At first it seemed like madness, a storyline out of a dystopian film. In Wuhan, a city of millions, only a couple of days before Chinese New Year, the streets were empty. People were confined to their houses. Masks became mandatory.

At the time it seemed like it couldn’t happen here. Yet without a vaccine or effective treatments, that was the only tool China had to control the spread of COVID-19. And as the virus spread, so did the lockdowns.

COVID-19 is passed on when humans do what humans do: interact, face to face. Lockdowns ask us to fight the virus by quelling our basic instinct.

“That’s why the virus is so effective, because it exploits our very humanness. Lockdowns ask us to be very anti who we naturally are,” says Dr Jason Thompson of the University of Melbourne, who worked on the Victorian government’s epidemic modelling.

Make no mistake, lockdowns come with enormous costs: damage to the economy, damage to mental health, damage to individual liberties.

Which makes it important to ask: do they actually work? Are there alternatives? How did Sydney manage to avoid one for so long, while Melbourne has had five? And what will the rise of Delta, and our growing number of vaccinations, mean for lockdowns in the future?

Do lockdowns work?

This seems like a straightforward question. It’s not.

The problem is in the framing – work to do what? What works for some people won’t work for others. Some prioritise the protection of life, others mental health, still others freedom or the economy. It might sound cold, but governments every day must weigh up the value of the lives of their citizens – for example, when deciding which drugs and medical treatments to subsidise.

“There are areas where trying to weigh human life is repugnant,” Dr Hugh Breakey, a moral philosopher based at Griffith University, told The Age and the Herald last year. “Lockdowns don’t fall into that easy case.”

Let’s focus on just one goal, the one governments cite when they send us into lockdowns: stopping the virus’ spread.

Do lockdowns stop the virus spreading? The best national and international evidence suggests they do.

“Lockdowns appear to be very effective – in Australia and all around the world,” says Dr Nick Scott, head of modelling and biostatistics at the Burnet Institute. This is particularly the case when compared to not locking down.

And University of Sydney COVID-19 modeller Professor Mikhail Prokopenko says: “I think it’s important to tell people that the restrictions that we all have been enduring have saved nearly 50,000 lives, in comparison to similar countries like the UK or US. Feel free to quote me on that.”

Liam Mannix’s Examine newsletter explains and analyses science with a rigorous focus on the evidence. Sign up to get it each week.

Three studies in leading scientific journals Nature, Science, and Nature Communications all conclude measures associated with lockdowns, such as closing businesses and banning people from gathering, do work to cut the spread of the virus.

This shouldn’t be a surprise. “[Lockdowns] stop movement. If you stop humans contacting humans, that transmission slows,” says Professor Caroline Miller, director of the Health Policy Centre at the South Australian Health and Medical Research Institute.

But do lockdowns save lives?

“Please know every individual effort you made today and everything you’ll do tomorrow and every day will save lives,” Victoria’s Premier Daniel Andrews told his state as it headed back into lockdown in June.

Is this true? A lockdown that stops a virus while leading to a large number of non-virus deaths – say, from suicide – is a failure on that metric. So could it be the cure becomes worse than the disease?

The data shows Australia and New Zealand experienced no excess mortality in 2020. To put that another way: we experienced a global pandemic that has killed almost 4.1 million people, and Australia saw no appreciable rise in deaths. South Korea, Taiwan, and Thailand – all lockdown countries – also experienced few to no extra deaths.

Brazil, Sweden, Russia and parts of the United States, which largely eschewed lockdown measures, had large numbers of deaths; in the States, COVID-19 became the leading cause of death for several months in 2020 and 2021.

In fact, some countries that locked down and managed to avoid a major outbreak actually reduced their total number of deaths from all causes. Fewer people died than an average non-COVID year, mostly because flu deaths were cut out.

“The best current evidence suggests lockdowns cut deaths from non-COVID causes. Suicide deaths and accidental deaths stay the same or go down in most cases,” says University of Wollongong epidemiologist Gideon Meyerowitz-Katz.

OK, but what about all the harms caused by lockdowns?

Overall, psychological distress increased during Australia’s lockdowns; Victoria’s mental health worsened more than other states, with rises in distress, loneliness and a fall in life satisfaction. However, there has been no increase seen in suicides, and as Victoria’s restrictions eased mental health improved.

There are other harms. People missed medical appointments, scans and checkups. In a report from September 2020, conservative think tank the Institute of Public Affairs claimed that lockdowns destroy “the dignity of work”. Indeed, Victoria’s unemployment rate jumped above 7 per cent in the middle of 2020’s lockdown. The already disadvantaged bear the brunt of lockdown the hardest; and people aged between 20 and 29 both carry the brunt of infections and the brunt of unemployment caused by lockdowns, despite having a much lower chance of dying from the virus compared to older people.

Government debt will dramatically increase, with some worrying future generations will have to pay that back (not all economists share that concern).

But Victoria’s economy appears to have recovered at a dramatic pace. Unemployment has now fallen below 4.5 per cent, one of the lowest marks in a decade.

The other question about harms is this: how much of this is caused by lockdowns and how much by the pandemic itself? That is very hard to disentangle.

Mental health in Melbourne during last year’s second wave worsened, but were people struggling with the lockdown or with the thousands of COVID-19 cases and hundreds of deaths? Were they avoiding medical appointments because of lockdown, or because they were worried about heading to a crowded hospital with a COVID ward?

A clever bit of research suggests the pandemic itself does play a big role. Two economists analysed mobility in border zones in the US, comparing states that had a lockdown against adjacent states that did not. People on either sides of the borders behaved almost the same, suggesting it is fear of the virus, not government orders to stay home, that most affects our behaviours.

What about Delta? Does that change the game?

Early evidence suggests Delta significantly undermines how well our lockdowns work.

If our goal remains the same – stopping the circulation of the virus and therefore saving lives – we may need to lock down harder than ever before.

Delta is about 60 per cent more infectious than the Alpha variant of the virus, which in turn was perhaps 50 per cent more contagious than the version of the virus that first emerged in Wuhan. The more infectious the virus, the harder it is to control.

We can see this in federal government modelling. On average, an outbreak in Sydney at the moment of the original strain would have an R-eff – how many people each infected person infects – of 0.91. That would quickly bring the outbreak to a stop.

With Delta, that number is 1.7, which means the outbreak will continue to grow.

“Delta has completely changed it, totally changed it,” says Professor George Milne of the University of WA, who is providing official COVID-19 modelling to the WA and Queensland governments.

Milne is so worried about Delta he is working on modelling what he terms “stage 5” lockdown measures, more draconian than any introduced so far, in which large numbers of people may be banned from leaving their houses altogether. “Nobody goes out the door.”

More worrying are what his models say about Sydney’s current outbreak: current lockdown measures simply will not be enough to bring case numbers down to zero.

That troubling claim is supported both by federal government modelling and the work of an independent modelling group led by Prokopenko. In a study that has not yet been peer reviewed, published this week, his team found 80 per cent of people in Sydney needed to comply with social distancing to stop the spread; at present, that figure sits at 40 per cent.

“People are saying the current lockdown measures in NSW will put a cap on cases. Our results do not suggest that,” says Milne.

All models are wrong, scientists like to say, but some are useful. They are only as accurate as the data we can put in, and Delta only emerged in March. Sydney’s current situation is the first time Australia has faced a significant Delta outbreak. It remains to be seen how well current restrictions can bring it under control.

Sydney managed to live for a long time without a lockdown. Why?

Experts offer two possible explanations. First, the earlier strains were easier to contain without a lockdown. And it is possible Sydney had better contact tracing and public health infrastructure, although – beyond repeated claims from politicians of “gold standard” contact tracing – there is little hard evidence to prove that is the case, says Dr Driss Ait Ouakrim, an epidemiologist and COVID-19 modeller at the University of Melbourne.

Which leads to the second explanation: luck.“I think they were lucky. I know there was a narrative about how their contact tracing was gold standard. There is no evidence for that. I think it was a large part luck,” says Ait Ouakrim.

Unwilling to accept that? Consider modelling by the Burnet Institute: introducing a single infection into a model of Melbourne and running it 1000 times suggests even with zero restrictions, more than half the time it goes nowhere – and the other half the time you get an outbreak. Such is the random nature of the virus.

“With the original strain, the majority of the time you were safe, but it also depends on how many times you roll the dice – and so now [with Delta] these numbers are a bit worse,” says the Institute’s Dr Nick Scott.

What about postcode-based lockdowns? Are there any other effective lockdown alternatives?

Before Victoria’s second wave , the state government tried a postcode-based lockdown. It did not work. An unpublished paper from Melbourne mobility researchers, shared with The Age, shows one potential problem: following lockdown, the number of people travelling into locked-down suburbs did not decrease dramatically in many surrounding areas, and for some regions the number of travellers actually increased.

NSW had more luck with a lockdown of the Northern Beaches in January. But that was partially down to unique geography, several experts told The Age. Much of the area is on a thin peninsula with only a few roads in and out, making limits easy to enforce.

Could we be more efficient with lockdowns, using only some tools and not others? Say, requiring mandatory masks but keeping businesses open?

Some studies suggest limiting gatherings and closing universities and businesses has a much stronger effect than telling people to stay home. Masks seem to play a key role – a government-funded study on Victoria’s second lockdown found masks played a big part in ending the epidemic – and they can be easily deployed before a lockdown is needed.

But it remains difficult for scientists to disentangle which interventions work best. Asked how far we are between not knowing how lockdowns work and being able to deploy them with surgical precision, Dr Jason Thompson estimates we’re about halfway.

“At a local level, you just do not know,” he says.

OK, so lockdowns work, but they suck. How long can we wait before we have to use them?

In some ways, a lockdown is using a hammer to crack a nut. Adelaide, for example, has locked down 1.7 million people to control just a handful of cases of the virus. Wouldn’t it be nice if we knew exactly when to go into lockdown and exactly how long does it need to last for?

Sadly, our current evidence base does not allow us to do that. We just don’t know enough to perfectly time a lockdown. By the time we realise things are serious, says Scott, “it is too late.”

Ait Ouakrim’s modelling now suggests the Delta variant means states should lock down as soon as spread is detected. “We are now beyond that debate, because of Delta. It’s so fast, that question has been answered now,” he says. “This is the message now: there is no room for ‘wait and see’.”

That is the current strategy adopted by many states: go hard and early.

A model built by the University of WA team providing forecasts to the WA and Queensland governments suggests doing just that is most likely to quickly suppress case numbers. Delaying restrictions by two weeks leads to about 700 more cases in a city Melbourne’s size; delaying by a month leads to about 14,000 more cases. A softer lockdown – equivalent to Melbourne’s stage-3 restrictions – would have allowed Melbourne’s second wave to bubble along for months. The stage-4 restrictions that were instead put in place imposed a night-time curfew, limits on the distance people could travel from their homes, and the closure of certain businesses.

A second benefit to a short, sharp, strategy: it’s plausible that lockdowns work better early, before “lockdown fatigue” sets in. “Lockdown fatigue definitely happens, and it seems to set in within two or three weeks of stage 4 being introduced,” says mobility analyst Rohan Byrne.

There’s a big problem trying to “wait and see”: it’s very hard, if not impossible, to be 100 per cent certain just how many people are infected early in a pandemic.

Milne remembers watching Daniel Andrews, early in the state’s second wave, saying he would “wait and see” before strengthening restrictions. “That just rang such large alarm bells for me. I thought ‘this is so wrong’.”

The problem, Milne points out, is it takes several days for people to show symptoms or get tested, meaning the number of cases is a lagging indicator. “You don’t have a good indication of how many infectious people are in the community.”

How will full vaccination rates change things?

Vaccines have long been seen as our path out of lockdown. In July, Prime Minister Scott Morrison said we would eventually be able to treat the virus “like the flu”. He outlined a four-phase plan to get back to normal. Phase 3 “means no more lockdowns”, he said.

But even though we have a vaccine for various strains of the flu, it is a more serious problem than we give it credit for: about 530 Australians die from it every year. In 2019, it was the ninth-leading cause of death.

Even if we vaccinated 80 per cent of Australians against COVID-19, allowing it to run through the country entirely unchecked would still lead to a significant number of deaths, Burnet Institute modelling suggests.

With a more contagious virus that cuts the effectiveness of our vaccines, such as Delta, still more people would die.

“With the Delta variant, even with high vaccine coverage we are going to require an ongoing public health response to COVID in the long term,” says Scott. “Because people can still become infected when they are vaccinated, even with high vaccine coverage there can be an epidemic among the unvaccinated population.

“Vaccines are unlikely to be the silver bullet that does away with restrictions.”

Liam Mannix’s Examine newsletter explains and analyses science with a rigorous focus on the evidence. Sign up to get it each week.