DABUS Carlos Dario
Reexamining the link between Instability and Growth in Latin America: a Dynamic Panel Data Estimation using K-MEDIAN Clustered Data
CECILIA BERMÚDEZ, CARLOS DABÚS, GERMÁN GONZÁLEZ
Latin American Journal of Economics
Pontificia Universidad Católica de Chile
Lugar: Santiago de Chile; Año: 2015 vol. 52 p. 1 - 1
We estimate a dynamic panel data model to empirically assess the relationship between instability and growth in Latin America over the last fifty years. We employ the inflation rate in logarithm and the volatility of the GDP growth rate as proxies of macroeconomic instability. Following the most recent literature, we test the hypothesis of a non-linear relationship between the instability proxies and the economic growth. However, these variables are likely to be contaminated by outlying observations, due to the several crises experimented by Latin American economies. The presence of outliers seriously distorts the results of classical estimations and leads to spurious regime-finding as outliers can be mistaken for a regime. However, this issue is rarely dealt with in the econometric literature. Our contribution is to use the k-mean clustering algorithm as a pre-estimation technique in order to mitigate the outlier problem and to properly identify different scenarios for the instability variables, based on the clusters means. We find that the clustering procedure indeed help capture the different growth performance of Latin American economies at different scenarios of the explaining variables. We conclude that inflation has a significant and positive effect at low levels of the variable, and becomes harmful to economic growth at scenarios with average annual inflation rate of 57%. As for the traditional instability measure, the GDP growth rate volatility has also a negative and significant impact on growth that is robust to all the specifications, with slight changes over the different data grouping presented in this work.