Improving Tax Administration in Weak States: Evidence from the DR Congo

How to cite

Danner, Sarah; Pomeranz, Dina; Weigel, Jonathan; Laroche, Arthur; Ngoma, Marina; Tourek, Gabriel. 2025. "Improving Tax Administration in Weak States: Evidence from the DR Congo." AEA RCT Registry. AEARCTR-0017355. https://doi.org/10.1257/rct.17355-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Sarah Danner* , Dina Pomeranz , Jonathan Weigel , Arthur Laroche , Marina Ngoma , Gabriel Tourek

* Primary investigator

Registry ID
AEARCTR-0017355
DOI
https://doi.org/10.1257/rct.17355-1.0
First registered
Last updated
Status
On going
Trial dates
2025-08-01 – 2026-07-31
Countries
Private
JEL codes
H2, H3, D73
Registry record
http://www.socialscienceregistry.org/trials/17355

Abstract

This project investigates four key levers of tax administration in a large-scale field experiment embedded in the 2025 property tax campaign in Kananga, D.R. Congo. First, we investigate the cost-effectiveness of in-person collection, a widely used approach in low-income countries that likely boosts compliance but is administratively costly and opens scope for corruption. Bill delivery occurs everywhere, but half of neighborhoods do not receive follow-up in-person collection visits; owners there must pay via other payment modalities. Second, we examine the introduction of mobile money tax payments. Exploiting the first-ever rollout of this payment option in Kananga, we provide information and instructions on paying taxes via mobile money to a random sample of property owners. Third, we test whether paying in installments can raise compliance and revenue, and to what extent this varies by access to mobile money payment. Finally, we cross-randomize a high-frequency personalized deterrence SMS treatment. We assess how these four levers shape compliance, revenue, and the cost-effectiveness of tax collection.

Experimental design

In many low-income countries, tax inspectors collect payments in person. This approach reduces transaction costs for taxpayers and may enhance willingness to pay taxes by strengthening tax morale or increasing the perceived risk of enforcement. However, it is administratively costly and creates opportunities for corruption. We explore these tradeoffs by experimentally assessing the cost-effectiveness of in-person tax collection in the context of a large-scale property taxation campaign in the D.R. Congo. All neighborhoods receive tax bills, but only half of them receive follow-up tax collection visits in which they can pay directly to tax inspectors. Property owners in all neighborhoods can pay at the bank or the tax authority.

Moreover, in the same tax campaign, taxpayers are for the first time able to pay via mobile money. Recent evidence from other sectors suggests that digitization can reduce leakage, improve accountability, and lower transaction costs (Muralidharan et al., 2016; Dodge et al., 2023). Mobile money, in particular, has emerged as a scalable tool for delivering transfers and payments in low-infrastructure environments (Suri and Jack, 2016; Suri, 2017). We exploit the novelty of this payment modality in Kananga and randomly provide information about mobile money payments at the property level. Our design allows us to compare mobile money with in-person collection, evaluating their impacts on compliance, revenue, and cost-effectiveness.

We explore two additional classic levers of tax administrators.

First, it is well known that compliance with the property tax — a bulky annual payment — is often constrained by liquidity constraints (Brockmeyer et al., 2023; Wong, 2020). Such constraints are even more likely to bind and hold down tax compliance in low-income countries (Bergeron et al., 2024). We therefore randomly provide property owners with the option to pay their property tax liability in installments. This intervention can increase compliance without increasing revenue if taxpayers only pay one installment (and thus essentially receive a discount). We thus expect limited effectiveness among owners who don’t receive information about mobile money. However, we expect that property owners with the option to pay via mobile money with installments will increase compliance by a large enough amount to also increase revenue.

Second, we implement a high-frequency personalized deterrence SMS treatment. We are not interested in horse-racing an enforcement message versus a public goods message like many past studies.
Rather, we constructed a single bundled treatment that, based on our reading of the literature, we expect to have the large possible impact on compliance (for a message treatment). The treatment includes a series of personalized SMS messages focused on deterrence messaging: reminders about penalties and other enforcement actions. We are interested in the interactions this intervention has with our other treatments. In particular, we expect this treatment to boost the cost-effectiveness of the mobile money payment option. We also expect it to have a positive interaction effect with installments.

Each of these interventions is embedded in a large-scale property tax campaign run by the Kasaï-Central provincial tax authority. The different interventions are fully cross-randomized, enabling us to identify the most cost-effective bundle of interventions in such a setting.

With this experimental design, we hope to answer the following questions.

First, is in-person collection cost-effective? How does its cost-effectiveness vary geographically and along the property value distribution? Despite the fact that in-person tax collection was used in virtually every society historically and remains common in many low-income societies today, we know of no well-identified evidence on its cost-effectiveness. Our study will bring evidence to bear on this question in a real-world policy experiment.

Second, can low-capacity tax administrations use digital payment systems such as mobile money as substitutes for, or complements to, in-person tax collection? Despite the prevalence of mobile payment
technologies in low-income countries — and a growing literature on the use of technology by tax administrations at various stages of the collection pipeline1(Okunogbe and Tourek, 2024; Dzansi et al.,
2022; Mascagni et al., 2021; Das et al., 2023; Okunogbe, 2021) — little is known about whether, and how, digital payment technologies can be deployed effectively in contexts where compliance depends heavily on in-person collection. Apeti and Edoh (2023) examines mobile money as a tax payment modality, offering cross-country evidence that mobile money adoption is associated with higher tax revenues — particularly from direct taxes and especially in lower-capacity tax environments. However,
we know of no causally identified micro-level evidence on the impact of mobile money on tax compliance.

Third, how can mobile money be combined with other policy instruments — options to pay in installments and/or personalized high-frequency deterrence messaging — to maximize compliance and
cost-effectiveness? A growing wave of empirical studies on the public finance in developing countries focus on tax administration (Jensen and Weigel, 2025). We contribute by investigating a series of policy levers — in-person collection v. mobile payment, installments — and their interactions that have yet to be experimentally studied in the empirical PF-Dev literature.

Design details

Randomization method
Randomization is done by a computer.
Randomization unit
Individual treatments were randomly assigned within delivery zones and property value bands. The facilitation treatment was randomly assigned at the delivery zone level within strata.
Clusters
1,359 delivery zones
Observations
48,450 tax-assessable and non-exempted properties
Arms
In-person Tax Collection: N= 23,114 Mobile Money Information: N= 17,871 Installments: N= 8,386 High-Frequency Personalized Deterrence SMS: N=17,807
Primary outcomes
Primary outcomes, measured in administrative records, include: 1. Tax compliance (any payment and complete payment) 2. Tax revenue (in levels, using Poisson regression, and logs) 3. Cost-effectiveness We will also examine how the treatments shape the payment channel used (mobile money vs. cash). Cost-effectiveness is assessed based on the cost data detailed in the full PAP. We plan to study heterogeneity by property value, compliance history, neighborhood and collector characteristics, and baseline perceptions of the government.

Cited by

No citation data available for this registration yet.

Keywords