Patronage Network Analysis

A survey measures individual attributes — whether a respondent received a gift, their income, their loyalty — and relates them with regression. Patronage network analysis instead measures the relations themselves: who is tied to whom, in which direction, and with what resource. This lets it locate brokers and choke points, recover the pyramidal architecture, and reason about how removing an actor would fragment the machine, none of which is visible from attribute data alone. The two are complementary: network structure explains the channels, survey or experimental data measure what flows through them.

Because the patron-client bond is fundamentally asymmetric. A patron supplies protection, jobs, and resources downward; a client supplies loyalty, votes, and deference upward. Symmetrizing the graph would collapse these into a single undifferentiated tie and erase exactly the hierarchy the theory cares about. Keeping the arcs directed lets the analysis distinguish patrons (high out-degree of benefits) from clients (high in-degree) and trace resource flows down the pyramid and support flows back up.

Not raw connectivity but structural position. A broker who sits between otherwise disconnected clusters — spanning a structural hole, in Burt's terms — has high betweenness and low constraint, meaning resources and information must pass through them and they can play one side against the other. Such intermediaries can capture rents, deliver or withhold support, and survive even when richer or higher-status actors are removed, which is why patronage network analysis treats brokerage measures, not just degree, as the key to identifying the machine's load-bearing actors.