Perceived Risk of Pregnancy and Demand for Contraception

Two information interventions are delivered through: (1) a video message from a medical doctor from the same region as the participant, (2) a sticker replicating an infographic message used in the video and (3) scripted messages delivered by the enumerator as part of a face-to-face interview. The sticker is stuck on a small gift which is expected to be consumed within one month of receipt. The control group sees a placebo video which forms part of the treatment video and receives the same small gift covered with a blank sticker instead.

2024-02-14

2024-05-17

Number of months using contraception in the past 12 months measured at endline. This includes any form of contraception/doing anything specifically to avoid pregnancy including effective traditional methods such as cycle-awareness methods and withdrawal.

The time period may be slightly longer than 12 months if the endline survey takes place more than 12 months after the intervention - this will be determined prior to the endline survey and registered accordingly. Use will be self-reported but there will be random checks at health providers (for consenting women). Respondents will be informed that there will be checks.

Secondary contraception outcomes:
1. Perceived probability of intended use measured right after receiving the information message.
2. Number of months using modern contraception in the past 12 months measured at endline.
3. Whether currently using contraception at follow-up.
4. Whether currently using a modern method at follow-up.

Fertility outcomes:
5. Whether the woman has had a new pregnancy between the intervention and the endline survey.
6. Whether the woman has had a new birth between the 9th month following the intervention and the follow-up survey.

Beliefs about pregnancy risk absent contraception:
7. Posterior belief about pregnancy risk absent contraception (from 0 to 20 out of 20 women) measured immediately after receiving the information message.
8. Posterior belief about own pregnancy risk absent contraception (from 0 to 20 out of 20) measured right after the intervention.
9. precision of the above belief (i.e., the difference between perceived maximum probability and minimum probability)
10. Belief update (i.e., the difference between posterior and prior belief) about own pregnancy risk absent contraception (from 0 to 20 out of 20) measured right after the intervention.
11. Endline belief about own pregnancy risk absent contraception (and absent current pregnancy, if applicable) (from 0 to 20 out of 20). Note that, since this risk may be affected by the pregnancy outcome, which in turn may be affected by the intervention, finding a significant effect on beliefs about own pregnancy risk immediately after the intervention but not at follow-up does not necessarily imply fading out.

Beliefs about husbands’ preferences – the relevant comparison here is control vs. the treatment group receiving information about husbands’ preferences:
12. Index of posterior beliefs about husband’s preferences measured right after the intervention. The index will be constructed using the methodology in Kling et al. 2007 and will summarize the following variables:
a. Perceived probability that there is at least one contraceptive method which the woman’s husband would approve of her using,
b. Precision of the above belief (i.e., the difference between perceived maximum probability and minimum probability),
c. Perceived desire of husband to delay pregnancy for at least one year (i.e., an indicator variable equal to 1 if the woman thinks that her husband either wants no more children or wants another pregnancy but only after at least one year), and 0 otherwise.
d. Perceived accuracy (out of 20) of the woman’s perception about her husband’s desire to delay pregnancy for at least one year.

13. Two standardized indexes, constructed using the methodology in Kling et al. (2007), of:
a. Posterior perceived husband’s preferences measured just after the intervention (using the variables 12.a and 12.c entering the index listed as secondary outcome 12)
b. Posterior perceived accuracy of women’s perception of her husband’s preferences measured just after the intervention (using the variables 12.b and 12.d entering the index listed as secondary outcome 12).

Secondary contraception outcomes:
1. Perceived probability of intended use measured right after receiving the information message.
2. Number of months using modern contraception in the past 12 months measured at endline. (The 12 months may be slightly extended, to match the duration covered by the primary outcome). Use will be self-reported but there will be random checks at health providers (for consenting women). Respondents will be informed that there will be checks.
3. Whether currently using contraception at follow-up: (a) self-reported and (b) elicited using a list experiment
4. Whether currently using a modern method at follow-up.

Fertility outcomes (self-reported):
5. Whether the woman has had a new pregnancy between the intervention and the endline survey.
6. Whether the woman has had a new birth between the 9th month following the intervention and the follow-up survey.

Beliefs about pregnancy risk absent contraception:
7. Posterior belief about pregnancy risk absent contraception (from 0 to 20 out of 20 women) measured immediately after receiving the information message. Question: “Out of 20 women of the same age and with an identical lifestyle as you, how many do you think would get pregnant within the coming 12 months, on average, if they did not use any contraception?”
8. Posterior belief about own pregnancy risk absent contraception (from 0 to 20 out of 20) measured right after the intervention.
9. precision of the above belief (i.e., the difference between perceived maximum probability and minimum probability, set to 0 if the woman says she is sure of her response, in which case we do not ask about minimum and maximum perceived probability)
10. Belief update (i.e., the difference between posterior and prior belief) about own pregnancy risk absent contraception (from 0 to 20 out of 20) measured right after the intervention.
11. Endline belief about own pregnancy risk absent contraception (and absent current pregnancy, if applicable) (from 0 to 20 out of 20). Note that, since this risk may be affected by the pregnancy outcome, which in turn may be affected by the intervention, finding a significant effect on beliefs about own pregnancy risk immediately after the intervention but not at follow-up does not necessarily imply fading out.

Beliefs about husbands’ preferences – the relevant comparison here is control vs. the treatment group receiving information about husbands’ preferences:
12. Index of posterior beliefs about husband’s preferences measured right after the intervention. The index will be constructed using the methodology in Kling et al. 2007 and will summarize the following variables:
a. Perceived probability that there is at least one contraceptive method which the woman’s husband would approve of her using,
b. Precision of the above belief (i.e., the difference between perceived maximum probability and minimum probability, set to 0 if the woman says she is sure of her response, in which case we do not ask about minimum and maximum perceived probability),
c. Perceived desire of husband to delay pregnancy for at least one year (i.e., an indicator variable equal to 1 if the woman thinks that her husband either wants no more children or wants another pregnancy but only after at least one year), and 0 otherwise.
d. Perceived accuracy (out of 20) of the woman’s perception about her husband’s desire to delay pregnancy for at least one year.

13. Two standardized indexes, constructed using the methodology in Kling et al. (2007), of:
a. Posterior perceived husband’s preferences measured just after the intervention (using the variables 12.a and 12.c entering the index listed as secondary outcome 12)
b. Posterior perceived accuracy of women’s perception of her husband’s preferences measured just after the intervention (using the variables 12.b and 12.d entering the index listed as secondary outcome 12).

Listing phase: 98 enumeration areas were randomly selected in each of the five study states and all households in the resulting 490 enumeration areas were listed to identify cohabiting women aged 18-37. Each woman was then screened individually to establish whether she met the following additional eligibility criteria: her husband has not been away more than one month at a time in the previous 12 months or, if he has, has been returning at least once a month in the past 6 months; she is not pregnant; she does not self-declare as infecund; and she either: (criterion 1) says she does not want to have another child or wants to wait at least 2 years before getting pregnant or (criterion 2) has a child aged between 4 and 12 months.

Experimental sample: We first dropped 7 enumeration areas which were deemed unsafe. One eligible woman per household was then drawn at random. The resulting sample of eligible women was then stratified by state and age group (18-24, 25-34, 35-37) and 1400 women per state selected randomly to form the experimental sample.

Randomization into three experimental arms (control, treatment 1 and treatment 2) at the woman level: The experimental sample was stratified into 36 strata formed by the following variables: criterion 2, whether the woman answered "17" or more during the screening interview when asked "Out of 20 sexually active women of reproductive age, how many do you think would get pregnant within the coming 12 months, on average, if they did not use any contraception?", her age group, whether she is muslim and, if muslim, whether she is in a polygamous union. We then randomized assignment to the three experimental arms within stratum, and assigned misfits so as to ensure no more than a difference of one subject per arm within each stratum.

Informed consent was/will be sought at each point of contact: household listing, screening interview, RCT enrollment (baseline), endline survey.

Not available

Randomization done in office by a computer, using the command randtreat due to Carril (2017).

Reference: Carril, A. 2017. Dealing with misfits in random treatment assignment. Stata Journal 17: 652-667.

individual (after retaining only one randomly selected eligible individual per household)

No

7,000 women

7,000 women

control: 2,337; Treatment 1: 2,334; Treatment 2: 2,329.