Polls, politics, elections, turnout and sampling, or Why polls really don’t make a bit of difference



I’ve stayed away from politics on this site, the main reason is I figure the surest way to offend visitors is politically, these days so many people seem so starkly divided into one political camp or the other, that it seems very easy to step on peoples political sensibilities. I am, however going to talk about politics, but mainly in the general sense to illustrate something I’ve thought about for quite a while and something that is grossly absent in most news coverage of polls and elections. With all science there are assumptions, things that you “assume” to be true so you can proceed with evaluating everything. In the science of polling, the words Sampling and turnout are quite relevant. (What?) In regards to polls, sampling is deciding what the makeup of your poll sample is, (percentage makeup by party affiliation) and trying to match that to expected turnout. But how do you predict turnout? Let’s take the following example of cartoonia….


First, let me mention that Donald Duck and Disney are registered trademarks of The Walt Disney Company, Daffy Duck and Merrie Melodies are trademarks of Warner Brothers. No mice, rabbits or ducks were harmed in the generation of this data, no offense is intended to the Daffy followers, or the Donald voters. It should be noted that Merrie Melodies is sometimes better known as Looney Tunes… however, well, it’s easier to talk about the Merrie Melody party than the Looney Tunes party with a straight face. No, I am not affiliated with either of the above companies (outside of possible stock ownership.) Finally, NO, there shouldn’t be any offence taken, or read into the numbers, percentages, party names, etc. All were arbitrarily chosen and no intent was made to label any real, living, person or party, blah blah blah.

Cartoonia has 1000 registered voters
There are two main parties – the Disney party and Merry Melody Party. The two parties have selected Donald (D) and Daffy (MM) as their candidates
Now, 37% of the registered voters identify themselves as (D) party members and 34% identify themselves as (MM) party members the rest are independent

RV 1000

Total D MM Independent
100 37 34 29 Percentage of registration
370 340 290 Total registered number of voters by affiliation
90.00% 90.00% 56% Daffy (supporting their own parties candidate)
With the independants, we see 56% supporting Daffy.
333 34 128 Loyalty to candidate Donald (actual #s)
37 306 162 Loyalty to candidate Daffy

495 49.9% Total Donald
505 50.1% Total Daffy

So, Daffy and the MM party should win by 10 votes… 505 to 495, but…. not everyone votes that is registered.

Hmmm…. let’s say 30% of total voters are expected to turn out, so only 300 individuals are expect to decide this race.
Who will turn out?
Will it be 37% Disney party, 34% Merry Melody party and 29% independent perfectly reflecting the actual voter registration?
Likely not.

Let’s say, for the sake of argument, 38% of our voter base turns out to be Disney party voters (good “ground campaign” and “get out the vote” efforts).
We’ll also say that Merry Melody turnout is pretty good as well, making up 36% of the electorate and the independants with no organized effort to motivate them make up the remainder
Turnout 300

38 36 26 Percentage of turnout
114 108 78 Real numbers
90.00% 90.00% 56% Daffy Same percentage of
support for their own candidates just to keep things simple.
103 11 34 Loyalty to candidate Donald (actual #s)
11 97 44 Loyalty to candidate Daffy

148 49.90% Total Donald
152 50.10% Total Daffy

A much closer election, and this time Daffy prevails, but it’s rare to see both parties turn out to exceed their actual registered base.
300

39 33 28 Percentage of turnout
117 99 84 Real numbers
90.00% 90.00% 56% Daffy Same percentage of
support for their own candidates just to keep things simple.
105 10 37 Loyalty to candidate Donald (actual #s)
12 89 47 Loyalty to candidate Daffy

152 50.10% Total Donald
148 49.90% Total Daffy

This time, it’s another close run, but the Disney party prevails with their candidate Donald due to very good turnout efforts and win by 4 votes.
Turnout can vary MUCH more widely than this, but there are other factors, maybe the party loyalty among those turning out is higher for one party than another?
Maybe the distribution of independants is different for those actually turning out compared to those in the registered voter pool.

Now, for the original percentages of how many in the Disney party support Donald and how many in the Merrie Melody party support Daffy, there is no way to know the accurate number unless we talk to each and every one of the registered voters (which we’ll assume we have.) The same is true of the independants. It should also be noted that for simplicities sake, I’ve kept the percentage of independants favoring Daffy the same in all turnout scenarios which is likely unrealistic, as is the 90% “intensity” or loyalty for each party in actual turnout.

So, the bottom line here is polls don’t matter, that is unless they can accurately predict the percentage and intensity (loyalty) of the voters that turn out, which is to say, it’s not an easy thing to do. Sometimes polls will be close, but as we’ve seen in recent years, sometimes polls will be wildly off. I think in the 2004 election there were several examples of polls within a week of the election that were off by 5-10% due to incorrect assumptions about turnout.

How do you predict turnout? Likely voters, who is a likely voter, one that votes every time the polls are open? are there enough of those for a decent poll? What about those that are only motivated once in a great while, they could shift the balance as well. So, this is why polls fail to predict the future with great accuracy consistently. Yes, they may get it right occasionally, but there’s an amount of luck involved as well as science.

So, sampling and turnout is what it’s all about and the only poll that really counts in ANY election cycle in cartoonia or in the real world is the one where you actually go out and mark the ballot. I’m not going to take this opportunity to suggest that you go out and blindly vote, but I am going to suggest you learn about each candidate running in each of the races and vote accordingly, try to get information from multiple sources and try to think critically about what they’re saying. It can be hard to cut through the crap of political slogans and soundbites and get to the actual policies that people want to see implemented (or philosophies of judging in the case of judicial candidates.) But, this country and our freedoms depend on an INFORMED electorate taking part to shape the countries future.

   Send article as PDF   

Similar Posts