Table 1.
Linear OLS regressions testing asymmetries in truth discernment and bias.
| Belief |
||||||
|---|---|---|---|---|---|---|
| RQ1 | RQ2 | RQ3 | ||||
| Ideology (-3 to 3) | 0.04 | (SE = 0.03, p = .228) | 0.04 | (SE = 0.04, p = .269) | 0.08 | (SE = 0.05, p = .144) |
| News item truth (ref: 0) | 0.86 | (SE = 0.10, p = .000) | 0.87 | (SE = 0.12, p = .000) | ||
| Ideology * truth | −0.09 | (SE = 0.10, p = .001) | −0.11 | (SE = 0.06, p = .053) | ||
| Congruence (ref: 0) | 0.14 | (SE = 0.06, p = .027) | 0.23 | (SE = 0.12, p = .052) | ||
| Ideology * congruence | −0.14 | (SE = 0.06, p = .019) | −0.13 | (SE = 0.09, p = .151) | ||
| Congruence * truth | −0.16 | (SE = 0.13, p = .226) | ||||
| Ideology * congruence * truth | 0.10 | (SE = 0.11, p = .361) | ||||
| Constant | 2.59 | (SE = 0.09, p = .000) | 3.16 | (SE = 0.08, p = .000) | 2.61 | (SE = 0.11, p = .000) |
| Observations | 10,431 | 7,234 | 7,234 | |||
| R2 | 0.05 | 0.01 | 0.06 | |||
Note: Dependent variable across models is the belief score of a subject-news item encounter. All models use clustered standard errors at both the subject and the item level and apply weights (age, gender, education). P-values refer to two-tailed tests. In model RQ1, the negative interaction of truth and subject ideology suggests that truth matters less for more conservative participants. In model RQ2, the negative interaction of congruence and subject ideology suggests that congruence matters more for liberals. Model RQ3 does not find any support for a three-way interaction.