March 07, 2004

The Zorro of Statisticians

By Steve Sailer

The emergence of ultra-low cost self-publishing on the Web has allowed amateurs to post essays of a brilliance that often leaves me, a full-time professional writer, green with envy. Heterodox insights articulately expressed can be found at sites like Gene Expression, 2Blowhards, Randall Parker's ParaPundit and FuturePundit, Paul Cella, Thrasymachus, Dienekes, Gideon's, and many others.

To all those who write without monetary compensation, I say:

"Knock it off! It's hard enough for me to make a living in this racket without you guys giving good stuff away free. Remember what Dr. Johnson said: 'No man but a blockhead ever wrote, except for money.'"

Okay, I feel better now. Thank you for letting me get that off my chest.

Nevertheless, I can't resist calling your attention to one of my unpaid competitors who is perhaps the most bracing social analyst nobody has ever heard of.

Occam's Razor, the scientific imperative to prefer the simplest explanation, has few more productive fans than the elegant and mysterious academic who writes under the pen name La Griffe du Lion. On his website at www.lagriffedulion.f2s.com, Griffe, the Zorro of statisticians, wields relatively simple quantitative tools like a lion's claw to slash through a host of the most perplexing sociological questions.

I've known Griffe via email for years, but I still don't know much about him, other than he loves all things Italian. He writes:

"There are very few moments in a man's existence when he experiences so much hostility, or meets with so little benevolence, as when he challenges fashionable perceptions of race… As for our efforts, we can be certain of only one thing—vilification. It could drive a man to pseudonymity."

By contemporary standards, Griffe is certainly guilty of crimethink. For example, here's one of his most important observations:

"The fundamental law of sociology is a summary of hundreds of observations. It asserts that:

"On large-scale tests of reasoning ability, the observed mean difference between non-Hispanic whites and African Americans is 1.1 + 0.2 standard deviation.

"The observation is so unerringly reproducible, it justly earns the appellation, law. Appropriately, we call 1.1 SD the fundamental constant of sociology."

It's crucial to note that Griffe's studies this difference not to make any group feel bad, but to find out what reality is. Only honesty offers us any chance to ameliorate our problems.

Indeed, Griffe's analytical tools allowed him to determine that one university minority program actually does work to raise black performance in the sciences—the Meyerhoff Scholarship at the Baltimore County campus of the University of Maryland (UMBC). He provides a detailed explanation of how this obscure commuter college gets more out of mathematically talented black students than do Ivy League universities.

As Claude Steele has famously pointed out, smart blacks typically earn worse grades at elite college than their SAT scores would predict. Yet the Meyerhoff Scholars do better.

At Ivy League colleges, even black students who would have gotten in without affirmative action tend to be below the class average in ability, which can be dispiriting. At lowly UMBC, in contrast, the Meyerhoff Scholars are among the best students on campus, and they know it. The program works hard to build esprit de corps to get its students to encourage each other to keep up the struggle for grades, just like a good football coach gets his players to set high standards for each other.

Griffe's analysis should be required reading by college administrators everywhere.

Some of Griffe’s other contributions to science:

"Racial Disparities in School Discipline
There are among us persons of so refined and delicate a nature that they cannot bear the guilt even of crimes they have not committed. Their shame is so great that they turn their considerable talents to serve the demagogues of bias. In this essay we analyze their efforts to document racial discrimination in school discipline, and humbly offer advice on how to improve their methods."

Using strikingly similar international Interpol and domestic FBI data on criminal assault rates, Griffe estimates that blacks exceed whites in aggressiveness by between 0.3 and 0.4 standard deviations. The American data on school suspensions shows the same differential. Fortunately, according to Griffe, one early childhood intervention program for inner city blacks, the Chicago Child-Parent Center Program, has been shown to reduce arrest rates 15 years later by an impressive 0.3 standard deviations. I don't know what makes this program more effective than so many others, but, clearly, it should be studied closely.

"The Death of Meritocracy  The noise has subsided, and with passions contained we look back at [anti-affirmative action measures] Prop 209 and Hopwood. Our goal: to check for compliance with the law….[W]hen we saw admissions data from the medical schools of the University of California and the Law School at the University of Texas, we found noncompliance so blatant that simple inspection revealed it…Under Prop 209, the UCLA Medical School admitted 51 blacks and Hispanics in 1997. The chance of that occurring without the use of preferences was 1 in 10³⁶⁴. (There are about 10¹⁰⁰ fundamental particles in the universe.)"

No comment necessary!

"The Color of Death Row  
For those who desire a dispassionate view of death-row justice, let them know that no axe will be ground here…let them also know they will discover that justice depends on geography, that much of America is fair, and that bias on death row affects mostly whites."

Griffe found that the most notable bias in death sentencing is found in the South, where white murderers are much more likely to be sentenced to death than black murderers. Just last month, a massive study by Cornell researchers came to the same conclusion: "The excess of the African-American percentage of murderers over the African-American percentage of death row is greatest where the conventional wisdom would least expect it—in the South."

"The Politics of Mental Retardation: A Tail of the Bell Curve Political movements have victims, and the cause of diversity is no exception. Whites, Asians and males are all casualties of the diversicrat, but his most deplorable incivility makes victims of the hapless.  If anyone should deny the politicization of mental retardation, let him confront the data presented herein."

Griffe discovered that the Department of Education and other sensitive sorts have reduced the percentage of children eligible for special education from 2.2% in 1977 to 1.3% in 1997. Why? Because retardation (as defined by the Supreme Court as an IQ below 70) is found about five times more often in blacks than whites. Liberal bureaucrats prefer to cover up this embarrassing fact by tossing a lot of retarded black kids into mainstream classes to sink or swim.

Griffe's personal Occam's razor is the normal distribution (a.k.a., the bell-shaped curve). He might sometimes push it too far in using it to model the extreme left and right tails of the population, where a more complex probability distribution would be more appropriate. (Here, for the statistical cognoscenti, is a discussion on Gene Expression of the technical issues.) Still, the Bell Curve works well enough for modeling a striking array of social phenomena.

Further, the simplicity of the normal distribution makes it hugely useful for quickly grasping how the world more or less works. Readers can easily perform bell curve calculations of their own in MS Excel using Normdist and related functions.

Griffe's most recent effort, "Closing the Racial Learning Gap," is a dramatized Socratic dialogue where Griffe's alter egos Mentor and Prodigy explain why the conventional measure of test score differences is misleading to Agnesina and Santo Thurston, who are promoting their book No Alibis. (Any similarities to Abigail and Stephan Thernstrom and their book No Excuses is wholly coincidental, especially the conclusion where Agnesina dumps her historian husband Santo because of his invincible innumeracy and runs off with young Prodigy).

In 2000, President Bush ran on the "Texas miracle" in education, then made it the inspiration for the federal No Child Left Behind act. Bush bragged to the Urban League last summer that his policies had eliminated 25/35ths of the white-black test score gap in Texas:

"In my state [Texas], 73 percent of the white students passed the math test in 1994, while only 38 percent of African-American students passed it. So we made that the point of reference. We had people focused on the results for the first time—not process, but results. And because teachers rose to the challenge, because the problem became clear, that gap has now closed to 10 points. Because every child can learn, you've just got to focus the attention and the resources when necessary."

Griffe, a terrible spoilsport, points out that Bush couldn't pass a test in elementary statistical reasoning.

Instead of subtracting the black percentile from the white percentile, he should have subtracted the black score in standard deviations from mean from the white equivalent.

Let me illustrate using stylized data. Say that when you start out testing, 84% of white students and 50% of black students pass. That's a one standard deviation gap.

Then, you make sure the pass rates rocket upward. Dan Seligman's new Forbes article—" Children Will Be Left Behind: George Bush's school accountability law is something of a fraud"—explains some of the techniques states use to game the system: You make the test easier, you teach to the test, you make sure the dumb kids don't show up on test day, etc. Who knows, maybe you even teach the kids better.

Scores on national tests conducted by independent testing agencies, such as the SAT and the NAEP, have been rising slowly. But these real gains are nowhere near as spectacular as those seen on many state tests made up and administered by the people who stand to benefit from rising scores. (Can you spell "conflict-of-interest?")

So maybe after a decade, you've managed to finagle the white pass rate up to 98% and the black pass rate to 84%. Wow, you've cut the racial gap from 34 points to 14!

But in truth, as Griffe explains, the racial gap is still exactly the same: one standard deviation. "If a test is made so trivial that nearly 100 percent of all students pass, group pass-rate differences must again be small, and go to zero in the low-difficulty limit," he explains.

Texas might have done a little better than this—raising the white pass rate from 73% to 97% and the black rate from 38% to 87%—but not much. (Griffe's statistical adjustments suggest the improvement was even less than it looks).

And the documented fraud in Education Secretary Rod Paige's Houston school district might cast even that narrowing of the gap in doubt.

Is it important to accurately measure the truth about such a sensitive subject? In his speech to the Urban League, President Bush made a strong case that it does:

"I know measuring and using the measurement system as a way to diagnose problems so you can focus on the problems works… Accountability tells you what's going right and it tells you what's going wrong and it shows you where the emphasis needs to be."

Perhaps when Education Secretary Paige's scandals become so odiferous that the President has to break his rule of firing only for disloyalty, never for incompetence, he should appoint Griffe to take Paige's place.

At least then we'd get to see what La Griffe du Lion looks like.