In estimating these regressions, use all the observations with class enrollment less than 40. You will need to first create a variable (call it "above40") as an indicator for the discontinuity (a class size above 40 would be denoted as 1, and a class size 40 and below would be denoted as 0). You also need to create a variable of normalized class enrollment (call it "x") that measures deviations of class enrollment from the 40-threshold. Look at Table 2 at the end of this document for more guidance. Also, make sure to restrict your binscatter and regression analysis to when school_enrollment is between 20 and 60