🧭 What This Paper Does

Granular geographic data open new possibilities for examining how neighborhoods form and how they shape political life, but the subjective nature of neighborhood boundaries creates measurement and modeling challenges. An open-source, map-based survey instrument is introduced to let residents draw their own neighborhoods, and a statistical model is proposed to analyze how respondent characteristics and local-area features determine these subjective boundaries.

🗺️ How Neighborhoods Were Collected

• An open-source survey tool asked respondents to draw neighborhood boundaries directly on a map.
• Two separate surveys were conducted:
• Voters in Miami, New York City, and Phoenix were asked to draw their subjective neighborhoods.
• New York City residents were asked to draw a “community of interest” for possible inclusion in their city council district.

🧮 How Subjective Neighborhoods Were Modeled

• A statistical model links the inclusion of small geographic units (census blocks) in a respondent’s drawn neighborhood to characteristics of the respondent and characteristics of the local area.
• Model performance was evaluated with out-of-sample prediction and compared to standard neighborhood measures.

📌 Key Findings

• Holding other factors constant, white respondents are more likely to include census blocks with higher shares of white residents in their drawn neighborhoods.
• Partisan alignment matters: Democrats and Republicans are both more likely to include co-partisan areas when delineating neighborhoods.
• The proposed model yields more accurate out-of-sample predictions of subjective neighborhood boundaries than common, standard neighborhood measures.

🔍 Why It Matters

• The approach offers a transparent, replicable way to capture subjective neighborhood perceptions at high geographic resolution.
• Results demonstrate that race and partisanship shape how residents perceive neighborhood boundaries, with implications for research on representation, local politics, and the use of granular geographic data in political science.