Increase in Percentage of Domestic Violence-Related Emergency Calls on Weekdays
This blog post is part of a series that explores data on domestic violence during COVID-19. We hope to contribute to the international conversation surrounding this issue and shed light on this subtler but equally pernicious “pandemic.”
Main points:
- Domestic violence-related emergency calls occur more frequently on weekends both before and after shelter-in-place orders were put in place. However, after the lockdown, this trend is weaker: a more even distribution of calls is observed across all days of the week.
- Most domestic violence calls are received between 4pm and 12am. Shelter-in-place orders did not affect the time of day at which police calls for service were made.
As shelter-in-place orders related to the COVID-19 pandemic have increased the amount of time families spend together in close quarters, emergency calls related to domestic violence have gone up. However, there has been little analysis as to whether temporal trends in reports made during the lockdown differ from those observed in the pre-pandemic period in terms of the time and day on which they occur.
If trends are consistent throughout the pandemic period, we would expect domestic violence emergency calls to follow the pattern observed in pre-pandemic analyses, with calls spiking on the weekends [1] and during the evening and night [2]. One explanation for this trend is increased alcohol consumption at night and on the weekends once the workday has ended [3]. However, researchers have also attributed these trends to the increased likelihood of both partners being at home in the evenings and over the weekend. Since the lockdowns have caused many families to stay home almost every day, day and night, one might expect these trends to weaken during the pandemic period.
In order to test these hypotheses, we collected police calls for service data for 22 US cities [4] and compared the number of calls per day of the week for 28 days before and after shelter-in-place orders were enacted in each area (Graph A). We also grouped the calls by time of day, dividing them into six four-hour periods (12am — 4am, 4am — 8am, etc.) and analyzed the number of calls in each period before and after each city’s lockdown (Graph B).
As shown on Graph A, we plotted the average proportion of calls per day of the week across all cities, along with the standard error. To take into account the variations in the absolute number of calls across the cities, the call values were first standardized for each city. Our results show that, although there were still more calls made over the weekend after the lockdown, the standard deviation decreased [5], indicating that the distribution of calls throughout the week is more even. Specifically, the proportion of calls received between Tuesdays and Fridays increased after lockdown, but decreased for Saturdays, Sundays and Mondays.
This result confirms our hypothesis that the weekend spike trend would weaken after the lockdown, potentially because partners spend more time together at home every day of the week. A decrease in calls observed on Mondays would further support our hypothesis about weekend spikes in domestic violence. We imagine a hypothetical situation in the pre-lockdown period, where harm is perpetrated toward a survivor over the weekend, when the perpetrator is home. In many such cases, the first time that the survivor can reasonably call the police is on Monday, when the abuser is gone for the workweek. During Covid-19, survivors will no longer have this break, which may indicate why the percentage of calls is significantly down on Mondays in the post-lockdown period.
Although the increase in weekend reports is weaker during the pandemic period, it is still apparent, possibly because abusers are distracted by virtual work during the week, essential workers are still home only on weekends, or because alcohol is still more likely to be consumed on the weekends once virtual work obligations have ended.
Virtual work is more likely to occur during the day, which may explain our finding that the time of day of reports is unaffected by stay-at-home measures. The error bars on Graph B, along with the fact that the standard deviations for the number of calls in each time period before and after lockdown only decreased by 4.5%, suggest that there is no real difference in the distribution. We see a similarly high percentage of calls occurring between 4pm and 12am before and after lockdown orders: 47.8% and 46.2% of calls respectively.
Our results suggest that law enforcement and other organizations that support survivors of domestic violence should be prepared to receive more calls for support during the week than they may previously have accounted for. These results also imply that calls may spike once quarantine ends, mimicking the pre-lockdown spike in calls on Mondays. We suggested above that this increase may reflect the fact that more domestic violence occurs over the weekends, and that survivors may report it on Mondays once their abusers have returned to work. The quarantine period may function as one long weekend by confining survivors in their homes, and thus we might expect that lifting restrictions will act like a Monday, giving survivors the alone time necessary to finally report abuse. Lastly, it is important to note that many domestic violence incidents go unreported altogether and therefore do not figure in our analysis [6].
Footnotes/ References:
[1] Johnson, I.M., 2007. Victims’ perceptions of police response to domestic violence incidents. J. Crim. Justice 35, 498–510. https://doi.org/10.1016/j.jcrimjus.2007.07.003
[2] Cohn, E.G., 1996. The effect of weather and temporal variations on calls for police service. Am. J. Police 15, 23–43. https://doi.org/10.1108/07358549610116545; Joshi, M., Sorenson, S.B., 2010. Intimate Partner Violence at the Scene: Incident Characteristics and Implications for Public Health Surveillance. Eval. Rev. 34, 116–136. https://doi.org/10.1177/0193841X09360323
[3] Joshi, M., Sorenson, S.B., 2010. Intimate Partner Violence at the Scene: Incident Characteristics and Implications for Public Health Surveillance. Eval. Rev. 34, 116–136. https://doi.org/10.1177/0193841X09360323
[4] Albany NY, Ann Arbor MI, Atlanta GA, Baltimore MD, Boston MA, Chandler AZ, Cincinnati OH, Detroit MI, Fulton County GA, Griffith IN, Los Angeles CA, Mesa AZ, New Orleans LA, Phoenix AZ, Sacramento CA, San Francisco CA, San Jose CA, Santa Monica CA, Seattle WA, Scottsdale AZ, St Paul MN, St Petersburg FL
[5] The standard deviation of a data set calculates how different the points within the set are from each other. A low standard deviation means that the points are close together, while a high standard deviation means that the points are more spread out.
[6] Reaves, B., 2017. Police Response to Domestic Violence, 2006–2015 (No. NCJ 250231).; Shepherd, J., 1990. VIOLENT CRIME IN BRISTOL. Br. J. Criminol. 30, 289–305. https://doi.org/10.1093/oxfordjournals.bjc.a048022; Tjaden, P., Thoennes, N., 2000. Prevalence and Consequences of Male-to-female and Female-to-male Intimate Partner Violence as Measured by the National Violence Against Women Survey. Violence Women 6, 142–161. https://doi.org/10.1177/10778010022181769
Reporting Contributions By: Naomi Mowat-Amiet, Sylvie Stoloff, Anant Pai
