The three major credit bureaus, TransUnion, Experian and Equifax, publish their auto finance data in spurts throughout the year, usually a week or two apart from one another.
But one thing always stops me when reporting the data: The trends are the same, but the numbers don’t match up.
For example, TransUnion said the 60-day delinquency rate for auto loans during the first quarter was 0.99 percent, while Experian said it was 0.6 percent. Equifax measured each month individually, but the average of its rates for January, February and March was 1.08 percent.
Loan debt balances also varied among the bureaus. TransUnion said loan balances totaled $941 billion during the first quarter, while Experian said balances reached $905 billion. Equifax didn’t conduct a quarterly analysis, but for the first four months of the year, it said loan balances came to $934 billion.
So, which numbers are accurate?
It turns out they all are. The bureaus just have different analysis mechanisms, varied access to lender data and dissimilar definitions of subprime.
Slight variations could be due to credit score differences, said Melinda Zabritski, senior director for Experian Automotive. For example, Equifax defines subprime consumers as those who have an Equifax Risk Score below 620, while TransUnion and Experian consider consumers with VantageScore 3.0 credit scores lower than 601 to be subprime.
Data also varies because lenders choose which bureau to send information to, said Clifton O’Neal, TransUnion spokesman. One lender may send its data to TransUnion only, while another may send its data to all three credit bureaus.
TransUnion pulls from all the data it receives. Equifax uses the same model; it calculates its data based on all the information lenders provide, rather than a sampling.
“If you’re sampling, you’re going to get a different number based on how you sampled,” said Dennis Carlson, chief economist at Equifax. “When lenders report to us, we report on all trade lines we see. We aggregate across all the loans that lenders have reported to us.”
Experian, by contrast, works with samples. Researchers take “snapshots” of 10 percent of lenders’ provided data, said Melinda Zabritski, senior director at Experian Automotive.
The bureaus receive essentially the same data, but differences in technology and match logic for compiling credit files can yield conflicting results, she said.
“You could have some pure differences because of that random sample,” Zabritski said.
Since Experian uses sample files and because each bureau receives data from different lenders, “it’s not going to be apples to apples,” she said. “But you should see consistent trend lines across the files.”