Most decision tree classifiers are designed to classify the data with categorical or Boolean class labels. Unfortunately, many practical classification problems concern data with class labels that are naturally organized as a hierarchical structure, such as test scores. In the hierarchy, the ranges in the upper levels are less specific but easier to predict, while the ranges in the lower levels are more specific but harder to predict. To build a decision tree from this kind of data, we must consider how to classify data so that the class label can be as specific as possible while also ensuring the highest possible accuracy of the prediction. To the best of our knowledge, no previous research has considered the induction of decision trees from data with hierarchical class labels. This paper proposes a novel classification algorithm for learning decision tree classifiers from data with hierarchical class labels. Empirical results show that the proposed method is efficient and effective in both prediction accuracy and prediction specificity.