Voter identification laws are cropping up around the country: 31 states had a voter identification requirement in the 2014 midterms, up from 14 states in 2000. These laws vary widely in the types of identification they accept, even in whether identification is required or merely requested. And many people don’t know whether they need identification to vote, or what type of identification to bring.

Opponents argue that these laws disproportionately impact minority voters, who are less likely to have required identification. Our new research in this month’s American Political Science Review shows that minorities face another hurdle: bias in the bureaucracy that implements these laws.

Roughly 8,000 local officials – county or municipal clerks and election boards – manage the nation’s election system. These officials train local poll workers, provide information, and interact with constituents with little immediate oversight from state officials.

Here is the problem: election officials themselves also appear to be biased against minority voters, and Latinos in particular. For example, poll workers are more likely to ask minority voters to show identification, including in states without voter identification laws.

Our experiment demonstrates another form of bias: how willingly election officials answer questions from voters.

In September 2012, we contacted essentially all local election officials with short e-mails asking questions about voting. These e-mails were identical except for the name of the sender. A randomly selected half of the officials in each state received e-mails from a Latino name like “Luis Rodriguez,” while the others received the same e-mails from a non-Latino white name like “Greg Walsh.” (A similar type of experiment has been used to measure discrimination in everything from housing markets to state legislatures.)

The e-mail asked a question about either what identification would be needed in the upcoming election, or whether the sender needed to have voted in the primary election to vote in the general election. Responses were kept anonymous, and the study’s design made it impossible to identify bias by any individual official. But it does allow us to observe overall bias in responses to Latinos.

The Latino and non-Latino e-mailers both received many helpful responses. But nationwide, Latino names received fewer responses than non-Latinos and were more likely to receive vague and uninformative information about identification requirements.

This tendency for Latino names to receive fewer responses was not significantly larger when we asked about voter identification. This tendency was similar in states with and without voter identification laws, suggesting that the bias extends beyond the voter identification issue alone.

This bias cannot be explained by the population of each official’s district or by how likely officials were to have real constituents with these names. Bias was also not explained by partisanship – Democratic and Republican officials, in blue states and red states, were all less helpful to Latino e-mailers.

However, we did find one group that did not discriminate against Latinos: officials from jurisdictions covered for “pre-clearance” and federal monitoring against discrimination under two provisions of the 1965 Voting Rights Act. In areas with these institutional mechanisms meant to guard against discrimination, officials didn’t discriminate. But one of these provisions was suspended by the Supreme Court in 2013.

Research shows that everyone from doctors and car salespeople to loan officers, even those who believe in fairness and equality, can be biased against minorities. Many local officials implementing new voter identification laws and managing our nation’s election system are no exception.

The implicit bias we found in our experiment means that the implementation of voter ID laws can affect minority voters differently from other voters, even if the laws weren’t meant to discriminate. Serious attempts to monitor against discrimination, like those in the now-defunct Section 5 of the Voting Rights Act, may be our best bet for ensuring that our electoral system remains fair to all voters as we prepare for elections next year.

Ariel White, Noah Nathan, and Julie Faller are PhD candidates in Government at Harvard University.