While toggling the dashboard for a multi-risk SIR model with optimally targeted lockdown developed since February by top MIT macroeconomists, it occurred to me just how primitive the modelling behind the lockdowns were.
SIR stands for susceptible, infected or recovered and all academic models are rightly based on the 1927 SIR model.
• Covid 19 coronavirus: New modelling reveals Australians face isolation restrictions for months
• Coronavirus: The maths of Covid-19 - and how you can change it
• Covid 19 coronavirus: Big Read: What you need to know about scary virus projections
• Covid 19 coronavirus: Interactive - How does New Zealand compare?
The MIT dashboard allows online users to select their own cost of death valuation. How much is each extra year of life and well-being lost by Covid deaths worth?
I toggled for lockdown effectiveness, infection and testing rates, and the fractions locked down those who were young, middle-aged and old. Daron Acemoglu and his MIT co-authors found that a much longer strict lockdown, just limited to the old, could halve the economic losses but nearly double the lives saved.
The Imperial College model, developed since 2005, was adopted by many governments.
Leaving aside coding quibbles and replication difficulties, it massively over-predicted
deaths and infection rates before and after accounting for changes in public behaviour.
The MIT model and other models by other top economists were cobbled together since mid-February. I do not know if they received any research grants. They just dropped everything, worked into the night learning a new literature to develop sophisticated models that included subtle behavioural changes and estimated the cost in jobs and GDP under different lockdown scenarios.
Governments needed to know what their interventions do after accounting for reasonable changes in normal people's behaviour. Google Map tracking of phones showed that most of the social distancing and collapse in travel and tourism occurred weeks before any government mandates.
Models worked up since February have quantified major social costs from premature lockdowns.
Alexis Akira Toda's April 21 model found that an early lockdown is suboptimal because it prevents the building of herd immunity against an even worse second wave while incurring big economic costs.
Krueger, Uhlig and Xie's April 23 model of substitution of spending and jobs between a restaurant sector and a pizza home delivery sector let the epidemic play out without government intervention.
The Swedish solution avoided 80 per cent of the decline in GDP but still saved 80 per cent of the lives as a lockdown. People fearing infection social distanced on their own initiative.
Chari, Kirpalani and Phelan's hammer and the scalpel model found last month that when targeted testing and isolation policies, such as in South Korea or Taiwan, are used instead of a widespread lockdown, the fall in GDP more than halved to 15 per cent and cumulative lives saved tripled.
Ciminelli and Garcia-Mandicó's May 19 analysis found our Alert Level 3 was more than enough.
The shutting down of non-essential in-person services greatly reduced deaths across Italy, but the closing of Italian factories 10 days later did not. Death rates were virtually identical in the average Italian municipality and in those with a higher share of employment in factories that were closed.
If we let people work most of it out by themselves, we do not need Cabinet meeting into the night deciding which toilets stay open for truck drivers and which shops close.
Cabinet lacked the wherewithal even to draw up, after weeks, a decent list of what should close or not at alert level 4.
If our Cabinet had the free online MIT dashboard on their laptops when they decided on the alert levels, they might have made quite different decisions.
It's another question for a Royal Commission.
• Economist Jim Rose blogs at utopiayouarestandinginit.com.