Polnet – June update

Near the end of June I took a break from working directly on my research in Williamsburg to go on a bit of an adventure in Bloomington, Indiana for the 6th annual Political Networks Conference. I’d written most of this up on the plane back from Indiana, but never got around to posting it, so here it is!

Another student, Drew, and I traveled to Bloomington a few days early in order to take advantage of the workshops offered before Polnet officially started. The workshops were really interesting – among other things, we learned the basics of network analysis and how to gather data from Twitter. Probably the most relevant workshop to my research was one on social influence and field experiments using networks taught by David Nickerson from Notre Dame. Having worked with Green and Gerber (the rockstars of voter turnout field experiments), Dr. Nickerson knows a lot about how to design field experiments involving networks and how to look for causality and the spread of behavior within networks, something that can get very complicated in incomplete or very large networks. (For anyone who is interested, I would highly recommend taking a look into this really cool study Dr. Nickerson did on how voting behavior can spread like a “contagion” –

Looking into whether my postcards have an effect beyond their recipients is something I have been thinking about looking into, but I just have not had time to fully consider how I might easily analyze it on top of everything else I am working on. I might go back to this once I get more campaigns to cooperate with my research and am comfortable with the sample size for the main part of my project. Regardless, this workshop will be helpful as I start thinking about how I will target my mailers. I plan to randomly assign households to the treatment and control groups, rather than individuals, but this means that voters living in the same household will receive one mailer with all individuals’ voting histories on it. I will need to think more about how this could complicate my analysis. If I count everyone in the household as having received the mailer, I may weaken my measured effect, since the mailer may only have been seen by one voter in that household. On the other hand, if I only address a mailer to one voter per household, should I also look into whether the mailers made that voter’s family members/housemates more likely to vote as well?


Drew, Professor Settle, and me in Indiana.

The rest of the conference was really interesting and it was great to talk to folks from all over about what graduate school is like. It was also my first ever trip to the Midwest!