Back to History Current Version

Public Disclosure and Property Tax Compliance: A Field Experiment in Kampala, Uganda

Last registered on January 05, 2022

Pre-Trial

Trial Information

General Information

Title
Public Disclosure and Property Tax Compliance: A Field Experiment in Kampala, Uganda
RCT ID
AEARCTR-0005328
Initial registration date
May 14, 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
May 17, 2021, 10:35 AM EDT

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

Last updated
January 05, 2022, 11:38 AM EST

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

Locations

Region

Primary Investigator

Affiliation
London School of Economics

Other Primary Investigator(s)

PI Affiliation
International Growth Centre

Additional Trial Information

Status
In development
Start date
2020-11-01
End date
2022-05-31
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Property taxes are an important source of local government revenue. However, compliance rates can be extremely low in some developing country cities. This project studies the effectiveness of policy measures to influence voluntary tax compliance. We consider a specific type of policy intervention: public disclosure of tax compliance. The central question is "Does public disclosure of tax behaviour raise compliance?". We consider two channels of effect through which disclosure could affect compliance. The first is a direct effect; "Do taxpayers change their compliance behaviour when they know that it will be publicised?", and the second is a knock-on effect; "Do tax payers change their behaviour when they are publicly notified of the behaviour of others?". Further, we compare the relative effectiveness of two different types of disclosure: publicly reporting tax delinquents (shaming) and publicly honoring tax compliers. In theory, shaming delinquents could be less effective than honoring compliance. First, honoring compliers may be more effective in a low-compliance equilibrium because shaming provides a relatively uninformative signal when delinquency is very common. Second, honoring compliers may be more effective at creating positive knock-on effects because reports of compliers raises the visibility of good role models while reports of delinquents could demoralise taxpayers. To study these questions we partner with the Kampala Capital City Authority to randomise variations of their proposed tax reporting scheme. In this experiment treatment notifications are administered by text message to roughly 80,000 unique phone numbers associated with roughly 200,000 tax-owing properties.
External Link(s)

Registration Citation

Citation
Manwaring, Priya and Tanner Regan. 2022. "Public Disclosure and Property Tax Compliance: A Field Experiment in Kampala, Uganda." AEA RCT Registry. January 05. https://doi.org/10.1257/rct.5328-2.1
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)
In collaboration with the Kampala Capital City Authority (KCCA), we conduct a randomized evaluation of their 'Enhancing Property Rates Compliance through SMS' intervention. In this experiment, treatments are notifications administered by text message to roughly 80,000 unique phone numbers associated with roughly 200,000 tax-owing properties. There will be two rounds of treatment messages (two interventions). In the first round, taxpayers will be notified whether and how their behaviour will be disclosed. Some will be told that they will be honored publicly if they comply, some will be told that they will be publicly reported if they do not comply, and some will simply be reminded to pay. Then, based on their compliance at the end of the financial year (June), they will have their behaviour disclosed in the way outlined in their treatment message. In the second round, taxpayers will receive the disclosure reports of others' behaviour. Some will receive no report, some will receive reports of delinquents, and some will receive reports of compliers. Following the notifications, compliance will be measured again using administrative records (December). Treatments in the two rounds will be completely cross-randomized. This allows identification of both the direct and knock-on effects of public reporting.
Intervention Start Date
2021-05-17
Intervention End Date
2021-12-31

Primary Outcomes

Primary Outcomes (end points)
The primary outcome is tax compliance. There are different ways to measure compliance. To be specific, we define a property as compliant if that property's annual liability is paid before June 30th, 2021 (the end of the financial year).
Primary Outcomes (explanation)
We focus on the June 30th date because it is the de facto deadline for tax payments. While the KCCA considers December 31st as the deadline, the vast majority of taxpayers do not recognize this deadline. According to our baseline survey, 45% of taxpayers do not know which month property rates are due, and another 43% believe that rates are due in June or July.

We measure compliance as having paid at least the annual liability because it is a straightforward measure of compliance, and implicitly accounts for some properties having larger liabilities than others. Also, it simplifies the problem of partial compliance. Almost no properties are ever partially compliant by this measure, wheras partial compliance is common at the property owner level (property owners pay tax on some of their properties but not others).

Secondary Outcomes

Secondary Outcomes (end points)
As a secondary outcome we consider the size of tax payments, i.e. the total sum of payments made towards a property before June 30th, 2021.

We will also consider heterogeneity of our primary outcome along two key dimensions: (1) the wealth of the property owner, and (2) perceived norms around tax compliance.
Secondary Outcomes (explanation)
We add the measure of total sum of payments to capture both heterogeneity in treatment effects by liability and any effects on payments beyond annual liability amount (e.g. debt carried over from previous years).

The first dimension of heterogeneity will be measured in the administrative data by taking the sum of property value for each owner across all of their properties. The second is more difficult to measure. Using the administrative data, we will proxy for the percieved norms around tax compliance with the observed compliance rate at the local village level. That is, we will compare treatment effects across villages of varying aggregate compliance rates.

Experimental Design

Experimental Design
Our experiment randomly assigns properties to treatment groups. Randomization is clustered at the phone number level, so that all properties with the same phone number receive the same treatment. We use machine randomization and treatments are administered digitally.
Experimental Design Details
We measure outcomes at the property level, so our observational unit is the property. Each property has a registered owner with contact details and some owners can have the same phone number as one another. Therefore we randomize our treatments at the phone number level to avoid spillovers across owners sharing the same phone number.

Randomization is done by a computer, i.e. with a script in Stata. We employ a block-randomized design, stratifying on past (2019/2020) compliance (zero or partial/full compliance) and ventiles of total property value. Both strata are measured at the phone number level. Therefore treatments are stratified by 40 strata. Phone numbers within each strata are randomly assigned to one of our 8 treatment groups. Properties are nested within phone numbers, so each property is assigned to a unique treatment group.

Treatments are administered digitally by the KCCA as follows. After randomization, the stata script generates lists of properties for each treatment group. The KCCA then uploads the list to a mass text messaging system that sends standardized messages to each phone number on the list.
Randomization Method
Randomization is done by a computer.
Randomization Unit
The randomization is clustered at the phone number level.
Was the treatment clustered?
Yes

Experiment Characteristics

Sample size: planned number of clusters
62,758 phone numbers
Sample size: planned number of observations
151,649 properties
Sample size (or number of clusters) by treatment arms
Control:16,692 Phone Numbers
Reporting (SMS): 8,386 Phone Numbers
Reporting (Web): 8,369 Phone Numbers
Recognition (SMS): 8,352 Phone Numbers
Recognition (Web): 8,341 Phone Numbers
Information (Enforcement): 5,579 Phone Numbers
Information (Reciprocity): 5,582 Phone Numbers
Information (Relationship Management): 5,576 Phone Numbers





Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Simple power calculations were conducted using administrative data from the 2019/2020 Financial year. For the primary outcome (property compliance), the projected mean is 0.056 and the standard deviation is 0.23. According to Stata’s power command, we would need a sample of 6,281 units per treatment arm to detect an effect of 0.05 standard deviations (about 1.15 percentage points) on the control group mean. For the three main treatment groups (reporting, recognition, and information) our projected sample size of about 16,500 units per treatment arm can detect effects of less than 0.031 standard deviations (less than 0.71 percentage points) on the control group mean. For the disclosure sub-groups (SMS and Web) our projected sample size of about 8,000 units per treatment arm can detect effects of less than 0.038 standard deviations (less than 0.88 percentage points) on the control group mean. For the information sub-groups (enforcement, reciprocity, and relationship management) our projected sample size of about 5,500 units per treatment arm can detect effects of less than 0.044 standard deviations (less than 1.01 percentage points) on the control group mean.
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
LSE Research Ethics Committee
IRB Approval Date
2020-01-15
IRB Approval Number
1048
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?
Yes
Intervention Completion Date
December 31, 2021, 12:00 +00:00
Data Collection Complete
No
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
No
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials