/* Google Analytics Code asynchronous */

Friday, May 30, 2008

New Estimates - Nothing to See Here

Two new concepts to incorporate, while updating the figures: Markos suggested a methodology for estimating the popular vote, yesterday. Poblano suggests an estimate that uses exit poll data (I had used another source).

The most important take-away is that all figures relating to a "popular vote" in the nomination race are estimates, unlike the general election. (The second most important take-away is that estimates that show no "popular votes" for Obama out of Michigan are pablum, to be wholly ignored).

WHAT TO DO WITH THE NON-BINDING PRIMARY FIGURES?

Unlike Markos, I've completely discounted the votes from non-binding primaries. There seems to be no reason to believe that they are highly representative, precisely because they are non-binding. Because they are after-the-fact, they are not even straw-polls (if memory serves, Washington simply stopped counting the votes at circa 90% done, even). Texas, however, is a binding, combined primary-caucus state. Something could be done, possibly, but it would involve eliminating double-counting, a bridge too far.

BETTER EXIT POLL DATA FOR MICHIGAN

I've incorporated the alternative exit poll data. One has to judge between incorporating actual votes cast and how much adjusting to do. Of course, one could do the full adjustment, for both Michigan and Florida. That's too much of a departure, for me. However, there is a strong case to be made for adjusting the turnout for Florida and Michigan:

By contrast, Michigan's turnout in its unsanctioned Democratic primary was 594,398, or only 7.8 percent of its eligible voting age population. That is about three-and-a-half times less than it "should" have been, based on the patterns in Indiana, Ohio and Wisconsin.

Except that adjusting one without the other magnifies any bias already in the numbers.

Clinton gets Ricky Martin...
DATA MAVENS TIP HAND ON PUERTO RICO LIKELIHOODS

Last, there appears to be better info floating around to help estimate Puerto Rico turnout. The margins I have seem reasonable, but the turnout could be as much as half of what I've conservatively estimated. If so, there is little convincing chance that Hillary comes out a leader in the popular vote estimate. I like to be sure of these things, before continuing on.

Despite these expert estimates, I'm not changing my guesses, for now, which are based on the number of people who voted in the race for the Governor last time (the primary in Puerto Rico is an open primary) and the theory that Bill and Hill could convince a lot that their vote really counts. At 500K, officials are expecting less than 1/3 of my estimate, so I'm very conservative ...

1. Only estimates for the popular vote can be made (no direct observation possible). Updates for Kentucky, Oregon, and various other adjusted/finalized figures.


StateObamaClintonDifference
Primaries, States15,862,33915,778,39783,942
MI, FL Adj; BHO=23%713,0091,199,295-486,286
MI Adj: BHO=29%

38,660
memo: MI Adj: BHO=35%*

32,712
memo: MI Adj: BHO=35%, HRC=46%*

54,718
Tally, adjusted:16,575,34816,977,692-363,684
Projections (OR, MT, SD):491,517357,083134,434
Subtotal, Tally & Projected17,066,86517,334,775-229,250




Caucuses, States & D.C.437,810184,153253,657
IA, ME, NV, WA Adj253,948186,93467,013
Tally:691,758371,087320,670
Caucus-to-Primary Adj

560,789
Subtotal, Tally, adjusted:2,908,4312,026,972881,459




Caucuses & Primaries, Territories19,37110,0709,301
Projections (PR remains)768,8001,153,200-384,400
Subtotal, Tally & Projected788,1711,163,270-375,099




Grand Total20,763,46820,525,017277,110
*figures presented as incremental impact, so you can add them to totals to further adjust.

FYI, These primary-state numbers foot to the RCP totals, except for Oregon (I have the most current figures). RCP figures have a much higher adjustment for the non-reporting caucus states of IA, ME, NV, and WA., of about 40K more. They don't show their work, so I don't know why.


Gory details, here [Zoho spreadsheet. printable].