Tuesday, November 02, 2010

Elections and Blood Tests

Earlier today, I made the case for low voter turnout and argued high voter turnout is overrated, wasteful and inefficient because we would almost always get the same exact election results with a significantly lower voter turnout.  I compared elections to a blood test - in both cases you only need a small sample, and the results won’t change with a larger sample.

Here's a good quote I found today to help illustrate the point:

“Next time someone tells you they don’t believe a small sample poll can possibly tell you anything, just say to them ‘OK, then. Next time you have to have a blood test, why don’t you ask them to take the whole lot?’”

British opinion pollster Nick Moon, in Significance, March 2010


25 Comments:

At 11/02/2010 9:19 PM, Blogger Jet Beagle said...

Professor Perry, I do not understand how that is a good analogy. Blood is circulated throughout the human body very rapidly. The blood cells which were isolated in one's finger a couple of minutes ago could very likely be found now in every organ and appendage of the body. Blood is relatively homogenous.

 
At 11/02/2010 9:27 PM, Blogger Mark J. Perry said...

This is all based on statistics and sampling theory.

A randomly selected sample of voters at a sufficient number (1,000 or more) should be homogeneous as well, and accuratley represent the entire population of voters. Most national polls are based on samples of 1,000 randomly selected people, and will accurately reflect the population.

 
At 11/02/2010 9:32 PM, Blogger Jet Beagle said...

This comment has been removed by the author.

 
At 11/02/2010 11:10 PM, Blogger nightlysok said...

And know we know how economists come by their theories.

 
At 11/03/2010 12:28 AM, Blogger Jet Beagle said...

Come on, Mark. I’m not arguing against the validity of scientific sampling. I don’t have an economics PhD, but I do have a degree in mathematics and an Ivy League MBA. I think I understand a little bit about statistical sampling.

I agree that a small scientifically-designed sample can be used to predict an election result. But I don’t think that is the same as arguing that lower turnout and higher turnout will produce the same results.

The issue, for me, is whether the additional voters in a high turnout response are representative of the voters who will vote in a low turnout election. Older voters have very similar participation rates in both low and high turnout elections. The additional voters in high turnout elections tend to be younger voters.

I found voter participation statistics in the 2010 U.S. Statistical Abstract. These statistics were compiled using scientifically designed samples. I am not rejecting sampling in making my argument. Rather, I am using Census Bureau sampling to show that the composition of voters varies significantly when overall turnout varies.

 
At 11/03/2010 3:24 AM, Blogger Emil Perhinschi said...

Voting is about legitimacy, not about statistical relevance. One vote is not relevant, but it can win an election ...

Democracy (direct or by representatives) is a ritualized civil war: instead of fighting it off and the strongest party hanging the heads of it's opponents in front of the party offices, supporters are counted and the winners are those that managed to mobilize more support. Probably more expensive than a civil war, but everybody gets to live to the next round and the losers get another shot :)

 
At 11/03/2010 5:23 AM, Blogger juandos said...

Actually the only case you made Professor Mark is the rationale for not voting 'if' one is only voting for someone (not something) where more than a few thousand people participate...

The rationale 'seems' fail on the local and county level and fail rather badly at that...

 
At 11/03/2010 7:41 AM, Blogger Scott said...

Blood doesn't possess free will.
Blood is different than people.

 
At 11/03/2010 8:50 AM, Blogger morganovich said...

statistics theory is all well and good, but you are making some heroic assumptions.

first off, you are assuming there is such a thing as a random sample when we all know that that simply is not true. you get huge skew from issues like who is home to answer a call or who wants to participate in a survey. second, you are assuming a homogeneity that does not exist. it varies a great deal, and statistical sampling may amplify rather than reduce such issues. third, ideas like "likely voters" are all well and good, but intent and performance wind up being quite different in many cases. people lie to polls for personal reasons (like not wanting to look like a non voting slacker) the polls have been wrong in many tight races over and over. inevitably, the polls will be wrong in a number of races. this is why we cannot use them.

sure, you can see a landslide coming, but in a tight race, polls miss all the time. a 4% margin of error in a 1 point race is pretty useless.

your analogy to blood is not a good one. pathogens do spread with homogeneity in many cases and many just need a tiny hint to be found to be conclusive, but even with these advantages, many things cannot be determined by blood because they are localized.

elections are the same way. are you telling me that we would be as well served by polls in the bush/gore election or this year's CA AG election?

you are falling into the classic economist trap of assuming the terrain is the map. we all know that price does not = marginal cost in many businesses because competition is not perfect. you are making the same sort of counter-factual claim here by making assumptions about samples sets.

i see what you are trying to say about statistical sampling, but you are taking it much too far here. as my statistics professor used to say (and hopefully still does) "it is more important to understand the limitations of statistics than their capabilities. this will prevent you from believing stupid things."

 
At 11/03/2010 10:26 AM, Blogger morganovich said...

here's another one:

http://www.sfgate.com/election-results/2010/11/02/CA/c/h_u_s_house/h_11_district_11/g_general/c/california.shtml

121 votes in difference. you really think a poll is accurate enough for this race? (7/100 of 1 percent)?

do the math yourself on how big a sample set you'd need to get that level of accuracy. it's pretty much the whole population, which, obviously, you could never sample. far more people vote than are willing to answer pollster calls.

elections like this are a clear demonstration of why this this "polls are as good as elections" argument is wrong.

you cannot possibly sample well enough to get the kind of accuracy you need to call this race.

this also underlines the fallacy of the "your vote doesn't count" camp.

if 0.001% of voters read your article and stayed home in this election as a result, that could have turned the race.

 
At 11/03/2010 10:28 AM, Blogger Rand said...

morganovich:

Socialism is a political pathogen that spreads through populations like a contagious disease.

 
At 11/03/2010 10:42 AM, Blogger James Fraasch said...

So let's see. The SAMPLE Nevada Senate race poll showed Angle with a 2.7% lead up until Friday.

Harry Reid won by 5.6%. That there is an 8.5% difference between the sample and the results.

Republicans sure would have loved to have used the "Sample" results instead of the actual results. Democrats are damn sure happy they had an actual election instead of taking a "sample" of likely voters.

Really, does anything else need to be said?

James

 
At 11/03/2010 11:49 AM, Blogger juandos said...

Harry Reid had Harrah help

 
At 11/03/2010 12:07 PM, Blogger Hydra said...

We heve enough example of hundreds of thousands of votes being cast and the result being determined by less than a hundred votes.

You can't simulate that with a sample of 1000.

I suppose you could claim it doesn't make much difference: either way roughly half the population is going to be pissed off.

 
At 11/03/2010 12:09 PM, Blogger Hydra said...

Socialism is a political pathogen that spreads through populations like a contagious disease.

=================================

Sounds like a successful pathogen.

It probably got that way by competing in a free market without regulation.

 
At 11/03/2010 1:35 PM, Blogger morganovich said...

hydra-

so did ebola and HIV.

that merely makes them successful competitors for resources. they, like socialism, still kill the host in the end.

 
At 11/03/2010 1:57 PM, Blogger juandos said...

"that merely makes them successful competitors for resources. they, like socialism, still kill the host in the end"...

Ooh! Ooh! That's going to leave a mark!

Good one morganovich!

 
At 11/03/2010 2:29 PM, Blogger ExtremeHobo said...

I think this whole thing misses the point entirely only because I would perfer an even smaller amount of people vote. Would you rather have the opinion of 500,000 informed voters or 300 million idiots?

 
At 11/03/2010 3:20 PM, Blogger morganovich said...

extreme-

so who gets to decide who is informed and who is an idiot?

 
At 11/03/2010 5:55 PM, Blogger Jason said...

Something, a very practical issue of sampling, that has not been addressed is the political screwery that would inevitably come with determining the sampling methodology, For instance, in the original post, if the first 16,000 voters determined the outcome, there would be a wild scrum to be one of the 16,000. Talk about unintended consequences,

Not to mention, in Florida in 2000 weren't there 5.8 million votes cast? And wasn't the election determined with 537 votes? In order to devine the will of the people you would need a sample size of...well greater than the number of people that actually voted given the uncertainty of the measurement (counting isn't always precise!).

Elections are messy, and not always certain, but they are the most straightforward method there is. Let's not fix it.

 
At 11/04/2010 5:59 AM, Anonymous Anonymous said...

I don't think you could take the first 16,000 voters unless people who get up that early represent the entire group. Do Democrats sleep later ;) Better to spread the 1600 over the number of hours the polls are open. We would still have statistics v. reality.

The Congressman I actively campaigned for won simply because he got more votes than the other guy. I wonder if more people who would have voted for the other guy stayed home because they were not one of the first 1600 to the poll and figured their votes would not count?

 
At 11/04/2010 6:57 AM, Blogger juandos said...

"so who gets to decide who is informed and who is an idiot?"...

Well morganovich sometimes we have to make that type of decision even if we don't want to...

We can't physically force anything but its still a decision that every individual needs to make...

Remember this clip Professor Mark posted previously: Is Obama A Keynesian? Rally For Sanity, 10/30/10

I'm frustrated and a bit angry that these clowns also have the right to vote...

Now a days (and for quite some time now) there has been NO excuse for that sort of abysmal ignorance...

 
At 11/05/2010 3:24 PM, Blogger ExtremeHobo said...

morganovich - I would say some sort of basic political test at the polls would whittle it down a bit and we can work from there.

My biggest peeve is probably that people can vote on referendums without ever reading them. Any referendum that sounds nice in one sentence always gets passed. As part of my above test you should have to list the basics of every referendum on the ballot before you can vote on them.

 
At 11/05/2010 5:57 PM, Blogger Ron H. said...

"My biggest peeve is probably that people can vote on referendums without ever reading them."

People can also vote for any candidate they choose without knowing anything about them, as was demonstrated in 2008.

Various voter tests have been tried in the past, but all have run afoul of the US Constitution, due to their discriminatory effects against one group of people or another.

So, even though one might believe that better outcomes would result from some sort of testing,, there can be no discrimination against ill-informed or stupid people.

 
At 11/08/2010 7:44 PM, Anonymous Anonymous said...

I'm a clinical pathologist and a statistician, and I emphatically state that the analogy between a blood sample and a voting or polling sample is incorrect.

Blood, for most analytes, is a homogenous mixture. You could take hundreds of small samples from a person and get essentially the same test results from each sample.

Voters are a heterogenous mix of people. Small voter sample sizes can yield results significantly different from the results of all voters. Also, choice of candidate and likelihood of voting probably are not independent variables. Eligible voters who favor Candidate A may be more likely to vote (for whatever reasons) than eligible voters who favor Candidate B. Because of these factors, a vote with a low turnout may not reflect the wishes of the majority of eligible voters.

 

Post a Comment

<< Home