Saturday, November 04, 2006

The Case for Low Voter Turnout

People often complain about low voter turnout in the U.S. - 50-55% in presidential election years, and 35-40% in non-presidential election years like 2006. See voter turnout since 1960 at this link. Notice that voter turnout has decreased in the last 45 years by more than 10 percentage points, from 63.1% in the presidential election of 1960 to less than 50% in 1996, and from 48.4% in 1966 to 36.4% in 1998.

Many people are probably upset by this trend, but not I, and here is why: In almost all cases, higher voter turnout would NOT have changed the outcome of the election, and so we get the same election outcome/results at a lower cost to society (measured in the opportunity cost of our time).

I have never heard those who complain about low voter turnout make the argument that higher voter turnout would CHANGE the outcome of the election, they usually just say that more people should vote for other reasons: exercise our right to vote, fulfill our civic duty, participate in democracy, etc.

But I have never heard anyone say "More people should vote because low voter turnout leads to unreliable results," or "more people should vote because that would change the outcome/results of the election. " Mostly, I think people would simply "feel better" if we had the same results with 80% turnout, compared to having those same election results with 40% turnout.

But think about it - would you feel any better about a blood test if they took two pints of your blood compared to 20 ccs? Probably not.

There are about 160 million registered voters in the U.S. From sampling theory, a sample size of 16,639 would accurately and reliably represent the entire population of 160m at a 99% confidence level, with an error of only 1%.

What this means is that if the first 16,000 people who vote when the polls first open in the morning are random voters who represent the population of voters, almost all elections are already decided by 9 a.m. or so in the morning. The rest of the voters are really just wasting their time, in the sense that their votes will not affect the outcome of the election. It is like increasing the blood sample - the blood test will still be positive or negative regardless of the sample size.

The possible exception to this would be an extremely close presidential election like 2000, where "every vote mattered" in a sense, but this was an extreme case. Of course, you don't know ahead of time which elections will be that close, so you could justify always voting, just in case.

But in almost all elections, I would argue that voter turnout has NO effect on the outcome of the election, and the same results would prevail with 20% turnout as 100% turnout. Although most people would "feel better" about the election with 100% turnout, that is irrational in my opinion. If the results are same, I would prefer the 20% turnout, because of the significant saving of time for the 80% who did not vote.

Voting is expensive when measured in its full cost: our time. An hour spent voting is an hour lost forever doing something else. As Gatemouth Brown said (see an earlier post): "My time is expensive, I gotta make it last."

I like low voter turnout, because the election results are almost always exactly same as for high voter turnout, and low voter turnout saves and conserves our most precious non-renewable resource: our time, and therefore it is socially more efficient than high voter turnout.


At 11/04/2006 6:45 PM, Anonymous Anonymous said...

I agree that the downward trend of voting in presidential elections is not alarming and that the direct societal costs of voting on election day are closer to the optimal if less than 80% of the population votes. However, I'd like to question your critical assumption in this post. You explicitly assume that those who vote are a random sample of a the potential voting population. The voters who show up at the polls between 7am and 9am are probably not representative of the country's voting population. They are probably more involved politically and more aware of political issues. This may be good for political literacy of the voting sample, but if democracy is about representativeness, the early morning voters, or for that matter, the 55% who do vote, are probably not representative. Moreover, the potential voting population is a heterogeneous one. Homogeneity of political orientation is implicitly assumed in you post, but I would argue certain groups in society are LESS likely to vote but MORE likely to vote a particular way. Low income voters may be inclined to vote for a Democrat, but getting to the voting polls may be more difficult for this group, thus lowering their probability of actually voting. One may argue that wealthy voters are less likely to vote because they have higher opportunity costs of their time and are more likely to vote Republican. I agree with the second part of this statement, but I know of no data that says that the rich are less likely to vote.

It would be interesting to look into a natural experiment that kept a certain percentage of voters from voting. Such a natural experiment would be difficult (impossible?) to find, but this is would we would need to test Mark's hypothesis that a 20% turnout yields the same results as a 80% turnout.

Interesting post.

At 11/04/2006 10:52 PM, Anonymous Anonymous said...

Polls work because a sample is representative of a group. It's a fractal thing; small parts of the whole show the same structure as the whole.

Voting is a form of polling. So, one could walk into any polling place and grab a hundred or so ballots and burn them without affecting an election outcome (given that one hundred is a small sample) since that sample has the same outcome as the election as a whole.

The best thing that could have been done with the 2004 Florida ballots spoiled by chads would have been to burn them all. If they were genuinely a random sampling of voters and contained the same ratio of Bush to Kerry voters as clean ballots then destroying them would make no difference to the outcome.

The only reason to fight for every precious chad is if they were not random and were not representative of the general vote but, were highly skewed to one candidate. How would that happen?

There are two companies that make punch card voting sets. I don't know which was used in Florida and that doesn't matter since they are comparable. There are three machines in the set. One is the voting punch which is entirely manual. Two is the electronic reader. Three is a machine used for testing and this is where it gets interesting.

The test machine is a power operated ballot puncher that is used to make test sets of ballots to check the readers. It is capable of punching through a stack of ballots. There can be problems trying to punch through too thick a stack of ballots in which case you get - wait for it - hanging chads.

Just sayin'

At 12/28/2007 6:12 AM, Anonymous Anonymous said...

You said: "What this means is that if the first 16,000 people who vote when the polls first open in the morning are random voters who represent the population of voters, almost all elections are already decided by 9 a.m. or so in the morning."

But since there is no reason to believe they are random voters, none of your statement makes any sense.

At 10/24/2008 7:17 AM, Blogger Scoprire Mi said...

your analogy to blood sample doesn't works for me at all. Definitely collecting say 2 pints vs 20cc of blood from the SAME person will not affect the reason. But we are not talking about getting MULTIPLE vote from the same person (which I am sure you know is not legal). Are you assuming that everyone 'thinks' the same???
I also don't understand your theory of results being decided by 9am. I do accept that I don't understand all this probability theories but when it comes to Human mind and instincts no math theories can be applied. Minds of 169 million cannot be represented by 16thou people even if they are RANDOM.

At 4/29/2009 4:32 PM, Blogger Unknown said...

I believe a large portion of the people that don't vote in our presidential elections don't do so because of the electoral college. If you are in a blue state voting red, why bother, and vice versa.

I don't think that these non-voters would not make a difference in the popular vote though. We're talking about somewhere in the neighborhood of 150 million votes not being cast. In the 2008 election, there was a difference of 8.5 million votes. That means there would have to be less than a 6% difference in voting among the now non-voting population to change the victor of the popular vote. I don't think that is too outlandish.

Of course this is only the popular vote. Maybe it wouldn't make a difference in the electoral college. Although, with a voter turnout similar to Italy or Australia, maybe we'd realize how many people are not being counted with the electoral college even after they vote.


Post a Comment

<< Home