The Effect of Local Road Maintenance Tax Cuts on House Values

• In Event: Local Transportation Policies

Abstract

Research Question & Motivation

Most empirical research on transportation infrastructure focuses on the effects of new infrastructure projects primarily in developing nations; little is known about the consequences of maintaining existing infrastructure in developed nations. In this paper, we analyze the maintenance of existing roads in the U.S. which avoids issues of endogeneity coming from the placement of new roads. We design a quasi-experiment to study local referendums introduced to renew road maintenance taxes, which are typically levied via property taxes, and ask: What happens to road quality and neighborhood house prices when voters cut local road maintenance taxes? We compare housing sale prices between similar areas that narrowly pass or fail road property tax levies and use satellite images to fine-tune an Artificial Intelligence (AI) model for classifying road quality.

Data
We assemble a novel panel of Ohio townships, villages, and cities by merging: (i) official local referendum results; (ii) audited financial reports of local governments; (iii) 7 million CoreLogic arms‑length home sales (1995‑2021); and (iv) over 50,000 pre‑ and post‑vote satellite images. We predict road quality using a fine‑tuned Vision‑Transformer model (GPT-4O by OpenAI) that classifies each image on a 0 (low), 1 (medium), 2 (high) scale.

Empirical Strategy
The paper implements a dynamic sharp regression‑discontinuity design. The running variable is the Percent of Votes Against the road tax levy; jurisdictions with 50% + ε votes fail to renew and lose the funding. Within the optimal Calonico‑Cattaneo‑Farrell bandwidth, covariates are balanced and the McCrary test rejects manipulation (p = 0.98). We estimate intent‑to‑treat effects for each year τ = −3…+10 relative to the election, clustering at the jurisdiction level and controlling for flexible vote‑share polynomials and demographic covariates. Robustness checks include alternative bandwidths, placebo cut‑offs, winsorization, and the exclusion of "contaminated" jurisdictions that later pass supplemental levies.

Findings

1. 
   

   Drop in Maintenance funds. Failed renewals reduce average road‑maintenance resources by $163,547, which is 11% of the overall budget for local roads.

   
   


2. 
   

   Road Quality Deterioration. The Vision‑Transformer model detects a 0.44‑point (15%) post-election decline in road quality only in jurisdictions that lost the levy.

   
   


3. 
   

   Capitalization into housing prices. Beginning four years post‑vote, median single‑family sale prices fall sharply. The cumulative ten‑year effect is −$15,350 (≈ 9%). Event‑study estimates display no pre‑trend and remain significant through year 9, tapering thereafter.

   
   


4. 
   

   Heterogeneity. Effects are concentrated in urban jurisdictions and in the upper quintile of the price distribution; cheap or rural houses show little response. 

   
   

Policy Implications
Short‑run tax relief for households is eclipsed by longer‑run private losses: savings for households from lower property taxes is outweighed by the relative reduction in house values in later years. Deferred maintenance thus resembles an unfunded liability, imposing stealth costs on homeowners and local tax bases. Policymakers considering levy roll‑backs—or states debating caps on local taxation—should consider these hidden capital losses against temporary savings.

Author

• Saani Rawat, University of Cincinnati
  Presenting Author