Some Tools for Robustifying Econometric Analyses

Abstract

We use automated algorithms to update and evaluate ad hoc judgments that are made in applied econometrics. Such an application of automated algorithms robustifies empirical econometric analyses, it achieves lower and more consistent prediction errors, and it helps to prevent data snooping. Tools are introduced to evaluate the algorithm, to see how configurations are updated by the algorithm, to study how forecasting accuracy is affected by the choice of configurations, and to find out which configurations can safely be ignored in order to increase the speed of the algorithm. In our case study we develop an algorithm that updates ad hoc judgments that are made in Cápistran and Timmermann's (2009) attempt to beat the mean survey forecast. Many of these ad hoc judgments are often made in time series forecasting and have hitherto been overlooked. We show that our algorithm improves their models and at the same time we further robustify the stylized fact that the mean survey forecast is difficult to beat. JEL classificatie is trouwens C52, mocht je dat nodig hebben.