I just read this piece from Cato attempting to discredit my recent policy report:
Among other things Andrew Coulson asserts that large degrees of selection bias taint my study of Private School Expenditure and that not only did I ignore this, but I hid it intentionally and skewed the results. Okay, I urge anyone to read the study and check the extent of methodological detail.
Regarding Coulson’s specific claims of selection bias:
Apparently, Coulson failed to read pages 24 and 25 of the report or view Figures 9 or 10 of the report where I detail precisely the extent to which the IRS 990 sample of 1,500 schools is representative of various private school affiliations as reported in the NCES Private School Universe Survey. Here, I show that the IRS filings represent 74.2% of private independent school enrollments for the states and labor markets studied and nearly a third of Christian Association Schools. I acknowledge the possibility of sampling bias, and make additional comparisons in Figure 21 (page 39) of the report, showing that the reported pupil to teacher ratios in the full NCES Private School Universe Survey align well with the pupil to teacher ratios in the schools which reported finances on IRS 990. Class sizes are a primary driver of per pupil spending.
On a more trivial note, elimination of schools with budgets of less than $500k has little or no effect on labor market, state or national averages which are weighted by the numbers of children served. Further, Coulson’s assertion that these schools are systematically lower spending per pupil is incorrect. They are all over the map.
To clarify the parenthetical regarding DC spending on Page 42. Current Expenditures per pupil in that year (2006-07) were about $14,300 in DC, and total about $20,200. It would be relevant to compare the total figure here. Nonetheless, the DC private independent schools in particular still outspent DC Public.
Finally, Coulson appears to totally miss the point of the study, which is not that private schools are somehow invariably high spenders. Rather, the point is that spending among private schools varies. It varies a lot, and it varies largely by religious affiliation with Christian Association Schools in particular spending much less than public schools. As such, when low voucher levels are set in existing voucher programs like that in DC, it should come as no surprise that most vouchers are used at religious schools. They (typical vouchers) simply aren’t even close to sufficient for private independent schools. Perhaps if the DC voucher program did allocate $28k per child (Coulson’s version of DC public schools spending per pupil), this would make a difference.
Also, regarding the sponsorship of the study – The study received a nominal honorarium of $4,000 from the Great Lakes Center and EPIC. Note, however, that from a public school teachers’ union perspective, there’s not much in there to cheer about. In fact, the study finds that even the higher spending privates do not necessarily pay their teachers more than public schools in the same labor market. Rather, the additional expense is in smaller class sizes which are largely a function of more diverse curricular offerings. Indeed the much lower spending private schools do have both larger classes and lower paid teachers.
UPDATE: Coulson updated his post to include the statement: “The religious private schools that do file Form 990 are thus a small self-selected group that is presumably seeking to maximize its revenue from charitable donations, and hence very likely biased toward higher spending schools.”
So, the implication is that my sample of private religious schools is biased toward the higher spending ones. Note that I have a near 30% sample of Christian Association Schools, and that my main conclusion about these schools is that they are very low spending and coupled with that have very weak teachers, relatively larger class sizes (than independents) and very low teacher salaries. Is he sure that he wants to argue that this group is likely a biased, self-selected group of high spenders among CAS schools. Again, I compare my 990 sample to all CAS schools in the NCES PS universe survey in Figure 21. I do point out that my very small Catholic sample does appear biased in this very way (those few that did report were somewhat different than “average” Catholic schools in Figure 21).