Lower Scores, Fewer Students

Mean scores on the SAT fell this year by more than they have in decades. A five-point drop in critical reading, to 503, was the largest decline since 1975 and the two-point drop in mathematics, to 518, was the largest dip since 1978.

Gaps among racial and ethnic groups continued to be significant on the SAT, including the new writing test, for which the first mean scores were released at the College Board's annual SAT briefing on Tuesday. The board also reported a small decline in the total number of people who took the test, and while board officials insisted at a news conference that the decline was across the board, they acknowledged later Tuesday that the board's own data suggest that the decline appears to be among students from the lowest income families.

The percentage of SAT test takers with family incomes up to $30,000 was 19 percent for the high school class of 2006, down from 22 percent a year ago. The share of SAT test takers from families with incomes greater than $100,000 was 24 percent, up from 21 percent a year ago.

Gaston Caperton, president of the College Board, provided a generally upbeat assessment of the year's results, saying that the new writing test was off to a strong start, both strengthening the SAT and encouraging high schools to focus on writing skills. He attributed the drops in SAT scores to a decline in the number of students who took the test more than once. Fifty-three percent of students did so, down from 56 percent the previous year. Repeat test takers tend to improve their scores, Caperton noted, and students tend to alter their test-taking behavior in years when the SAT undergoes major changes, as was the case this year.

In light of these changes, he said he wasn't concerned about the one-year drops, although he remained seriously concerned that too many students are not taking rigorous courses in high school that lead to their doing well on the SAT and in college. He said that the average drops in SAT scores didn't even amount to a single additional question being answered incorrectly.

A reporter at the briefing asked Caperton why in previous years -- as SAT scores inched upward -- he had implied that those increases were signs of real progress, while he was playing down the impact of larger decreases. Caperton said that "I think we tend to overemphasize a few points here or there."

Christine Parker, who runs the SAT and ACT preparation programs for the Princeton Review, said that she was struck by the tone of the College Board's materials on this year's scores. "It's pretty clear that the board is on the defensive about these decreases," she said. She thinks that one reason the retesting totals are down is that more students are taking the ACT and the SAT and figuring out which score will help them the most with colleges, rather than simply retaking the SAT.

Many high school guidance counselors -- not to mention SAT test takers -- complained that the addition of the writing test made the SAT too long, and there has been much discussion of whether "SAT fatigue" contributed to the decline in scores.

But Wayne Camara, vice president for research and psychometrics at the board, said that the duration of the test had "no impact" on student scores, and that College Board officials have examined the rates at which students answer questions correctly or incorrectly or don't answer at all during all portions of the test. No link is evident between how long a student has been taking the test and the quality of answers, he said. The College Board has said that it will study the idea of letting students take different parts of the SAT at different times, and Camara said Tuesday that any determination on that idea was at least a year away.

As has been the case in past years, clear gaps were evident by racial and ethnic groups, with Asian and white students doing much better than other groups.

Mean SAT Scores by Ethnicity, 2006

Also consistent with past years, men outscored women -- 505 to 502 on critical reading and 536 to 502 on mathematics. But women had higher mean scores -- 502 to 491 -- on the new writing test. In some areas, subgroups of women outperformed men. For example, black women outscored black men on critical reading.

In most recent years, the total number of people taking the SAT has generally increased, but that was not the case this year, when there was a slight drop -- of just under 10,000 students -- out of a total of more than 1.4 million students who took the exam. During the press briefing, College Board officials insisted that the decline was not significant and that data indicated that it was across the board and not linked to any demographic group.

College Board data, however, show that the share of SAT test takers from the lowest income groups declined this year, while the share from the highest income group increased.

SAT Population by Income Level, 2005-6

The shares of test takers for those in the three categories up to $30,000 as well as those in $40,000-$50,000 declined this year, while there were increases for $70,000-$80,000 and those from families with incomes over $100,000.

The ACT -- which has been seeing increases in test takers, many of them people who also take the SAT -- uses slightly different income levels for its demographic comparisons. But ACT data show that there have not been notable changes among the share of test takers from various income groups, and that a much smaller share of students (10 percent) comes from families with incomes greater than $100,000.

Camara, in an interview after the briefing, acknowledged that the numbers are striking enough to suggest that the decline in test takers may be primarily from certain economic groups, but he said more study would be needed. He said that many students incorrectly report family income so he is skeptical of reading too much into answers on that question. Camara said he pays more attention to the question about parents' educational background.

But there too, the College Board's data suggest that the disappearing test takers are not coming from a broad cross section of the population. From 2005 to 2006, the percentage of SAT test takers whose parents' highest degree is a high school diploma or an associate degree declined while the percentage of SAT test takers whose parents have bachelor's or graduate degrees increased.

Camara said it was important to figure out what these drops mean because of the need to avoid having "students fall through the cracks."

One reason that economic demographics are important to the College Board is that the SAT mean scores follow a consistent pattern in which increases in family income correlate directly with scores.

SAT Mean Scores by Income Level, 2006

This year was the first with the writing test, with the most interest in the essay portion of that test. Essays are graded by two readers, providing scores on a scale of 1 to 6 for a maximum of 12. The College Board released the following information about the first year of essays and their scoring, based on overall averages and an in-depth study the board conducted of a sample of essays:

• Most essays received very similar scores from the two readers, with 97 percent of essays having scores that differed by one point or less. (Those with larger gaps had third readers.)
• The average essay score was 7.2 out of 12, with women leading men 7.4 to 7.1.
• Longer essays on average received slightly higher scores.
• Half of the essays used first person, but average scores were slightly higher for those who did not use the first person.
• Only 8 percent of essays used the standard five-paragraph essay structure.