Facebook leans red, TikTok trends blue, and YouTube is a swing state: Here's how social media usage is linked to the 2020 election

• Social media is well established as a major factor in American politics, but how voter preferences align with different platforms is less understood.
• Preferred social media platforms and how one uses them — such as simply browsing or posting regularly — are indicators of how Americans plan to vote, according to a series of Insider polls taken from August through October.
• Those who say they check Twitter at least daily are 20 percentage points more likely to say they'll vote for Joe Biden than the overall respondent.
• People who post or share links on Facebook at least once a week are the most heavily aligned with President Donald Trump, with that group being 10 percentage points more likely than the average respondent to say they'll vote for his reelection.
• YouTube is the closest platform to a "swing state," with some differences between those who post comments frequently and those who just watch videos.

Both social media preferences and behavior are aligned with who Americans plan on voting for in the 2020 election, according to a series of Insider polls conducted from August through October.

Self-described Twitter and TikTok users lean more toward voting for Democratic nominee Joe Biden, while active Facebook users are more likely to back President Donald Trump.

YouTube is a mixed bag, with those who post or comment more frequently tending to be more likely Trump voters, while those who simply browse the site to watch videos prefer Biden.

People who said they checked Twitter at least once a day had the highest lean toward Biden compared to the overall set of respondents, while Trump performed best with those who said they post or share links on Facebook at least once a week.

Here's a breakdown of the numbers:

This data is from an aggregation of nine polls taken between August 8 and October 12, particularly from respondents who said they were likely to vote. All told, the polls combined to 10,077 respondents, 8,623 of whom indicated they were registered to vote and 8,321 of whom said they would likely do so.

Respondents were asked who they intended to vote for in November as well as a number of other questions. This chart shows the difference between the overall margin of support for Trump or Biden among likely voters as a whole and the margin of support among respondents who said they agreed with the description of their social media usage.

• Twitter users skewed left compared to the overall respondent pool, with those who check it at least once a day being the most likely to say they'll vote for Biden — 20 percentage points more likely than the average respondent.
• Less frequent Twitter users are 12 percentage points more likely to say they'll back Biden at the ballot box.
• Facebook is the red state of social media platforms, with weekly posters 10 percentage points more likely to say they'll vote for Trump and those who check it once a day 5 percentage points more likely.
• TikTok users are more in the Biden column, with both daily and weekly users more likely to identify as voters for the Democratic nominee. This may very well be due to the younger-skewing audience of the platform, and it could shift over time as it expands.
• The polling also found that across all platforms, people who say they post on the sites are less likely to be Biden supporters than those who say they simply browse.
• This phenomenon is best demonstrated on YouTube: weekly commenters are 3 percentage points more likely to say they're behind Trump, while those who just check YouTube daily are 7 percentage points more likely to be in the Biden column. The viewers skew blue, the commenters skew red.

SurveyMonkey Audience polls from a national sample balanced by census data of age and gender. Respondents are incentivized to complete surveys through charitable contributions. Generally speaking, digital polling tends to skew toward people with access to the internet. SurveyMonkey Audience doesn't try to weight its sample based on race or income. Polling data collected 1,150 respondents August 7, 1,106 respondents on August 11, 1,128 respondents August 21-22, 1,073 respondents August 29, 1,161 respondents September 4, 1,107 respondents September 15, 1,017 respondents September 18, 1,122 respondents September 28, 1,176 respondents October 5, and 1,130 respondents on October 12. All polls carried approximately a 3 percentage point margin of error individually.