Back to History Current Version

Evaluation of the Di Lottery RCT (Burkina Faso)

Last registered on February 06, 2018

Pre-Trial

Trial Information

General Information

Title
Evaluation of the Di Lottery RCT (Burkina Faso)
RCT ID
AEARCTR-0002718
Initial registration date
February 06, 2018

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
February 06, 2018, 6:17 PM EST

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Region

Primary Investigator

Affiliation
Mathematica Policy Research

Other Primary Investigator(s)

PI Affiliation
University of Ouagadougou II

Additional Trial Information

Status
On going
Start date
2009-07-31
End date
2020-12-31
Secondary IDs
Abstract
Access to irrigated land is a key factor in determining a household's income and its food security in rural parts of Africa. The construction of large-scale irrigation perimeters comprising networks of primary, secondary and tertiary canals and often times pumping stations is one avenue to providing access to irrigated land on a large scale. From 2009 to 2014, the Millennium Challenge Corporation (MCC) invested around 100M USD into the construction of an irrigated perimeter in the Boucle du Mouhoun region of Burkina Faso, the Di perimeter. After the original owners of the land were compensated through higher quality but smaller plots of land, MCC distributed 710 hectares of irrigated land via a public lottery, called the Di Lottery.

Five hundred and three Di Lottery winners were randomly selected among 1,528 participants. Lottery winners received leases for 710 hectares of irrigated land in the Di perimeter, training in agricultural technologies for irrigated land, and starter kits (land preparation, materials and inputs). Lottery winners are the treatment group. The 1,025 lottery participants who did not win the lottery constitute the lottery control group. In addition, there are applicants who were not admitted to the lottery because they did not meet an application score eligibility cutoff. These are a second control group for an RD design. Based on baseline data collected by a previous evaluator prior to the lottery, baseline observable characteristics of lottery winners and non-winners are balanced and have similar means.
The lottery offered plots of 2 hectares for rice or 1 hectare for polyculture, and participants disclosed their plot preferences prior to the lottery. About one-third of the households that participated in the lottery received land.
We will analyze the effect of winning the Di Lottery on agricultural practices, production, total agricultural income, overall household income and land tenure security by using the framework of a randomized control trial (RCT). Our analysis will incorporate lottery participants' plot preference, will interview members of both parts of split households, and will administer survey questions on communal land rights. We will undertake two rounds of quantitative data collection. The first round will be in October 2017. It will cover the 2017 planting season. The final round will be in 2020.

At a cost in U.S. dollars of about $39,700 (without Compact administration costs) or $45,000 (with Compact administration costs) per hectare, the benefit of winning the lottery ranks as one of the largest stakes in any RCT that we know of. For more detail, please see Ksoll et al. "Evaluation of the Burkina Faso Agriculture Development Project: Design Report." Washington, DC: Mathematica Policy Research, June 2017.
External Link(s)

Registration Citation

Citation
Bambio, Yiriyibin and Christopher Ksoll. 2018. "Evaluation of the Di Lottery RCT (Burkina Faso)." AEA RCT Registry. February 06. https://doi.org/10.1257/rct.2718-1.0
Former Citation
Bambio, Yiriyibin and Christopher Ksoll. 2018. "Evaluation of the Di Lottery RCT (Burkina Faso)." AEA RCT Registry. February 06. https://www.socialscienceregistry.org/trials/2718/history/25591
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details

Interventions

Intervention(s)
Access to irrigated land is a key factor in determining a household's income and its food security in rural parts of Africa. The construction of large-scale irrigation perimeters comprising networks of primary, secondary and tertiary canals and often times pumping stations is one avenue to providing access to irrigated land on a large scale. From 2009 to 2014, the Millennium Challenge Corporation (MCC) invested around 100M USD into the construction of an irrigated perimeter in the Boucle du Mouhoun region of Burkina Faso, the Di perimeter. After the original owners of the land were compensated through higher quality but smaller plots of land, MCC distributed 710 hectares of irrigated land via a public lottery, called the Di Lottery.

Five hundred and three Di Lottery winners were randomly selected among 1,528 lottery participants, with an additional 650-701 applicants not admitted to the lottery. Lottery winners received leases for 710 hectares of irrigated land in the Di perimeter, training in agricultural technologies for irrigated land, starter kits (land preparation, materials and inputs) and land tenure documents. Lottery winners are the treatment group. The 1,025 lottery participants who did not win the lottery constitute the first control group. The 650-701 applicants who were not admitted to the lottery are a second control group. The lottery offered plots of 2 hectares for rice or 1 hectare for polyculture, and participants disclosed their plot preferences prior to the lottery. About one-third of the households that participated in the lottery received land.
We will analyze the effect of winning the Di Lottery on agricultural practices, production, total agricultural income, overall household income and land tenure security by using the framework of a randomized control trial (RCT). Our analysis will incorporate lottery participants' plot preference, will interview members of both parts of split households, and will administer survey questions on communal land rights. We will undertake two rounds of quantitative data collection. The first round will be in October Februay/March 2018. It will cover the 2017 planting season, roughly between June 2016 and May 207. The final round will be in 2020.

Intervention Start Date
2009-07-31
Intervention End Date
2014-07-31

Primary Outcomes

Primary Outcomes (end points)
We specify three primary outcomes (total agricultural production, total agricultural income, and total household income) for which we will adjust for multiple hypothesis testing. Following suggestions by Schochet (2008), we choose these three outcomes as they are priority outcomes for the Di Lottery. We consider the analysis of secondary outcomes (e.g., use of inputs, components of total household income, migration) and the subgroup analysis as exploratory and will not adjust for multiple hypothesis testing, as suggested by Schochet (2008).
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Investments in child education; land tenure security and access to communal land outside the perimeter
Secondary Outcomes (explanation)
Education expenses by Di Lottery beneficiaries versus control, years of education of children of beneficiaries and in beneficiary households; hectares of communal land to which households have access outside the perimeter

Experimental Design

Experimental Design
Recruiting applicants and selecting Di Lottery beneficiaries was a multi-step process. The lottery was announced over the radio, with posters and through postings in local administrations (town halls). The ADP invited individuals to apply for the lottery who were (1) not PAPs; (2) 18 years of age or older; (3) residents of one of the six provinces of the Boucle du Mouhoun region (Kossi, Banwa, Mouhoun, Bale, Nayala, Sourou); and (4) able to list at least two other people age 15 or older who could cultivate the plot of land with the applicant. Initially, 2,178 of the applicants were determined to meet all four criteria. However, applicants were allowed to contest the decision and ask for a review. As a result, an additional 51 applicants were deemed eligible, bringing the total to 2,229 eligible applicants.
Among eligible applicants, admission to the lottery was determined through a points-based system designed to (1) select applicants with higher expected benefits (for example, applicants received more points when they owned certain machinery); and (2) meet distributional objectives (for example, female applicants and younger applicants received additional points).
Applicants provided the following information as part of their application package:
-Number of household members aged 15 and above who would be available to could help with work on the land
-Applicant's experience with irrigation
-Any ownership of irrigated land in other AMVS perimeters
-Participation in MCC-sponsored training activity
-Age and gender of the applicant
-Type of agricultural equipment owned by the applicant
-Level of debt
-Location of residence
Applicants were aware of the four eligibility criteria, the information they were scored on, and the associated scores, but they were not initially aware of the exact threshold that would determine participation in the lottery. In order to make the selection process transparent, all application information was made public in multiple locations (for example, at local town halls). The accuracy of the application documents was verified by the training implementer, for example, by cross-checking debt with farmers' cooperatives and land ownership with water-user associations.
The cut-off for participation in the lottery was set at 60 points. Given the number of male and female applicants scoring 60 points or more, this cutoff made it highly probable that at least 20 percent of beneficiaries would be female. Applicants with a score of 60 and above participated in the public lottery, which was held in February 2014. The lottery was conducted by the national lottery company LONAB (Loterie Nationale Burkinabe) and validated by a cabinet of lawyers.
Because there were two standard types of plots for which leases were distributed, applicants had to rank the two types of plots as their first and second choices. One plot was two hectares of land suitable for cultivating rice, and the other was one hectare that was suitable for polyculture. Just under 10 percent of the applicants stated that they would not accept a second choice.
We compare outcomes of lottery winners to outcomes of non-lottery winners. We account for different probabilities of selection because the probability of winning the lottery depended on preferences for rice and polyculture plots. In addition, in some households there were multiple applicants for the lottery. We also drop multiple applicant households in some analyses.
Experimental Design Details
Randomization Method
Applicants with a score of 60 and above participated in a public lottery. The lottery was conducted by the national lottery company LONAB (Loterie Nationale Burkinabe) and validated by a cabinet of lawyers. To conduct the lottery, LONAB used a standard tombola set-up: each lottery participant was assigned a number, and balls with these numbers were put into a large container. In addition, two additional tombolas contained numbered balls identifying the available rice and polyculture plots. We call these the rice and polyculture tombolas.
In order to choose lottery beneficiaries, the following process was repeated until there were no plots remaining. To start the process, a volunteer from the public would draw the name of a lottery participant from the main tombola. If the selected participant's first choice of plot was available, a plot of this type was selected for the participant from the rice or polyculture tombola, whichever was the participant's first choice. If the selected participant's first choice was not available, but he or she had indicated a second choice, a volunteer drew an available plot from the tombola containing plots that were the participant's second choice.
If the first choice was not available, and the applicant had not indicated a second choice, the person did not receive a plot of land, and he or she was instead on the waiting list. In this case, the selection process started from the beginning. Of 1,528 participants, 5083 beneficiaries were selected.
After all plots had been assigned, 150 additional names were drawn to constitute a waiting list. If an applicant did not start working on the plot he or she received, applicants from the waiting list were meant to benefit from the unoccupied land. At the time of this writing, however, no applicants from the waiting list have benefited from a vacant plot.
Randomization Unit
Lottery participant
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
1
Sample size: planned number of observations
1,528 Lottery participants and between 650 and 701 RD controls
Sample size (or number of clusters) by treatment arms
503 treatment, 1025 RCT control, 650/701 RD controls
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
To be successful, an impact evaluation must have a large enough sample. For the Di Lottery RCT, the sample size is limited to the number of Di Lottery beneficiaries and the control group. Through a statistical power analysis, we have calculated the magnitude of potential impacts we could detect given the Di Lottery sample size, which we refer to as minimum detectable impacts (MDIs). Table IV.4 in Ksoll et al. (2017) provides MDIs for agricultural income for the entire sample and 20%, 50% and 80% sub-samples and details of the calculation. For the entire sample, the MDI (% of the mean) is 15.52% and the MDI (% of anticipated Di profit) is 3.52%. Given the sample sizes in the Di Lottery RCT, the second row of Table IV.4 shows that the MDI for income is 16 percent of agricultural incomes in a sample of farmers from the same region. This is a reasonable increase; for 70 percent of treatment households, the additional land received through the lottery makes up 25 percent or more of the land they cultivated at baseline. As the irrigated land is, by all accounts, much higher in quality, we found that a 16 percent change in agricultural income is plausible. In a review of the empirical evidence on the effects of irrigation on poverty, Hussain and Hanjra (2004) found several studies that reveal that a 50 percent income gain was attributable to the introduction of irrigation. This is corroborated by a rigorous evaluation in Mali that found that small-scale irrigation raises farmers' agricultural revenues by 30 percent (Dillon 2011). Both of these studies focused on projects that supported irrigation but did not make more land available to farmers for cultivation. Because the Di Lottery does give farmers more land for cultivation and also provides irrigation for this new land, the impacts found in these previous studies are likely to be the minimum level we would expect under the Di Lottery. The MDIs for females are smaller than the impacts described in Hussain and Hanjra (2004) and in Dillon (2011) and are only 8 percent of the expected additional profit from one hectare of land on the Di perimeter. We therefore intend to conduct a subgroup analysis even for the relatively small subgroup of female lottery winners.
Supporting Documents and Materials

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
IRB

Institutional Review Boards (IRBs)

IRB Name
Centre National des Statistiques
IRB Approval Date
2017-12-29
IRB Approval Number
AP2017001CNSCS4
IRB Name
Health Media Lab
IRB Approval Date
2017-09-14
IRB Approval Number
N/A
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials