Principal-agent analysis 1) can show that politician-dominance is consistent with well-known facts (although there is nothing deductively necessary about this conclusion; one might conclude, as Junko Kato does, that there is substantial agency slack; see also Kathleen Bawn, Spring APSR, on some conditions under which politicians might be inclined to give bureaucrats more or less autonomy), and 2) can suggest new places to look--specifically, the nature of the rules that govern the relationship between politicians and bureaucrats (including availability of sanctions), and the capability for and costs of monitoring where the contracts are incomplete.

Because politician-dominance is often observationally equivalent to bureaucratic dominance, it is difficult for one to disprove the other. The best test is yet to come: what will happen when different parties come to power? Will the bureaucrats go along with new policies, or will they successfully stonewall?

Best regards,
Frances Rosenbluth