Overview

An amendment to limit voting rights was proposed to the Minnesota constitution in 2012. With many similar proposed laws popping up across the US and heavy support in Minnesota, a coalition of Minnesota organizations set out to convince voters to defeat the amendment. Concurrently, an effort to defeat an anti-gay marriage constitutional amendment was underway, and swallowed resources and attention that the voter ID amendment may have otherwise gotten. In May 2012, a poll showed 80% of Minnesota voters in support of the amendment.

What are the resources?

The main thrust of this organizing campaign was how to strategically organize volunteers, voters, and money to produce a simple majority against the amendment.

The campaign carefully chose voters to persuade and aggressively recruited volunteers to reach more voters. Initially, the campaign used the public voter file and roughly categorized voters based on their voting history. Factors like party registration and likelihood to go to the polls played into decisions about resource selection, usually called list segmentation in this context.

Volunteers were tasked with moving voters who were pro-amendment but likely to jump categories, and to then document whether the interaction yielded movement on the 1-5 scale.

Why are the resources being organized?

The voter-volunteer interaction is designed to produce data about voters to be folded into algorithms that produce subsequent call lists.

The persuasion interactions turned the voter file into increasingly meaningful data with growing granularity. The voters that could initially be thrown into only 5 categories could be segmented much further: towards the end of the campaign, we could create a list of all registered Democrat veterans living in zip 55407 with valid phone numbers with whom we’d had at least two conversations. Conversations with voters enriched allowed for a sharpening of volunteer efforts towards more strategic persuasion.

How much are the resources organized?

Voter data was organized iteratively throughout the campaign as more and more data came in.

Volunteers noted additional descriptions such as veteran, having tribal ID, disabled, or unable to renew driver’s license, or otherwise personally affected by the amendment. Sometimes this identity “tagging” was used to match volunteers of a particular identity to similar voters for more effective conversations.

During the campaign, all volunteer-voter interactions were designed to describe voters in greater detail, which helped automatically filter out voters not receptive to conversations about voter identification. Based on certain behaviors and outcomes during volunteer-voter interactions, some voters were labeled Do Not Call, Wrong Number, or Deceased, handily eliminating them from future lists.

After the election, the campaign data could be used again, but the archived data was all that remained. The volunteers and voters don’t persist as campaign resources once votes are cast.

When are the resources organized?

Volunteers were trained in rapid voter classification before door-knocking or calling. Based on responses to the script, volunteers categorized voters with whom they interacted with a ranked resource description. Voters were given a number between 1-5 depending on whether they were likely to vote for or against the amendment, or somewhere in the middle. Their rankings were kept on file and those ranked anywhere from 2-4 received repeat calls during the course of the campaign to gauge the effectiveness of the persuasion tactics. The campaign decided to save resources by deeming anyone who was strongly in favor of the amendment as “immovable”.

Removing the extreme voters made sense at the beginning when the odds of beating the amendment were low and limited resources had to be used extremely frugally. The cost-benefit analysis changed once the campaign garnered more attention, and the campaign shifted to spend resources to reach these extreme voters: the supportive ones to ensure that they hadn’t been swayed by the opposition, and the opposing ones to begin the hard task of swaying them.

Who does the organizing?

The coalition of organizations behind the vote no efforts orchestrated the campaign’s organizing system.

The slate of volunteers working on the campaign grew in granularity as the campaign moved forward. Volunteers specialized by taking on particular roles within the campaign and called or visited voters within their district of Minnesota. Some climbed the hierarchy of leadership. As the campaign progressed, more categories of leaders were added to the leadership structure.

Other considerations

Because this organizing system is all about people organizing people, resource description is particularly difficult and subject to bias. All the “data” that results from persuasion interactions is necessarily fuzzy and one 4 is not necessarily equal to another 4 in likelihood to vote a particular way. However, a no vote from one person is interchangeable with a no vote from another. Conducting effective conversations with the right individuals has the potential to yield scores or hundreds of no votes because of the social capital and persuasive power of some individuals.

Interestingly, if a volunteer reached a wrong number with a receptive human at the end of it, they were advised not to spend time persuading them. Because the lists are produced ahead of time and individual volunteers didn’t have access to the voter file, they couldn’t match a receptive ear on the other end to a particular voter. The campaign couldn’t rely on conversations that weren’t documented.

Furthermore, it would be a problem of authentication even if they had access: how would they prove the voter’s identity, or resolve the name problem of many potential voters by the same name? Some volunteers flouted this advice and tried to convince the unknown individual to vote against the amendment anyway. Whether these undocumentable conversations resulted in wasted time is unknown, but in the end, voter ID was defeated.

Automatic description at this relatively small scale is impossible due to scarcity of financial resources: it took the massive efforts of many to gauge the political sentiments of others and make predictions about future behavior. At the scales of national elections, mountains of data and data scientists made it possible for the Democratic party to segment the voting public, test message effectiveness, and conduct remarkably successful fundraising, $5 at a time.