AMERICAN.COM

After being resolutely glum about the limits of educational improvement here, here, and here, it’s time for me to get in touch with my inner optimist. We can’t make our kids much smarter than they are naturally, but we can do a hugely better job of teaching them stuff. If you get away from the worst schools in the big cities, I think the central problem with the public schools is not poor teachers, but the curriculum teachers are given to teach, especially in elementary and middle school. There’s a great solution, packaged and ready to go, that doesn’t cost any more money than we’re already spending. It’s E.D. Hirsch’s Core Knowledge Curriculum. The Manhattan Institute’s Sol Stern has a wonderful description of it, and of the heroic work of E.D. Hirsch Jr. over the past quarter century. I commend it to all who might have influence with their local Board of Education. Install the Core Knowledge Curriculum in your local school, and some very large proportion of your complaints about public education will go away.

Educational romantics are the Jean-Jacques Rousseaus of the education debate, possessing a naïve (in my view) faith in the plasticity of human cognitive abilities and an equally naïve credulousness when they read the extravagant claims that are made for this or that demonstration program. When I made these points in my book Real Education, I was intrigued to find that I attracted more criticism from the Right than the Left, with Jay Greene and Ben Wildavsky being prominent examples.

The latest entry is Marcus Winters over on the National Review website, who is upset that my “dangerous idea” that too many people are going to college has been gaining momentum, when, in Winters’s view, we should look upon the number of kids who should go to college as open-ended.

What Winters and the other educational romantics won’t do is confront the huge literature about the kind of linguistic and logical-mathematical abilities that are intrinsic to success in a genuine college curriculum. Those abilities are essentially the cognitive skills measured by IQ tests. As it happens, I just published an article entitled “Intelligence and College,” in the maiden issue of National Affairs,  that takes up the topic systematically. Here are some empirical points drawn from that article, and documented in more detail in Real Education that I don’t think qualify as “controversial” among people who are familiar with the literature on academic ability—IQ if you prefer—and success in college.

1. By adolescence, what you see is what you get in academic ability. There is still a lively empirical controversy about how much IQ can be changed by outside interventions in preschoolers, but not in high-schoolers. Among the best programs, you’re looking at improvements in the region of 0.2 standard deviations on an exit test, and those fade to triviality when retested two or three years later.

2. A common operational definition of “college readiness” in the literature is a 65 percent probability that a youngster will get a 2.7 grade point average in his freshman year—not a demanding standard in an age of grade inflation and soft courses. In a study based on 165,781 students at 41 major colleges, the combined SAT score that predicts a 65 percent chance of a 2.7 freshman GPA is 1180. It is a score that only about 9–12 percent of American 18-year-olds could get if all of them took the SAT.

3. Both the 1979 and 1997 cohorts of the National Longitudinal Survey of Youth (NLSY), consistent with a half-century of collateral data, show that the mean IQ of whites who get a BA is 114–115, a range that demarcates the top 16–17 percent of the distribution. (The story about IQ and college experience among blacks raises a host of ancillary issues that I won’t try to deal with here.)

4. Both the 1979 and 1997 NLSY cohorts indicate that the 50-50 break point for successfully completing a BA among those who are self-selected to try to attend a four-year college is an IQ of 105, which cuts off the top 37 percent of the distribution.

5. We’re currently giving out BAs to about 35 percent of all 23-year-olds.

So what is the evidence that lots more kids could successfully complete college if we were to try harder? Don’t tell me about a miracle school where inner-city kids are doing just as well as suburban kids (as Winters does, about the Carl Icahn school in the Bronx), unless you’re prepared to give me interpretable data that, in reality, none of these miracle stories can offer. And don’t tell me that short-term increases of 0.2 standard deviations that fade out within a few years do the trick.

One last thought: If we really have the best interests of young people at heart, when do we start counting the costs—emotional, financial, and in opportunities—of a dropout rate from colleges that is in excess of 40 percent?

I don’t want to reduce the number of students who get more education after high school. I don’t want fewer students on college campuses. I just don’t want them to be stuck there for four years in the straitjacket of a program leading to a bachelor’s degree. The BA evolved in the 19th century for purposes that have nothing to do with the needs of today’s economy or today’s students. In fact (okay, so I’m not being dispassionate here), the BA has become the work of the devil.

Mark Schneider nailed one problem posed for No Child Left Behind (NCLB) by the new math results—the gains in math prior to NCLB were larger than they have been since. There’s another problem Mark didn’t take up: No progress in achieving the most highly touted objective of NCLB, closing the gap in black and white test scores. From 2004 to 2008, the difference in scores went up by a trivial 2 points for 9-year-olds, down 2 points for 13-year-olds, and was unchanged for 17-year-olds.

To examine trends over a longer period of time, the National Center for Education Statistics has implemented an easy and flexible online tool, the NAEP Data Explorer, that makes it easy to ask whatever questions you wish. I created tables with the scores and standard deviations for blacks and whites, then calculated the black-white gap in terms of standard deviations. (For people who aren’t familiar with standard deviations but are familiar with the SAT math test, a standard deviation is around 115 SAT points.) Here’s what the plot looks like:

The NCES changed the assessment format in 2004, reporting scores for both the old and new version, hence the funny discontinuities in the graph at 2004. But the overall story is pretty simple. In the early years, the gap narrowed,  most impressively for 17-year-olds. For students ages 9 and 13, not much has changed since then. The narrowest gap for 9-year-olds was reached in 1986 and has moved within a narrow range since then. The narrowest gap for 13-year-olds also occurred in 1986, followed by a rise through the 1990s, then a decline in the 2000s to about where it was in 1986. The narrowest gap for 17-year-olds occurred in 1990, followed by a widening. It is not an encouraging picture.

Now suppose you approach the data differently. Instead of looking at the trend by the year the test was administered, look at it according to the year the children were born. This means that we have multiple sets of scores for some years. For example, children born in 1971 are represented at all three ages (9, 13, and 17) through the National Assessment of Educational Progress (NAEP) tests administered in 1980, 1984, and 1988.  If you average the black-white difference across age groups, this is what you get:

The narrowing of the black-white gap extends from children born in 1961 through children born in 1973, and it was substantial—from 1.2 standard deviations to about .8. Then the trend goes flat, with a few spikes, for children born over the next 26 years.

You can’t use this graph as proof of anything for a variety of technical reasons (the systematic differences in scores for different age groups have to be taken into account). I present it because it is consistent with many other datasets, analyzed in more sophisticated ways.

A few years ago, my colleagues (and friends) Bill Dickens and Jim Flynn published a widely publicized paper arguing that the standardizations of the major IQ tests show that the black-white IQ gap dropped by 4 to 7 IQ points between 1972 and 2002. Those years refer to the years that the tests were administered. As far as I can tell, their findings are almost completely consistent with my alternative proposition, to wit: the black-white gap in cognitive test scores narrowed for people born from the first decades of the 20th century into the early 1970s, and has not dropped for people born thereafter.

The timing of the narrowing is hard to pin down, but the greatest portion of it might well be concentrated among people born in the 1920s, 1930s, and 1940s. The evidence that the narrowing stopped for children born since the mid-1970s is much more detailed—from the NAEP, the SAT, the standardizations of the Wechsler adult IQ test and of the Stanford-Binet IQ test, and, most powerfully, the extensive testing of the thousands of children born to women of the 1979 cohort of the National Longitudinal Survey of Youth.

There was also narrowing between the 1979 and 1997 cohorts of the National Longitudinal Survey of Youth, but all we know is that it occurred among people born sometime between 1964 and 1980—we don’t know what birth years account for the narrowing. The only data that directly contradict my proposition (that I’m aware of) come from the standardizations of the Wechsler IQ test for children. The latest results from the NAEP are one more bit of grist for my side of the argument.

So I solicit evidence that the black-white gap in test scores has narrowed for people born since the mid-1970s, and suggest that, if the balance of the data continues to point in the other direction, it behooves social scientists to explore why the narrowing was so substantial before then and so flat thereafter.

Another case of two bloggers talking past each other. I write a post saying that miraculous test score increases—the kind that Andrew Coulson originally cited for Ben Chavis’s school—never survive scrutiny. He responds with evidence that scholarly reviews of the literature often show more than marginal increases for voucher and charter school programs. He uses .15 standard deviations as the demarcation line between a marginal and more-than-marginal effect. To get a sense of the size of a .15 standard deviation increase: On the SATs with their 800-point top score, .15 standard deviations amounts to about 15 SAT points. Not the kind of increase that would get Kaplan or Princeton Review much business if that’s what they promised. But I won’t argue with Coulson’s characterization of the literature. In fact, I agree with his conclusion that “these are more than marginal improvements, and they are part of a consistent pattern.” But claims of “more than marginal improvements” are not what bother me. I had in mind the claim that Coulson made for Chavez—a huge turnaround of student achievement, whereby kids from the inner city suddenly erase the usually-observed achievement gaps. Such claims are made frequently by many people, and all I’m requesting is a citation of a study that meets the criteria I listed in my last post.

There are a couple such studies with comparison groups, such as the one for the DC voucher program. In the case of oversubscribed schools where students are chosen by lottery, those comparison groups are excellent. But they’re not enough. Put aside the topic. Suppose we were looking at the evaluation of a government social program that claimed huge reductions in delinquency by young people who were diverted from incarceration into some touchy-feely psychological program, and it had a control group of delinquents who were incarcerated. People like Coulson (and I as well) would scoff if these claims were made without considering the subsequent criminal records of the program dropouts, or without having at least a few years follow-up to see if the effects lasted. But when it comes to programs we like (I, like Andrew, am a proponent of market-run education), somehow these basics for assessing the effects of a program can be set aside, and we can leap upon whatever evidence comes to hand. It shouldn’t work that way—sauce for the goose, sauce for the gander, and all that.

The Cato Institute’s Andrew Coulson, who prepared the graph that I posted a few days ago, has responded that I’m wrong in saying that school reform can produce only marginal improvements in test scores, citing Ben Chavis’s accomplishments with some charter schools in California as proof to the contrary.

A couple of things that we aren’t arguing about: First, I’d send my own children to the American Indian Public Charter School that Ben Chavis ran. The descriptions I’ve read (I can’t put my fingers on the full-scale profile I read, despite Googling) convince me that he’s a great principal and ran a fine school. Second, I was referring to marginal test score improvements on a large scale, not what is possible in a single terrible school. But only a few percent of the nation’s students attend the worst-of-the-worst schools, and national or even state-wide test score results are driven by the huge majority of students who attend schools in the normal range, where improvements are modest.

However, I do disagree with Coulson in an important respect, and I might as well lay it out: I have yet to see a single, solitary example of a school that produced dramatic improvements in test scores that stood up to scholarly examination. They always have evaporated—not completely, but down to small puddles. Here are the questions that a scholarly examination must ask:

1) Were the tests conducted by the same people who reported the results?
2) Were the students tested representative of the entire student population (or were certain kids mysteriously absent that day)?
3) Are the results compared to those of a legitimate control group?
4) Were there practice effects from teaching to the test?
5) Has attrition been taken into account? (The failure to include the subsequent performance of the kids who dropped out of the program or school is usually the single most decisive artifact of inadequate evaluations.)
6) Was there a test for fadeout two or three years after the exit test? (Fadeout of initial results has been universal when such tests have been conducted.)

My prediction: the miraculous results of Ben Chavis will also evaporate down to marginal improvements when all of those questions have been answered about his schools. I’m not saying that he did a bad job or didn’t produce better education for his students, or that he’s undeserving of our applause. But, just as TINSTAAFL, TINSTAMTSI.

Okay, Andrew—and Rick Hess, Jay Greene, and Paul Peterson, and my other friends and colleagues who think I’m a grinch on this: give me the citation that proves me wrong.

One of the most oversold social policy interventions, running neck and neck with sending unqualified kids to college to try for a bachelor’s degree, is early childhood intervention. The programs used for the optimistic claims are primarily Perry Preschool and Abecedarian, both of which were extremely small-sample programs evaluated by the same people who ran them and riddled with serious methodological questions. The advocates of early childhood intervention seldom even mention the Infant Health Development Project, a much larger, independently evaluated replication of Abecedarian for low birthweight babies, which produced extremely modest results—no significant results at all, some argue. If you want more description and references, see chapter 2 of Real Education. But I’ve been making similar points for years, without visible effect.

A great idea has surfaced for blowing the claims for early childhood intervention out of the water. Congressman Jared Polis has proposed creating securities whereby people could invest in early childhood education and participate in the eventual returns. I think we all ought to get behind this idea, and thereby prompt some unsentimental hedge-fund guys to take a hard look at the claimed returns for early childhood intervention and the data that are being used to support those claims. I predict their technical conclusion will be “You can’t be serious.” Maybe someone will pay attention to them.

I suppose Rick Hess is right about the good intentions behind the president’s national address to schoolchildren, but I’m getting less and less worried about intentions and more and more worried about the creepiness factor in this presidency. What I liked about Barack Obama during the campaign, and what made me believe he had a chance of being a seriously good president, was his persona—smart, good humored, reasonable, disciplined. Where did that guy go? I do not throw “narcissistic” around casually, but I’m hard put to find another word to describe his disconcerting sense of what presidents are supposed to do, with his plan to deliver a national address to elementary and secondary students on September 8 being the latest case in point.

Presidents have gone to visit schools from time out of mind. That’s fine. But a national speech? Has any other president ever done anything like this? I can’t think of one, and can’t imagine that any other president would have thought it was an appropriate thing to do. I suspect they would have thought it a little presumptuous. A national speech to schoolchildren is something that a Dear Leader would do. Something that somebody who thinks he is God’s gift to children would do. Does it really not even occur to Mr. Obama that millions of parents are going to be unhappy about this?

And Mr. Obama has so many enablers who are ready to help build the cult of personality, as in the “Menu of Classroom Activities” that the Department of Education asks elementary school teachers to use on the day of the president’s speech. An excerpt:

Students could discuss their responses to the following questions:
What do you think the president wants us to do?
Does the speech make you want to do anything?
Are we able to do what President Obama is asking of us?

Read the whole thing to get the gestalt of it. The “Dear Leader” analogy is not all that far-fetched.

To those Obama supporters who put this kind of reaction down to angry conservatives, ask yourselves: quite apart from your political views, if George W. Bush had proposed to make a national speech to schoolchildren, complete with lesson plans, isn’t “creepy” a word that would have come to mind?

The ACT—the competition to the SAT as a college-admissions test—announced in its annual report released today that only 23 percent of high school students who took the ACT had a 75 percent chance of getting a C or higher on all of four entry-level college courses  (English composition, reading, mathematics, and science). The ACT can figure this out by comparing ACT scores of high school seniors with their performance as freshmen.

What does this mean? In thinking about that question, start with another datum: about 40 percent of all 18-year-olds, representing about half of high school graduates, enroll in a four-year college.

Now think about this: That 23 percent figure refers to students who took the ACT, a self-selected group. But hundreds of thousands of students who go to college did not take the ACT (or the SAT). Statistically, they must be expected to get substantially lower scores, if required to take the ACT, than those who self-selected to take it. The proportion of all college-bound students who have the academic preparation and ability to succeed in college is a fraction of the 23 percent that meet the ACT’s definition of college readiness.

How small a fraction? In a book I published last year, Real Education, I used the College Board’s analysis of college readiness to calculate an answer to that question. In an excellent little College Board study by Jennifer Kobrin, the definition of college readiness was a 65 percent chance of getting a 2.7 GPA in the freshman year—a C+/B– average. Of students who take the SAT (another self-selected group), about 22 percent could meet the SAT score benchmark in the verbal test and about 19 percent in the math test. I estimated the percentage of the entire 18-year-old cohort that could get those scores if everybody took the test. Three independent methods led to estimates converging on 9 percent to 12 percent, with a best guess of 10 percent, in an age when about 40 percent of all 18-year-olds go to four-year colleges.

The ACT results point to the same ballpark. The ACT definition is easier than the College Board’s in one sense (a 2.0 GPA is easier to get than a 2.7 GPA) but more demanding in another (asking for a 75 percent chance instead of a 65 percent chance). On balance, I’d bet that the ACT’s definition is a tad softer, meaning that the ACT results are marginally more depressing than the SAT’s analysis.  

Sending 40 percent of our children to four-year colleges when perhaps 10 percent of them can be expected to do well—not just survive, but do well—is stupid. I suppose I should choose a more judicious word, but let’s face it. “Stupid” is the mot juste.