Weekend Reads | Racial Disparities in IRS Audits

by Kevin Schofield

This weekend's read is a report from the Stanford Institute for Economic Policy Research looking at racial disparities in whom the Internal Revenue Service (IRS) decides to audit.

According to the IRS, overall, 0.54% of tax returns are audited, roughly 1 in 200. That number has been dropping in recent years, often because of reduced staffing. The agency claims that the method it uses to select tax returns to audit is race-blind, and, in fact, it doesn't even collect information on taxpayers' race. But is it really race-blind in practice? This matters, because an audit can be a significant hardship on a taxpayer: On top of the time and energy that go into preparing for and dealing with an audit, auditors routinely freeze tax refunds until after a pending audit is completed — so the IRS holds on to your money until it decides to let it go. Further, audits can create disincentives to enrolling in aid programs if doing so is perceived to increase your likelihood of being selected for an audit.

The IRS' audit-selection mechanism, in theory, tries to predict which taxpayers have a higher likelihood of underreporting the amount of taxes they owe. That alone isn't race-specific, and a taxpayer's race isn't one of the pieces of data fed into the algorithm, so it's hard to argue that the agency's selection process intends to target particular races. But the courts have recognized that discrimination can take two forms: It can be based on intent to discriminate, or it can be a "disparate impact," meaning that, regardless of intent, the impact on certain individuals is different from others in a way that consistently and measurably disadvantages them. So even a "race-blind" algorithm can be discriminatory if the result of using it creates disparities that disadvantage people of a specific race.

Since the IRS doesn't collect data on taxpayers' race, it doesn't have an easy way of identifying whether its audit selection process is discriminatory — nor did the Stanford researchers. But they do have two useful pieces of information from tax returns: the taxpayer's name, and their address. It turns out that a combination of a person's name and their voting precinct (which can be determined by their address) can be used to predict a person's race with high accuracy; the researchers used this, along with (at the IRS' invitation) lists of people who have been selected for tax audits, to build a model of the demographics of who gets audited.

They found that Black people get audited more frequently than people of other races — over the past decade, around three times more frequently.

A high-level analysis points to one factor: the Earned Income Tax Credit (EITC). Claiming the EITC makes it significantly more likely a tax return will get audited: For example, in 2014, 1.45% of returns claiming the EITC were audited, versus 0.31% of returns where the EITC was not claimed.

A skeptic could argue that the racial disparity is simply because more Black people claim the EITC (eligibility is based upon income and number of dependents). And while that is true, the researchers found that only a small portion of the racial disparity is explained by that. It turns out that even among the taxpayers who claim the EITC, Black people are audited much more often.

And it turns out that the racial disparities are even more pronounced when looking at further subgroups who claimed the EITC: For example, among single males with dependents, 9.11% of Black people were audited compared with 3.05% of non-Black people. This proves the disparities aren't because of income or family composition; when controlling for those factors, the racial disparities still exist.

So given that Black EITC claimants are audited at a higher rate, we must ask: Is it warranted? Do those Black taxpayers underreport the amount of taxes they owe at a higher rate than their non-Black counterparts? According to the researchers, the answer is yes — but not at a substantially higher rate, and certainly not enough to warrant the significant racial disparity in who gets audited.

The auditors then look at a related question: What is the goal of the IRS' audit selection process? Is it trying to maximize the likelihood that a particular return underreports taxes, to maximize the total dollar amount underreported, or to focus more attention on faulty claims for refundable tax credits like the EITC? They ran models of each of these and found that the models focused on maximizing the number of underreported returns and claims for refundable tax credits produced higher audit rates of Black taxpayers, but the model prioritizing the total dollar amount underreported returned a lower rate of Black taxpayer audits.

To sum up: Researchers found that the IRS' process for selecting tax returns to be audited doesn't explicitly target Black people, but because it seems to prioritize maximizing the overall number of returns that underreport, the likelihood of underreporting, and tax credits like the EITC normally used by poor people — rather than trying to maximize the amount of money the agency could recover through auditing returns — the result is a disparate impact that targets Black people with more audits.

The report concludes by pointing out that these are policy decisions — including how the federal government wants to allocate its auditors toward different classes of tax returns — and that it's well within the IRS' ability to make different policy decisions that lead to a more equitable outcome.

Measuring and Mitigating Racial Disparities in Tax Audits

Featured image by Andrey_Popov/Shutterstock.com; edits by the Emerald Team.