Mike Reno, a Michigan blogger and elected school board official, critiques my recent study of school district size and spending here. My 411 on his analysis:

His chief objection to the study is that I use per capita income as a proxy for a community’s level of demand for education. Mr. Reno “cannot agree that all wealthier communities have higher standards than less affluent communities.”

This is a non sequitur. The fact that income has been found to be strongly correlated with educational demand does not mean that all wealthier communities will have higher educational demand than all low income communities. It simply means that, on the whole, the correlation between the two is strong and positive. There are surely exceptions.

I use income per capita as my main proxy for a community’s demand for educational services because that has been the norm in econometric research on school district spending since such studies got going in the late 1950s. It is not, however, the only such proxy that I used. I also control for the share of the district’s population actually enrolled in public schools (which captures the extent to which community members and their relatives are personally affected by the system — and hence stand to gain from higher spending on it). In addition, I tested out a control variable which was an index of parental level of education, also commonly used as a proxy for educational demand (the more education you have, the higher your expectations for the education of your children).

As I explain in the paper, the coefficient for the public school enrollment per capita term is unexpectedly negative, and the parental level of education term was statistically insignificant and so dropped from the final model. Hence, after controlling for the other dozen or so variables in the model, no proxy for educational demand explains a substantial share of the variation in district spending, and one actually has a negative relationship to spending.

A second of Mr. Reno’s concerns is my observation that student achievement is not, on average, related to per-pupil spending, after controlling for other factors. He observes that spending can matter when the stars align and a public school district has efficient leadership. This is not an objection to the overall pattern. It simply illustrates that, on the whole, public school districts do not have efficient leadership. The lack of substantial correlation between public school spending and student achievement is a very well established research finding. In most cases, public schools are simply not efficient. As I explain in my paper, the evidence suggests a reason for that pattern: the current system provides incentives for public officials to spend as much as they can, and, on the whole, they apparently heed those incentives.

Mr. Reno’s final concern is that I do not deal with any systematic variation in the service mix offered by districts of varying size. Presumably, he is suggesting that larger districts are able to offer a wider range of services, and that this explains their somewhat higher spending. A key conclusion of my paper, however, is that size does not matter all that much in determining district spending. Realistically, breaking up large districts would reduce total spending by about 1 percent (that would be for a hypothetical savings of $200 million from district breakups, out of a budget of about $20 billion).

As I point out in the paper, what really determines district spending is the ease with which officials can raise per pupil revenues. The easier it is, the more they spend. Is it any wonder spending is out of control?