Assessing partisan bias in voting technology: The case of the 2004 New Hampshire recount

Abstract

Following allegations that Accuvote optical scan machines used in New Hampshire during the 2004 presidential election produced unusually low vote totals for Democratic candidate John Kerry, third party candidate Ralph Nader requested and funded a hand recount of ballots cast in eleven New Hampshire precincts. Using statistical methods well-suited for identifying election irregularities, we find no evidence of systematic biases among New Hampshire's Accuvote machines. Nor do we find evidence of other technology-related tabulation problems in the state. Our findings explain why the New Hampshire presidential recount did not substantiate alleged Accuvote discrepancies, and indeed it recovered more votes for George W. Bush than it did for Kerry. More generally, our analysis demonstrates methods that can help avoid false allegations about vote fraud while enabling concerned citizens, election administrators, and researchers to find and remedy real election irregularities.