Juliacon 2024

The deferred acceptance (DA) propensity score approach is a highly credible causal inference method often used in education research due to its ability to approximate a randomized control trial (RCT) closely. However, this method is typically very complicated and computationally expensive to run due to the requirement of specialized knowledge about intricate school assignment systems with a lot of iterative steps and the need for a large number of simulations, i.e., assignment runs. No published package is available in any programming language ecosystem that implements and runs the method, although ad hoc code exists in SAS and R. We developed a package in Julia to run the DA propensity score approach that will be easy for applied researchers in education and impact evaluation research to use and implement and that automatically parallelizes tasks. We plan to benchmark our package against the previous ad hoc implementations to assess package performance. Future work will include incorporating other causal inference methods for a more complete package that allows users to test their own causal inference methods against the DA propensity score approach, all with an eye towards reducing the computational burden of these tests while providing credible causal inference.