Door to door canvassing has been a staple of the Obama campaign. In fact, in states like North Dakota it seems to be the whole campaign. Looking at the campaign's events page for North Dakota, almost every event is a door to door canvass. Last sunday's Fargo Forum listed a bunch of community festivals, but it doesn't look like the campaign will be making any of them. Issue oriented community events are a good political strategy too, but I don't see much of them scheduled. It appears that the campaign is betting everything on canvassing- but does canvassing work?
I've done about a dozen shifts canvassing for Obama in Iowa, Minnesota, South Dakota. and North Dakota. In previous campaigns I've canvassed for Dean in Iowa as well as other candidates. In most cases I've felt that at most I'd moved maybe one vote per shift. That's not much of a return for piling another four hours or so of abuse on my tired old body. One canvassing session in inner city north Minneapolis even brought me into contact with a gangbanger who had no interest in voting, but was seriously peeved that I was trespassing on his turf. Then there's the time I spent all day on Iowa dirt roads trying to catch farmers who weren't home or were too busy to caucus anyway.
On Iowa caucus night in 2004 and 2008 I observed caucuses in precincts I had canvassed. In 2004 I saw a few familiar faces, but they were regular party activists who would have been there whether I'd canvassed them or not. In 2008 my canvassing may have encouraged a young mother with newborn child and her mother to attend the caucus- but she was pretty highly motivated so I have to give her most of the credit. But we still came in 3rd behind Hillary and Edwards in that precinct. Fortunately the Edwards supporters formed an alliance with us and we shared both delegates with them, freezing Hillary out.
A couple weeks back I did a canvass shift in North Dakota. After four hours I was only half way through the list. I may have won over one voter, but I probably antagonized several. I could see why the burn out rate among canvassers was so high. The Obama staffer assigned to that turf said that it was rare to have even one volunteer join him in the canvassing, and he usually ended up walking the turf by himself. I've heard rumor that the Obama campaign has as many as 100 staffers in North Dakota. That number is probably high, but given that many of the canvasses listed on the Obama website have no one signed up, the campaign effectively has maybe 100 canvassers on the ground in North Dakota. Reputedly the Obama campaign has set a quota of 330 "contacts" per week for staffers, so the campaign may be contacting 33,000 North Dakota voters a week. But there's a universe of over 400,000 North Dakotans eligible to vote, and even with those 100 canvassers going full tilt it'd take about three months to attempt one contact with them. Unfortunately the election is two months away, and early voting begins in just a few weeks.
So essentually the Obama campaign is relying almost entirely on canvassing in North Dakota- I haven't even heard an Obama ad on KFGO, the talk station that features democrats Joel Heitkamp and Ed Schultz as well as the usually republicon blowhards. So is canvassing effective, and effective enough to win a state almost all by itself?
About now some Camp Obama graduates will chime in and claim that canvassing will boost the vote for the candidate by 8% or so. I've even heard numbers as high as 17% claimed, citing a Michigan study. Being a Camp Obama dropout (actually couldn't afford the travel and housing costs to attend), I missed that bit of indoctrination. But the numbers seemed iffy to me- in some southwest Minneapolis and Iron Range precincts turnout is so high that we'd have to "canvass" the graveyards to cause that big an increase.
So being better at stats than doorknocking, I read the studies, all three of them. The first was done by Green and Gerber during the 1998 election in New Haven. There canvassers were able to contact 37% of the voters on their list and they observed a 2.2% increase in turnout among democrats they contacted versus controls. Then they performed some statistical hocus-pocus to come up with that 7% improvement number- they assumed that the democrats they couldn't reach would have responded the same as the ones they could, and extrapolated thusly that 7% number. I've done canvasses where over half the voters on the walk list had moved, passed on, or were hard core republicans- trying to extrapolate a treatment effect to voters they couldn't even find is a statistical leap of faith one expects of republicons, not political scientists. Green, Gerber, and Nickerson then studied the effects of canvassing in several 2001 municipal elections. They reported a 2.1% improvement in turnout among the 30% of the targeted voters they contacted. Through statistical sleight of hand they extrapolated that to the 8% number bandied about. And the much quoted Michigan study? That study of the 2002 elections produced that wild 17% number- but the researchers, Nickerson, Friedrichs, and King, noted that the sample size was too small to be statisticly significant.
In conclusion, canvassing probably works. But in the limited studies we have, it only increased voter turnout by a couple percent. Given how labor intensive canvassing is for such limited benefits, no sane campaign would rely on it as heavily as the Obama campaign is in North Dakota. Canvassing has it's place in the campaign toolbox, but the Obama campaign needs to utilize other tools like targeted local media, events, visibility, etc.. We also need a lot more study of the effectiveness of canvassing- does it work better in low or high turnout elections? Is canvassing more effective with different voter populations than others? Given that the Obama campaign won't be able to canvass all of North Dakota's voters, it'd be helpful if they'd at least randomly assign some of them to treatment and control groups so we can measure the effectiveness of canvassing efforts.