I sorted the states by percent proficient in 2005 and then looked at two things for each state: their rate of unionization (based on data from http://teachersunionexposed.com) and how much they spend per student. If unionization is bad for kids (as the argument goes, since it places teachers' needs above kids' needs) then we'd expect to see performance to drop as the percent of unionization increases. Similarly, if the amount of money spent did not matter for education, then we'd expect to see no correlation between the performance and dollars spent per student. As it turns out, performance not only doesn't decline but it even increases with an increase in unionization. Similarly, an increase in kids' performance accompanies an increase in per-student dollars spent (actually, this should not be too surprising given that its likely the rate of unionization will generally correlate with dollars spent).
Because my goal is not to be an ideologue, I will readily admit that correlation does not equal causation. In fact if I were to attack this data, I'd use that as one point, the other being the representative nature of NAEP scores. Here is a good discussion on why NAEP state score comparisons should be done with caution. Regardless, correlation is a useful indicator in and of itself, and if the graphs were not consistently indicative of these correlations I'd be much more likely to question whether NAEP scores mis-represent the population by virtue of their representative nature.
A couple clarifications. Because it is difficult to 'see' a slope when the rate of unionization data is graphed directly on a state by state basis (variation between states that are 100% unionized vs ones that are 0% unionized dont provide much value) I did a running average of 10 states to get a more generalized trend to the data. In other words, the first rate of unionization value is an average of the lowest 10 state's rates (1-10), the next is an average of the next 10 (2-11) etc.
NB: the reason there is a dip near the right hand side of the math scores is that there are two states with low single-digit unionization (Texas and South Carolina) that score at about the 75th percentile among a number of states that are almost 100% unionized.
Please feel free to provide an alternative interpretation.
Graphs:
