A Projection-Based Framework for Integrating Machine Learning into Macro finance Models

This research introduces a supervised machine learning projection method to approximate expectations in the optimality conditions of economic models. Building upon the parameterized expectations algorithm (PEA) with stochastic simulation, the proposed neural network-based method efficiently addresses challenges such as multicollinearity and the curse of dimensionality. The study extends existing optimal debt management models to include four maturities, uncovering insights about medium-term maturities’ role in fiscal policy. Our findings demonstrate that the optimal policy actively leverages medium-term maturities to stabilize financial income during economic shocks without increasing total debt.