An experimental investigation of pure voting power

We evaluate the accuracy of power indices by experimentally measuring the political power embodied in blocks of votes <i>per se</i>.  The experiment incorporates several players interacting in online chat rooms under supervision.  The chat rooms and the selection process reduce or eliminate extraneous political forces leaving logrolling as the primary political force.  Results show that two standard power indices reflect voting power while other power indices and proportionality do not.