An early definition of clientelism emphasized the exchange of votes for favours, over a long period of time, among actors with asymmetric power, the clients having little power. Politicians would reward a portion of their supporters with public resources in return for electoral support. Scholars have found this definition increasingly wanting: first, clients can offer politicians financial contributions and other non‐monetary resources, not just votes. Second, clients could be rather ‘powerful’. Third, the sale of one's vote in exchange for a benefit to which the client is not otherwise entitled qualifies as corruption.
It is more fruitful to understand clientelism as a type of principal–agent relationship. Clientelism involves three actors, a principal, an agent, and a ‘client’. Typically, a client (say, a politician's supporter and financier) transfers resources over which he has control to the agent (the politician). The agent will then transfer resources he obtains from the principal (the electorate) back to his client. The criterion of allocation is particularistic, rather than universalistic: clients are rewarded with public contracts, appointments and the like not because of merit or qualifications but prior support. Given the nature of this exchange, the relationship between agent and client tends to be long‐term.
Contrary to corruption, the clientelistic exchange is done in the open and contravenes neither a legal provision nor a custom: the US president openly appoints trusted friends and supporters to ambassadorships around the world. To the extent that such an allocation breaches a legal provision and is done secretly, clientelism turns into corruption, the exchange occurring in an illegal market. If the exchange goes counter to public sentiments, it still qualifies as clientelism although the public frowns upon it.