Reasons to be skeptical about poll numbers in the 2020 Presidential Election
When you are skimming through a news article and a percentage number catches your eye, it is easy to take it at face value; for example, if a poll states that 75% of people prefer prawn and cocktail flavor crisps, you could deduce that this flavor is the most popular. Remember this analogy, as it will be used later.
Even with election polls being as popular as they are as an indicator of a candidate’s predicted success, there are many variables that can impact upon these numbers, thus making them a bit less reliable than they initially appear. The most famous example, of course, being the polls surrounding the 1936 electron between Franklin D Roosevelt and Alf Landon; polls conducted by The Literary Digest pointed at an almost certain win for Landon, but when the voting was tallied, Roosevelt had won.
This is an extreme example but many polls used today suffer from similar issues. In this article, it will be explored why it is best to be skeptical about poll numbers in this year’s Presidential election.
Larger sample does not equate to better accuracy
So, back to the crisps analogy; suppose this survey was conducted on a sample of 200 students in a cafeteria. That is a fairly good sample size, but does it make it more accurate?
If an electoral poll boasts having a huge sample that makes it the best predictor, it may be using a cheaper and more error prone sampling method, which may be making it a very unreliable predictor of the final outcome.
And, of course, this is before the sample itself is examined in more depth. In an ideal world, each poll or survey would have a representative figure of the population being explored, but as many polls represent a particular group, they do not account for the everyday concerns of the common registered voter. Ergo, you may believe that 75% of people prefer prawn flavoured crisps, but the survey was conducted in a cafeteria where it was the only crisps that were available.
Different polls conduct different surveys
Anyone who has studied a subject which involves collecting data to represent a population is aware of the problems which may arise with wording, or even the type of survey used.
Let’s say there were 2 media outlets who wanted to discover which flavor of crisps were the most liked; 1 media outlet used a phone in poll where they asked their callers to explain why they liked a certain flavored crisp. The 2nd outlet used an online survey to gather the information, through a general online questionnaire.
As you can already see, these are 2 very different approaches to answering the same question, and both present their own problems with accuracy; for the first outlet requires the use of a telephone and a television to watch the call in show, so may only represent people who watch TV at a certain time of day.
The 2nd type is more likely to reach younger voters as it is an online survey, potentially reducing the input for older people who may not have access to a computer, thus creating a result biased in their favour. And, in the US elections, different media stations use different polls and sampling, creating an end result which has more than a few issues.
Confidence affecting outcome
Many statistical studies have found that if a certain candidate is given the best odds of winning, then those who support that candidate may be so assured in their victory, that they don’t vote.
If you were told time and time again by political analysts that your favourite flavoured crisp was going to win in an election, this may prevent you from taking part in an actual vote.
But if you were approached in a poll before the vote, you would still state that a certain flavour was your favourite but if asked to vote, you would already have ‘inside information’ from the analyst that the race was effectively over and your favourite flavour had won.
Scale this interesting piece of human psychology, and the end result is a poll which states one thing, and an election result which says another.
In fact, some political analysts state that due to this factor and its impact on accuracy of an election outcome, traditional polling estimates and margins of error should be used to gauge popularity of one candidate over another.
And so, this upcoming year, take each Presidential election poll you read with a pinch of salt; they have different samples, different ways of measuring that sample and as is the way with the prawn flavored crisps, those who took part in the poll may be so confident that their favorite flavor is everyone else’s, that they won’t actually cast their vote at all.