Update: While counties provide a fair level of resolution, the precinct data could provide an additional level of detail. This time around I used the precinct data, from the MN Secretary of State's Office, for both the Senate and Presidential elections. The resulting calculation was performed identically to the method described in the following article, just with additional data.
More after the jump...
The precinct calculation yielded 17,831 undervotes in precincts won by Obama, while the county result ranged roughly 500 votes higher at 18,328. This deviation initially appears worrisome, for Franken, but the final tally shows a staggering drop off in the number of undervotes necessary for a Franken victory. An undervote total of just 3,280 votes, evenly distributed across each precinct and inline with the current trend, delivers Franken the votes necessary to overcome his 206 vote deficit. These 3,280 votes represent just 13.2% of all potential undervotes, as identified in the preceding article.
Given this result, I have no idea if 13.2% exists in the realm of likelihood, but its better news, for Franken, than our previously determined 40% threshold. Tomorrow (or the next day) I plan on exploring the historical correlation between overvotes, undervotes, and the Presidential to Senate ratio; I'll report back with my findings.
-------------------End Update-------------------
With each passing day, Norm Coleman's lead, currently 206, slips further into obscurity as the final canvassing concludes. No amount of crying, here, here and here, by Norm Coleman will change the fact that the final canvassing total often differs from the election night tally; as evidenced in a document released by the Franken campaign. If you want further evidence, a quick glance to our previous post, illustrates a perceived inequality with relation to the expected trend and the actual result.
4:21 AM Proj Actual
Diff.
Franken 80,600 92,990 88,620 -4,370
Barkley 20,550 23,415 22,652 -914
Coleman 55,840 63,650 61,513 -2,137
Drawing from the table, Coleman may continue to bleed votes as these four northern counties continue to finalize their vote tally in preparation for the automatic recount. But even if Franken cannot close the 206 vote advantage Norm Coleman currently holds, Franken may be able to make up ground in the undervotes. Undervotes occur when the ballot reading optical scanner fails to record a vote in a given race, in this case the Senate race.
Using data provided by the Minnesota Secretary of State for the presidential and senate elections, I tabulated a massive grid with data from each of Minnesota's 87 counties.
I took this data and calculated the discrepancy between Presidential votes and Senate votes, it turned out to be 24,830; in other words, 24,830 people cast a vote for President, but did not cast a machine defined vote in the Senatorial election. This number represents the absolute maximum number of undervotes that could possibly exist (technically it could be more, but its statistically improbable); in reality it is likely fewer. There are three scenario's that could result in a vote being marked as an undervote by the optical scanner; perhaps the person legitimately did not vote for a Senate candidate, there may have been a random malfunction or perhaps the voter failed to adequately mark their Minnesota ballot as depicted below:
The three above instances clearly demonstrate the intent to vote, but would still none the less be classified as undervotes. It is instance which could very well allow Franken to surpass Coleman upon the conclusion of these undervotes in the impending recount.
Using the previously collated data table, I concluded that there is a maximum of 18,328 undervotes present in counties won by Obama, but this isn't news, the Associated Press beat me to it, but I still have more to offer. This tilt towards Obama translates into a rather larger lead for Franken in counties that have a high number of undervotes. If a uniform distribution of undervotes exists in each of Minnesota's 87 counties and the current trend holds in each county, Franken could pull ahead if just slightly less than 40% of the potential undervotes are infact undervotes. Franken will likely win the recount if the trends hold and there are 9,848 undervotes evenly distributed across the state.
Donate to Al Franken to help with recount efforts.