Regression-Discontinuity Design is a non-experimental method to estimate the impacts of social welfare programs in situations where the treatment assignment is determined by whether an observed variable (running variable) is above or below a known cutoff point. The main idea behind RDD is that individuals whose running variable is just above or just below the cutoff are similar, and so, any differences in the outcome between the two groups (just below and just above) may be attributed to the treatment. Despite the existence of many recent works on the RDD subject, the existing software that implements the state of the art RDD procedures is unfit to deal with very large datasets, which are increasingly more common. The purpose of this work is to present a Parallel Computing Approach refactoring of a well-known RDD algorithm and apply it to a dataset from Bolsa Familia with more than 13 million observations. The objective of the RDD study is to evaluate the impact of the program on the labor choices of needy 18-year-old Brazilian boys and girls. We verified that the parallel approach was able to process Bolsa Familia data in a reasonable time and that it outperformed other existing RDD implementations.