The export promotion strategies have received substantial attention in the development literature. A number of empirical analyses have employed the Granger’s causality test based on vector autoregressive (VAR) models to investigate the export-led growth hypothesis (ELGH). But as Granger (1988) pointed out, traditional Granger’s or Sims’ tests are likely to provide invalid causal inferences if the time series are I (1) and cointegrated.
The primary purpose of this paper is to reexamine the relation between export growth and economic growth in the APEC countries by taking into consideration the issues of unit roots, integration, and cointegration. So, we apply the methods of the cointegration approach and error-correction models (ECMs) to re-investigate the causality relationship between exports and economic growth for the APEC region.
The secondary purpose of this paper is to investigate whether the export promotion strategy is valid for any developing economies. There are eighteen members in APEC now. According to their economic development level, we can classify them into three groups: developed countries, newly industrialized countries (NICs), and developing countries. We try to see if there are any differences in the export promotion strategies of the three groups.
The empirical results show that bi-directional causality between export growth and economic growth receives strong empirical support in the NICs (Hong Kong, South Korea, Singapore, Taiwan). The export promotion strategies are more successful in NICs than in developed and developing countries.
In addition, we compare the empirical results of the Granger’s causality test in VAR models and ECMs. We conclude that if time series are I(1) and cointegrated, the use of causality tests in ECMs is more encouraging than in VAR models because the traditional time series technique could have missed some of the forecastability and hence reached an incorrect conclusion about non-causality of the mean. On some occasions, causations could be present, but would not be detected by the testing procedure used (the VAR model). But if the time series does not exhibit a cointegration relationship, we cannot use the Granger representation theorem to apply the ECMs, so we can only use the VAR model to test causality relationships in this case.
