Technology and Local Tax Capacity in Ghana

Last registered on March 01, 2021


Trial Information

General Information

Technology and Local Tax Capacity in Ghana
Initial registration date
February 28, 2021

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
March 01, 2021, 10:41 AM EST

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



Primary Investigator

Harvard University

Other Primary Investigator(s)

PI Affiliation
International Growth Centre
PI Affiliation
International Growth Centre
PI Affiliation
Boston University

Additional Trial Information

In development
Start date
End date
Secondary IDs
This project evaluates the potential for technology to alleviate constraints on local tax capacity. Working in collaboration with a local government and a local private firm in Ghana, the project randomizes the availability of technology for tax collectors in the field. Collectors face several constraints to achieve their objectives of delivering bills and collecting tax payments which technology may partly remove. In addition, we study whether the presence of technology impacts property owners' belief about capacity of the state and engagement with local officials.
External Link(s)

Registration Citation

Dzansi, James et al. 2021. "Technology and Local Tax Capacity in Ghana." AEA RCT Registry. March 01.
Experimental Details


The project randomizes the presence of a technology tablet for collectors which are working in the field to deliver bills and potentially collect tax payments (payments can be made by transfer rather than in person). The tablet helps collectors navigate in the field and provides precise information on the location of the intended bill recipient. In addition, the tablet generates a digital ID upon any partial tax payment, which is shared with the property owner and the central finance office instantaneously.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Our main outcomes of interest are bill delivery and tax compliance. These are the core outcomes that the technology is meant to help improve. Collectors in the field face difficulties to locate tax-payers precisely. In addition, property owners may either unwilling or (temporarily) unable to comply with property taxes. The technology is hypothesized to positively alleviate both of these sets of constraints, and improve bill delivery and tax collection.
Primary Outcomes (explanation)
We measure tax compliance through two main variables: a dummy for whether any tax payment is made; and, a continuous measure of tax amount paid.

Secondary Outcomes

Secondary Outcomes (end points)
We are interested in several additional outcomes. At the collector level, we investigate the impact of technology on productivity, effort, and organization. At the household level, we investigate impacts on beliefs and views about local tax collectors and the local government. At the bill level, we study impacts on leakage.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
The sample contains 60 tax collectors that are assigned by the local government and the local private firm to work on the 45-day tax campaign which forms the basis for the evaluation. Each collector is randomly assigned to a collection unit, which is a bundle of 150 property parcel units that are geographically adjacent. In turn, each collector-collection unit pair is randomly assigned to either the treatment group or the control group. The treatment group is provided with a tablet, the control group is not. All other aspects of the collection process (including remuneration) are held constant between groups, and correspond to the status quo process that the local government and the private firm have designed prior to the intervention and are otherwise using in the present moment to collect taxes.
Experimental Design Details
More details are provided in the analysis plan
Randomization Method
Randomization done in office by computer
Randomization Unit
Each collector is first randomly assigned to a collection unit. Each unit contains 150 geographically adjacent property tax bills. In turn, each collector-collection unit is randomly assigned to either treatment (presence of technology) or control (absence of technology).
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
There will be 60 clusters, which correspond to the number of collector-collection unit pairs.
Sample size: planned number of observations
The experimental sample contains 9,000 property tax bills in the administrative data. In addition, the household survey will collect responses from 5,000 property owners.
Sample size (or number of clusters) by treatment arms
There will be 30 collector-collection unit pairs in the technology treatment, and 30 collector-collection unit pairs in the control group. There will thus be 4,500 tax bills in both the treatment and the control groups.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Power calculations were based on effect sizes obtained from the pilot conducted in September-October 2019 as well as administrative records of property tax bill values and past payments in the study sample. As noted, the primary outcome of interest is property tax bill delivery by collectors. Based on the results of the pilot, the power calculations assume a minimum detectable effect (MDE) of 0.91 standard deviations. With a sample size of 30 collectors in each arm, estimated power is above 0.9 for the main hypothesis test of impact of technology on bill delivery. For compliance effect of technology on property taxes collected, power is lower at about 0.77. We note that this result is based on effect size from the pilot that lasted only 30 days when the collectors had not completed their tax collection recovery efforts. At the bill level, we use effect size of amount of property tax payment of about 0.09 standard deviations. Using the distribution of tax payments made by the households in the previous period from the administrative records, estimated power to detect a similar effect is above 0.8.
Supporting Documents and Materials

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

Request Information

Institutional Review Boards (IRBs)

IRB Name
Harvard University
IRB Approval Date
IRB Approval Number
Analysis Plan

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

Request Information


Post Trial Information

Study Withdrawal

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

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials