BLUNDERS BLAMED ON SHORT TIME; COMPANY OFFICIALS ALSO CITE PROGRAM ERROR FOR BAD DATA.

Byline: Terri Hardy Sacramento Bureau

A national testing company committed a major programming blunder and incorrectly calculated exam scores for fluent and nonfluent students throughout California because of inadequate quality control and too little time, testing officials said Wednesday.

The president of Harcourt Educational Measurement said the company would revamp the way it checks results and promised corrected data would be delivered by July 15 - throwing into disarray a planned statewide release of the much-anticipated results.

``We're aware that this problem is very, very serious and we regret our error deeply,'' said Harcourt President Joanne Lenke, whose San Antonio, Texas, company is receiving $22 million from California school districts for its testing services. ``Had we had more time, we could have detected the problem and corrected it.''

When asked whether the company's quality-control system was flawed, Lenke said: ``Obviously it is.''

Both Harcourt and state Superintendent of Public Instruction Delaine Eastin said the overall test results of the Stanford 9 achievement test are accurate, showing a 2- to 4-point increase in scores statewide. Individual student scores that are to be sent home to parents are also correct.

But the test errors are a major debacle nonetheless.

``It's most unfortunate,'' said Scott Hill, Eastin's chief deputy, who said state officials have growing doubts about the testing company's competence. ``Obviously, the concerns brought forward do not provide a high level of confidence and our expectations are extremely high.''

Added importance

Never before have test results been so crucial in the state.

The exam results will be the major criteria used when poorly performing campuses are selected for extra help under Gov. Gray Davis' school improvement plan. Also, every school will be ranked according to students' performance.

The Los Angeles Unified School District has its own accountability plan under way. Schools with lagging test scores face serious consequences, including the transfer of employees and the takeover of schools by the district's headquarters staff.

Scores also are being closely scrutinized for the nonfluent students as the state implements Proposition 227, which essentially bans bilingual education and places students in one-year English immersion courses.

``The stakes are very high,'' Hill said. ``That's why it is imperative that these numbers are credible and reliable.''

Results were skewed when Harcourt grouped the scores of nonfluent and fluent students. The company incorrectly placed into the nonfluent group about 225,000 students who had been in bilingual classes but were subsequently deemed fluent in English.

Big gains gone?

The resulting figures indicated large increases in the test scores of the limited-English-speaking children - results which were widely publicized in early release of data by some school districts as an indicator that the elimination of bilingual education under Proposition 227 was working.

Eastin said she still thinks there will be increases in those test scores but expects they will be about 5 to 10 percentile points lower than reported by Harcourt.

School district officials from the Anaheim City School District and San Jose Unified School District noticed problems with the numbers during their own evaluations of the nonfluent group.

Although Harcourt's Lenke said the company did not find out about the problem until state officials called the firm Tuesday evening, a top administrator from San Jose Unified said he notified the company about the problem earlier in the day.

``I told them that they showed more LEP students in their analysis than I did,'' said Aaron Buchanan, director of education accountability for San Jose Unified. ``I told them their report didn't match with our numbers. I was concerned, very concerned.''

Eastin noted that these were not the first problems with Harcourt. The company failed to deliver tests on time to 200 districts in California, and in Arizona was weeks late in reporting test results.

Doesn't add up

Lenke defended her company, saying that in most cases it was the fault of schools or districts for turning in damaged or incomplete testing forms. In Arizona, Harcourt negotiated a revised delivery date because of document problems, she said.

Ann Bancroft, a spokeswoman for Gary K. Hart, California's secretary of education, said the state Board of Education will be evaluating Harcourt's performance and has the authority to withhold funds from the testing company.

Bancroft said Hart, who was out of the state, was ``disappointed'' when he learned of the testing snafu.

Harcourt also has posted a performance bond with the state. Lenke said she would not comment on whether the company will turn over those funds to the state.

Harcourt's explanation of the foul-up was contradictory and vague.

On Tuesday, Ed Slawski, senior research scientist for Harcourt, said the company knew there were problems with the data before being contacted Tuesday night by the state Department of Education.

``We did the counts and they didn't work, the numbers didn't add up - that's the first thing you check when you look at data,'' Slawski said Tuesday.

But in a telephone press conference the next day, Lenke said the testing company was unaware of the problem before being contacted by the state.

Differing defenses

Company officials said they weren't tipped off to problems even when they saw large increases in nonfluent student scores.

``Some gains are expected. It's difficult to estimate what those gains might be,'' Lenke said.

At first, Harcourt's president said company officials would have had a difficult time detecting the programming problem, and typical quality-control measures wouldn't find such a glitch. Then, she said that if they had enough time, they could have found the error.

If the company had two weeks to evaluate the data, it would have been able to look at the numbers of nonfluent students this year, and match them with last year's figures, she said.

``We would have seen a bump and we would have questioned that,'' Lenke said.