Give me a U, give me a V, give me an L!: How effective are countercyclical policies in shaping the output dynamic during recessions

In this paper, we argue that studying recession events through the shapes observed in sequences of GDP data can help to avoid methodological complications and offer new insights to traditional inquiries. With a set of 147 recession events from 77 countries, we analyze whether the shape of the output dynamic might be affected by the application of countercyclical policies. Firstly, we apply a machine learning technique to discover and cluster the shapes (or ‘shapelets’) that prevail in empirical spells. Secondly, we use a multinomial model to study fiscal and monetary interventions, in which we specify the categorical variable with a set of statistically different shapelets. Not only do we find strong empirical evidence that it is possible to overcome a recession through countercyclical policies, but also that there are non-linear effects that make it more likely when the strength of these policies crosses certain thresholds.