Yale political scientists Donald Green and Alan Gerber deserve a lot of credit for the high turnout on Tuesday. For a long time, the effectiveness of different methods of voter contact was an open question. It was too hard to control for all the different factors affecting the turnout decision. So they conducted a series of controlled experiments in real-life elections, which showed that face-to-face contact is the most effective way to mobilize voters. Their work has been very influential, and parties and interest groups are now using it to mobilize people on a scale not seen in decades.
Alan S. Krueger recently highlighted some of the key findings from their new book Get Out the Vote in the Times:
-Door-to-door canvassing, though expensive, yields the most votes. As a rule of thumb, one additional vote is cast from each 14 people contacted. That works out to somewhere between $7 and $19 a vote, depending on the pay of canvassers - not much different from the cost of that three-pack of underwear. Canvassers who matched the ethnic profile of their assigned neighborhoods were more successful.
-The effect of leaflets on turnout has not been evaluated as thoroughly as canvassing, but results from two partisan campaigns indicate that one vote was generated for every 66 leaflets hung on doors. In another experiment, just one vote was added for every 200 nonpartisan leaflets. Over all, leafletting costs $14 to $42 a vote. (A salutary aspect of the book is that one, two or three stars are placed next to the central findings to signify the degree of confidence the authors have in the results. This is only a one-star result.)
-Direct mail is less cost-effective than leaflets. Mailing costs totaled around $60 for each additional vote cast. Telephone calling is also not highly effective, with the cost per vote ranging from $200 for heavily scripted calls to $45 for more personalized calls. Even worse, recorded messages and e-mail had no detectable impact on turnout.
-Some candidates mail negative messages to their opponent's supporters to discourage voting. Mailing a negative message depresses votes, but at a very low rate. The cost per vote diminished was about $300. (This is another one-star finding.)
In just-completed research, Professor Green and Lynn Vavreck of the University of California, Los Angeles, placed 5,500 get-out-the-vote commercials on cable networks across randomly selected geographic areas in four states shortly before the general elections of 2003. The ads hardly affected turnout, although the estimated impact necessarily entails much statistical uncertainty. A similar conclusion was drawn from a study of ads in the 2000 presidential election. Thus, commercials may persuade viewers to support a candidate, but they do not appear to affect whether they vote.
Krueger ends with this:
A great deal of knowledge about turnout strategies has been gleaned from experiments in election campaigns, but many gaps remain. Professors Green and Gerber conclude their book in a novel way, by giving office seekers a step-by-step guide on how to conduct scientific experiments on their own. This is not just fanciful thinking: two campaigns have already taken the bait and conducted randomized experiments. The 2004 election promises to be the first to exploit scientific research on voter turnout on a national scale.
Along the same lines, a Times article today on John Kerry's Internet fundraising notes the campaign's extensive use of experimentation:
[Kerry's campaign] also spent a lot of time testing which wording in e-mail messages and on the Web site drew the most contributions. With 2.6 million supporters on the campaign's e-mail list and a Web page averaging 250,000 daily visitors during peak times, even small increases in the percentage of people who donated could equal large gains.
"You start adding those nickels up and it makes a dramatic, dramatic difference," Mr. Ross said.
The campaign learned that fund-raising letters do poorly on Monday. E-mail messages are best sent around 11 a.m., after people have cleared their mailbox of unwanted "spam." And contributions swell at lunchtime on both coasts, when people spend time online.
Mr. Ross's team also tested e-mail subject lines. On the day of Mr. Kerry's convention speech in July - which was also the last day the campaign could raise private money before switching to public financing - the campaign sent out a long letter and a shorter letter, some carrying the subject line "this is it" and some saying "last chance." The short version with the "last chance" heading did best and was delivered en masse.
The Web page was also engineered to bring in money. One example was the "splash page," the first thing that new visitors see. At one point, Mr. Ross and his colleagues had 30 versions of the page up on a wall. They tested photos until they settled on a picture of Mr. Kerry flashing the thumbs up. They tested headlines until they chose "Make history with us."
Even a small contribution button toward the bottom, which was bringing in more than $75,000 a day at its peak, was maximized. The campaign tested four different versions before finding that the label "contribute before deadline" increased the number of donations by 35 percent.
"We have no problem testing our own assumptions," Mr. Ross said. "We don't do anything based on a guess."
Call it the Moneyball-ization of politics (after the brilliant Michael Lewis book on how the Oakland A's use statistics to stay competitive with better-funded teams). In the economic and political realms, people are moving from informed guesses to testing their assumptions against hard data. It's going to change everything.
Comments