The current analysis suggests that voter preferences only partly determine election outcomes and subsequent policy choices. In the game model, voters do not behave strategically, but parties, candidates and contributors are all constrained by their correct anticipations of what voters are likely to do. To the extent that voters' behavior in the model reflects their preferences over policy outcomes, one can say that the game solution is an example of ``the rule of anticipated reactions,'' a relation between outcomes and preferences that some would say indicates that voters are powerful (Nagel 1975; Arnold 1990). In the game, the best measure of voters' preferences is the concentration parameter, g: when g<0, the benefits from service are widely dispersed and voters prefer to have post-election service increase, and when g>0, any benefits from service are concentrated on a few and voters prefer to have service decrease.
The ideal for voters, then, is to have dispersed benefits and service as high as possible. In two of the four types of campaign in the game solution, that is what they get. The choices of service concentration and challenger quality that place the campaign in region IV or region V of Figure 1 result in service provision levels that strictly increase for as long as the campaign is considered to run after either the challenger (in region IV) or the incumbent (in region V) drops out. In the former case, voters are getting the best policy outcomes but are also being given the least choice--i.e., no choice--in the election. When there is a competitive race with the incumbent running the voters have already lost out, because in that case g>0 but the amount of post-election service will be positive.
It does not follow, however, that voters are best off when the campaign is not
competitive. In the other kind of uncompetitive race, with
(g,h)=(.0425,0), the unpredictability of the dynamics in region VI
of Figure 1 muddies the picture a bit, but as long as h=0 the
outcome is quite unfavorable for voters. For in this case, once the
challenger disappears (i.e., p=1), contributions to the incumbent and
therefore the amount of post-election service from the incumbent increase very
rapidly. Voters dislike this service, but there is nothing they can do to
stop it. The game model suggests that the reason for the incumbent to be
running unopposed is much more likely to be such an unfavorable dynamic than
the one that gives voters their ideal outcome. Given that the incumbent is
unopposed, the probabilities for the mixed-strategy equilibrium of the
first-stage game suggest that the unfavorable dynamic is roughly seven times
more likely than the ideal one to be the reason.
The dynamics of anticipated reaction are most likely
to have driven the candidates and the financial contributor into a de
facto conspiracy against the voters.