The StarTribune last week posted a note suggesting that John Lott's analysis (noted here) did not add up. It had entirely different numbers than John's analysis had had. I assumed John had seen it, and indeed he did. It turns out the data included in the StarTribune report held information on two precincts (Richfield W3 P1 and Alexandria W1 P2) that had not reported any presidential results at all. The StarTribune analysis tried to include these data to show that the changes in the Franken-Coleman race and in the Obama-McCain race were comparable. Having established that the Richfield and Alexandria entries were the source of the difference, John writes (reprinted here with his permission):

[M]y piece makes it very clear that I was writing about corrections of typos, not slightly late reported results. The point of these numbers was to measure the corrections made that could possibly have been due to fraud. That does not include counties that simply reported their results very slightly after the time you set as the cut off.

My son, who is working at Fox News this fall, called the auditor of Douglass county, where the Alexandria precinct is located. The auditor said that there had been a network malfunction and so the numbers were not sent until Wednesday morning. The auditor said he thought this change was very different from those that other counties had in correcting typos.

So, the fact is that among precincts that had successfully reported results, Obama gained just 106 votes. Franken gained 435 votes. Those are the relevant numbers for trying to get an idea if any fraud may have occurred.

If you wanted to evaluate something different from what I did, fine, make it clear. But don't accuse someone of getting it wrong and imply that you have no idea where the numbers came from. Explain your reasons for your approach and why it is better. Assuming that you didn't understand what I had written and didn't see that my piece was talking about correcting typos, I suggest that you try to contact me or other authors when that happens.

I trust that you all will now correct your piece.

Nobody believes that zeroes were the right entries for those two precincts for president. But the StarTribune has run a confusing story here with very little thought put into it and in the process allowed some "editor for computer-assisted reporting" to engage Dr. Lott claiming "[h]is numbers are simply wrong." The editors do not come off better for it. I hope John isn't waiting for a correction though. They aren't too good at that sort of thing.

As I stated last week, it's not a question for statistics to answer if three typos are the stuff fraud is made of. To be blunt, we just don't know and you wish you had better reporting. I had one correspondent (who has asked to remain anonymous) reported to me that rural precincts in Saint Louis County faxed rather than called in their results. That would make some sense, as a fax leaves a trail that can be verified. Why isn't this sort of thing being reported? How about if the StarTribune did that rather than passing off interoffice email musings as editorials?

We are expected to take these typos on faith in the good behavior of election officials. Most of them are good people; I know one, and his life this month has been a living hell. I hope that both the Coleman and Franken campaigns can remember to treat those officials as human beings. We invest in an automatic system of reporting in part to not have aspersions cast on these good people. As I told that election official, the one thing you want is assurance that, at the end of the day, we all agree that we got the best count we could get. It would be nice if the media would help.

Labels: elections, Minnesota