Read the study, not a press release. It said quite explicitly that cloth masks had no effect, and surgical masks had an extremely minor overall effect (11.3% reduction in symptomatic seroprevalence), with an absolute effect size of 0.09%:

https://www.poverty-action.org/sites/default/files/publicati...

The Bangladesh study does not say that cloth masks have no effect.

It said that it can’t support an impact of cloth masks usage on infection rates, but associates them with reduced severity/symptoms.

That’s a notably distinct proposition from “no effect.”

"Symptoms" are self-reported. Symptomatic seroprevalence is objective.

What this should tell you is that "symptoms" are a bad endpoint for a randomized-controlled trial, not that if the good endpoint doesn't show you what you want, you should rely on the bad one.

11.3% reduction in the base of an exponential process is huge.

Sure, but that is not what the study said, and not at all what the observed difference means.

They found a tiny effect in symptomatic seroprevalence at a single point in time.

If 11.3% fewer people were infected overall up to that point in time, how many people are likely to be infected at T+1? 11.3%. It's not even a trick question.

But again, that's not what the study measured. You're extrapolating incorrectly.

First, the study conflates distancing and mask-wearing. They admit this in the paper (there is a significant increase in distancing amongst the mask wearers). There is also a large change in the size of the populations that could easily swamp the effect size (i.e. the size of the mask population was something like 9% bigger than the control).

Second, just because you observe effect size X at time T does not mean that you will see an equivalent effect at T+1. It also doesn't mean that you will see a compounding effect. In human terms: maybe you get a surge of behavior at the start of the intervention that doesn't continue, due to burnout or non-compliance. Or maybe you got lucky at time T, and randomly saw a population that yielded a result. You can't make the assumption.

But, that's not what you want to measure. You want to measure whether an intervention works if people adhere to the protocol. Again, as I asked you in another comment, would you expect masks to work if people don't wear them?

> that's not what you want to measure. You want to measure whether an intervention works if people adhere to the protocol.

No, you want to measure if an intervention works if you do everything reasonable to enact it. You don't get to invent an alternate reality where people are 100% compliant and there are chocolate rivers and gumdrop trees.

Otherwise, I have the perfect diet plan that will end all obesity, worldwide: don't eat if you're fat. Simple! If it fails, it's because you didn't follow my brilliant advice.

No, these are not public policy studies. These are medical studies. Do they measure the effectiveness of an experimental medication on people who stop taking it? No. If I were to measure things as you propose, I guarantee you I could conduct a huge study showing that exercise doesn't make people healthier, just because most people won't adhere to an exercise program. That has no bearing on the effectiveness of the intervention.

The point of these studies is to show people that the intervention works if people do it. And, yes, there are public policy studies showing that mask mandates don't work, but that doesn't mean they don't work for the people actually wearing them. It just means people who refuse are cutting their noses off to spite their faces. As noted in another comment, this isn't the beginning of the Third Reich, and should not be controversial at all.

All drug studies are public policy studies.

If you show that a drug works in a petri dish, that's completely irrelevant if it's so toxic that people don't take it because of side-effects.

If you show that full-face respirators work on a mannequin in a box, that's completely irrelevant if real-world people don't wear full-face respirators because they suck.

> If you show that a drug works in a petri dish, that's completely irrelevant if it's so toxic that people don't take it.

> If you show that full-face respirators work on a mannequin in a box, that's completely irrelevant if real-world people don't wear full-face respirators because they suck.

And these examples are straw men. That's completely irrelevant to the point.

I just noticed: my comment is missing the word "fewer". It should have read "how many fewer people are likely to be infected at T+1?"

Why is poverty-action.org more reputable than Stanford?

This is the study the Stanford press release was talking about.

In a game of telephone, the further you get from the original source the less reliable the information becomes.

https://en.wikipedia.org/wiki/Game_of_telephone

I would take any study performed during the heat of the moment with a huge grain of salt. Especially in an environment where saying anything but “masks work” gets you labeled as some wacko spreading misinformation.