Table 3.

Does receipt of concordant messages increase willingness to vaccinate?

Panel A. Effect of concordant score on intent to vaccinate
	Ordered logit
	Intent to vaccinate self	Intent to vaccinate child
Concordant score	0.018	0.032
	$[-0.041,-0.076]$	$[-0.055,-0.120]$
Cut 1	−2.399***	−1.710***
	$[-2.694,-2.103]$	$[-2.018,-1.402]$
Cut 2	−1.959***	−1.389***
	$[-2.248,-1.671]$	$[-1.691,-1.086]$
Cut 3	−1.782***	−1.219***
	$[-2.068,-1.496]$	$[-1.517,-0.920]$
Cut 4	−1.336***	−0.688***
	$[-1.618,-1.055]$	$[-0.980,-0.397]$
Cut 5	−0.880***	−0.289*
	$[-1.158,-0.601]$	$[-0.577,-0.002]$
Cut 6	−0.103	0.384**
	$[-0.377,-0.172]$	$[-0.097,-0.672]$
Subsegments	24	11
Observations	2621	1032
Notes: 95% confidence intervals in brackets using heteroskedasticity-robust standard errors. Each regression includes subsegment fixed effects. Outcome ranges from 1 (highly unlikely) to 7 (highly likely). Concordant score is the number of message attributes customized for that respondent’s segment memberships, plus an additional unit if treated with Spanish parallel text if Latinx. +$p<0.10$, * $p<0.05$, ** $p<0.01$, *** $p<0.001$.
Panel B. Margins of coefficient on concordant score
		$\mathrm{\Delta }$ Prob. of each reply: 1 is “highly unlikely to vaccinate,” 7 is “highly likely”
	$N$	1	2	3	4	5	6	7
Self	2621	−0.0026	−0.0005	−0.0002	−0.0003	−0.0002	0.0002	0.0035
Child	1032	−0.0058	−0.0005	−0.0002	−0.0003	0.0001	0.0008	0.0057

Notes: The marginal change in the likeliness of reporting the given category of vaccination intent due to an increase of one concordant element, based on the ordinal logit estimates in Panel A.