NEW: today we've published the culmination of a weeks-long @ftdata team effort, summarising everything data has shown us about the virus over the year so far, from Wuhan to the autumn resurgence ig.ft.com/coronavirus-gl…
Free to read, with a wealth of #dataviz on the key trends.
A huge amount of work from a vast cast of people went into this, over many weeks.
Consider this our rolling credits screen:
1) @caletilford & @aendrew built a beautiful interactive experience, designed by @carolinenevitt who also drew up a bespoke colour palette for the series
4) @manib0g was a masterful cat-herder, keeping us on our toes and making sure everything happened when it needed to
And 5) in the director’s chair as our Spielberg/DuVernay was @theboysmithy, who somehow managed that alongside making charts, writing words & doing 12 other jobs
• • •
Missing some Tweet in this thread? You can try to
force a refresh
This is why the Japanese approach of avoiding the three Cs (enClosed spaces, Crowded spaces, Close contact) has been so effective.
Even if someone is highly infectious, you hugely constrain the amount of spreading they can do if you limit the number of people they’re mixing with
We still don’t know exactly what causes super-spreading. Is it mainly that someone is especially infectious, or mainly that a moderately infectious person spends time in an enclosed & crowded space while infectious?
Whatever the answer is, avoiding the 3Cs can only help.
I’ve noticed a lot of people slipping up on how they interpret UK Covid-19 prevalence & testing data, so here’s a very brief thread on how to interpret figures from different sources, and what caveats each source does and does not come with:
• Pillar 2 community testing: these are the bulk of cases picked up at the moment. Case and positivity rates here *could* be influenced by where and who is being tested, so e.g patterns in this data with age, deprivation etc could be skewed by who is getting tested
• @ONS infection survey: these tests are random, and designed to be representative of the overall population.
Therefore trends and patterns in this data *are not* due to e.g certain locations or groups of people being more likely to get tested.
1/ Footfall in central London is still down 69%, but has picked up elsewhere
2/ This is driven by working patterns, but that in itself plays out in two distinct ways:
First, job type. Staff are returning to the workplace at very different rates in different sectors, and the sectors with the most remote working today are clustered in cities, esp. London
3/ Workers in retail, hospitality can’t do their jobs remotely and have returned to the workplace. They’re popping out for lunch or drinks near work and maybe shopping centrally before going home.
In big cities, office-workers are still at home, leaving the high streets empty.
The most effective way to keep Covid in check and return to semblance of normality (far more so than blanket restrictions) is to have as many people as possible being tested, regularly & regardless of symptoms.
For government to be discouraging people from getting tested is wild
If there’s still a shortage of testing (or in this case test-processing) capacity, that’s a problem to be solved on the supply side, not the demand side.
Blanket restrictions (which do the most economic damage), are what countries do when their testing apparatus is inadequate.
The “overreaction to a 'casedemic' is killing our economy/cities” crowd are tilting at a false dichotomy where our only options are:
"Keep restrictions in place to limit transmission, hurting econ & cities" or "It’s overblown, let’s get back to normal and save econ/cities"
When @christinezhang joined the FT as our US elections data reporter, this is exactly the kind of piece I was excited to see:
Read her brilliant explainer on how US polling methods have changed since 2016 and why this makes 2016 vs 2020 comparisons tricky ft.com/content/b32976…
2 big differences between 2016 and today:
1) As the link between US education levels and partisanship has grown, it's become vital for pollsters to weight by education.
Most didn't do this in 2016, but are doing now, so the same Biden lead today would have looked larger in 2016
Here's a more detailed example: there was little to no partisan slant of education before 2016, but in 2012 the gap opened up, so new edu-weighted methods (pink) now give consistently smaller Dem leads than old methods (green)