Do machine learning methods (random forest and lasso) predict U.S. IPO withdrawals more accurately than conventional regression models (logit)? Dr Annika Reiff analyses this question in her article "Comparing the Prediction Performance of Random Forest, Lasso, and Logit in the Context of IPO Withdrawal". Results show that random forest outperforms both logit and lasso in in-sample and cross-sectional out-of-sample predictions when the training and test sets are drawn from the same time period. However, when models are trained on past data and tested on future observations, all models fail to accurately predict IPO withdrawal. The main reason for this finding is the presence of "concept drift" - a change in the relationship between predictors and IPO withdrawal over time. Concept drift occurs at multiple points in time, affects various predictors, and persists even when accounting for economic shocks, institutional changes, or different prediction horizons. These findings suggest that the generalizability of previous results on IPO withdrawal is limited.
Dr Annika Reiff's article has been published in the journal Intelligent Systems in Accounting, Finance and Management.
