COVID and the case for non-pharmaceutical interventions

The use of non-pharmaceutical interventions has been a source of persistent controversy during the COVID pandemic.  Opposition to NPIs was the motivating impulse behind The Great Barrington Declaration (GBD), which called for an immediate end all “lockdowns” and the use of “focused protection” to “protect the vulnerable”.  The GBD was trumpeted by the American Institute for Economic Research, a previously little-known organization that “educates people on the value of personal freedom, free enterprise, property rights, limited government and sound money”, and then amplified through new and existing libertarian think tanks and blogs. 

Here I want to examine three arguments used by opponents of lockdowns that have the potential to lead us astray as we think about lessons learned and planning for future pandemics.  These arguments are: 

  1. there was a pre-COVID expert consensus against non-pharmaceutical interventions (NPIs, “lockdowns”),
  2. our experience with COVID has shown that NPIs are ineffective, and thus totally unjustifiable on cost/benefit grounds, validating the pre-COVID consensus against NPIs, and
  3. because the case against NPIs is so strong, we should foreswear the use of NPIs in future pandemics.

These claims are all false and potentially dangerous.

What was the pre-COVID thinking on NPIs?

Opponents of NPIs have repeatedly claimed that there was a pre-COVID consensus against lockdowns.  See here, here, here, here, here, here, here, here, here.  This argument is rhetorically attractive to the GBDers and their allies because it allows them to claim the mantle of science for their preferred policy without the bother of producing evidence about the costs and benefits of different NPI policy options. 

This argument is misleading.  There certainly were people who argued against the use of most NPIs, and some government plans for pandemic response argued against general masking, social distancing, and school closures (e.g., the UK here).  On the other hand, some scholars argued that NPIs might be effective and should be considered in the event of a major pandemic.  One thing that most did agree on was the lack of a strong evidence base concerning the effectiveness of NPIs.    

Perhaps the most frequently cited evidence for an alleged consensus against NPIs is this WHO report on pandemic influenza from 2019.  This report argues that the evidence for the effectiveness of NPIs is generally weak, but there is still reason to think some NPIs may be helpful at slowing a flu epidemic in at least some circumstances.  WHO conditionally recommended mask wearing by asymptomatic individuals, improved ventilation, school closures, workplace measures (including possibly closures), and limiting crowds, often based more on mechanistic plausibility than empirical studies of policy effectiveness. 

It is true that the WHO report recommended against contact tracing and quarantine of exposed individuals, despite some low-quality evidence of effectiveness, due to concerns about infecting other household members and the financial impacts of missing work, and the unsubstantiated claim that resources could be used more effectively elsewhere.  Given that the report acknowledges that delaying cases can be valuable in some circumstances, and that contact tracing and quarantine can delay cases, this document hardly provided strong reasons to reject contact tracing and quarantine of exposed individuals across the board in 2019.  In any event, we now need to update our prior beliefs about NPIs in light of our experience with COVID-19. 

What have we learned from COVID-19?

During the COVID-19 epidemic many countries implemented NPIs of various sorts.  Opponents of lockdowns claim studies show that NPIs were ineffective at reducing the prevalence of COVID-19.  This is misleading. 

There are real difficulties with studies that try to measure the effects of “lockdowns” using variation in policy and mobility and COVID outcomes across localities.  Measurement error is severe for both outcome variables and inputs, policies vary from place to place in their form and degree of enforcement, the disease process is incredibly noisy, it is difficult to disentangle the effects of voluntary distancing from the effects of NPIs and lockdowns, etc.  Results are also sensitive to changes in model specification. 

These statistical challenges do not show that lockdowns are ineffective.  They just highlight the well-known difficulty of identifying causal pathways with very poor data and a very complex and noisy disease process.  Critics of lockdowns bootstrap these empirical difficulties into an argument that NPIs do not work by conflating absence of evidence that NPIs are effective with evidence of that NPIs are not effective (see p. 43 here).

The methodological and data problems that arise when we study NPIs are important, but we should not let a technical debate over statistical inference divert our attention away from the bigger picture.  Our experience with COVID-19 has taught us that social distancing can be very effective at limiting the spread of a pandemic respiratory virus, even if it is unclear if voluntary behavior changes alone can be effective or legal restrictions are needed.  We have also learned that stringent NPIs can keep the prevalence of disease low for an extended period.  

This is important new information.  Prior to 2020, many were skeptical that social distancing and NPIs could control a flu epidemic.  In fact, the measures put in place to fight COVID-19 crushed the flu worldwide.  This should inform our planning for future pandemics.  Social distancing – whether caused by government orders or voluntary – also may have played a role in bringing COVID-19 spikes under control.  (With the possible exception of the recent Omicron wave, it seems doubtful that the rapid declines in cases we have seen following many surges was due to reaching herd immunity or weather changes.)

The evidence that stringent NPIs can control disease is clearest in places like New Zealand and Australia, which successfully brought outbreaks of the highly contagious Delta variant under control through lockdowns.  South Korea kept cases quite low for two years.  If you believe the official statistics, China has managed to contain the hyper-contagious Omicron variant through harsh lockdowns.  In these places it is implausible to think that voluntary distancing would have worked to keep the virus in check.  The prevalence of disease was simply too low to induce enough voluntary distancing to prevent exponential growth in disease.

I am not claiming that the COVID-19 restrictions in these countries were justified.  We can debate which NPIs were worthwhile and which were not.  (I personally believe that China’s war on Omicron is likely to fail, and that it is inhumane and frankly insane, and an excellent illustration of the value of democratic political institutions.)  My point is simply that we now have incontrovertible evidence from many countries that even a highly infectious respiratory virus can sometimes be controlled through a combination of voluntary distancing and NPIs.  This should inform our thinking going forward.

Should NPIs be off the table in pandemic planning?

Some GBD-aligned thinkers say that we should never use NPIs in pandemic management, and that use of NPIs should not be included in pandemic planning.  For example, the WHO recently published an interim document on updated pandemic guidance that briefly mentioned NPIs.  In response, Will Jones, editor of The Daily Sceptic, wrote an article arguing that this is “alarming” and a “disaster” because we know from previous WHO guidance that efforts to curb the spread of a respiratory virus “will not be possible and will waste valuable resources”.  You can find similar calls for a bright-line rule against lockdowns here and here.

This position is extreme and hard to understand.  Of course NPIs should be part of our pandemic control toolkit, and we should think hard – in advance – about when they are justified and how to implement them in the most effective and least disruptive way.  Pre-pandemic planning must include NPIs. 

Again, this doesn’t mean that all COVID NPIs were justified, or that we should indiscriminately lockdown the moment someone dies from a new strain of flu or a novel coronavirus.  But every pandemic is different, and it is easy to imagine cases in which different NPIs should be on the table. 

The next pandemic pathogen might be much more lethal than COVID, which would increase the benefits from NPIs.  It might spread through droplets that are effectively curtailed by masks.  It might spread primarily through young children and be containable through school closures.  Handwashing and surface cleaning is not important for COVID, but it may well be useful in a flu pandemic.

Just as important, if we focus on the future rather than re-litigating the past, technological progress will continue to make NPIs more effective and less costly.  For example, more effective and cheaper masks will become available, remote work will become a more feasible option for more and more workers, and remote schooling will improve.  Even more important, faster development of vaccines and anti-virals seems highly likely.  The faster vaccines and treatments can be developed, the stronger the case for stringent NPIs to keep viral spread in check at the beginning of a pandemic.

The bottom line is this:  There was no pre-pandemic consensus against NPIs, and experience with COVID shows that NPIs can be much more effective against respiratory viruses than we believed in 2019.  NPIs can clearly play a role in curbing pandemics, and we need to study them to figure out when they are likely to be effective and how they can be implemented in the least disruptive way possible.