Class Size, Dropouts, & the Windy Atlantic

In Monday’s Wall Street Journal, I argued that America has too many public school employees, and has wasted those employees' talents on a mass scale. Jordan Weissmann, an associate editor with The Atlantic, disagrees, accusing me of running “roughshod over a lot of important nuance.” As it happens, no nuances were injured in the composition of my piece.

Let’s consider Mr. Weissmann’s cruelty-to-nuance claims in turn. First, he feels that I ignored “significant evidence that smaller classrooms do indeed improve student performance,” citing two sources. The first source is an unsigned web-page by the “Center for Public Education” that is so biased in its selective coverage as to not be worth serious consideration. The second is a scholarly paper by Alan Krueger, author of one of the two best-known literature reviews of the subject.

What Weissmann doesn’t mention is the work of Eric Hanushek, author of the other best known literature review on class size. Krueger contends that class size reduction is usually educationally beneficial and cost effective, Hanushek argues the contrary on both points. It’s easy to compare their evidence and arguments because both contributed at length to the book: The Class Size Debate, published by the left-of-center Economic Policy Institute. It is a testament to how comfortable Hanushek is with the strength of his case vis-à-vis Krueger’s that he links to a full .pdf of that book from his own web pages at Stanford University. I understand why. When Hanushek looked at the most methodologically sound estimates—those that measure changes in student performance over time instead of at just a single point in time—he found that 89% show either no statistically significant advantage or a significant negative effect to smaller classes. To arrive at his opposing conclusion, Krueger had to, among other things, overweight the lower quality studies.

Hanushek’s conclusion is also more consistent than Krueger’s with the national U.S. data. The average American classroom has gotten substantially smaller over the past 40 years (by about 7 students) but achievement at the end of high-school is essentially flat. The only way to counter this evidence is to claim—usually without systematic basis—that children must be so much more difficult to teach today that the gains we would have seen from smaller classes have been eclipsed by this reduced “teachability.” The only systematic study of “teachability” trends of which I am aware does not support that claim—finding net “teachability” to have been mostly flat over time, with some improvement in the past decade.

Hanushek’s conclusion has also been supported by new, large-scale research, published after his and Krueger’s reviews. Harvard researchers Antonio Wendland and Matthew Chingos reported in 2010 that Florida’s state-wide class size reduction had “no discernible impact upon student achievement,” but has so far cost the state roughly $28 billion.

Some journalists are aware of the evidence that smaller classes generally do not improve outcomes. Consider, for example, this bit of reporting from last December:

Think of the ingredients that make for a good school. Small classes. Well-educated teachers. Plenty of funding. Combine, mix well, then bake. Turns out, your recipe would be horribly wrong, at least according to a new working paper out of Harvard…. The study comes courtesy of economist Roland Fryer, an academic heavyweight who was handed a MacArthur Foundation "genius award" earlier this year…. Fryer found that class size, per-pupil spending, and the number of teachers with certifications or advanced degrees had nothing to do with student test scores in language and math.

In fact, schools that poured in more resources actually got worse results.

Who is the astute journalist who wrote these words and from whom Jordan Weissmann could learn a few lessons? You guessed it… it was Jordan Weissmann, writing just seven months ago. How soon we forget.

Next, Weissmann claims that “dropout rates, for instance, have fallen by almost half since the 1970s.” Presumably he is unaware that this statement and the table he cites in support of it do not reflect reality. The “dropout rates” published in that NCES table are statistical fabrications of the nation’s education bureaucrats, looking to placate the public with the help of the remarkably credulous education media. You needn’t take my word for it. That is the finding of the left-leaning, Nobel-prize-winning, cited-by-President-Obama-with-approbation economist James Heckman. Heckman’s 2007 study, with Paul LaFontaine, is still the definitive work on the subject (though it was not the first to report the truth). Here is what Heckman and LaFontaine established through a painstaking analysis of the nation’s graduation data:

(a) the true high school graduation rate is substantially lower than the official rate issued by the National Center for Educational Statistics [the one cited by Weissmann]; (b) it has been declining over the past 40 years; (c) majority/minority graduation rate differentials are substantial and have not converged over the past 35 years; (d) the decline in high school graduation rates occurs among native populations and is not solely a consequence of increasing proportions of immigrants and minorities in American society;

They also note that the post-NCLB uptick in graduation rates probably does not imply a genuine improvement in educational outcomes:

NCLB gives schools strong incentives to raise graduation rates by any means possible. When monitoring was implemented in 2002, minority retention [a.k.a. “flunking”] dropped sharply and graduation rates turned upward, especially for minority groups. A similar pattern is observed following the publication of A Nation at Risk. Whether these represent real gains or are an indication of schools cheating the system in the face of political pressure remains an open question for future research, although the timing suggests strategic behavior [i.e., “cheating”].

The italics and the text in square brackets in the above quotations are mine.

The fact that public school systems report falsely rosy “dropout rates” is not a secret. Anyone who spends 60 seconds on Google will discover it. It’s even been reported in such popular media outlets as… Mr. Weissmann’s employer, The Atlantic. That page on the Atlantic’s website actually links to the very same Heckman and LaFontaine study I link to above. Heck, it’s even been mentioned in The New York Times (though they’ve managed to protect their most die-hard readers from cognitive dissonance by restricting coverage of these findings to David Brooks’ column).

Weissmann wraps up his blog post with a foray into the art of mind-reading:

I doubt Coulson truly believes we really have too many teachers in this country. He hints at so much in his last paragraph, writing, "While America may have too many teachers, the greater problem is that our state schools have squandered their talents on a mass scale.” Why the hedge? My guess is….

Kudos to any readers who correctly predicted that both Mr. Weissmann’s belief and his guess were wrong. The reason that I can make no certain statement about the ideal size of the U.S. education labor force is that no one can predict the allocation of human and capital resources that will occur in future in a free market. That said, there is reason to expect fewer teachers will be required under market conditions since our public school monopolies have been on a hiring spree 11 times faster than enrollment growth for forty years. Moreover, on-line learning and educational software options are only getting better and more numerous, and this should lessen demand for classroom teachers. Against those forces we have to consider that families may choose to invest some of the resulting savings from employing fewer classroom teachers in one-on-one tutoring, which is generally accepted as highly effective if, at present, too expensive to be much used.

One thing I can say with some certainty, based on the world-wide research literature comparing different sorts of school systems within countries, is that whatever particular allocation of teachers and capital resources the market arrives at will be more efficient than the gross, unproductive staffing bloat that has been perpetrated by state schooling. And, as I explained in that linked study, the existing small niche of non-profit private schools in the United States does not constitute a free education marketplace. A further explanation of the difference, should anyone find it necessary, can be found in this piece by economist John Merrifield.

To quickly correct some of Weissmann’s remaining errors: I am on record as not faulting teachers’ unions as the cause of our nation’s education woes. Their predations (e.g., contributing to the system’s demonstrably unproductive employment bloat) are a symptom, not the disease. And while some public school teachers are obviously overpaid, others have been equally obviously underpaid. The problem with public school teachers’ salaries at present is that they are allocated based on time-served and credentials (neither of which is consistently related to student achievement) rather than performance. Markets tend to compensate employees based on performance and so this problem, too, will likely be solved by liberalizing America’s education sector through programs like K-12 education tax credits.

This is probably all the time I will have to debunk the various flawed criticisms that were offered in response to my WSJ piece, so I thank Mr. Weissmann for conveniently collecting most of them in one place.