Statistical Tools For Detecting Electoral Fraud

RockDoctor writes "A recent paper published in PNAS describes statistical techniques for clearly displaying the presence of two types of electoral fraud (PDF) — 'incremental fraud' (stuffing of ballot boxes containing genuine votes with ballots for the winning party) and 'extreme fraud' (reporting completely contrived numbers, typically 100% turnout for a vote-counting region, with 100% voting for the winning party). While the techniques would require skill with statistical software to apply in real time, the graphs produced in the paper provide tools for the interested non-statistician to monitor an election 'live.' Examples are discussed with both 'normal' elections, fraud by the techniques mentioned, and cases of genuine voter inhomogeneity. Other types of fraud, such as gerrymandering and inhibiting the registration of minority voters, are not considered."

Comments

Re:Voters' intent

[artor3]

Franken won by ~225 votes. Most districts only had 10 or 20 ballots that Coleman was disputing. Is your theory that the omniscient democrats launched a state-wide conspiracy to reject a dozen Coleman votes here, two dozen there, somehow knowing that that would tip the scales?

No, you are just parroting what you have heard from your chosen masters, who have been having a temper tantrum for the better part of four years.

Re:Frad Detection Algorythm

[mwvdlee]

So basically, if it's like a typical Florida election but Democrats win, it's fraud?