Tweedledum, Tweedledee: A Statistical Correlation of Ballot Challenges in the MN Senate Recount
The frivolousness of some of the ballot challenges issued by the Norm Coleman and Al Franken campaigns during the U.S. Senate race recount has been well documented – both through the media’s publication of many such ballot scans as well as critical public statements of the campaigns by Secretary of State Mark Ritchie.
The idea that either campaign was issuing challenges based on a particular set of standards throughout the recount was dismissed early in the process when, as each day passed, the number of challenges issued increased from each campaign in a pattern that can only suggest mimicry. (Rising from approximately 2.5 cumulative ballot challenges per 10,000 ballots on Day 1, to about 3 on Day 2, to about 7.5 on Day 3, to about 12.5 on Day 4 to about 18 on Day 5 to about 21.5 on Day 6 to about 36 on Day 7; ballot challenges increased by both campaigns on each day during this span).
The most likely reason for this mimicry – regardless of whether one campaign initiated the process of an elevated number of ballot challenges early on or not – is that neither campaign wanted to appear to have lost votes at the end of the recount. That would be viewed as momentum – even if an artificially concocted momentum – that might be leveraged in the media and public to influence the outcome at the margins in the Canvassing Board process (or at any future court challenges). An influx in Franken challenges of (legitimate) Coleman votes was thus met by a nearly proportional increase in Coleman challenges of (legitimate) Franken votes in kind (and vice-versa).
The net result was a total of 6,655 challenges that were almost evenly distributed between the two campaigns: 50.7 percent by Coleman and 49.3 percent by Franken. Is this a coincidence? Consider this:
Smart Politics conducted a bivariate correlation to measure the strength of the relationship between the number of Coleman challenges in each county and the number of Franken challenges. The results were nearly as close to 1:1 as you will find in statistics:
The correlation of Coleman challenges and Franken challenges in the 87 counties is .977, significant at the .001 level. This suggests an extremely strong and positive relationship: the more one candidate issued challenges in a particular county, the more the other candidate responded likewise.
On the one hand, it makes sense there would be a strong correlation: one would expect counties with larger vote totals to have a larger number of challenges by each campaign. But the degree of similarity was sometimes jaw dropping:
· In Sherburne County: 434 challenges for Coleman (49.7 percent), 440 for Franken (50.3 percent)
· In Ramsey County: 213 challenges for Coleman (51.4 percent), 201 for Franken (48.6 percent)
· In Dakota County: 356 challenges for Coleman (48.0 percent), 386 for Franken (52.0 percent)
Moreover, the two campaigns had the exact number of challenges in 11 of the state’s counties: Big Stone, Chisago, Lake of the Woods, Lincoln, Norman, Otter Trail, Polk, Red Lake, Redwood, Rice, and Wadena.
But the mimicry did not end with the recount.
After Ritchie’s plea to the two campaigns to withdraw frivolous challenges, Franken announced last Wednesday he was withdrawing 633 challenged ballots. The Coleman campaign then responded on Thursday with an announcement it was withdrawing 650 ballots – nearly an identical number as Franken’s (though, perhaps not surprisingly, just a slightly larger amount to project an image that his campaign is just a tad less frivolous than Franken’s).
For all the reasons stated above we should not read too much into the fact that Coleman’s post-recount, pre-Canvassing Board margin (192 votes) is nearly identical to his margin before the recount began (206 votes). That said, it should also not be surprising – barring some unusual developments in the Courts or by the Canvassing Board – that Coleman emerge the victor by a similar margin.
Great post on this undecided seat. It sounds like it’s politics as usual. The more one candidate pushed (in this case challenges votes) the more likely his opponent will push back. Classic.