The Transactional Records Access Clearinghouse at Syracuse accused the DOJ's Executive Office for Immigration Review of "silently but systematically deleting records" in datasets provided to the research center through the Freedom of Information Act, without disclosing that cases had been withheld.
The report also said that "garbled" and "inconsistent" data from EOIR on immigration court cases showed the agency had failed to conduct standard verification measures to check for accuracy. When TRAC flagged the problems, the agency stood its ground, claiming that FOIA does not require federal agencies to "certify the accuracy of data contained in responsive documents," the report said.
"It is deeply troubling that rather than working cooperatively with TRAC to clear up the reasons for these unexplained disappearances, the agency has decided to dig in its heels and insist the public is not entitled to have answers to why records are missing from the data EOIR releases to the public," TRAC's report said.
Kathryn Mattingly, a spokesperson for EOIR, said that the agency does not delete records, and that the missing records are "simply withheld from release to TRAC under a FOIA exemption."
"TRAC does not disclose their methodology and to the best of our knowledge, the EOIR data release is accurate and up-to-date," she wrote in an email to Law360.
According to TRAC, EOIR releases monthly batches of updated data to the research center in response to the center's FOIA requests, which TRAC used to maintain public data on the immigration courts system, tracking the growing immigration court backlog nationwide and by state.
The immigration court backlog recently surpassed 1 million cases, according to TRAC. This growing backlog has prompted the Trump administration to roll out a number of policies aimed at encouraging immigration judges and members of the immigration courts' appellate board to decide cases and appeals faster.
But last month, TRAC claimed that it discovered "serious inconsistencies" with the latest EOIR dataset, which was supposed to track cases through September 2019. The first time that EOIR set the dataset, the data was "incorrectly formatted," making it difficult to match variables to values, the report says.
TRAC flagged the issue, and EOIR sent the center several more versions, each with their own issues, the research center said.
After continuing to notice omissions, TRAC compared EOIR's final version of the September data with the dataset from the month prior and with the September 2018 dataset, and discovered even more cases were missing.
One thousand five hundred applications for relief that were present in the August dataset were absent in September's. Between the September 2019 dataset and the September 2018 dataset, nearly 900,000 applications for relief were missing.
TRAC also discovered several thousand more records that had disappeared in the last month, including records of court proceedings, charges filed by the U.S. Department of Homeland Security, scheduled hearing records as well as records related to a system that flags special immigration cases, including cases for minors and people forced to wait out their immigration court proceedings in Mexico.
The research center urged EOIR to check its datasets against past releases to make sure no records had slipped through the cracks, a practice it said is "industry standard for agencies managing large databases."
And if EOIR is withholding cases deliberately, it should keep track of those cases and disclose the numbers of cases being withheld, TRAC said.
This isn't the first time EOIR has bungled its data, TRAC said. In 2016, the U.S. solicitor general notified the U.S. Supreme Court that EOIR had made "several significant errors" when calculating data on how long certain categories of immigrants are held in immigration detention.
The high court had used that data when deciding Demore v. Kim
in 2003, which upheld mandatory detention for immigrants in removal proceedings.--Editing by Katherine Rautenberg.
Update: This story has been updated with a comment from EOIR.