Forecasting U.S. recessions with probit models

In this important and highly researched topic, recession forecasting in the United States, we aim to review previous benchmark works on the econometric literature to draw inspiration and try to improve prediction using different methods. The most widely used variable, and also the one by far with the highest prediction power, is the yield curve. We try to add prediction power to this model with two approaches: using other financial variables and trying different kind of models. We also try to improve the forecasts of previous works with approaches using the Lasso variable selection method and with General Autoregressive Score (GAS) models. These models were able to achieve meaningful prediction power in both short and long time forecast horizons.