Using CrowdTangle, Media Matters compiled a list of public Facebook pages that posted content using one of 21 keywords describing or referring to political topics, including the names of U.S. politicians, government officials, and government agencies, or the general state of U.S. politics from January 1, 2021, to December 31, 2022. Such pages were associated with news and media outlets, media figures, politicians, political parties, and issue-focused advocacy groups, among others. 

Of the over 4 million posts including the political keywords, Media Matters isolated the posts that made up the top 95% of the the total interactions for all political posts from 2021 and 2022. There were 2,276 pages that met this threshold and were subject to further analysis.

Three researchers independently assessed these 2,276 pages for two factors. They determined: (1) if the page targeted a U.S. political audience and (2) the page’s ideological alignment (left-leaning, right-leaning, or ideologically nonaligned). Each page was reviewed individually and given a final code if two of the three researchers independently awarded it the same code. Pages that did not achieve this level of consensus were reviewed again individually by two additional coders who then reconciled discrepancies.   

To be considered as targeting a U.S. political audience, a page had to be written in English and focus primarily on the politics of the United States rather than as a topic among other international news. We also excluded pages that clearly or explicitly focused on entertainment or lifestyle content without political connotations.

We determined the ideological alignment of a page first by self-identification: Looking to see if an ideology was clearly stated in a page’s title, about section, profile, or header pictures. 

We coded pages as left-leaning if they identified as Democrat, liberal, left or left-leaning, pro-Biden, pro-Obama, pro-choice, pro-racial equality and justice, or pro-environment. We coded pages as right-leaning if they identified as Republican, conservative, right or right-leaning, pro-Trump, pro-Second Amendment rights, pro-life, pro-Confederacy, or libertarian. Pages that explicitly identified as being against one of the positions were coded as the opposite ideology. For example, anti-Biden pages were coded as right-leaning.

We coded pages as ideologically nonaligned if they represented legacy media, local media, or government agencies and also displayed no clear or self-identified ideological characteristics from the lists above. Pages for U.S. embassies abroad were removed from the sample due to an overall lack of content focused explicitly on U.S. politics. We relied on a Pew Research Center definition of legacy media as any news organization that was not “born on the web,” including print newspapers and television and radio broadcasting organizations.

All pages that did not explicitly state one of the above positions were analyzed and classified using the following criteria:

• If a page did not self-identify its ideology but was owned by an entity that did outwardly express an ideology (according to the same guidelines), the page in question was coded based on the owner's additional pages.
• If the page’s most recent 3-5 posts clearly expressed an ideology, the page was coded accordingly.
• If a page’s 10 most recent posts with links shared content from three or more media outlets or pages that self-identified as a particular ideology, the page was coded accordingly. Commentary on these links were considered in conjunction. For example, if a page consistently shared an outlet to criticize its content, the page in question was coded with the opposite ideology.
• If a page was affiliated with a known right- or left-leaning media outlet, including that outlet’s journalists or personalities, the page was coded accordingly. Subsidiaries of these known outlets were also coded based on their parent outlet. 

Any page that could not be coded using any of the criteria or was not coded in the same way by a majority of the coders was not included in the final sample. The resulting list excluded 790 pages, accounting for only 16% of the posts and 10% of all interactions in the original political search. The final lists included 1,486 pages, each double-coded by ideology.