Predicting Domestic Violence from Liquor Store Activity Using Machine Learning

A new research paper from the University at Buffalo investigates the relationship between alcohol outlets and domestic violence (DV) in various neighborhoods in a US city, based on anonymized GPS location data from a mobile phone network.

The article concludes that liquor store activity is more strongly associated with domestic violence than any other place where liquor can be purchased, indicating that the correlation between liquor outlets and domestic violence may be more nuanced than recent studies suggest:

“Remarkably, the liquor store visitation rates variable shows the highest positive contribution to the DV rate compared to the other variables. This result suggests that liquor store attendance is an important predictor of neighborhood-level DV.

“Interestingly, visits to two other types of alcohol outlets (i.e. pubs and breweries) have negative coefficients, which is consistent with the analysis of correlation.

“This result indicates that visits to these outlets are, in fact, associated with lower DV rates. This result is surprising, as a higher density of drinking places in a neighborhood was often thought to be related to higher DV [rate]’

From the article, a mapped visualization illustrating visits between liquor stores and home neighborhoods over a week in 2019, before the advent of data-disrupting COVID. Source: https://arxiv.org/ftp/arxiv/papers/2203/2203.04088.pdf

The authors theorize a possible near-inverse relationship between attendance at more prestigious alcohol-related venues and domestic violence, finding that visitors to venues such as breweries and bars are likely to have higher socio-economic status , which reduces their likelihood of being involved. in incidents of domestic violence.

The paper is titled The role of visits to alcohol outlets derived from mobile phone location data in improving the prediction of domestic violence at the neighborhood leveland comes from three researchers from Buffalo’s math, medicine, and geography departments.

Data and delimitation

The researchers’ objective was to determine whether anonymized mobile location data could be used, in conjunction with available crime statistics, to predict incidents and frequencies of domestic violence at the neighborhood level, insofar as these may be related to the availability of alcohol purchases in the locality.

The first task was to define neighborhoods, which was accomplished using census block groups (CBGs), the smallest geographic units available that still retain complete demographic and socio-economic data, as provided by the American Community Survey (ACS).

The anonymized mobile location data came from SafeGraph, a commercial company that provided researchers with data covering the entire United States for free. The dataset covers 3.6 million points of interest across states, including those that sell alcohol, such as liquor stores, bars, wineries and breweries.

The study covers a period leading up to, but not included, the statistically disruptive COVID outbreak, and the locations of residences of liquor outlet visitors were inferred by SafeGraph based on nighttime locations over the six weeks. previous ones. Neither the exact trajectories of visits nor the exact domicile of users are disclosed in the data – a sampling that corresponds to the “generalized” boundaries of the neighborhoods identified in the ACS data.

The trajectories studied in the work do not necessarily correlate with home > liquor-store > home visits, but may represent paths to another home setting, with a stop at a liquor store along the way.” width=”1000″ height=”271″/><noscript><img decoding=The trajectories studied in the work do not necessarily correlate with home > liquor-store > home visits, but may represent paths to another home setting, with a stop at a liquor store along the way.

Although the available GPS data encompasses the entire country as a potential area of ​​investigation, the researchers chose Chicago, due to the relatively well-labeled information available through the City Police Department’s Citizen Law Enforcement Analysis and Reporting (CLEAR) system. Chicago.

The CLEAR dataset includes crime types (such as arson or assault), along with the time, location, and latitude/longitude pairs of the incident, among other data points. It also contains a binary value for ” Domesticated “indicating whether visiting police attributed the incident to the Illinois Domestic Violence Act of 1986.

The researchers note that a number of crimes unrelated to “domestic violence” (including arson) have been erroneously assigned to this category and filtered out these incidences.

On the left, Chicago's city limits, along with its Census Block Groups (CBGs), on the right, extracted incidents of domestic violence.

On the left, Chicago’s city limits, along with its Census Block Groups (CBGs), on the right, extracted incidents of domestic violence.

Only occasions when a visit to a liquor store lasted four minutes or more were counted. Additionally, the researchers concede that there is no way of knowing whether the visit resulted in a purchase, nor the extent or type of purchase. Additionally, they note that domestic violence is considered one of the least reported crimes, which contributes to the approximate nature of the results obtained in the study.

Four approaches

The data was studied using four approaches: Random Forest (RF), Ordinary Least Squares (OLS), Deep Neural Network (DNN), and Geographically Weighted Regression (GWR). OLS and GWR are statistical rather than machine learning methods.

Overview of experiences, covering four methods and four categories of alcohol-related points of interest in SafeGuard data, with reference to domestic violence incidents in CLEAR data.

Overview of experiences, covering four methods and four categories of alcohol-related points of interest in SafeGuard data, with reference to domestic violence incidents in CLEAR data.

The results from all four methods indicate that the relationship between liquor stores and domestic violence varies significantly across neighborhoods, although the researchers ultimately conclude that “increased visits to liquor stores are associated with increased rates of VI in most Chicago neighborhoods”.

Correlation with DV incidents, based on the four types of alcohol outlets studied in the article: a) liquor stores, b) drinking places (i.e. bar), c) brewery and d) winery.

Correlation with DV incidents, based on the four types of alcohol outlets studied in the article: a) liquor stores, b) drinking places (i.e. bar), c) brewery and d) winery.

The document contains a more detailed overview of the socio-economic and demographic indicators linked to the analyses.

The authors state*:

“Our results provide additional insight into the relationship between alcohol consumption and DV. Of the four types of liquor outlet visits, liquor store visits have the strongest association with increased DV in the neighborhoods where liquor store visitors live.

“The routine activity theory of crime suggests three conditions necessary for a crime to occur, namely: a motivated offender, a potential target, and the absence of anything to inhibit the offender’s behavior. Liquor store visits suggest that purchased alcohol will be consumed at home when all three conditions are met: the offender and the potential target are in close proximity to each other and the alcohol provides the [disinhibitor].’

The authors further suggest that the results of studies such as these could contribute to future policy decisions regarding the regulation of liquor stores in areas where their relationship to domestic violence seems evident from the data, with possible interventions, including restricting permitted hours of sale and limiting the number of liquor licenses.

* My conversion of the authors’ inline citation to a hyperlink.

First published March 9, 2022.

Sherry J. Basler