On the Estimation of Okun’s Coefficient in Some Countries in Latin America : A Comparison between OLS and GME Estimators

We explore Okun's coefficient in several Latin American countries for the period from 1995-2017 and compare the results obtained using ordinary least squares (OLS) and generalised maximum entropy (GME) estimators. There are several advantages in considering the GME estimator over traditional regression approaches. First, it can estimate the parameters of an equation without imposing constraints on the probability distribution of the errors. Second, empirical and simulation studies available in the literature showed that GME worked well for ill-posed problems (e.g., when a model estimate is performed using a small sample of data). Among the main findings, we confirm the inverted relationship between changes in the unemployment rate and real gross domestic product growth in the explored countries except for Perù. Okun's coefficient and the associated confidence intervals obtained by applying GME were very close to those obtained from OLS. Therefore, we did not observe a gain when using the GME estimator rather than the classic OLS approach