- Housing Prices and Crime Perceptions: How Do Partisanship and Media Exposure Affect the Housing Market?
• Short abstract: This paper examines how crime perception and media depictions of crime affect property values and how these effects vary across neighborhoods with different partisan compositions. It develops a Bayesian framework of belief updating and combines ZIP-code-level housing prices, crime perception surveys, GIS methods, election data, and machine-learning analysis of television news transcripts. To address measurement error and endogeneity, the paper employs a Battese–Harter–Fuller small-area estimator, exploits spatial discontinuities at media market boundaries, and estimates hedonic price models with Murphy–Topel standard-error corrections. The findings show that higher crime perception lowers property values, while the effects vary systematically with neighborhood partisanship and exposure to crime-related content on FOX News.
- Heterogeneity in Probability Perception: How Does Partisan Media Shape Our View on Crime?
• Short abstract: This paper examines how partisan media coverage of crime affects perceptions of neighborhood crime and how these effects vary by political affiliation. It develops a Bayesian framework of belief updating and combines survey data, GIS methods, and machine-learning analysis of television news transcripts. To address endogeneity, the paper exploits variation in local television offerings across Designated Market Areas (DMAs) within a spatial discontinuity framework and estimates the model using a two-sample two-stage least squares (TS2SLS) approach with analytical standard-error corrections. The findings reveal substantial heterogeneity in the effects of partisan crime reporting across respondents and neighborhoods with different political affiliations and under different presidential administrations.
- Dynamic Panel Estimation for Unbalanced and Nonconsecutive Panels: Evidence from the Effect of Crime Perception on Property Prices
• Short abstract: This paper develops a novel estimator for dynamic panel data models with highly sparse, unbalanced panels containing nonconsecutive observations. The estimator combines ideas from the long-difference literature with a causal matching strategy to eliminate individual fixed effects and constructs observation-specific instrument sets based on the nearest available lag. It is designed for settings in which conventional dynamic panel estimators are infeasible due to irregular observation patterns. The estimator is applied to study the dynamic relationship between crime perception and neighborhood property-price appreciation, revealing stronger effects during the Biden administration than during the Trump administration.