See post #124.
The breakdown of voting districts into finer gradations (yes precinct is finer than county in that sense) leads to more coarse data in smaller groups. The 'proportional assignment' is more coarse (= less smooth, less averaged, more raw) at the precinct level than at the county level.
Semantics aside (and the above are the proper statistical semantics) the issue comes to accuracy and what constitutes 'illegal' or improper ballot.
Thanks for the clarification. I had understood you to mean by, "more 'coarse' data," to be effectively, "coarser data," until the clarification. I see what you mean.
Yes, raw precinct data can yield far more evidence of fraud than does county data. For example, during the famed Florida recount in precinct 144F the Democrats had a 267% turnout while the Republicans had an 11% turnout.