Prof. Hiwatari raises several excellent points.

1) The P-A approach draws on the industrial organization literature (see, for example, the edited volume by Putterman, or the primer by Jean Jacques Tirole) for criteria by which to judge the impact of rules, contracts, capabilities, and the like. I am not familiar with Prof. Ito's study, but he is certainly correct in stating that bureaucrats possess superior expertise and information compared to politicians. This is especially true in parliamentary countries where politicians do not have to worry about competition from the executive branch and therefore don't bother to keep large legislative staffs. On the other hand, politicians are informed by the clients when the clients' interests are negatively affected. In fact, politicians may even structure the administrative process (establishing shingikai, for example) to give their favored clients "fire alarms" to pull when they want politicians to give an issue some attention. The very existence of fire alarms should usually be enough to keep bureaucrats within the bounds of these clients' interests.

2) I completely agree on the importance of comparative studies to avoid nontestable country-specific explanations.

3) The last couple of years of coalition government have been interesting, but it is hard to use them as decisive tests for a few reasons. a) The Ozawa/Hata/Hosokawa group was stymied by the Socialists; b) now the LDP is virtually back in the driver's seat; c) most importantly, there has been no election yet under the new rules. If electoral institutions matter, the people who are elected under different rules should appeal to voters on a different basis. In the Japanese context, there should be less emphasis on catering to organized groups of supporters. But transitions take time. Britain's electoral rules changed in the mid-19th century (1832-46 or so); the corn laws were finally repealed a decade later.

Nonetheless, the years of coalition government give us some clues. The conventional wisdom is that bureaucrats have been in total control, but one could also see them as years of policy inertia or policy drift. Moreover, there have been some interesting episodes of politicians sanctioning bureaucrats. Kumagai forced the MITI AVM [Admin. Vice Minister] to resign, Tanaka Makiko fired her deputy at the STA [Science and Tech. Agency], and the LDP demoted MOF's Taya
--probably for being too cooperative with the Ozawa group. The P-A approach looks at these kinds of events as "out-of-equlibrium"--in other words, if the situation were more stable and bureaucrats could be more certain about the consequences of their actions, they would have avoided getting in trouble in this way.

In the final analysis, I think more empirical work is called for! I'm interested in looking at, for example, the extent of factional ties with particular groups of bureaucrats, and what is happening to those ties now that the LDP is no longer in one piece.