Forecasting stability and growth pact compliance using machine learning

Abstract

Abstract The 2011 reform of the Stability and Growth Pact (1996) strengthened the European Commission's monitoring of EU member states' public finance. Failure to comply with the 3% limit on public deficit triggers an audit. In this paper, we present a machine learning based forecasting model for compliance with the 3% limit. We use data from 2006 to 2018 (a turbulent period including the Global Financial Crisis and the Sovereign Debt Crisis) for the 28 EU member states. After identifying 8 features as predictors among 138 variables, forecasting is performed using a support vector machine (SVM) algorithm. The proposed model achieved a forecasting accuracy of nearly 92% and outperformed the logit model used as a benchmark.