Friday, January 11, 2008

Were the polls really wrong about Obama in New Hampshire?

In my experience, polls are disturbingly accurate. My first thought about the early call in New Hampshire for McCain and the lack of a similar call on the Democratic side was that the early results in the latter were in clear conflict with the exit poll information. And when that happens, fraud must be considered. Ben Moseley writes:
I ... put together a spreadsheet of the Democratic results of the NH primary for each town with almost all but a few towns reporting ...

What the informal statistics show is that Hillary Clinton received a 4.5% boost [from the poll numbers] in towns using Diebold counting machines compared with towns that didn't. Cnversely, Obama ... showed a 2.5% decrease [from the poll numbers] in the Diebold towns. ... [Boost and decrease from what?]

The possibility of election fraud is important to consider because of the predictions heading into NH primaries. All the polls were showing Obama with at least a 7 point lead over Clinton, with a few showing a double-digit lead, which is no surprise considering Obama's win in Iowa over Clinton, who placed third in the caucuses.

Update I: Some more statistics from the data show that Obama in non-Diebold towns garnered 38.7% of the vote to Clinton's 36.2%. The results in Diebold towns show the exact opposite: Clinton with 40.7% of the vote and Obama with 36.2%. Not only are the positions swapped but the informal statistics have the second place candidate holding 36.2% in both cases, which may or may not be purely coincidence. ...

Update II: Another thing to notice is that the Diebold machines returned a 7-point difference (+4.5 for Clinton, -2.5 for Obama) which is exactly what the polls had been predicting [except in favor of Obama].
Complete data are available at An analysis showing that the Diebold difference could in fact simply be an urban/rural difference is available at (although that does not answer why the vote differed so from the polls.

And anyone who notes that it's Republican operatives that control the Diebold machines need only remember that it is very much in the interest of the Republicans to have Clinton as the Democratic nominee. She is not only quite defeatable as a divisive candidate with a ton of baggage, she is also pretty much Republican anyway.