Table 3. Effect of messages received in previous round on prevention in current round.

	Outcome: Probability of Prevention in Current Round
	(1)	(2)	(3)	(4)
A. Number of Messages Received in Previous Round	1.09
	[0.98,1.21]
B. Number of Messages in Previous Round Reporting Prevention Decision and Outcome		1.14*
		[1.01,1.29]
C.Number of Successful Prevention Messages Received in Previous Round			1.53**
			[1.25,1.86]
D. Number of Failed Prevention Messages Received in Previous Round			1.1
			[0.91,1.33]
E. Number of Successful Risk Messages Received in Previous Round			0.57**
			[0.42,0.79]
F. Number of Failed Risk Messages Received In Previous Round			0.94
			[0.69,1.27]
G. Number of Messages Received in Previous Round Reporting Prevention				1.32**
				[1.15,1.52]
H. Number of Messages Received in Previous Round Reporting Risk-Taking				0.76*
				[0.60,0.96]
I. Number of All Other Messages Received in Previous Round		1	1.01	1
		[0.86,1.16]	[0.89,1.16]	[0.88,1.15]
P value: C = D			0.01
P value: E = F			0.01
P value: G = H				0.00
Mean of Dependent Variable	0.71	0.71	0.71	0.71
Number of Observations	30555	30555	30555	30555

Notes: Results are from logit regressions estimating the association between messages and information received in the previous round and the odds of prevention in the current round, with a fixed effect for each individual (each observation consists of an individual-round). All regressions control for the effectiveness of the randomly assigned preventive technology. The sample includes observations from all sessions although individuals only had the opportunity to receive messages in the third session. The “All other Messages Received” category varies with each regression. Coefficients are in terms of odds ratios and 95% confidence intervals, based on standard errors clustered at the individual level, are in brackets.

*p<0.05

**p<0.01