The key notion is that the populace is divided into three categories:
- The selectorate is the subset with a say in selecting the leader. This group is small in a monarchy or military government and large in a democracy.
- The winning coalition is the subset of selectorate whose support the leader needs to stay in office. This group is small in a rigged-election autocracy and large in modern democracy.
- Those who are not part of the selectorate are disenfranchised. Literally.
To stay in power, a leader can provide private goods — like land, money, and monopoly rights — to his supporters in the winning coalition, or he can provide public goods — like roads, bridges, and rule of law — to everyone in society. The larger the winning coalition the more efficient it is to pay them off in public goods.
Where this gets interesting is in the incentives.
Members of the winning coalition want to get the most benefits possible, but they realize that they are always at risk of being removed from the winning coalition if they become too “expensive” for the leader (or challenger) to keep — so they prefer systems that increase their chance of staying in the winning coalition. This means they prefer systems where the winning coalition is a big proportion of the selectorate — like an electoral system that requires majorities instead of pluralities — but where the selectorate is as small as possible.
The leader, on the other hand, wants the selectorate as big as possible and the winning coalition as small as possible, to keep them competing for his largesse.
The disenfranchised naturally want the selectorate expanded to include them, but they also want the winning coalition expanded as much as possible, so that they might benefit from any public goods used to reward the large winning coalition.
I recommend all of the Bruce Bueno de Mesquita EconTalk Podcasts.