« [SSJ: 273] RE Critique of P-A Approaches | Main | [SSJ: 275] RE A Banking Kabuki? »
September 12, 1995
[SSJ: 274] P-A Theory
From: Ellis S Krauss
Posted Date: 1995/09/12
In light of both Michael Thies' and John Campbell's quite stimulating comments
on principal-agent theory applied to Japan and the McCubbins-Noble [hereafter
MN] article, I thought I would elaborate on my rather stark previous comments
that P-A theory in this article is tautological, untestable, and nonfalsifiable,
as well as my other criticisms of it.
First, before critiquing the article, let me establish my lack of inherent bias
to deductive theory per se or to the substantive goals of the article or the
authors themselves, lest responses be tempted to question the motivations for my
criticism and divert attention from the substance of my critique.
As most know, my work with Michio Muramatsu going back over 12 years [and first
expressed in publication in an APSR article in 1984], was one of the first to
challenge the "bureaucracy dominant" model of Japanese politics. Indeed, if one
goes back to our "patterned pluralism" piece in the Yasuba and Yamamura
POLITICAL ECONOMY OF JAPAN, vol. 1, [esp pp. 533ff] and a summary of the model
in the Okimoto and Rohlen volume , INSIDE THE JAPANESE SYSTEM, one will find as
I recall both the argument that LDP politicians play a significant role in
policymaking and that party leaders are especially important in settling 'turf
conflicts' when issues transcend bureaucratic jurisdictions. So I am sympathetic
to the general argument of the article
Second, I have no problem with deductive theory, used properly. Although I
prefer to work a bit more inductively, I recognize that deductive theory such as
P-A theory is necessary to the [social] scientific process especially in the
early stages of enquiry--generating hypotheses for testing--and in the latter
stages--making sense out of and discovering patterns in the empirical findings.
Third, I think much of Greg Noble's prior, non-rat cho work is among the best in
the field of Japanese politics. I don't know McCubbins' prior work, but
colleagues in American politics whom I respect think highly of him.
Thus my objections to the McCubbins-Noble article are not at all based on
aversion to either to deductive theory in general nor to the attempt to prove
that LDP leaders have a role in policymaking in Japan. Nor did I read this piece
with any prior negative conception about the authors. Rather it is based on my
intellectual reactions to the substance and "methodology" of the piece. I offer
the following critique not with malice but in a genuine desire to try to figure
out what the fuss is all about and why I seem to see real problems with the
analysis but apparently others do not:
1)The article attempts to prove, not that party leaders do in fact have
influence and show us how, but rather really only that the "null hypothesis" of
"abdication theory" is wrong; namely, that bureaucratic agents cannot make
policy that legislators, if fully informed, would not prefer to existing policy.
This is a total straw man. As John Campbell has pointed out, no one has argued
that bureaucratic influence in Japan is based on their ability to make policies
that are anathema to politicians. Bureaucratic influence in Japan is based
rather on their ability to a)to influence in the formulation process, directly
through persuasion and information provided to politicians, and indirectly
through ability to influence the recommendations of shingikai that have
influence on politicians and public opinon, the preferences, pespectives, and
goals and ideas of party leaders and politicians, and b)the ability to influence
in the implementation process how policies adopted by parliament actually gets
enforced. Most especially, that in Japan the LDP has written particularly
general guidelines in the laws, allowing bureaucrats within that framework to
make important decisions ["administrative guidance" and all that] as to how
those policies are implemented, and thus in effect to determine somewhat what
impact policies actually have. The only thing this article "proves" is that
bureaucrats aren't likely to openly defy the law and Constitution and go
completely against the policies of the LDP and parliament. No one ever said they
did.
2)The article makes some extremely debateable and tenuous ad hoc assumptions
about the ability of politicians to keep "accurate tabs" on bureaucratic
behavior, and to gather and understand technical information about policies
independently of bureaucracy, for example that constituents and the private
sector will keep Diet members informed as to the content and impact of policies
bureaucracies are pursuing, and shingikai also can perform this role of
monitoring bureaucracy. Yet in the areas where bureaucratic influence is
generally acknowledged to be greatest, namely the economic bureaucracies of MOF
and MITI, there are real questions as to whether politicians and their
constituents care that much about these policy arenas since they have little
direct impact on constituency demands and needs. Indeed, I thought this was one
of the important implications of Rosenbluth and Ramsayer's analysis of the
electoral system in Japan--that it oriented politicians to the "low policies" of
roads, bridges, dams, and rice, and not to the "high policies" of macro and
micro economic policy. Thus if even constituents and party politicians don't pay
much attention to those policies, the ad hoc and very tenuous assumptions on
information feedback upon which
much of the argument is based upon would not necessarily hold.
There is also empirical evidence that the attitudes and goals of MOF and MITI
bureaucrats are different from those in the more politicized ministries. Putnam,
Aberbach and Rockman's work in Europe indicates that BOTH politicians and
bureaucrats play many different roles in policymaking and in an article in
BRITISH JOURNAL OF POLITICS, Aberbach, Rockman, Muramatsu and I presented data
on surveys of bureaucrats that showed that the economic bureaucrats indeed had
greater levels of commitment to, for example, managing social change and
coordinating interests [normally roles for politicians] than the bureaucrats in
other agencies.
Finally, although the article treats shingikai as monitoring agents of
politicians, in fact it completely ignores the fact that they are set up,
affiliated with, and report to, not the politicians, but to the ministries.
3)The article is tautological and actual rests almost completely on invoking
formal authoritative power to prove its point, is unfalsifiable and doesn't seem
designed to be capable of testing by empirical data either.
As I understand the argument, it goes something like this:
1)Majority party leaders and parliaments have formal authority to make policy
2)They delegate bureaucrats as their agents to formulate and implement those
policies.
3)Party leaders would certainly know if bureaucrats deviated from what they
wanted [see critique above] and would not allow it; bureaucrats know this and
therefore anticipate what ruling parties want and give it to them.
4)QED: party leaders actually have the power and evidence of bureaucratic
influence on policy is "illusory" or only an "appearance." Evidence to the
contrary are "the exception that proves the rule."
Let me translate the above in perhaps cruder terms: the assumptions are that
formal and legitimate power makes party leaders "principals" --this
predetermines the conclusion that they are powerful; agents can't influence
principals; any empirical evidence to the contrary is merely "appearance";
anecdote and allegory are sufficient for "evidence;" and our argument could be
test if we chose to, but this is really unnecessary because the case is proven.
4)I find this form of P-A theory disturbing because it really seems to take us
back to the pre-1960s when "formal-legal" theorists thought you could understand
politics by looking at a Constitution and the formal legal powers given by it.
This P-A approach is dressed up as pseudo-scientific deductive theory but in
fact, really rests upon the pre-behavioral approach assumption that formal
authority determines actual power. While some of the most interesting research
on Japan in recent years [Upham, Pharr, Haley, Samuels, etc] has shown that what
makes Japan quite distinctive is exactly the disparity between formal, legal
powers and the way process actually works, the P-A approach assumes away any
such possibilities.
5)By arguing against the influence of zoku as well as that of the bureaucracy,
the article undermines one of the best empirical arguments for why politicians
might be more powerful than bureaucrats, as zoku would have the most expertise
and also the most information and conduits to the private sector regulated by
bureaucrats.
6)When all is said and done, what really do we learn from the conclusions? That
party leaders may actually have more power than we've given them credit for. Ok.
I'll possibly buy that. But where's the empirical evidence backing up the
deductive theory? Where's the test[s]? Anything else? One could come up with the
same hypothesis [and really that's all it is as the article doesn't really
"prove" anything], as John Campbell pointed out, from reading a lot of the prior
literature on policymaking in the field, and without the P-A trappings.
Sorry. I'm totally underwhelmed. I'd really appreciate it if someone would
defend this piece and explain to me why it is not tautological and
unfalsifiable, and why this approach moves the field forward or is important.
Ellis Krauss
Approved by ssjmod at 12:00 AM