The best evidence I've seen that Bernie Sanders's political revolution might be possible

Bernie Sanders's presidential campaign is based on a simple theory: There is a reserve army of liberal voters who've sat out past elections but who stand ready to support a more stridently leftist Democratic nominee.

By getting these historic nonvoters to turn out, Sanders claims, he could win the general election, maybe take back the House and Senate, and have an organized public ready to pressure Congress to pass a democratic socialist agenda.

So far, this idea of a leftist political revolution has been widely dismissed as implausible by many liberal commentators — and I share a large part of their skepticism. But new research by Stanford political scientists Simon Jackman and Bradley Spahn has convinced me that at least one big part of it is correct: There really is a reasonably large segment of the American population that most political campaigns aren't reaching.

It's a segment that's disproportionately black and Latino and decidedly more liberal than the American public as a whole. If they were turning out, it could conceivably push the American electorate to the left.

It's getting them to the polls that's the hard part.

How campaign lists silence poor, liberal Americans

Jackman and Spahn's key insight is that campaign contacts with voters usually spring from those voters' presence in voter databases sold by companies like Catalist (for Democrats), Data Trust (for Republicans), TargetSmart (also for Democrats), and L2. If people aren't in those databases, they aren't targeted by campaigns.

Jackman and Spahn confirm this by matching up voter databases with respondents to the American National Election Survey (ANES), a widely cited in-person survey of about 2,000 voting-age people conducted every election cycle.

They found that you can break down those people into four categories:

• 70 percent are registered to vote and listed in voting databases.
• 7 percent were on the lists but unregistered to vote.
• 12 percent were mislisted: They were registered or on a commercial database, but at a different address from the one in the database or voter registry.
• 11 percent were unlisted: not in the commercial databases.

Campaigns basically only contact the 70 percent of voters in the first category:

But they are very different, both demographically and politically, from unregistered, mislisted, or unlisted voters. The latter are significantly more likely to say that the US should increase federal welfare spending, and mislisted and unlisted voters overwhelmingly supported Barack Obama in 2012 over Mitt Romney, by much bigger margins than among registered voters:

They're much younger, poorer, and likelier to be uninsured than registered voters as well, with unlisted people boasting average incomes some $30,000 below those of registered voters:

Mislisted and unlisted voters are also disproportionately black and Latino, relative to registered/listed voters:

So the people campaigns reach out to are "whiter, older, wealthier and more conservative than the citizenry," Jackman and Spahn conclude. "It seems clear that the well-off and the already-powerful are the beneficiaries of this new political institution."

The issue: Turning them out is hard and expensive

The key for Sanders's political revolution to succeed, then, is to find some way to mobilize the 23 percent of voters who are mislisted or unlisted. But though Jackman and Spahn's research confirms that this group exists, and is more liberal than America as a whole, it's unclear if campaigns could cost-effectively reach them and persuade them to vote.

A recent study by Notre Dame's David Nickerson makes this point well. Between 2004 and 2007, he conducted field experiments in six cities (Detroit, Tampa, Kalamazoo, Denver, Memphis, and Louisville) meant to reach previously unregistered voters and sign them up.

In each case, a local nonpartisan group hired canvassers to walk down certain streets and knock on every door. If every resident in the house was registered, the canvasser moved along. If someone was unregistered, the canvasser would help them register. If no one answered, the canvasser would do another sweep or two to try to reach them. Nickerson assigned "treatment" streets to be canvassed and compared them with nearby "control" streets that didn't get canvassed. In three of the experiments, he also made sure that wealthy, middle-class, and low-income streets were all canvassed, so he could compare effectiveness among them.

The canvasses did increase the number of registered voters — but not by a lot. On average, each treated street led to 10 newly registered people, a 4.4 percent increase. But only two of those people would go on to vote. The increase in registration was greater on poor streets, and the increase in turnout was greater on rich streets.

"Registering likely supporters is not a cost effective activity for campaigns to pursue," Nickerson concludes. "A dedicated registration effort would cost roughly $60 per vote. … This sum is substantially higher than $13 for phone calls and $21 for canvassing or leafleting [registered voters]."

Moreover, campaigns are likely to be less competent about it than the nonpartisan groups Nickerson worked with. A paper last year by Harvard's Ryan Enos and Yale's Eitan Hersh looking at the Obama 2012 campaign found that "individuals who were interacting with swing voters on the campaign’s behalf were demographically unrepresentative, ideologically extreme, cared about atypical issues, and misunderstood the voters’ priorities."

That might be less true with liberal nonvoters than moderate swing voters, but it suggests that maintaining message discipline when attempting a canvass-based mass mobilization of poor voters would be pretty difficult.

One thing that could mobilize this population is mandatory voting. In Australia, this has had the kind of effect that Sanders's political revolution is attempting to have: The country's compulsory voting policy has boosted the vote percentage of the left-of-center Australian Labor Party by anywhere from 5 to as much as 7 to 10 points.

But short of a drastic, politically implausible solution like that, it's hard to see anyone successfully mobilizing the lumpenproletariat Sanders is pinning his hopes on.