Progressivity, Fairness, and Tax Capacity: Evidence from the D.R.Congo
How to cite
Ngoma, Marina Mavungu; Weigel, Jonathan; Bergeron, Augustin; Laroche, Arthur; Naritomi, Joana; Tourek, Gabriel. 2024. "Progressivity, Fairness, and Tax Capacity: Evidence from the D.R.Congo." AEA RCT Registry. AEARCTR-0014314. https://doi.org/10.1257/rct.14314-1.2
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
The innovation of progressive taxation in the early 20th century accompanied some of the largest increases in tax revenue in Europe and the US and is a core feature of most tax systems in today's developed countries. In developing countries, however, states often use simplified tax instruments that are comparatively less progressive. This project will explore the randomized introduction of progressive property taxation in the D.R. Congo, a low-income country with weak fiscal capacity. In collaboration with the provincial government of Kasai-Central, the study will compare neighborhoods in the city of Kananga assigned to (1) a progressive property tax rate schedule, (2) a proportional schedule, or (3) the status quo flat fee schedule. We will study the effects on total revenue, household compliance, perceptions of fairness and tax morale, and general attitudes toward the government. Cross-randomized household-level information treatments will help distinguish between ability-to-pay and tax-morale mechanisms.
Intervention
We partner with the Provincial Government of Kasaï-Central to study the rollout of more progressive property taxation at scale in the city of Kananga, Democratic Republic of Congo.
The first step of this project is to assist the government with the ``computer assisted mass assessment'' (CAMA) of all buildings in Kananga, following a procedure that has been widely used in sub-Saharan African cities. We first collect high-resolution drone imagery of the city and used AI-based roof detection, supplemented by human verification, to estimate the surface area of all buildings in the city. We then combine this information with a survey recording building characteristics (e.g., wall and roof type) and compound characteristics (e.g., land area, fence type, outbuildings). We then estimate a model in a training data set of 1,500 randomly sampled properties assessed by valuation experts in the government cadastral office. Finally, we predict values using the estimated coefficients.
Armed with estimated building values, the second step is to choose the system of property tax rates given the government's objective of increasing revenue while minimizing welfare losses and raising the perceived fairness of taxation. This is where the randomized controlled trial comes in. We will compare neighborhoods assigned to proportional property tax rate, a progressive system of tax rates, or the status quo fixed fee tax schedule. The proportional and progressive treatments correspond to competing notions of tax equity: the former equalizes the property tax burden as a share of wealth, while the latter treatment places a higher burden on wealthier properties. We study the impact of these treatments on compliance, revenue, and perceptions of fairness.
Experimental design
The 106,111 assessable buildings in Kananga will be randomly assigned to either the existing flat fee property tax schedule (control), a proportional tax rate schedule (treatment 1), or progressive tax rate schedule (treatment 2).
1. Buildings assigned to the flat fee schedule (control)—the status quo in Kananga— will face the same annual liability within two coarse value bands determined by the materials of the principal construction.
2. Buildings assigned to the proportional tax schedule (treatment 1) will face the same (flat) tax rate. The liability of each building will be a constant share of its estimated value.
3. Buildings assigned to the progressive tax schedule (treatment 2) will face higher tax rates according to the building’s value band: i.e., more valuable buildings will face higher tax rates than less valuable buildings.
There are two potential mechanisms through which greater progressivity may impact compliance and revenues. First, according to an ability-to-pay mechanism, a tighter correspondence between tax liability and wealth (property value) could increase the likelihood that property owners pay if ability to pay and wealth are correlated. Second, according to a tax-morale mechanism, a progressive rate schedule could increase compliance and revenue by improving the perceived vertical equity and fairness of the tax system and thus strengthening non-pecuniary motivations to pay. The tax system's perceived fairness and vertical equity are thought to be essential components of tax morale in rich and poor countries alike. If the arrival of progressive taxation similarly causes citizens to view the tax system as fairer than the status quo, they might also become more likely to pay.
To disentangle these mechanisms, we randomize information embedded in tax bills delivered to property owners. In particular, we randomize the following messages on tax bills: (1) a building's location in the distribution of building values; (2) the location in the distribution of building values and the full distribution of tax rates; (3) a detailed description of how building values were estimated; or (4) a placebo message noting the importance of paying the property tax.
In addition, we embed several cross-randomizations at the property level to further help understand mechanisms and rule out alternative interpretations. First, randomly selected property owners will have an additional enforcement message on the back of their tax bill to study the interaction of enforcement beliefs with the intervention. Second, tax collectors will be assigned a list of property visit appointments each day. We will examine compliance during these visits to study the impact of the intervention absent collector targeting.
To collect data about the intervention, independent enumerators will additionally conduct a baseline, midline, and endline survey.
The treatment is assigned at the neighborhood level. We also implement a stratified randomization. The strata were constructed by first grouping areas based on geographic communes and then splitting these groups at the median property value within each polygon. Within each polygon, we further stratified individual treatments based on the specific polygon and the property value bands.
Design details
Cited by
No citation data available for this registration yet.