Two years ago, the CDC reasonably suggested that the general public wear masks, as they may be protective against Covid, and it was thought (and now known) that Covid spreads asymptomatically. While studies regarding influenza had not been clear on whether public use of masks would help, people generally spread the flu only after symptoms, so the inconvenience of wearing a mask could be avoided if symptomatic sick people just stayed home. This is not the case with Covid.
So the CDC made the mask recommendation, but of course it was early in the pandemic and no studies had been done on the effectiveness of mask-wearing.
That changed relatively quickly. The CDC's morbidity and mortality weekly (I love that name!) published a case-control study of Covid transmission (See this link: https://www.cdc.gov/mmwr/volumes/69/wr/mm6936a5.htm?s_cid=mm6936a5_w ) in the late summer of 2020. A case-control study is a type of study where people with a disease are matched with similar people (age, overall health, etc.) without the disease, to see what may have caused, or at least be correlated with, getting the disease. It's a common way to deal with the fact that randomized studies are sometimes difficult (e.g., we cannot tell people to smoke or not smoke at random, and then see if they get cancer). For this study, patients who had been to health clinics for Covid-like illness who did not have Covid were compared to people who went to those clinics who tested positive for Covid.
The study looked at a number of things, and found that mask use was not a statistically significant factor in getting Covid. It was not a big study (250 people total) but it was big enough to find that other activities had a statistically significant correlation with getting Covid; e.g., going to restaurants, going to bars where people were not social distancing or observing precautions, and having a close contact who had Covid. Keep in mind that lack of statistical significance does not mean there is no effect -- it just means that the natural variation is large enough and the observed effect is small enough that it could be reasonably attributed to chance.
Shortly thereafter, a second study of close contacts of people with Covid (published in Lancet: https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30833-1/fulltext ) indicated that masks were not a statistically significant factor for transmission whether worn by the person with Covid or the close contact of that person. Once again, it was not a large study, but many other factors were found to be statistically significant (e.g., speaking to the infected person for more than 30 minutes, sharing a car, using the same bathroom, and sharing a meal with the infected person).
In late 2020, a third study came out (published in March 2021 in the Annals of Internal Medicine: see this link: https://www.acpjournals.org/doi/10.7326/M20-6817). Unlike the above two studies, this one specifically looked at mask-wearing, and was a randomized control trial--the researchers randomly asked people to wear or not wear a mask, and then observed whether they got Covid or not. The study found no statistically significant difference in Covid rates (1.8% vs 2.1%) in mask vs non-mask-wearing participants.
The fourth study, from Bangladesh, came out more recently (December 2021): doi: 10.1126/science.abi9069. This was a randomized study and is the only one of the four that shows a statistically significant effect of mask wearing. However, that effect was only found for surgical masks (not cloth). The effect of surgical mask-wearing was found to be quite small: an 11% reduction in infection rates (this despite an almost 4-fold increase in mask-wearing in those villages that were given masks and encouraged to use them properly). The study also finds that people wearing the masks had a small but statistically significant increase in their social distancing behavior. This effect, called a confounding effect, makes it impossible to know whether the small reduction in Covid infections in the mask-wearing group was due to the masks or due to the increased social distancing, or to some combination of both.
So, of four decent studies involving mask use, 3 show no statistically significant effect and one shows no statistically significant effect for cloth masks but a small but statistically significant effect for surgical masks (with the caveat that the study is not adjusted for a confounding effect of increased social distancing among mask wearers).
So what data does the CDC site provide to support its recommendation to use masks?
The CDC site has a bullet-point list of 9 human studies that it highlights to support its recommendation. The first is the Bangladesh study described above.
I reviewed each of the other 8 studies and they:
a) have no control group at all (i.e., not a real study, because they do not correct for Covid rates of those not wearing masks). For example, the CDC cites a "study" consisting of the experience of two masked hairdressers who had Covid. The CDC notes that these two hairdressers interacted with a large number of clients without transmitting Covid to any of them; and/or
b) fail to control for key factors (like vaccination) known to have a substantial effect on transmission.
A recent article in The Atlantic reviews the deficiencies of several of these studies in detail (https://www.theatlantic.com/ideas/archive/2022/01/kids-masks-schools-weak-science/621133/ ). Additional reviews of these and other studies, showing masks are not effective (mostly focusing on schools) is also provided on the urgency of normal website.
But what about N-95 masks?
Recently, there has been some acknowledgement that cloth and surgical masks are not very effective. As a result, some organizations and institutions are beginning to require N-95 masks. However, public use of these masks have not been studied at all in the context of preventing Covid. It is unscientific and bad public policy to promote an untested measure, especially given that similar measures have been shown to be ineffective. It's akin to promoting a stronger version of a drug without studying it when a study of the original version failed to show it worked.
What does all this add up to? Mask use by the public has little or no effect on Covid transmission. The CDC recommendation for mask use by the public is not supported by the science.