October 6, 2005

The NYT finally reveals a big truth about school testing

"One Secret to Better Test Scores: Make State Reading Tests Easier"

PARENTS are delighted when state test scores go up. Obviously, their children are getting smarter and the teachers are doing better. Politicians are ecstatic; their school reforms must be working. Indeed, during his re-election campaign, Mayor Michael R. Bloomberg has repeatedly cited the rise in the city's 2005 fourth-grade test results (up 10 percentage points in English to 59 percent at grade level, and up 9 points in math to 77 percent) as proof that his school programs are a success. "Amazing results," he said, that "should put a smile on the face of everybody in the city."

However, those in the trenches, the teachers and principals, tend to view the scores differently. While they would rather be cheered than booed, they know how much is out of their control.

Take Frances Rosenstein, a respected veteran principal of Public School 159 in the Bronx. Ms. Rosenstein has every right to brag about her school's 2005 test scores. The percentage of her fourth graders who were at grade level in English was 40 points higher than in 2004.

How did she do it? New teachers? No, same teachers. New curriculum? No, same dual-language curriculum for a student body that is 96 percent Hispanic and poor (100 percent free lunches). New resources? Same.

So? "The state test was easier," she said. Ms. Rosenstein, who has been principal 13 years and began teaching in 1974, says the 2005 state English test was unusually easy and the 2004 test unusually hard. "I knew it the minute I opened the test booklets," she said. [More]

The corrupt hole at the heart of the No Child Left Behind law is that states get to make up and manipulate the tests by which the federal government evaluates their progress toward meeting NCLB goals. That should have been obvious all along that NCLB was utterly undermined by this built-in conflict of interest, but the truth was that NCLB is just a feel-good fantasy law. Mandating that all students score as "proficient" on achievement tests by 2013 is simply a demand for the states to corrupt the test.

By the way, one of the more subtle purposes of making the tests easier it to make it look like the racial gaps are narrowing. That's because journalists always measure the racial gap by the simple minded use of percentage point differences rather than by standard deviations. For example, if in Year 1, 84% of the whites 50% of the blacks pass, then the racial gap will be reported in the newspapers as 34 percentage points. Then say the test is made much easier so that 98% of the whites and 84% of the blacks pass. Well, then the racial gap has "narrowed" from 34 to 14 percentage points, so, whoopee, cash bonuses for everybody! But, of course, the racial gap is still about one standard deviation.

Underground statistician La Griffe du Lion pointed out that the racial gap is always going to be zero in percentage point terms for a test so hard that everybody scores 0% or for a test so easy that everybody scores 100%. The racial gap in point difference terms will be largest when scores fall in the middle, so administrators have an incentive to start off with hard tests where scores fall in the middle of the range, then make them progressively easier to make it look like the kids overall are getting smarter and the racial gap is closing.

My published articles are archived at iSteve.com -- Steve Sailer

No comments: