[UPDATEs: Fans of agonized logic will want to read the whole screenplay that follows; others can skip to "HAMLET OPINES, AGAIN!" at the bottom, where a case for Krugman is made. But by whom?!? [But it is never over! Don Luskin tells us that "two out of three" was also reported at the time as "two out of four"].]
Let's start with Patterico:
Okay, stop. Isn’t Krugman saying the same exact thing he said in his correction? Paul Krugman initially said “Gore won two out of three” — and corrected that statement today to “Gore won two out of three.” Call me crazy, but this appears to be the same exact claim.
I’m really starting to wonder whether Paul Krugman is looking at a different 2001 study of undervotes by a consortium including the Miami Herald than I am. Every time I look at my link to the USA Today article on the study, it says Bush won 3 out of 4 times...
OK, I agree with Patterico - it is a pretty cryptic correction that simply restates the original claim with no explanation, and I have a hrd time believing that this is what the public editor had in mind (although maybe he enjoys getting e-barrages). However, Patterico's "different study" suggestion triggers a bell.
Patterico is linking to a USA Today story from May 2001 describing the results of the Miami Herald/USA Today consortium. However, Krugman cites an April 2001 study, and his original comment seems to have been based on a book review, rather than any research.
Well. This article tells us that (a) the Miami Herald/USA Today consortium published results in April, and (b) those results could have been re-interpreted to point to "Gore Wins Manual Recount of Undervotes".
Now, why the April stories are not trumped by the May reporting is between Krugman and his editors - I would have hoped that in a reality-based community, if a paper publishes a final version in May, columnists would not cite the preliminary April conclusions four years later. Or, if an editor requested a clarification or correction, we would get something other than a restatement of the original proposition.
However, pending the moment when I track down the specific Miami Herald stories, let's say this: it is quite possible that the Miami Herald published a story and follow-up in April 2001 that, when condensed into a table-pounding book about Republican election malfeasance, came out as "Gore won two out of three manual recounts of undervotes".
[And that is the case]
MORE: More cites here - we are 404 on the links, but the headlines are certainly suggestive. Please note that  was published *after* :
 Recounts could have given Gore the edge
Broward, Palm Beach checked
The Miami Herald
Published Thursday, April 5, 2001
BY SHARI RUDAVSKY AND BETH REINHARD
 Review Shows Ballots Say Bush
But Gore backers have some points to argue
The Miami Herald
Published Wednesday, April 4, 2001
BY MARTIN MERZER
Enough. I am declaring victory [prematurely] and going home - this article from April 20 explains the whole thing. Now, why the USA Today article from May is hanging around, and which takes precedence, I don't know. Nor do I know what "the truth" is, or what an accurate journalistic disposition of this question might be.
But on a much narrower point - Krugman is certainly basing his "Two out of Three" on this April logic, which tracks with the "different study" suggestion. Although actually, it is the same Miami Herald/USA Today study modified to recount the already-counted counties under a less restrictive standard. And the jump to "Two out of three" eludes me. Let's just say, *IF* Krugman has a coherent defense, these two additional scenarios described by the Miami Herald will be a key part of it.
Here is what Krugman coulda shoulda written, and it *might* be defensible:
On April 4th, 2001, a newspaper consortium led by the Miami Herald and USA Today reported that in three out of four scenarios involving a manual recount of undervotes, Bush won the Florida election. Under these scenarios, the already-recounted and certified results for four disputed counties were included with a state-wide hypothetical recount performed by the consortium's accounting firm.
However, on April 5th, the Miami Herald reported separately on another plausible set of scenarios, in which *all* votes, including the four disputed counties, were recounted by the accounting firm under consistent standards. In these cases (which were legally unlikely), Gore was the winner in two
out of three scenarios.
Here is what Krugman wrote in his attempted explanation last week:
About the evidence regarding a manual recount: in April 2001 a media consortium led by The Miami Herald assessed how various recounts of "undervotes," which did not register at all, would have affected the outcome. Two out of three hypothetical statewide counts would have given the election to Mr. Gore. The third involved a standard that would have discarded some ballots on which the intended vote was clear. Since Florida law seemed to require counting such ballots, this standard almost certainly wouldn't have been used in a statewide recount.
Since Krugman does not distinguish between the consortium conclusion, as reported originally by both the Miami Herald and the USA Today, and the later "what-if" reported by the Miami Herald, a clarification was clearly in order. However, a careful reading reveals this key phrase - "Two out of three hypothetical statewide counts". Ah ha! As we have been saying, the "three of four go against Gore" story cited by Patterico (and me) does not involve a *hypothetical statewide* recount - it relies on the actual recount for the disputed counties.
Well, here is the attempted clarification, the cause of Patterico's puzzlement:
Corrections: In my column last Friday, I cited an inaccurate number (given by the Conyers report) for turnout in Ohio's Miami County last year: 98.5 percent. I should have checked the official state site, which reports a reasonable 72.2 percent. Also, the public editor says, rightly, that I should acknowledge initially misstating the results of the 2000 Florida election study by a media consortium led by The Miami Herald. Unlike a more definitive study by a larger consortium that included The New York Times, an analysis that showed Al Gore winning all statewide manual recounts, the earlier study showed him winning two out of three.
That did not exactly lend clarity. Well, e-barrage Byron Calame: Public@nytimes.com
But in e-barraging him do note that, although he ought to be filing newspaper columns rather than legal briefs, Krugman seems to
be technically correct have available two scenarios supporting his position, although the leap to "two out of three" is missing. Why this is still mysterious after two attempts suggests a breakdown in the communication process.
Still Missing - Do any of these articles follow-up articles actually say "Two out of three"? Hmmph. A quick re-read indicates this possible interpretation - Gore winning under two loose standards that recounted the missing counties, and losing under the rules chosen by the Herald.
IGNORANCE WAS BLISS: The indefatigable Patterico points me to a copy of the April 5 Miami Herald story, reproduced here sans the Beavis and Butthead background. I see where Krugman gets two new scenarios in which Gore wins, but I can't conjure a "two out of three" from this.
Well. After far too many words, let's just say that *if* Krugman has an explanation, these two scenarios will be a part of it.
HAMELT OPINES, AGAIN! A careful reading of the version of the April 4 story found by Patterico in the Google cache yields this:
The Florida Supreme Court ruled in 1998 that canvassing boards must examine ``damaged or defective'' ballots for voter intent and further defined the term ``defective ballot'' as ``a ballot which is marked in a manner such that it cannot be read by a scanner.''
Though some canvassing boards in optical-scan counties conducted such reviews on Election Night, most boards around the state did not.
Had all canvassing boards in all counties examined all undervotes, thousands of votes would have been salvaged in Broward County, Palm Beach County and elsewhere long before the election dispute landed in court -- and the outcome might have been different, The Herald found.
In that scenario, under the most inclusive standard, Gore might have won Florida's election -- and the White House -- by 393 votes, The Herald found. If dimples were counted as votes only when other races were dimpled, Gore would have won by 299 votes.
But if ballots were counted as votes only when a chad was detached by at least two corners (the standard most commonly used nationally), Bush would have won by 352 votes.
That is two out of three for Gore, for statewide manual recounts under the 1998 standard.
This bootleg version of what is meant to be the original April 4 Miami Herald story breaks these same scenarios out in more detail and presents this same two of three result under the heading "The Statewide Standard".
That is a pretty clear boost for Krugman, on content anyway. Why was this so mystifying? If we had been able to find that darn April 4 story right away, this controversy might never have developed.
And in the fullness of time, my source for that bootleg version may become clear. Or, the copyright folks may take it away - stay tuned!
FINAL ANSWER - per Don Luskin, the Miami Herald also subdivided the scenarios favorable to Bush and reported the result as "two out of four".
NY Imes bio page