Table 2.

Estimated treatment effect of framing on click-through behavior

	Treatment Effects				Moderated Treatment Effects
	No Interaction No cov. Adjustment	Interaction No cov. Adjustment	Interaction cov. Adjustment	Interaction cov. Adjustment	Conditional on Gender	Conditional on Regional Party Affiliation	Conditional on Regional Smoky Days
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Narrative	0.225***	0.206***	0.215***	0.214***	0.249***	0.198***	0.329***
	[0.2, 0.25]	[0.171, 0.241]	[0.18, 0.251]	[0.178, 0.25]	[0.205, 0.292]	[0.159, 0.237]	[0.258, 0.4]
EDU Messenger	0.022	0.001	0.003	0.006	0.003	− 0.001	0.006
	[−0.003, 0.046]	[− 0.036, 0.038]	[− 0.034, 0.04]	[− 0.031, 0.043]	[− 0.035, 0.039]	[− 0.041, 0.04]	[− 0.031, 0.044]
Narrative*EDU	–	0.038	0.032	0.022	0.032	0.022	0.021
		[−0.011, 0.087]	[−0.018, 0.082]	[− 0.028, 0.073]	[− 0.018, 0.082]	[−0.028, 0.072]	[− 0.029, 0.072]
Narrative*Female	–	–	–	–	−0.067**	–	–
					[−0.116, − 0.017]
Narrative*Republican	–	–	–	–	–	0.053	–
						[−0.001, 0.108]
EDU*Republican	–	–	–	–	–	0.022	–
						[−0.032, 0.077]
Narrative*Smoky Days	–	–	–	–	–	–	−0.003***
							[−0.004, − 0.001]
Intercept	−4.374	−4.363	−4.495	−5.113	−4.514	−5.1	−5.176
Covariate: Age	No	No	Yes	No	Yes	No	No
Covariate: Gender	No	No	Yes	No	Yes	No	No
Covariate: Party Affiliation	No	No	No	Yes	No	Yes	Yes
Covariate: Reginal Smoky Days	No	No	No	Yes	No	Yes	Yes
Observations	1,838,100	1,838,100	1,801,959	1,789,380	1,801,948	1,789,380	1,789,380

Estimated treatment effect of framing on click-through behavior, and estimated treatment effects conditional on gender, regional party affiliation, and regional smoky days. The estimated effect is the change in the log odds of click-through behavior (CTB). Logistic regression estimation. Dependent variable is the CTB. The odds of CTB can be measured as CTR/(1-CTR), where CTR is the click-through rate. 95% confidence intervals in brackets. * p < 0.05, ** p < 0.01, *** p < 0.001