Sen. Daniel Patrick Moynihan once puckishly said that data indicated that the leading determinant of the quality of public schools, measured by standardized tests, was the schools' proximity to Canada. He meant that the geographic correlation was stronger than the correlation between high test scores and high per-pupil expenditures.
Moynihan was a smart guy, so I suspect that's not what he meant with the Canada crack, given how education spending actually relates to geography in this country. Click through for a moment to Matt Miller's recent Atlantic article, for example, and scroll down to the map showing per-pupil expenditures broken down by county. The light-colored (low-spending) counties are clustered in the South, while the dark-colored (high-spending) counties are disproportionately in the Northeast, upper Midwest, and northern parts of the country in general, i.e. the places closer to Canada.
There are a whole bunch of different causal factors working in the same direction, but my view is that Carey is mostly wrong here. It's true that, all things held equal, an increase in education spending will lead to an increase in test scores, as Robin Chait notes. But Moynihan's point was that this effect is basically swamped by the reality that states close to Canada tend to have unusually excellent schools, even when you control for things like income, race, and per-pupil spending. This is not a recent development. Check out this 1992 New York Times article, which looked into the question of why states like North Dakota, Montana, Iowa, Nebraska, Minnesota, and Wisconsin had such great scores on standardized tests. The basic answer is that it's hard to explain, except these areas tend to have low dropout rates, high levels of parental involvement, a culture of less television watching, and so on. North Dakota, the highest-scoring state, ranked 49th in teacher salaries.
What do these states have in common? Well, it's really cold for much of the school year, and there's not much else to do. More importantly, though, these states are all part of David Hackett Fischer's "Greater New England" region, the homogeneous, white, Protestant northern tier of the country settled by New England Yankees and northern European migrants, which I've referenced before. This region is sort of the goody two-shoes of America in a variety of quantitative social-science measures: Great test scores, very low crime rates, a historical aversion to violence (nearly all the states with no death penalty are Greater New England states), a tradition of clean, nonpartisan reformist politics (which Michael Lind rightly suggests helps explain Barack Obama's success there). This effect manifests itself in bizarre other ways, too, like the oft-cited statistic that Minnesota, for instance, gets fantastic health outcomes despite spending far less per patient than other regions of the country. I'm inclined to think that the correlation Moynihan was identifying owes primarily to this aspect of regional culture. This doesn't, as Will apparently believes, mean that there's nothing we can do, or that increasing per-pupil expenditures won't have any effect. But it does suggest that some places start out with a huge, built-in cultural advantage.