For those of you who didn’t hear it on NPR, Nature just published a fascinating study by Robert M. Bond, Christopher J. Fariss, Jason J. Jones, Adam D. I. Kramer, Cameron Marlow, Jaime E. Settle & James H. Fowler on an experiment to increase voter turnout via Facebook messages. Some, of 61 million Facebook users were randomly assigned to receive a message on their Facebook profile with links to polling place information and a button to indicate to their friends that they had voted. Most (but not all) Facebook users also received information on how many of their friends had voted (a “social message”). The remaining users constituted the control group.
The results show that especially the social message had a small but significant impact on voter turnout (including verified turnout using voter rolls). Moreover, the messages did not just influence the receivers but also their friends and friends of friends.
The massive sample size allows the researchers to identify statistically significant effects even if the substantive size of these effects is small (and believable). Such a sample size (well maybe not this big) is probably needed to examine small intervention (just one Facebook message). It is a useful reminder though that when political scientists say that some campaign event doesn’t matter we can’t really exclude that these campaign events have small effects that are simply not registered as statistically significant due to relatively small sample sizes. Multiple small effects could make a real difference.
[Cross-posted at The Monkey Cage]