(NOTE: UPDATE 9:55pm EDT - Additional graphs added below.)Some things are so obviously true that they needn't ever be discussed. For example, the more you pay teachers, the more our students will learn. So why bother even looking up the data to prove something so self-evident? Well, what the hell.
The American Federation of Teachers publishes an
annual survey of average teacher salaries by state as a service to their members. The last survey was from 2006/07. Now of course living standards differ across the country, but if we normalize these salaries to median household income by state, that should give us a good measure of teacher pay relative to the state as a whole. Household median income from the 2008 American Community Survey is available
here.
The NAEP reading scores for 8th graders should give us a good measure for how well students in each state have learned to read, a basic job of teachers. To avoid the confounding factor of race, I've restricted the scores to white students only. NAEP data is available
here.
So what do we find? (Click for larger image.)

There is actually a negative relationship between relative salary and reading score - albeit a weak one with an R2 of 0.1. But there's no support for the assumption that higher teacher pay leads to better educational outcome.
Here's the graph with labels for each state so you can find how your state fares.

It would appear that in some states teachers are real bargains - New Jersey, Maryland, Massachusetts and Connecticut stand out. California not so much, and Arkansans are getting a real raw deal. West Virginia as usual stands dismally alone. For my home state of New Jersey, where Governor Christie is locked in a savage battle with the NJEA, this data could be used to argue that the state could clearly afford higher teacher salaries, or that cutting back on compensation is unlikely to have a particularly adverse impact on education. California teachers, on the other hand, in their battle with budget constraints haven't a leg to stand on.
UPDATE: At Steve Sailer's request, I've updated the graphs to reflect teacher salaries relative to white incomes, as most teachers would probably measure themselves against this standard rather than the state as a whole. Using White-Non-Hispanic median family income, there is some substantial shifting around of some of the states, though the overall story is the same - the relationship is still negatively correlated (R=-0.32, R2=.11. For example, as Steve expected, California teachers are no longer the 3rd highest paid but are now tied for 20th (because of the dichotomy in white vs. Hispanic incomes). A number of heartland states are really overpaying their teachers it would appear, while some of the Northeast states are getting their teachers at bargain prices. And note the gap between Rhode Island and neighboring Mass. When it comes to viewing this as indicating what a state could afford to pay their teachers, the total income (vs. white only) is probably a better measure to use.


Data behind the graphs is
here.