Sustainability of external debt: further evidence from non-linear framework

Abstract

This study investigates sustainability of external debt under a two-step non-linear framework. The first step uses a general linearity test proposed by Harvey and Leybourne (2007) to determine the linearity property of external debt. The second step applies a non-linear ADF unit root test proposed by Kapetanios, Shin, and Snell (2003) on the non-liner processes and the linear ADF test on the linear processes to examine the sustainability of external debt. The analysis of 36 debt and 55 current account ratios identifies strong evidence of non-linearity and sustainability. The results indicate superior performance of the non-linear unit root test over the ADF test in determining the stationary property of the data.

Keywords:

• external debt sustainability
• stationarity
• linearity test
• non-linear unit root test

JEL Classifications:

• C4
• C5
• F4

Acknowledgements

We are grateful to David Harvey, School of Economics, University of Nottingham, for providing us with the GAUSS codes for the linearity test. We also acknowledge Matiur Rahman, Kofi Amoateng, Miguel Leon-Ledesma, and anonymous referees for helpful comments, suggestions, and support at various stages of the manuscript preparation.

Notes

1. The choice of b, for a given level of significance, ensures that the asymptotic critical values of the modified Wald statistic coincide with the critical values of a distribution for the corresponding significance level. Harvey and Leybourne (2007) provide three values of b (0.2406, 0.2272 and 0.2034) for 10%, 5%, and 1% significance levels respectively.

2. From the WDI source, based on the availability, we use the longest time span data for external debt service payments for each country. Existing studies, examining the sustainability of external debt measures of developing countries, largely use data either of similar or shorter length. Sample size for each country is shown in Tables 1 and 2.

3. The H-L linearity test statistics from the rescaled debt measures are 7.8^∗∗∗, 11.99^∗∗, and 12.67^∗∗ for Columbia, Panama, and Paraguay respectively. While the KSS test statistics are, respectively, –0.75, –3.63^∗∗, and –3.06^∗∗.

4. One referee raised the issue of robustness of the KSS test to the presence of extreme observations. We examine this issue by excluding the first three observations from Columbian debt data appear to be exhibiting sudden upward shift. The conclusion from the KSS test remains the same. Future studies may explore this matter in more detail.