Voter Fraud in New Hampshire?

There are multiple online news sources, from obscure to heavyweight, reporting the possibility of voter fraud in the New Hamster Hampshire primaries this past Tuesday. But let’s keep our shirts on (especially you; yes, you), and think rationally about this. I’m going to be teaching my intro stats class next week, so bear with me while I warm up the stats brain.

First, what are the data that suggest the possibility of voting irregularity? I’ve seen 2 items:

1. Clinton beat Obama in the NH precincts where votes were cast using Diebold ^TM voting machines, but nearly the exact opposite happened in precincts where votes were hand-cast¹.
2. In the GOP primary, at least a couple dozen Ron Paul votes were uncounted in Diebold^TM precincts, as well (after announcements that he received zero votes, several individuals said they had voted for him, after which the voting-management types said, “Whoops! So you did!”).

Let’s focus on #1. The allegations of irregularities are fueled into the flame of near-accusation by longtime knowledge that the Diebold^TM machines are easily hacked, and by the fact that the hand-voting results are much, much closer to the exit poll. Let’s look at those numbers:

By itself, this is nothing more than a coincidence. However, it’s the kind of coincidence Columbo would have jumped on with both feet. The differences are like 5%, but with an N of more than 200,000, numbers like these are extremely unlikely to have happened purely by chance². As long as there’s no difference in the distribution of hand-voting vs. computer-voting precincts in NH.

But there’s the double rub³. “Extremely unlikely” is not “impossible.” Even more important is the distribution question. If the hardcopy votes are somehow not a representative sample of all NH votes, then it’s quite plausible that no irregularity occurred.

What if affluent, suburban precincts had computer voting machines, while poorer, urban precincts had paper ballots? Maybe all the White Soccer Moms voted with computers, and all the ethnic minorities and disaffected humanities majors and computer scientists voted by hand. I’m not saying this actually happened; but things like this have happened regularly in the history of American politics.

In fact, I think I’ll copy this post to my stats-teaching files. It will make lovely fodder for an introductory discussion of sampling. :D

Statistical sleuthing aside, I think the larger issue involves the use of the electronic voting machines. The paper votes are the only ones available for a recount. Cause if the computer voting machines were hacked, then re-counting the hacked votes doesn’t really do much good. The “paper trail” is as compromised as the final result.

1. Note: tallied by optical scanners, from a paper ballot [↩]
2. Anyone wanna dare me to do the chi-square test and calculate the p-value? Huh? Huh? Chicken? [↩]
3. Mmmm. double rub. [↩]