A lot of the work of politics often seems to be about persuading people of your position. But in practice it is hard to know when people are really persuaded by political arguments and when they simply align with the position of others because it's in their interests to do so. Experimental approaches can be helpful for sorting these things out since the messages can sometimes be separated from the content of messages.
In one project in Sao Tome e Principe, Martin Sanbdu, Will Masters and I examined an ambitious exercise in deliberative democracy to assess the extent to which deliberative communication leads to convergence of political positions in small groups. Disappointingly we found that outcomes of group deliberations are very strongly determined by who happens to be facilitating the discussions suggesting that participatory processes may be much more open to manipulation than we tend to think. Paper here. [See also: Data and Stata Do File | Note on the interpretation of the R squared | CI around the R-squared ]
In a later project in Canada with Torun Dewan and Daniel Rubenson we worked with the British Columbia group campaigning for the introduction of single transferable vote system for the province. To assess the relative merits of different strategies the campaign randomized which messengers went where and what messages they delivered. There, unlike the Sao Tome experiment, we found evidence for strong campaign effects with both message-based and endorsement-based campaigns seeming to work. Surprisingly though but we find a muted role for idiosyncratic features of prospective persuaders. Paper here. [Replication and other material]