What have we learned about the coronavirus?
It's been 6 months since the pandemic hit us all. Are there any facts we can take to the bank?
Welcome to Plugging the Gap (my email newsletter about Covid-19 and its economics). In case you don’t know me, I’m an economist and professor at the University of Toronto. I have written lots of books including, most recently, on Covid-19. You can follow me on twitter (@joshgans) or subscribe to this email newsletter here.
It seems like Covid-19 is a constant debate about government policy. Should we require masks? Should we reopen schools? Should we allow restaurants? Should we quarantine people if they travel? If we knew more about what each of those policies would get us, perhaps these choices would be easy. But delve deep enough and what drives policy-makers is more a judgment call on being seen as prudent versus being seen as prescriptive. When there is uncertainty, we just cannot know. But when things are costly whichever way you look at them, there is a tendency for policy-makers to sell policies under a veneer of certainty.
Six months have passed. We surely know more about how this virus spreads than before. In some cases, what we learned updated our thinking. We have learned that masks can help reduce spread as can good ventilation. But on many dimensions, things remain decidedly murky.
Today I want to present the results of a recent paper by Andrew Atkeson, Karen Kopecky, and Tao Zha who tried to find some facts that we could take to the bank to inform our policy-making. I should warn you that these facts are stylised. But the authors have documented them across many countries and also, data permitting, at the higher resolution of US states. In this sense, they seem to be more universal and related to characteristics of the virus itself rather than the culture, institutions, or state competence of countries.
They document four facts. Let’s go through them.
Fact 1. The growth rate of daily deaths from COVID-19 fell rapidly everywhere within the first 30 days after each region reached 25 cumulative deaths.
Daily deaths is one of our better indicators of what happened. This is because many Covid-19 cases went unrecorded and what’s more, we don’t know how many we missed. But we tend to record deaths more accurately.
It turns out this picture from New York is universal.
New York had a very bad outbreak and a large number of deaths. But a month after they had reached 25 deaths in the region, the growth rate in deaths peaked. Now there were still a rising number of deaths after that point but there was an inflection point and the growth rate started to decline and eventually become negative. In other localities that had a significant number of deaths, the same pattern could be observed. The chief question is: why a month? That is the universal thing going on here.
Fact 2. After this first period of rapid decline, the growth rate of daily deaths in all regions has hovered around zero or slightly below.
After the growth rate declined, it then stuck the landing — at zero. In other words, there were deaths each day but they were neither growing in number nor declining. Why didn’t they keep on going? Instead, they reached some level and then just kept on going at a smooth rate. It is like society decided they could bear a certain number of deaths per day and that was it.
Fact 3. The cross-regional standard deviation of growth rates of deaths fell rapidly in the first 10 days of the epidemic and has, subsequently, remained low relative to its initial level.
Different regions had different experiences with the virus. But the big differences were all in the first couple of months. Once you got to a month after 25 cumulative deaths, the experience was pretty similar. In other words, the growth rate in deaths varied enormously initially and then not so much later on — indeed, as Fact 2 says, they fell to zero. So whatever experience regions had early on, they ended up at a point where the number of new deaths steadied out.
Fact 4. When interpreted through a range of epidemiological models, Facts 1 - 3 imply that both the effective reproduction numbers and transmission rates of COVID19 fell rapidly from widely dispersed initial levels during the 30 days after cumulative deaths reached 25. After this initial period of rapid decline, the effective reproduction number has hovered around one everywhere in the world.
This fact is demonstrated here.
Using death rate data, the authors confirm what we have seen in infection data (imperfect though that might be) — the effective reproduction rate goes towards 1. That means, after 30 days of a significant outbreak, you ended up with a situation where on a daily basis there were as many people becoming infected with the virus than who were recovering. This is something I documented in an earlier post but without the rigour of this study.
Policy Debates
Given that policies have varied widely across countries, the universality of the pandemic experience documented here should give us pause. This is what the authors say:
One of the central policy questions regarding the COVID-19 pandemic is the question of which non-pharmaceutical interventions governments might use to influence the transmission of the disease. Our ability to identify empirically which NPI’s have what impact on disease transmission depends on there being enough independent variation in both NPI’s and disease transmission across locations as well as our having robust procedures for controlling for other observed and unobserved factors that might be influencing disease transmission. The facts that we document in this paper cast doubt on this premise.
What they are saying is that if this is the way the pandemic goes, then we have virtually no hope in working out that any given policy choice is correct or not. Moreover, it is hard to know whether interventions — the experience of which differed between regions quite markedly — had any effect on transmission rates.
That is certainly one message here. But the other message is that humans are humans. If you have policies, people will follow them and protect themselves. If you don’t have policies, people will protect themselves of their own accord. In other words, policy debates are potentially meaningless. What we are arguing about is whether the people who don’t protect themselves should be forced to do so but our experience now tells us that forcing them to do so won’t necessarily help reduce the spread of the virus.
That said, there are missing elements. First of all, if people moved after 30 days to protect themselves, how did they know whether to do that? Enacting a policy actually can communicate that message as can normal information flows that are much better as a result of the internet and mobile phones. Were there situations where information wasn’t travelling or expected not to travel and so policy-makers stepped in to fill the gap?
Second, not every region got to 25 cumulative deaths. One that didn’t was New Zealand. Another was Taiwan. Why? Because they acted ahead of the 30-50 day ‘time for action.’ In other words, this analysis is telling us that the virus had a ‘cat out of the bag’ quality. In order to get ahead of the virus, you have to get ahead of the people. Very few countries did that; a theme I will pick up again in the next newsletter.