A tale of three campaigns

In 1980, a citizen petition was passed under the initiative and referendum law restricting the ability of towns in Massachusetts to increase property tax rates. In the almost quarter-century since then, every Massachusetts town has been cutting back on expenditures for schools, roads and community services in order to keep under the 2.5% limitation, while also struggling to satisfy steadily increasing state and federal spending mandates. Proposition 2.5 allows communities to override its provisions by a majority vote in an election. This has spawned a minor political industry in Massachusetts – the override campaign. In 2003, more than 70 towns in Massachusetts attempted overrides. Williamstown, Massachusetts, attempted several overrides during the 1990’s, most of which failed. The first significant success (to raise more than a few tens of thousands of dollars) happened in 1998 when a citizens group was organized to mount a supporter identification/get out the vote drive. Overrides in 1999 and 2003 also passed, helped again by the efforts of organized citizen groups. Given the repeated early failures, it iis clear that without the organized campaigns, the three successful overrides would not have passed.

The three successful overrides were similar in their size (all in the range of $500,000-$600,000), and all were supported by organized citizen campaigns. However, the differences between them are quite interesting to study. In each case, town residents were contacted by volunteers - other town residents - to inform, persuade and ultimately to identify support. This was followed by a get out the vote effort targeted at those identified supporters. The total town population is roughly 8,500, with roughly 4,700 registered to vote.

In 1998, we handed out geographically delineated call/walk lists to everyone we could find - 85 volunteers - and in theory, we should have contacted every vote-eligible resident. The call lists were assigned, as much as possible, to volunteers who lived in the same neighborhood as those people being called. Despite the 100% coverage of the vote-eligible population on call lists, we actually only succeeded in contacted about 60% of those people. Why? Simply put, volunteers do not like to do cold-calling. This distaste is the major reason less than 2/3 of the assigned calls were actually completed. Of those residents contacted, only 30% were, in the end, willing to be cataloged as supporters. This was the case, despite the fact that most of the residents were contacted by people who were, nominally, their neighbors, and also despite the fact that in the actual voting, 58% supported the override. There can be little doubt that, all else being equal, being contacted by a neighbor is better than being contacted by a rank stranger from outside the community. Yet outreach from rank strangers, in fact, is how the majority of political field operations are done.

In the 1999 override, I was in charge of the Supporter ID/GOTV effort, and I was determined not to use cold-calling. I created a system I called a friend-to-friend campaign. I printed out the resident list for the town, and convened a meeting where about 45 people went around the tables where a printed resident list was arrayed and picked names of people that they thought, based on their personal knowledge of them, were potential supporters. I instructed them to avoid people who they thought were unlikely to support. The meeting ended with about 1,200 names picked for call lists, which I felt that would not be sufficient (our goal was to identify 1,000 supporters). Over the next couple weeks, we managed to bring in about 15 more volunteers, have them to go through the resident book (something I revised and printed out every couple days), and sign up to call people. Each stage of the process involved keyboarding in all the picks (I used a Claris Works database) and then printing out call lists for each volunteer. Results of contacts were physically or electronically sent to me, where again they were keyed in, and we thereby kept track of the progress of the campaign and were able to generate mailings to identified supporters. This different kind of campaign is reflected in the 1999 numbers. We had a lot fewer volunteers most likely because recruiting and activating them was more difficult - they had to go to where the resident book was, and physically search through it, to create their call lists. Despite this, we actually identified more supporters in 1999 than we did in 1998. 57% of contacts resulted in identified supporters and 85% of assigned calls were actually completed. The volunteers were much happier too.

In 2003, we used a rudimentary web application I designed and Jonathan Leamon coded, which simply put the friend to friend type of campaign, used in 1999, online. The resident database could be searched, and call lists built. Results of contacts could be recorded, and progress could be viewed by all, in real- time. Captains could monitor the work of the people they were supervising. In cases where a volunteer did not have web access, his captain acted as facilitator - either doing the data entry directly, or setting up that person with someone who could do the online work. We began with about 55 volunteers, however the Supporter ID effort doubles as a recruitment campaign (as it did in both earlier campaigns), and with the ease of adding and effectively utilizing new volunteers, our numbers quickly increased to almost 100. The 2003 numbers speak for themselves. Not only did we activate more volunteers, but they created larger call lists, while returning an identified supporter percentage equal to the 1999 campaign.

To put this all another way, with the web system facilitating a friend to friend campaign, we were able to recruit and utilize 2% of the vote-eligible population as volunteers, and with that 2%, come very close to saturating the potentially supportive population (late-comers to the volunteer ranks complained that almost every person they could think of was taken), without wasted effort and alienation or mobilization of the opposition.

I call this a “friend to friend” campaign. Instead of laying siege to the population and wearing it down with uncomfortable and unwanted approaches from strangers ringing doorbells or calling during the dinner hour, you grow the campaign from the inside, through the web of relationships inside the community.

Comparing friend to friend with demographic targeting. My experience is that cold-calling performs poorly compared to “warm calling”. But what about the fact that through careful targeting you can restrict your approaches to those most likely to be supportive, and thereby improve the rate of return vs. effort substantially? Given that you can use interchangeable anyones to do cold calling or doorbelling; doesn’t the increased size of your potential volunteer pool offset the poorer performance per volunteer? Also, while any individual person has a limited number of friends (a variable number, but definitely not infinite), a motivated volunteer is only limited by hours in the day for contacting strangers.

Targeting requires that assumptions be made about the relationship between information known about individuals and their likelihood to support your cause. Intuitively, it makes sense that a black urban middle-class registered democrat will support the more left wing candidate, for example. This seems simple enough, but in fact most people do not present such clear-cut targeting profiles, and many of those who do, do not in fact turn out to be supporters!

Here is an interesting anecdote: In our third override campaign, after 80 or so volunteers had picked call lists and started reporting that they could not find any friends who had not already been taken, I created a targeted list from the unpicked population of those people who had been identified as supporting one or both of the previous overrides that we had data for – roughly 200 people. We weren’t guessing that these people might be sympathetic to the idea of raising property taxes – we knew from past experience. I assigned a dozen or so high-school aged “shock troops” to cold-call these people. I was surprised to discover that their success rate was no better than the rate for our 1998 campaign, about 20%. These kids had received training on making cold-calls, and were enthusiastic (not at all unlike the van loads of college-break kids for Howard Dean descending upon Iowa as I write this).

This suggests that targeting made little difference in the rate of success in identifying supporters. I don’t believe that is true generally, especially if you are dealing with a much more heterogeneous population than that of a small New England town. But it does call into question the conventional wisdom that very high quality demographic data is the most critical asset for a political campaign. I would argue that the most valuable asset is a volunteer pool that is well networked into the “target” population, and how well the campaign organizers understand how to develop and use those volunteers.

But what about the fact that a given volunteer can call many more strangers than they have friends, even assuming they do have friends in the target population? In my experience, this is truer in theory than it is in practice. Cold-calling is distasteful for the majority of people. It takes a rare degree of motivation to perform a large number of cold calls well. Centralized phone banks actually help significantly since they provide motivation and emotional support to the volunteers as they are being hung-up upon by irate callees. But you can only utilize so many people in phone banks due to infrastructural costs, and the need to get volunteers to a given place at a given time.

Clearly, it is better that a person have the political conversation with someone they know and trust, rather than being approached by a stranger. The success rate is more than double. The volunteers do less work, and easier work; so they are happier. The target population is a lot happier: it is not put under siege from outside their community walls; they do not have to deflect desperate strangers during their dinner hours. By utilizing the existing network of relationships within a community, a friend to friend campaign is a community effort that strengthens the social fabric. It also inspires the population to become a community of activists, rather than a passive “market”.

How does the friend to friend campaign work in the real world? The organizational framework required for friend to friend may be somewhat different from traditional techniques that use demographic targeting, precinct walking and phone banking (or target list cold-calling). However it is not incompatible, and in fact both kinds of methodologies can be, and probably should be used simultaneously in the same target population. The only thing that friend to friend requires, that may not be provided in some traditional campaigns, is task ownership. When you put a friend on your call list, you own that particular task until your contact is completed, or you remove him from your list. It is quite easy to put traditional campaign methodologies into this model, and indeed they would likely benefit from it. For example, a phone bank software module would create a call task when the call sheet is instantiated, and release the task when the call sheet is closed. A precinct walk list would be owned by the precinct walker (or the organizer of a van load of precinct walkers), and can be released when the results are recorded.

One thing I've learned in working on multiple campaigns is that your volunteer workforce is an asset not to be taken for granted. And the value of good will with the electorate should likewise not be underestimated. Stranger-to-stranger campaigns turn people off and alienate them - both "your" people, and your target population. Then the next time around, you find yourself wallowing in the bad blood of the last campaign. Your population is more polarized and hostile, and you have difficulty recruiting activists.

Clearly, your mileage may vary depending upon the nature of the community and the political question at hand. But it is clearly suggested by the experience I have described that an organized friend to friend campaign produces more identified supporters per volunteer; when augmented with web tools that every volunteer can use to build call lists and record responses it facilitates recruitment and effective utilization of a larger number of volunteers.

Pat Dunlavey, 12/24/2003