Abstract
The headache any researcher faces while using Twitter data for social scientific analysis is that we do not know who tweets. In this article, we report on results from the British Social Attitudes Survey (BSA) 2015 on Twitter use. We focus on associations between using Twitter and three demographic characteristics—age, sex, and class (defined here as National Statistics Socio-Economic Classification [NS-SEC]). In addition to this, we compare findings from BSA 2015, treated as ground truth (known characteristics), with previous attempts to map the demographic nature of UK Twitter users using computational methods resulting in demographic proxies. Where appropriate, the datasets are compared with UK Census 2011 data to illustrate that Twitter users are not representative of the wider population. We find that there are a disproportionate number of male Twitter users, in relation to both the Census 2011 and previous proxy estimates; that Twitter users are predominantly young, but there are more older users than previously estimated; and that there are strong class effects associated with Twitter use.
Highlights
In the past decade, the social sciences have undergone a revolution in response to the challenges of utilizing “big data” for social scientific analysis
With reference to RQ1, the analysis of British Social Attitudes Survey (BSA) 2015 in this article has demonstrated associations between Twitter use and sex, age, and NS-SEC for British users: Men are proportionally more likely to use Twitter than women relative to the male/female split of the UK population; the age distribution of Twitter users is younger than the age distribution of the UK population; and certain occupational groups are more likely to use Twitter than others—notably NS-SEC 1 and 2, characterized by managerial, administrative, and professional occupations
What this article is unable to answer is why differences in Twitter use are associated with these demographic characteristics
Summary
The social sciences have undergone a revolution in response to the challenges of utilizing “big data” for social scientific analysis. The data generated by social media platforms such as Twitter are “big” as defined by Kitchin and McArdle (2016) and present particular problems for researchers through their size, the speed at which data are generated, their variety (text, images, audio, videos, hyperlinks), their exhaustivity (populations rather than samples), tight and strong resolution and indexicality, strong relationality built on networks, and high extensionality and scalability (Kitchin & McArdle, 2016) Underlying all of this is the question of veracity in regard to the authenticity of both the message being conveyed (Williams, Burnap, & Sloan, 2016) and, of principle importance for this article, who is producing the content. We provide UK Census 2011 data as a baseline from which to judge to what extent Twitter users are representative of the population of interest
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
