Laws are passed by legislatures but implemented by bureaucracies, and the bureaucracy is not a neutral conveyor belt. Public choice models the bureaucrat as it models everyone else — an agent with interests — and asks what an agency, left to its incentives, will do. The classic answer, due to William Niskanen (1971), is uncomfortable: it will tend to grow too big.
The Niskanen model
Niskanen modelled a bureau as a monopoly supplier of a service to a single buyer — the legislature or 'sponsor' that funds it. Two features drive the result. The bureau has a monopoly (the sponsor cannot easily buy defence or tax-collection elsewhere), and the bureau knows its own true cost function while the sponsor does not. The bureau also derives utility from a larger budget — more staff, status, perks, and mission. Put these together and the bureau, which effectively makes the sponsor a take-it-or-leave-it offer of a whole output-for-budget package, can extract far more than a competitive supplier could.
Niskanen's oversupply result
Because the bureau proposes an all-or-nothing budget and the sponsor cannot see the true cost, the bureau expands output past the efficient level — where marginal social benefit equals marginal cost — toward the point where total benefit equals total budget. In the starkest version the bureau drives output to roughly twice the efficient quantity, exhausting the entire social surplus in budget. The waste is not (mainly) theft; it is over-provision — a bureau doing more of its thing than is worth doing, because its incentive is size and its information advantage lets it indulge it.
The principal–agent frame
Modern treatments restate this as a principal–agent problem. The politician (principal) wants the service delivered at least cost; the bureau (agent) wants budget, autonomy, and an easy life; and the agent has private information about effort and cost. The gap between what the principal would choose with full information and what the agent delivers is 'agency slack' — the bureaucratic counterpart of managerial slack in a firm with weak shareholders. Everything in oversight design — audits, performance targets, competition, transparency — is an attempt to shrink that slack.
The critiques, and what survives
- Migué & Bélanger (1974) — bureaucrats may maximise discretionary budget (the slack between the budget and the minimum cost of the required output) rather than total size, which they spend on perks, easy workloads, and pet projects rather than raw output. The distortion changes shape but doesn't vanish.
- Dunleavy (1991) — senior bureaucrats often prefer 'bureau-shaping' (small, elite, policy-focused units) to running large, troublesome, budget-heavy delivery agencies. This explains the appeal of contracting delivery out while keeping the prestigious core — and complicates the simple size-maximising story.
- Empirics — pure doubling of output is rarely observed, and political oversight, competition (including from the private sector), and budget rules constrain bureaus. But the core insight is robust: information asymmetry plus an agency that benefits from its own expansion produces a systematic tilt toward oversupply and slack unless actively checked.
Why this matters most for parastatals
The Niskanen problem is sharpest where the monopoly and information asymmetry are sharpest and oversight is weakest — state-owned enterprises and statutory agencies that face no competition, control their own performance data, and answer to a distracted ministry. The chronic over-staffing, soft budget constraints, and mission creep of many African parastatals are a textbook Niskanen pattern, not merely bad management. (The SOE reform problem is taken up in the Industrial Policy course.)
Exercise
A statutory authority that licenses and inspects a sector requests a 40% budget increase, citing the need for more inspectors to protect the public. It is the sole provider of this function, and only it holds the data on inspection workloads and outcomes. (1) Analyse the request through the Niskanen model — why can't the sponsoring ministry simply trust the stated need? (2) Reframe it as principal–agent: what is the agency slack and why is it hard to observe? (3) Migué-Bélanger and Dunleavy each suggest a different reading of what the agency really wants — give both. (4) Recommend three oversight mechanisms that attack the information asymmetry rather than just cutting the number, and explain how each helps.