Thousands more Oklahoma students were held back in early grades than what the U.S. Department of Education reported, according to newly released state data.

The state Education Department’s data is likely more accurate because it is pulled directly from the public-school enrollment database, whereas the federal data is collected from schools in a biennial survey.

Parents and educators are worried about the high number of retained students because research shows those students are more likely to drop out of school. It’s also more costly to the state than providing educational support within the students’ grade.

The federal data came from the department’s Civil Rights Data Collection, which includes information from every public school and district in the U.S. on topics like student suspensions and discipline, law enforcement in schools and advanced math and science courses offered. Oklahoma Watch used the data for a Dec. 14 story that reported early-grade retentions are happening more often in Oklahoma than in all but one other state.

The federal data, released in April for the 2015-16 school year, shows 3,977 Oklahoma kindergarten students were held back. State data for the same year shows 4,909 kindergarten students held back, or nearly 1,000 more students (23 percent more). The state counted 1,104 more first-grade students, 542 more second-grade students and 375 more third-grade students who were held back that year compared with the federal data.

The state Education Department doesn’t typically collect and report student retention data. But Oklahoma Watch in October requested a count of students enrolled in the same grade for two consecutive years, and the information was provided Friday.

There are several possible explanations of why the numbers on retention in the two data sets differ. The definition of retained students could vary. Also, which schools are counted could vary; charter schools and juvenile justice facilities are examples of sites that may be included in one data set but not in another.

But if previous news reports are any indication, the main discrepancy is probably human error. Mistakes get through due to lax data entry, either by accident or from a miscalculation.

Errors in the Data

Journalists and researchers have found a number of errors in the Civil Rights Data Collection.

In one example, NPR and Child Trends, a nonpartisan research organization, found an error in the number of school-related shootings. The Civil Rights Data Collection reported such incidents in 240 schools in 2015-16, much higher than in previous reports. When NPR and Child Trends attempted to verify the number, they were able to confirm just 11 incidents. The ACLU looked into it and found two school districts mistakenly reported each of their schools having a shooting that year, accounting for 63 of the reported shootings.

In another example, Education Week found that the number of school desegregation cases doubled between 2013-14 and 2015-16 following a sharp drop two years prior. Federal education officials couldn’t explain it.

Also, South Dakota, which is using data from the Civil Rights Data Collection in its school report cards, found inaccuracies in school-related arrests, reported the Rapid City Journal in December.

At least one Oklahoma district found its retention data was misreported. It tried to correct the information but found out too late.

Kathy Dunn, assistant superintendent of Mid-Del Public Schools, said the retention numbers published by the Office of Civil Rights are higher than what Mid-Del reported. When the district attempted to appeal, they were told the window for corrections had passed.

“With our best eagle-eyes in the district studying the data, we have not been able to find any explanation for why the report data is different than what now shows on their website,” Dunn said.

So why do journalists and researchers keep using the data? In short, it covers subjects of interest to the general public, such as equity, school safety, discipline and chronic absences, and most states are not collecting data in those areas, said Paige Kowalski, executive vice president of the Data Quality Campaign, a nonprofit organization advocating for better education data.

While the U.S. Department of Education has been collecting data on key education and civil rights issues since 1968, only in recent years has the general public had access to it and the ability examine it.

The main incentive encouraging accuracy now is knowing that schools will be held to account by the public, she said.

“That busy school secretary is now going to spend a little more time to ensure the data submission is correct, because now there’s an incentive,” Kowalski said.

Kowalski said her group encourages states to collect the information for the Civil Rights Data Collection and submit it on local schools’ behalf. States could improve the accuracy and frequency of the data in this way, and it would be a “huge service” to local districts, she said.