Identifying Bias

Groundbreaking research by USC Dornsife’s Christian Grose, associate professor of political science, and doctoral candidate Matthew Mendez has shown that lawmakers who support voter ID laws are more likely to show racial bias against Latino constituents.

“We wanted to find out if we could detect bias among legislators toward certain groups of people affected by voter ID laws,” Mendez said. Such laws require registered voters to show government-issued ID, such as a driving license, before they can vote.

“We decided to focus on Latinos, because while there’s a lot of talk of Latinos being affected by voter ID laws, there’s substantially less about what legislators think of Latinos themselves in relation to these laws.”

Featured in The Washington Post and other publications, the study found strong evidence that “discriminatory intent underlies legislative support for voter identification laws.”

The findings also raise questions about the constitutionality of voter ID laws, which the Supreme Court affirmed in 2007.

Supporters argue that voter ID laws are necessary to combat voter fraud. Opponents point to research showing that voter fraud involving impersonation occurs so rarely it is almost negligible. Moreover, they argue, such laws disproportionately affect minority and low-income groups, as well as the elderly.

Voter ID laws are being passed in more and more states, Grose said.

“In 2007, the Supreme Court ruled that such laws are acceptable because there’s a compelling state interest to protect against voter fraud, even if there are discriminatory effects against, for example senior citizens, Latinos, or African Americans,” Grose noted. “But what is not legal is laws passed with the intent to discriminate against minorities. Since our study shows that legislators who support these laws do respond differentially to minorities, that raises some legal questions about the law that are potentially very big indeed.”

To test bias among state legislators, Grose and Mendez developed a pioneering field experiment. In the two weeks leading to the Nov. 4, 2012 general election, they sent e-mails to 1,871 state legislators in 14 states with the largest Latino populations in the U.S. The e-mails read as follows:

Hello (Representative/Senator NAME),

My name is (voter NAME) and I have heard a lot in the news lately about identification being required at the polls. I do not have a driver’s license. Can I still vote in November? Thank you for your help.

Sincerely,
(voter NAME)

Grose and Mendez sent one group of legislators the e-mail from a fictional voter they named “Jacob Smith.” The other group received it from fictional voter “Santiago Rodriguez.” In each group, half the legislators received e-mails written in Spanish, while half received e-mails in English.

The study was designed so none of the states included required driving licenses in order to vote. This meant legislators could theoretically have responded to the e-mails with a simple “yes.”

Grose and Mendez then measured how many legislators responded in each category.

The results showed that lawmakers who had supported voter ID requirements were much more likely to respond to “Jacob Smith” than to “Santiago Rodriguez,” thereby revealing a preference for responding to constituents with Anglophone names over constituents with Hispanic ones. They also showed legislators were more likely to respond to English than Spanish-language constituents.

Among voter ID supporters, the responsiveness to Latino constituents was dramatically lower than to Anglo constituents. Even within the Spanish language constituents’ requests, the Spanish speaker with an Anglo name was responded to nine percentage points more than a Spanish speaker with a Latino name. The latter received virtually no response from the voter ID supporters, with a response rate of just one percent.

Among both Republican and Democrat lawmakers who do not support voter ID, Spanish-language constituents with both Anglo and Spanish surnames received almost the same rate of response at around 12 percent, with no statistical difference in preference for the Anglophone or Hispanic name, Mendez said.

And while an Anglophone name bias could still be observed among non-voter ID supporters who received English-language e-mails, it was much smaller — a difference of 6.9 percentage points compared to a difference of 17.5 percentage points for the supporters of voter ID laws.

“I was shocked by the magnitude of the bias,” Mendez said. “I went in thinking we might find a gap of some kind, but I wasn’t prepared for an almost 20 percentage point gap among those who support voter ID laws based on the name of the ethnicity of the constituent.”

“Research shows everyone has biases,” Grose said. “However, in politics, we tend to believe that logically those biases should be overcome by an elected official’s interest in seeking out reelection. Thus one might imagine that legislators would respond to constituents without regard to ethnic and racial backgrounds because if every person is a potential voter there’s no reason not to do so. However, our study shows that is not the case.”

The results of the study also have implications for the quality of representation for Latino constituents.

“The same elites who propose and support legislation to restrict Latino voting rights also provide less non-policy responsiveness to Latino constituents, at least in the context examined here. This means that the quality of representation is poor for many Latino constituents,” the researchers wrote.

Grose and Mendez noted that while the majority of legislators who support voter ID are Republican, the majority of Republicans in the study sample did not support voter ID.

“The people who are not responding to Latinos and who favor voter ID are a numerical minority in the sample in general, and also a minority within the Republican party, and so the findings of our study are not driven by partisanship,” Grose said.

Grose has authored an article published last month in The Annual Review of Political Science arguing there should be more such studies conducted on elected officials.

“Considering legislators worthy of experimentation in this kind of study and also working with public officials to convince them to collaborate with us in similar experiments with their colleagues so we can learn about politics and governance is something we encourage in our political science and international relations Ph.D. program,” Grose said.

“We didn’t conduct this study to either support or oppose voter ID per se, but I would like to see the results being used by the legal and policy community.”