How do machine learning algorithms perform in predicting hospital choices? evidence from changing environments

Researchers have found that machine learning methods are typically better at prediction than econometric models when the choice environment is stable. We study hospital demand models, and evaluate the relative performance of machine learning algorithms when the choice environment changes substantially due to natural disasters that closed previously available hospitals. While machine learning algorithms outperform traditional econometric models in prediction, the gain they provide shrinks when patients' choice sets are more profoundly affected. We show that traditional econometric methods provide important additional information when there are major changes in the choice environment.