Abstract
This paper provides a link between causal inference and machine learningtechniques - specifically, Classification and Regression Trees (CART) - inobservational studies where the receipt of the treatment is not randomized, butthe assignment to the treatment can be assumed to be randomized (irregularassignment mechanism). The paper contributes to the growing applied machinelearning literature on causal inference, by proposing a modified version of theCausal Tree (CT) algorithm to draw causal inference from an irregularassignment mechanism. The proposed method is developed by merging the CTapproach with the instrumental variable framework to causal inference, hencethe name Causal Tree with Instrumental Variable (CT-IV). As compared to CT, themain strength of CT-IV is that it can deal more efficiently with theheterogeneity of causal effects, as demonstrated by a series of numericalresults obtained on synthetic data. Then, the proposed algorithm is used toevaluate a public policy implemented by the Tuscan Regional Administration(Italy), which aimed at easing the access to credit for small firms. In thiscontext, CT-IV breaks fresh ground for target-based policies, identifyinginteresting heterogeneous causal effects.