Indoors and outdoors: weather and the spatial displacement of crime
Matt Ashby
Is any relationship between weather and crime stronger for offences outdoors?
Data
Daily counts of police-recorded crime in five cities from 2009–18
Daily records of maximum temperature and hours of drizzle, rain and snow
Model
Bayesian Markov chain Monte Carlo generalised linear mixed model using the R package MCMCglmm
Model
- level 1: lagged crime count
- level 2:
- maximum temperature,
- hours of drizzle, rain and snow,
- month,
- day of the week
Odds ratios are relative to …
- assaults on a Sunday
- in January
- with no drizzle, rain or snow
- and mean maximum temperature for that city
| city | offense type | indoors | outdoors |
|---|---|---|---|
| Austin | assault | 1.024 (1.017–1.032) | 1.071 (1.059–1.083) |
| non-vehicle theft | 1.005 (0.999–1.011) | 1.004 (0.997–1.012) | |
| property damage | 1.009 (0.998–1.021) | 1.027 (1.014–1.040) | |
| Chicago | assault | 1.025 (1.021–1.029) | 1.089 (1.084–1.093) |
| non-vehicle theft | 1.016 (1.012–1.020) | 1.033 (1.029–1.037) | |
| property damage | 1.037 (1.032–1.043) | 1.042 (1.037–1.048) | |
| Fort Worth | assault | 1.026 (1.017–1.036) | 1.051 (1.033–1.069) |
| non-vehicle theft | 1.007 (1.000–1.014) | 1.001 (0.992–1.012) | |
| property damage | 1.009 (0.998–1.021) | 1.041 (1.024–1.058) | |
| Louisville | assault | 1.044 (1.034–1.053) | 1.072 (1.054–1.089) |
| non-vehicle theft | 1.021 (1.013–1.029) | 1.046 (1.035–1.057) | |
| property damage | 1.036 (1.022–1.049) | 1.030 (1.014–1.047) | |
| New York | assault | 1.016 (1.011–1.020) | 1.088 (1.082–1.094) |
| non-vehicle theft | 1.005 (1.001–1.009) | 1.037 (1.032–1.042) | |
| property damage | 1.014 (1.008–1.020) | 1.036 (1.030–1.042) | |
| … |
Relationships between assaults and weather are larger for crimes outdoors
for other crimes relationships are less clear
These slideslesscrime.info/talk/esc-2019
Questions or commentsmatthew.ashby@ucl.ac.uk
Free alerts for new articles in criminology journals
lesscrime.info/crimpapers