As Australia prepares to gradually relax the non-pharmaceutical interventions (NPIs) that have led to our success in containing COVID-19 — from locking down the border and mandatory quarantining of returning citizens, to social-distancing and the shuttering of pubs and restaurants — it is crucial that we have the correct narrative about what is behind us, and what lies ahead.
Narratives — stories, really — are important because they help us make sense of complicated fact patterns. And as we continue to battle COVID-19, there is another battle taking place: for the narrative describing our response to the pandemic.
In light of our success various commentators that were “lockdown sceptics” from the start — like Adam Creighton, Henry Ergas and Janet Albrechsten — have suggested we could go “too far” or even reinterpreted the success of NPIs in Australia as evidence that we may never have needed lockdowns at all, in the case of Creighton.
Some have said that epidemiological models are “wrong” and should not be relied on. Their preferred narrative is that we, as a nation, panicked and caused additional economic harm for no reason.
These claims are clearly false given what we have seen overseas, everywhere from Italy and Spain to the UK and US. Looking at the relatively “good” observed outcomes in Australia, 7054 cases and 99 deaths to date, and concluding that NPIs were unnecessary is a rookie error.
It’s akin to looking at a community with lots of police on the beat and a low crime rate and concluding that if all police were removed the crime rate would stay low.
Australia’s relatively good COVID-19 outcomes are precisely because of our interventions. And countries with bad outcomes are due to their interventions coming too late.
As leading epidemiologist Jodie McVernon told The Australian’s Paul Kelly: “I think in Australia we really don’t appreciate what we have been spared. We have not lived in a situation where we’ve used ice rinks as makeshift morgues or seen refrigerated trucks outside hospitals. These things are happening in other high-income countries.”
Whether we adopt the “McVernon narrative” or the “Murdoch narrative” remains to be seen. But that choice will affect how we respond to the remainder of this crisis, and to the next one. And that choice of narrative depends on how we understand the path of the pandemic in Australia to date.
One good way to do so is by reference to epidemiological models that predict various health outcomes like infections, hospitalisations, and deaths depending on what containment measures are put in place.
These models estimate the probabilities of different outcomes and report statistics like the “expected” (or average) outcome, but also what might happen with a 5% or 10% probability. This is because of the inherent uncertainty about key features of the virus such as the proportion of infections severe enough to require hospitalisation, and how containment measures reduce the replication rate of the virus.
So if we see better-than-expected health outcomes, we should not view that as inconsistent with a good model. Similarly, we should also remember that models often tell us that things might have been a lot worse than how they turned out.
That is particularly important in the case of a deadly virus that can grow exponentially, like the one we currently face.
Avoiding relatively low-probability but catastrophic events is the right policy response, even if, by their very nature, we don’t often see those horrible events occur.
We need to remember that low-probability events actually happen. The famous US election forecaster Nate Silver gave Donald Trump a 30% chance of being elected president in 2016, compared to Hillary Clinton’s 70%. We all know what happened. But just because Trump is president, it doesn’t mean that Silver’s model was wrong.
Similarly, a professional golfer has a 19% chance of missing a five-foot putt. So you would bet that, faced with such a putt, they will make it. But if they miss that doesn’t mean your model is wrong. The key is to remember your assessment of the probabilities before you know the outcome.
One long-recognised reason that we are bad at doing this is what social psychologists call “hindsight bias” — a concept first demonstrated by Baruch Fischhoff when he was a PhD student of the celebrated psychologists Daniel Kahneman and Amos Tversky, who pioneered the study of heuristics and biases.
A key flaw in most people’s reasoning is that we perceive events that occurred in the past as being more predictable than they actually were. Often it is relatively easy to construct a narrative that rationalises events after the fact, but which we couldn’t articulate before the fact.
In The Undoing Project, Michael Lewis recounts Tversky telling an audience in 1972:
All too often, we find ourselves unable to predict what will happen; yet after the fact we explain what did happen with a great deal of confidence. This ‘ability’ to explain that which we cannot predict … represents an important, though subtle, flaw in our reasoning. It leads us to believe that there is a less uncertain world than there actually is.
And it is this belief in a more deterministic world that is so dangerous, as it can lead us to not take appropriate precautions against dangerous future events. Other recent viruses like SARS and MERS killed fewer people than many models predicted, perhaps leading to a sense of complacency. No lesser figure than Bill Gates warned about a deadly global pandemic in 2015, but few of his precautionary measures were put in place.
We will have other pandemics in the coming decades, and we may yet face a second wave of this pandemic in Australia. Epidemiologists worry about things like an avian flu with a 60% fatality rate. And rightly so. Australia’s story so far is one of catastrophe avoided, not overreaction.
Richard Holden and Bruce Preston are professors of economics at UNSW Business School and the University of Melbourne, respectively.