Overreaction vs underreaction


There has been a rather hefty amount of speculation about the work that various countries, governments, municipalities, religious places, universities and frat houses have been taking in making sure that we're all successfully protected against Covid. In rather straightforward journalistic narrative fashion, it tends to ride a wave of superlatives.

Some reactions:

  • "We're doing nothing and need to do way more."

  • "Those who disagree with the lockdowns are morons and killers."

  • "The only sensible course is to open everything up, otherwise it's tyranny."

  • "Freedom....!!!!"

  • "We have to be more thoughtful about our individual actions, but we should all agree that governments are not doing enough."

And to add joy to this staccato of opinions, there is also a surprising amount of punditry that argues we're doing too much, too little, too little of the right things, too much of the wrong things, and every combination thereof. There's Bryan Caplan recently stating that the cost of Covid prevention is way too high. Robin Hanson vehemently agrees. Tyler Cowen disagrees. And on it goes.

The toughest part is not the disagreement, or the clearly stated assumption on almost all sides that they have stumbled onto an obvious truth that everyone else is too wilfully stupid to ignore.

First, to respond, my biases upfront.

  • I think we've done far too little to actually prevent things in the West. Lockdowns, when imposed, have mostly been anaemic post the first ones in Spring. Rules imposed have been byzantine and unenforced, relying on public goodwill to get anything done.

  • The natural human reaction in most places, except for those who seem to like protesting things, is towards some combination of laziness and reasonable rule following. Yes, wear a mask when in a store. No, don't wear it walking around outside. Yes, don't stay overnight at someone else's house. No, feel free to go have a beer. And while this is suboptimal from a mathematical modelling standpoint, it's also rather obviously not irrational.

  • The costs we are bearing as a society is not entirely one of mandated lockdowns that are making us poorer. It's one where every individual takes a decision saying "damn, this seems like a big deal since a million and half people have died, so maybe I should take care", and that depresses economic functioning. Government mandates help, but it's a catalyst and not the main determinant. We've seen places like Sweden, which didn't impose any lockdowns, have similar economic hits as places which did.

People aren't morons - if there's a killer virus roaming about, they're gonna stay home. If you don't believe me, imagine there's a terrorist ninja assassin roaming about who will kill 1-2 people out of every 100 he meets, and see if you feel safe stepping out!

The arguments against doing more prevention seem to revolve around an extremely simplistic analysis of "Cost of lockdown" vs "Benefit of QALY lives saved." I have no methodological or moralistic quibble here. But the math is wrong! Cost of lockdown is not the delta GDP lost, it's delta GDP lost due to societal action (which is being argued against) vs GDP that would've been lost anyway.

And it also rather directly impacts the QALY lives saved. How could it not? If you save 1m more lives due to lockdown, then naturally the costs and benefits increase side by side.

So how do we solve this rather simplistic equation?


Enter spreadsheets. I tried to do something basic. Assess a "death impact", equivalent to R score, which increases if everyone just goes about their lives as normal. Then there's a "Panic reaction", which is the amount of panic you can induce in society through governmental and media actions.

Then we introduce some nice circularities where Death affects panic (obviously), Panic affects people's actions (again, obviously), and Death affects people's reaction too (same comment). So finally there's the actual action people take as a result of Death impact and Panic reaction.

And here I've created the world's most simplistic and obvious sensitivity table to our panic reactions and death impacts. I indexed the starting point here to 100, just to see the level of drop you would expect based on varied policy reactions.

If the panic reaction is high, then the death impact will be low, but it affects the GDP a lot! And if the panic reaction is very low, but the death impact is high, it still reduces the GDP by a massive chunk.

Seems like our silly three variable model tells us what our common sense also tells us. You should try to reduce panic and also reduce deaths. So far, so genius. But how? Well, some thoughts:

  • Maybe this should lead us to believe that not all Panic is the same. Maybe it can be directed in the right places without the cognitive complexity increasing so dramatically that it loses efficacy?

  • Maybe the government should help economy transitioning into the 'new normal' by significantly helping those who are being hurt by the transition, and that will reduce the sensitivity of People's Action being affected by Panic.

  • Maybe the government should directly help people so that the "People's action" isn't being taken despite the Panic, by helping them get a floor underneath them.

  • Maybe the government should do more to reduce the Death impact by ... well by R&D I suppose, pushing it forward however you can.

Ultimately whether you believe in more test and trace (reduce Panic), or masking (reduce Panic and reduce sensitivity of People's Action to panic), or full lockdowns on everyone (reduce Death impact), you're still playing in the same field.

The economy would be better if we were all ignoring the doom and gloom worries and just living our lives, or we might be killing 20x more people worldwide, exactly as some models predicted pre-lockdowns. There are only a limited number of actions available. Wishing it weren't so doesn't change the fact. Yes life would be nice if we could all follow masking rules and test and trace. But we don't. There's no such thing as a free lunch.