Interesting Analysis of Minnesota Senate Race Vote Totals
John Lott has an interesting analysis of recent corrections of "typos" in the vote totals in the Minnesota Senate race, found here. Its seems like the corrections have disproportionately favored Democrat Al Franken. Sounds like this is something that will bear watching.
However, given the small actual number of typographical errors (a hundred here, a hundred there - not a ballot here, a ballot there), we aren't even talking about that improbable a result even if Obama's numbers changed in a way quite dramatically from Franken's.
Either Lott failed to disclose something helpful in his analysis, is intentionally deceiving his audience, or made a high-school level goof. In any event, even if Lott's analysis was correct, it would hardly prove his point, because we aren't talking about a systematic undercount, but instead a handful of typographical errors.
And ALSO in any event, if there was the kind of fraud that Lott alleges in his original piece, then a recount would be the surest way to discover or discount it. But then what is the purpose of the original supposed fraud, as a recount was going to occur regardless?
"Looking at the precincts that had entered in a number and then changed it between Wednesday, Nov. 5, and Monday, Nov. 10, Obama picked up 106 votes. Adding together the changes for Richfield (908), Alexandria (107), and the corrected typos (106) yields 1,121 votes, exactly the number that the Star Tribune reported."
Wrong, of course. Once again, the difference between net votes and total votes is one that either eluded Lott, or one he intentionally ignores to try to trick his readers. It's also possible that the Minnesota Tribune was in error when it said net - but that is not an allegation Lott has made at this point.
If Lott is going to be part of the media, its important he gets it right, too.
Besides he explicitly said that this doesn't mean there was fraud.
Besides he explicitly said that this doesn't mean there was fraud.
All very true. So what is his point?
John Lott might file suit against you for defamation, just like he did with the guys who wrote Freakonomics. Just because he's a discredited crank who's been caught red-handed falsifying data doesn't mean he's harmless.
Based on what he posted on his site, I think that both the Strib and Professor Lott are partially correct and partially wrong on the numbers. Professor Lott is correct (and the Star Tribune is incorrect) in saying that Obama’s gain from the recount is not 1121 votes because some of them were not votes from a recount but rather the result of two districts that reported late.
However the Star Tribune appears to be correct in saying that the recount still benefitted Obama more than Franken who picked up 459 votes even after you correct for the two districts that reported late. From Professor Lott’s post:
By Lott’s numbers then, Obama netted 434 votes (908 minus 474) in Richfield W-3 P-01 and 107 in Alexandria W-1 P-2 (although I think that number’s wrong because Lott’s article doesn’t say how many votes were for McCain in that district would reduce Obama’s net). That means then that the maximum total of votes that Obama gained from the two late reporting districts (as opposed to from the recount) is 541 which if you subtract that from the Star Tribune’s reported 1121 gain, still leaves Obama with picking up 580 votes from the recount which is more than Franken’s 459 votes.
You've grossly misrepresented or misread Lott. Lott clearly wrote that Obama gained 106 votes over McCain as a result of precincts reporting "corrected" typos and a 908 + 107 vote advantage over McCain as a result of two precincts reporting late. The total gain to Obama was exactly 1,121 votes--the same number reported by the Tribune. It seems reasonable to assume that the Tribune did, in fact, conflate gains to Obama from both typo "corrections" and late reporting.
Lott's point is that we should focus on the "corrections", where there seems to be an extraordinarily high shift in favor of Franken. A shift that is much larger than that from corrected tallies in any other race in Minnesota. That seems disturbing and worthy of further investigation to me.
I wrote 908+107+106= 1,121, not that 908 + 107 = 1,121. My math was correct.
My piece several times makes clear about net vote gains. Note also that the Star Tribune can only get to its total of 1,121 by looking at net changes.
What spurred my conspiracy theory was this:
Not what you state.
Finally, the recount cannot resolve issues of fraud that might have occurred in the earlier ballot process. For example, if ballots were altered or added.
Fortunately, we trust the Secretary of State from MoveOn to save Al from Mean Old Norm. Al Franken '08 - the kind of serious leadership this nation needs!
As far as I can tell, the Minnesota Star Tribune didn't say that typographical errors gave Obama 1,121 net votes (ie, 1,121 more total gross votes than his previous total). The Star Tribune said that his net increase was 1,121. That is, that Obama netted 1,121 net votes.
As far as John's conspiracy point goes, typograhpical errors are too few in frequency to expect a systematic problem. Even if there weren't obvious reasons for the fact that Franken's differences exceeded a bunch of local races with fewer total votes and precincts, when there was little reason to be completely careful with the other races (none, afaik, being within the margin of error), we just aren't talking about a large enough number of staistical events for coincidence to be ruled out as too unlikely.
Nor does John's point explain how a recount wouldn't obviously undo any typographical "correction" that wasn't accurate in the first place.
You should re-read the original Star Tribune article. By "net" gain, the Star Tribune is not, as you were, subtracting McCain vote additions from Obama new additions. It was simply adding new Obama votes from the recount, and subtracting lower Obama numbers from the same recount (from other precincts). There was no consideration of McCain votes in their estimation of "net" gain.
And it appears that, in fact, Prof. Lott is exactly correct. If one ignores the 2 non-reporting precincts, then a recount shows 106 votes more for Sen. Obama, but nearly 2.5 times more originally miscounted or uncounted votes for Franken, which would seem non-intuitive, at the least.
Were the numbers somehow similarly slanted toward a specific Republican candidate, what would your reaction be? Whatever they might have been, I doubt "Big deal" would have been among the likely finalists.
I don't consider myself angry left, and sure enough I am EXTREMELY pro-Second Amendment. But I'll tell you, Lott's work in this field is NOT a credit to our cause, just as Kellerman and Bellesiles are not a credit to the other side's arguments. I await with relish my philosophical opponents to trot out Kellerman in a debate. And they await with relish Lott's 2+ million number to come out so they can attempt to discredit it. I often find that I have to preemptively throw out Lott's numbers to disarm my opponent...ahem.
I apologize, Mr. Lott, if that's really you on this thread. But I find your work in DGU to be a hindrance in the advancement of our cause when I'm debating people on the other side of that issue. Unfortunate.
Q1 Did these two precincts also report late for Franken/coleman, and if so are these votes part of Franken's 459?
Q2 I Alexandria were there zero votes for McCain?
Since it's posted under his own name, probably not.
What do you think, Mary Rosh?
According to the editors of the Minneapolis Star-Tribune, the state's largest daily newspaper, the "conservative statistics" differ starkly from the reality-based data used by everyone else. The paper's editor in charge of computer-assisted reporting compared the actual numbers and found that "58 precincts changed their Obama totals for a net gain of 1,121 votes (46 precincts added another 1,268 votes for Obama, 12 precincts took away 147 votes)." while "48 precincts changed their Franken totals for a net gain of 459 votes (37 precincts added 569 votes for Franken; 11 precincts took away 110 votes)." In short, the Lott accusation, originally featured on Fox, was "simply wrong." Perhaps the old aphorism should be changed to "lies, damned lies, and conservative statistics."
Perhaps Lott's numbers come from a hard drive that he lost. Did Lott's computer eat Coleman's votes?
John Lott is to careful, honest analysis what Sarah Palin is to smart.
No, unfortunately even the recount doesn't bear watching, since only the inevitable court case will matter in the end.
"John Lott" turns up here whenever John Lott comes up as a topic of conversation. Until proven otherwise, I'll assume that it's the real John Lott - it's too small a community and the VC is too "well regulated" for it to be likely that an imposter John Lott would last very long.
So, my take is that he's the real deal.
What we don't know is how many other nyms he also posts under. Or why any reputable publication will print his screeds. Bellesiles doesn't get published anymore, and for good reason. Why is Lott able to get away with well publicized academic dishonesty?
Are the Franken returns similar, ie is he overlooking these two late reporting precincts in his presentation?
At any rate, this doesn't seem terribly worrisome yet. First, the comparison is only between Franken and Obama, which could be distorted for various reasons (e.g., since Obama clearly won the state, the recounters may be putting less focus on counting his votes correctly).
Second, I assume there is transparency, at least eventually. So that outsiders can go back and check the recounters, as they did in Florida 2000. A more precise examination should expose any problems. Is there such transparency in Minnesota?
I think it was those two black panthers from Philly that intimidated people into voting Dem, but they made typos as a signal to others their true allegiance.
Yes, normally only disreputable leftist rags like Salon would pay attention to these sorts of dishonest sockpuppetting twits and print their screeds on a regular basis.
Good Day, sir!
Sincerely,
Ellison
Thomas Ellers
Ryan
the cabana boy
Wilson
Rick Ellensburg
Although the media reports are less than defining on the subject, it appears that the 2 precincts that reported late (the ambiguous 1,015 votes in the presidential race) only reported late in the Presidential election results. Apparently, at least the Senate race from those precincts was reported within the timeline established by the state. But I would qualify that by saying that I can't find anything in print that categorically states as much. But the numbers as reported in the Senate race seem to preclude the Senate votes being likewise reported late.
As as the genuine recount progresses, apparently Franken has a net gain vs Coleman of 5 votes, the result of Coleman losing a total of nine votes, to Franken's loss of only four. This is very early in the recount. A number of ballots have been sent to the appellate authority for final disposition, but it seems as though no one anticipates any changes from the initial rulings.
Plus I wish I had the fellow's credentials.
Justin is clearly a moron who utterly misrepresented Lott's argument in the article and then tried to hedge in his next post, there are several posters who clearly didn't even read the full text and Nuncio, I'm sorry you disagree with the guy on another issue but you're not really addressing any of the substantive points in this article and its pretty intellectually vapid (one of the hallmarks of this site, is addressing everyone's arguments at face worth, regardless of whether they're the deluded one of Yoo or Gore or somebody) to act that way.
Lott may be a pretty dubious scholar in some areas, but he's laying out a pretty basic argument here. We can continue attacking him but is downright lazy to not address the substance.
I always wondered what side of the line you guys were on w/respect to reasonable posters--at least we know which now.
do you have any links with respect to whether the senate results were also not reported
if we take the circumstances in lott's argument as fact, it seems a network problem would have prevented all results from coming in
The same reason why Lawrence Tribe, Charles Ogletree, Doris Kearns Goodwin, etc. are.
I'm not an expert on this controversy, but I think there are a few possible answers to this.
(1) If you're talking about scholarly publications, they don't publish his "screeds," they publish his peer-reviewed studies. If you're talking about papers publishing op-eds, those are usually relatively simple arguments that can be evaluated without relying on specific, original empirical claims (op-eds don't present "my own research that may be faked).
(2) It's not clear to everyone that he engaged in "academic dishonesty." He shilled for his work under a pseudonym; that may be unethical but it's not clear to me that it rises to the level of something that should preclude further scholarly work. As far as the allegedly invented survey, I've spent a couple hours researching that and haven't been able to find anything conclusive, although I'll admit it looks bad. (Did students ever come forward and say they participated? I wasn't able to find data on that.)
It seems appropriate to me that when someone's clearly an able and prestigious scholar, and it's not clear whether or not he falsified data, you keep publishing him and simply subject him to additional scrutiny.
IIRC, there was virtually no ambiguity with the Bell. situation. I may misrecall, though. I would love to see a post from Lindgren with his opinions about the two situations side by side.
If you believe the numbers, say HERE or HERE as just representative, then Franken picked up something like a total 435 votes (this ignores negative numbers posted by Coleman) from typo errors in the election boards' reporting. 3 precincts, none of which was one of the two which were late in reporting the Presidential numbers (all were identified by name), added 448 positive votes to Franken. The net of all other precincts in the state, plus or minus for Franken alone, was a minus 13, resulting in the 435 number. It seems unlikely that the 2 precincts that reported late and had a Democratic vote total in the Presidential race of something like 1,015 votes could possibly be a part of the lower numbers (statewide total minus the 3 mentioned of -13) in the Senate race.
I did notice that the numbers tended to vary by small amounts (448 votes from 3 precincts vs 446 in another article), but were generally consistent, at least within an margin that excluded something like 1,015 votes. I did read yet another article discussing the same 3 precincts, the state-wide totals from the remaining precincts, and more, but I can't find that particular article now.
Was I clear about the rationale, excluding the 2 late-reporting precincts?
I'm not sure which of the Two Harbors precints (Lake County) ended up with a correction of 246 votes. But in those precinct Franken ended up with 273, 243, 346 and 348. Franken with 246 fewer votes in any of these makes no sense either.
The manual recount will sort all of this out.
I didn't realize VC had became a place where we resorted to personal or tangential attacks when an actual argument is presented.
Justin is clearly a moron...
So Anderson, loki13, you've pretty much decided to go for broke and lose all credibility by just engaging in personal attacks huh?
i've never seen someone so efficiently destroy their own credibility. less than 2 min must be some kind of record.
Justin is clearly a moron...
Brian K, less than 2 minutes? More like 0.3 seconds.
i stand corrected :)
Like you, I don't entirely trust the word of John Lott, but it's time to attack his arguments, not him. Check his data. Check his math. Be skeptical of pro-Lott comments. But the Mary Rosh jokes got old a long time ago, especially from people who, like me, post under a pseudonym.
He's paid enough for the Mary Rosh stuff. It was an arrogant, boneheaded move. But it's really old news and it distracts his critics from focusing on his arguments.
It seems to me that the appropriate statistical comparison between the corrections in Obama's and Franken's vote totals should be in the number and kind of corrected tallies, not in the total votes changed.
I eagerly await Lott's more sophisticated analysis on this issue. I'm holding my breath.
To who said I didn't read the Star Tribune article, that's right - it's not publicly available. I was relying on John Lott's description of what it said. To the degree John Lott said "gained 1,211 net votes" when he meant "gained 1,211 total votes" or "netted 1,211 votes," I apologize for misreading him but the redundancy was what caused this error.
And as I said before, and others have said since,
1) the recount makes this all moot
2) the number of errors is only a handful, regardless of its effects on a vote total, and that is what we are looking for when determining whether this is too coincidental to be true.
BTW, I think I've posted here long enough to verify that I'm not a JL sock puppet. :-p
It occurred to me that Steven Levitt has admitted to egregious dishonesty in relation to the Lott situation. Would Lott's critics have Levitt silenced, as well? I hope not. I admire him and continue to enjoy his work despite a single instance of failure of character.
Is there anyone here who doesn't know that somehow somewhere enough errors and unreported votes will somehow be found to put Franken over the top? In the State of Washinton in 2004 just enough votes were found (3 weeks later) and there is no doubt in any Democrat's mind that in Florida in 2000 enough votes would have been found. That is why they so hated the SCOTUS decision. They know there would have been enough votes to elect Gore. I was raised in Chicago and understand the way the game is played.
Al Franken will be the next Senator from Minnesota.
So what are typos? Failure to save a cell, the 0 errors? Inversion, 19 for 91? Dropping a digit 109 for 1090?
I apologize, Mr. Lott, if that's really you on this thread.
Since it's posted under his own name, probably not.
What do you think, Mary Rosh?
So . . . are you saying the whole "sock puppet" thing is bad? See, I had never heard this before.
Does this mean I should shut down my blog? (I pose-&-post as a guy named "Andrew Sullivan.")
Magnificent!
(BTW, are you really a hoosier? I went to Manchester College and did radio in Ft Wayne for several years...)