The Monkey Cage

I don't think its surprising that email has a small turnout effect, but I also seriously doubt that campaigns use it that way. Its much more likely to be used as a fundraising tool among people who are already potential supporters or as a way to remind people about scheduled events, etc. With that in mind, I wonder if the kind of randomized experiments that have been getting used in this work reach the same population that the campaigns see as their targets. (This is not meant as a knock on David's work, which is interesting.)

John, thanks for the advertising. It is cool to see a blog I read mention my work.

Scott, to answer your two questions:
1) Campaigns do use email as a get out the vote tool. The Kerrey campaign sent over 7 emails in 2004 encouraging supporters to vote. Naturally, they use email to perform a number of other activities such as provide information directly to voters, recruit volunteers, solicit donations, and announce events. Once the email list is compiled for all those other purposes, it is essentially costless for a campaign to hit "Send." So most campaigns do use email to encourage turnout. Initially, I imagined that few people thought email was effective at increasing turnout. I have received enough correspondence from political consultants offering anecdotes about "effective" email campaigns, that I now think at least a few consultants think email works.

2) Your point about targeting is a two-sided coin. Experimenting with a campaign, a researcher targets the same population because the campaign is doing the targeting. That is, the researcher simply extracts the control group from the targeted population. However, external validity then comes into question. How would people NOT targeted by the campaign respond to the treatment? Whether this is a different quantity of interest or heterogeneous response to treatment depends upon which question the researcher thinks is most interesting.

In an ideal world, the campaign would agree to target everyone and then the researcher could examine whether people who would not ordinarily be targeted responded differently. I have never encountered a campaign willing to do that. Unfortunately, tax laws prevent the researcher from coordinating with the campaign and expanding the universe directly. So the question as to whether people not targeted by a campaign would respond the same way is difficult to study.

Thanks for the publicity and the comment.

David

Thanks for coming a long to respond, David. A quick response, of sorts.

I regret not being a little more careful in raising my initial question. Its not so much that campaigns don't use email to remind people to vote, but I suspect that they believe the yield is much higher on other activities (spreading information, raising money, etc.). In that sense, it seems to me that the emails aren't really seen as GOTV tools as much as, say, they are fundraising tools.

That said, your comments about the consultants are interesting and are evidence against what I'm saying. As a research matter, though, I still wonder if your results would differ in the dependent variable were something else. (This would be a great place to look for indirect mobilization, as you do in the APSR paper!)

Fair comments on the other issue. I won't engage here because it would probably bore the hell out of everyone else.