The Krauss-Thies dialogue is very useful in clarifying arguments, in part by
showing that much was semantic. With regard to the alleged shift in the early
1970s, now most everyone seems to agree that politicians became more influential
at least in the sense of articulating goals and getting them achieved, but that
they really did not increase their authority. Some of the dispute centered on
what people mean by "power." We should hear from TJ Pempel on this point: his
first published article was about the concept of power, written (at cost of
great envy among his colleagues) when still a graduate student at Columbia.
Thies' explanation about "explaining" budget outcomes is also very well put and
right on target. I would add, though, that some such analysis are content to
find a statistically significant relationship between, e.g., partisan control of
Congress and defense vs. welfare budget growth. I would take the methodological
purist position and say that statistical significance just means it probably
didn't happen by chance and entitles you to talk about a relationship, but you
have to go on from there and tell us something about how strong it is. If the
Dems coming in means that defense goes down by $1 million, it may be
statistically signficant but who cares? (I can't help adding that if you are
looking at 200 budget items, and your criterion for significance is .05, ten of
them are going to look significant purely by chance).
I bring up all this boring stuff because I haven't yet seen an analysis of
political influences on the Japanese budget--urban-rural or factional
influence--that gets by these tests in a convincing way.