UMaine study finds connection between partisan politics and effective face mask production during pandemic

Face masks have been an indelible part of the COVID-19 pandemic. Examining how, where and of what face masks were made can provide insight to how the personal protective equipment impacted — or was impacted by — the society around it. A study published by new faculty Matthew and Natalia Magnani in the Department of Anthropology at the University of Maine found that specifics of homemade face mask production were different depending on a state’s political affiliation and changed throughout the country with increased political polarization. To study these connections, they used an innovative anthropological tool: big data.

Anthropologists, archaeologists and other researchers interested in studying material objects and their roles in reflecting or shaping the world around them have long excavated, collected or physically observed materials. However, such scholars have yet to take full advantage of large, complex digital datasets afforded by the internet, which can provide an even more complete picture of the material world with more samples than physical collections.

To show the potential of online data sets and how they could transform the understanding of the material world, a group of researchers led by Matthew Magnani studied the production of face masks across the United States in 2020 and 2021, during the height of the COVID-19 pandemic.

“Significant interest in the social meanings of mask use was apparent early on in the pandemic. Our team — including Jon Clindaniel from University of Chicago, and Natalia Magnani also starting at UMaine — wondered how we could develop new anthropological tools to consider the significance of their production. We wanted to hone an approach that would allow us to look across the entire United States, and capture changes in mask making over time,” Matthew Magnani says.

The researchers used Alura, a market analysis application specifically designed to analyze craft sales from the website Etsy. The software gathered data about the characteristics of the face masks sold by Etsy users, like where the sellers were located, what material the masks were made of and what other tags the seller included to describe the product. The researchers looked at this information at multimonth intervals to analyze the production of face masks across the country. They coupled these data with an analysis of state mask policies and how mask wearing became politicized over time. 

The results showed clear linkages between the changing nature of the production of face masks and partisan politics, particularly when it comes to materials that decrease the face masks disease mitigation effectiveness. For example, masks made in states where electoral votes were counted in favor of Democrat Joseph Biden in the presidential election were more likely to boast characteristics and tags that touted their safety, while states that went for Republican Donald Trump did so at a lower rate. The study is careful to highlight that periods of increased polarization reduced apparent mask efficacy across the board, independent of political leaning.

Low rates of effective mask making sometimes co-occurred with more relaxed public health measures. For example, just 38% of masks made in South Dakota were found to be associated with disease-mitigating attributes; South Dakota, notably, was one of the few states without a mask mandate throughout the pandemic. In other cases, there were outliers — Nebraska had no mask mandate, and yet followed only Colorado in masks produced evoking functional vocabulary — but in general, politics had implications for how well a state’s personal protective equipment was produced. 

Unlike the rates of overall mask efficacy, however, the researchers found no correlation between political affiliation and the production of intentionally ineffective masks — for example, those that were made of only a single layer or mesh, lace or other breathable fabric. Production of such masks were more associated with political events of the time. For example, Republican states where the data showed no intentionally ineffective masks were produced — Alaska, Missouri, Nebraska and North Dakota — didn’t implement statewide mask mandates, suggesting that intentionally ineffective masks could not be used as an effective form of political dissent in these places.

In a similar vein, “antimasks,” those that complied with mask mandates but contained protest messages, showed no clear political affiliation in the 15 states with sellers that produced them. They also represented a relatively small portion of the total masks produced, but the number of states producing them did increase over time as masking became more politically polarized.

Nationally, increasing political polarization led to an overall decrease in the effectiveness of masks produced. As the election cycle was in full swing, the data revealed plummeting efficacy in masks produced across the country into November 2020, going as low as 51%. However, that efficacy bounced back to 68% by May 2021 after a few months of the Biden administration. 

The differences in red and blue states were nuanced, and the researchers found that looking at these changes on a fine temporal scale was important to understanding them. For example, while Republican states tended to produce less effective masks on average over the course of the study, the analysis of changes over time in Democrat-held regions revealed that the politicization also negatively impacted effective mask production across the country. Time-averaged assemblages of this data would have drowned out the meaning.

“Our study demonstrates the deleterious effects of political polarization for public health as they manifest through the production of personal protective equipment, suggesting that divisive partisan rhetoric led to the manufacture of physically less effective masks across the country,” Magnani says.

The results demonstrate how effective mining internet data can be in understanding the changing distribution and social significance of material culture. Going forward, the researchers hope that the tools used for this method will be refined to make them more efficient and precise. 

“We are contributing to a new trend in the study of material culture — all the stuff that surrounds us — using big data. These studies have the potential to condense years, or even lifetimes of fieldwork and data collection into months of research and a few clicks of a mouse. Moving forward, we will apply these methods to rethink the way we look at the things that surround us on an unprecedented scale, from the hundreds of millions of objects sitting in museum collections, to the Amazon delivery boxes at the front door,” Magnani says. 

A FirstView of the study is available on the Cambridge University Press website. The full study will be published in October 2022 in the journal American Antiquity.

Contact: Sam Schipani,