Bureaucracy in Practice

Weber's account of bureaucracy (see 4.3.4) describes what organisations claim to be: a hierarchy of offices, governed by written rules, staffed by qualified appointees, acting impersonally.

Every element of that is real, and none of it is how the work gets done.

This lesson is about the gap — and about the fact that the gap is not a failure of implementation but a structural necessity, which can be demonstrated by the simple expedient of removing it.

The week they followed the rules.

A large processing office handles applications — for a benefit, a permit, a licence; the substance does not matter. Thirty thousand cases a year, sixty caseworkers, a published procedure manual running to four hundred pages.

The manual specifies everything. Which documents must be present before a case can be opened. Which checks must be completed in which order. Which categories of query must be referred to a supervisor. The wording of standard letters. The point at which a case must be returned to the applicant as incomplete.

And the office does not work that way, and has not for years.

Caseworkers open cases with documents missing, because the missing document is usually the one that takes six weeks to obtain and can be chased in parallel. They ring each other rather than referring upward, because the person two desks away dealt with an identical case last month and the supervisor has not done casework for nine years. They keep informal notes on which employers' payslips are reliable and which are not. They know which of the standard letters produces a phone call and which produces silence, and they use the second when they need the applicant to respond.

None of this is in the manual. Some of it contradicts the manual. And the office clears its caseload.

Then a dispute arises over pay, and the union calls a work-to-rule.

Nobody stops working. Nobody is even slightly obstructive. They simply do exactly what the manual says, in the order it says, and nothing else.

Cases are returned unopened for missing documents, restarting the clock. Every query goes to a supervisor, who has six hundred queries and no capacity. Nobody rings anybody. The standard letters go out in their standard form.

Within three weeks the backlog has doubled. Within six, the office is in the news.

And the crucial observation: not one rule was broken. The organisation was brought to a halt by the exact application of its own written procedures.

Which is the demonstration this lesson exists for. The formal structure could not run the organisation for six weeks. What runs it is the informal system — the workarounds, the shared knowledge, the discretion, the judgement about which rules to apply when — and that system is invisible, unrewarded, unrecorded, and doing most of the work.

If the rules cannot run the organisation, what are they for?

The tempting conclusion is that bureaucracy is inefficient and the rules are obstacles. That conclusion is common, it is the basis of a great deal of management reform, and it is wrong in an important way.

Because the alternative to being processed by a rule is being processed by somebody's opinion of you.

Weber's contrast was not between bureaucracy and efficiency. It was between bureaucracy and patrimonial administration, in which office is personal property, decisions are made by favour, kinship and payment, and there is no basis on which anyone can demand consistent treatment.

Against that, the impersonal rule is a protection, and it is worth stating what it protects.

That your case is decided by the same criteria as everyone else's. That the decision is written down and can be reviewed. That the person deciding is qualified and can be replaced without the decision changing. That you do not need to know somebody.

Every one of those is an achievement, and the people who benefit most from it are those with the least ability to know somebody — which is why the critique of bureaucracy has a distributional edge that is rarely noticed. Personal discretion advantages the connected.

So the question is not whether rules are good. It is: what happens to an organisation once it is governed by rules? — and that is an empirical question with a well-developed answer.

What happens to organisations governed by rules

Five findings, each from a study, and together they constitute the field.

One — goal displacement. Rules exist to produce reliable outcomes. To produce reliability, adherence must be emphasised. Emphasising adherence makes the rule an end in itself.

The result is the familiar figure who applies the regulation correctly and defeats the purpose it existed for — and Merton's point is that this is produced by the very devices that create reliability, not by a defect of character. An organisation that trains people to follow rules precisely has trained them into an incapacity to notice when the rule does not fit.

Two — cooptation. An organisation that must survive in an environment will absorb the groups that could threaten it — giving them representation, consultation, a place on the board.

And it is changed by them. Selznick's study of a large public agency found that in securing local support, it took on local interests and priorities that substantially redirected what it did. The organisation kept its stated mission and its actual programme drifted, in a direction nobody chose and that the formal structure did not record.

Three — the vicious circle of rules. Gouldner's study of a gypsum plant found a pattern worth stating as a sequence.

Management, uncertain about performance, introduces rules. Rules specify a minimum standard. Workers, treated as needing supervision, deliver the minimum the rules specify. Performance falls to the specified level. Management, observing this, introduces more rules.

Each step is rational and the sequence is self-defeating — and the rule that was meant to raise performance defines its ceiling.

He also distinguished three ways rules exist in practice: rules nobody enforces or obeys and everyone is content with; rules both sides accept as legitimate; and rules imposed by one side and enforced through punishment. The same written rule can be any of the three, which is why reading an organisation's manual tells you very little about it.

Four — the informal system is what makes the formal one work.

Blau's study of a public employment agency found that agents were required to consult supervisors on difficult cases and forbidden to consult each other, since peer consultation was thought to spread errors.

Agents consulted each other constantly, against the rule.

And the ones who did performed better. The informal consultation network handled the cases the formal channel could not absorb, spread accurate knowledge faster than the hierarchy did, and reduced the anxiety that made agents avoid difficult cases altogether.

Five — power accumulates wherever uncertainty remains.

Crozier's study of a French tobacco monopoly examined an organisation in which almost everything had been specified. Production was standardised, pay was by formula, promotion was by seniority, and discretion had been eliminated almost everywhere.

Almost. The machines broke down, unpredictably, and only the maintenance workers could fix them.

They had enormous power — over production workers, over supervisors, over the plant's schedule — and they defended it by refusing to write anything down. There were no maintenance manuals. The knowledge was held personally and transmitted orally.

Crozier's general claim is the useful one. In a rule-bound organisation, power belongs to whoever controls the last remaining zone of uncertainty — and the more comprehensively an organisation is specified, the more concentrated that power becomes.

Look for what cannot be routinised, and you will find where the power is.

Street-level bureaucracy

The most useful single idea in this lesson, and it applies to a large fraction of all public employment.

Consider the people who actually deliver public services: teachers, police officers, social workers, benefits assessors, housing officers, nurses, immigration officials, court clerks.

Three things are true of all of them.

Their work cannot be fully specified, because it involves human situations that are too various to enumerate. A rule cannot say what to do about this family. So they must exercise discretion — not as a privilege but as a condition of the job existing.

Demand always exceeds resources, and it does so structurally rather than temporarily: services that are free at the point of use generate demand up to the limit of supply, so the queue is a permanent feature.

And they are accountable for outcomes they do not control.

Lipsky's conclusion follows from those three, and it is the important part. People in this position necessarily develop routines, simplifications and rationing devices to make an impossible job possible. And because policy is only ever realised through their decisions, those coping mechanisms become the policy.

What is written in legislation is an intention. What a claimant, pupil or patient actually receives is the output of a person with too many cases, a partial rule, and a set of habits developed to survive the workload. The street-level worker is where policy is made, whatever the constitutional theory says.

And the coping mechanisms are identifiable, which makes this a research programme rather than an observation.

Rationing. Queues, waiting lists, restricted opening hours, requirements to attend in person, forms that take three hours. Demand is limited by cost to the applicant rather than by rule.

Creaming. Where performance is measured by success rates, workers select the cases most likely to succeed — the job-ready applicant, the compliant patient, the pupil near the grade boundary. The people hardest to help are systematically deprioritised by a system measured on results (see 7.2.2 on Goodhart, and 9.1.3 on cream-skimming).

Routinising. Treating cases as instances of a small number of types, which is cognitively necessary and produces the experience of not being listened to.

And psychological adjustment: redefining the job to something achievable, and redefining clients — as deserving or undeserving, motivated or not — in ways that make the rationing bearable to the person doing it (see 8.6.2).

And the modern extension, which is the sharpest current idea in this area.

Every interaction with a bureaucracy imposes costs on the person navigating it, and those costs are of three kinds.

Learning costs: finding out that the entitlement exists, what the criteria are, and how to apply.

Compliance costs: gathering documents, completing forms, attending appointments, providing evidence repeatedly, waiting.

And psychological costs: the stress of the process, the stigma of claiming, and the loss of autonomy in being assessed.

Together these constitute administrative burden, and two findings about it matter.

Burden reduces take-up substantially, and the reduction is not random: it falls hardest on people with the least time, the least literacy, the least stable housing, the worst health and the least experience of dealing with institutions — which is to say, on the people the programme exists for.

And burden is frequently a deliberate policy instrument. Where a benefit cannot be cut politically, its take-up can be reduced administratively — by requiring more evidence, more frequent reassessment, in-person attendance, or shorter application windows. Rationing by hassle is a policy choice that appears in no legislation and has distributional effects as large as many that do.

Which means: when comparing two systems' generosity, the entitlement rules are only half the story. The other half is what it costs to obtain what you are entitled to.

The escape attempts

What happens when organisations try to stop being bureaucratic.

The recurring proposal is to flatten hierarchy, remove rules, and organise around teams, networks and shared purpose.

And there is a classic finding about what happens, which came not from management research but from a movement that tried it seriously.

A group operating deliberately without formal structure — no offices, no elected roles, no procedures — discovers that it has an informal structure instead. Influence accrues to those who know each other, who have time, who are confident, who have been there longest, and who are friends with whoever is at the centre.

The difference is not that the structureless group has no hierarchy. It is that its hierarchy is unaccountable, because it does not officially exist. There is no procedure for challenging a decision nobody officially made, and no way to remove someone from a position they do not officially hold.

Formal structure is not the opposite of informal power. It is the thing that makes informal power contestable — which is the same argument as the one about patrimonialism, arriving from the other end of the political spectrum.

And the second escape attempt is subtler: rules moved into systems.

Much apparent debureaucratisation has relocated rules rather than removed them. A procedure written in a manual becomes a field that cannot be left blank, a workflow that will not advance, a scheduling algorithm, a productivity metric, a route that must be followed.

This is bureaucracy with a shorter feedback loop and less discretion, not less bureaucracy. The discretion that street-level workers used to exercise — and which, as above, is what made the work possible — is precisely what a specified system removes. Which predicts that algorithmically managed work should exhibit exactly the pathologies above in intensified form, and 9.3.3 examines whether it does.

And the third: measurement. Where trust in professional judgement declines, it is replaced by inspection, audit, targets and reporting. This creates an entire secondary activity of demonstrating compliance, which consumes the time of the people being audited, generates its own specialists, and — by Goodhart's law (see 7.2.2) — corrupts the measures it relies on.

The costs of that secondary activity are real and are almost never counted against the savings it is meant to produce.

Because "cut the red tape" and "we need proper procedures" are both slogans, and the useful questions are underneath them.

Four questions for any organisation.

What does the rule prevent? Every rule was written after something happened. Before removing one, find out what. A great many were written after a scandal, and removing them re-creates the conditions for it.

Where is the discretion, and who holds it? Not in the manual — in the practice. Follow a case and note every point where a person decided something the rules did not determine.

What cannot be routinised? That is where the power is, and it is frequently not where the organisational chart says it is.

And what does it cost to obtain what you are entitled to? The burden is a policy, whether or not anybody wrote it as one.

Two things to hold together, which is the difficult part.

The impersonal rule is a genuine protection, and the strongest beneficiaries are those with no other way to be treated fairly. A world of pure discretion advantages the connected, and the anti-bureaucratic mood in politics rarely notices whom it is proposing to disadvantage.

And the rules cannot do the work, which the work-to-rule demonstrates in six weeks. The informal system is not a corruption of the formal one; it is the condition of the formal one operating at all — and an organisation that treats its workarounds as failures to be eliminated is dismantling the mechanism by which it functions.

The competent position is to know which rules are load-bearing, which are residues, and which exist to be worked around — and that knowledge is held by the people doing the job, is transmitted orally, and disappears when they leave.

A work-to-rule halts an organisation without a single rule being broken — which demonstrates that the formal structure cannot run it and that the informal system does most of the work.

The alternative to the rule is not efficiency but patrimonialism — decision by favour, kinship and payment. The impersonal rule guarantees consistent criteria, a reviewable decision, an interchangeable decider, and no need to know somebody — and it benefits most those with the least access to personal connection.

Five findings constitute the field. Goal displacement: the emphasis on adherence that produces reliability turns the rule into an end. Cooptation: absorbing external groups to survive redirects what the organisation does, while the stated mission is retained. The vicious circle of rules: rules specify a minimum, workers deliver the minimum, management adds rules — with the rule that was to raise performance defining its ceiling; and the same written rule can be unenforced, jointly accepted, or punitively imposed. The informal system makes the formal one work — agents forbidden to consult each other did so anyway, and performed better. And power accumulates in the last remaining zone of uncertainty — the maintenance workers who could fix the machines and refused to write anything down.

Street-level bureaucracy follows from three conditions: work that cannot be fully specified, demand structurally exceeding resources, and accountability for uncontrolled outcomes. The coping mechanisms become the policy — rationing, creaming (systematically deprioritising the hardest to help wherever success rates are measured), routinising, and redefining the job and the clients.

Administrative burden — learning, compliance and psychological costs — reduces take-up most among the people a programme exists for, and is frequently a deliberate instrument: rationing by hassle, appearing in no legislation.

And the escape attempts fail in instructive ways. Structurelessness produces an unaccountable hierarchy rather than none — formal structure is what makes informal power contestable. Rules relocated into software are bureaucracy with less discretion, not less bureaucracy. And replacing professional judgement with audit creates a secondary activity of demonstrating compliance whose costs are never counted.

Ideal type of bureaucracy — hierarchy, written rules, specialisation, impersonality, qualified appointment, salaried career.

Patrimonial administration — office as personal property, decision by favour; the contrast that makes bureaucracy an achievement.

Goal displacement — the rule becoming an end in itself through the emphasis that produces reliability.

Trained incapacity — competence in a procedure producing inability to notice when it does not fit.

Cooptation — absorbing external groups for survival, and being redirected by them.

Mock, representative and punishment-centred rules — unenforced, jointly accepted, and imposed.

Informal organisation — the unwritten practices without which the formal structure does not operate.

Zone of uncertainty — what cannot be routinised, and therefore where power concentrates.

Street-level bureaucracy — front-line discretion under structural scarcity, whose coping mechanisms constitute the actual policy.

Creaming — selecting the cases most likely to produce good measured outcomes.

Administrative burden — learning, compliance and psychological costs of obtaining an entitlement; unequally borne and often deliberate.

Tyranny of structurelessness — the unaccountable hierarchy that emerges where formal structure is refused.

Audit society — the growth of verification as a distinct activity replacing professional trust.

One — find the workaround. In any organisation you know, identify one thing everyone does that the procedure does not describe. Then ask what would happen if it stopped.

Two — trace a case. Follow one application, referral or request from start to finish and mark every point where a person decided something the rules did not determine.

Three — locate the uncertainty. Ask which role in an organisation you know cannot be covered by anyone else at short notice. That is where the power is.

Four — cost the burden. For one entitlement in your country, count the steps, documents, appointments and hours required to obtain it. Then ask who is least able to supply them.

Five — check a rule's origin. Take one rule you find absurd and find out when it was introduced and why. Some are residues. Some are load-bearing. The difference is not visible from the rule.