An Improvement of Quality of Statistical Matching for Survey Data Using Dynamic Caliper

Nowadays, matching is a widely used technique to estimate program net effects. The goal of the method is to establish a counterfactual state by choosing from the control pool a group that is similar to those in the treatment group. In this article we propose a modification of the matching with caliper procedure. The novelty in our approach is setting the caliper value as a fraction of estimated propensity score. The simulation results and examples are presented. Using Deheija and Wahba (1999) data benefits of the proposed approach are stressed. The obtained results indicate that proposed approach is more efficient than the one traditionally used.