Extremely Online? Ideological and affective polarisation on British political Twitter

In my last post, I talked about how politicians have gradually became more emotive in their Twitter rhetoric over time. Are these politicians preaching to the converted? Were Donald Trump's tweets encouraging the January 6th Capitol riots speaking to an already politically-extreme audience? After all, members of ultra-nationalist groups like the Proud Boys and Oath Keepers were found to have played a key role in the Capitol riots. From Trump's point of view, of course, Twitter’s extremism stems entirely from the 'other' side:

While it might be a stretch to claim that Silicon Valley’s Twitter has ever been a hub of revolutionary Marxism, it’s worth considering whether the dominant voices on social media tend to come from the fringes of the political spectrum. Political polarisation, in both its ideological and affective forms, is a problem. When these divides deepen, policy gridlock often follows, making it harder for legislators to compromise and get things done. Polarisation can reinforce prejudice, erode trust in democratic institutions, and even fuel political violence. The graph below traces its development in the United States, but it's also a growing phenomenon in the United Kingdom.

Some argue the seeds of polarisation were sown in a pre- social media era, when the emergence of hyper-partisan broadcast media like Fox News helped reinforce biases and negative perceptions of the opposing party. These outlets provide content which aligns with the ideological preferences of their audience. This, in turn, leads to 'selective exposure' where individuals consume media which simply reinforces their pre-existing beliefs.

Have social media amplified political divisions or helped bridge them? Logging in to Twitter, even in the pre-Elon Musk era, it often felt like the former. The most strident and controversial viewpoints, generating the greatest amount of agreement or opprobrium, would generally appear at the top of one's feed. This is no accident of design; most platforms are refined to maximise user engagement, pushing our latent psychological buttons by promoting sensational and divisive content. This tends to generate more clicks, shares, and comments. Compared with 'legacy' partisan media, most social media platforms allow an even greater degree of self-selection - and the ability to curate a personal network of like-minded individuals.?These echo chambers make it even easier to express a political identity and reinforce group norms. Social media platforms have become increasingly fertile ground for the spread of misinformation and fake news, which distort perceptions of political issues and individuals. Fake news tends to spread faster and reach more people than factual information, leading to more frequent and intense disputes over a 'ground truth'.

It's tricky to unpick the causal direction of this relationship, however. Do social media simply reflect the polarisation already present in society? After all, highly-polarised, highly-motivated (and often narcissistic) individuals are drawn to social media to express and validate their views. The negatively-charged, and often false, content produced by these individuals spreads further and faster, fostering hostility towards opposing groups and deepening socio-cultural divisions. Whether social media is driving political extremism, or vice-versa, we still know relatively little about who uses social media to talk about politics and the nature of the content they share. Are they more ideologically extreme, and more hostile to people on the other side? Anecdotally, at least, we might expect this to be the case. We already know that social media users are not representative of wider populations. For example, in the UK, they overall tend to be more left-wing and more politically-engaged. On Twitter, users in the United States and Spain are majority male and disproportionately live in urban areas. Surprisingly, nobody has conclusively identified whether those people who discuss politics on social media are more ideologically and affectively extreme, and whether there are differences in this extremity between partisan groups.

How did we approach this question?

We took a nationally-representative snapshot of the British population over two six-month periods (September 2019 - February 2020 and May 2021 - October 2021),?using data from an original survey fielded in July 2021 by YouGov, and linked this survey data to respondents' actual online activity on Twitter. We pulled a random sample of maximum 25 tweets from each active user in each study period, giving us a total of 10,910 tweets. These tweets were hand-coded by three research assistants using 19 standardised coding categories, including evaluations of whether a tweet or retweet was political, whether it represented support for (or opposition to) each of the main UK parties, and support of (or opposition to) the UK's decision to leave the European Union. We categorised 'political Twitter users' as those users who tweeted about politics at least once during both periods.

To measure ideological attitudes, we averaged agreement with a series of statements along three dimensions: 'left-right' economic attitudes, liberal-conservative 'social' attitudes, and attitudes towards the European Union. Responses ran from 'strongly disagree' to 'strongly agree', recorded 1-5 on a Likert scale. These were averaged and ordered so that all high scores indicate more left-wing, more socially-liberal, and more pro-EU. A measure of extremity was then calculated by taking ideological averages on each dimension, for each respondent, and subtracting the average of all responses in our sample. To measure affective attitudes, we took a standard 'thermometer' score; the difference between two 0-100 ratings, indicating feelings of favourability/unfavourability towards voters on either side of either the Conservative/Labour or Leave/Remain divide. Higher scores equal greater warmth' or favourability, so a greater difference between the scores given to in-group and out-group equals greater affective polarisation. We then ran a series of OLS regressions modelling political Twitter users against non-political Twitter users from their partisan group.

What did we find?

Put simply, we found that political Twitter users are more ideologically extreme than those who don't use it to talk about politics - but only those who identify as Labour partisans and those on the 'Remain' side of the Brexit debate. Similarly, when we looked at actual Twitter activity, more ideologically extreme Remain supporters were more likely to tweet politically, while affective extremity is linked to a greater number of political tweets by Leavers and Remainers.

Why did we see these patterns?

Why were more extreme partisans from these groups disproportionately represented? It could be partially attributed to self-selection effects; perceptions of Twitter being more left-leaning or more pro-EU may have made it a more appealing space for Labour and Remain partisans to discuss politics. If echo chambers do exist, we might expect attitudes to become reinforced over time, as well as attracting new partisans to become politically active on the platform. Obviously, this perception has shifted beyond all recognition since Elon Musk's takeover of the platform, and it is now considered to heavily favour those towards the 'right' of the ideological spectrum.

Given the timing of the study, our findings also raise questions about some of the underlying political dynamics at play. On most of our measures, Brexit appeared to be a more salient and divisive fault-line than traditional party politics. There certainly seems to be a clear divide between political winners' and 'losers', particularly in relation to the EU referendum, where many Remainers felt a degree of anger, not only about the result, but about the process itself. High-arousal emotions such as anger can motivate people to participate in political activity; social media represent a low-cost method of political participation, and a convenient outlet for these emotions. This chimes with our finding that greater affective extremity correlates with a greater number of political, and negatively-partisan tweets.

The asymmetry we found on Twitter could also be a function of a government-opposition dynamic. For the vast majority of Twitter's existence, the Conservative Party was in government. Knowing that opposition parties are more likely to use negative rhetoric than incumbent parties, it's perhaps unsurprising that Labour supporters were likely to be more motivated to take to Twitter to attack the government. Again, this appears to be supported by our finding that more ideologically-extreme Remain and Labour Twitter users are more likely to share political, and negatively-partisan, tweets. Being dominated by individuals who are more ideologically and affectively polarised - even in the relatively benign pre-Elon Musk era - this highlights the worrying potential for social media platforms to exacerbate polarisation.