Special Post on the New Australian Lockdown

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.


Popping in today for an extra post devoted to Australia. Thanks to an escape through quarantine, Victoria had a significant Covid-19 outbreak from June onwards that resulted in a very strong lockdown for six weeks. Now, there is a plan for re-opening that is very cautious. The goal of this is to eliminate the coronavirus from the state and with it, hopefully, the whole country. This is a goal that I very much approve of given how uncertain the timeline is for a vaccine and how difficult it is to continually manage the pandemic. But, the plan has to make sense and it has to be implemented with a view to minimizing the economic and other costs. From what I can see, I am worried about the judgment going into the specifics of the Victorian approach.

Let’s start with the numbers. Here is the graph of new Covid-19 cases in Victoria over time.

Lockdowns were implemented in stages and then at the beginning of August, Victoria went all in with a very strong lockdown. There was an 8pm curfew, face masks outside the house both outdoors and indoors, everything closed but essential services and then police enforcement of all of this. This seems to have worked. In fact, surprisingly well. Usually, when you are on an upswing, it can take two or more weeks to see an effect. But in Victoria the decline was dramatic.

Reopening Plan

Given that the six weeks of major lockdown was nearing an end, today the Victorian government announced its reopening plan. Here are the highlights:

Residents will remain in stage 4 for another two weeks, but with some small extra freedoms. These include:

  • Easing curfew from 8:00pm to 9:00pm

  • Allowing 2 hours of exercise, up from one

  • Allowing outdoor public gatherings of two people or one household

  • Creating "social bubbles" for people who live alone

  • Reopening playgrounds

Step 2, 3 and 4 of the plan are all subject to health advice, and depend entirely on daily case increases going down.

If new cases meet the required thresholds (read more about those here), step 2 will begin on September 28 allowing:

  • Public gatherings of five people from two households

  • Staged return of some students to school, and childcare reopens

  • More workplaces can reopen

  • Pools can reopen and personal training sessions with up to two clients allowed

  • Outdoor religious gatherings with five people and a leader allowed

Step 3 could look like this from October 26:

  • Curfew abolished, no restrictions on reasons or distance to leave home

  • Up to 10 people can gather outdoors, and up to five people from another household can visit homes

  • More progression on school years 3-10 returning

  • Hairdressing, retail and hospitality can reopen conditionally

  • A staged return to outdoor, non-contact sport for adults

And from November 23, subject to all the necessary requirements, step 4 includes:

  • Allowing up to 50 people to gather in public and up to 20 visitors at homes

  • Hospitality to reopen with limits, retail and real estate to reopen

  • Up to 50 people at weddings and funerals

  • Further return to community sport

The triggers for transitions are less than 50 daily cases on average over 2 weeks, 5 daily cases and less than 5 mystery cases and then no new cases for 2 weeks.

There are some good things about this plan. First, schools are opening relatively soon and before pubs. Second, playgrounds are opening quickly as are allowances for social bubbles. And big gatherings are last. I would have made stronger distinctions between indoors versus outdoors given that the Australian summer is upon us but whatever.

The issue is, why is the pace of reopening so slow? One reason could be that the government is trying to avoid disappointment of things being extended. But playing those games seems second order to providing clarity on the plan. It seems to me that the plan is slow because it is based on some modeling. So let’s look at that modeling.

The Victorian Model

The model used by the government was published the other day in the Medical Journal of Australia. So it is peer-reviewed. But peer-review only tells us that the model is accurate for what it claims to do, not whether it is the right model for the decisions being made.

The model is an ‘agent-based’ epidemiological model. That means that unlike the SIR model that many of us economists have been learning that explicitly lists equations to describe aggregate behaviour and information flows, this one is a computer simulation based on the interaction of agents. What that means is that you run the simulation over and over again as agents randomly run into each other and see how the pandemic progresses. That can be a useful approach but it is heavily dependent upon a critical assumption: that agents spread the virus by interacting with neighbours but that the distribution of agents over space is, in order to make this computationally viable, pretty smooth. This means that such models don’t divide up into groups neatly and have the quality that if there is a little bit of the virus somewhere, it eventually ends up everywhere. So what you want to do as you adjust the model to get elimination is to eliminate the cause of transmission in the model — people moving.

Not surprisingly, that is what the Victorian government is doing. It used a model that is calibrated but fundamentally based on people moving around and then it decided to stop people doing that because, not surprisingly, in the model that will work and, indeed, it is the only thing that works.

Actually, stopping people moving does most of the work. Mask wearing does things as well. But somewhat frustratingly, they didn’t run a standard lockdown with 90% masks. But since this is an agent-based model, it probably didn’t do as well.

Is this the right model?

Recall that the premise of the agent-based model is that people interact with neighbours and are linked in a fairly smooth, albeit probabilistic, manner. Is this the case for the spread of the coronavirus. Here is a map of the pattern of outbreaks across greater Melbourne where the strongest lockdowns are in place.

This is the pattern right now but I have been watching this one and it has been the same throughout. The outbreak is concentrated in the west of Melbourne and is hardly present in the East. It won’t surprise you to learn that, generally, that the West is not as wealthy as the East. But as you move away from the city centre, the wealth drops in all directions.

This is not the pattern that would justify an assumption of smooth spacing. This is a pattern that suggests that people interact more intensively within their own local areas than in chains that have the same probability of transmission city-wide. It also suggests that if you are going to have a stringent lockdown and need resources to make that work, there are places where it is more important to put those resources.

Other factors

A couple of other things worth observing. First, it doesn’t look like there are many google searches for symptoms that would suggest that the government is missing people in testing.

Second, there doesn’t seem to be any hospital capacity constraints that might have caused concern. Finally, summer is coming. If you let them, people will be outdoors more and from what we know about the coronavirus, that really reduces spread.

What Should Victoria Do?

For one, it should try other modeling approaches and compare their outcomes with the agent-based one. There is too much economic cost to additional months of lockdown not to do this. As I outlined a few weeks back, what you want to do is ensure that you have taken into account network patterns in a city as that changes the approach you would take.

Second, I would make reopening either postcode based or local government area based. That way you can monitor whether low prevalence areas have outbreaks from reopening and use that to inform the pace of reopening. Taking the whole of Melbourne as your unit for triggers seems crazy to me. The best case for it is based on people regularly traveling far distances across the city. The worst case is some notion of fairness.

Third, I would encourage people to be outside as much as possible. No mask mandate outdoors. A relaxed approach to gatherings outdoors etc. This makes sense and is going to make the job of enforcing the more important directives much easier. People are doing that anyway if you look at the Google Mobility reports.

If I had to guess, I suspect a cadence of reopening plans of two weeks with triggers based on finer locations rather than a month based on the whole of Melbourne would make more sense.

Finally, and I can’t emphasise this strongly enough, test and trace! Lockdowns alone won’t get to 0. But when cases are low, aggressive identification and isolation of infectious people will do the trick. Victoria is in striking distance of doing that while avoiding economic pain. They should double down on it.


What did I miss?