Gordon Tullock argues, in this short PBS video, that a rational agent should not bother voting. It comes across as a partial defense of free riding. The argument mostly assumes that a rational agent has perfect information about other voter preferences and voter participation data, and is therefore in a position to estimate the value of a single vote. Although the probable value of a single vote can be estimated from historical data, there is always a degree of uncertainty attached to this value (admittedly, that should be fairly low, but the information is nothing close to perfect). However, for many people the cost of voting is lower (driven lower by allowing for voting in super convenient locations) than the cost of bearing with this uncertainty (driven higher by induced paranoia) and therefore could vote without violating RCT. Voters do behave as rational agents, in that they react to real time information. For instance, when results from the US east coast helps lower the uncertainty of an outcome, voter participation in the US west coast drops (or adjusts. I forget my reference.); indicating that the adjusted cost of uncertainty is then lower than the cost of voting.
Gordon Tullock is a busy man. Also, he seems fairly confident about the value, or the lack of thereof, of his vote. That drives up his cost of voting, while cutting his cost of uncertainty. Rationally, he shouldn’t vote.
Those are the easy questions. The real question for some is that if all of the outcomes of a given process are not to their liking, what do rational agents do?