From Machine Learning to Structural Econometrics with Discrete Variables – MLSEDV

Machine Learning (ML) is experiencing an exceptional development in many fields. The success of ML has several reasons, including the ability to capture more complex relationships than those implied by linear and non-linear parametric or semi-parametric models. An important method in ML is the tree method, whose main idea is to create classes of homogeneous individuals, which in practice corresponds to a relevant discretization of the variables of interest. The main idea of the project is to study structural time series and structural micro-econometric models and to apply them to discretized variables to use the power of ML. The project will develop the theory in time series, micro-econometrics and spatial models. Special attention will be given to the identification of the shocks in the dynamic case. An advantage of the approach in micro-econometric models is the ability to relax some strong assumptions used to study non-linear and non-separable models. The project will face some difficulties that will be addressed: the misspecification's consequences of the discretization; the definition and the interpretation of the structural parameters of the discrete models; the optimal way to discretize the continuous variables; the parsimony of the model. For the later problem, the solutions are grouping of classes; the use of factor models; the sparsity and regularization methods; the parametrization of the model; and working with partial identification. We will do several empirical analysis like the effects of uncertainty on economic growth; the asymmetric responses to shocks during booms and recessions; the asymmetric effects of oil price shocks; the systemic risk management; the housing market in France; the forced technological adoption and spatial externalities in agriculture; the first world war’s impact on the change of location of the French industry; the entry games; and the impact of school building programs on socio-economic outcomes in sub-Sahara countries.