Tests Fail the Grade

Test scores measure results. So, it’s obviously critical that the scores be accurate if educators and parents are to assess how public school students are performing. Last week, the testers failed the children of California.

The new school accountability system, proposed by the governor and approved in special session by the Legislature, depends on measurable results for judging the performance of students, teachers and schools. That intensifies concerns about the accuracy of this year’s scores on the required Stanford 9 test and the validity of certain math questions.

The flawed test results overshadow a glimmer of good news: a slight rise in overall reading and math scores, with the most significant progress in the primary grades where class sizes have been reduced.

Until Harcourt Educational Measurement, the publisher of the Stanford 9 test, gets its results for California absolutely right, Gov. Gray Davis, the Legislature, the State Board of Education, educators, parents and taxpayers who underwrite the cost of testing the state’s students, must keep up the pressure, including withholding a $2.3-million performance bond, 10% of Harcourt’s fee. Harcourt, its June 30 deadline shattered, now promises accurate results by July 15. State experts should at least seek possession of the scores a few days earlier to make sure they are right before they are finally released.

Delaine Eastin, the state superintendent of public instruction, rightly declined to post the inaccurate scores on the state web site. The defective results, which showed an impressive gain by limited-English children who are in English-immersion classes because of Proposition 227, have already taken on a political life of their own.

The error was discovered not by those who designed or scored the tests, but by educators in the Anaheim and San Jose school districts who saw that the numbers didn’t add up. They correctly deduced that students who had already been designated as fully English-proficient had been lumped with students still struggling to learn English. That mistake could have been caught earlier if state testing experts had been allowed to monitor the work of the private contractor, but old political rivalries in Sacramento prohibited that simple safeguard.

Eastin also wisely asked about math problems on a special section of the test aligned with California’s rigorous new academic standards. How is it, she asked, that students who did poorly on the overall test were more likely to correctly answer difficult math questions than students who excelled overall? Good question. How, too, can the downturn in ninth-grade reading scores, a repetition of 1998 results, be accounted for given the impressive performance of the same students when they were in the eighth grade? Another good question.

The Stanford 9 mess has major consequences, not least the doubts that may be in the minds of parents when they review the scores of their children. As California demands more from students and teachers, results from standardized tests must be on time and absolutely trustworthy.