A study worked on by academics at several major institutions in collaboration with the U.S. Department of the Treasury found that Black Americans are anywhere from 2.9 to 4.7 times more likely to be audited by the IRS.

The working paper was published January 31 by the Stanford Institute for Economic Policy Research.

“The disparity is unlikely to be intentional on the part of IRS staff,” according to a summary fo the study.

“Rather, as the team’s research demonstrated, the racial disparity in audit selection is driven by a set of internal IRS algorithms that Goldin likens to the recipe for Coca-Cola. That is: It’s completely secret,” it added.

However, despite the fact that the IRS keeps many details under wraps about how exactly its algorithms select people to audit, the research team was able to identify the disparity (in tax returns from 2014, Black Americans were 2.9 to 4.7 times more likely to be audited than other groups) and suggest possibilities for what could be behind it.

The study found that the IRS was more likely to flag Black Americans through its “Dependent Database,” which uses risk scores and other factors to create a database of taxpayers who might have not-reported income, then generates an automatic letter audit.

The study was not able to say why, but it did show that “the bulk of the observed racial disparity” was related to by-mail, lower-effort audits (rather than ones that take place in person).

One reason for the disparity is the increased auditing of people who claim the Earned Income Tax Credit, (EITC) which gives a tax break to families of low or moderate income. That disproportionately affects Black taxpayers.

However, the largest racial disparity actually occurs because, within the group of people who claim EITC, the IRS is also much more likely to select Black Americans.

Black Americans faced 45% of audits among those who claim EITC — but were only 21% of the total group.

“The racial disparity in audit rates persists regardless of whether EITC claimants are male or female, married or unmarried, raising children or childless,” per the summary.

The study proposed a few other explanations, borne out by its data analysis, from the possibility that the racial bias is introduced by a potential focus on the number of instances of unreported tax income (versus how much actual income is not reported — i.e., attempting to find the largest number of reported income).

Another was the fact that the IRS audits people who report income from a business less often because those audits are more expensive, and that also disproportionately results in Black taxpayers facing audits.

The business income situation feeds into a larger trend of the IRS lacking the funds to audit high-income taxpayers, and thus focusing on lower-cost audits, which “have a racially disparate impact.”

The study used U.S. Census data to predict the likelihood a taxpayer identifies as Black, as the IRS does not collect data on race.