Downloading polarisation: the effect of internet consumption on affective polarisation in the United Kingdom

Affective polarisation captures partisan animus – the emotional rift between political opponents rooted in people’s tendency to have a more positive affect towards co-partisans than political opponents. Affective polarisation has increased in the US and in the UK, but there is still much debate on the causes of this phenomenon. Recent studies suggest that internet usage may increase affective polarisation through exposure to polarising content, but the theory is ambiguous. Moreover, extant research focuses overwhelmingly on the emblematic US case, and it often falls short of establishing a causal relationship. Finally, it is unclear whether affective polarisation is growing because of stronger positive sentiment towards the co-partisans, greater animosity between political opponents, or both simultaneously. Against this background, I propose to identify the causal effect of internet usage on affective polarisation in the United Kingdom. I assemble a dataset merging newly released large-N geocoded panel data on individual behaviour with postcode-level data on internet performance. This enables me to leverage exogenous variations in internet usage at the individual level, where internet usage in each wave is instrumented by time-varying local-level measures of internet performance – which depend on local infrastructure developments and are thus independent of individual characteristics. I find that internet usage does increase both conventional affective polarisation and cross-cutting polarisation around Brexit identities. However, internet performance does not appear to have an effect on individual internet usage in the studied time period. Further investigation reveals that performance has a negligible effect on aggregate internet usage, using registry data at postcode level, the lowest level of aggregation. Finally, I show that even relative micro-level aggregate measures of internet performance hide substantial underlying variation.