Predictors of TFP growth in European countries

This paper aims at identifying relevant indicators of TFP growth in EU countries during the recovery phase following the 2008/09 economic crisis. We proceed in three steps: First, we estimate TFP growth by means of stochastic frontier analysis (SFA). Second, we perform a TFP growth decomposition in order to get measures for technical progress (TP), changes in technical efficiency (CTE), in scale efficiency (CSC) and in allocative efficiency (CAE). And third, we use BART—a non-parametric Bayesian technique from the realm of statistical learning—in order to identify relevant predictors of TFP growth and its components from the Global Competitiveness Reports. We find some indicators to show quite stable relationships with TFP growth. In particular, indicators that characterize technological readiness, such as broadband internet access, are outstandingly important in order to predict technical progress. The inflation rate is a major predictor of TFP growth in lower-income new EU members. Our results identify areas in which further action could be taken in order to increase economic growth. It becomes obvious that machine learning techniques might not be able to replace sound economic theory but they help separating the wheat from the chaff when it comes to selecting relevant indicators of TFP growth.