Problem 1: asking the wrong question

Estimating short-term excess mortality effects cannot tell us whether lockdowns were a good idea.

The conventional approach to policy evaluation is to identify, estimate and weigh against one another all plausible costs and benefits of a policy. It goes without saying, but is probably worth noting, that unless the proven societal benefits of a policy exceed its costs, it should not be implemented in the normal course of events. This approach, based on well-established methods of cost–benefit analysis (CBA), is not mentioned in CL, nor is it acknowledged that their own analysis can only provide a small input to the core question.

Based on demonstrating the absence of excess short-run mortality, the authors seem to try to normalise lockdowns as a robust scientific policy, advocating working from home during ‘the winter months’. CL fails to reference the serious CBAs of Australian and other countries’ lockdowns that had already been produced prior to the receipt by BMJ Global Health of CL on 26 March 20222–4 and which concluded that the costs of lockdowns overwhelmingly exceeded their benefits. Illustratively, Allen reviewed over 100 studies in 20215 and concluded that ‘using a midpoint estimate for costs and benefits, the reasonable estimate for Canada is a cost/benefit ratio of 141. It is possible that lockdown will go down as one of the greatest peacetime policy failures in modern history.’

Problem 2: short-run excess deaths are an incomplete measure of human welfare losses in 2020–2021

As noted by Sabhlok,6 ‘This debate is not just about deaths. It is about mass scale, society-wide torture. Lockdowns are mass imprisonment and cause mental anguish on a mammoth scale’. A countless myriad of non-fatal harms to millions add up. Lockdowns impacted all sectors of the community, including children. Even a public health evaluation of only the short-run effects of lockdowns would need to count all such costs and all benefits, rather than being restricted to mortality alone.

Suspending the activities of everyday life naturally reduces the risks of death carried by some everyday life activities (including driving cars, going to pubs and having elective surgeries or other non-emergency medical interventions). Thus, lockdowns would inevitably reduce some deaths in 2020 and 2021 that would otherwise have occurred for these reasons (eg, car crashes, accidental homicides and medical misfortunes). Yet each of these activities is a calculated risk undertaken in normal times. We drive our cars, see our family (including visiting our grandparents), and opt for risky medical procedures because of the expected benefits of these activities. Living carries risks and constraining the act of living carries costs entirely ignored in CL.

Problem 3: long-run welfare costs (and other costs) of lockdowns are ignored

A preliminary cost–benefit analysis for Australia presented by Foster to the Victorian state parliament in August 2020 suggested that lockdown harms were far greater than their benefits. In 2022, this analysis was expanded to become Foster with Sabhlok7 (hereafter FS), which showed that lockdowns’ harms exceeded their plausible benefits by at least 68 times.

Only about 1% of the lockdown harms estimated in FS were attributed to a short-term increase in non-COVID-19 deaths in 2020 and in 2021. Even if these deaths are excluded, the net harms of lockdowns still immensely overwhelm any plausible benefits.

Reductions in happiness (see problem 2 above) during lockdowns constitute around 44% of FS’s estimated total costs of lockdowns. The long-run cost of the increase in debt—which will crowd out future government expenditure, carrying significant health impacts—amounts to 49% of the lockdown costs estimated in FS. Other long-term harms include costs not yet well quantified so not counted in FS, such as the loss of trust in institutions and within society, the development of unhealthy habits, and a loss in productivity.

Problem 4: the link between mobility and mortality cannot be due to reductions in COVID-19 deaths

The authors of CL attempt to demonstrate the effectiveness of lockdowns by regressing short-term weekly excess mortality against weekly mobility data (their Figure 3), concluding that in weeks in which people did not go to retail shops, transit stations or workplaces, fewer people died.

Given the incubation period of the COVID-19 virus8 and the average time from symptom onset to death with COVID-19,9 this inference is implausible. One week is not enough time for the virus to transmit, incubate in a new host and then cause death. Any correlation in these two measures is more likely due to reductions in events like traffic accidents and homicides, and/or due to the political cycle combined with the natural ebbing and flowing of virus deaths: when virus deaths were rising, governments tightened restrictions and even though these restrictions had little to no impact on viral trajectory, once they were in place viral deaths started to decline because that is what was going to happen anyway, regardless of government policy.

Problem 5: other misleading statements and representations

CL misrepresents the policy context at the start of the COVID-19 era, the data later on, and what those opposed to lockdowns said about them, and the authors also self-contradict on a crucial point.

First, CL fails to highlight the starting position of Australia that lockdowns are not a viable policy option. Illustratively, the Victorian Pandemic Plan of 10 March 202010 took a risk-based approach and ‘focused on protecting vulnerable Victorians’. It explained that ‘older Victorians and people with chronic diseases are known to be at greater risk of COVID-19 infection,’ saying that it would ‘ramp up risk reduction activity [for] at-risk groups’.

Second, CL misrepresents the Great Barrington Declaration11 (GBD). CL cites the following from the GBD: ‘Current lockdown policies are producing devastating effects on short and long-term public health … [including] … worsening cardiovascular disease outcomes, fewer cancer screenings and deteriorating mental health’. CL clips off the next words, namely: ‘—leading to greater excess mortality in years to come’. Contrary to the implication of Clarke and Leigh’s statement that ‘at least in terms of short-run mortality, some of the concerns that led to the Great Barrington Declaration have not eventuated,’ the GBD authors never claimed that there would be short-term excess mortality from lockdowns. Instead, they referred to overall impacts on mortality ‘in years to come,’ based mainly on the non-fatal ‘deterioration’ in short-run health measures.

Third, Clarke and Leigh contradict themselves by first claiming that ‘lockdowns have proved effective in reducing transmission and deaths from COVID-19,’ but then noting that in 2022, many ‘prevented’ COVID deaths occurred. In the authors’ own words, COVID-19 deaths were prevented only ‘in the first 2 years of the pandemic (this changed in 2022, after lockdowns were lifted)’. Thus, it is not clear that lockdowns reduced total COVID-19 deaths among Australian residents. At best, lockdowns delayed the spread of COVID-19 for a short period.

Finally, the authors claim that a resurgence of the virus occurred ‘after lockdowns were lifted’, yet COVID-19 deaths had been occurring in Australia even during lockdowns. The extreme Melbourne lockdown was finally lifted on 22 October 2021, but COVID-19 deaths had been rising well before then. Daily COVID-19 deaths did increase once vaccinated tourists were allowed through Australian borders, but what the postopening path of deaths suggests most clearly is an embarrassing ineffectualness of the COVID-19 vaccines—which were billed as the solution to COVID-19 and lockdowns, because they would allegedly protect Australia from severe COVID-19 consequences. This fuller picture is omitted from CL’s description of the path of COVID-19 deaths.