Why Social Media Data Don’t Tell Us Very Much About Public Opinion
When social scientists want to learn something about public opinion, they traditionally conduct surveys, using representative samples. Representativity means that even with relatively small numbers of participants, if selected at random, the opinions expressed by the sample can be generalized to the population as a whole, with a margin of error that can be estimated. The last couple of years, social media have become a popular forum for discussing political opinions, for example on Twitter and on the Facebook pages of news media, politicians and political parties. Since some of these expressions of citizens’ political opinions are relatively easily accessible, using those to examine public opinion, in addition to or as an alternative to surveys, seems like an excellent idea. But is it, though? I argue that political opinions we find on social media will not be representative of the society as a whole, and here are some of the reasons why:
Patterns of social media presence and use
First, one precondition for expressing political opinions on social media is that a person needs to be online – but in many countries, this does by far not apply to the whole population. The proportion of onliners in Norway (98% in 2017) is exceptionally high, but it is much lower in most other countries, also in Europe. And even in countries with high Internet penetration, the variation of time of Internet use leads to overrepresentation, usually towards the younger, more educated, and more politically interested citizens.
Second, far from all onliners use social media, and the popularity of various social media platforms strongly differs. In 2018, for example, 83% of all Norwegians had a profile on Facebook, but only a minority of 29% had a profile on Twitter according to Ipsos. Again, these proportions are much lower in other countries, particularly for Twitter.
Third, not all social media users express their political opinions on social media. Many users do not even come into contact with political issues on social media at all. As with the Internet in general, people with extreme (political) opinions are more inclined to express these on social media, while many (political) moderates refrain from doing that. As a result, the minority of politically extreme users will be overrepresented in political discussions on social media, while the majority is moderate, but will be underrepresented.
Fourth, we cannot know for sure where the authors of different comments come from – and if they are real humans or perhaps social bots. If we, for example, investigate Norwegian Facebook pages, probably not all comments we find there were written in Norway or by Norwegian citizens, which is why they do not reflect public opinion in Norway.
Access to data and consequences for representativeness
Fifth, as researchers, we can only access comments on publicly available social media profiles – and even then, collecting these data can be a violation of terms of use of different social media platforms. Consequently, many researchers collect data from Twitter since the tweets are publicly available – but most likely, Twitter users are even less representative of the general population than Facebook users: Those active on Twitter are typically younger, higher educated, and men outnumber women strongly. Moreover, politicians and journalists are highly overrepresented among Twitter users, making Twitter kind of an elite medium.
Last but not least, some strategic actors (e.g. political parties) delete comments that are not in line with their political positions on their social media pages, and we as researchers cannot know whether any, and which comments were deleted.
Political opinions expressed on social media anything but representative
Due to these reasons, political opinions expressed on social media are anything but representative for the opinion distribution among a country’s population. Thus, social media analysis cannot be a substitute for survey-based measurement of public opinion. Nonetheless, analyzing social media data can provide interesting information regarding processes of public opinion formation and opinion expression. The specific – and often rather extreme – political opinions of those onliners (that might even be social bots) that have a social media account and express their political opinions there are certainly atypical. Comments are further biased because they must pass the filter of strategic actors and be accessible for the researchers. Still, these atypical users may influence public opinion to some extent. Taken together, social media provide researchers with a distorted perception of the opinion climate in the population rather than with representative information about the public opinion. When investigating public opinion, representative surveys should still be the main go-to method for researchers, even in the age of social media. What social media can certainly offer, however, is a platform that can give us unique insights into the dynamics of online debates among citizens. Therefore, using them to study the dynamics of online debates is interesting, important, and a fruitful avenue for further research.
Melanie Magin
Melanie Magin (M.A., Dr. phil.) is an Associate Professor in Media Sociology at the Department of Sociology and Political Science at NTNU in Trondheim. She is a member of various research groups, such as Elections, Values and Political Communication (EVPOC) and Digitalization and Social Life.
