The Minnesota Recount

As a very hectic week settles down a bit, I can give you a little more information and perspective on the Minnesota US Senate Race recount. There are a number of misconceptions circulating about this process that I can dispel, and I have a pair of predictions for you: Taking the same exact data, we have the Democratic Party Line and the republican Party Line, wherein "line" means trend line on a graph.

[Update: See this new analysis suggesting that the Republicans cheated]

Let's start with the prediction. Given all the available data, we can now estimate what is going to happen as a direct result of recounting all the votes in this race, not counting changes that may occur because of challenged ballots or previously denied absentee ballots. Look at the chart:

Here's how to read this chart. The X-axis is number of votes that were cast for either Franken or Coleman counted. To date, as of Friday PM, about one and a half million of these votes were counted, and there are just under two and a half million votes in total. Note that this does not include votes that were cast for other candidates, no candidate, or votes that are thrown out (all of which together are officially termed "other").

The Y-axis is the shift from the pre-recount gap of 215 votes. The zero line on the Y axis is negative 215 votes, or, the difference at which this is a tied race.

Only three days of counting are shown here, those being the first three days. I'll make a new graph at the end of today, Saturday. (Yes, there is some counting being done today).

I've drawn two trend lines on this graph, both of which are really just there for fun. You know that two points define a line, and three points define a statistical analysis! Here, the blue line is a standard least squares regression line using the three dots to estimate the outcome. This is what I call the Blue Line, or the Democratic Party Line. This shows that when the last votes are recounted, the gap between Coleman and Franken is going to be about 10 votes, with Coleman still winning.

The Red Line, also known as the Republican Party Line, is drawn using a second order polynomial equation. As you know, a second order polynomial equation estimated from three points will join the three points perfectly to the line and produce some kind of curve. In this case, we have a trend line that is actually rather off the charts. The Republican Party Line (which we shall also refer to as Norm Coleman's Latest Wet Dream) predicts that at the end of the counting, the Coleman-Franken gap will grow to approximately 500 votes.

The Real Line, which I have not drawn here but may after more data are available, could be a line that starts out steeply and then get less steep, and ends up with a gap of something like fifty votes or so. The idea here, which I do not entirely buy (but I'm analyzing this and will report back later) is that there will be more of a shift towards Franken in the more Franken-friendly areas of the state. This may be true, but these shifts may also be the result of the nature of the voting machines in each area, or other non-partisan factors. The fact that the rate of shift has diminished as Franken friendly areas were finished and there was a shift to Coleman-friendly areas has been suggested, but not demonstrated. Stay tuned to this blog, as I will be testing this hypothesis before lunch today!

On to the misconceptions: I just want to address two related ones at this point. One is that the challenges being issued by the Franken and Coleman camps have been frivolous. The other is that these challenges are changing the vote count.

In order to understand these issues, you need to know a little about how the process works.

A small number (two, as far as I know, but this may vary) election officials take all the ballots out of one or more boxes, from a single precinct, and divide them into three piles: Franken votes, Coleman votes, and Other votes. Then they count the number in each pile. This number is then compared to the machine count. If the numbers are different, some searching may be done to see if this difference can be explained. For instance, if you go back through all the votes for the candidate who has more votes than the machine said, you may find a ballot with a clear vote for that candidate (so the clerk put the ballot in that candidate's pile) and a single dot in one of the other bubbles. So, it is a reasonable guess that the machine counted this as an over vote (i.e., "other") instead of for the candidate.

That ballot, which has effectively moved from the "other" to a candidate's pile now may or may not be challenged by the side that effectively lost a vote in this case. If it is not, that means a simple change in the recount. If it is challenged, it means no change in the recount (of the two candidates), but rather, a reduction in the 'other' category.

In total there are five hypothetical piles of ballots as these votes are being counted. Coleman, Franken, Other and two challenge piles. In truth, there are so few challenges that there are no such piles. What really happens, is this: If there is a challnge (or two or three) in a given precinct, the lawyers come over, the political reps come over, and the head of the election team (county official) comes over, and they talk about it. A challenge that is really off will typically be refused by the election official, but even a liberally interpreted challenge made by one side or the other ill be accepted. The problem here is that no one knows at this time what the standard that the state is going to use, or that a court may decide, will be. If a court, later on, decides that a filled in bubble and a dot in another bubble is always an over vote, but if one side in this struggle habitually makes such votes as challenges but the other side does not, then the process is not fair.

So, the two misconceptions:

1) That the challenges being issued by both sides are changing the vote count. Not really. The challenges are adding to an ambiguous pile that will be sorted out later. The vote count changes you are seeing here are mainly votes being removed or added because of bad counting by machines. The vote count changes that may occur because of challenges are not being considered yet. We don't know yet what effect that will have on the final number.

2) That the challenges are often frivolous. My understanding from my own observations and conversations with people working in two other counties is that this is not accurate. It is accurate to say that the challenges are somewhat liberally interpreted, but that is fair given what I said above.

Two other quick items: It is the case that the Franken campaign is asking for an added step: They would like to have both campaigns go through the challenges one more time, having seen and discussed the overall nature of the beast to date, to withdraw many of the challenges. If the Coleman campaign goes along with this, you'll see the number of challenges drop somewhat. Currently, I estimate that there will be a total of about two thousand challenges. Not bad considering the number of votes and what is at stake here. If the challenges are sorted through and some are pulled, I'd guess that this number would go down to about 400.

The other item is totally a rumor, so don't put much stock in it: In one vote recount location, the head election official had to chuck out the Coleman team head (or lawyer, not sure) because s/he was making too much trouble for everyone else. That is probably an oversimplification of what happened.

That is all for now. Anybody else out there counting votes????