Murder is Up, Crime is Down
Does Domestic Violence Play a Role in the Paradox?
Editor’s Note: The idea for this commentary began when Antonnella Hooper of CreditCards.com wrote to us about her online guide for women in abusive relationships that “offers advice on the steps to take towards financial and physical freedom while outlining helpful resources for continued support.” She asked us to put the link to her page on the Critica website, which we do here and on our home page, and Critica COO Catherine DiDesidero suggested we do this in the context of a piece on domestic violence during the pandemic. That in turn led us to ask the question: what do scientists do when we need to answer a question during an ongoing event?”
There is now widespread agreement based on reasonably good data that the homicide rate has increased in U.S. cities since the beginning of the pandemic. At the same time, overall crime rates have decreased. It is unusual for these to go in opposite directions and the apparent anomaly has sparked many theories. Figuring out whether any of them are true is an exercise in understanding what to make of predictions and how to use data while events are still unfolding.
The question is, of course, whether there are sufficient data to back any of the factors proposed as causal for the observed increase in homicide rates. For example, we were especially interested in seeing whether an increase in domestic violence during the pandemic could account for the increase in homicides. But the question can be asked about any number of other putative factors, like increased gun sales, dissatisfaction with the police leading to less policing, the bad economy, and overwhelmed hospitals less able to save the lives of victims. A review in Vox last month by German Lopez is an excellent place to find the pros and cons of these and more theories weighed.
Gathering data during any ongoing event is always difficult. Even in a well-controlled randomized controlled trial (RCT), data “snooping” while the study is still being conducted is frowned upon. Besides having implications for the integrity of the study and later statistical analyses, it is always possible that what seems like a finding at one point in the study turns out to be a non-existing one later on, and vice versa.
Nevertheless, sometimes even with an RCT it is necessary to break in at some midpoint and see how things are going. In the case of an RCT, this is done whenever there is a chance that either giving or withholding a treatment might be harmful to the study participants. For many studies, ethics committees require what are known as data and safety monitoring boards (DSMB), a group of independent people who look at the data at several predetermined time points during the study and determine if any harm can be detected and if a clear finding is already apparent so that continuing the study would be unlikely to yield any further findings.
For example, suppose a new medication for a serious heart condition is being tested against an existing drug to see if the new medicine works better. Because patients are randomized to one or the other medication and neither patients nor the researchers know who is given which drug, it is possible that the new drug is causing some adverse effect that could only be detected by “breaking the blind” and having a look at the data. Similarly, it may be that the new drug is so clearly a bit worse than the existing drug that even continuing the study and adding more participants would not possibly lead to a statistically significant finding in favor of the new drug. In either case, the DSMB has the authority to order the study stopped—continuing it exposes the participants to unnecessary risk since the study is clearly “futile.”
In a similar manner, we sometimes cannot avoid looking at data during an ongoing health emergency, even though we know that a retrospective look when the event is finally over may yield very different conclusions. Remember how we were initially told that cigarette smoking protects against serious complications of COVID-19? Other studies say exactly the opposite. Still, as in the case with the RCT described above, we may want to know if there are already danger signals, however weak, that should be acted upon before the emergency runs its course.
If domestic violence is fueling an increase in homicides, for example, we need to know that as soon as possible in order to do whatever we can to increase prevention. That might take the form of adding personnel to social service agencies that deal with domestic violence, controlling gun sales during the pandemic (a move that would meet certain resistance), increasing the number of helplines available for women who are in potentially dangerous situations, and increasing the number of shelters for women who are threatened or already victims. That means we have to do some data snooping. In this case, however, things are certain to be much messier than would be the case with an RCT.
Certainly, one can easily imagine that the pandemic would increase domestic violence for a number of reasons. During times of extreme stress, domestic violence increases. Another potential reason is that gun sales are up, a phenomenon associated with increased domestic violence. People are trapped at home with each other, making it potentially harder for women to get away from violent male partners (the majority of domestic violence is committed by men against women). Social service agencies that ordinarily help victims of domestic violence may not be operating at full capacity during the pandemic. Early in the pandemic, the U.N. Population Fund issued an interim Technical Report in which it predicted that:
The COVID-19 pandemic is likely to undermine efforts to end gender-based violence through two pathways:
○ Reducing prevention and protection efforts, social services and care
○ Increasing the incidence of violence
Similarly, an NPR report on June 1 stated that “One consequence of COVID-19 is a projected global increase in domestic violence, including intimate partner violence.”
But predicting something because it seems logical that it will happen is not the same thing as rigorously observing it actually happening. If we do some data snooping about domestic violence rates since March 2020, we find that so far data are “scarce” regarding any increase in domestic violence during the pandemic and mostly come from observations reported in the media. An example of the latter is an April article in the Guardian reporting that “At least 16 suspected domestic abuse killings in the UK have been identified by campaigners since the Covid-19 lockdown restrictions were imposed, far higher than the average rate for the time of year….” A paper in an academic journal noted that “Upticks in domestic violence were reported in many regions soon after stay-at-home directives were announced.”
Strikingly, there are also findings of a decrease in reports of domestic violence. This may simply reflect increased difficulty abuse victims have in being able to call for help because they are isolated at home with their abusers. A very informal search for articles in medical and scientific journals in the search engine PubMed, which is operated by the National Library of Medicine, revealed more than a thousand articles in the last year on domestic violence in general, but we did not find any with more than anecdotal data about rates of domestic violence during the pandemic. Although there is some evidence of murder-suicides increasing specifically in which a male is perpetrator, there are indeed few data collected with any rigor.
One study that was published on August 13, after we performed our search, does suggest that the incidence of victims of domestic violence reporting to a single Boston hospital’s emergency department (ED) was higher during the pandemic period so far compared to similar periods in the last few months. Moreover, the severity of the physical injuries was worse since March 2020. Although an important paper, it does represent only one hospital’s experience and therefore does not by itself make the case.
The data aren’t any more solid for the many other reasons given for the increase in homicides. It is important to note that even with this increase, the rate of homicides in U.S. cities is still below what it was ten years ago. American cities are significantly safer now than they used to be. We can see, however, that getting a grip on even one of the potential reasons is complicated.
There are different ways to get data on domestic violence—reports from helplines, social service agencies, and emergency departments are among them, but each of these requires that someone makes the report. That means that there is inevitably going to be a lot of missing data. Even if scientists design a survey with good scientific properties, it is hard to know what the comparable “before” rate should be. The standard statistic usually given is that one in four women and one in nine men are victims of domestic violence. As mentioned above, however, domestic violence is believed to be significantly underreported, so unless the researchers happened to have used the same methods in the year before the pandemic, it may be difficult to know if the rate during the pandemic has really changed, especially if underreporting is even more widespread now.
Perhaps a better way would be to count the number of deaths from domestic violence using publicly available FBI crime statistics. This too may entail some technical difficulties. In 2019, a rise in deaths from domestic violence was reported, from 1875 in 2014 to 2237 in 2017. If the rate continued to increase in the year before the pandemic, we would then need to demonstrate a significant increase in the rate of rise during the pandemic to show that domestic violence is a contributor to the overall increase in homicides.
Of course, the increase in homicides observed during the pandemic is likely to be caused by multiple factors. That will make it only that much more difficulty to see a clear signal from any single factor, like domestic violence. And—hopefully--this increase in homicides itself could be a transient phenomenon. If so, depending on what time frame is used, researches may observe very different phenomena.
These considerations show some of the difficulties researchers have in looking at data in the middle of any event, be it a well-controlled study or an international disaster. In the case of understanding whether domestic violence (or any other factor) is responsible, we have myriad technical difficulties including underreporting, lack of a clear baseline, and multiple contributing factors. The result of all of this is that at present we cannot be sure whether or not domestic violence or any other of the proposed factors is a contributor to the surge in homicide rates since the pandemic began.
It is thus always problematic to listen to predictions of what might happen when biological systems, like human beings, are involved. Yes, it seems fully logical that the pandemic would increase rates of domestic violence and therefore of deaths due to domestic violence, but in science what is seemingly logical often turns to be either only part of a picture or even completely wrong.
Public Policy Implications
What can interim data therefore add to the conversation about homicide rates and domestic violence? We have one solid piece of data—that homicide rates went up following the beginning of the pandemic. We have two hypotheses: that domestic violence rates have also gone up and that these are at least partially responsible for the increase in homicide rates. We have an incomplete dataset to address these hypotheses, although we must be clear that this also means we cannot rule either hypothesis out. Moreover, we do have anecdotal data that domestic violence has increased.
Domestic violence is obviously a terrible thing, so would there be any harm in accepting the theory and anecdotal data that it is getting worse and taking action? We can think of two possible issues. First, increasing staff at hotlines and social services agencies and creating more shelters would obviously cost money at a time when governments and nonprofits are strained for funds. Second, focusing on one potential factor could mean that less attention is spent on addressing others, such as reducing gun sales and improving policing.
Despite these potential concerns, we think policymakers should assume for now that domestic violence is a worsening problem and, while balancing other priorities, reinforce education, outreach, and service efforts to help its victims. While we should of course call for more research and better data, we find the threat of increased domestic violence and intimate partner murder to be sufficiently compelling from this bit of data snooping to warrant appropriate action.
We may be wrong, of course. But resources for women suffering from abuse have always been under funded and for some completely unavailable. Hence an increase in funding now would never be wasteful given the enormity of the problem.
When anyone predicts that something will surely occur, as they have about domestic violence, we always need to ask if that prediction is based on more than just logic, because the data sometimes defy logic and common sense. As long as we know we have entered an area of uncertainty, it still may be necessary for us to use the scraps of information we have in order to save lives.