Hybrid seminar: Predicting Firm Exits with Machine Learning: Implications for Selection into COVID-19 Support and Productivity Growth

Evaluations of business support programs often face the problem of accurately predicting which firms leave the market. We address this problem, by first showing that a machine learning model better predicts firm exits than conventional methods. This improvement is mostly due to the use of high-dimensional firm data. Second, we provide new insights on the short-term selection effects into COVID-19 support in 2020 in the Netherlands. We find that firms with a lower predicted exit probability are more likely to use support. But because the benefits of support - as measured by prevented exits - are increasing in the probability to leave the market, we conclude that COVID-19 support also reduced the cleansing mechanism of economic crises but less than previously thought.