From the WSJ comes this story about Russia’s plan to install web cameras in polling stations for the March presidential election.

Russia Friday launched the presidential election website, www.webvybory2012.ru, that will allow web users to access video recorded at any of the approximately 92,000 polling stations across the country. One camera will give a full panorama view of each polling station and a second camera will be directed at the ballot box.

[…]

The website allows users to select as many polling stations for monitoring as they wish, although only until Election Day. Users will be able to monitor the election from 12 a.m. to 8 p.m. Moscow time. For an hour, recording will continue but nothing will be shown to observe the secrecy of the ballot. Starting at 9 p.m., when voting closes in Kaliningrad, Russia’s westernmost province, the service will show ballot counting and viewers will be able to see video from 8 p.m. local time.

Webcams in polling stations aren’t a bad idea by themselves, but I have a lot of problems with the way this is being implemented. My first concern is that it may contribute to the common development practitioner practice of assuming impact from an output. As with all transparency initiatives (making public records available, etc..) it’s not enough to simply produce the information and assume civil society will use it.  Often times they won’t. There are many similar “citizen monitoring” projects being done through the Ushahidi platform, which produce neat maps. Often the assumption is that people will actually do something with that map. I don’t want to bash Ushahidi too hard as I think it can do interesting things, but producing data should not be viewed as a behavior-changing impact of an intervention. It’s just an output that we hope will lead to the behavior change.

The webcams do, however, remind me of an innovative experiment done in Afghanistan: In 2010, local election monitors took photographs of the final tally sheets in local Afghan polling stations, which was shown to reduce fraud by 60%. The Afghanistan experiment was done through a Randomized Control Trial (RCT), which brings me to my next problem with this experiment.

It’s always difficult to determine if election monitoring actually reduces fraud (although Susan Hyde has done great work showing that it can). This is for the simple reason that we don’t know the counterfactual level of fraud if the observation wasn’t there. Because of this, I think it would be much smarter if  – instead of trying to put webcams in nearly every polling station – they randomly assigned the web cameras to certain stations. This would  allow us to measure if the intevention was actually effective or not. Aside from the fact that so many webcams will make monitoring of any of them less effective, not randomizing the cameras will make it impossible to actually determine impact.  Of course this assumes the actual goal of the project is to reduce fraud and not just give the appearance of transparency.