Table 4.

Impact on contraception adoption (N=122).

Model detailsa	Tried intervention	Adopted contraception
	First stage regression estimateb (95% CI)	Intent-to-treat estimatec (95% CI)	Instrumental variables estimated (95% CI)
Assigned to treatment	.31 (0.19 to 0.44)	.13 (−0.01 to 0.26)	N/Ae
Tried intervention	N/A	N/A	.41 (−0.03 to 0.85)
Mean in control group	.02	.16	N/A

^aModels include the following controls: an indicator for mode of follow-up survey administration and several baseline characteristics, including age, number of children born, and indicators for having attended postsecondary schooling, past use of family planning, being married or in a union, and nulligravida.

^bThe first stage regression estimate is the coefficient on assignment to treatment from an ordinary least squares regression of intervention use on assignment.

^cThe intent-to-treat estimate is the coefficient on assignment to treatment from an ordinary least squares regression of contraception adoption on assignment.

^dThe instrumental variables estimate is the coefficient on intervention use in a two-stage least squares regression of contraception adoption on assignment and intervention use.

^eN/A: not applicable.