Understanding progressives’ presidential picks
June 26th, 2007
Jack
Jack Santucci was FairVote staff from 2005-07.
Part two in our analysis of IRV exit poll results from Take Back America 2007. For more posts on the conference, click here.
We’ve crunched the numbers (warning - Excel file 61kb) from our 2007 Take Back America straw poll. FairVote has run such instant runoff polls - or “second choice” as they’ve come to be known - at TBA since 2003. This year Politico.com joined in with a second choice poll of its own:
The poll establishes Obama and Edwards as a two-man top tier among the liberal segment of the Democratic base, said pollster Stan Greenberg, because they were also the most popular second choices for the nomination.
“If you look at this, you see Obama’s [supporters'] second choice and Edwards [supporters'] second choice are each other — in this group, the two of them form the top tier.”
Those results may indicate that a majority of the activists surveyed are looking for a choice other than Clinton.
In general, our results and conclusions mirror Politico’s - Obama won the instant runoff - but there are some key differences:
1) We included Gore on our ballot.
2) We asked voters to rank beyond their second choice, which allows for, one, deeper examination of breadth of candidate support and, two, fuller speculation on whom the supporters of X vote for when X isn’t in the race.
3) At least among this sample (admittedly not the most representative of Democrats), there is no clear frontrunner. Progressives are divided.
No clear frontrunner
The progressive landscape has changed quite a bit from 2003 when Dean was a clear leader in both first (42%) and total second choices (27%).
For Greenberg, Obama and Edwards make up the top tier of the field, based on results from TBA. This, he says, is because supporters of each overwhelmingly list the other as their second choice.
While we did find a 43% plurality of Obama supporters ranking Edwards second, there was no definitive leader for second-choice support among Edwards voters. Among Edwards supporters, Hillary Clinton picks up the most second choices (but only 27%).
Where Edwards really picks up support is among Gore voters, capturing 44% of their second choices.
As far as first choices go, there is no overwhelming leader. Gore and Obama lead the field with support in the 20-point range, while Kucinich and Edwards follow with support in the teens.
Hillary’s shallow roots
For all the buzz about Hillary Clinton (various writings on various walls and the near-certain first woman President) the results of this poll tell a different story, at least among this small and left-leaning sample. Not only is the former First Lady light on core support, coming in fifth with only 12% of first choices, but she lacks breadth of support to an almost equivalent extent, coming in fourth with 15% of overall second choices. Gore beats her for third place in second choices, followed by Obama and Edwards respectively.
Thinking about “total support,” or first choices plus second choices (a “Bucklin index,” if you will) Hillary (with 18) again comes in fourth, well behind Obama (with 28), Gore (with 27) and Edwards (with 26).
The world without Gore
IRV polling also lets us simulate what would have been in the absence of any candidate. This is especially useful here because Politico did not include Gore on its ballot while we at FairVote did.
First-choice standings were as follows:
Biden: 2
Clinton: 8
Dodd: 1
Edwards: 11
Gore: 16
Kucinich: 10
Obama: 14
Richardson: 3
Feingold (on write-in): 1
Assuming ‘IRV logic’ and ’second-choice poll logic’ are strategically equivalent in voters’ minds (I speculate on that a bit here), we can redistribute Gore’s ballots to those voters’ second choices for a breakdown of:
Biden: 2
Clinton: 8 +4 = 12
Dodd: 1
Edwards: 11 +7 = 17
Kucinich: 10 +1 = 11
Obama: 14 +3 = 17
Richardson: 3 +1 = 4
Feingold (on write-in): 1
In a world without Al Gore, Edwards is the consensus candidate, picking up 7 additional ‘breadth’ votes. In this scenario, Edwards ties with Obama, who picks up just three votes from Gore supporters. Despite picking up 4 more votes, Clinton still lags 5 votes behind either Edwards or Obama. And Edwards and Obama tie.
The value of IRV polling
Second choice polls seem to be the new thing. National Journal’s Hotline was running them as early as last May, and now the Politico asks for subsequent choices at major conferences like TBA. Two DailyKos polls asked readers for their second choices (April 2007 and June 2007), and the Des Moines Register ran a second choice poll this May, though it didn’t report the results. At the very least, ranked poll reporting should come with tables providing distributions of second choice support.
Overall, though, IRV polls have value over second choice polls because they let one take the analysis further. Many candidates enter; one (wo)man leaves. As the field narrows, knowing about voters’ third, fourth, fifth choices (and so on) lets us better gauge where consensus is likely to fall.
And next-choice polls in general have value for politicans and prognosticators alike. This post is itself crude prognostication. What politicians can learn from ranked polling, however, is which opponent’s base to court as primary day looms.
Other posts by Jack
- Understanding progressives' presidential picks - June 26th, 2007
- Ontario MMP vote hits U.S. blogosphere - June 22nd, 2007
- Fatah calls for proportional voting in Palestine - June 22nd, 2007
- Fifth annual Democratic IRV straw poll - June 21st, 2007
- McCain: Front-loaded primary schedule is "bizarre" - June 19th, 2007
- Back at Take Back - June 18th, 2007
- Turkey, women and PR lists - June 6th, 2007
- Ontario MMP video roundup - May 31st, 2007
- Open thread: Ranked voting activism - May 28th, 2007
- Proportional apportionment vs. proportional representation - May 25th, 2007
July 1st, 2007 at 7:46 pm
Excuse me Jack,
When you tallied the results for the IRV demonstration, did you save the individual ballot ranking information? Perhaps we might be able to see what the Condorcet results could have been as well? Just wondering.
July 4th, 2007 at 12:54 pm
Joel, I do have the individual ballot information. It will take some time to computerize it all, however.
July 6th, 2007 at 12:42 pm
The Condorcet winner (using OpenSTV to count ballots) was Obama. In case of ties (Condorcet cycles), I set the tiebreaker function to use IRV on the Smith set. The tiebreaker was not needed.
July 6th, 2007 at 12:45 pm
The Approval winner, also using OpenSTV, was also Obama.
Note that the software does not differentiate among rankings. Any ranked candidate is considered “approved.”
July 6th, 2007 at 2:33 pm
I’m not sure that OpenSTV’s definition of approval voting is very useful, even as a quick-and-dirty comparison.
Did your poll question(s) include something you can use as an approval threshold, and that you can tally by hand? By the way, as someone who once designed survey questionnaires for a living, I think there might be some cross-contamination if you ask the same people for both approval/disapproval and rankings in the same survey.
July 6th, 2007 at 3:50 pm
Agreed and agreed. It’s a lot like trying to model PR results on the basis of plurality results.
But the thought experiment is fun.
July 9th, 2007 at 8:49 pm
It is interesting to note that someone needs to see ALL of the votes in order to compute who an IRV winner is. For anyone to analyize the election results they would need to analyize all of the ballot entries for the race.
This is because votes tallies in IRV are not summable.
So if, for example, it was discovered that 1000 ballots were missing from the election one could not simply tally the results of those 1000 ballots and add it to the other results. One would instead need to re-analyize ALL of the ballots and rerun eliminations which potentially could be very different with the extra ballots.
Contrast that with almost any other single-winner voting system where you could just add up subtotals (by candidate for most systems, by candidate pair for Condorcet methods) to determine who the winner
would be.
July 9th, 2007 at 9:10 pm
The political significance of Greg’s point might be that manual recounts (read: paper and hands) could be a pain under IRV.
OpenSTV did the math very quickly. I wonder if one could just add new ballots to the datafile and have results immediately.
July 15th, 2007 at 8:50 pm
Jack: I wonder if one could just add new ballots to the datafile and have results immediately.
Exactly. Greg’s summability is only important to those who believe that ballots should be counted separately in each precinct for election integrity reasons. I happen to think that precinct-based counting is probably a bad idea — precisely for election integrity reasons.
Even so, as long as the number of precincts or voting centers is manageable, you can count IRV separately. It’s done in Northern Ireland (and possibly in the Republic of Ireland as well, I’m not sure). After each round, each voting center reports it’s current totals to a central location. The central location figures out whether someone has won or whom to eliminate, and feeds the information back to the separate voting centers. Each round takes a little longer because each location has to wait for all of the others at each round.
July 16th, 2007 at 5:09 pm
I think that Jack has correctly identified the political cost. There is a tension between the “let’s have manual recounts” movements and the “let’s promote IRV movements” that has not been explored sufficiently.
July 17th, 2007 at 2:04 pm
Let’s stop talking in the passive voice and explore the tension for a minute, then.
There is not much of one.
1) It is highly unlikely that paper-only elections become the norm in America again.
2) Even if they do become the norm, that does not signficantly reduce the potential for fraud. To steal an electronic election, hack the comptuer. To steal a paper election, incinerate the ballots.
3) If those 1,000 ballots are found and a recount is needed, just do the recount. Ireland hand-counts STV elections (more complicated than IRV) in a day. Takoma Park hand-counted an IRV election in just a couple of hours.
The choice is between (1) more quickly knowing the result in a choiceless election where third candidates are branded as spoilers or (2) having real choices in an election where the recount takes a little bit longer.