The work of David Spiegelhalter was part of the network’s first year of reflection. He works at the Winston Centre for Risk and Evidence Communication. Their strap line is inform not persuade, which is not a bad summary of the principles set out in the last part of How to be Wrong.
Spiegelhalter has been critical of what he calls the ’numbers theatre’ at the heart of government application of data to policy. In a recent article in the Financial Times (16th of April ‘Risk is a very loaded term’) he is quoted as saying:
“They were reeling out lots of big numbers which I knew were desperately unreliable, and it gave a spurious sense of precision and importance to these very flaky numbers. …. It was an appalling lost opportunity, because there was a public hungry for proper detail, who were sacrificing so much, and yet they were getting fed this stuff.”
In the FT article he reflects on risk being applied more to the downside than to the upside of life. There is risk of winning the lottery. There is the risk of being happy today. He welcomes work that counts both harms and benefits.
There is a nice example in the FT piece. (The FT are the leaders in data communication, and their article comes with a beautiful graphic. But its lives behind a firewall that I cannot share. So we have to make do with the original graphic broadcast on the Winston Centre website).
It juxtaposes on the left the benefits (the number of ICU admissions prevented every 16 weeks) with, on the right, the risks (the number of serious harms) associated with the AstraZeneca vaccine. It is pretty obvious from these data why it might be limited to the over 60s (or over 50s), even though the downside for those 20-29 remain minimal.
One more thing about Spiegelhalter. He is happy to say when he’s wrong. At the beginning of the pandemic he was sceptical about the potential death rate. Initially he estimated that around 15 per cent of people died of Covid would have died anyway. He later revised this figure down to five per cent.
https://wintoncentre.maths.cam.ac.uk
https://www.ft.com/content/76e93737-b2f2-4b07-9afe-a2cc21745eaf (firewall)