In this Paper, we test how politicians respond to information about their constituents' preferences. Our general empirical framework is the following: in the 2015 Canadian election, we conducted the largest ever study of public opinion in Canada, soliciting the opinions of nearly 40,000 Canadians on ~50 issues. Since that time, we have been systematically sharing these data with Members of Parliament on relevant issues. Our general approach is to randomly assign who receives this information on an issue-by-issue basis, and to then observe how this changes their vote, their speech in parliament, and their interactions with constituents. For this Paper, we focus in particular on whether negatively or positively framed public opinion data (i.e. how many constituents support or oppose a policy) has an effect on how politicians respond to this information. The issues we use in this case are free trade, support for struggling industries, and taxation. Our outcome variables are parliamentary votes, speeches, and responses to constituents in a large scale audit experiment with real voters.
