Abstract
Three recent trends in economics stand out: (1) the rise of empiricism, (2) general acceptance of the causal inference framework in econometric analysis, and (3) increasing adoption of the machine learning approach and its greater interaction with econometrics. This study aims to discuss the evolution of econometrics over these main trends and to understand the nature of the interaction between econometrics and machine learning. In its relatively short history, econometrics has made important breakthroughs and has also experienced methodological and paradigmatic shifts. More recently, the main purpose of econometric analysis is to develop methods for unbiased/consistent and efficient estimation of causal economic relations. On the other hand, (supervised) machine learning aims to develop algorithms for solving estimation/prediction and classification problems. The machine learning approach generally provides more successful predictions since it can exploit the bias-variance trade-off optimally compared to the econometric approach where unbiased/consistent and asymptotically efficient estimation is the principal aim. It can be said that the interaction between econometrics and machine learning is shaped by the phenomenal predictive success of machine learning algorithms. This ongoing interaction has resulted in the development of new econometric methods for causal inference and the improvement of the existing ones.