Home / Articles / US Election 2016 – Social Desirability Bias and Betting on Elections

US Election 2016 – Social Desirability Bias and Betting on Elections

If you have been following our twitter since we launched a few days ago, you would have seen us go 3/3 in terms of our recommendations, and even though Trump to win the US election was paying over $5.00, of all of our recommendations this was the one we were most confident in.

If you know what to look for, elections are the easiest market to predict. By far. And this US election was no different.

If you have been closely following the US election betting markets over the last few months you would have seen them swing wildly as the narrative changed from FBI investigations to offensive remarks about women. They swung as each new scandal hit the news, but ultimately they did a very poor job of predicting the eventual outcome, just like the polls they were based on.

The first thing you have to understand is that the people that set and bet on political betting markets have no special insight beyond a reasonable understanding of politics, their understanding of what constitutes fair odds is based only on the data they have available to them – mainly political polls. That isn’t to say they don’t take into account a margin of error, as well as the possibility that vote-affecting events will occur in the lead up to the election, but underpinning it all is how well a candidate is polling.

In many cases betting markets both reflect and predict polls, as the first thing to change when a scandal comes out is the odds, which will then correct once polling data follows. But what the polls, and therefore odds don’t take into account, not for Brexit, nor for the 2016 US election, is that the polls they are based on are susceptible to an intrinsic bias that boils down to two things:

  1. Some people will poll one way but vote another
  2. The sample of people polled is not a good representation of the people that will vote.

Whilst it is quite possible that Trump voters, given the very different age and ethnic makeup to Clinton voters, would be less likely to be answering polls, this really depends upon where and when the poll was conducted, and how people were targeted. We could see no compelling reason why polls would significantly favour one candidate or the other, particularly given the veracious nature of US polling where some polls even attempt to get a representative (rather than random) sample.
So, if we assume there is no major difference in the demographic of people polled, why would a person poll for Clinton but vote for Trump you ask?

We propose one key reason:

Social Desirability Bias

Which is a bias that posits people will answer polls in a way that will be viewed favourably by others- to present themselves in the best possible light in an effort to deceive others (and possibly themselves). Given Trump’s controversial stance on a lot of issues, and unpopularity with demographics likely to be conducting polling (young adults) we predicted that there would be a significant number of people lying at the polls about the fact that they would be voting for Trump.

If this hasn’t completely convinced you, look up the ‘Shy Tory Factor’, which is a name given to the phenomenon of consistent poor polling of the UK conservative party relative to the share of the actual vote.

The prevailing theory with the Shy Tory Factor is that the more vocal liberal ‘leftists’ have assumed the moral high ground and shamed certain demographics (such as the wealthy older generation) for voting for what has been branded ‘selfish’ conservative policies, such as reduced income redistribution and lower taxes, to the point where they seek to avoid stigmatisation by outwardly identifying assupporting liberal candidates. Of course, in the secrecy of the ballot box they do not feel this pressure.

Election Betting - Tips for Betting on politics
Predictive modelling based on polling data had one fatal flaw- not understanding that the Trump Street vote did not reflect his true popularity

When you are next considering betting on an election, we strongly suggest you take polling data as your starting point and then ask yourself if this is representative of how the people will vote, and if not why it would be different and to what extent. Picking an election result can be done to a much higher degree of confidence than any sport, and is truly the easiest betting market to predict. You just need to understand the polls, and how biases might mean they differ from the actual vote. Based on Trump and Brexit you have no excuse for not being aware of one key factor- that controversial policies and politicians may poll worse than safer, ‘moral high ground’ options, but that doesn’t necessarily mean that people aren’t voting for them in the safety of a confidential voting booth.

Check Also

betting cricket india australia kohli smith

Betting Preview: 2nd Test India v Australia

First Test Recap No one saw the result in first test coming. The Australian team …

One comment

  1. Loving this blog. Agree about Trump, we see it a bit in Aus as well with small right wing parties always doing better than expected at the Polls

Leave a Reply

Your email address will not be published. Required fields are marked *