The Trump administration recently began implementing a “no-match” policy of flagging administrative records that don’t match what the Social Security Administration (SSA) has on file.

If you’ve been following the news in Georgia, you might think this is a reference to a tactic used by Georgia Secretary of State Brian Kemp to purge minorities from the state’s voter rolls right before the gubernatorial election he narrowly won against his Democratic opponent, Stacey Abrams. Not exactly, but both are related—and have civil rights implications.

Voter suppression efforts in Georgia and elsewhere have used a range of strategies to purge voters, including striking those whose names loosely matched those of dead people and felons and those whose names didn’t exactly match SSA or drivers’ license records (so much for consistency). In a legal challenge, civil rights groups noted that an estimated 80 percent of voters affected by the “exact match” policy were African American, Latino, or Asian American.

Now the Trump administration is reviving a failed policy of sending letters to employers informing them of apparent discrepancies between employee W-2 forms and SSA records. In this case, the ostensible goal is checking tax forms, not voter registrations, but both efforts use SSA data as a validity check and both disproportionately impact immigrants and people of color.

SSA’s stated purpose is ensuring that paperwork is in order so workers receive the Social Security benefits they’ve earned. But there are better ways of doing this—including by informing the workers themselves of any discrepancies. In 2001, when SSA sent letters to both workers and employers, the former was nearly three times as effective in prompting corrections.

Employers can already electronically check W-2s against SSA data to make any necessary corrections before the forms are sent out. So what’s the harm in contacting employers by mail? At best, mass mailing no-match letters will add to the already-long wait times at Social Security field offices as people scramble to add missing hyphens or update their married names. At worst, it will lead to many employers wrongly suspecting that some of their employees don’t have work authorization. And while John Doe is unlikely to lose his job as a result, Juan Dominguez might, even if they are both U.S. citizens.

This is hardly an unforeseen problem. A version of the no-match policy on steroids was suspended by a judge in 2007 who determined that it would “result in the termination of employment to lawfully employed workers… because, as the government recognizes, the no-match letters are based on SSA records that include numerous errors.” Unlike the current policy, the Department of Homeland Security under George W. Bush put forward a regulation that would have required employers to fire workers on suspicion of lacking work authorization if they weren’t able to resolve discrepancies within 90 days. The Bush policy was opposed by a broad range of groups, including the AFL-CIO and the U.S. Chamber of Commerce, and was ultimately rescinded by the Obama administration.

Current no-match letters do warn employers against taking action against workers based on what in most cases is an innocent oversight or clerical error on the part of the employee, the employer, or SSA. But this disclaimer won’t dissuade employers inclined to be suspicious of foreign-born workers. No-match letters can also be used as a union-busting tool—by giving employers a convenient excuse to fire employees who support or engage in efforts to unionize, as happened at a Smithfield meatpacking plant, a Cintas laundry, and other workplaces before the earlier no-match policy was suspended. They can also be used to threaten workers who speak up about unpaid wages or unsafe working conditions.

Immigrants are far from the only group who’ll be adversely affected. Women who change their names after marriage or divorce, people with composite or hyphenated names, and people with less common names that are liable to be misspelled are more likely to be flagged, as are elderly and less-educated workers who are more likely to use nicknames. Low-income workers are also more likely to change jobs often, increasing the scope for error.

But immigrants are at highest risk, not only because they’re more likely to be suspected of not having work authorization, but because many of the above factors disproportionately apply to them. Social Security cards issued as part of the immigration process are more likely to contain errors than those issued at birth. Immigrants are more likely to have uncommon names or use different naming conventions, such as using compound names or listing surnames first. They’re also more likely to lack high-school educations and to have low-paying, transient jobs. Finally, in most parts of the world, dates are listed in a different order than in the United States, leading to inconsistently recorded birth dates.

Discrepancies between SSA records and other documents may have been caused by applicants, employers, or government agencies—and often by SSA itself. In a 2009 review, Social Security’s Office of the Inspector General found that SSA’s Help America Vote Verification (HAVV) program, in which SSA partnered with states to check voter registrations, had a much higher no-match rate than other programs used by states and employers to verify voter registration or tax records. The HAVV program had a 31 percent no-match rate, while other programs had no-match rates ranging from 6 to 15 percent. The Inspector General found that the HAVV program didn’t even provide consistent responses when the same applicant data were resubmitted! In other words, the evidence is overwhelming that SSA records should not be used for purposes that they were never intended for.

Which brings us back to voter purges. In a partial about-face, Georgia Governor Brian Kemp signed a bill into law last week that, among other things, would no longer automatically strike voters from the rolls based on differences between voter registrations and SSA records. Despite this positive development, the law was ultimately opposed by Democrats because of security concerns regarding the use of electronic voting machines. In any case, Georgia has a history of reintroducing voter suppression tactics with minor revisions over U.S. Department of Justice objections, so history suggests that we’ll see another attempt at a voter purge in Georgia that’s barely disguised as “administrative recordkeeping.” Like SSA no-match letters, it’s a bad idea that never seems to go away.