At this point it is well-known that widespread use of face masks indoors is an effective way of preventing spread of the Covid-19 disease1. One might naturally assume, based on that, that putting in place laws or regulations requiring the use of face masks would be effective at preventing spread of the Covid-19 disease. Not so, it appears!
A new paper (pre-print, so YMMV) argues that mask wearing prevents Covid-19 spread, but mask mandates do not [MedRxiv]. Why? Because all else being equal, it does not appear that mandates are effective at driving additional mask wearing. Tyler Cowen snarkily called it, “the best mask-wearing study so far”.
I may not go so far, but it reveals an important and uncomfortable truth: governments have extensive power in many ways, but very limited abilities to do the type of social engineering that effective infectious disease control may require.
Compliance Matters
The highest-level finding is reflected in the title of the paper, “Mass mask-wearing notably reduces COVID-19 transmission”. There are evidently no bonus points for clever titles in the computational epidemiology world. But the finding is convincing and the effect sizes are very large: when everyone wears masks in public it seems to reduce spread by about 25%:
This is perhaps not surprising - it has become the conventional wisdom that masks prevent Covid-19 spread. What is more surprising is their look at mask mandates, which appear to have no relationship with the spread of Covid-19! The reason is because mask mandates do not appear to cause increase in mask-wearing, which in most places seems to occur well before the mandates are actually put in place.
The key insight here is that rather than the mandate causing adoption of masks, the causation might run the other way: widespread acceptance of masking provides political “cover” for governments to mandate their use. That is more intuitive when you consider the states that do not appear in this dataset because they never instituted a mask mandate (“censored data” in statistics-speak). Those are states like Tennessee or South Dakota [AARP] where it may not surprise you to learn that many people never wore masks at all [NYT].
So when the authors attempt to actually disentangle the effects of mask usage vs. mask policy - a complex topic I will not attempt to cover here - they find that all of the Covid prevention comes from the adoption of masks. Since the mask mandate does not appear to drive mask adoption, it has little effect on Covid transmission.
Policy doesn’t create policy impact
I love this finding because it’s a powerful reminder that decisions and policies are not self-executing. It’s possible to take a clear insight: “masks prevent Covid transmission”, make the obvious inference that “making people wear masks will prevent Covid transmission”, and then in turn that “we should pass a law making people wear masks”. It’s the clearest, most obvious, no-brainer, simple policy evaluation in the world and yet it appeared to have had no detectable real-world effect.
This phenomenon popped up all over the place during the pandemic, on both sides of the policy fight2. Mask mandates may not have had much of an effect, but neither did aggressive attempts to encourage people to go out and mingle. Jurisdictions like Florida, where the governor was aggressive about pushing people to behave as normal, still saw declines in business activity similar to those in more restrictive polities [Track the Recovery]. The reason? People stayed home in order to not get sick!
All too often, people talk about decisions and policies (in government or business) as though setting forward a law or making a decision will make change happen. This paper is a good reminder that change never happens from words on a page, but from people changing their behavior. One may cause the other, but there’s no guarantee they have to go together.
Like an “ATM Machine”, I’m aware that “Covid-19 disease” is redundant. Well, here we are.
I would have a hard time articulating what exactly the sides were.