if one is primarily interested in testing zero-restrictions on the coefficients, rather than in testing for the presence of unit roots or for cointegrating relations, the MWALD test is appropriate. Of particular notice that the MWALD test requires that the true lag length exceed the order of integration. Moreover, this procedure is much attractive not just because the MWALD test is computationally simpler than the traditional F-test for Granger non-causality. But also because the MWALD test has the finite sample performance of this Monte Carlo based testing procedure in size and power. Zapata and Rambaldi (1997) verify that the MWALD test is superior to both the LR test of Mosconi and Giannini (1992) and the WALD of Toda and Phillips (1993, 1994) in samples of 50 or more observations. In what follows, we thus rely on the Toda-Yamamoto (1995) MWALD tests to make the causal inference among the variables in the VAR system. The whole procedure may be summarized as follows: First of all, we establish the maximum order of integration (d max ) and the optimal lag length (k) of the VAR model. Second, we estimate the (k + d max )th-order VAR model in levels. Because Rambaldi and 6

8 Doran (1996) show that the seemingly unrelated regression could easily compute the MWALD test, we estimate the specified VAR system by the least squares and seemingly unrelated regression methods, respectively. Once the congruency of the VAR duly examined through the standard diagnostics tests, the non-causality test is formulated as a zero restriction on the first k coefficient matrices, ignoring the last d max lagged vectors in the model. The MWALD test statistic asymptotically follows a chi-square distribution having the usual degrees of freedom. 4 Data and Estimation Results 4.1 Data Description A plenty of empirical studies have investigated the causal relationship between the budget and current account balances for annual data set of developed and developing countries. Moreover, most of these studies employ the Granger causality test in bi-variate VAR models. According to the above-mentioned literature, however, both the interest rate and the exchange rate play important roles in a channel through which the budget deficit affects the current account deficit (Abell, 1990; Ibrahim and Kumah, 1996). Owing to the omission of relevant explanatory variables, these results from bi-variate frameworks may have been biased. In order to keep away from such possible distortion of the causality inferences, this paper examines the possible causal relationships by utilizing a multi-variate setup, which contains budget deficit (BD), current account deficit (CAD), the interest rate (IR) and the exchange rate (EXC). This paper adopts both annual and quarterly macroeconomic data for five North European countries, Four Asian Tigers and the United States. These ten countries are Denmark, Finland, Iceland, Norway, Sweden, Hong Kong, Korea, Singapore, Taiwan, and the United States. The regional countries included in our sample are primarily 7

9 selected on the basis of data availability and historical background, while the United States is embedded for comparison purpose. To be specific, Table 1 summarizes the related literature on the twin deficits for the ten countries. It is clear for us to see that there are very few papers regarding the North European countries and the bi-variate framework is most utilized. Also, this study is not the first one to conduct the Toda-Yamamoto (1995) MWALD test, but is the first one to apply this procedure to a four-variable setup. With the exception of Taiwan and Hong Kong, all of the annul data are drawn from the World Economic Outlook databases (WEO) of IMF, while the quarterly data are collected from the International Financial Statistics (IFS) of IMF. For Taiwan data, they are obtained from the AREMOS databases, and for Hong Kong, they are from the Census and Statistics Department of Hong Kong and the INTLINE International Macroeconomic Statistical Databank. Due to the data availability, the annual data cover the period from 1980 to 2007, while the quarterly data are not symmetric for all countries. The budget and current account variables are scaled by GDP, the exchange rate is measured by the nominal effective exchange rate, and the interest rate is a proxy of policy measure via the discount rate. For details, we exhibit a listing of the ten countries as well as the corresponding sample periods in Table 2. In order to apply the MWALD test for Granger non-causality, it is necessary to determine the maximum order of integration of each of the series being studied. To do this, the standard unit root tests are used: the Augmented Dickey-Fuller (ADF), Phillips and Perron (PP), and Kwiatkowsi et al. (KPSS) tests. The test results of unit roots for all series of each country indicate that the maximum order of integration is one, I(1); that is, d max = 1. In selecting the optimal lag length for a VAR system, both the Akaike information criterion (AIC) and the Schwarz Bayesian criterion (SBC) are first used. Since these criteria yield different lag lengths, Sim s modified 8

10 log-likelihood ratio (LR) test is then employed and diagnostic tests examined. To identify the direction of any causality between BD and CAD, the four-equation VAR model is estimated using the least squares and seemingly unrelated regression methods, respectively. 2 After that, the MWALD test is then applied to examine whether the coefficients of the optimally lagged variables are jointly equal to zero. In particular, the diagnostic tests shall suggest that the VAR model, on which the MWALD test is based, is quite well-specified. 4.2 Estimation Results The results of the MWALD test for each country are illustrated in Figure 2. The graphs are arranged by the regional economy. Since the adoption of the quarterly data is much apart from the previous researches, this section will focus on analyzing the results from the quarterly data. The arrow, and denotes that the null hypothesis of the non-causality is rejected at the 10%, 5% and 1% significance levels, respectively. The order of the estimated VAR system is determined by the sum of the optimal lag length (k) and maximum of integration (d max ), which are shown in the orders column. The specified VAR model is estimated using the least squares (LS) and seemingly unrelated regression (SUR) methods, separately. Some countries have a few observations so that their results from the SUR method show near singular matrix. Moreover, comparing the available outcomes from the LS to those from the SUR, this demonstrates both are very similar in terms of the significantly direct causal link between BD and CAD except Sweden. Let s look at the results from five North European countries first. Among these 2 For comparison, we also estimate the bi-variate VAR system using the least squares and SUR methods, respectively. Except that Hong Kong has the bi-directional causality relationships between BD and CAD, others confirm the twin deficit hypothesis which the uni-directional causality of BD leading to CAD although this is against the study for Sweden which posits the Ricardian equivalence hypothesis. 9

11 five North European countries, there are three out of five countries (Finland, Iceland and Sweden) consistent with the twin deficits hypothesis which indicates uni-directional causality from BD to CAD. The outcomes are in conformity with that of Afonso and Rault (2009) for Finland but against that of Kouassi et al. (2004) for Sweden. 3 To be specific, the exchange rate mediates the relationship between the twin deficits for Finland while the interest rate does so for Iceland. The results for Denmark show that the uni-direction of causality runs reversely from CAD to BD and is mediated either by the exchange rate or by both exchange and interest rates. On the other hand, the consequences for Norway indicate a predominantly bi-directional causality even at the 1% significance level. Among Four Asian Tigers, the results for both Korea and Taiwan are consistent with the causal link of twin deficits hypothesis, implying that BD leads to CAD. This confirms the bi-variate outcomes of Lau and Baharumshah (2006) for Korea as well as that of Chang (2004) for Taiwan. Of particular interest, the indirect causal link for Taiwan is completely reversed to run from CAD to BD through the exchange rate and interest rate when the four-equation VAR system is estimated by the SUR method. In a multi-variate framework, the strongly bi-directional results for Singapore are contrary to that of Lau and Baharumshah (2006). A two-way causality is detected between the twin deficits, giving credence to both twin deficits and current account targeting propositions in which budget cuts improve current account and this further leads to a further reduction in budget deficit. Following the Feldstein and Horioka (1980), we may further check whether or not the investment and saving are highly correlated in Singapore. Lastly, using the quarterly data after 1997, the causal link for Hong Kong runs from CAD to BD. 3 In a bi-variate framework, Kouassi et al. (2004) declare that the twin deficits are independent for Sweden, which is consistent with the Ricardian equivalence hypothesis. 10

12 Recent declines in the United States fiscal and current account balances have sparked renewed debate over the twin-deficit hypothesis, which argues that a larger fiscal deficit, through its effect on national saving, leads to an expanded current account deficit. The empirical results for the United States in this paper are parallel to the literature; however, the indirect causal link is possibly mediated by the interest rate, not by the exchange rate. 5 Concluding Remarks Utilizing the simpler Granger non-causality procedure proposed by Toda and Yamamoto (1995), this paper provides broader evidence on a debating causal linkage between the budget and current account deficits for five North European countries, Four Asian Tigers, and the United States. Througout the empirical investigation, this study finds some support for the twin defiicts hypothesis although the strength of the relationship varies across countries. Yet, none of the economies studied in this paper is supportive of the validity of th ericardian equivalence hypothesis. From the policy perspectives, this analysis will enhance the robustness of our empirical findings by potentially revealing both specific and general information on the vastly structural imbalances of the economies studied. Especially, adding the realistic likely costs of the financial tsunami to the previous national debt puts the national debt to Gross Domestic Product (GDP) ratio at over 100% in both the United States and most of North European countries. Lately government bailouts as well as a fall in the tax revenues due to a decline in business in export sector, tends to support the causality from current account to budget deficits. Thus, since the causality relationships could be confirmed, policymaker might effectively make the twin deficits under control and keep economic growth sustainable. 11

13 References Abell, J. D. (1990), Twin Deficits during the 1980s: An Empirical Investigation, Journal of Macroeconomics, v. 12, iss. 1, pp Afonso, A. and C. Rault (2009), Bootstrap Panel Granger-Causality between Government Spending and Revenue, Department of Economics, Technical University of Lisbon, Working Papers No. 2008/39. Alkswani, M. A. and H. A. Al-Towaijari (1999), Cointegration, Error Correction and the Demand for Money in Saudi Arabia, Economia Internazionale, v. 52, iss. 3, pp Anoruo, E. and S. Ramchander (1998), Current Account and Fiscal Deficits: Evidence from Five Developing Economies of Asia, Journal of Asian Economics, v. 9, iss. 3, pp Baharumshah, A. Z., E. Lau, and A. M. Khalid (2006), Testing Twin Deficits Hypothesis Using VARs and Variance Decomposition, Journal of the Asia Pacific Economy, v. 11, iss. 3, pp Baharumshah, A. Z. and E. Lau (2007), Dynamics of Fiscal and Current Account Deficits in Thailand: An Empirical Investigation, Journal of Economic Studies, v. 34, iss. 5-6, pp Bordo, M. (2006), Globalization and Imbalances in Historical Perspective, Federal Reserve Bank of Cleveland Working Paper No.13. Coughlin, C. C., M. R., Pakko, and W. Poole (2006), How Dangerous is the U.S. Current Account Deficit? The Regional Economist, pp Edwards, S. (2001), Does the Current Account Matter? National Bureau of Economic Research Working Papers No Enders, W. and B.-S. Lee (1990), Current Account and Budget Deficits: Twins or Distant Cousins? Review of Economics and Statistics, v. 72, iss. 3, pp Feldstein, M. and C. Horioka (1980), Domestic Saving and International Capital Flows, Economic Journal, v. 90, iss. 358, pp Garcia, A. and J. Ramajo (2004), Budget Deficit and Interest Rates: Empirical Evidence for Spain, Applied Economics Letters, v. 11, iss. 11, pp Gordon, R. J. (1986), The American Business Cycle: Continuity and Change, in National Bureau of Economic Research Studies in Business Cycles Series, v. 25, pp. xiv, 868. Chicago and London: University of Chicago Press. Granger, C. W. J. (1988), Some Recent Developments in a Concept of Causality, Journal of Econometrics, v.39, iss. 1/2, pp Hatemi-J., A. and G. Shukur (2002), Multivariate-Based Causality Tests of Twin Deficits in the US, Journal of Applied Statistics, v. 29, iss. 6, pp Ibrahim, S. B. and F. Y. Kumah (1996), Comovements in Budget Deficits, Money, Interest Rates, Exchange Rates and the Current Account Balance: Some Empirical 12

14 Evidence, Applied Economics, v. 28, iss. 1, pp Kalyoncu, H. (2007), Budget and Current Account Deficits in Asian Countries, Empirical Economics Letters, v. 6, iss. 2, pp Kasibhatla, K. M., M. N. Johnson, J. Malindretos, and A. C. Arize (2001), Twin Deficits Revisited, Journal of Business and Economic Studies, v. 7, No. 2, pp Khalid, A. M. and T. W. Guan (1999), Causality Tests of Budget and Current Account Deficits: Cross-Country Comparisons, Empirical Economics, v. 24, iss. 3, pp Kim, C.-H. and Donggeun Kim (2006), Does Korea Have Twin Deficits? Applied Economics Letters, v. 13, iss. 10, pp Kouassi, E., M. Mougoue, and K. O. Kymn (2004), Causality Tests of the Relationship between the Twin Deficits, Empirical Economics, v. 29, iss. 3, pp Laney, L. O. (1984), The Strong Dollar, the Current Account, and Federal Deficits: Cause and Effect, Federal Reserve Bank of Dallas Economic Review, pp Lau, E. and A. Z. Baharumshah (2006), Twin Deficits Hypothesis in SEACEN Countries: A Panel Data Analysis of Relationships between Public Budget and Current Account Deficits, Applied Econometrics and International Development, v. 6, iss. 2, pp Leachman, L. L. and B. Francis (2002), Twin Deficits: Apparition or Reality? Applied Economics, v. 34, iss. 9, pp Lutkepohl, H. and H. E. Reimers (1992), Granger-Causality in Co-Integrated VAR Processes: The Case of the Term Structure, Economics Letters, 40, Mann, C. L. (2002), The U.S. Current Account Deficit Sustainability, Journal of Economic Perspectives, v. 16, iss. 3, McKinnon, R. I. (1980), Exchange Rate Instability, Trade Balance, and Monetary Policy in Japan, Europe, and the United States, in P. Oppenheimer (ed.), Issues in International Economics (pp ). Boston: Oriel Press. McKinnon, R. I. (1990), The Exchange Rate and the Trade Balance: Insular versus Open Economies, Open Economies Review, v. 1, iss. 1, pp Megarbane, P. (2002), Slovakia s External Current Account Deficit: Why so Large and Is It Sustainable? IMF Country Report 210. Miller, S. M. and F. S. Russek (1989), Are the Twin Deficits Really Related? Contemporary Policy Issues, v. 7, iss. 4, pp Mosconi, R. and C. Giannini (1992), Non-causality in Cointegrated Systems: Representation Estimation and Testing, Oxford Bulletin of Economics and Statistics, v. 54, iss. 3, pp Obstfeld, M. and K. Rogoff (2004), The Unsustainable Current Account Position Revisited, National Bureau of Economic Research Working Paper No

15 Obstfeld, M., and K. Rogoff (2005), Global Current Account Imbalances Exchange Rate Adjustments, Brookings Papers on Economic Activity, v. 1, pp Ozturk, I. (2008), Testing Twin Deficits Hypothesis: Empirical Evidence from Turkey, Empirical Economics Letters, v. 7, iss. 1, pp Rambaldi, A. N. and H. Doran (1996), Testing for Granger Non-Causality in Cointegrated Systems Made Easy. Department of Econometrics, University of New England, Working Paper in Econometrics and Applied Statistics No. 88. Rosenswieg, J. A. and E. W. Tallman (1993), Fiscal Policy and Trade Adjustment: Are the Deficits Really Twins? Economic Inquiry, v. 31, iss. 4, pp Salvatore, D. (2006), Twin Deficits in the G-7 Countries and Global Structural Imbalances, Journal of Policy Modeling, v. 28, iss. 6, pp Sinai, A. (2006), Deficits Expected Deficits, Financial Markets, and the Economy, North American Journal of Economics and Finance, v. 17, iss. 1, pp Summers, L. H. (1988), Tax Policy and International Competitiveness, International Aspects of Fiscal Policies, pp , National Bureau of Economic Research Conference Report Series. Chicago and London: University of Chicago Press. Toda, H. Y. and P. C. B. Phillips (1993), Vector Autoregression and Causality, Econometrica, v. 61, iss. 6, pp Toda, H. Y. and P. C. B. Phillips (1994), Vector Autoregression and Causality: A Theoretical Overview and Simulation Study, Econometric Reviews, v. 13, iss. 2, pp Toda, H. Y. and T. Yamamoto (1995), Statistical Inference in Vector Autoregression with Possibly Integrated Processes, Journal of Econometrics, v. 66, iss. 1-2, pp Wissem, A. (2007), The Twin Deficits : Are They Really Twins? An Empirical Investigation of the Case of a Small Developing Economy, The Icfai University Journal of Applied Economics, v. VI, iss. 1, pp Zietz, J. and D. K. Pemberton (1990), The U.S. Budget and Trade Deficits: A Simultaneous Equation Model, Southern Economic Journal, v. 57, iss. 1, pp Zapata, H. O. and A. N. Rambaldi (1997), Monte Carlo Evidence on Cointegration and Causation, Oxford Bulletin of Economics and Statistics, v. 59, iss. 2, pp

16 Table 1 Summary of Related Literature Five North European Countries Denmark Number of Variables Methodology Causality None Finland Number of Variables Methodology Causality Afonso 3: BD, CAD, EXC Bootstrap panel Granger-causality BD CAD Rault (2009) Iceland Number of Variables Methodology Causality None Norway Number of Variables Methodology Causality None Sweden Number of Variables Methodology Causality Kouassi 2: BD, CAD Toda-Yamamoto test (1995) BD / CAD Mougoue Kymn (2004) Four Asian Tigers Hong Kong Number of Variables Methodology Causality None Korea Number of Variables Methodology Causality Anoruo 2: BD, CAD Granger causality test CAD BD Ramchander (1998) Lau 2: BD, CAD Toda-Yamamoto test (1995) BD CAD Baharumshah (2006) Kim 3: BD, CAD, EXC Toda-Yamamoto test (1995) CAD BD Kim (2006) Kalyoncu (2007) 2: BD, CAD Granger causality tests BD CAD Singapore Number of Variables Methodology Causality Lau 2: BD, CAD Toda-Yamamoto test (1995) BD CAD Baharumshah (2006) Kalyoncu (2007) 2: BD, CAD Granger causality tests BD / CAD Taiwan Number of Variables Methodology Causality Chang (2004) 2: BD, CAD Granger causality tests BD CAD 15

17 United States Number of Variables Methodology Causality Abell (1990) 4: BD, CAD, IR, EXC VAR model BD CAD Zietz 2: BD, CAD Small simultaneous-equation model BD CAD Pemberton (1990) Rosenswieg 3: BD, CAD, EXC Five-variable VAR model BD CAD Tallman (1993) Enders 2: BD, CAD Unconstrained VAR model BD / CAD Lee (1990) Kasibhatla 2: BD, CAD Granger causality tests BD CAD Johnson Malindretos (2001) Leachman 2: BD, CAD Error correction models BD CAD Francis (2002) Hatemi-J 2: BD, CAD ( ) Granger causality tests BD CAD Shukur (2002) 2: BD, CAD ( ) CAD BD Kouassi 2: BD, CAD Toda-Yamamoto test (1995) BD / CAD Mougoue Kymn (2004) Salvatore (2006) 2: BD, CAD Multiple regression model BD CAD 16

18 Table 2 List of Countries Studied Country Frequency Sample Period Specific Variable Five North European countries Denmark Finland Iceland Norway Sweden Four Asian Tigers Hong Kong Korea Singapore Taiwan Annual (28) Quarterly 1998Q1-2007Q4 (36) Annual (28) IR: Average cost of CB debt Quarterly 1978Q1-2007Q4 (120) IR: Average cost of CB debt Annual (28) IR: General loans, bonds Quarterly 19798Q1-2007Q4 (40) Annual (28) Quarterly 1996Q1-2007Q4 (48) BD: Net operating balance / gdp Annual (28) IR: Bank rate (end of period) Quarterly 1994Q1-2006Q4 (52) IR: Bank rate (end of period) Annual (28) IR: Savings deposit rate Quarterly 1999Q1-2008Q1 (37) IR: Discount rate EXC: Market rate, period average Annual (28) EXC: Market rate, period average Quarterly 1976Q1-2000Q3 (99) EXC: Market rate, period average Annual (28) IR: Saving rate Quarterly 2003Q1-2006Q3 (15) IR: Saving rate Annual (28) IR: Rediscount rate EXC: Market rate, period average Quarterly 1984Q3-2008Q (97) IR: Rediscount rate EXC: Market rate, period average United States United Annual (28) States Quarterly 1975Q1-2007Q4 (132) Notes: For the annual data, BD: General government balance (percent of GDP), CAD: Current account balance (percent of GDP), IR: Discount rate (end of period), and EXC: nominal effective exchange rate. But for the quarterly data, BD: Cash surplus or deficit (percent of GDP), CAD: Current account (percent of GDP), IR: Discount rate (end of period), and EXC: nominal effective exchange rate. 17

19 Figure 1 Four Possible Types of Relationships between twin deficits 18

20 Denmark ( ) Orders LS SUR 4+1=5 (AIC) (SC) (HQ) Near singular matrix Finland (1978Q1-2007Q4) Orders LS SUR 4+1=5 (AIC: 26) Iceland ( ) Orders LS SUR 4+1=5 (AIC) (SC) (HQ) Near singular matrix Norway (1996Q1-2007Q4) Orders LS SUR 1+1=2 (SC: 6) 19

21 Sweden (1994Q1-2006Q4) Orders LS SUR 1+1=2 (SC: 6) (HQ: 6) Hong Kong (1999Q1-2008Q1) Orders LS SUR 1+1=2 (SC: 3) Korea (1976Q1-2000Q3) Orders LS SUR 1+1=2 (SC: 3) Singapore ( ) Orders LS SUR 4+1=5 (AIC) (SC) (HQ) Near singular matrix 20

22 Taiwan (1984Q3-2008Q3) Orders LS SUR 3+1=4 (AIC: 7) The United States (1975Q1-2007Q4) Orders LS SUR 7+1=8 (AIC: 20) (HQ: 20) Notes: denotes that the null hypothesis of the non-causality is rejected at 10% significance level, denotes that the null hypothesis of the non-causality is rejected at 5% significance level, and denotes that the null hypothesis of the non-causality is rejected at 1% significance level. Figure 2 Empirical Results 21