The Campaign Disinformation Divide: Believing and Sharing News in the 2019 UK General Election

Figures & data

Table 1. Mixed linear regression model explaining perceived headline accuracy using the PCA factors for media use (surveys 1 and 2 combined).

Predictor	b	SE	p
Survey (1 = survey 2)	−.215***	.02	< .001
Participation in both surveys (1 = yes)	−.250***	.06	< .001
Gender (1 = men)	−.071***	.02	< .001
Age	.001*	.00	.028
Number of cars (proxy for income)	.012	.01	.213
Low education vs. medium education	−.051*	.02	.020
Low education vs. high education	−.121***	.02	< .001
Likelihood of voting in 2019 general election	−.001	.00	.954
Use of professional media for election news (PCA)	−.063***	.01	< .001
Use of social media for election news (PCA)	.124***	.01	< .001
Recognizing the headline (1 = yes)	.739***	.01	< .001
Incorrectly recognizing a placebo headline (1 = yes)	.071***	.02	< .001
Headline covers policy issue important to respondent (1 = yes)	.054***	.01	< .001
Headline attacks preferred party/position (1 = yes)	−.305***	.01	< .001
Headline false vs. true (1 = true)	.358***	.01	< .001
Headline false vs. true (1 = true) X Use of professional media for election news (PCA)	.164***	.01	< .001
Headline false vs. true (1 = true) X Use of social media for election news (PCA)	−.188***	.01	< .001

Note. Nakagawa’s marginal R2 = .226. Displayed are only fixed effects but model includes random intercepts for each participant, n_observations = 41,415, n_respondents = 3,730, lower n due to missing values on control variables. See SM5-10 for robustness checks. * p < .05, *** p < .001.

Figure 1. Perceived accuracy of true vs. false headlines by use of professional and social media for getting news about the election (surveys 1 and 2 combined).

Note. Uses of media are PCA factors. Perceived headline accuracy adjusted for the influence of all other control variables, n_observations = 41,415, n_respondents = 3,730, lower n due to missing values on control variables, colored areas around bold lines indicate 95% confidence intervals.

Table 2. Mixed linear regression models explaining intention to share headlines using the PCA factors for media use (survey 2 only).

	Model A			Model B			Model C
Predictor	b	SE	p	b	SE	p	b	SE	p
Participation in both surveys (1 = yes)	−.190	.14	.180	−.141	.14	.304	−.139	.14	.309
Gender (1 = men)	.076	.04	.060	.096*	.04	.014	.097*	.04	.013
Age	−.001	.00	.367	−.002	.00	.229	−.002	.00	.231
Number of cars (proxy for income)	−.040	.02	.094	−.043	.02	.062	−.043	.02	.061
Low education vs. medium education	−.159**	.05	.003	−.145**	.05	.005	−.144**	.05	.006
Low education vs. high education	−.235***	.05	< .001	−.217***	.05	< .001	−.215***	.05	< .001
Likelihood of voting in 2019 general election	−.021*	.01	.023	−.019*	.01	.029	−.019*	.01	.033
Use of professional media for election news (PCA)	−.019	.02	.423	−.003	.02	.889	−.003	.02	.888
Use of social media for election news (PCA)	.474***	.03	< .001	.446***	.03	< .001	.443***	.03	< .001
Recognizing the headline (1 = yes)	.231***	.01	< .001	.077***	.01	< .001	.090***	.01	< .001
Incorrectly recognizing a placebo headline (1 = yes)	.428***	.05	< .001	.414***	.05	< .001	.406***	.05	< .001
Headline covers policy issue important to respondent (1 = yes)	.053***	.01	< .001	.049***	.01	< .001	.046***	.01	< .001
Headline attacks preferred party/position (1 = yes)	−.170***	.01	< .001	−.114***	.01	< .001	−.113***	.01	< .001
Headline false vs. true (1 = true)	−.058***	.01	< .001	−.122***	.01	< .001	.109**	.04	.009
Perceived headline accuracy	/	/	/	.195***	.01	< .001	.209***	.01	< .001
Headline false vs. true (1 = true) X Use of professional media for election news (PCA)	.036*	.01	.011	.008	.01	.554	.014	.01	.322
Headline false vs. true (1 = true) X Use of social media for election news (PCA)	−.050***	.01	< .001	−.015	.01	.275	−.013	.01	.354
Headline false vs. true (1 = true) X Perceived headline accuracy	/	/	/	/	/	/	−.070***	.01	< .001

Note. Nakagawa’s marginal R2 _Model-A = .242, Nakagawa’s marginal R2 _Model-B = .270, Nakagawa’s marginal R2 _Model-C = .271. Displayed are only fixed effects but model includes random intercepts for each participant, n_observations = 20,658, n_respondents = 1,878, lower n due to missing values on control variables. See SM5-10 for robustness checks. * p < .05, ** p < .01, *** p < .001.

Figure 2. Intention to share true vs. false headlines by use of professional and social media for getting news about the election (only survey 2).

Note. Uses of media are PCA factors. Intention to share headlines adjusted for the influence of all other control variables, n_observations = 20,658, n_respondents = 1,878, lower n due to missing values on control variables, colored areas around bold lines indicate 95% confidence intervals.

Figure 3. Intention to share true vs. false headlines by perceived headline accuracy (only survey 2).

Note. Intention to share headlines adjusted for the influence of all other control variables, n_observations = 20,658, n_respondents = 1,878, lower n due to missing values on control variables, colored areas around bold lines indicate 95% confidence intervals.

Data Availability Statement

The data described in this article are openly available in the Open Science Framework at https://osf.io/xfked/?view_only=f3df4b38d106460e84eaf86352b1a3e3.