Income convergence dynamics in ASEAN and SAARC blocs

Abstract

This paper examines whether convergence of real income exists among the Association of South East Asian Nations (ASEAN) and South Asian Association for Regional Cooperation (SAARC) countries for the period covering 1970–2009. Univariate Lagrange Multiplier (LM) unit root tests with structural break(s) are employed to check for the incidence of stochastic convergence, which is necessary for conditional convergence as proposed in the neoclassical model. We further examine the presence of β-convergence, which is considered as the sufficient requirement for conditional convergence. Test results suggest convergence among ASEAN members, with the absence of convergence in SAARC member countries.

Keywords:

• income convergence
• ASEAN
• SAARC
• structural breaks

• JEL Classifications: C22
• O47
• O57

Notes

1. Any evidence for the existence of either unconditional (absolute) convergence or conditional convergence is an indication of β-convergence of output across countries. Another form of convergence is the sigma () convergence, which focuses on the dispersion of cross-country output over time. According to Quah (1993) and Friedman (1992), convergence should be evaluated directly by investigating the variance of output across countries, as against evaluating convergence indirectly through the sign of β. Therefore, β-convergence is a necessary but not sufficient condition of-convergence. Despite the highlighted drawback, β-convergence continues to enjoy most attention of convergence papers as it provides information regarding structural parameters of growth models, whereas -convergence usually does not provide such information (Islam, 2003).

2. Closely related to conditional convergence is the club convergence. However, it is not easy to empirically distinguish between the two concepts (Durlauf & Quah, 1999; Islam, 2003).

3. The arrangement was to ensure that all members provide duty-free access; exclusive tariff preferences or deeper tariff preferences for the export; the removal of non-tariff barriers; and the removal, where appropriate, of para-tariff barriers of the goods generating from the least developed member states.

4. There are specifications with break(s) in level(s) only. Sen (2003a) argued that specification which provides for both breaks is better than specification with only break(s) in level(s). Sen (2003b) used Monte Carlo simulations to validate the superiority of specifications that provide break(s) for both level(s) and trend(s).

5. To decide the lag length of k we apply the process suggested by Ng and Perron (1995). Beginning with an upper bound k_max on k, k = k_max is selected if the existing lag is significant. If not, k is reduced by a unit the lag is significant. If none of the lags are significant, then k = 0. In the empirical section, we set k_max = 4 and employ the 10% value of the asymptotic normal distribution, 1.645, to evaluate the significance of the last lag.

6. We report the β-convergence results of countries with stochastic convergence only.