A link to the study pre-analysis plan allows forecasters to read further if they want additional information.
Study: The Impact of a Multidimensional Program on Nutrition and Poverty in Burkina Faso
Authors: Adrien Boguen and Andrew Dillon
Authors: Adrien Boguen and Andrew Dillon
It is important to inform respondents about the intervention context.
Study introduction: A large number of people living in Burkina Faso live in poverty. This problem is compounded by the Sahel environment, which leaves households vulnerable to food insecurity. This impact evaluation examines the effects of a multi-faceted anti-poverty program in rural Burkina Faso.
Respondents should know who the intervention is targeting.
Target population: The program targets ultra-poor and poor households with a breastfeeding, pregnant women and/or a child under the age of five who are experiencing malnutrition. Ultra-poor and poor households were identified before randomization using quantitative and qualitative targeting methods. On average, 21 households per village were selected to be eligible for the program.
Respondents should know the sample size in the experiment, and the level of randomization.
Randomization: This study is a cluster randomized controlled trial. Randomization takes place at the village level: 168 villages were randomized into one of three treatment conditions or a control condition. Note that treatments varied slightly based on region. The main components of these three treatments are described below:
This study involves cash transfers. To understand the magnitude of the transfers, we present them in FCFA, PPP adjusted USD, and as a % of monthly transfers.
We also highlight the treatment timing (since the transfers were distributed during the lean season). We also highlight the treatment timing (since the transfers were distributed during the lean season).
We also highlight the treatment timing (since the transfers were distributed during the lean season). We also highlight the treatment timing (since the transfers were distributed during the lean season).
Treatment 1: Unconditional cash transfer. The cash transfer lasted a total of two years. In the first year, households received 20,000 FCFA per month for four months (~40% of monthly household consumption; ~100 USD, PPP adjusted), for a total annual transfer of 80,000 FCFA (~400 USD, PPP adjusted). In the second year they received 15,000 FCFA per month for four months (~30% of monthly household consumption;~75 USD, PPP adjusted), for a total annual transfer of 60,000 FCFA (~300 USD, PPP adjusted). Cash was distributed during the lean season between planting and harvesting (June-September).
Treatment 2: Unconditional cash and asset transfer. The cash transfer is as above. These households also receive a coupon for animals which can be exchanged at markets. The coupons varies by area in two ways. First, whether a household receives a coupon for sheep or poultry varied based on the suitability of the village for raising each animal, as determined by a local implementing partner. Second, the amount of the voucher varied by region: poultry coupons were worth either 25,000 or 44,000 FCFA, and sheep coupons were worth either 90,000 or 114,000 FCFA.
Treatment 3: Unconditional cash and asset transfer and nutrition intervention. The cash transfer and assets intervention are as above. These households also receive nutritionally fortified flour for children aged 6 to 23 month. Pregnant or lactating women receive bread flour on a monthly basis.
Treatment 2: Unconditional cash and asset transfer. The cash transfer is as above. These households also receive a coupon for animals which can be exchanged at markets. The coupons varies by area in two ways. First, whether a household receives a coupon for sheep or poultry varied based on the suitability of the village for raising each animal, as determined by a local implementing partner. Second, the amount of the voucher varied by region: poultry coupons were worth either 25,000 or 44,000 FCFA, and sheep coupons were worth either 90,000 or 114,000 FCFA.
Treatment 3: Unconditional cash and asset transfer and nutrition intervention. The cash transfer and assets intervention are as above. These households also receive nutritionally fortified flour for children aged 6 to 23 month. Pregnant or lactating women receive bread flour on a monthly basis.
A figure can be used to summarize the experimental findings and break up text blocks
This table highlights the key components of the interventions in a condensed format. Respondents are able to revisit this when they provide their predictions.
Respondents should know when outcomes are measured relative to treatment delivery.
Timing: The second wave of the intervention is estimated to be fully delivered by November, 2020. Endline is estimated to take place in April-May, 2020.
When we collected forecasts, the interventions had not been fully delivered. Additionally, some respondents who were not supposed to received interventions had been treated. We summarize this information in the table below. More generally, participants should be made aware of any uncertainty in treatment implementation.
Treatment delivery: The table below depicts the proportion of households in each group that received an intervention midline (after the first year of treatment should have been delivered).
- For the bold values, treatment should be 100% if all intended recipients were treated. For example, only 75.3% of households in T1 (cash) had received some cash at mindline.
- Red text indicates households that should not have received a treatment. For example, 1.8% of households in the control group had received some cash at midline.
Respondents should know the sample size in the experiment, and the level of randomization.
Randomization: This study is a cluster randomized controlled trial. Randomization takes place at the village level: 168 villages were randomized into one of three treatment conditions or a control condition. Note that treatments varied slightly based on region. The main components of these three treatments are described below:
Respondents should be provided with a comprehensive description of the predicted outcome.
Given issues with implementation fidelity, we emphasize that we are collecting predictions of expenditure differences based on assignment.
Given issues with implementation fidelity, we emphasize that we are collecting predictions of expenditure differences based on assignment.
One outcome we are interested is monthly household food consumption expenditure. This includes purchased food, home produced food, food received from other household members, friends and in the form of in-kind payments.
Please predict the difference in monthly health expenditure between households assigned to each of the three treatment groups and the control group.
Please predict the difference in monthly health expenditure between households assigned to each of the three treatment groups and the control group.
- This link takes participants to the two page description.
- We provide the control groups mean and standard deviation as a reference.
- An example can be useful for helping ensure respondents understand their predictions.
Notes:
- Click here for a reminder of the interventions and study background, which will open in a new window.
- Reference: At midline, the average monthly household food consumption expenditure in the control group was about 200 USD (ppp adjusted; 40,000 FCFA) with a standard deviation of about 130 USD (ppp adjusted; 26,000 FCFA).
- As an example, if you enter $13 it means you think average monthly household food consumption expenditure will be $13 USD higher in the treatment group. If you enter -$13 it means you think average monthly household food consumption expenditure will be $13 USD lower in the treatment group. If you enter 0 it means you think the program had no impact.
When providing details of an intervention, respondents may want to know how successful implementation was. For example, did all teacher attend all six training sessions? If this information is not yet known, it scan be good to state it explicitly, or to provide bounds (see the example for Bouguen and Dillon).
We bound numeric responses at +-1 SD to avoid confusion among survey respondents.
We chose not to randomize elicitation order for this question, since there is an intuitive ordering.
Here we bounded our slider scale at +-2 SD, but elicited predictions in raw units. To avoid acquiescence, the slider must be moved to continue to the next question (the default is not 0).
We chose not to randomize elicitation order for this question, since there is an intuitive ordering.
Here we bounded our slider scale at +-2 SD, but elicited predictions in raw units. To avoid acquiescence, the slider must be moved to continue to the next question (the default is not 0).
