Selected Publications (full list in CV)

Published in Social Problems

With Alyssa Goldman and Christopher Browning

Data and Method:

• Longitudinal individual GPS location tracking and Ecological Momentary Assessment (EMA) in Chicago Metro Area.
• Multilevel models.

Highlights:

• Explicitly test the effect of local crime level on older adults' mobility behavior, particularly in time allocation inside vs. outside the home space.
• Findings suggest that older adults living in higher crime areas spend less time at home, on average. The effect is mediated by higher perceived unsafety at home.
• Neighborhood characteristics can permeate the boundaries of the home, adversely affecting an already vulnerable population in ways that may exacerbate inequality in community engagement, collective efficacy, and health.

Published in Journal of Quantitative Criminology (2024)

With Yanji Zhang, Guangwen Song, and Yongyi You

Data and Method:

• Grid level street view images, court judgments, and location-based service population in central Beijing, China.
• Deep learning and difference-in-difference (DID).

Highlights:

• Examine the causal impact of small businesses on street theft and the underlying mechanisms, testing and advancing our understanding of classic criminology theories (routine activities, crime pattern, eyes on the street)
• The treatment units (a mass loss of small businesses) showed a significant reduction in street theft compared to the control units. Ambient population and social activity played a mediating role in promoting and deterring crime, respectively, with the former dominating.
• While small businesses exercise a certain amount of natural surveillance power, as a whole, they function primarily as crime generators.

Published in Crime and Delinquency (2023)

With Yanji Zhang, Guangwen Song, and Chunwu Zhu

Data and Method:

• Grid level court judgments, socio-economic and ambient population LBS data in the urban core of Beijing, China.
• Machine learning and SHAP interpreter.

Highlights:

• Compare the performance of multiple machine learning algorithms in predicting the spatial pattern of theft.
• Reveal nonlinear and spatially heterogeneous associations between environmental features and theft.
• Summarize six relation types of such associations at the global scale and cluster grids according to feature contribution at the local scale.

Published in Cities (2023)

With Guangwen Song, Lin Liu, Luzi Xiao, Yuhan Wu, and Han Yue

Data and Method:

• Grid level mobile phone ambient population data and crime statistics in a major Chinese city.
• Negative binomial regressions.

Highlights:

• Focus on the internal heterogeneity of ambient population by income and its impact on theft.
• Results highlight the heterogeneous segments of the ambient population, coupled with the dynamic human mobility patterns, structure the convergence of the crime triangle in nuanced yet crucial ways.

Published in Annals of American Association of Geographers (2022)

With Yanji Zhang, Guangwen Song, Lin Liu, and Chunwu Zhu

Data and Method:

• Grid level court judgments, socio-economic and ambient population LBS data in the urban core of Beijing, China.
• Dissimilarity index (D) at micro units (grids) and negative binomial regressions.

Highlights:

• Move focus to the impact of education segregation on crime at micro scale in activity spaces (vs. dominant residential racial segregation at the global scale).
• Both residential and activity space segregation significantly increase the risk of theft and violence, with activity space–based segregation more consequential.
• Results highlight the elevated influence of segregation on safety beyond the residential space, especially for areas clustered with the less educated ambient population.

Published in Journal of Quantitative Criminology (2021)

With Guangwen Song, Yanji Zhang, Wim Bernasco, and Lin Liu

Data and Method:

• Grid level court judgments, socio-economic and ambient population LBS data in the urban core of Beijing, China.
• Negative binomial regressions.

Highlights:

• Focus on the internal heterogeneity of ambient population by distinguishing residents, employees, and visitors and their differential impact on theft.
• The effects of ambient population on thefts vary by its composition in terms of social roles.
• Larger ambient populations imply larger theft frequencies. The effect of visitors is stronger than the effects of residents and employees. The effects of residents and employees vary over the course of the week.

Published in Sage Research Methods: Doing Research Online (2022)

With Kathleen Cagney, Christopher Browning, Erin York Cornwell

Data and Method:

• Methodology introduction and literature review.

Highlights:

• A graduate-level guide on location data.
• Topics covered: basic elements of GPS data, how to prepare for GPS data analysis, different sources of GPS data, common data preprocessing tasks and challenges, GPS data augmentation, and major strands of social science research using such data.

Non-spatial and Spatial Inequalities in Exposure to Physical Disorder across Chinese Cities Using Urban Visual Intelligence

Manuscript available upon request

Under Review

With Yanji Zhang, Yongyi You, Guangwen Song, and Luc Anselin

Data and Method:

• 33 million Street View images across 360 Chinese cities from 2013 to 2022.
• Observed and perceived physical disorder using computer vision methods (object detection and image regression).
• Fixed effects models.

Highlights:

• One of the first large-scale projects in China to examine different forms of inequalities and their evolutionary patterns across cities
• Non-spatial and spatial inequalities follow distinct geographical and evolutionary patterns.

Under Review

With Yanji Zhang and Guangwen Song

Data and Method:

• Observed & perceived physical disorder from 107,858 street view images (computer vision).
• 1972 burglary arrests in a major Chinese city.
• Discrete choice models.

Highlights:

• Examine how neighborhood social contexts (concentrated disadvantage) structure burglars' decision making.
• Offenders' spatial knowledge about the target context moderates decision preferences.