The impact of openness and domestic performance on growth using nonparametric estimation

Abstract

This article distinguishes openness from domestic performance as different growth determinants and uses nonparametric estimation to reveal their direct effects on GDP. On average, openness promotes growth, but its effect on growth is negative when the levels of openness and domestic performance both are very low. The effect of domestic performance on growth is always positive averagely and locally, but its effect on growth is decreasing when the economy experiences the initial stage of globalization and domestic performance. Our result makes a useful contribution to the globalization debate because the distinction between openness and domestic performance shows how different economies can perform and improve in each aspect. Among the developing economies, openness is a necessary and prerequisite condition for growth, but the achievement of a high level in domestic performance is also important for growth.

Acknowledgements

The authors are indebted to earlier comments from Cheng Hsiao, Anne Harrison, Sven Arndt, Michael Funke, Steven Durlauf, Chris Milner, Gianluca Grimalda, Olivier Blanchard, Tony Cockrill, Daniel M. Bernhofen, Toke Aidt, Nikolaus Wolf, Daniel J. Henderson, Aart Kraay, Triggvi Herbertsson, Axel Dreher, Shiyi Chen, conference participants in St. Petersburg University and Fudan University and seminars participants in the University of Cambridge, Nottingham University and University of Warwick. Research funding from Hong Kong APEC Study Center, City University of Hong Kong (Strategic Research Grant project number 7002433) and National Natural Science Foundation of China (Grant: 70971143) and research assistance from Julian Chow and Joseph Y. Y. Wong are gratefully acknowledged. The usual disclaimer applies.

Notes

¹The data for the two indices and their rankings among our sample countries are available from the authors upon request.

²Here we only present the nonparametric estimation of the panel data model with fixed effects because the results from random effects model are similar.