Who Won the Facebook Referendum? Engagement, Tone and Nationalism on UK Party and Campaign Pages, February to June 2016

Between the announcement of the EU referendum on 20 February 2016 and polling day on 23 June 2016, the Leave and Remain sides fought a large part of their campaign on Facebook. I analyse 17,657 posts from 92 UK pages, covering the official and unofficial campaigns and the main party, leader and MP pages, which together drew 33.5 million interactions. Pages aligned with Leave produced 65% of the engagement from 23% of the posts. Most of that lead comes from four large pages (Leave.EU, Vote Leave, Nigel Farage and UKIP), and it shrinks but stays positive, with wide uncertainty, when audience size is held constant. I check the text coding against a sample of 360 posts coded by an AI model (Claude) from a written codebook. Against that reference coding, the dictionary measure of negativity is precise but misses about 42% of negative posts, so it understates how negative both sides were. Corrected for this, 37% of Leave posts and 42% of Remain posts were negative among campaigns, leaders and parties (difference -5 points, 95% CI -19 to +8). Across all pages Leave looks more negative, but that gap comes from one individual MP page, Douglas Carswell's. Leave pages were more focused on immigration (17% against 5%) and British national identity, and less focused on cooperation. Negative posts drew more engagement on both sides, by 15% on Leave pages and 41% on Remain pages when each side is modelled separately. Whether the premium differed between the sides depends on how the model is specified, so I do not claim a difference. The evidence does not support the claim that Leave won engagement by being more negative. Preprint, not peer reviewed. This version corrects the description of the validation sample (coded by an AI model, not by hand), the claim about the size of the negativity premium on each side, and several descriptive statements; no estimate changed.