Where in Russia do people oppose the war? I tracked the use of protest hashtag #нетвойнеI ("no war"), and found opposition in all of Russia’s 50 largest cities, in all 11 time zones and in 83 out of 85 federal subjects.
In Vladimir Putin’s Russia, holding a blank piece of paper can land you in jail. Helped by
@OvdInfo
, I analysed the latest data on anti-war protest in Russia.
What is the pandemic's true death toll? After months of work, we are launching our daily updating estimate of excess deaths around the world. With official deaths at 4.5m, I estimate the true tally to be between 9.3m and 18.1m, as of today
It has been months in the making, but it is finally ready. For this week's issue of
@TheEconomist
, we have used thermal imaging and radar to provide an entirely new view of the war
I was just named Data Journalist of the Year! I won the award for what the judges called "pioneering" work using satellites to cover war. Huge thanks go my colleagues at
@TheEconomist
who make my work possible. More here
Very happy to say that
@TheEconomist
's Data team has launched an open "call for papers" for
#research
on any topic. We are looking for exciting and important new studies - if that sounds like your work, please send it to us
What would "veganuary" for everyone, all year round, look like from space? I had a look at the numbers, and found that it could free up roughly 1/3 of all habitable land.
More Russians are asking Google “how to leave Russia” than have done so in the last 18 years since such data became public. We used search data from the country to track discontent:
Ukraine is outgunned and outnumbered, but not, I find, as much as some numbers suggest. As clashes continue, it appears to have inflicted more casualties in the first 24 hours than Russia suffered over eight years of engagements in Syria.
Has Twitter's algorithm treated politicians on the right unfairly? The data, as far as I can tell, suggests yes. It has tended to give them a greater boost.
For this week's issue of
@TheEconomist
, I had a look at a gigantic multi-year experiment run by
@Twitter
. (If you are reading this, you probably participated.)
Our Ukraine war tracker is now live. Using data from satellites and a custom machine-learning-based algorithm, it updates several times daily. You can check it here:
Our model of excess deaths around the world allow fair comparisons. I found that adjusting for population, the death rate is now about eight times higher in America than in the rest of the rich world.
Do top football players who get covid-19 ever fully recover? Among the best of the best, even a tiny drop in performance matters. We looked at the latest research:
If politicians want to loot, they often bring flags.
For this week's cover of
@TheEconomist
, I investigated links between paranoid nationalism and corruption.
How do you compare foods’ greenhouse-gas emissions? My solution is bananas. Check out your favorite food - by weight, calories, or protein - in our interactive chart
No reliable data is available on how many people continue to attend protests. But the number of arrests appears to have slowed. Between March 9th and 20th, there were 1,065 documented arrests, compared with 14,009 in the two weeks prior.
There is no question about the bravery of Iranian protesters. But what is less clear is how widespread the protests are, if they are still growing - and how they might end. I looked at data on tens of thousands of events to find out (charts:
@_rospearce
)
Is violence on the Lebanon / Israel border spiraling out of control? I built a system to automatically map the fighting. It suggest that currently, the answer is no.
I wrote a short piece about Omicron. By my rough estimates, informed by input from
@TWenseleers
, Omicron infections are now in the hundreds of thousands.
An updated look at the latest Russian war dead estimates - Russia has probably lost more soldiers in the past 17 months than it did in the preceding 70 years.
The situation in Sudan has been getting worse, fast. How bad? The limited information coming out of the country tell of shelling, mass graves, and burned villages. I looked at what the satellites could tell - and it was much worse than I imagined.
Opportunity:
@TheEconomist
's Data team has an open "call for papers" for
#research
on any topic. We want to share the insights of exciting and important new studies - if that sounds like your work, please take 2 minutes to send it to us
In this week's "Off the Charts", I write about how
@TheEconomist
uses voting data to detect election fraud. Wonder where we found it? Sign up for the data team's weekly newsletter here:
@shashj
Perhaps of interest:
@jonasgahrstore
shared estimates of Russian casualties with Putin directly a few days ago in their call (Putin had no comment).
Happy to share that
@TheEconomist
's Data team has an open "call for papers" for
#research
on any topic. We are looking for exciting and important new studies - if that sounds like your work, please send it to us
For ten days, there were no death counts reported from Gaza. The reason was however a lack of data rather than dead. I had a look at the new numbers and their reliability. They imply total dead could now be over 20,000, of which over 8,000 are children.⬇️
We are
#hiring
! The Economist's data team is offering a one-year paid traineeship to an aspiring political data journalist. Details at the link below. If you think the right person might be someone in your network, do share:
Our cover this week. The methods to help malnourished children and mothers are proven and cheap. Implement them at scale, and their benefits could, I project, be in, and last, over 400m lifetimes
Massive activity all along the front line currently, as well as at the Belgorod border: over war-related* 1200 high-temperature events were detected by our system in the past 24 hours. No sign in this data that things are slowing down
That includes deaths to covid-19, but also those lost to overburdened hospitals - and lives saved due to fewer ill with influenza. Rather than trying to distinguish between types of deaths, our approach is to count all of them
I had a look at the sequencing data to track the spread of the delta mutation. Evidence suggests it may soon be the most prevalent strain in America, France, Germany, Italy, Mexico, South Africa, Spain and Sweden, among others.
I see many claims floating around, but according to Twitter's own data from an experiment involving 58m users, their algorithm has in the past favored political parties on the right. ⬇️
Do you want your research to be written about in
@TheEconomist
? Please send it to us. Our submission platform is 100% open and details are at the link - please feel free to share this with you network
@emollick
@halhod
I think dual use is mathematically always possible. Just optimize for the negative of the good objective. But how "bad" the result is will depend on training data for such outcomes.
Ukraine war activity - in terms of fire events probably corresponding to heavy shelling - is now at the highest detected level since the invasion began. See where + caveats here:
I wrote about the Delta variant, rapidly gaining dominance at the expense of other mutations of sars-cov-2 (chart shows why). June 11 PHE brief found increased transmissibility, hospitalization risk, and resistance to vaccines relative to other mutations.
Where countries build their military bases suggest what they plan to do with them. I had a look at every military airfield built or expanded in China in the last decade--and what I found is good news
A bit of good news amid much gloom: tools to fight malaria are getting better. For this week's issue of
@TheEconomist
, I calculate that using them could avert 20m deaths over the next two decades.
Estimating excess deaths for every country in the world, for every day, for multiple years, is a complex process. We are 100% transparent: you can view all our data, estimates, changelog, and code here
How many Russian soldiers have died in Ukraine? Estimates vary, so I collected all of them. By my calculations, the latest figures imply 1% of Russian men under 50 have been killed or wounded in Ukraine in the past two years.
The big push of Ukraine's counter-offensive appears to be over, my data suggests. Activity, as far as I can detect, is down 90% compared to late September.
Huge thanks to all our sources, which we list both in our methodology and live page. That page, designed by the excellent
@martingz
and including all this data, will update every single day going forward. You can find it here:
A common tale of the pandemic has also been one of "waves": but our data suggest that globally, there has been a gradual rise, with one exception: India's delta wave, which we estimate killed millions (global wave below)
These arrest figures are all lower bounds. Without access to internal government tallies, should these even exist, the OVD-Info’s data can only capture part of the picture of Mr Putin’s crackdown on dissent.
@olliecarroll
FWIW we also detected abnormal fire activity within the city for the first time in weeks recently. Though most fire activity picked up by satellites is north of the city
One month ago, we called on researchers to send us their best work. Today, I'm happy to say that an
@TheEconomist
is featuring a "call for papers" submission for the first time: a fascinating study of "wolves’ passports", "enemies of the people", and skills of former prisoners.
It matches this to over 100 other indicators, and then uses a machine learning algorithm to project excess deaths where they are unknown. That was, for instance, the case for India as a whole
Beyond it, the model uses total mortality data from 82 countries, as well as subnational data from Jakarta. Much of it collected by
@ArielKarlinsky
in his datasets
In their study, researchers at Twitter found that the algorithm favors right-wing parties: giving tweets from these parties' legislators a larger boost than it does for those on the left. Audience with algorithm compared to with "old-school" feeds below (see e.g. Britain, US):
For all my projects, I have uploaded full replication code and data on Github (unless I legally or ethically can't). You can find the code and data behind our excess death model here: (23/23)
Using such other data, we can also view the pandemic death counts in new ways. Here are two: estimated excess deaths per population over 65 years old, and implied infection fatality rates
Huge thanks
@Rukmini
, Sumitra Debroy,
@thej
, Srinivasan Ramani, and Vignesh Radhakrishnan who have collected total mortality counts from India, which helped our models be more accurate
@emollick
@halhod
Interestingly, good training data on bad outcomes is often very important in producing good outcomes (to avoid them). Avoiding such uses seems a hard but interesting problem to solve
All this work allowed me to estimate excess deaths that were unknown. Official data suggest 3.3m have died so far: I found the most likely number to be 10m, with a likely range of 7-13m (20/23)
For 99% of Twitter users, a deep learning algorithm decides. What difference does that make? Our analysis suggest more right-wing politics and less reliable media sources.
@LuisRMartinezA
@TheEconomist
A pleasure to feature this
@LuisRMartinezA
. To everyone: I am always on the lookout for excellent and important new research to share with our readers - please send papers my way.
You can read the full piece here - which includes an introduction to this work through the battle for Antonov airport, maps of destruction in individual cities, and more on what the fire data tell
I developed a machine learning algorithm that uses pre-war patterns in fires to investigate those which have taken place since the fighting began. It classifies fires in excess of plausibly normal activity as related to the war
@fhuszar
Very interesting, thanks. I've started seeing things from the opposite angle - if humans can do something, why shouldn't a (sufficiently well-resourced) DL-model be able to?
Those joining the movement have included Ivan Urgant (who saw his tv-show taken off the air) and
@lachesisq
, who wrote on Twitter: "History has seen many Black Thursdays. But today is blacker than the others. #нетвойне
#saynotowar
"