Analyzing E-Learning Adoption via Recursive Partitioning

The paper analyzes factors that influence the adoption of e-learning and gives an example of how to forecast technology adoption based on a post-hoc predictive segmentation using a classification and regression tree (CART). We find strong evidence for the existence of technological interdependencies and organizational learning effects. Furthermore, we find different paths to elearning adoption. The results of the analysis suggest a growing "digital divide" among firms. We use cross-sectional data from a European survey about e-business in June 2002, covering almost 6,000 enterprises in 15 industry sectors and 4 countries. Comparing the predictive quality of CART, we find that CART outperforms a traditional logistic regression. The results are more parsimonious, i. e. CARTs use less explanatory variables, better interpretable since different paths of adoption are detected, and from a statistical standpoint, because interactions between the covariates are taken into account.