I often recommend people set their mental compass to search for the scientific consensus on an issue. This is great in theory but this consensus is often an intangible idea. Where do you find it when it exists? How do you recognize it? You may trust that respectable media outlets will choose to interview the right scientists to give voice to this consensus. On the issue of curfews, however, those scientific voices are all over the place.
In the context of the ongoing COVID-19 pandemic, a curfew can be defined as a stay-at-home order during specific time periods with the goal of reducing the transmission of the coronavirus. I spent a day reading news articles about the worth of this particular measure, and the scientists and physicians interviewed simply could not agree. Some argued there was no evidence curfews worked. Others put forward the notion that they must work because they make intuitive sense. One justification I heard for them is that they signal to the population that things have gotten very serious.
As it turns out, determining if curfews do indeed work is very, very difficult. Calculating their plausibility takes us down crossroads littered with assumptions, and the evidence we do have is imperfect and piecemeal.
The curfew thought experiment
Popular culture, especially crime-solving shows featuring very smart protagonists, has taught us that most riddles are a simple deduction away from getting solved; in reality, solving problems often relies on inferences, and that’s a problem in and of itself. Inferences are best guesses but they may not always be right. Let’s imagine a city in the grips of the coronavirus. Starting tonight, we will impose a curfew. Will it work?
If we prevent people from congregating in supermarkets and pharmacies in the evenings, surely we will put a stop to potential outbreaks! Shoppers may decide to make fewer trips and simply buy more and stock up on the trips they do make... but they might make the same number of trips but in a shorter window of time each day. A supermarket might conceivably see the same number of customers but packed into fewer hours. If one of them is infectious, it could lead to a worse outbreak.
As we continue to think about the issue, we may infer that restricting nighttime activities will be beneficial because these activities often involve alcohol, especially in younger populations, and alcohol impairs judgment. But if our city has already closed bars and restaurant dining rooms, will adding a curfew do more to curtail this behaviour? Will people meet inside of houses and apartments instead? Will the police be able to tell when a small group is gathering inside after hours? Are the outbreaks in our city really coming from people congregating after work or are they mainly coming from schools and workplaces? At what time should the curfew begin? Should it only affect people of a certain age?
Answering these questions in the abstract forces us to make assumptions, and depending on the ones we choose to make, we will reach different conclusions. So we need data. But even here, there is entanglement afoot. We can’t simply look at COVID-19 case numbers before and after the start of a curfew. Curfews are not solitary creatures; they tend to be part of a set of interventions that includes lockdowns, school closures, mask ordinances, and work-from-home orders. In fact, a model often used to illustrate how these interventions work is the Swiss cheese model. Each slice represents one intervention, and each slice has a couple of holes here and there, but when you layer them you build a uniform block impenetrable by the virus. So if cases drop after enacting a curfew, was it because of the curfew or because of the curfew layered in with the other measures? Did the curfew add anything?
I hope these questions make it clear that armchair-opining about the worth of a real-world intervention like a curfew in the midst of a public health crisis that is being solved in real time by an ever-changing combination of measures is not wise. Luckily, we have some data on the effectiveness of curfews. Unfortunately, the data is fragmentary.
Goin’ mobile
The data begins with models. This is when researchers build a system within a computer that is meant to behave as close to the real world as possible. By programming enough rules into this system, scientists can see what the impact of a curfew would be on the spread of a virus. The main problem with models is that they are always incomplete: we can never fully simulate our world with every single human being in it. Models are thus simplified versions of the world and assumptions must be programmed in them to see how the system behaves.
A that predicted that infections, severe cases and deaths from the coronavirus would be delayed by enacting a curfew, but as the authors admit, a curfew was never the only measure tested in the model. So did the curfew add anything? A, this one out of Luxembourg, predicted that an 11pm curfew would have a relatively small impact, but the model itself did not consider how people would truly behave: would they still go shopping earlier or would they altogether cancel their evening trip to the grocery store? Imperfect models are not getting us anywhere.
Some of the best data we have on the effect of curfews comes from a technology that did not exist during the infamous 1918 flu pandemic: cell phones. Research teams have combed through publicly available data sets featuring anonymized mobility data from users of mobile devices like cell phones to see if their movements changed after a curfew was put in place. This is often done by adding up the time spent outside of home and seeing how it changes over time.
Even though this is real-world data, there are plenty of caveats when reading these analyses. Not everyone has a cell phone and not everyone who has a cell phone manually turns on the ability their phone has to keep a history of where they have been, so these mobility studies only track a specific segment of the population that may (or may not) behave differently from the rest. Importantly, mobility studies can tell us if curfews impacted the distances people travelled, but not if those changes translated into fewer infections. We can assume that if people travel outside of their home less, they will be at a lower risk of catching a virus, but this is not always tested for. Also, mobility data in the studies I read was not broken down by hour, only by day, which makes it hard to confirm if people travelled more during the day to compensate for the evening curfew.
In Greece, a 9pm curfew was put in place in November of last year, but in February of this year, the region that includes Athens saw its curfew moved up to 6pm. This expanded curfew according to mobile phone tracking data. The authors hypothesized that these stores were more crowded during the day, possibly facilitating the spread of the virus.
Meanwhile, in France, a nighttime curfew has had, with mobility being at its lowest for the wealthy and for people with white-collar jobs. Another study of the French situation looked not at mobility data but at the 7-day average number of confirmed COVID-19 cases and how it changed over time as measures like curfews were enacted. The authors looked at their data and suggested that on the number of cases, but given that curfews, school holidays, lockdowns, bar and restaurant closings, and the limitation of social gatherings were all tangled together in a short period of time, it is challenging to isolate the impact of the curfew itself. For that, a better country might be Canada.
On January 9, 2021, the province of Quebec imposed a curfew, but its neighbour to the West, Ontario, did not. This allowed a team of Canadian researchers to watch this natural experiment and look at mobile data to see what would happen. They published their findings in a preprint that has not yet been peer reviewed. It turns out that compared to Ontario, there was. When the city of Montreal was pitted against Toronto, that decrease was 39%. These reductions however were not as high in lower socioeconomic neighbourhoods or in places with a large number of essential workers. The authors mention that this cell phone data came from about 1% of the population only and that daytime mobility between the two provinces was not entirely comparable, as Quebec schools reopened on January 11 but Ontario schools lagged behind by a month.
These same authors and a few more colleagues looked at cell phone mobility data for the past year in Ontario and reported that. The more mobility, the more cases.
Can we conclude anything from all of this?
The jury is out
It is a real puzzle to figure out the specific effect of a curfew on COVID-19 numbers, and I don’t think there is a consensus yet on a curfew’s utility.
A few large studies have looked at what they call “non-pharmacological interventions,” meaning ways to control the pandemic besides drugs and vaccines, and they have attempted to extract as much information from as many countries as possible to rank the effectiveness of various public health measures. One such study concluded that it is the combo punch of many interventions that is necessary, but they ranked in terms of effectiveness. Another study (preprint only, not yet peer-reviewed) disagreed: they calculated that. Study #3 argued that but confessed: “Individual strategies could not be evaluated independently.” Imagine melting a pack of Skittles and trying to pull the green ones from the mess. That’s the challenge.
Lastly, there are downsides to curfews. People experiencing homelessness, already highly vulnerable, may fear catching the virus by being forced to seek shelter during the curfew. And having to stay home at night has in many countries, which has large repercussions especially on women and children.
I would love to point to a scientific consensus in the debate over curfews. I would argue, however, that there is no consensus at the moment. Decisions are being made, I suspect, based on fragmentary data, precautionary principles, and intuition. Sometimes, that is unfortunately all that can be done.
Take-home message:
-It is very difficult to separate any impact a curfew might have on coronavirus infections from the impact of other public health measures, like lockdowns and bar and restaurant closings.
-A few studies indicate that curfews probably do help reduce the transmission of the virus, but these studies contain many caveats.
-Curfews can have harmful consequences on vulnerable populations, like people experiencing homelessness or domestic violence.
