Export-led growth: are the results robust across methodologies and/or data sets? A case study of Latin America

Abstract

We examine the export-led growth (ELG) hypothesis from a different perspective asking if the choice of data and/or methodology drives the results. We apply the Granger causality test modified by the corresponding error correction model using real export data from two common sources: the International Monetary Fund and the United Nations Commodity Trade Statistics. Additionally, to determine if the level of deflation could be a factor, we deflate the data at a disaggregated level and by a single export unit price index. Outcomes are compared to those obtained through the Granger causality procedure developed by Toda and Yamamoto (1995 Toda, H and Yamamoto, T. 1995. Statistical inference in vector autoregressions with possibly integrated process. Journal of Econometrics, 66: 225–50. [Crossref], [Web of Science ®] , [Google Scholar]) and Dolado and Lütkepohl (1996 Dolado, JJ and Lütkepohl, H. 1996. Making wald tests work for cointegrated VAR systems. Econometric Reviews, 15: 369–86. [Taylor & Francis Online] , [Google Scholar]). We use sixteen Latin American countries to test our hypotheses. The analysis reveals inconsistencies in the results both by selection of data and methodology suggesting much of the debate regarding ELG, at least for Latin America, is fuelled by data and/or methodology choice, putting into question previous studies.