1. create two datasets from orignial dataset - dataset with policy =1 and dataset with policy = 0
2. use joinby to link all of possible matched observations
3. Check for the closest 4 observations based on FS
1. When you create the two dataset, rename the FS and other variables to different name, for example FS to FS1 if policy = 1, and do not need to change the matching variable name.
2. After the joinby, generate the difference between FS and FS1, you may take the absolute value for the difference
3. bysort EAR SIC2 DP (difference): keep if _n == 4