Table 3 Coefficients, z values, and two-tailed p values for the individual-level difference moderation analyses, derived from the linear regression models described in the text.

From: Accuracy prompts are a replicable and generalizable approach for reducing the spread of misinformation

 

Moderation of accuracy prompt effect on discernment

Relationship with baseline discernment in control

Lucid/YouGov

MTurk

Lucid/YouGov

MTurk

b

z

p

b

z

p

b

z

p

b

z

p

Female

−0.006

−0.814

0.416

−0.005

−0.380

0.704

−0.001

−0.160

0.873

−0.019

−2.175

0.030

White

0.014

1.746

0.081

−0.011

−0.773

0.440

0.006

0.966

0.334

0.042

1.899

0.058

Age

0.001

3.182

0.001

0.000

0.840

0.401

0.001

5.661

<0.001

0.001

2.051

0.043

College degree

0.001

0.139

0.890

−0.011

−0.919

0.358

0.037

5.116

<0.001

0.030

2.371

0.018

Conservatism

0.009

0.623

0.533

−0.031

−1.693

0.090

−0.120

−4.531

<0.001

−0.069

−2.220

0.026

Republican

−0.007

−1.410

0.158

−0.022

−3.057

0.002

−0.070

−4.209

<0.001

−0.022

−1.270

0.204

Republican (no independents)

−0.005

−0.867

0.386

−0.022

−2.526

0.012

−0.086

−4.018

<0.001

−0.036

−1.781

0.075

Trump 2016 voter

−0.001

−0.087

0.931

−0.027

−2.744

0.006

−0.053

−3.031

0.002

−0.014

−0.803

0.422

Value accuracy

0.004

0.315

0.753

0.028

2.755

0.006

0.097

6.404

<0.001

0.087

7.449

<0.001

Cognitive reflection (CRT)

0.038

1.986

0.047

0.030

2.371

0.018

0.062

5.351

<0.001

0.061

4.117

<0.001

Attentiveness

0.068

4.090

<0.001

   

0.067

7.628

<0.001

   
  1. The left half of the table shows the coefficients for the 3-way interaction between headline veracity, condition, and the individual difference - which captures the extent to which the individual difference moderates the treatment effect on sharing discernment. The right half of the table shows the coefficient on the 2-way interaction between headline veracity and the individual difference - which captures how the individual difference relates to baseline sharing discernment in the control condition. Results are shown separately for the more representative samples from Lucid or YouGov, versus the convenience samples from MTurk. Coefficients with p < 0.05 are bolded. No adjustments were made for multiple comparisons.