Nate Silver is Not God and other political musings

In the wake of the election, the big nerd story is the perfect stats-based prediction that Nate Silver of the 538 blog made on the results in every single state. I was following the blog and like all, am impressed with his work. The perfection gives the wrong impression, however. Silver would be the first to point out he predicted Florida as very close with a slight lean for Obama, and while that is what happened, that's really just luck. His actual prediction was that it was too close to call. But people won't see that, they see the perfection. I hope he realizes he should try to downplay this. For his own sake, if he doesn't, he has nowhere to go but down in 2014 and 2016.

But the second reason is stronger. People will put even more faith in polls. Perhaps even not faith, but reasoned belief, because polls are indeed getting more accurate. Good polls that are taken far in advance are probably accurate about what the electorate thinks then, but the electorate itself is not that accurate far in advance. So the public and politicians should always be wary about what the polls say before the election.

Silver's triumph means they may not be. And as the metaphorical Heisenberg predicts, the observations will change the results of the election.

There are a few ways this can happen. First, people change their votes based on polls. They are less likely to vote if they think the election is decided, or they sometimes file protest votes when they feel their vote won't change things. Vice versa, a close poll is one way to increase turnout, and both sides push their voters to make the difference. People are going to think the election is settled because 538 has said what people are feeling.

The second big change has already been happening. Politicians change their platforms due to the polls. Danny Hillis observed some years ago that the popular vote is almost always a near tie for a reason. In a two party system, each side regularly runs polls. If the polls show them losing, they move their position in order to get to 51%. They don't want to move to 52% as that's more change than they really want, but they don't want to move to less than 50% or they lose the whole game. Both sides do this, and to some extent the one with better polling and strategy wins the election. We get two candidates, each with a carefully chosen position designed to (according to their own team) just beat the opposition, and the actual result is closer to a random draw driven by chaotic factors.

Well, not quite. As Silver shows, the electoral college stops that from happening. The electoral college means different voters have different value to the candidates, and it makes the system pretty complex. Instead of aiming for a total of voters, you have to worry that position A might help you in Ohio but hurt you in Florida, and the electoral votes happen in big chunks which makes the effect of swing states more chaotic. Thus poll analysis can tell you who will win but not so readily how to tweak things to make the winner be you. The college makes small differences in overall support lead to huge differences in the college.

In Danny's theory, the two candidates do not have to be the same, they just have to be the same distance from a hypothetical center. (Of course to 3rd parties the two candidates do tend to look nearly identical but to the members of the two main parties they look very different.)

Show me the money?

Many have noted that this election may have cost $6B but produced a very status quo result. Huge money was spent, but opposed forces also spent their money, and the arms race just led to a similar balance of power. Except a lot of rich donors spent a lot of their money, got valuable access to politicians for it, and some TV stations in Ohio and a few other states made a killing. The fear that corporate money would massively swing the process does not appear to have gained much evidence, but it's clear that influence was bought.

I'm working on a solution to this, however. More to come later on that.

Ballot Propositions

While there have been some fairly good ballot propositions (such as last night's wins for Marijuana and marriage equality) I am starting to doubt the value of the system itself. As much as you might like the propositions you like, if half of the propositions are negative in value, the system should be scrapped. Indeed, if only about 40% are negative, it should still be scrapped because of the huge cost of the system itself. My anecdotal review of famous propositions shows a mostly negative score. Many were bad to start with; others people saw were failures with time. I'm not just talking about any one person's personal opinion. This election in California three propositions whose goal was to reverse the actions of earlier propositions. In one case, the death penalty repeal, though it lost, was sponsored by the authors of the original proposition to establish the penalty! I am hoping that somebody will do, or has done, an objective analysis of proposition success in hindsight.

Up close, it surely looks bad. It's very clear that money can do a lot when it comes to propositions, though it must be spent with care as it can backfire. In California, the proposition to label genemod foods clearly went down to money. (One might agree that the money was largely right here, in that the law, backed by the organic food companies with a specific exception placed in it for them, was a badly written law, but the money allowed that message to get out strongly and defeat the proposition.)

Another bad proposition was Prop 35, which was billed as anti-pedophile. Write a proposition that way and it can't lose, nobody is going to put big money to be seen as pro-pedophile, and in fact donations to No on 35 were $0. The ACLU, EFF and other principled orgs came out against it and in fact we got a temporary injunction against it today due to its unconstitutional restrictions on speech, but it had no chance of losing at the polls.

Also very bad are the sneaky "tax to support motherhood" propositions, which work too often. Prop 30 was one example this time, though it only barely passed so voters may be seeing through this strategy. These propositions arise when the politicians want to raise taxes to spend on unpopular programs. They can't ask the public to support the unpopular programs (duh) so instead they pretend that the most vital, popular programs are threatened and due to be cut any day. And so they go out to ask the public to fund schools, or levees, or parks. And of course the public wants to fund those. In effect, they hold the schools hostage. The money comes in and funds the motherhood cause, and that saves them from having to spend as much general revenue on the motherhood cause. Without the propositions, they would never cut things like schools and parks, and they never really intend to. But the tax increases the public thinks are going to support those things really free up general funds to be used for other things the politicians want.

The hostage taking was particularly blatant with prop 30, which is why it barely passed. They actually passed a budget with cuts, and then said that unless the tax increase was approved, the cuts would be automatic and children would go uneducated. Right. The teachers' unions support these things, even though they are partly the ones taken hostage -- just as the hostage probably wants you to pay the ransom.

It seems a good policy on ballot propositions would be to vote no, unless they are repealing other propositions. Certainly on things that the legislature could and should easily do on its own. This is where things like the marijuana legalization fit. In these cases, the state house is (under federal pressure) not willing to actually represent the will of the people, and so the people get their say. That's the goal of the proposition system. I just wish it worked that way more than on rare occasions.


Interesting post Brad.

As someone who lives and is forced to vote by law, in a jurisdiction without propositions, I think I'd like to see propositions on the ballot paper. For one thing it would definitely see gay marriage legalized in Australia. You're generally right though in that we think these ideas are great right up until cynics take over and then it becomes a mess.

Regarding Silver, he seems like the kind of guy who won't get overly carried away with himself. Hopefully I'm right on that. You can bet he's going to be making bookmakers nervous in the future, however...

The hard challenge is to have a proposition system which can get you the ones you like but avoid the abuses. Perhaps with the lessons of California and other states things can be improved. But people have a very hard time doing the math I describe. They will think that a system that gives them 50% propositions that are good for the state and 50% bad is a good idea, when it's not.

What you want to avoid are:

  • Propositions that provide strong financial benefits to private parties. These will cause huge funding efforts by those parties, and the vote will tend to mirror the money. No proposition should be such that it will make billions for a group of private interests -- that's a pretty strong limitation.
  • Harder to accept is you may not want propositions that will cost companies billions, they will spend it to stop it. Problem is there are a lot of propositions of this type, almost all business regulation will show up here.
  • The legislature should be given the chance to vote on the proposition if it is within their power to enact. Only if it fails there should it become a ballot proposition. This limits us to things the legislature can't or won't do.
  • You need some limit on fundraising and bond-issuing propositions sponsored by the government. In California, local tax increases that have a specific purpose require 2/3 vote which they almost never get, but even then many of the trick propositions for motherhood issues -- normally schools -- get past this rule.

Add new comment