Non-parametric estimation in regression discontinuity design

Regression discontinuity is a quasi-experimental impact evaluation method of assessment which is used in cases where control and effect groups can not be generated before the implementation of the program. In this method, the control group and effect group are determined according to a criterion called the threshold. Individuals, which close to threshold value are assumed to have the closest characteristic to each other, except for the application of the program. The proportion of individuals assigned to an effect or control groups determines the design of the method. In regression discontinuity design, the assignment is completely based on individuals position according to threshold. Graphical representation is used to visualize discontinuity in the first step. As the effect of the program is evaluated, the bandwidth is calculated to inclusion of how many individuals are close to the threshold to analyze. Within this bandwidth, the regression models of the individuals above and below the threshold are estimated in the threshold value, whether there is a break or discontinuity. Discontinuity is considered as the effect of the applied program.