Linking Taxes to Benefits: The Compliance Effects of a Landmark Project

The municipality is building the Túnel de Hornos, a large local infrastructure project financed through municipal revenues, and we partnered with the local government to inform property taxpayers about this tax–benefit link. We are mailing letters to the universe of roughly 100,000 property tax accounts that made the tunnel salient and explicitly connected it to funding from the property tax. The core variation is in framing and personalization: relative to a baseline letter with standard payment information, treatment letters highlight the tunnel as a municipal investment and emphasize the recipient’s geographic proximity to the project to make the individual benefit more concrete. We then measure how these informational frames affect compliance and payment behavior, and whether responses vary with distance to the tunnel.

2026-01-15

2026-01-31

The primary outcome is property tax compliance, measured using administrative records at both the extensive and intensive margins. Our main indicator is timely payment, defined as a binary variable equal to one if the account makes a payment on or close to the due date in a given month.

Tax compliance will also be captured through additional outcomes, including the probability of making any payment within the tax period, the share and amount of pre-existing debt settled, and enrollment in a debt repayment plan.

The analysis focuses on treatment-effect heterogeneity by distance to the tunnel. We estimate treatment effects pooling Treatments 1 and 2, and also separately by treatment arm. We compare treated accounts to pooled controls as well as to each control arm, which also allows us to assess potential spillovers affecting non-treated accounts located on treated sidewalks.

We will further analyze heterogeneity by baseline characteristics, such as prior compliance and assessed property values. In addition, we examine whether effects vary with sidewalk baseline compliance, defined as pre-intervention compliance rates aggregated at the sidewalk level.

We will analyze treatment effects separately by distance to the tunnel, comparing taxpayers who are closer versus farther away (using pre-defined distance bins defined as 0-1.5km, 1.5-3km,3-5km,+5k. This heterogeneity is central because the intervention is explicitly about making the benefits of the project salient, and proximity is a natural proxy for expected individual benefit. If the tax–benefit linkage matters, effects should be stronger among those with higher exposure to the project (closer properties). At the same time, the sign is not mechanically pinned down: some taxpayers may react positively to learning that their taxes finance a visible local investment, while others may react negatively if they perceive the project as irrelevant, poorly targeted, or not worth the cost, even if they live nearby. We will also report pooled effects across all distances.

We will explore heterogeneity by baseline compliance (past payment history). We expect larger effects among taxpayers who have a margin to adjust, such as occasional or late payers, and smaller effects among never-payers and always-on-time payers. Never-payers may be less responsive to informational content because nonpayment is likely driven by constraints or persistent disengagement, while consistently compliant taxpayers have limited room to increase compliance.

Not available

We randomly assigned properties to control, treatment 1 or treatment 2.

We implement a two-stage randomization to define who receives each treatment. First, we randomly assign sidewalk segments to treatment arms (Control, Treatment 1 or Treatment 2), ensuring that nearby households within the same segment receive the same assigned arm. We define sidewalks as street segments: both sides of a street block between two adjacent cross streets. Second, within each treated segment, we randomly assign individual property tax accounts to receive either the control letter or the segment’s assigned treatment letter. This design preserves balance across arms while controlling for within-segment contamination and allowing us to measure treatment effects accounting for the cluster-level assignment on the unit.

Yes

5,188 sidewalks.

101,540 properties

Control: 1,040
Treatment 1: 2,075
Treatment 2: 2,073

The experiment is divided into four subsamples based on sidewalk distance to the tunnel: 0–1.5 km, 1.5–3 km, 3–5 km, and more than 5 km. The outcome of interest is a binary indicator equal to one if the account makes a payment in a given month. Effects are measured in percentage points. Minimum detectable effects (MDEs) are computed for the full sample and separately for each distance range. Table Minimum Detectable Effects by Group T vs C (Pooled) T1 vs C0 T2 - T1 Full sample 0.0135 0.0250 0.0198 By distance to tunnel 0–1.5 km 0.0284 0.0477 0.0443 1.5–3 km 0.0232 0.0395 0.0314 3–5 km 0.0241 0.0422 0.0374 >5 km 0.0268 0.0455 0.0379 Notes: Entries report minimum detectable effects (MDEs). Columns “Pooled T vs Pooled C” and “T1 vs C0” correspond to regression-based summary intention-to-treat contrasts. The “T1 vs C0” column is equivalent to a T2 vs C0 comparison, where C0 refers to control accounts on on-treated sidewalks, controlling for potential spillovers. Column “T2 - T1” reports the difference between two active treatment arms and captures the incremental effect of distance personalization among treated units.