Error May Have Skewed Statewide Test Results

Firm admits misclassifying 300,000 test-takers' English skills. L.A. scores are affected.

The private company that is paid $34 million to administer statewide achievement tests admitted Tuesday that it had misclassified about 300,000 students as not being fluent in English, a blunder that may have skewed school district reports across California for the past month.

The effect of that error could be seen in scores released Tuesday by the Los Angeles Unified School District.

The district initially reported stunning, double-digit gains in the Stanford 9 scores achieved by students who were not yet fluent in English.

Later, because the overall scores had inched up only slightly, officials realized the gains couldn’t be real.

The misclassification doesn’t affect individual student scores that have already been sent home to parents or the overall trends in district scores.

Reached Tuesday night, officials for Harcourt Educational Measurement attributed the problem to programming errors. Ed Slawski, a senior research scientist for Harcourt, said the problem was discovered Monday.

But a state Department of Education spokesman said the company did not notify the state or school districts. Instead, officials began to suspect a problem and contacted the company.

Despite the error, test scores for the statewide program involving 4.2 million children in grades two through 11 were still to be released today.

State officials said they were bound by state law to post the data on the Internet although the scores for limited-English-speaking students may have been inflated. Scores for English-proficient students could be affected as well.

Doug Stone, a spokesman for the Education Department, said some of the numbers will be correct. “At this point, we don’t believe that the error has affected the accuracy of the overall statewide numbers,” Stone said.

Stone said Harcourt had told the state that it would take until July 15 to rectify the situation.

“This may have been an honest mistake, but it was a real mistake and we do expect Harcourt to fix it and fix it expeditiously,” Stone said.

Over the past two weeks, school districts from Orange County to Sacramento had begun reporting scores for their students in reading, math, writing and spelling. The numbers seemed to show dramatic gains for the one in five California students who are not fluent in English.

The Oceanside Unified School District reported that test scores for students still learning English had doubled or even tripled.

Such scores had been widely seen as vindication by backers of Proposition 227, the anti-bilingual education measure approved by voters a year ago. The approval of that law had banned most lessons not taught in English, and educators across the state had predicted that it would lead to an educational disaster.

Proponents of the proposition had exulted as the scores trickled out.

Ron Unz, author of the proposition, had boasted that it was “the most rapidly successful education reform program in history.”

When informed of the error Tuesday night, Unz said he remained confident that the corrected numbers would bear out such predictions.

“All the anecdotal evidence preceding the release of these test scores had certainly made a rise seem possible,” Unz said. “I would expect that, once the correct numbers are found, we would see a very strong rise in Oceanside. I would expect to see a rise in L.A. Unified and the other districts, though perhaps not as dramatic.”

Harcourt officials would not concede Tuesday night that scores for individual school districts were affected by the error.

“I can’t say with any certainty,” Slawski said. “It’s quite possible the district results are correct.”

However, problems were immediately apparent in L.A. Unified’s scores.

In some grades, limited-English-proficient students were reported to have made gains of eight or nine percentile points, although there were no overall gains for those grades. Since more than half of those taking the exam were limited-English-proficient, that would mean that the scores of native English-speaking students had declined by an even more unlikely amount.

Eva Baker, a nationally known testing expert at UCLA, said the dramatic gains reported for non-fluent students across the state had left researchers puzzled.

“We’ve been mystified by how these numbers could have been true,” she said. “We were hoping to find an explanation that would make some sense.”

She said parents, teachers and politicians “shouldn’t be disappointed by the reality of the test scores” if they show only modest gains. “This reinforces the idea that real gains in learning are achieved over the long term by teaching the important content the test measures.”

Supt. Ruben Zacarias said L.A. Unified’s overall gains were encouraging but that he is not satisfied.

“I have said no child should get past the third grade without reading at grade level,” Zacarias said. “We’re not there, but we’re headed that way.”

The district’s scores improved slightly in reading but remained in the bottom third nationally in every grade level. Math scores were higher, with five grade levels at or above the 35th percentile.

State education officials used the snafu to reiterate their view that the state ought to have a single contract with the test publisher. As it stands now, each of the state’s 1,000 school districts contracts directly with the company.

* * *

The Times will publish selected school-by-school data when it becomes available.

* * *

Times education writer Louis Sahagun contributed to this story.

Comments are closed.