I am a doctoral researcher in economics, on the job market 2024-25. My focus is on developing a machine learning Transformer framework for macroeconometric analysis.

My job market paper proposes a new Transformer algorithm for estimating non-linear dynamic factors from time series data, requiring minimal identifying assumption. The results are improved substantially on small datasets by using a Kalman filter factor model as prior information to guide the estimation. Prior information is supplied into the loss function through a new regularization term. The results can be interpreted with Attention matrices, which show the impact of each variable's every lag on the output. A Monte Carlo esperiment demonstrates that the Transformer is more accurate than the Kalman filter, when the data deviates from linear-Gaussian. In an empirical application the Transformer is used to construct a coincident index for the real economic conditions of the United States.