Design Deficiencies and Lost Votes
In 2010, tens of thousands of votes in New York did not count due to overvotes — the invalid selection of more than one candidate. This report demonstrates how the lack of adequate overvote protections disproportionately affected the state's poorest communities, suggests commonsense reforms, and examines national implications.
In 2010, New Yorkers voted on electronic, optical-scan voting machines for the first time. Citizens went to their polling places on Election Day, filled out paper ballots and fed them into the brand-new optical scan machines.
But tens of thousands of their votes did not count. Specifically, about 20,000 voters in New York State did not have their votes for governor counted because the machines read their choices as “overvotes” — the invalid selection of more than one candidate. Even more votes were lost in other contests — 30,000 to 40,000 more. In a presidential year, with nearly twice the turnout, we expect that the number of votes lost because of overvoting would more than double, possibly resulting in more than 100,000 lost votes.
In modern history, New York has never seen so many lost votes due to overvoting. Unlike the new optical scan voting system, New York’s old lever machines did not allow overvoting. But even so, the numbers of lost votes due to overvoting in 2010 were far greater than they should have been. Overvotes are almost always unintentional. A well-functioning voting system, even one that includes optical scan equipment, should have overvote rates very close to zero.
A great irony of this new problem is that the federal mandate to purchase new machines was specifically meant to reduce overvotes nationwide. Tens of thousands of votes were voided as overvotes in 2000, in places like Florida on punch card and other voting systems. The Help America Vote Act (“HAVA”), passed by Congress in 2002, requires that new voting systems used in polling places in United States must:
(i) notify the voter [when she] has selected more than one candidate for a single office on the ballot;
(ii) notify the voter before the ballot is cast and counted of the effect of casting multiple votes for the office; and
(iii) provide the voter with the opportunity to correct the ballot before the ballot is cast and counted.
The Brennan Center, NAACP New York State Conference and other civil rights and good government groups have argued that New York’s overvote protections did not satisfy these requirements and predicted in a lawsuit filed in 2010 that these inadequate protections would lead to such high overvote rates. Specifically, they pointed to a message voters would see if the machine could not discern the voter’s intent; the groups argued this message would confuse voters, making it more likely they would cast invalid votes, and less likely that they would correct their ballots to ensure they were accurately counted.
As we demonstrate in this paper, the lack of adequate overvote protections had a disproportionately negative impact on the state’s poorest communities. Lost votes due to overvoting occurred far more frequently in areas with higher populations of low-income residents, people of color, and immigrants.
Black and Hispanic voters were at least twice as likely to lose votes due to overvoting as non-Hispanic whites. Shockingly, in two Bronx election districts, nearly 40 percent of the votes cast for governor were voided as overvotes.
The good news is that the New York State Board of Elections has agreed to adopt a better overvote warning when a voting machine cannot discern voter intent, hopefully in time for the November 2012 election: such a warning will inform the voter of the problem in plain English (“you have filled in too many ovals”), and clearly explain the consequences of casting an overvote (“your vote will not count”).
This should significantly reduce the number of overvotes in 2012, but it will not eliminate the problem. There is more that our public officials, and especially our state legislators, could do. In this report, we discuss how commonsense solutions, like requiring boards of elections to publish precinct-level election results, can improve detection and correction of machine-related problems. Critically, we also explore how better ballot design requirements can reduce overvotes.
Finally, we examine the national implications of our findings in New York.