Rafael Henrique/SOPA Images/Sipa USA via AP Images
Even though Black taxpayers make up about 20 percent of EITC claimants, they undergo more than 40 percent of the audits.
For some African American taxpayers, and especially people claiming the Earned Income Tax Credit, the stepped-up IRS taxpayer resources assistance could be the difference between an error-free return and a major stress-inducing IRS audit.
The Inflation Reduction Act delivered a landmark $80 billion boost to the chronically underfunded agency over the next decade, with $3 billion allocated to tax filing assistance and education. The Treasury Department has steered $426 million into IRS taxpayer services since February, which led to the hiring of more than 5,000 customer service representatives to take calls in the run-up to Tax Day, a long-overdue development after years of staffing cutbacks.
A recent Stanford University-Treasury Department research study highlighted sharp racial disparities in IRS audits. African Americans undergo at least three times as many audits as other taxpayers. In particular, households claiming the EITC tax break for low- and moderate-income taxpayers are audited more frequently than others—and even though Black taxpayers make up about 20 percent of EITC claimants, they undergo more than 40 percent of the audits.
These taxpayers can also run into more hurdles that can put them at risk for repeated IRS scrutiny. Accessing additional tax filing help, including phone consultations, in-person assistance centers, and free tax help from trained volunteers, could potentially keep some EITC filers out of what Kathleen Bryant and Chye-Ching Huang of the New York University Law Tax Law Center call the “doom loop” of closer IRS scrutiny. “These disparities are in large part sort of the downstream consequences of inequitable access to free and reliable and simple tax filing resources,” says Bryant.
Their new report offers a set of “next steps” and stresses that uncovering and addressing these racial disparities hinges on the IRS providing some visibility into its audit algorithms in order to understand possible biases. IRS chief Daniel Werfel has pledged to have a report on racial audit disparities issued within 60 days of taking the reins at the agency. He was confirmed as commissioner last week.
It is possible, according to Bryant and Huang, who also review the Stanford-Treasury conclusions, that over time the algorithms may “learn” to replicate existing problems—that is, to steer the Black EITC filers, the people least likely to be able to navigate the system, into a “doom loop” of correspondence audits. These audits, which require a taxpayer to furnish certain documents by mail, are less labor-intensive than in-person audits conducted in IRS offices or homes, which may have contributed to their appeal to the pre-IRA resource-starved agency.
EITC filers are disproportionately Black, and the rates for correspondence audits are higher for Black filers than for others filers. Given the complexity of IRS audits, combined with the difficulties trying to reach an understaffed agency by phone and by mail, not surprisingly, the IRS winds up penalizing these taxpayers for some type of noncompliance, including loss of EITC eligibility. But the question facing the agency now is what are the assumptions that have produced a system that disproportionately scrutinizes poor Black taxpayers.
The plain fact is that the rich are where the real unpaid tax lies—1 percent of filers create about 20 percent of the tax gap.
It’s reasonable to expect to get more tax payments from money spent on auditing, says Huang, but there are other outcomes to consider. “To the extent that the audits themselves are something that only some people can make their way through, even if they are completely compliant with their taxes, then you might start introducing bias in those ways,” says Huang.
The IRA allocated $15 million to report on the feasibility of a direct electronic IRS filing system; such efforts have been vociferously opposed by private tax preparers. EITC claimants often use “unenrolled” tax preparers, who lack the professional credentials required of typically more costly “enrolled” preparers such as attorneys, CPAs, and private-sector preparation firms, and cannot appear before the IRS to answer questions about the return. Overall, tax prep fees eat up between 13 and 22 percent of EITC benefits.
These unenrolled preparers also happen to make the most errors on EITC eligibility claims. In 2013, a federal judge ruled that the IRS, which had undertaken an initiative to regulate tax preparers, did not have the authority to regulate credentialing, which would have established education requirements and testing for tax preparers.
There is little known about the IRS-designed algorithms and other processes that parse out who gets audited. Using some machine-learning technologies, however, apparently helps the agency make quick work of some types of error detection—which raises questions about the possible use of data typically associated with race, such as ZIP codes, in audit algorithms.
Dr. Sanjiv M. Narayan, an artificial-intelligence and machine expert and professor at Stanford University School of Medicine, has explained machine-learning biases this way: “Bias in AI occurs when results cannot be generalized widely. We often think of bias resulting from preferences or exclusions in training data, but bias can also be introduced by how data is obtained, how algorithms are designed, and how AI outputs are interpreted.”
The Stanford-Treasury researchers did not have access to the IRS algorithms. Instead, they modeled sets of alternative audit selection policies to illustrate how audit algorithms might be reconfigured. The IRS does not collect data on the race of filers, so the Stanford-Treasury researchers collected certain public data, such as home addresses and first and last names, and then looked at public data on Black and other filers to estimate that Blacks are more likely to be audited than other people with similar household demographics.
Understanding the assumptions behind the IRS processes could help the agency focus more closely on the amounts of money unpaid by a taxpayer, which captures other filers, instead of the mere existence of tax underreporting, which tends to ensnare Black filers. And monitoring algorithms, according to Bryant and Huang, would require annual or periodic disclosure of disparities by race to track increases or decreases in racial disparities.
More taxpayer assistance for people who cannot afford private tax preparation services or tax attorneys to do their returns could possibly pull some low-income people out of the doom loop, but the real questions are how many Black EITC filers would be subject to continued scrutiny, and why the IRS expends its finite resources going after small-dollar Black taxpayers. The problem may rest with many policymakers’ biases around behavior—a presumption that people who claim EITC credits have a propensity for tax evasion—rather than the converse, that more low- and moderate-income people, particularly ones without access to good-quality tax preparation services, can make honest mistakes.
The phrase “tax gap” refers to the difference between estimated owed tax and what is actually paid. The plain fact is that the rich are where the real unpaid tax lies—1 percent of filers create about 20 percent of the tax gap. But most of them are not subject to the doom loop regime since there is no presumption of tax evasion, that is, criminal intent. Bryant and Huang note that the IRS and Treasury should devote more resources toward instances of tax noncompliance that shield higher-income tax filers from audits rather than simple taxpayer errors, and begin to address the complexities of a racially biased regulatory regime.
