Clientelism Analysis

The defining feature is contingency. Programmatic redistribution gives benefits by transparent rule regardless of how the recipient votes; patronage is a diffuse, long-term obligation. Clientelism is a direct, conditional, iterated exchange in which the benefit is tied to political support and can be withdrawn for defection. Stokes's model shows that this contingency is what produces 'perverse accountability,' where the machine holds the voter to account rather than the reverse. Empirically, the test is whether receiving a benefit tracks observed or expected support, not merely need.

Through monitoring by brokers embedded in dense social networks. Even without observing the ballot, a broker who lives in the community can infer loyalty from rally attendance, local reputation, and everyday observation, and can target those too poor or too exposed to risk losing benefits. Stokes models the credibility of the implicit threat as rising with network density: where voters are observable, the machine's promise to reward compliance and punish defection becomes credible, so clientelism flourishes where social ties are tight and falters where voters can hide.

A naive vote-buying logic predicts targeting persuadable swing voters, but Stokes finds machines often reward already-loyal and easily monitored individuals. The reason is enforcement: the machine can only credibly condition benefits on people whose behavior it can observe, and dense ties to committed clients make compliance verifiable. Rewarding the monitorable sustains the long-run relationship and deters defection, even if it leaves some persuadable voters untargeted. This monitoring-driven targeting is one of clientelism analysis's signature empirical predictions.