Wednesday, November 5, 2008

First Mistake

Welcome to the kick-off entry of Least of All Evils! Here, I intend to advocate in favor of preferrential voting in general (and Condocet ranked pairs in particular), with the hope of increasing its use in the United States at large, but especially in my home district and state.

I don't think today I'll get much attention from Democrats or third-party voters; the first is too busy celebrating, the second is perhaps too accustomed to their position at electoral whipping-boy. So we start with the Republicans, as I think they're the ones looking for answers today. So:

My condolences, Republicans; you lost.  Now is a time for reflection, a time to look back, a time to analyze, a time to determine what went wrong, and what you can do differently next time to fix it.  Please, allow me to offer an possible answer to these questions.  Your first mistake, was in selecting John McCain.

Wait, hold on; don't go!  I don't mean it like that! Give me the chance to explain. Selecting John McCain was a mistake, but not in the sense of "It was a mistake for you to pick him."  It was a mistake because you didn't pick him, and yet he was your nominee anyway.  You see, it's not your fault.  You didn't do anything wrong.  You went to your caucus or your primary (or in some cases, both), and you voted.  You voted based on a careful analysis of the candidates positions; you voted what you knew in your heart was right; you voted for who you thought was the best man for the job.  The fault was not in how you voted, but in how you voted.

John McCain, by and large, was not the candidate most members of the Republican party had as their first pick for the nomination.  And yet, he still got the job.  How?  There are a lot of details that build up to arrive at the answer, but the short version comes in two parts: you used a bad system, and Mike Huckabee.

Going in to the Iowa caucus, way back on January 3rd, all the talk was about Mitt Romney and  Mike Huckabee and which of them was going to win the kick-off of the primary season.  John McCain didn't even make it into the top three there, losing not only to Huckabee and Romney, but even to Fred "I'll do it if you make me" Thompson.  McCain of course turned it around five days later and won New Hampshire.  But what do we mean when we say that, "won"?  Let's take a closer look.

John McCain "won" New Hampshire with 37% of the vote.  Compare that to yesterday's returns in Alabama, where no one is surprised Barack Obama lost by a huge margin, but he still took 39% of the vote. Now, I know what you're saying: You're saying that the New Hampshire primary was different because there were a lot more than two contenders.  But that is precisely the point, and in  fact, is the point of this entire blog.  That point being, the methods we use to choose our elected officials do a surprisingly poor job when we use them to pick between more than two options, but there are ways to fix that short coming.

Let me offer you a hypothetical.  Suppose Mitt Romney and Mike Huckabee hadn't run in the primaries, but instead there was another candidate, let's call him Mital Rockabee.  Rockabee, had he been running instead of these two, would have gotten all the votes that either of them would have otherwise gotten.  How would the primaries have gone differently?  Due to the vagaries of election laws, Rockabee would not have gotten only the sum of Huckabee and Romney's delegates, but would have also received a large chunk of McCain's delegates; enough to make all the difference.

I've written a torturous state-by-state analysis (which you'll find in the comments) if you're a detail-oriented type of person, but feel free to skip it if you're comfortable taking my word for it.

Following the results of Super Tuesday, Romney suspended his campaign, at which point he had 270 delegates, and Huckabee had 176, versus McCain's 680.  With the 259 (or more) votes that would have shifted away from McCain due to vote-tallying minutia, Rockabee would have had 705 delegates, versus a McCain total of 421.  At that point, Rockabee would certainly stay in the race (like Huckabee did).  It's hard to do any kind of what-if analysis from this point on. Did more voters figure that, since McCain had it in the bag, they'd vote for their #2 favorite and try to get him a VP slot?  Or did more of them decide to jump on the party bandwagon, an show their support by adding it to the presumptive winner?  It's difficult to say.  But with that kind of lead, it's certainly possible that Rockabee would have gone on to assume the nomination.

But what does that mean, since there obviously is no Mital Rockabee? Rockabee represents a rough estimate of how a prefferential voting system would have adjudicated the election, given the assumption that every Huckabee and Romney voter preffered either of those two options over John McCain. (Is this a fair assumption?  Hard to say without data. Hit up the comments with your thoughts!) Had the primaries used such a system, instead of Huckabee spoiling the election for Romney by spliting the vote, Romney likely would have defeated Huckabee and McCain and become the Republican nominee. This is suggesting that, in essence, the Republican party as a body, didn't want John McCain; it wanted Mitt Romney.  But due to the use of these poor methods for measuring the will of the voters (which, I should mention, are the same poor methods used for virtually ever election in this country) a candidate you didn't want was put forward as your selection.

Now, I'm not going to guarantee that Romney would have defeated Obama yesterday.  But you wanted him, and picking someone you didn't want, that was your first mistake.

Tomorrow, we'll start talking about how to fix it.


  1. First state of interest, on January 8th, New Hampshire. Each state allocates their delegates in a slightly different way. In New Hampshire, the tally up all the votes across the state. Any candidate who gets less than 10% of the total is dropped. Among the remaining votes, the state's 12 delegates are divided proportionally, with any left over fractions going to whichever candidate had the greatest votes. McCain got 37.71% of the vote, Romney 32.17%, and Huckabee 11.44%; all other candidates had less than the 10% cutoff. So the actual primary vote gave 5 delegates to McCain, 4 to Romney, and 1 to Huckabee, plus the 2 delegates that got fractioned up going to McCain for a total of 7. But had Rockabee been in there, not only would he have gotten Romney's 4 plus Huckabee's 1 delegate, he'd have taken the 2 fractional delegates as well, and won the state. 2 down.

    Next up, eleven days later, is South Carolina. South Carolina uses a system that seems to be rather common. It breaks the delegates into two parts. The first part is chosen by congressional district (of which South Carolina has 6), and it's winner-takes-all; whoever gets the majority of votes in each district, gets 2 delegates, a total of 12. The states other 12 delegates are also winner-takes-all, but chosen by the whole state. Mike Huckabee came out on top in half of the districts, and McCain in the other half, but McCain won the state-wide tally 42% to 38%, and so won 18 of the states 24 delegates. But Rockabee? The Huckabee-plus-Romney totals, district by district, beat McCain across the board, and Romney's 20% state-wide gives a clear 58% versus 42%, meaning Rockabee would have swept the state, docking McCain 18 delegates. 20 down.

    Ten days later and we take a look at Florida. Florida's system is very simple to understand: whoever gets the most votes, gets all the delegates. McCain got 36%, and Romney only 31%. But Huckabee took 13%, which would have given Rockabee a total of 44%, handily defeating McCain, and taking all 57 of Florida's convention votes. 77 down.

    One week later. Super Tuesday, so called because such a large number of states have their primaries. Usually, after Super Tuesday, it's pretty clear who's going to win. We'll go quick, since I think you'll get the idea by now. Delaware, like Florida, is a true winner-take-all. Rockabee would have beaten McCain, and cost him 18 delegates. It's the same story in Missouri, for 58 more. Oklahoma plays the same game as South Carolina, with a combination of district-level and state-wide winner-take-all components. I can't find district-level totals, but Rockabee would certainly have taken Oklahoma's 23 state-wide delegates, and probably several of the 9 district-level delegates McCain gained. And again in Georgia; although Huckabee won the state's state-wide delegates, Rockabee would have run up the score and taken all 9 district-level delegates McCain would have won. Finally, it's the big-boy on the field, California, who also uses the combination system. McCain won all but 3 of this huge states 119 delegates, but Rockabee would have taken not only the 14 state-wide, but 69 more at the district-level. Tennessee, Alabama, Illinois (not to mention Michigan) would have also had a few more delegates jump to a Rockabee decision (but like Oklahoma's district numbers, it's hard to determine how many). But that's irrelevant, because we can certainly account for at least 182 more votes on Super Tuesday, for a total shift of 259.

    At the end of the primary period, McCain had just 1,401 delegates locked-in. That's only 210 more than the bare minimum, and that margin can be eaten up by the end of Super Tuesday!

  2. Oh, a thought: it's not necessary that Romney voter's second choice be Huckabee. As long as Huckabee voter's second choice was Romney, then he still would likely have prevailed under any preferential system.

  3. Been reading more, and thinking. Maybe not ranked pairs. Maybe beatpath. Readin a lot about range and approval, too (such as claims that range voting is more likley to find a Condorcet winner than any Condorcet method with strategic voters). Interesting.