Read on Twitter
Today I've posted a new pre-print to @socarxiv analysing initial data on how crime in 16 large US cities has changed during the COVID-19 pandemic.
Pre-print: https://osf.io/preprints/socarxiv/ep87s
Thread with some results …
Pre-print: https://osf.io/preprints/socarxiv/ep87s
Thread with some results …
It's common to analyse crime changes by simply looking at year-on-year or month-to-month changes, but this can be seriously misleading because it ignores existing crime trends, seasonal patterns and (most importantly) that crime counts are typically very noisy.
So to understand whether any changes are out of the ordinary, I used seasonal ARIMA models to forecast how much crime we'd expect without coronavirus, then compared those forecasts to actual crime counts.
HT @mitchoharawild for the excellent {fable} R package!
HT @mitchoharawild for the excellent {fable} R package!
Some results. First, serious assaults in public places. We'd expect these to decrease as people stay at home, but actually they've stayed about the same.
Why? Serious violence is highly concentrated among a few people; maybe those most likely to be involved ignore lockdowns?
Why? Serious violence is highly concentrated among a few people; maybe those most likely to be involved ignore lockdowns?
Serious assaults in residences (imperfect proxy for domestic violence). Many worried this would go up during lockdown, but in all-but-one city it has varied only as much as we'd expect normally.
Why? Maybe reporting issues, although that's less likely with serious assaults.
Why? Maybe reporting issues, although that's less likely with serious assaults.
Residential burglary. Down substantially, but only in some cities. So in the first four weeks of Chicago's lockdown, about 560 fewer homes were burgled than expected. But in other cities (e.g. San Francisco) there's been no change. Why is one of many unanswered questions.
Non-residential burglary. Might expect this to go up because many businesses premises are closed and unattended, but so far there's no difference (Minneapolis excepted).
Why? Maybe places most likely to be burgled (e.g. drug stores) are those that are still open. More research!
Why? Maybe places most likely to be burgled (e.g. drug stores) are those that are still open. More research!
Theft of motor vehicles. All over the place: significantly up in Philly and LA, significantly down in Chicago and Tucson.
This may depend a lot on where cars get stolen from: lockdown may mean more cars available to steal in residential neighbourhoods but fewer downtown.
This may depend a lot on where cars get stolen from: lockdown may mean more cars available to steal in residential neighbourhoods but fewer downtown.
Finally, theft from vehicles. Down in most cities, although again with variation.
One of the main findings in this research is that no patterns were consistent across all cities. This isn't surprising: context matters. It's a reminder we need to study more and different places.
One of the main findings in this research is that no patterns were consistent across all cities. This isn't surprising: context matters. It's a reminder we need to study more and different places.
It's obviously very early to draw conclusions about crime changes, but it's also useful to help focus future research question and data collection.
The preprint hasn't yet been peer reviewed yet, so should be read critically (but then everything should!) https://osf.io/preprints/socarxiv/ep87s
The preprint hasn't yet been peer reviewed yet, so should be read critically (but then everything should!) https://osf.io/preprints/socarxiv/ep87s
