Challenges in accurately measuring public opinion

Abstract

Welcome to our Special Issue focusing on "The challenges of accurately measuring public opinion." This is a very important topic, especially, in the wake of the British Polling Council/Market Research Society (BPC/MRS) inquiry into the performance of the UK polls in predicting a win in 2015 for the Conservative Party and other occasions where pollsters have been accused of misreporting the public's mood, as in the earlier Scottish referendum on independence and the 2016 Presidential Election in the United States. Such accusations also raise the question of pollsters' responsibility to society if they are misinforming stakeholders in the election process. The defense that they are recording views at a certain point in time, predicting turnout is increasingly difficult, and these findings are subject to some level of error (however, this might be measured!) rather than providing an accurate prediction of future behavior becomes a weak argument, where pollsters methodologies are found to be flawed as in 2015. However, all this has been covered in depth in my previous Editorials. Opinion polling has always fascinated me, especially, as it is a rare example where commercially commissioned research projects come under the spotlight and transparency about methodology is a prerequisite. It is also important because the results of the polls can be openly compared with the actual result of the outcome it is supposedly attempting to predict. By virtue of this publicity, research companies that undertake polling are also the "brands" in the research sector the public are most aware of. Overall, this all contributes to the high level of social responsibility that falls on the shoulders of pollsters. As I write this Editorial, Kantar has just published an international review of polls conducted in 2017, showing that pollsters enjoyed a much more successful year in terms of accurate predictions (https://www.research-live.com/article/news/pollsters-called-correct-results-in-2017/id/5032402). This is obviously good news after all the bad press, but there is no doubt that the changing political landscape in many countries in recent years makes polling increasingly difficult. However, as the review underlines, all polls are subject to some margin of error, in whatever way "error" might be calculated. But, as Peterson argues in his paper in this issue, "error" is a multidimensional construct in survey design. We simply cannot ever deliver the absolute degree of precision in forecasting voting intentions that pundits would ideally like, so there may well be disagreement in some quarters

Sacred Excess: Organizational Ignorance in an Age of Toxic Data

“…The appropriate response to this paradox is then to spend more money on research. and polling industry organizations address this problem by calling for the creation of more data about a problem that was caused by data creation (Moncey, 2016;Skibba, 2016).…”

Perceptions of the rights and capabilities of people with intellectual disability in the United States

“…Also awareness of, and engagement with Special Olympics was added as a further predictor as this was of particular interest to the commissioners of the study. To examine consistency in responses across the United States for data collected over two waves of the study. Most opinion polls have a recognized margin of error of around ± 3% although this is a somewhat simplistic approach and is hard to defend (Mouncey, 2018). However, a more thorough test would be to examine the findings with a second comparable sample using the same methods and also to compare the findings from across selected samples within a country as well as nationally.…”