Welcome to our weekly newsletter highlighting the best of The Economist’s data journalism. In Off the Charts we go behind the numbers to show you how our data team gathers, analyses and visualises data for our Graphic detail section and beyond.
In Graphic detail this week, we look at how the pandemic has affected birth rates. Unsurprisingly, we find that countries that have had high covid-19 infection rates have seen their birth rates fall the most. Tougher lockdown measures and declines in economic output correlate with falling birth rates, too. But there was one unlikely measure that tracked changes in fertility most closely: visits to the park. We explore what’s going on.
In America, Liz Cheney, a conservative congresswoman from Wyoming, has been ousted by the Republican Party over her stance on the legitimacy of last year’s election. According to our poll with YouGov, most Republican voters won’t miss her.
A poll from Gallup shows that 43% of Americans do not support the death penalty—the highest share since the 1960s. Nevertheless, South Carolina’s Senate has confirmed a vote that will bring back the firing squad. The number of executions in America has been falling in recent years—17 executions were carried out in 2020, a 29-year low.
What could divide Americans more along party lines than the use of firearms? Perhaps the pandemic has produced an answer to that question: face masks. New guidance issued by America’s CDC recommends that everyone who has been fully vaccinated for two weeks should not need to wear masks in public, except on public transport and planes. The unexpected shift in advice has caused a stir in a country where masks are as much a political statement as an epidemiological necessity.
Joe Biden, America’s president, may declare America’s “independence from this virus” on July 4th, but the pandemic still rages in many parts of the world. In our data-driven cover briefing, we estimate that between 7m and 13m people have died since the pandemic began. Below Sondre Solstad, one of our data journalists, explains how we built a statistical model to arrive at a realistic toll of the pandemic.
If you have any comments about this newsletter, want to point us towards interesting data, or just say hello, you can reach us on email@example.com.
In this week’s cover briefing, we give our estimate of the true death toll of covid-19. Our statistical model suggests that between 7m and 13m people have died across the globe as a result of the pandemic. This number is two to four times more than the current official death toll of 3.3m deaths. Why are these numbers so different?
For many, the two key metrics of the covid-19 pandemic are official cases and deaths. These numbers are readily available for many countries and territories on a daily basis. But actual infections in any given country will always outnumber official cases, even in countries that have the capacity to do lots of testing. Some people will be asymptomatic and others may choose not to get tested.
These concerns might be surmountable in a world where testing capacity was comparable over time and between countries. But that isn’t the case: Britain, with a population of 67m, now tests over 6m people a week; Nigeria, home to 206m people, has tested fewer than 2m people altogether.
Deaths might appear easier to capture. In theory, some countries may be able to test everyone who dies to see whether they had covid-19. But dying with covid-19 is not the same as dying because of it. Meanwhile, the pandemic’s true toll should also capture people who die of ailments other than covid-19, perhaps because hospitals are at capacity, who wouldn’t ordinarily have done so in pre-pandemic times.
There are other ways to count the dying, some of them creative. For our excess-death tracker, we have relied on burials in Jakarta, for example. And sometimes we get tip-offs. A coffin-handle importer based in Haiti told us that while there were few official deaths registered in his country, his imports were twice as high as he’d expect between June and November. Unfortunately, we were unable to verify his claims with other data and so we could not gather more accurate excess-mortality data for Haiti.
Instead, to get a better idea of the true death toll in countries like Haiti that don’t publish excess-mortality statistics, we had to build a statistical model. My data-team colleague James Tozer pioneered excess-death reporting last year and together we collected partial data on deaths for 79 countries and territories, using sources such as the World Mortality Dataset and the Human Mortality Database, two academic projects.
It took several months of collating and coding to create our model of excess deaths to capture the true toll of the pandemic. The model uses 121 variables that might conceivably be related to excess deaths, such as covid-19 test positivity rates, the share of population over 65, government covid-19 policies on mask-wearing, and people’s mobility. It also incorporates official counts of covid-19 deaths, however imperfect they may be.
We used a machine-learning algorithm to find relationships in these data and the excess-deaths we had been able to collect. Finally, we used these patterns—our machine-learning model—to estimate excess deaths in the countries where they were unknown. You can read a deep-dive into our model’s methodology and the underlying data and code are available on our GitHub repo.
The truth, as our model suggests, is that most of those who have died either directly or indirectly from covid-19 since the pandemic began lived in poorer countries. About two-thirds of them never entered official covid-19 death tolls. Only by departing from those counts—however convenient they might be—can we understand the true toll of the pandemic.
This email has been sent to: firstname.lastname@example.org. If you'd like to update your details please click here. Replies to this email will not reach us. If you don't want to receive these updates any more, please unsubscribe here.