Congo's Progressive Property Tax Experiment Raised Revenue by 56%

Experimental design

The study is a citywide field experiment conducted in partnership with a provincial government in the Democratic Republic of Congo, with a pre-analysis plan registered on the AEA RCT Registry (AEARCTR-0014314) in September 2024 and IRB approval from the University of California, Berkeley. Neighborhoods were the unit of randomization, assigned to either a proportional property tax schedule (a single rate applied to all assessed property values) or a progressive schedule (marginal rates rising with assessed value, structurally similar to an income tax bracket system but applied to property assessments instead of income). Randomizing at the neighborhood level, rather than letting owners self-select or applying the schedules city-wide sequentially, lets the researchers isolate the causal effect of the tax schedule design from confounding factors like neighborhood wealth, existing compliance culture, or enforcement capacity that happened to vary across the city for unrelated reasons.

Cross-randomized information treatments

Layered on top of the schedule randomization, the researchers also cross-randomized information treatments — different messaging sent to taxpayers about the rates. Some messages emphasized the taxpayer's own rate; others emphasized the rates other neighborhoods or property tiers faced; others framed the schedule in terms of its overall fairness. The point was to separate three possible channels through which a progressive schedule could affect compliance: taxpayers reacting to their own bill, taxpayers reacting to relative treatment versus others (a comparison or "fairness" channel), or taxpayers reacting to perceptions of the system's legitimacy as a whole. The finding was that compliance responses were driven overwhelmingly by a taxpayer's own assigned rate. Information about what others paid, or framing around the fairness of the schedule as a whole, made little independent difference. That's a methodologically useful result on its own: it suggests governments designing tax communications in similar settings can focus on making an individual's own liability and payment process clear, rather than investing heavily in fairness messaging or social-comparison framing.

Effective tax rates and the regressivity puzzle

One counterintuitive finding concerns effective tax rates — actual taxes paid as a share of property value, as opposed to the statutory rate written into the schedule. Effective rates declined as property value rose, and this pattern was most pronounced under the progressive schedule, not the flat one. In other words, in the schedule's initial rollout, the properties with the steepest statutory rates were also the ones with the biggest gap between what was legally owed and what was actually collected — a form of effective-rate regressivity sitting underneath a progressive rate structure on paper. This is an important distinction for anyone evaluating tax reforms from statutory design alone: a progressive-looking schedule can still be regressive in practice if compliance and enforcement don't scale with the size of the liability.

Enforcement changes the picture

After the initial experiment, the provincial government scaled the progressive schedule citywide and paired it with enforcement effort specifically targeted at high-value properties — the segment where the earlier effective-rate gap was largest. That combination reversed the pattern: statutory and effective rates converged, meaning the tax burden shifted more genuinely onto wealthier property owners once enforcement, not just rate design, targeted the top of the distribution. The methodological takeaway is that progressive rate design and progressive enforcement are complements, not substitutes — a progressive schedule without matching enforcement capacity at the top of the distribution risks looking progressive on paper while remaining close to flat in practice.

Study credibility and funding

The paper was pre-registered before data collection, reviewed by an institutional IRB, and disclosed funding from the Financing of Local Government Revenue Initiative (LoGRI), the Fund for Innovation in Development, USAID, the Weiss Fund, Harvard's Dean's Competitive Fund for Promising Scholarship, the Philomathia Foundation, the International Growth Center, the Swedish International Development Cooperation Agency, and the World Bank's Datax program — a mix of academic and development-finance funders with no single funder positioned to influence the result in a particular direction. The authors are affiliated with NBER's Public Economics and Development Economics programs; the paper circulated as NBER Working Paper 35536, issued July 2026.

Limitations and what's still open

The results come from one city in one country, and property tax administration varies enormously across (and within) developing countries depending on cadastral quality, local political economy, and the specific enforcement technology available. Whether a 56% revenue gain and the same top-and-bottom mechanism would replicate in a different administrative environment — a different country, a different city with a more or less developed property registry — is an open empirical question the authors don't claim to have answered. The result should be read as strong evidence that progressive property taxation is administratively viable in at least one low-capacity setting, not as a universal claim that progressivity always beats flat rates everywhere; the enforcement-pairing result in particular suggests the revenue and equity gains are conditional on state capacity being deployed deliberately at the top of the distribution, not automatic.

How to read the 56% number correctly

It's worth being precise about what the 56% figure does and doesn't say. It is a relative comparison between two randomly assigned groups of neighborhoods within the same city, over the same time period, collected by the same administration — not a before-and-after comparison, and not a claim about how much more revenue progressive taxation could raise in the abstract. That randomized, same-city, same-period design is what allows the causal claim: absent the schedule difference, the two groups of neighborhoods would be expected to generate similar revenue, so the 56% gap is attributable to the rate structure itself rather than to which neighborhoods happened to be richer or already better at complying.

Why the compliance response matters as much as the rate design

The paper's finding that low-value property owners paid more often under lower rates — enough to offset the smaller amount each owed — is a compliance-elasticity result, not just an arithmetic one. It implies that at the bottom of the property-value distribution in this setting, the binding constraint on revenue wasn't the rate itself but whether people paid at all. A schedule that lowers rates for the least valuable properties while raising them for the most valuable ones can, under the right compliance response, out-collect a uniform rate set anywhere in between — which is a substantially more actionable insight for a finance ministry than simply being told to "tax progressively."