Research paper: "Relationship between visits to liquor stores and domestic violence in Detroit after the COVID-19 lockdown"

Hi all, with my co-authors Aaron Chalfin and Monica Deza, we have a new working paper on the relationship between visits to liquor stores and domestic violence in Detroit after the COVID-19 lockdown. The short summary is that while the relationship between alcohol consumption and domestic violence is known, we find that the lockdown amplified that relationship almost 3 fold in the city.
Appreciate any comments and feedback, and we look forward to presenting it at the SafeGraph Seminar.

very interesting

@Shooshan_Danagoulian_Wayne_State_University I just finished reading through your paper and this is incredibly interesting. I would be curious to see if this is the same for other metropolitan areas as well.

Looking forward to having you on for the seminar to get a more in depth understanding.

Thank you, Spencer! It was made possible by the amazing detailed SafeGraph data. We thought about doing it in other cities, but domestic violence data is particularly hard to come by, and if anyone has access to city 911 calls or domestic violence numbers, we can repeat the study in other cities.

Would you like to repeat the studies for other cities? The Placekey community has some partnerships with various city, county and state governments. Happy to put some feelers out if you like.

Also the communities own @Kivan_Polimis_Office_of_Harris_County_Judge_Lina_Hildalgo
Might be able to offer some pointers for locating that data for Harris County if it’s available.

Hi @Shooshan_Danagoulian_Wayne_State_University, I recently came across a Denver crime dataset that you might be interested in. It has a column for “Offense Type ID,” which has the value “aggravated-assault-dv”. I’m not sure if this means “exclusively domestic violence”, or if it’s “aggravated assault and domestic violence” or what, but it may be valuable to you either way. Great work on this research, thank you for sharing :slightly_smiling_face:

Thank you, Ryan! I will look into it.

@Shooshan_Danagoulian_Wayne_State_University I will ask around this week for the availability of 911 calls or DV data

Thank you @Kivan_Polimis_Office_of_Harris_County_Judge_Lina_Hildalgo!

@Shooshan_Danagoulian_Wayne_State_University Interesting study. It makes sense that you used Poisson regression for the modeling framework. Did you try other methods, like panel regression and different time lags? Were the results consistent with cross-validation? Anyways, interesting study.

We tried Poisson with different combinations of fixed effects, as well as linear regression with fixed effects. These fixed effects take advantage of the panel nature of the data – so the poisson results are, in essence, panel regression results. While in some cases the noise increased in our estimates, reducing the statistical significance of our findings, the magnitude and sign of the estimates remained consistent. We also incorporated 2 lags of the alcohol visits to see if going to the liquor store has effects on domestic violence in up to 3 days after the visit. As expected, we found even larger effect – that is, going to the liquor store is strongly associated with domestic violence up to three days after the visit. Hope you can come to our seminar on Tuesday.