Automating Public Distribution Programs - Experimental Evidence from Ghana's PREMIX

Last registered on January 01, 2024


Trial Information

General Information

Automating Public Distribution Programs - Experimental Evidence from Ghana's PREMIX
Initial registration date
July 14, 2023

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
July 19, 2023, 2:33 PM EDT

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

Last updated
January 01, 2024, 5:16 AM EST

Last updated is the most recent time when changes to the trial's registration were published.


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

Request Information

Primary Investigator

University of California, Berkeley

Other Primary Investigator(s)

PI Affiliation
University of Ghana
PI Affiliation
Georgia State University
PI Affiliation
World Bank

Additional Trial Information

On going
Start date
End date
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
We study Ghana’s Premix Fuel Program – a large “conditional” transfer program subsidizing premix fuel for fishermen in Ghana. Subsidized fuel is hoarded, creating artificial shortages and prices that defy the program’s spirit. We partner with the National Premix Fuel Secretariat, which manages the Program. The Secretariat/Government of Ghana is preparing to rollout a nationwide program that will automate access to subsidized fuel to monitor its distribution and reduce leakages. Some experimentation already took place, so the technical feasibility of the process is not in question. Following our interactions, the Secretariat is willing to randomize across 234 villages (which will be in cohorts) the switch from village committee-based distribution of PREMIX fuel to Automated PREMIX fuel Dispensing and Monitoring System that come with personalized canoe identification cards. We proceed as follows. First, multi-cohort randomization across the 234 villages (315 beaches) will determine the timing of the rollout. Second, survey of fishermen and villagers to track: fuel usage, leakage, poverty, and economic/environmental/social outcomes the automation may improve, including village-wide spillovers and general equilibrium effects. We implement innovative measures for fuel diversion and quality: (i) administrative data on amount of fuel delivered to the village versus survey data, including estimates of black-market prices and quantities. We deploy Quality Assurance Teams to measure the quality of premix fuel in circulation across villages. We also use high-frequency measures of air pollution by (i) installing outdoor monitors and (ii) use available satellite and remote sensing data.

We compare two distribution alternatives within Ghana’s PREMIX fuel assistance program:
(i) Village committee-based distribution (status quo) vs (ii) Automated dispensing and monitoring system (new monitoring technology).

We will use this to document how monitoring technologies improve the distribution and poverty effects of public assistance programs by helping to bring in (or include) poor excluded beneficiaries. Indeed, when and how "changes" in the delivery (or distribution) of public assistance programs reduce poverty is a significant yet poorly understood issue.
External Link(s)

Registration Citation

Annan, Francis et al. 2024. "Automating Public Distribution Programs - Experimental Evidence from Ghana's PREMIX." AEA RCT Registry. January 01.
Experimental Details


The Secretariat (or government partner) has provided listing of all the landing beaches (N=234 villages, where the beneficiaries are located and the automation program will be deployed). We will randomly assign these fishing villages via a multi-cohort village-level field experiment to examine the general equilibrium effects of the automation program. "Digitization" (or automation) consists of modernized stations, to be constructed and introduced nationally, where fishermen can access subsidized PREMIX fuel. This station includes an (i) automated dispenser system (ADS), which only operates with personalized-biometric cards that will be issued to all canoes / owners, (ii) an office for station manager and landing beach committee (LBC), (iii) toilet facility; along with (iv) a live CCTV camera. Its central goal is to prevent leakages or diversions in the PREMIX program. "Cohort" denotes the set of villages (which contain landing beaches) that the Secretariat has operationally chosen to automate at different milestones -- averaging around 50 villages per cohort.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
1) Premix fuel delivery and usage, including detailed information about business outcomes (e.g., lines of business: fishing vs non-fishing; employment; revenues; assets; profits; boat or canoe failures; fishing rates, etc.), beneficiaries’ perceptions about and preferences for the digitization program, likelihood of having access to premix fuel in the village, political affiliations and perceptions, prices beneficiaries pay for fuel (both premix and non-premix fuel), prices of fish, etc.

2) Leakages across fishing villages by (i) comparing (the Secretariat's) administrative data and (the PIs) survey data, including measurement of black market activities.

3) Poverty: The villages are low-income and poor environments. We field questions to directly examine poverty. We will adapt a recently developed measure of poverty, called the “Simple Poverty Scorecard”, that is rigorous, inexpensive, simple, and transparent (for details, see Schreiner 2015). We will complement this with data on:
*(i) Household expenses (food, bills, education, health, durables / appliances / accessories, personal care, durables) and
*(ii) Subjective and objective inclusion/exclusion indexes reflecting whether: every fisherman has the opportunity to official PREMIX, LBC leaders avoid playing favorites, fishermen in village treated fairly and/or full in access to official PREMIX, fishermen excited that LBC leaders will welcome and support them, fishermen feel like beneficiaries of program, fishermen perspectives are counted/included in PREMIX decision making by LBC leaders; fishermen have not had access to PREMIX for for long time and/or denied access to the official PREMIX.

4) Quality of premix fuel in circulation across villages by deploying NPA’s Quality Assurance Team to randomly conduct quality checks and certification tests at across select villages.

5) Air pollution and carbon emission across villages by deploying NPA’s Quality Assurance Team to install and monitor outdoor pollution trackers across select villages. We will complement this with available satellite and remote sensing data.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
In total:
1) Treatment: 60% of villages as Treatment villages (Automate: apply the automation technology "EARLIER" to enforce monitoring of fuel subsidies and avoid diversions of premix fuel; N=~140).
2) Control: 40% of villages as Control villages (Status Quo = no automation until "LATER" after 3-4 months per cohort; N=~94). For each cohort of fishing villages, after 3-4 months, all the control villages will be also digitized. This delayed automation of control villages agrees with the Secretariat’s resource and capacity constraints and have hence agreed to enroll their automation program according this design.

We will be able to detect automation effects by measuring differences between the Control and Treatment arms. We will also be able to detect village-wide spillovers and equilibrium effects by (i) comparing beneficiaries to non-beneficiaries in Treatment villages; (ii) measuring differences between villages with more vs less beneficiaries in Treatment villages, including (iii) measuring changes in prices across nearby commercial fuel outlets.
Experimental Design Details
Not available
Randomization Method
Computer software and covariate-constrained randomization (we have small number of clusters/villages per cohort), ensuring balance on relevant characteristics (Bruhn and McKenzie [2009]; Ivers et al. [2012])
Randomization Unit
Randomization is at the village-level.
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
Number of fishing villages: 234
Number of districts (larger administrative units containing multiple villages): 26
Number of regions (larger administrative units containing multiple districts): 9
Sample size: planned number of observations
Number of fishing villages: 234 *Number of (direct) beneficiaries: 234 villages x about 5 canoe owners per village = around 1,170 owners *Number of (direct) beneficiaries: 234 villages x about 5 canoe employees/workers per village = around 1,170 employees *Number of non-beneficiaries (indirect): 234 villages x about 5 non-beneficiaries per village = around 1,170 individuals not engaged in fishing business.
Sample size (or number of clusters) by treatment arms
140 villages (Treatment);
94 villages (Control program)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

Institutional Review Boards (IRBs)

IRB Name
University of Ghana
IRB Approval Date
IRB Approval Number
ECH 164/ 22-23
IRB Name
Georgia State University
IRB Approval Date
IRB Approval Number