may not be the easiest way: try

egen x=count(1), by( county)
gsort county - income2
by county : gen index=_n<round(x/2,1)

the index variable will be 1 for the observations with the highest income2 value within a county cut by 50% of n counts within a county.  There can be ties, so you may need to change the code depending on how to handle the ties.

Thank you very much for your answer！！！ It does not give me the direct answer, but it does help a great deal in getting what I want!! Thanks again!