When politicians spread disinformation on social media, do citizens believe it? Evidence from two experiments

What happens when politicians spread disinformation online? We investigate whether citizens, when exposed to a politician’s tweet containing disinformation, update their belief in the false claim. In line with prior work, we hypothesize that exposure increases citizens’ belief, but that the effect is contingent on (1) citizens’ prior exposure to the false claim: when citizens repeatedly encounter the same false claim, this enhances their belief in the false claim (2) citizens’ preceding evaluation of the false claim: if citizens previously categorized the false claim as unbelievable, renewed exposure to the false claim has no, or a negative effect, on subsequent belief (3) citizens’ evaluation of the politician: citizens are more likely to accept disinformation if they evaluate the politician more positively (4) the presence or absence of a warning message. Given that social media platforms such as Twitter have begun implementing warnings on posts with contested or false claims, we expect that such warnings decrease the impact of exposure on disinformation belief. To evaluate our propositions, we present novel data stemming from two online, survey-embedded experiments in Flanders (Belgium). Experiment 1 used a large non-representative sample (N=1,460) whereas Experiment 2 replicated the first experiment on a smaller, representative sample (N=805). Both experiments exposed participants to a tweet containing disinformation on COVID-19, systematically varying the politician (a left-leaning politician, a centre politician, or a right-leaning politician) and the presence or absence of a warning message. Using a panel design in both experiments, we examine the change in participants’ belief in disinformation pre and post exposure. Results indicate that citizens’ belief in the disinformation increased after exposure. Contrary to our expectation, prior exposure decreases post-exposure belief, suggesting a reversal of the truth effect. However, prior belief—or lack thereof—largely explains this counterintuitive finding: it is amongst highly sceptical citizens that repeated exposure decreases belief. In line with expectations, we find that the evaluation of the politician moderates the effect of exposure. Finally, we find that warnings are somewhat effective in reducing the effect of exposure, but that they are nowhere near able to offset the increase in disinformation belief caused by exposure.