Slicing and Dicing

Mitt Romney is the President of white male America.

-- Maureen Dowd, New York Times Columnist

Amid the post-Presidential election head shaking, while analysts are slicing the electorate into statistical racial, ethnic and economic slivers to try to determine why Americans voted the way they did . . . I noted the passing of a man who incited a major and discomfiting political debate over the real differences between racial and ethnic groups. 

Arthur Jensen died just a few days before the 2012 Presidential election at the age of 89. He was the University of California, Berkeley, psychology professor best known for his efforts to prove that the average member of some ethnic groups come into the world equipped with less “general intelligence” than the members of other ethnic groups, including whites.
In his book, Straight Talk about Mental Tests he enumerated two of his tenets:

(1) In human populations there is a well-recognized trait that can be called general mental ability or intelligence, in which differences among persons can be measured with a fair degree of accuracy by appropriate tests (often called IQ tests);

(2) Observed differences among persons in this trait are largely attributable to genetic inheritance.

He did not include in the enumeration another of his conclusions, but neither did he back away from it: Because the average IQ scores of some minority groups – including African Americans and Hispanics – are measurably lower than whites and Asians, most members of those ethnic groups are born with less mental ability than whites and Asians.

In 1980, I spent an uncomfortable 15 minutes on live television interviewing Professor Jensen after the publication of his most ambitious book, Bias in Mental Testing. In the book, Jensen phrased his conclusions carefully and narrowly, wrapping them in 786 of pages of statistical analysis and complex calculations that were virtually incomprehensible to the average reader. But a good portion of the popular press and some groups with questionable agendas grabbed hold of Jensen's conclusions and ran with them. Period. Too bad. Deal with it minority people, you can't fight science.

Jensen's argument was seductive in its mathematical complexity. There is no debate about his statement that African-Americans and Hispanics consistently have lower average scores on certain sections of standardized tests than other groups, including people with Northern European and Asian ancestors. But he set off fireworks by resurrecting a theory and terminology created by psychologist Charles Spearman in 1905, concluding that the statistical clusters of test scores within ethnic groups are an accurate measure of the group's hereditary mental capacity.

In our televised interview, Jensen spent much time emphasizing the dangers of reading too much into his conclusions. Unfortunately, too much time was spent parsing words, and I'm afraid the interview generated very little heat or light. But on paper, though the mathematics strewn throughout his book is intimidating, Jensen's conclusions are comfortably simple and intuitive.

The outcome of an IQ test is straightforward – a single number that is meant to represent a person's innate mental capacity. It's a number that can and has been used to compare one person's mental worth to others. The best schools, the best opportunities are often made available to those with the highest numbers.

The use of numbers to quantify intelligence is hardly new. Even before the development of standardized tests, intelligence researchers devised formulas that ranked innate mental capacity based on the physical size of the brain. Looking at those numbers, researchers deduced that women could never be expected to function at the same intellectual level as men.

In 1903, statistical researcher Léonce Manouvrier described the effect those numbers had on the debate about women's place in society:“Women displayed their talents and their diplomas. They also invoked philosophical authorities. But they were opposed by numbers. . . . These numbers fell upon poor women like a sledge hammer . . . Theologians had once asked if women had a soul. Several centuries later some scientists were ready to refuse them a human intelligence.”

In first four decades of the 20^th century, American developers of “I.Q.” tests had an overt political agenda. Their test results were used in a successful campaign to get the U.S. Congress to tighten immigration quotas to keep the country's gene pool from being polluted by ethnic and racial groups with a hereditary lack of intelligence. Among those excluded because of their limited intelligence, as demonstrated by I.Q. tests, were Jews and Asians. That policy had tragic implications for many Jews seeking refuge from the Holocaust during World War II. There is bitter irony in the fact that Jews and Asians are among the groups with the highest average “general intelligence” scores on today's standardized tests.

Until I began preparing for the interview with Professor Jensen, I had never given serious thought to the question Jensen's research and writings brought to the forefront: Can we rely on modern intelligence tests to tell us something useful about a group of people? Professor Jensen, in claiming the thing standardized tests measure to be innate “general intelligence” implies the answer is yes.

But can it be used at a group level to predict performance or success in school, work, or life in general?

In a report sponsored by the Ralph Nader's organization, author Allan Nairn produced data showing that the results of standardized Scholastic Aptitude Tests (the “SAT”) had a higher correlation with the household income of a test-taker's parents than with future academic success. In fact, parental income turned out to be a better predictor of how well an individual will do on standardized tests than ethnicity.

In a 2009 paper entitled, “Reflections on a Century of College Admission Tests*,” former University of California President Richard Atkinson and Saul Geiser of UC-Berkley's Center for Studies in Higher Education concluded simply, “High-school grades are the best indicator of student readiness for college, and standardized tests are useful primarily as a supplement to the high school record.”

Still, psychologists and social scientists are uncomfortable dealing with criteria where they can't control all of the variables. Some schools have higher standards than others, some schools offer more activities, and some teachers pack their grade books with A's, others with C's and D's.

In theory, a person making hiring or college admission decisions based on standardized test scores should be bias-free because the decision-maker's mind is uncluttered with biases about whether the owner of an individual score is black, white, Asian or Latino; rich or poor, short or tall, attractive or homely. The flaw in that theory is, the test itself seems to have a good general idea about which groups a test-taker belongs to.

In their paper Atkinson and Geiser cited a study of admission applications for the University of California that found when UC applicants were ranked by SAT scores, only half as many Latino, African-American and American Indian students appeared in the top third of the applicant pool as when the same students were arranged by high school grades. Whatever their other virtues, standardized tests appear to put some racial and ethnic groups at a competitive disadvantage not supported by other criteria.

In The Mismeasure of Man, Harvard biologist Stephen Jay Gould observed that, historically, efforts to rank intelligence by group “invariably find that oppressed and disadvantaged groups – races, classes, or sexes-- are innately inferior and deserve their status.”

If Jensen had called his statistical clusters something like “ethnic\racial testing differential” his work would likely have been largely ignored outside of the testing community because it would have been considered a characteristic of the test. But calling it “general intelligence” assigns the characteristic to an entire, identifiable group of people, providing a reason to regard that group as qualitatively different from the norm.

Leaping back to the subject of electoral politics, remember the 47 percent of all Americans who were characterized as “takers”? Pundits have spent much of their post-Presidential election energy defining individual ethnic groups by how they voted. Certainly, common attitudes within an ethnic group is a valid thing to study, but I would suggest that we not allow ourselves to become distracted from looking at the most obvious characteristic that defined the largest group of voters in the 2012 general election: They favored a politically moderate President and rejected candidates who were on the extreme fringe of the political spectrum.

As analysts and statisticians sort individuals into narrower and narrower groups, I like to keep in mind one characteristic we all share. When things go wrong and it's fight or flight time, we instinctively divide the people around us into just two groups: Us and them.

This year, Republican strategist Karl Rove discovered how costly it can be to use the wrong criteria to define us.

* The Atkinson/Geiser paper is available online at: http://cshe.berkeley.edu/publications/docs/ROPS-AtkinsonGeiser-Tests-04-15-09.pdf