With all ballots having been counted in Montana and South Dakota, we present our final version of the popular vote scenario tester.
With around 36 million votes having been cast between the two leading Democratic candidates, the voting totals wound up being remarkably close. They are within about a percentage point of one another, almost whichever scenario you pick.
However, we can probably narrow things down from the 972 scenarios that we described before to just a relative handful, all of which center around Michigan. At this point, there is very little argument not to include Florida at full weight. Nor is there any credible argument not to include all caucus states as best as we are able, even if our vote estimates have a margin of error around them. The more exotic scenarios like counting the Texas caucus vote or excluding Puerto Rico and other territories can probably also be discarded.
That leaves eight potential ways that we can handle Michigan.
1. Ignore Michigan entirely. That gives Obama a win by 155,782 votes.
2. Count Michigan at 100 percent and give no votes to Obama. That gives Clinton a win by 172,527 votes.
3. Count Michigan and give all uncommitted votes to Obama. Obama +65,641.
4. Count Michigan and give all uncommitted and write-in votes to Obama. (Note that we have included a new option to treat Michigan's 27,694 discarded write-in votes as uncommitted). Obama +93,335.
5. Count Michigan and allocate uncommitted votes based on the preferences of uncommitted voters in exit polls. We have that total at Obama +6,961, however it is so close that it can essentially be considered a tie.
6. Count Michigan and allocate uncommitted and write-in votes based on exit polls. That gives Obama a "safe" win by 28,008 votes.
7. Count Michigan and allocate all officially-recorded votes based on exit polls. This may be a truer reflection of voter preference because roughly 20 percent of Hillary Clinton's voters indicated in exit polls that they'd prefer to have voted for another candidate. Under this scenario, Obama wins by 90,398.
8. Same as above, but also include write-in votes in the total that we divide among the candidates. This is actually my preferred solution, because write-in voters were almost certainly included in exit polls even if they weren't included in Michigan's official tally. Counterintuitively, Obama's margin goes down slightly if we take this approach (because we are giving the majority share of a slightly larger pie to Clinton). But we still have Obama winning the national popular vote count by 87,351.
Wednesday, June 4, 2008
Final Popular Vote Estimates: Obama wins 7 of 8 Michigan scenarios
-- Nate Silver at 2:08 PM
Labels: michigan, popular vote
16 comments
At this point, there is very little argument not to include Florida at full weight.
You mean besides the fact that we aren't supposed to include either Florida or Michigan?
This is what I meant when I said the other day that the DNC screwed up. Their primary objective should have been to find a solution that satisfied both states and both sets of candidate supporters. They failed miserably. They should have granted Hillary supporters 90% of the vote (some sanction was necessary for the states' violations), they shouldn't have taken her delegates in Michigan, and Obama still would have won the nomination. Now Obama wins the nomination but he's got the Furies chasing him.
At this point, there is very little argument not to include Florida at full weight.
1) Obama never got a chance to make his case to the Florida voters.
2) Hundreds of thousands of Florida voters stayed home because they were told their votes would not count.
I hardly think that qualifies as "very little argument." The DNC compromise was done for political reasons, and it hardly confers upon Florida the legitimacy of a real primary like the one in PA.
The more exotic scenarios like counting the Texas caucus vote or excluding Puerto Rico and other territories can probably also be discarded.
I agree about Puerto Rico, but how is it exotic to include the Texas caucus, where approximately 1 million people showed up because they were told their votes would count? Every contest in which delegates were selected should be included, because voters were told these contests mattered. It seems strange to exclude just one. The weighting should be roughly (1/3) caucus vote and (2/3) primary vote, as established by the Texas Democratic party.
Lisa--
I really wish people would stop repeating the "Obama took Clinton's delegates" misinformation. The delegate compromise Michigan presented had nothing to do with the Michigan primary. They basically admitted the primary was a farce and shouldn't count, and created a compromise delegate allocation to use *instead*.
A perhaps more realistic proposition is that Obama *gave* Clinton five of *his* delegates, because a 50/50 split would have been rather more fair considering no valid primary was held.
I don't see how you can possibly disregard the Michigan write-ins. The popular vote is a moral argument, not a technical one. We want to know which candidate had the support of the most voters. Can there be any question that voters who wrote in Obama's name supported Obama?
Lisa-
Now that Obama has the nomination in hand, I expect he'll give back those 4 delegates in time for the convention.
Stephen, I never said that Obama took anything. And I'm not a very good Clinton apologist. I'm just thinking strategically for the DNC...the point should not have been trying to reverse engineer the motives of the voters...it shouldn't even have been to come up with something "fair." By the time the R&B committee met, the nomination was already determined in favor of Obama, and they should have realized that the *only* goal should have been to come up with a plan to mollify the supporters of the runner-up and to devise a solution in which there was little or no possibility of a "we wuz robbed!" claimed from the Clinton voters.
That's just my take.
Lisa:
You're 100% right from a strategic point of view, but the RBC was also considering the consequences in future primaries of letting Florida and Michigan off the hook. A 50% delegate penalty was the least they could get away with. Still, it would have been best for them to award Michigan's delegates 73-55 with the penalty to satisfy Clinton supporters. If they were really bold, they could have left the 55 delegates for 'uncommitted', since the superdelegates would have put Obama over the top anyways. But that might have pushed Obama supporters over the edge.
Lisa--
Sorry, your "they shouldn't have taken Clinton's delegates in Michigan" comment just reminded me too much of the whining about Obama "stealing Clinton's delegates". My apologies if that wasn't what you intended to say.
Regardless, aside from the effort to mollify Clinton's supporters, it would be very difficult to achieve party unity if Clinton could make a more persuasive argument about the popular vote. If they'd respected Michigan's failed primary, they would not only have been essentially asserting that they have *no standards* on what constitutes a viable contest, but they would have been legitimizing Clinton's claim to a popular vote lead.
No matter what the outcome, many Clinton supporters would cry bias and unfairness. Even had they given Clinton everything she'd asked for, if Obama had then won the nomination the cries of those evil, unfair delegates taking away the nomination from the will of the people would've been even louder. I think they achieved the best possible result to give Clinton more than she could reasonably ask for without damaging Obama's legitimacy.
The argument for not including the Texas caucuses is that it's a double counting of votes. Every single person who voted in the Texas caucuses was required to have voted in the primary earlier that day. To include those caucus results in the overall popular vote count would not be discounting anybody's vote and would be, by definition, counting single individuals' votes twice.
RealClearPolitics notes that if the Washington primary votes are used instead of the estimated caucus votes, Obama's margin shrinks by 50,000. In any case, this is a remarkably close election - your preferred method provides a margin of 0.25%. Who would have given odds that the popular vote would be effectively tied after 57 states and territories voted separately over 21 weeks?
Nate -
first, i like your analysis of Michigan's primary based on exit polls. i wanted to say, though, that while it's a reasonable compromise to award write-in ballots based on exit polls, and a reasonable assumption that those votes occurred proportionally on election day, i think that in the specific case of our primary this year that this is not necessarily accurate.
confusion leading up to the primary was great, of course, but the push on the MDP's part to clarify what real options people had on their ballots was not in full swing until one or two weeks prior to election day. absentee ballots, however, were distributed 4 or more weeks prior to election day. although it's anecdotal evidence, i have heard many more accounts of people writing in names on their mail-in ballots, than in the voting booth. by that time, we had the word out about Uncommitted. exit polls, then, would probably represent a LESS than proportional number of the ruined ballots.
second, although you may have cleared this up at some point, i can't tell if you're distributing write-ins to ALL the candidates based on the exit polls, or all except those on the ballot. again, anecdotal, but even being in a position to hear a lot of stupid things people did with their ballots this year, i haven't heard of anyone writing in Clinton - or Kucinich or Gravel for that matter. any reason to think that these votes were intended for candidates OTHER than those provided on the ballot?
I have problems with your Michigan scenarios.
1. You award Obama all uncommitted and write-in votes. Edwards was still a viable candidate at the time.
2. Since when do we count illegal write-in votes? If you start down this road, you might as well count disqualified ballots from every other contest.
3. Anything awarded based on exit polls seems to completely abrogate the meaning of "popular vote." If this metric has any meaning, it is for actual, legally cast votes. What you're doing is trying to better reflect popular will. But, that's something a little different from popular vote.
4. My personal count: Michigan's election was a farce--it doesn't count.
5. Florida is a tougher case because turnout was much closer to normal and the results are much closer to reality. But, since there was no campaigning and people were told their votes would not count, I tend to think it's vote total should also be excluded.
6. All caucuses should count, even those for which we don't have precise numbers but do have estimates. Of course, we then have to admit that the resulting totals are estimates as well.
7. Territories are sanctioned contests under the rules. Of course their votes should count.
8. What to do with Texas is a mess. I like the (2/3 * primary vote) + (1/3 * caucus vote) solution, as it avoids double counting and reflects the relative weight assigned to the primary/caucus.
All of this (plus SO much more) show that "popular vote" is way too arbitrary of a measure in which to place much, if any, meaning. Especially for a race as close as this one.
Another Mike, what's wrong with just counting the Texas Primary. If the principle is that ea