Do Place-Based Policies Reduce Retail Vacancy? Quasi-Experimental Evidence from New York City

• In Event: Student Research Panel: Housing Markets, Climate Adaptation, and Place-Based Investments

Abstract

Storefront vacancy has become one of the most visible symptoms of urban retail distress, yet rigorous causal evidence on which place-based policies actually reduce it remains scarce. Evaluating policy effectiveness is methodologically difficult: programs targeting distressed retail corridors are selected for high prior vacancy, creating severe confounding. This paper addresses that challenge using NYC's Storefront Registry (2019–2025), a building-level panel tracking vacancy at approximately 49,500 buildings containing over 65,000 ground-floor retail storefronts.

Which place-based interventions reduce retail vacancy, and how sensitive are the conclusions to research design? We evaluate five interventions: Open Streets pedestrianization, commercial rezonings, Business Improvement District (BID) formation, protected bike lane installation, and historic district designation.

We apply two identification strategies. Spatial regression discontinuity (RD) designs exploit sharp geographic boundaries (BID borders, Open Streets limits, historic district lines, rezoning edges, and Opportunity Zone boundaries) to compare storefronts just inside and just outside each boundary. Difference-in-differences (DiD) designs with propensity score matching (PSM) exploit the temporal rollout of each intervention as quasi-experiments, constructing matched control groups from untreated buildings with similar pre-treatment vacancy, built environment characteristics, and neighborhood demographics.

All five spatial RD boundaries produce statistically null estimates: vacancy operates as a continuous spatial process at approximately 800 ft that renders sharp boundary comparisons uninformative. DiD with PSM tells a more differentiated story. Open Streets pedestrianization reduces vacancy by 2.8 percentage points (p = 0.024); decomposing by persistence reveals that streets maintaining permanent pedestrianization drive the entire effect (-7.0 pp, p = 0.035), while streets that reverted to car traffic show no lasting benefit. Commercial rezonings reduce vacancy by 4.0 percentage points (p = 0.028), driven by SoHo/NoHo's use-restriction relaxation. BID formation is null in pooled estimates but shows an emerging negative trend by the sixth panel year (-3.5 pp, p = 0.067). Protected bike lanes show no effect after matching; a staggered Callaway-Sant'Anna estimator confirms the null (ATT = +0.010, p = 0.583). The unmatched bike lane estimate was positive and significant (p = 0.005), demonstrating that corridors receiving bike lanes were already on worsening vacancy trajectories before installation. PSM changed the substantive conclusion in three of four DiD designs.

Deploying both spatial and temporal causal designs on a single outcome reveals that null RD estimates can coexist with significant DiD estimates when policy effects flow through economic activation rather than boundary proximity. Pedestrianization and use-restriction relaxation reduce vacancy, while amenity designation and bike infrastructure do not. The persistence decomposition demonstrates that sustained commitment to pedestrianization, not temporary experimentation, generates lasting retail benefits.

Author

• Sean Andrew Chen, Rutgers University - New Brunswick
  Presenting Author