Election 2016 Social Desirability Bias / Overview: Summarized Findings
The conditions are ripe for people to feel social pressure to vote a certain socially desirable way. Americans do not appear to be actually changing who they vote for, simply changing who they say they will vote for when in front of other people. In private, like in an anonymous poll, and presumably in the voting booth they will revert to their true leanings.
According to many sources, the 2016 US presidential election is the most emotional, violent, and important (or at least perceived as such) in decades. The two main candidates also have the worst unfavorable ratings in memory. Add to this a very opinionated media, an unusually vocal sitting president, and a sharp increase since the last election of the use of social media (which can make an individual’s political opinions extremely public and permanent), we get an environment in which many people will look for a socially desirable answer in place of their actual opinion.
From research conducted comparing live polls, anonymous polls, and when appropriate primary results, the following conclusions are made:
- Obama’s Approval Rating – is artificially high due to an extremely strong Social Desirability Bias with live polls putting his approval rating at an astounding 11 percentage points higher than anonymous polls in August 2016. The current bias appears to have been created by Obama’s aggressive anti-Trump attacks and the media’s mostly supportive coverage of Obama and negative coverage of Trump,
- Approval of Congress – was severely impacted by Social Desirability Bias right after the mid-term elections which saw the Republicans perform unexpectedly well. Live polls showed average disapproval at a rate of 11 percentage points worse than anonymous polls after the election.
- Republican Primaries – showed that Social Desirability Bias played against Trump, on average his live polls were approximately 6 percentage points lower than anonymous polls. The bias was the worst for Trump in traditionally Democratic states of the Northeast and Rustbelt where on average his actual primary results were 10 percentage points higher than live polls.
- Trump versus Clinton Polls – in August and September, Trump has an approximate 4 percentage point negative bias against Clinton in head-to-head polls, inferring that live polls provide a significant yet superficial boost to Clinton.
- Clinton Bonus Bias – Clinton actually benefits in some demographic groups from Social Desirability Bias. Her recent live poll figures were 6 percentage points higher among women and 10 percentage points higher among African-Americans when compared to anonymous polls.
- Third Party Candidates – have benefited from Social Desirability Bias as many look to ‘hide’ by publicly supporting lesser known candidates which are seen as more neutral than the two main candidates in order to avoid social condemnation, the positive bias as measured by the difference in aggregate support in anonymous polls for Johnson and Stein minus live polls amounts to 4 percentage points.
- Undecideds – is extremely high during this election cycle, anonymous poll undecided levels are 7 percentage points higher than live polls.
The difference between what people admit in front of others and what they really believe appears to have increased when it comes to US politics. Though Social Desirability Bias has always likely existed in politics, the conditions seem to have become more supportive recently. The beneficiaries appear to be those topics and people supported aggressively by the media and general opinion makers. The opposite is true for those depreciated by the same.
As it stands currently, the beneficiaries of Social Desirability Bias are Obama, Clinton, and the Democratic Party. For Trump and the Republican Party, Social Desirability Bias is a significant liability.
As it impacts the 2016 US presidential election, you should note that reporting on the election and most of the well-known polls are likely artificially supportive of Clinton. Specifically, the following corrections should be made:
- Female Vote – live polls on average are overestimating the support of women for Clinton, use anonymous polls or use a discount on support for Clinton when using live polls to correct for this,
- African-American Vote –live polls on average are overestimating the support of African-Americans for Clinton, use anonymous polls or use a discount on support for Clinton when using live polls to correct for this,
- Trump vs Clinton – focus on anonymous polls, or use live polls only after discounting Clinton’s support,
- Northeast and Rustbelt – live polls on average are discounting Trump to a significant degree in these traditionally solid Democratic states, Trump could pull off upsets if anywhere near Clinton in the polls in these states,
- Third Party Candidates – live polls on average are providing a significant boost to these candidates, either use anonymous polls to estimate support or apply discounts on live polls,
- Undecideds – are likely much higher than live polls suggest implying that this race still contains a significant amount of potential surprise left.