U.s.-japan and U.s.-canada Bilateral Real Trade Balances: an Empirical Exploration

U.S.-Japan as well as U.S.-Canada bilateral real trade balances, real exchange rates and relative real GDP is non-stationary in levels revealing I (1) behavior. There are evidence of co integration, long-run bi-directional causal flows among variables with no clear evidence of feedbacks in different lag structures. INTRODUCTION Japan and Canada are the most important trading partners of the United States of America (U.S.). Japan has the second largest economy of the world despite being in chronic recession since 1988.This is next to the U.S. mega economy ($11 trillion). Historically, U.S. has persistent and yawning trade deficit with the rest of the world. A bulk of it is with Japan alone. From time to time, it became an irritant in the U.S.-Japan economic relations. There has been a significant pressure on Japan from the U.S. to reduce its bilateral trade surplus through voluntary export restraint and import expansion policies. But a huge trade deficit still persists and even continues to swell despite gradual opening of the Japanese market to U.S. products. This is attributed to a much faster pace in U.S. economic growth than in Japan. Canada tops the list of major trading partners of the U.S. This country is overly dependent on the U.S. market for its exports and imports (some 90 percent). It occupies a unique position as a U.S. trading partner because of its geographic proximity with the U.S. and being a member of the North American Free Trade Agreement (NAFTA). In the post-NAFTA period, the U.S.-Canada trade volume continues to rise at a considerable pace. Given the paramount importance of Japan and Canada to U.S. foreign trade, the issue of the dynamics of U.S. real trade balances with these countries merits an in-depth inquiry. The primary objective of this paper is thus to address the aforementioned issue within the co integration framework. The remainder of the paper is organized as follows. Section II provides a survey of the related literature. Section III outlines the empirical design. Section IV reports the empirical results. Finally, section V concludes. Southwestern Economic Review 126 SURVEY OF RELATED LITERATURE Among a host of macroeconomic explanators of bilateral exports and imports, relative GDP and exchange rates have gained prominence in the mainstream empirical literature on this subject of paramount importance. Numerous studies focused on the functional forms of imports and exports [Burgess, (1974), Houthakker and Magee (1969), Kreinin (1967), Leamer and Stern (1970), Magee (1975), Ball and Marwah (1962), Khan and Ross (1977), Goldstein and Khan (1985), Boylan et al., (1980), Kohli (1990)]. The major conclusions that emerge from these studies on the functional forms are either linear or log-linear. Subsequently, [Volker (1992), Feenstra (1995), Sawyer (1999), Matscebayashi (1999)] re-examine similar issues and draw similar inferences. The depreciation of the dollar against foreign currencies confers a price advantage for U.S. exports relative to those of the U.S. trading partner nations. Yet, it is by no means clear that the volume of these exports will increase by more than the decrease in their dollar prices. Unless the export volume increases by more than the export prices fall, the resulting dollar value of the exports will drop. Furthermore, exports to these countries depend not only on export prices, but also on their real economic growth. The U.S. aggregate nominal trade data over the last three decades reveal that nominal exports and nominal imports are more closely correlated with each other than with the movements in the nominal external value of the dollar. The correlation between total nominal exports and total; nominal imports during the 1972 to 2002 period at 0.98 was incredibly high while the correlation between nominal exports and nominal imports as a percentage of nominal GDP was 0.77. As a result, the nominal depreciation of the U.S. dollar since 1972 has been closely correlated with expansion in U.S. total nominal exports and equally correlated with the growth in U.S. total nominal imports leading to no reductions in the U.S. trade deficit (Blaine, 1996). Furthermore, Husted (1992) maintains a similar view that no matter how U.S. total exports and total imports are measured, they are always highly correlated with each other. But this may not be necessarily true for the bilateral real trade among the U.S., Canada, and Japan. Rahman, Mustafa and Burckel, (1997) find no long-run association but find some short-run relationship between the yendollar real exchange rate changes and the U.S.-Japan real trade balance. Rahman and Mustafa (1999), using annual data from 1970 through 1995 and the ADF test of co integration, find no co integrating (long-term equilibrium) relationship between these variables. However, they find some short-run bidirectional Granger causality. Despite a wide publicity and profound ramifications for U.S. open-door trade policy towards Canada, and Japan, a few studies have addressed the issues of real export, real import and real exchange rate linkages in depth while numerous casual observations and detailed descriptive accounts using nominal data are readily available. Formal modeling and data analyses at the theoretical and empirical levels are limited in number. More recently, BahmaniOskooee and Ratha (2004) considering 18 major trading partners of the U.S. is unable to discover any J-curve pattern in the short run. But favorable effects of real depreciation of the dollar on the U.S. trade balance are confirmed in most cases. U.S.-Japan and U.S.-Canada Bilateral Real Trade Balances: An Empirical Exploration 127 There is also an evolving literature about estimates of U.S. and/or Japan exports and import functions using a co integration approach [Caporale and Chui (1999), Carone (1996), Clarida (1994), Geglowski (1997), Hamori and Matsubayashi (2001), Hooper and Marquez (1995), Marquez (1999), Parsons (2002)]. To state briefly about the latest ones, Hamori and Matsubayashi (2001) empirically analyze the stability of the Japanese import demand function using the concept of co integration. The results do not support the presence of a stable co integrating relation among real GDP, real imports and relative import prices. Parsons (2002), tests for co integration of Japanese real imports of semiconductors, real gross domestic expenditure (GDP) and relative prices. This paper concludes that rapid growth of imports during the 1980s and 1990s was attributed to growth in real GDE and change in relative prices independent of the Voluntary Import Expansion (VIE) policy. Marquez (1999) investigates longperiod trade elasticity’s for Canada, Japan and the U.S. The elasticities are found to be inconsistent with the view that income and prices affect imports. EMPIRICAL DESIGN The causal relationship between real exchange rate and real trade balance may be explained in terms of "income" and "substitution" effects. Theoretically, there are two possible connections between the balance of trade and the exchange rate. One is the "income effect" in which the dominant influence runs from the trade balance to the exchange rate. The other is the "substitution effect" in which the effect stems from exchange rate to the balance of trade. With the income effect, an exchange rate appreciation in real term will lead to substitution of exports for imports and hence a deterioration in the balance of trade. However, the income effect generally dominates the substitution effect through the business cycle because it takes longer for prices than income to effect trade flows. To provide theoretical motivation for this empirical work, the following export and import functions are specified: Vx = f (e, y*).................(1) Vm = f (e, y)..................(2) Where, Vx = real value of exports in local currency, e = bilateral real exchange rate (units of home currency per U.S. $), y* = foreign real GDP, Vm = real value of imports in local currency, and y = real domestic GDP. Both e and y* exert positive influences on Vx. As home currency depreciates against U.S. $ in real term, the exports become cheaper in foreign markets resulting in higher exports. At the same time, a surge in foreign real GDP boosts foreign demand for goods causing exports to rise. Conversely, depreciating local currency against U.S. dollar in real term makes foreign goods more expensive in local markets. As a result, imports decline. Also, an increase in domestic real GDP spurs domestic demand for goods resulting in higher imports. Southwestern Economic Review 128 From equations (1) and (2), the real trade balance (TB) is formulated in ratio form for scaling purposes ( Bahmani-Oskooee and Brooks, 1999) as follows: Vm Vx = F (e, y y * ) Or TB = f (e, y y * ) (3) A univariate analysis is conducted to investigate the stationary properties for each time series by implementing the most commonly used ADF test (Dickey and Fuller, 1981), and its opposite counterpart, the KPSS test (Kwiatkowski, et al., 1992). To clarify further, the ADF test is about data nonstationary assuming a unit root in each time series while the KPSS test is about data stationary assuming no-unit root in each time series. A time series variable to be non-stationary, I) its variance is time-variant and it goes to infinity as time approaches infinity, ii) it depicts no long run mean-reversion, and iii) theoretical autocorrelations do not decay but the sample correlogram dies out slowly in finite samples. To be co-integrated, all variables must have the same order of integration (Engle and Granger, 1987). They reveal I (1) behavior, if stationarity is induced on the first differencing of the level data. Next, the co-integrating relationships (the tendency for variables to move together in the long run) between or among the variables are determined by using the VAR approach as developed in Johansen (1988, 1991), and Johansen and Juselius (1990, 1992). The appropriate lag-length (P) is selected with the aid of the FPE criterion (Akaike, 1969) to reduce the problem of autocorrelation in residuals or to ensure that the errors are white noise. This helps outcome the problem of over/under parameterization that may induce bias and inefficiency in the estimates. The analysis commences with a congruent statistical system of unrestricted reduced forms as follows: