Knowledge Without Taxation? Experimental Evidence from a Tax Education Program (Jensen, Kumar & Poensgen, 2026)

Summary

"Knowledge Without Taxation? Experimental Evidence from a Tax Education Program," by Anders Jensen, Nikhil Kumar, and Konstantin Poensgen, circulated as NBER Working Paper 36002 in November 2026, reports a randomized experiment on property-tax education in Ablekuma Central, an urban district in the Greater Accra Region of Ghana. Its abstract states the headline: the intervention "raises tax knowledge, which spreads through social networks to untreated neighbors, but does not increase tax payments."

The tax studied is the municipality's existing annual property rate, which the local government sets "on the property valuations (distinguished by property use and geographic area)" (Section 2.1, p. 7). The paper tests no land tax and does not say whether land or buildings is the valued base. Ghana's property rate falls on buildings, as King (2026) describes. The paper's relevance to the Georgist case is therefore indirect: it is evidence on what informing taxpayers can and cannot do for compliance with a recurrent local property tax where enforcement is weak.

Setting and Design

• Setting. The experiment ran in the fall of 2024 in a subset of the district's collection units (p. 15). Compliance with the property tax in Ghana "typically ranges between 5 and 15%," and in the control group 12 per cent of owners reported any payment at endline while 31 per cent reported receiving a bill (pp. 7-8; Table 3, p. 41).
• Sample. There were 24 frontline tax officials, the entire set available at the time, and 48 collection units grouped into 24 pairs of about 60 properties each (p. 15-16). The baseline survey covered 1,193 property owners; the endline covered 1,108, an attrition rate of 7 per cent that was balanced across treatment groups (p. 21 and Table A.5). The sample is limited to residential and mixed-use properties; purely commercial properties were excluded, as were the wealthiest and poorest units (pp. 15, 20).
• Baseline ignorance. Of owners, 92 per cent could not name a single property characteristic determining their liability, and 93 per cent could not name the institution responsible for valuation (pp. 1, 9).
• Four groups, not two. Collection-unit pairs were randomly assigned to a control group or one of three treatment groups, and within treated units about half of the surveyed owners were randomly selected for education (pp. 15-16). In the first treatment group (TG1) only owners were educated. In TG2 officials were also told, on their maps, which two areas contained many educated owners. In TG3 officials additionally received the same workshop. Results for owners are reported separately for the "TG-treated" (educated) and "TG-untreated" (not educated, but living in a treated unit).
• The intervention. Educated owners received a concise preview at the end of the baseline survey, an invitation to a workshop delivered by a private-sector tax lawyer, and a hotline number. The content was developed with Accra's Institute of Local Government Studies and aimed to balance rights and responsibilities and to be neutral about both compliance and tax minimization (pp. 12-13). It was purely informational and did not act on owners' behalf. Only about 23 per cent of educated owners attended a workshop (p. 17).

Findings

• Knowledge rose for educated owners. At endline the index of declarative knowledge rose by 0.16 standard deviations among TG-treated owners (standard error 0.07), driven by a 0.40 rise in the process index (Table 1, p. 40).
• Knowledge spread to neighbors. Owners in treated units who were not themselves educated scored 0.32 standard deviations higher on the practical "taxpayer action" index (standard error 0.12), and were 11 percentage points more likely to have talked to anyone about property tax, against a control mean of 4 per cent (Tables 1 and 2, pp. 40-41). The authors attribute the spillover to conversations with educated neighbors.
• Payment did not move. Neither group differed from control on receiving a bill, receiving an official visit, or making any payment: the coefficients on any payment are -0.02 (untreated) and +0.01 (treated), standard errors 0.04, against a control mean of 0.12 (Table 3, p. 41). The payment-amount estimates are likewise statistically indistinguishable from zero.
• How firm the null is. The payment and bill outcomes are self-reported by owners. The authors note that the control group's 12 per cent is "very close to the actual rates of payment observed in Ghana" in similar settings, and that response bias is a concern for levels. Official daily reports and tablet GPS records corroborate only the absence of effects on visits and registrations, not on payments (Tables 4-5, p. 42). The authors also caution that a light-touch, informational intervention may be too weak to move payment, and that only 24 officials were involved.

Mechanisms the Authors Document

• Beliefs about enforcement fell. Control-group owners estimated that 56 per cent of residential owners in the municipality pay the property tax, against about 12 per cent actually reporting payment. Untreated owners in treated units lowered that estimate by 6.0 percentage points (standard error 2.5) and educated owners by 4.4 (standard error 2.6) (Table 8, p. 44; p. 8). The authors read this as owners correcting an inflated view of the state's enforcement capacity.
• Awareness of lawful requests, without uptake. Owners became more likely to know that they can ask for a bill reduction, a revaluation, or payment in instalments (for educated owners, +0.27, +0.22 and +0.17 respectively). Requests actually made did not change (Table 7, p. 44). This is awareness of legal ways to lower or defer a liability, not evasion.
• More engaged, not more collaborative. Educated owners report more active dealings with officials but do not rate officials more favorably; the authors find "no movement towards more co-productive, collaborative relationships" (p. 4). Tax-morale measures were unchanged (Table 8, p. 44).
• The authors' framing. They conclude that "knowledge is a double-edged sword: it enables taxpayers to engage more effectively with the state but also allows them to recognize the constraints on the state's enforcement capacity," and that education may work best "as part of a broader reform agenda" that also strengthens enforcement capacity (p. 37).

Relation to the Georgist Case

The paper is about compliance with an existing property rate in a low-capacity municipality. It supplies no evidence on land value, site-value taxation, or the incidence of a land tax. Its bearing on Georgist argument is by analogy; the authors make no such claim.

A land-based tax would need the same administrative preconditions as any recurrent property tax: bills that reach owners, officials who collect, and credible follow-up. The experiment's result is that supplying owners with information about the tax did not substitute for those preconditions, and that it lowered owners' estimates of how likely others are to pay. That pairs with King's account of why Ghanaian property rating remains weak, where authority to tax sits with local assemblies while authority over land information is divided. It also fits the DRC experiments, which found that enforcement capacity caps the revenue-maximizing property-tax rate and that a progressive schedule raised revenue only with targeted enforcement. The authors cite the first of these as related work. Brockmeyer et al., using Mexican natural experiments, find that raising property-tax rates raises welfare while heavy-handed enforcement can reduce it, a reminder that the enforcement margin cuts both ways.

Caveats: it is a working paper that has not been peer reviewed; it covers one municipality, 24 officials and a small number of collection-unit clusters; and the null on payment rests on self-reported data.

See Also

• When Tax Authority and Land Authority Diverge: Property Tax Reform in Ghana (King, 2026)
• Bergeron, Tourek & Weigel: The State Capacity Ceiling on Tax Rates
• Tourek et al.: Progressive Property Tax in the DRC
• Brockmeyer et al., Taxing Property in Developing Countries
• Property Tax in Africa (Franzsen & McCluskey)

Sources

1. Anders Jensen, Nikhil Kumar & Konstantin Poensgen (2026), "Knowledge Without Taxation? Experimental Evidence from a Tax Education Program," NBER Working Paper 36002 (November 2026). nber.org/papers/w36002 — full text read in the authors' own openly hosted copy (cover date November 2026, abstract identical to the NBER page); page numbers cited are that copy's printed page numbers, and table numbers should be checked against the NBER-hosted PDF, which was not accessed at last review (2026-11-09) — used for the setting, sample and four-group randomized design (pp. 7-8, 12-17, 21), baseline ignorance (pp. 1, 9), the knowledge, spillover, public-finance and mechanism estimates (Tables 1-3, 7-8, pp. 40-44; Tables 4-5, p. 42), the self-report caveat (p. 27) and the authors' conclusions (p. 37). Unrefereed working paper; empirical claims B-claim, authors' interpretations attributed to them (D-claim). The authors are academic economists at Harvard (Tier 1 standing under EDITORIAL §4c).
2. Godswill Dormenyo King (2026), "When tax authority and land authority diverge: property tax reform and institutional persistence in Ghana," Journal of Institutional Economics 22, e59 — see the wiki's King page — used only for the statement that Ghana's property rating is building-focused, which is not made in the Jensen-Kumar-Poensgen paper (A-claim, via that page).
3. Augustin Bergeron, Gabriel Tourek & Jonathan Weigel (2024), "The state capacity ceiling on tax rates: Evidence from randomized tax abatements in the DRC," Econometrica 92(4) — see the wiki's Bergeron, Tourek & Weigel page — used for the comparison with the DRC enforcement-capacity result; cited by Jensen et al. as related work (B-claim).