آثار تنوع بخشی پرتفوی بلوک های تجاری: مورد پیمان نفتا

This study investigates the evolving nature of North American Free Trade Agreement (NAFTA) stock market interdependencies and their association with diversification gains from the perspective of US investors. The issues are addressed for both short- and long-run interdependencies through correlation of stock market returns and cointegration of stock market prices. The basic findings include: (1) the existence of a long-term relationship (a cointegration relation) which is time-varying and statistically unstable and (2) diversification gains with cointegration not consistently lower than without cointegration. Thus, per-unit-of-risk diversification gains to US investors from NAFTA stock markets are determined by return volatilities, return correlations and domestic market performance. Based on increased return volatilities and return correlations and the very small per-unit-of-risk diversification gains even when the US stock market performs poorly, US investors’ diversification gains have diminished since the implementation of NAFTA.

The passage of the North American Free Trade Agreement (NAFTA) in 1994 has reduced (if not completely eliminated) barriers to trade and capital flows and thus reinforced economic interdependencies among its member countries: Canada, Mexico and the US. Inasmuch as stock markets should reflect prospects for the underlying economies, fewer institutional constraints and increasing economic integration in the NAFTA area potentially can lead to stronger short- and long-run comovements or interdependencies of stock market prices and returns in the three NAFTA countries. Conceptually, these relationships should affect the magnitude of diversification gains from cross-border investments in NAFTA stock markets. This study is intended to investigate the evolving nature of NAFTA stock market interdependencies and their association with diversification gains from the perspective of US investors. These two aspects of NAFTA equity markets have not been sufficiently examined or quantitatively verified in prior empirical work. Numerous studies (e.g., Koch and Koch, 1991, Koutmos, 1996 and Michaud et al., 1996) have consistently detected international stock market linkages in the short-run through contemporaneous correlation or determination, lead–lag or Granger–causal relations, or volatility transmission across national stock market returns. The evidence has been particularly strong during and after tumultuous events including the US stock market crash of 1987 (Masih and Masih, 1997) and the Asian Financial Crisis of 1997 (In et al., 2001). Informational efficiency of international stock markets (Eun and Shim, 1989), contagion effects during market turmoil (Climent and Meneu, 2003) and financial market globalization including deregulation and improvements in communication technology (Park and Fatemi, 1993) are among possible factors contributing to increasing short-run interdependencies across national stock markets. Evidence of long-run international stock market interdependence, however, appears comparatively conflicting. Cointegration analyses based on Engle and Granger (1987) and Johansen (1988) have been the techniques typically employed in prior studies to investigate this interdependence. If cointegration is detected across national stock market price indices, they are expected to comove together over time to retain cointegrating or long-run equilibrium relations. Blackman et al. (1994) and Masih and Masih (2002) relate cointegrating relations across various national stock markets primarily to financial market globalization with findings especially pronounced during periods of economic uncertainty such as the US stock market crash of 1987 (Parhizgari et al., 1994) and the Asian financial crisis of 1997 (Ratanapakorn and Sharma, 2002). Conversely, Kanas (1998) and Climent and Meneu (2003), among others, detect very weak or no evidence of cointegration within similar groups of national stock markets. More recently, with monthly data covering the period January 1979–June 2002, Phengpis and Apilado (2004) investigate cointegration of stock market price indices from major European Monetary Union (EMU) countries relative to non-EMU countries. Their results clearly indicate two cointegrating relations within the first group but none within the latter group, implying that the absence of cointegration across national stock markets cannot be generalized across groups of countries. In fact, strong economic interdependence can emerge as an important contributing factor to cointegration of stock markets across national boundaries, a finding supported by similar findings in earlier studies of EU stock markets such as those of Serletis and King (1997) and of Latin American stock markets such as those of Chen et al. (2002). Studies directed toward NAFTA effects on stock market integration provide differing results and implications. With monthly data from November 1987 until March 1997, Ewing et al. (1999) find no evidence of cointegration among stock market price indices in the three NAFTA countries (even with the inclusion of a dummy variable to account for the implementation of NAFTA in January 1994). In a related study, Ewing et al. (2001) report no evidence of daily volatility transmission among NAFTA stock markets during the pre-NAFTA period (1992:06:02–1993:12:31). They however detect significant volatility transmission from the US to the Canadian and Mexican stock markets, but not vice versa, during the post-NAFTA period (1994:01:03–1999:10:28). These two studies jointly imply that the passage of NAFTA has brought about greater short-run linkages, but not necessarily long-run equilibrium relations among the members’ stock markets. In contrast, other related studies report evidence of cointegration among NAFTA stock markets, especially after the passage of NAFTA in 1994. Darrat and Zhong (2001), using weekly data from 1989 to 1999, find no evidence of cointegration during the pre-NAFTA period but evidence of cointegration during the post-NAFTA period. Gilmore and McManus (2004), who study only the post-NAFTA period (1994–2002), report a cointegrating relationship based on both weekly and monthly data. These studies collectively suggest that NAFTA implementation has reinforced equity market integration and linkages among the member nations. Further, Atteberry and Swanson (1997) test for short- and long-run relations among the three NAFTA stock markets for each annual sub-sample from 1985 to 1994 with daily data. The results indicate consistently strong contemporaneous determination (especially between the US and Canada) but intermittent existence of cointegrating relations from one year to the next. They suggest that the exceptionally strong statistical significance of cointegrating relations in 1993 and 1994 potentially reflect economic uncertainty affecting investors’ perceptions of the effects of NAFTA passage on domestic economies similar to the effect of the US stock market crash of 1987. The first issue to be addressed in this study is the evolving nature of short- and long-run interdependencies through correlation of stock market returns and cointegration of stock market prices among the three NAFTA stock markets. Specific attention is given to time-varying relationships. The study period is January 6, 1988–December 31, 2003, and the relationships are analyzed with weekly data: (1) over the full sample period; (2) during the pre- and post-NAFTA periods; (3) over 10- and 5-year moving or rolling windows, where the first observation is dropped while an additional observation is included to keep the sample size and test power constant as the estimation window rolls forward. In so doing, this study provides an update and significant extension of previous literature on NAFTA stock market interdependence. First and conceptually, longer time spans employed should enable more accurate inferences concerning long-run equilibrium relations, especially during the post-NAFTA period. If cointegration across national stock markets is related to strong economic interdependence among the subject countries (Phengpis and Apilado, 2004), it would take time for identifiable interdependence to emerge after the degree of institutional impediments to trade and capital movements within the NAFTA area was lessened. Hence, Ewing et al. (1999) may not have been able to detect cointegration due to an insufficient time horizon covered in their sample after the passage of NAFTA. Second and more importantly, dynamic analysis based on a rolling window approach provides deeper insights into the possibility of ever-changing long- and short-run linkages. It is possible that even with NAFTA the degree of economic ties among its member countries is not so robust (compared to EMU countries, for instance) that a stable cointegrating relationship of its stock markets over time will result. Rather, the relationship may be time-varying and sporadic in that it is discernible statistically during periods of economic uncertainty (e.g., Atteberry and Swanson, 1997 and Ratanapakorn and Sharma, 2002). With this possibility, statistical evidence of cointegration based on static tests during post-NAFTA periods such as in Darrat and Zhong (2001) and Gilmore and McManus (2004) may in fact be sample-dependent and may lead one to incorrectly conclude that NAFTA passage has resulted in greater integration and linkages among the member countries than actually occurred. Further, return correlations among NAFTA stock markets can change drastically (if not increase) over the course of time due possibly to gradual effects of NAFTA and ever-changing news and information transmitted to the markets. Because the exact dates on which considerable changes in long- and short-run relations actually transpired are not known a priori, dynamics of these relations may not be completely or appropriately captured in arbitrarily chosen sub-samples such as pre- and post-NAFTA periods and other non-overlapping intervals. Only a few previous studies (none of which was directed toward NAFTA) have considered these time-varying possibilities. Longin and Solnik (1995) show that the unweighted correlation of returns between the US stock market and six major foreign stock markets appears unstable over time. Barari (2004) investigates equity market integration in Latin America by computing time-varying integration scores based on the Capital Asset Pricing Model (CAPM)'s betas over moving estimation windows. Pascual (2003) employs rolling cointegration tests for long-run equilibrium relations among stock markets in France, Germany and the UK over rolling windows of selected time intervals. This study takes advantage of these methodologies by investigating return correlations and cointegrating relations dynamically over rolling 10- and 5-year windows. This approach can reveal the evolving and time-varying characteristics of short- and long-run relations in NAFTA stock markets, characteristics which have not been detected by methodologies employed in prior studies of these markets. Based on portfolio theory and empirical evidence, the degree of interdependencies across national stock markets affects the magnitude of international diversification gains. These gains to US investors can be measured from an increase in average return when switching from the US stock market to an international efficient portfolio which maximizes average portfolio return at the level of risk (standard deviation) of US stock market returns (e.g., Li et al., 2003). As a result, return correlations or short-run comovements between national stock markets are important inputs in identifying this portfolio and resultant gains. Numerous studies have shown that diversification gains to US investors from international stock portfolios are substantial (e.g., Errunza, 1977 and Harvey, 1991) and especially so if emerging markets are included due to their relatively low return correlations with developed markets (e.g., Bailey and Stulz, 1990 and Li et al., 2003). None of the prior studies have quantified these gains from diversification within the NAFTA area. Further, even if return correlations are low between stock markets, international diversification gains can be minimal in the presence of cointegration because, in that case, stock market prices trend or comove together over extended time horizons (e.g., Blackman et al., 1994 and Chen et al., 2002). Conversely, considerable benefits should be possible in non-cointegrated markets (e.g., Kanas, 1998 and Gilmore and McManus, 2002). Previous studies however have not constructed international stock portfolios for NAFTA countries to quantitatively verify this conceptually convincing claim. Additionally, if return correlations and cointegrating relations are time-varying in nature, diversification gains to US investors from NAFTA stock markets measured over pre-determined sample periods would yield inaccurate measurements and inferences. Measurement of gains through time is more conceptually appropriate and consistent with the dynamics of short- and long-run relations. Thus, the second issue to be investigated in this study is the magnitude of potential diversification gains to US investors from NAFTA stock market investments over time via a rolling window estimation methodology. With this approach, this study provides answers to several inquiries which have not been addressed in prior studies of international stock market diversification. First, diversification gains to US investors, specifically from NAFTA stock markets, are quantified. Second, whether the implementation of NAFTA has brought about cointegrating relations across NAFTA stock markets and resulted in smaller diversification gains to US investors (based on unverified claims by Blackman et al., 1994 and Chen et al., 2002, among others) is investigated in a comparative framework between the pre- and post-NAFTA periods. Third, if short- and long-run interdependencies are time-varying, the gains are likely to differ from one period to the next. Thus, an investigation over rolling windows should allow better insight into the connection between interdependencies among NAFTA stock markets and potential diversification gains to US investors from these markets over time rather than the usual assumption of intertemporal stability across pre-specified subsamples. The remainder of this study is organized as follows. Section 2 describes the data and delineates the econometric methodology employed, with estimation results presented in Section 3. Section 4 delivers conclusions.

This study adds two new dimensions to prior empirical work on short- and long-run interdependencies of NAFTA stock markets and on potential diversification gains to US investors from international stock portfolio investments. The data are stock market price indices in Canada, Mexico and the US covering the period from January 6, 1988, to December 31, 2003. The first dimension is the evolving and time-varying characteristics of NAFTA stock market interdependencies. Through return correlations and cointegrating relations, short- and long-run interrelationships are analyzed dynamically over rolling 10- and 5-year windows in addition to investigations over the full sample, the pre-NAFTA and the post-NAFTA periods. This is in contrast to previous studies which test only the full sample and non-overlapping, pre-determined subsamples—an approach which cannot properly or completely capture the dynamics of short- and long-run interdependencies because changes in these relationships are not known a priori. The second dimension provides insight into the magnitude of potential diversification gains to US investors from NAFTA stock market investments. While stock market interdependence and diversification gains in an international context have been investigated separately for other groups of markets, the two aspects have not been examined together. This joint approach provides better understanding of the relationships between interdependencies and gains, especially over time via rolling window methodology. The results from static analysis over the full sample, during the pre-NAFTA period and during the post-NAFTA period, consistently indicate the absence of long-run interrelations among NAFTA stock markets through a cointegrating relationship. These results support those in Ewing et al. (1999) who report no evidence of cointegration even if NAFTA passage in 1994 is taken into consideration as a dummy variable. However, the results differ from Darrat and Zhong (2001) and Gilmore and McManus (2004) who report evidence of cointegration during different pre-determined post-NAFTA periods. Thus, it is possible that long-run interrelations among NAFTA stock markets are sample-dependent, or specifically, time-varying and require dynamic rather than static analysis such as rolling cointegration methodology for accurate interpretations. This paper's dynamic analysis for 10- and 5-year rolling windows indeed identifies the presence of a time-varying, statistically unstable cointegrating relationship among NAFTA stock markets. Strongest significance of cointegration occurs over the window which covers two major tumultuous events, the 1994–1995 Mexican peso crisis and the 1997–1998 Asian financial crisis. Discernible cointegration evidence during periods of international economic turmoil suggests that stock market prices in each nation are responding to the same set of adverse news and information rather than reflecting economic fundamentals. This implication is consistent with the conclusion in Atteberry and Swanson (1997) who relate a cointegrating relationship with looming uncertainties in NAFTA stock markets. The static and dynamic results collectively imply that increased intra-NAFTA trade and capital flows, without other coordinated policies to help align domestic economic variables, have not integrated the three NAFTA economies to an extent sufficient that a time-invariant, statistically stable cointegrating relationship among their stock markets has evolved. Without dynamic analysis to reveal the time-varying nature of cointegrating relationships, researchers erroneously may infer, when relying on static analysis over fixed post-NAFTA periods such as in Darrat and Zhong (2001) and Gilmore and McManus (2004), that linkages among NAFTA stock markets are stronger than they are in reality. In contrast, short-run interdependencies among NAFTA stock markets appear to have strengthened based on increases in the magnitudes of pair-wise return correlations over time. These increases may result from various contributing factors such as relaxation of trade barriers as a result of NAFTA.9 Increases are especially pronounced between Canada and Mexico and between the US and Mexico, but less strong between the US and Canada, the two countries between which trade and capital flows barriers had been less restrictive and where pair-wise return correlations had been relatively high even before NAFTA. The increased return correlations with Mexico are also consistent with higher return volatilities for stock markets in Canada and in the US after NAFTA. Per-unit-of-risk diversification gains to US investors are appreciably lower in the periods following NAFTA passage, reflecting increases in return volatilities and in short-run interdependencies among the markets through return correlations. This is true for the entire post-NAFTA period and for 10- and 5-year post-NAFTA investment horizons irrespective of timing of portfolio formation. When the US stock market performs poorly, even a small positive average return on the US-equivalent-risk NAFTA efficient portfolio can result in considerable gains to US investors through return differential (ΔR). This portfolio, however, cannot deliver per-unit-of-risk gains in the magnitude obtainable during the period prior to NAFTA. In summary, the findings indicate that, because a cointegration relation is time-varying and statistically unstable, long-term trends may not necessarily nor consistently explain the magnitude of per-unit-of-risk gains to US investors through time. Gains during the period when cointegration is present are not necessarily lower than gains during the period when cointegration is absent. Moreover, even when cointegration is detected at the same level of statistical significance during two different periods, per-unit-of-risk gains can vary appreciably. Thus, it is not appropriate to conclude that the absence (presence) of cointegrating relations across national markets necessarily implies that appreciable (minimal) international diversification gains can be acquired, especially if these relations are intertemporally unstable. It appears that per-unit-of-risk diversification gains to US investors from NAFTA stock markets are determined by return volatilities, return correlations and domestic market performance. Given the increased return volatilities and return correlations and the very small per-unit-of-risk diversification gains even when the US stock market performs poorly, US investors’ diversification gains have diminished since the implementation of NAFTA.