Conditions Ripe for Social Desirability Bias

Election 2016           Social Desirability Bias / Overview:  Conditions Ripe for Social Desirability Bias

Poll bias is alive and well.

Different polls should show more or less similar results if there is no bias involved.  If polls begin to vary considerably for an apparent inexplicable reason, more research is needed to understand why.  It is somewhat shocking that more resources have not been placed towards this during the 2016 election season as polls are varying for specific topics and candidates depending on how the data is collected.  This should be a huge red flag.

There are multiple ways of collecting data for polls.  Traditionally, most mainstream polls use live interviewers using telephones to ask people standardized questions.  An alternative that has been growing in popularity uses methods that do not use live interviewers but instead ‘robocalls’ / IVR (Interactive Voice Response) calls or internet collection.  Many times, the questions asked are remarkably similar but the end results are considerably different.  This is very odd.

The difference appears to be that individuals feel constrained when speaking with a live interviewer.  In contrast, those answering via a more anonymous method answer in a more unfiltered way.  By taking the difference between the ‘live-interviewer’ polls and ‘anonymous’ polls we can indirectly measure the degree to which people are changing their answers possibly due to feeling more constrained.  A difference between a few polls really does not mean anything – but if approximately the same difference continues over time for the same topic or person in a variety of polls, it is likely a bias issue.

Social Desirability Bias is a social sciences theory and academic area of study.  It supposes that some people will change answers to surveys or polls if they believe that a more socially acceptable response makes them look better.  This phenomenon has been confirmed for many areas including personal finances, diet, recreational use of drugs, exercise, sexual practices, and political leanings.  Essentially it has been shown to exist in areas where the respondent feels real or imagined social pressure to respond in a certain way.

So, an individual might feel uncomfortable admitting to something and change an answer to ‘correct’ for this.  There are ways around this bias problem.  More anonymous data collection methods decrease the tendency for individuals to feel social pressure and to provide more truthful responses.  Additionally, academics use a series of questions designed to detect inconsistencies to help to identify biased responses.

Social Desirability Bias exists in politics.  It is often referred to as the Bradley Effect after Tom Bradley who lost the 1982 Los Angeles Mayoral race, even though he was leading in the polls.  The idea is that voters said in polls they would vote for Bradley because they felt social pressure to publicly support the African-American candidate but then did not follow through on such claims as he lost the election.  This racial explanation, if correct, apparently did not impact polls regarding Obama as there was no discernible difference between polls, results, and exit polls.

A focus on race seems to completely miss the point and referring to the Social Desirability Bias in politics as the Bradley Effect seems misleading.  People tend to change their answer when they feel social pressure and that social pressure needs to be reasonably strong.  Social pressure can be created in many ways.  Using a racial explanation to describe such a large phenomenon is misleading, especially when race was shown to not have an impact on how respondents answered polls regarding Obama.

As Social Desirability Bias relates to the 2016 election, there is considerable speculation that Trump is receiving artificially low poll numbers and Clinton artificially high numbers as people are changing their responses due to social pressure.  This fits the definition of Social Desirability Bias.

The theory states that greater social pressure will result in more greatly modified responses.  Before looking at data for proof that a bias exists, we should stop to see if there are any signs that people could feel social pressure to modify their responses either to pro-Clinton or anti-Trump during this 2016 election season.

 

  • Another Groundbreaking ‘First’ Candidate – Clinton is the first major-party female presidential candidate in the US and there is considerable coverage surrounding this fact, especially after such a feel good environment in 2008 when the US elected Obama, the first non-white, there seems, by many, to be attempts to draw parallels between electing the first female in 2016 and the first non-white in 2008, as such many feel social pressure to support Clinton as a way of ending sexism,
  • “Special Place in Hell” – was the comment that former Secretary of State Albright used to comment on women who might not support Clinton, this is just one example but there is an element of shaming of women who do not support Clinton,
  • ‘Racist’ Candidate – Trump has been repeatedly accused of racist tendencies making it potentially socially difficult for minorities or whites fearful of this label to publicly support him,
  • ‘Misogynistic’ Candidate – Trump has been repeatedly accused of ‘hatred of women’ making it potentially difficult for women, who make up the majority of the registered voters, or men afraid of this label to publicly support him,
  • Sitting President’s Involvement – Obama has entered the race like no previous sitting president, aggressively attacking Trump and imploring Americans not to elect him, as a fairly popular president Obama’s statements create an unique form of social pressure against Trump,
  • Obama vs Trump Public Feud – rare is the election where there is such a personal, let alone political, feud apparent between one of the candidates and the sitting lame duck president,
  • Traitor to the Cause – many Democrats feel that not electing a Democrat in 2016 would essentially be a vote against Obama’s legacy, as the first African-American president there is an underlying accusation of racist leanings of those not openly supporting Clinton,
  • “Insult to my legacy” – Obama confirmed the general Democratic feeling that African-Americans must support Clinton when he stated that “I will consider it a personal insult, an insult to my legacy, if this (African-American) community lets down its guard and fails to activate itself in this election.” Obama’s clear request puts considerable social pressure on African-Americans,
  • Divided Country – by most accounts, the US has become more politically divided over the last few decades making choosing a side or even changing a side much more disruptive than it had been, which provides further incentive for some to choose the most socially acceptable candidate,
  • Media Bias – by most accounts, many or even most of the major media outlets have a distinct political preference and have left neutral journalism behind making coverage more pointed and talking-points more radicalized on both sides, there are more liberal-leaning media outlets creating a greater negative bias towards Trump,
  • High Unfavorable Ratings – both candidates have extremely high unfavorable ratings making a tactic of shaming supporters of opponents effective, which likely creates enough social pressure to at least temporarily change voting preference in a poll,
  • Identity Politics – has become a major strategy for politicians to target votes within specific demographics which places social pressure within those groups to vote with their group, Democrats have much more effectively used identity politics and those target demographics likely feel pressure to support Clinton, this is especially true in this election for the female, Hispanic, and African-American demographic groups,
  • Important Election – both sides view this election as the most important of the last few decades or from the start of the data, creating more social pressure around voting intentions and being seen as supporting the ‘right’ candidate,
  • Violent Election – violence has been shown on both sides during this election, much more than in recent memory, the balance seems to be against those supporting Trump,
  • Job Losses – there have been multiple examples of individuals losing jobs during this election cycle due to support for a particular candidate but seemingly more against those supporting Trump,
  • Breaking Social Ties – surveys show that in 2016 the election is so emotional that many in the US would cut social ties with friends due to voting intention that is if they found out their friend planned to vote against a preferred candidate they would simply ‘unfriend’ them. Liberal voters score higher on this metric which implies that more Trump supporters, if they knew or inferred such a consequence, could keep their voting intentions quiet or private or avoid admitting Trump support to forgo losing friends,
  • Social Media Megaphone – social media makes any position taken potentially very public and permanent, much more so than in previous elections, people have more of an incentive to stay as neutral as possible or to go with the most social acceptable response,
  • Sidewalk Chalk Reaction – pro-Trump messages written with temporary sidewalk chalk caused traumatic responses on multiple college campuses, such extreme reactions to a seemingly innocent statement could make Trump supporters feel more pressure to remain anonymous, especially those living in predominantly Democratic environments,
  • Political Sign Stealing – there have been numerous cases of political yard signs being stolen or vandalized, though appearing to have hit both sides, the number of news articles appears to show a national bias against Trump in terms of problems with political signs,
  • Strong Partisan Tendencies – it was not so long ago that states would periodically swing between Democrats and Republicans, in the recent past however the lines have firmed and most states seem to strongly lean in favor of one, those in certain geographic areas also likely feel more social pressure such as those in ‘Blue States’ to support Clinton and ‘Red States’ to support Trump,

 

There appears to be a very strong argument that a relatively high level of social pressure exists in the 2016 US presidential race.  Certainly, there seems to be pressure to support Clinton for women and those who voted for Obama.  And, there seems to be pressure not to support Trump as there are intense accusations regarding his character and as there are very negative reactions towards public support of him.

So, before even looking at data, we can more or less expect there to be an element of Social Desirability Bias involved.  The real questions then become what is the size of this bias and will it be enough to make polls inaccurate?