EXPORT LED GROWTH vs. GROWTH LED EXPORTS: LDCs EXPERIENCE
Bahmani-Oskooee, Mohsen, The Journal of Developing Areas
Previous studies that investigated the export led growth hypothesis, relied upon Granger or Sims causality detection approach. Since these approaches employ first differenced variables, any inference could be short-run in nature. In this paper we consider the relation between exports and economic growth to be a long-run phenomenon. After applying Johansen's cointegration technique to establish the long-run relationship between exports and output (in presence of other factors), we rely upon weak exogeneity tests proposed by Johansen to establish exogeneity of exports or output. Annual data over 1960-99 period from 61 countries are used for analysis. The results are country specific and there is no uniform pattern.
JEL Classifications: F10
Keywords: Export Growth, Income Growth, Cointegration
One of the areas in international and development economics that has received a great deal of attention by researchers is the relation between export growth and economic growth. Based on economic theory, one could easily postulate that an increase in exports leads to an increase in real GDP through the well-known multiplier effect. On the other hand, increase in real GDP could lead to realization of economies of scale and cost reduction that could, in turn, boost exports.
Due to lack of long time-series data, early studies employed either crosssectional data or panel data to establish the relation between export growth and economic growth. The list includes Kravis (1970), Michalopoulos and Jay (1973), Voivodas (1973), Michaely (1977), Balassa (1978a, 1978b, 1982, 1985), Heller and Porter (1978), Tyler (1981), Feder (1983), and Kavoussi (1984). In general, these studies provide support for the positive relation between export growth and economic growth. However, these studies have well-known weaknesses, including uncertainty about exogeneity of export variable. Furthermore, they do not incorporate country specific factors into analysis due to the nature of the data (cross-sectional).
To overcome the deficiencies associated with cross-sectional studies, for any country that enough time-series observations are available, researchers employed timeseries data and using ordinary least squares estimation method tried to establish the relation between export growth and economic growth. The list includes Krueger (1978), Ram (1987), Salvatore and Hatcher (1991) and Sengupta (1993). These studies did not investigate whether causality is from export growth to economic growth or vice versa. Given the evidence of unit root in most macro variables, they used non-stationary data. Thus, their estimates could be spurious.
Since introduction of Granger's (1969) and Sims' (1972) concept of causality within time-series framework, researchers shifted their emphasis toward investigating causality between export growth and output growth. The list includes Jung and Marshall (1985), Chow (1987), Ahmad and Kwan (1991), Bahmani-Oskooee et al. (1991), Ahmad and Harnhirun (1992), Hutchison and Singh (1992) and Dodaro (1993). In general, these studies have failed to provide strong support for export led growth (ELG) or growth led exports (GLE) hypothesis. One reason, as pointed out by Bahmani-Oskooee and Alse (1993) is that they have ignored to incorporate the cointegrating properties of exports and output in their testing procedure. Engle and Granger (1987) show that when two variables are cointegrated, there is an additional channel through which one variable can Granger cause the other variable and that is through what they call an error-correction term. Indeed, when Bahmani-Oskooee and Alse (1993) employ granger causality test inclusive of an error-correction term, they do provide strong support for bi-directional causality between exports and real GDP in almost all countries that they consider. Subsequent authors such as Kugler and Dridi (1993), van den Berg and Schmidt (1994), and Ahmad and Harnhirun (1995) who employed cointegration and error-correction modeling, also provided some support for bi-directional causality. …