A Statistical Method for 2012 Election Vote Fraud Allegations


Home / A Statistical Method for 2012 Election Vote Fraud Allegations

How to Extract Evidence from Opinion Polls

“George Gallup, pioneer of opinion polling” : Image via Quibik and The Gallup Organization

Some evidence regarding election fraud may come from opinion polls, assuming the pollsters accurately sampled truly eligible voters, rather than using the possibly excessive voter registration list.

In a statistical analysis, the polling data should be confined to Wood County, Ohio. This pro forma analysis uses Gallup’s “reputation” as a national pollster, and their prediction for Ohio overall. The justification is that Wood County’s 2012 election result was not terribly different from Ohio, although many counties did indeed differ significantly.

Gallup’s statistical “reputation” is the standard deviation in predicting the percentage margin of victory: 3.12 across 20 elections.

In the lead-up to the U.S.A. election 2012, on Oct. 16, Gallup had reported that Obama had a 52-to-49% “favorable” rating lead over Romney, and both were tied at 46% “unfavorable.” Unfortunately, this poll sampled registered voters, and was not Gallup’s final prediction.

Let’s say that Gallup had predicted Obama to win Wood County Ohio by 1%. The official  election result was a 4.2% margin; the error is 3.1%. That is almost equal to the 3.12 standard deviation for Gallup in national elections.

“Gallup Rasmussen and Silver Poll Reliability 2012” : Image by Mike DeHaan

A statistician would not reject the null hypothesis if the disagreement between prediction and data is only one standard deviation.

Also, the corrected 10% margin of victory for Romney is farther from the Gallup prediction, and less supported by other polls. The difference would be 11/3.12 = 3.4 standard deviations. The usual “confidence interval” is limited to 2 standard deviations, so one would reject this alternative hypothesis. (In other words, the suggestion that over 8,000 people registered and voted falsely in Wood County does not bear up under statistical analysis.)

Wood County Election Results: A Sober Second Look

One should examine other evidence to support or contradict one statistical result.

Prior to the 2012 vote, Ohio was considered a “swing state” by many pollsters, predicting a close race. The state-wide result was indeed close.

However, some Ohio counties had large margins for Romney; so one could imagine the same in Wood County.

Historic data also helps. From “Election Summary Report 2008 Presidential General Election”: Obama had 52.61% versus 45.50% for John McCain, for a 7.11% margin, of 65,164 total votes cast by 101,891 registered voters.

In 2008, Obama received 34,285 votes versus 31,596 in 2012; total votes decreased also. These results do not suggest an injection of 8,045 excess votes for Obama in 2012.

Diligent readers may compare the Wood County margin of victory to that of Ohio and the entire United States for each election for which the data is available. Does the standard deviation suggest the 2012 result is consistent with past elections?

Leave a Comment