Chillin' like a statistician...

For the last couple days, I've been running some elementary statistical
tests (ANOVA analysis, two-sample t-tests) for this Paper that Krister
wants to write about a "race to the bottom" in decentralized forestry
regimes.

The race to the bottom argument says that, essentially, when capital
becomes more mobile (it's easier to move capital in whatever form
around) or when political authority is decentralized to sub-national
governments, local governments compete for investment and income by
lowering their regulations and taxation. The end result is that nobody
attracts any more investment, because everybody's taxes and regulations
go down, and everybody ends up worse off because regulations (like labor
and environmental regulations) decline, as do government services
(things like road maintainence law enforcement, and education) because
tax revenue declines along with declining tax assessment rates.

Aside from insane Laffer-curve nonsense, this makes a certain amount of
sense.

So, basically, what I'm doing is comparing forestry regulations, changes
in regulations, and local officials' knowledge of neighbors' forestry
regs through the use of several subjective survey questions.

In the end, I think the evidence pretty strongly supports the argument
that we want to make--there is no race to the bottom. The paper may
also present some weak evidence for a race-to-the-top dynamic (instead
of cutting regulations and taxes, municipalities compete for investment
by operating more efficiently and providing better services), but that's
a stretch for the data we're using.

Essentially, what we're doing is comparing the means of these questions
across the three countries--if a Race to the Bottom is taking place, you
would expect that regulations would be stricter in the least
decentralized country and less strict in the less decentralized
countries, municipal officials would have more knowledge of their
neighbors' forestry regulations in the most decentralized regimes, and
the stringency of regulations would have declined the most in the most
decentralized regimes. None of these things is the case, in several
different iterations of these tests.

In general, I think the paper is pretty strong. My concern is that the
relatively unsophisticated statistical techniques we're using won't be
taken seriously. What I wanted to do was use GIS software to generate a
spatial lag with which to test the idea. But Krister wants to get it
out quickly, and it's true that that sort of stuff will take a billion
years.

Maybe for the dissertation. It's a good excuse to spend months looking
at satellite pictures on Google Earth.