México
This article presents a critical appraisal of three different econometric techniques commonly employed to analyze causal relationships among economic series. Our results indicate that the empirical application of the Granger causality test, the Engle-Granger cointegration test and the Hausman test for causality performed with small samples suffers severe size distortions, and therefore that the results should be taken with caution.
Furthermore, we show that these tests produce better results if the series are differentiated. Our results are applied to the series for consumption and GDP in Mexico and the US and suggest that these series are cointegrated in the case of the US only (causality and cointegration are different). We comment upon these results in relation to the conclusions of Guisan (2004), and other related studies, in which several methods are used to analyze the bilateral causality between consumption and GDP in Mexico and the US and where it was found that cointegration and Granger causality tests may fail to detect the true causal relationships.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados