GOTV - Part 12 - "In-Your-Face" Feedback

by: Grebner

Sun Feb 27, 2011 at 04:59:18 AM EST

Our conventional GOTV methods provide little or no feedback in the gray, conventional sense, of "reports on how things went", and I suppose it might be a good thing to distribute such information.  But that's not what I'm talking about here.  I don't care how many volunteers showed up or how many precincts were walked.  When I talk about "feedback" I'm talking about something with much sharper edges.

What I mean by feedback is letting voters, volunteers, consultants, and organizations know that we remember what we asked them to do, and now the election's over we've checked up on them, and here's their report card.  To some we say, "Congratulations!  Thank you!  We KNEW we could count on you!"  To others, things like: "Where were you on Election Day?"  "Why didn't your mother mail her absentee ballot back?" Or, "We spent $10,000 and 200 volunteer hours on your project, but it appears to have produced only 20 additional voters."

In short, the February after each even-year November election ought to be a time of much angry shouting, excuse-making, pontificating, and maybe even a little proud boasting.  Whereas today, nobody has any clue which individuals did a good job, which techniques were successful in turning out voters, or what money was well-spent.  As I've argued before, our GOTV management and methodology is reminiscent of Soviet Russian agriculture.  Plenty of rhetoric and medals, but not much wheat.

If we want to move toward a regime which is optimized to turn out as many of our voters as possible given limited resources, tight feedback loops are essential.  Each targeted voter needs to be identified in our database as a target, and each intervention applied to each voter needs to be noted similarly.  As soon as individual voting record data is available (January 15, typically) we need to start "walking back the dog" so we can assign credit and blame.

Small randomly chosen groups need to be identified on the database to be concealed from our various projects, in order to serve as statistical controls.  (Because we have so many, overlapping, programs, many controls will actually receive some GOTV contact - just not the full dose.)

A model must be specified in advance, which tentatively assigns probabilities of voting to each registered voter, in order to ensure the after-election analysis isn't tainted by hind-sight.

The largest amount of feedback will be to the voters themselves, with absentee and non-absentees handled separately.  As we've already shown (Get Out the Vote) giving feedback to voters about their past voting behavior is FAR more powerful than merely reminding them to vote.  Interestingly, telling a voter that we noticed they DID cast a ballot is about equally powerful in increasing subsequent voting as telling them we noticed they DID NOT. 

In the case of what I've called "micro-volunteers", who may for example have agreed to help a family member get to the polls, feedback on their success is likely to have three separate impacts.  First, it will increase the amount of effort the devote in the future, by making clear their promise to us wasn't forgotten.  Second, in many cases, it may tip them off their family member has been quietly delinquent - information they might never have gotten without our intervention.  Third, it will help convince both the micro-volunteer and the would-be voter their relationship to the Democratic Party is a continuing one, and didn't end with last Election Day.

In the case of volunteers or paid staff assigned to larger areas, such as entire precincts, the analysis is more difficult to perform and convey.  Unless we are very meticulous, a specific door-to-door canvasser's efforts will not "pop out" statistically from the random noise of individual decisions to vote or not vote.  That's one reason I prefer to assign each volunteer with a specific short list of individual voters (or households and street addresses) so they can later be provided with the same list, annotated with whether each person actually voted or not.

Finally, at higher levels, the question is whether the targeted groups turn out at a higher rate than the small, concealed, control groups.  For some programs (particularly those involving media) this comparison will be difficult to make.  But in my opinion, the fact it's difficult to measure effectiveness should be taken as a warning there may not BE any effectiveness to measure.  In my experience, if something really makes a difference, the impact is rarely concealed from all detection.

Why bother with creating feedback mechanisms?  After all, if we already know that what we're doing is optimal, setting up, recording, and analyzing feedback just "wastes" resources.  But if we DON'T KNOW what works - and we don't - the resources "wasted" to gather and distribute information pay off handsomely.  

• Feedback methods are the most powerful means we know for encouraging voting.
• Feedback can motivate some volunteers, weed out slackers, and reinforce the belief among the productive that their contributions are noticed and appreciated.
• Feedback to larger organizations will tell them which activities have real payoffs, and which are a waste of resources.
• Feedback allows small pilot projects to be designed and tested, so the successful ones being scaled up in subsequent elections.