The Internal Revenue Service audits Black taxpayers at 2.9 to 4.7 times the rate of non-Black taxpayers, mainly due to automated algorithms that flag discrepancies in claims for tax credits, according to a new academic study that looked at racial disparities in tax audits.
The
"Our results highlight how seemingly technocratic choices about algorithmic design can embed important policy values and trade-offs," said the study, conducted by a team of researchers from Stanford University, the University of Michigan, the University of Chicago and the Treasury Department.
However, the researchers do not attribute the disparity in audit rates to racial bias on the part of IRS auditors.
"We have no reason to think that this is a result of intentional discrimination, as the IRS does not observe race and the vast majority of these audits are conducted by mail," said Daniel Ho, a law professor and director of Stanford's Regulation, Evaluation, and Governance Lab, in an email. "While we do not have access to IRS's exact selection protocols, we show that a focus on refundable credits can in fact drive this disparity. Our models show that if the IRS treated dollars evaded equally — regardless of whether from eligibility for refundable credits or underreporting of high income — the disparity would reduce substantially."
An IRS spokesman declined to comment and directed inquiries to the Treasury Department, which did not immediately respond to a request for comment.
The study bolsters earlier studies from Syracuse University's Transactional Records Access Clearinghouse finding that the IRS tends to audit more low-income taxpayers than upper-income families since it's easier to identify discrepancies in their tax filings when compared to information returns (
Treasury officials have pledged to use funds from the Inflation Reduction Act to step up audits on high-income taxpayers.
"Equitable enforcement of our tax laws is a top priority for the Administration, and resources provided by the Inflation Reduction Act will enable the IRS to upgrade technology and hire top talent to go after wealthy tax evaders," a Treasury spokesperson said in an email to
"Historic challenges and underfunding have led to audit rates for those at the top of the distribution decreasing more than the correspondence audits of those at the bottom in the last decade, which should change," Treasury deputy secretary Wally Adeyemo wrote in a letter last fall to then-IRS commissioner Charles Rettig, according to
However, that doesn't totally explain the racial disparity. The new study noted that the task of selecting which taxpayers to audit is largely an exercise in predicting which taxpayers have underreported tax obligations that an audit would uncover. It noted that modern machine learning methods offer the potential to enhance the efficiency of tax audits by improving their predictive accuracy, but a growing body of literature on algorithmic fairness warns that policies based upon such predictions could inadvertently reinforce disadvantages against historically marginalized groups such as Black taxpayers.
"Such concerns are particularly acute for tax audits, which can exacerbate financial strain for the lowest income taxpayers — whose tax refunds are typically frozen while an audit is in place — and can dissuade individuals from participating in safety net programs for which they qualify," said the study.
Lawmakers in Congress have taken notice after news of the study was first reported Tuesday by the Times. "For years, Ways and Means Democrats have raised alarms over audit rate disparities between low-income and wealthy taxpayers, and today's findings confirm that stark racial disparities exist as well," said Rep. Richard Neal, D-Massachusetts, the ranking Democrat on the House Ways and Means Committee. "This is unacceptable, but a consequence of algorithmic tools that exacerbate racial biases in our institutions. The Committee's Racial Equity Initiative has focused on biased algorithms in our health care system, and it's clear we must address the discrimination at the IRS. I look forward to working with Secretary Yellen to turn over a new page at IRS, spurred by our multiyear investments from the Inflation Reduction Act. All taxpayers deserve fair treatment, and Ways and Means Democrats won't stop until that's a reality."
The academic researchers partnered with the Treasury Department on the study, so that may eventually lead to a lower rate of bias in tax audit selection. However, the Treasury and the IRS are limited in what they can do because they don't ask for information about a taxpayer's race on their tax forms.
To circumvent what the researchers call the "selective labels problem," they leveraged nearly 72,000 audits of randomly selected taxpayers to investigate the effects of the IRS's audit selection policies. To address the problem of the missing race label, they used a technique known as Bayesian Improved First Name and Surname Geocoding, which can guess a taxpayer's race based on their full name and census block group from the U.S. Census.
They estimated that the audit rate for returns filed by Black taxpayers was between 0.81 and 1.34 percentage points higher than the audit rate for non-Black taxpayers. That disparity was substantial compared to the base audit rate of 0.54% for the overall U.S. population.
The results suggest some potential avenues through which the IRS may be able to alleviate racial audit disparities, according to the researchers, such as shifting from more classification- to regression-based prediction algorithms, equally prioritizing underreporting from refundable credits and other sources, and expending agency resources to accommodate auditing more complex EITC returns.
"At the same time, some of the factors we identify are shaped by forces outside the IRS's control," said the study. "For example, Congress determines the information reported to the IRS by third parties — which shapes the distribution of noncompliant taxpayers that classification models identify — the rules governing credit eligibility — which may contribute to more mistakes among Black taxpayers due to racial differences in family structure — and IRS funding levels — which shapes the ability of the agency to allocate resources to complex cases."
