How can we fix bias in email spam filtering? Look to the Post Office.

Eric Wilson is the managing partner of Startup Caucus, a Republican campaign technology incubator.

Not all spam is created equal, it turns out: A recently published academic study, which analyzed more than 300,000 political emails sent during the 2020 election, found evidence of a political bias in how the nation’s most popular email inbox providers filter messages. Google’s Gmail was 50% more likely to designate emails from Republicans as spam than messages from Democrats. And on the flip side, both Microsoft Outlook and Yahoo were significantly more likely to filter out messages from Democrats than from Republicans.

Error-prone filtering from the most popular email service in the country may have a major impact on elections in the United States. Online fundraising relies primarily on email solicitations, and the Republican National Committee estimates it may have lost up to $2 billion in contributions since 2019 because fundraising messages were sent to spam filters. Beyond fundraising, campaigns are also hampered in their ability to combat disinformation, share accurate information about voting and encourage supporters to go to the polls.

Google said that correlation is not evidence of causation and that “political affiliation has absolutely no bearing on mail classifications in Gmail.” Microsoft likewise chalks the discrepancy up to technology that isn’t perfect.

Sorting algorithms are notoriously a black box, and in this case for some good reasons — if any rules were made public, they would be immediately circumvented. But because these companies do not make their spam filtering algorithms available for scrutiny, we cannot know for certain whether or not some political factors — intentionally or otherwise — contribute to an email’s classification. But we can agree that if a voter opts in to receive email messages from a political candidate or party, they should be able to receive them.

Inbox providers have these spam-filtering measures in place to protect their users from unsolicited messages, scams and other malicious emails. Federal law, however, exempts political messages from the definition of spam — although best practices dictate that most campaign email marketers abide by the provisions of the law. The mere fact that political emails are routinely delivered to spam folders is troubling giving the priority we place on First Amendment protections.

Somewhat surprisingly, we can look to the United States Postal Service for a solution to the political bias of spam-filtering algorithms. Campaign-related postal mail from candidates and political parties receives priority and expedited handling with a special “Tag 57” — also known as a red tag because of its color — attached to it as it moves through the system. Post office employees know the red-tagged mail should not be delayed.

A similar system of identification already exists for email, but has yet to be applied to political campaigns. There are a handful of open standards built on DNS records that can verify the identity of email senders and their ownership of the domain they are using. Email inbox providers such as Google, Microsoft and Yahoo rely on these records to authenticate the messages their users receive and protect them from cyber scams like phishing and spoofing. Email marketers know that authentication with DNS records helps improve the likelihood that their messages make it past spam filters and into inboxes.

To address the challenge of political bias in email filtering, we need to create a new type of authentication that identifies messages from political candidates and parties to email inbox providers. This will ensure that subscribers see the messages they are signing up for and that political communications are not treated the same way as commercial messages.

A nonpartisan coalition of industry and political representatives can develop a new, open standard that would be made available to qualified candidates and parties. Platforms like Facebook and Google already have systems in place that verify the identities of authorized representatives for political advertising purposes. Eligible campaigns would update their DNS records and inbox providers would ensure that emails from the authenticated domains would reach the inbox of the intended recipients.

This system should not make political campaigns immune from email marketing best practices. The privileged delivery status should only be for communications that recipients have opted in to, and recipients should always have the ability to unsubscribe.

The quality and relevance of campaign emails would increase with this new standard, as candidates and parties wouldn’t have to rely on gimmicks or hyperbole to convert the few supporters who rescue political emails from their spam folder. Campaigns would also be more secure, as this verification system can be used to prevent cyberattacks from bad actors as well.

Email is foundational to the internet and, as such, is an important channel for communicating with voters. Campaigns face strong headwinds when educating voters about elections. Our media landscape is more fragmented than ever, local news is on the decline and some major online platforms even ban all paid political advertising. Protecting opted-in political email communication can be accomplished, and will buttress civic engagement.