Lexi Gray, Lauren O’Shea, Adam Cohen, Kyle Wolfe
Social desirability bias has long existed, both in the United States and in other countries. People are always afraid of being judged for choices not deemed socially acceptable, even in a phone poll by someone they have never and will never meet.
Historically, this effect has badly thrown off polls. For instance, in the United States, social desirability bias against African-American candidates has historically been prevalent, although there is some debate on whether or not there is still an effect anymore. For instance, Tom Bradley, an African-American mayor of Los Angeles, ran against a white candidate, George Deukmejian, for the governorship of California. Although Bradley lead significantly in polls before the election and exit polling, to the point where newspapers declared him the winner, he narrowly lost. Social desirability bias explains this effect perfectly, as many poll respondents could have felt that they would be judged for voting against a black candidate, even if they planned to do so.
Social desirability bias is not a uniquely American phenomenon, nor is it necessarily tied to issues of identity, such as race or gender. For instance, Brexit polling showed strong effects of social desirability bias.
In the above graph, we can see that Remain is much more heavily favored in telephone polls than in online polls. Social desirability bias affects telephone polls more than online polls because the feeling of being judged by an interviewer is greater than when filling out a form online. In the case of Brexit, many interviewees could have not wanted to reveal their support for leaving the E.U., a very controversial decision.
Social desirability bias occurs when an individual answers a poll, survey, or other mechanism of public feedback with answers he or she perceive to be socially desirable regardless of his or her own true opinion. This bias can happen quite often around sensitive questions regarding race, gender, gay marriage, and other demographics. Rather than admit personal beliefs and biases regarding these sensitive topics, some people may resort to giving answers that appease the individuals asking them. The importance of environment also plays a key role in determining the potential for social desirability bias within a poll; for example, in a classroom setting at Bowdoin College, a student might shy away from admitting to support Donald Trump due to the college’s reputation as a liberal school. In this instance, by not stating their true beliefs, the student instead appears to favor Hillary Clinton, thus giving the optimal social response but not the true view he or she holds.
Social desirability bias is also more frequented in polls that involve personal conversations with a pollster or contain limited anonymity in the structure of the polling. In these cases, because an individual must take ownership of his or her response, he or she may give an answer they believe the pollster wishes due to feared social judgment and perceived ignorance. Social desirability bias can further impact a poll if the interviewer’s demographic is being questioned within the poll; for example, if a woman asks an individual how they feel about the idea of a woman acting as President of the United States, the respondent may answer that they feel positive about this topic even though, in reality, they may oppose the idea. Therefore, the subject of the poll, the environment in which the poll is conducted, the amount of interpersonal contact involved in the polling, and certain qualities and characteristics of the interviewer can all contribute to the social desirability bias present within a poll.
Are Voters Afraid to Openly Support Trump?
Donald Trump has been doubted by many at every stage of the 2016 presidential election. Experts predicted that he would have no chance in the Republican primary, running against established politicians like Jeb Bush, Ted Cruz, and Marco Rubio. On July 1, 2015 the experts were right. Trump was polling in 7th place. However, by the end of the month Trump was in first place, and went on to secure the nomination by a sizable margin. Trump is clearly a popular candidate and has certainly posed a threat to Clinton in the general election. If he has such a large base of support, why are people continually doubting the legitimacy of his shot at winning? How many people you know openly support Donald Trump? The answer could be none, or could be a lot, either way it is likely less than the actual amount of Trump supporters that you know. The reason for the disparity between perception and election results for Trump campaign is the effect that social desirability has in a person’s response to political surveys. This is especially relevant to Trump given the polarization his platform evokes and the stigma surrounding his supporters.
According to RealClearPolitics, Trump’s favorability rating is currently 38% (RCP 2016). He has a history of making derogatory statements about women, minorities, and people with disabilities. As a result, to some people a vote for Trump has come to represent racism, sexism, and general ignorance. In more liberal communities across the country, like Bowdoin, the backlash against Trump supporters is significant. The impact of the hesitancy voters show in revealing their true intentions is concerning, specifically in the context of the accuracy of polls. Apart from the difficulty in obtaining a bias-free random sample of representative voters, if people are not honest, the polls are worthless. Trump’s true chances could be severely understated if a significant portion of his likely voters are truthfully responding to surveys due to fears of being judged.
In a NYT opinion piece, Thomas Edsall agrees that certain polls may under-represent Trump supporters because, “many voters are reluctant to admit to a live interviewer that they back a candidate who has adopted such divisive positions” (NYT 2016). There is evidence that more people agree with Trump’s views than polls or public perception suggest. Based on a American National Election Studies, favoritism of a white-centric America has support across all political parties.
This poll suggests that there is a strong basis of likely support for Trump that may not be captured through in person polling. The question is how to capture true voter sentiment in order to more accurately predict elections. Not surprisingly, Edsall notes that Trump does better in online polls, where there is little or no risk of external judgment, than in person-led interviews. Are online polls more trustworthy? While online polls can mirror the privacy of election day, they are not without their own issues of uncertainty. A strength of online polling is its convenience for responders. This can also pose an issue. The people who are going to fill out an online survey may not be the same people to show up to vote on election day. Voting takes an extra level of effort that online polls do not present. It is important to consider the effect that social desirability may be having on Trump’s polling results. Understanding the risks of all types of polling will make it easier to understand the strengths and limitations of each survey.
Is the United States Actually Ready for a Female President?
Are we currently overestimating support for Hillary Clinton or is the social desirability effect a thing of the past? The literature is conflicting. Some studies state that there was never a social desirability effect toward women running for office, and other studies claim that there was and continues to be a large effect that may impact the 2016 election. We took a look at biases against female candidates, how the office the candidate is running for can affect these biases, and finally how these biases may affect Hillary Clinton.
Many people express their support for female candidates publicly; however, this does not mean that they vote for these candidates when they enter the voting booth. A study conducted by Matthew Streb and colleagues in 2008 investigated social desirability effects and support for female presidential candidates. The study used a list method to determine subjects’ true feelings about a female becoming president. The study found that 26% of participants would be angered by a female president; this is a stark contrast to the 91% of people who claimed on that they would vote for a qualified female candidate. These results suggest a social desirability bias. People may outwardly support female candidates because they believe it is socially unacceptable to express their true opinion. This false reporting can cause the polls to overestimate the support for female candidates (Steb, 2008).
In order for a social desirability effect to exist, there must be biases present that are perceived as socially unacceptable. Do voters actually view male and female presidential candidates differently solely based on gender? A study conducted by Jessie Smith in 2007 sought to answer this question. Participants in the study were provided with one of three identical resumés, with the only difference being that one had a female name, Karen, one had a male name, Brian, and the other had a gender neutral name, Terry. Participants rated their impression of the candidate based on factors including presidential potential, skill level, and accomplishments. The resumé that was marked with the name Brian was aligned with more positive statements than the resumé that was marked with the name Karen. Participants viewed Brian as having better presidential potential and thought he did better in his career when compared to Karen (Smith, 2007). Most people would not openly say that, all being equal, a man would do a better job than a woman, but it appears that many people may actually feel this way.
Table from (Smith, 2007)
There have been many studies conducted that deny the existence of a social desirability effect for female candidates; however, these studies are referencing women running for positions that are not the President of the United States. Is it possible that voters are accepting of a female senator or governor, but not of a female president? Smith’s study found that when comparing subjects’ views on presidential candidates to views on senatorial candidates, there is a marked difference. The study showed that participants demonstrated no statistically significant biases toward female senatorial candidates, indicating that gender biases may be restricted to the presidential office (Smith, 2007).
In 2000, Clinton won the New York senatorial election against Rick Lazio by a larger margin than expected. This is the opposite of what we would expect had the social desirability effect been present. This was the case for her 2006 election as well. Although there has been no social desirability effect demonstrated in Clinton’s past election cycles, this does not mean that she will not experience a social desirability effect in the 2016 presidential election. Until now, we have never had a female candidate become the nominee for a major party in the United States, so we do not have historical data about the social desirability effect as it relates to women running for president. The data we do have suggests that voters may not have biases against female candidates in general, but that there are biases specifically against female presidential candidates. This may create a social desirability effect where one was not previously seen and would result in an overestimation of support for Clinton.
How Can We Eliminate the Social Desirability Effect?
Rather than predict the outcome of this election, we wanted to conclude with three methods that best eliminate social desirability bias from election polls and surveys. The first is “list ranking,” which was used in the Streb study, “Social Desirability Effects and Support for a Female American President.” List ranking increases the anonymity in responses by eliminating the ownership as to which listed items a respondent either agreed or disagreed with. However, by implementing both a control and an experimental group in the polling, a researcher is able to conclude that mean differences between two groups can be attributed to the additional listed item given to an experimental group. For the Streb study, this additional item was whether or not having a female president would upset the participant. Another technique to decrease social desirability bias in polling is to make the survey as impersonal as possible. This can be done through polling online. If a researcher chooses this method, he or she must be aware of the addition of response and selection biases that can arise from online polls. The last method our group wished to mention was creating polls that are as similar to the actual voting process as possible. This can be done by creating mock ballots with a personal space designated for individuals to cast votes. By mimicking the decreased human interaction of the polling and additional privacy available to the respondent, a pollster may best be able to capture the voting that will occur on election day.
Smith, Jessie. “No Place for a Woman: Evidence for Gender Bias in Evaluations of Presidential Candidates.”BASIC AND APPLIED SOCIAL PSYCHOLOGY, 29(3), 225–233 Copyright # 2007, Lawrence Erlbaum Associates, Inc.
Streb, Matthew. Burrell, Barbara. Frederick, Brian. Genovese, Michael. “Social Desirability Effects and Support for a Female American President.” Public Opinion Quarterly, Vol. 72, No. 1 2008, pp. 76–89.
Authors, By LSE. “Polls Apart: Why We Need to Treat All EU Referendum Polling with Caution.” LSE BREXIT. February 22, 2016. Accessed October 04, 2016. http://blogs.lse.ac.uk/brexit/2016/02/22/polls-apart-why-we-need-to-treat-all-eu-referendum-polling-with-caution/.
Edsall, Thomas. “How Many People Support Trump but Don’t Want to Admit It?”. New York Times. May 11, 2016. http://www.nytimes.com/2016/05/11/opinion/campaign-stops/how-many-people-support-trump-but-dont-want-to-admit-it.html?_r=1
RealClearPolitics, Trump Election Poll http://www.realclearpolitics.com/epolls/other/trump_favorableunfavorable-5493.html