Trade, growth and income

Abstract

This paper uses the instrumental variable threshold regressions approach of Caner and Hansen (2004) to investigate whether the trade's contribution to the standard of living and long-run economic growth varies according to the level of economic development. The empirical evidence shows that greater trade openness has strongly beneficial effects on growth and real income for the developed countries but significantly negative effects for the developing countries. The heterogeneity in the relationships suggests that greater international trade and integration may foster uneven development and hence contribute to more diverging economies. In addition, the link of trade to economic performance is found to work through both capital accumulation and productivity growth channels. Finally, the evidence shows that real effects of trade also depend on the level of financial development, inflation and trade openness.

Keywords:

• trade openness
• economic growth
• economic development
• instrumental variables
• threshold regression

JEL Classifications:

• C21
• F10
• O11

Notes

1. See Grossman and Helpman (1990, 1991), Rivera-Batiz and Romer (1991), Barro and Sala-i-Martin (1997) and Baldwin, Braconier, and Forslid (2005).

2. See Romer (1989), Ades and Glaeser (1999), Alesina, Spolaore and Wacziarg (2000), and Bond, Jones and Ping (2005).

3. See Sachs and Warner (1995) and Rajan and Zingales (2003).

4. See Edwards (1993) and Lopez (2005) for the detailed survey and references therein.

5. Theoretically, while Ben-David and Loewy (1998), Mountford (1998) and Spilimbergo (2000) postulate that trade is more beneficial to low-income countries, Stokey (1991), Krugman and Venables (1995) and Bhagwati (2002) propose that trade does not necessarily close the gap between the rich and the poor economies. Empirically, while Ben-David (1993, 1996) and Sachs and Warner (1995) show freer trade contributes to convergence, Slaughter (2001) finds no strong systematic link between trade liberalization and convergence.

6. It is noted that the Hansen methodology tests for a general structural break between the poorer and richer countries based on the entire coefficient vector, not just on the trade coefficient.

7. Rodriguez and Rodrik (2000) and Brock and Durlauf (2001) question whether such geographical variables could have effects on growth in their own right and whether this alone could explain the significance of the instrumental estimate of trade constructed out of them. Geography may influence health, endowments or institutions, any one of which could affect growth. However, Frankel and Rose (2002) repeat the instrumental variables approach of Frankel and Romer (1999) and show that the basic conclusion is robust to the inclusion of geographical and institutional variables in the growth equation; that is, trade openness indeed plays a role even after allowing for geography.

8. Dowrick and Golley (2004) also reach similar conclusions that the impact of trade on growth varies in both sign and magnitude with the level of economic development for the period 1980–2000 by using the interactive term between trade and the level of initial income.

9. Although it is common practice to use the Barro-type cross-country growth model, Lee, Pesaran and Smith (1997) note that estimating cross-section growth regressions, or growth regressions using observations based on data averaged over long periods, makes it impossible to consider either the complex dynamic adjustments involved in the countries' output processes or the heterogeneity of growth rates across countries. The use of a threshold regressions model may partly mitigate the problem by allowing different growth experiences for different groups of countries.

10. The Levine et al. (2000) dataset has been widely used by many researchers to examine the relationship between financial development and economic growth. Since the list of variables included in the dataset is quite comprehensive, it allows us to control for omitted variable bias in the relationship between trade and growth. However, the drawback of the use of the Levine et al.' s (2000) data is that the sample period (1960–1995) is out of date. Hence, the period is extended into 2000 with data from the World Development Indicators, World Bank.

11. The total factor productivity growth variable is taken from Beck, Levine and Loayza (2000) and is available only for the 1960–1995 period.

12. This study also tests whether there is a second or third threshold in the relationship. The results show that there exists only one threshold in the relationship, justifying that two-regime threshold regressions are appropriate.

13. This test assesses if the instruments have independent effects of growth, income, investment and productivity growth, respectively, beyond their ability to explain cross-country variation in trade openness. The test statistic is obtained by running the residuals from the second stage regression on the instruments and multiplying the R ² from this regression by the number of observations. Under the null hypothesis that the instruments are not correlated with the error term, this statistic is , where j is the number of instruments, and k is the number of variables instrumented for.

14. For a robustness check, this study also finds consistent evidence using average tariff as an openness indicator. The results are not reported but available upon request.

15. The description below closely follows Caner and Hansen (2004).