Margin Of Error: Breaking down the polls

Margin Of Error: Breaking down the polls

Data on non-citizen voters offers questions, not answers

Posted November 3, 2014

Logo for the Margin of Error blog.

— Concerns about non-citizens voting are nothing new. As WRAL News noted back in 2012, hard evidence that non-citizens are voting in North Carolina is rare, despite fears to the contrary.

As part of these fears, the State Board of Elections recently completed an audit of over 10,000 registered voters. A voter watchdog group, Voter Integrity, claimed they were non-citizens. Although the audit failed to uncover past voting by non-citizens, it suggested that about 100 currently registered voters might be non-citizens.

I am writing about this issue because a recent survey found that about 14 percent of non-citizens said they were registered to vote. As reported by two political scientists on the blog, "The Monkey Cage," voting by non-citizens appears to be substantial. They estimated between 2  and 6 percent of all non-citizens voted in 2008 or 2010 and inferred non-citizens might be affecting election outcomes.

I am skeptical about that research, and here’s why. It depends on answers to survey questions from the Cooperative Congressional Election Study. This survey was designed to measure political attitudes and behaviors of citizens, not non-citizens. Thus, very few non-citizens are included in the incredibly large samples. While most surveys of the U.S. include around 1,000 adults, these included over 30,000 and 50,000 adults, depending on the year. Since non-citizens make up just 1 percent of these samples, though, I doubt they are representative of non-citizens.

The goal of drawing a sample is to have this smaller group of people look exactly like the larger population they are meant to represent. Non-citizens who think it is worthwhile to take surveys about U.S. politics, though, are likely more politically active and attentive than their counterparts who are not part of the study. Rather than recruiting respondents via probability methods – in which each person had an equal chance of being asked to be a participant in the study – the CCES allows respondents to opt-in at their own discretion.

This method is problematic for the same reasons that surveys on web pages do not generate valid results without adjusting for known population parameters such as age, gender, education, etc.

Second, the controversy is based on a question that is subject to measurement error. There are other types of survey error besides the margin of sampling error. Respondents were asked, "Are you registered to vote?" As two different political scientists noted in a subsequent Monkey Cage post, that is not the same thing as asking, "Are you registered to vote in the United States?"

It is likely that many non-citizens who are registered to vote in their home country, but not in the U.S., answered yes. Indeed, of the 75 non-citizens in the 2008 survey who said they were registered to vote, 15 were verified as not being registered at all. Another 46 could not be matched to actual voter rolls and verified one way or another. As a result, only 14 of the 75 could be verified as registered.

Third, these claims about voting by non-citizens are being made based on incredibly small absolute sample sizes. In 2010, for example, just 13 out of 55,400 respondents claimed to be non-citizens and voted. I think it is a mistake to heed claims based on unrepresentative samples whose sizes are equivalent to rounding errors. To be sure, some non-citizens have surely voted, but these data are ill suited for confidently estimating the exact percentage who do.

In terms of policy implications of illegal voting by non-citizens, two things I've read stand out.

One, as the second Monkey Cage post suggested, it is easier to make sure votes of non-citizens don’t affect election outcomes by making it easier for actual citizens to vote. If so, illegal votes would be an even smaller share of the total vote and have less of a statistical ability to tip the balance of an election. Second, instead of requiring voters to produce state identification that doesn't preclude being a non-citizen, a superior requirement would be mandating a national identity based on verified citizenship status. Political realities, however, make the latter option unlikely, and recent laws in many states arguably make it more difficult, not less, for actual citizens to vote.


Please with your account to comment on this story. You also will need a Facebook account to comment.

Oldest First
View all