How Institutions Reproduce and How They Change

This Part has examined six institutions — education, family, work, religion, health and media — and the same puzzle has appeared in every one.

Each institution persists in a recognisable form across generations, survives the departure of everyone who staffs it, resists deliberate reform, and reproduces outcomes that many of the people inside it do not intend and would not defend.

And yet institutions do change — sometimes slowly, occasionally fast, and in ways that are patterned rather than random.

This lesson draws the common thread. It is about what makes an institution durable, and what makes it, nonetheless, change.

The keyboard, the gauge, and the argument they are usually used to settle.

Two objects are used, endlessly, to illustrate how institutions get stuck.

The keyboard layout on which most of the world types was, according to the standard story, designed to slow typists down — to stop the mechanical arms of early typewriters jamming — and has persisted for a century and a half after the mechanical problem it solved disappeared, because everyone learned it, so everyone teaches it, so everyone learns it.

And railway track gauge: the distance between the rails on much of the world's railways traces, through a chain of copied decisions, back to the axle-width of carts, itself shaped by the ruts of earlier roads. A measurement set for one reason persisting long after the reason is gone.

These are the standard illustrations of path dependence, and this course has already flagged the honest problem with the first (see 7.5.4): the specific claim that the keyboard is significantly inferior to alternatives, and was retained despite this, has been disputed — the studies purporting to show a better layout are weak, and the layout may be roughly as good as any.

Which is exactly why the examples are worth revisiting here rather than simply repeating. The mechanism — early choices constraining later ones through increasing returns — is real and well evidenced across institutions. The famous illustrations of it are shakier than the mechanism. And keeping those two things apart is the discipline this whole course has been trying to teach.

So set the keyboard aside and take a better-evidenced case from this Part.

A catchment-based school admission system (see 8.3.3, 9.1.3). Once it exists, house prices near good schools rise to reflect the school. Once prices have risen, the families who bought in have a large financial stake in the system continuing — their house is worth what it is because of it. They are also, being the families who could afford to buy in, the best-resourced and best-organised constituency in the area. Any reform that severs school from address threatens their wealth, and they will resist it effectively.

Nobody designed this to be self-protecting. But once the arrangement capitalised into asset prices, it grew a constituency with the resources and the motive to defend it — and that is why it persists, long after anyone would design a school system this way.

That is path dependence with the mechanism intact and the illustration solid, and it is the model for everything in this lesson.

What is an institution, such that it can persist?

An institution is not a building or an organisation. It is a stable pattern of rules, roles, expectations and practices that structures a recurring area of social life — a settled way of doing something, that people orient to, that has consequences for departing from it, and that outlives its current occupants.

Marriage is an institution; a particular marriage is not. The school is an institution; a particular school is an organisation that instantiates it.

And the defining feature — the one that makes reproduction a puzzle — is that institutions are simultaneously constraining and enabling (see 5.8.2 on structuration). They constrain: you cannot easily marry three people, teach without a curriculum, or work outside the wage relation. They enable: they make coordinated action possible at all, by supplying shared expectations that everyone can rely on.

The puzzle is that this stability is produced by nobody in particular and defended by mechanisms that require no defender — which is 4.2.1's social fact, and it is the thread through this entire Part. The question is: by what specific mechanisms?

The mechanisms of reproduction

Five, each of which has appeared in this Part, now named.

One — increasing returns and lock-in. The more an arrangement is used, the more it pays to keep using it — through sunk investment, coordination (it pays to do what others do), learning (people get good at the existing way), and adaptive expectations (people plan around it continuing). The cost of switching rises the longer the arrangement runs, until it exceeds the benefit of any alternative, even a better one. This is the catchment school; it is credential requirements (see 9.1.4); it is the employment classifications of 9.3.3.

Two — the growth of a constituency. Any arrangement that distributes advantage creates people who benefit and who will act to protect it — and, critically, the beneficiaries are frequently the best-resourced and best-organised, because the arrangement made them so (see 8.1.1 on opportunity hoarding). Reproduction is not automatic; it is actively defended by an interested party the arrangement itself produced.

Three — legitimation. An institution persists more securely when it is seen as right, natural or inevitable rather than merely as one option among others. The school's sorting appears fair because its criteria are public and universal (see 9.1.1); the medical classification appears technical rather than moral (see 9.5.2); the distribution appears to reflect merit (see 8.3.3 on misrecognition). An arrangement that is believed to be just requires no coercion to maintain — which is hegemony (see 5.5.1), operating institution by institution.

Four — the taken-for-granted. The deepest mechanism, and the one that requires no belief at all. Most institutional reproduction happens not because people endorse the arrangement but because they never consider the alternative — it is simply how things are done, cognitively invisible, reproduced through habit and script (see 5.8.1 on habitus). You do not decide each morning to use money, speak your language, or send children to school; you enact institutions without noticing them. This is the sociological-institutionalist point: much of what looks like rational adoption is the copying of what is standard, adopted because it is legitimate rather than because it works (see 9.3.1 on rules copied between organisations).

Five — complementarity. Institutions fit together, and each is held in place by the others. The wage relation fits the family form fits the school schedule fits the housing market fits the care system (see 8.5.3, 9.2.3). You cannot easily change one because the others are built around it — which is why piecemeal reform so often fails, and why the whole configurations of 8.8.1 differ between countries rather than individual policies.

These five are not competing theories. They are five mechanisms that operate together, and any durable institution is held by several at once — which is why institutions are so much harder to change than the people who want to change them expect.

Why reform disappoints

Four recurring reasons, each documented across this Part.

One — advantage relocates. Close one channel and the advantaged use another (see 8.3.3). Equalise access to a level of education and advantage moves to the next level (see 9.1.4). Remove a stated requirement and the informal version persists (the removed degree requirements of 9.1.4). A reform aimed at one mechanism, in a system held by five, reroutes rather than removes.

Two — the informal system absorbs the formal change. An institution is not only its rules; it is the practices that make the rules workable (see 9.3.1). A reform that changes the rules without changing the practices is worked around — the street-level workers, the discretion, the unwritten system all persist through the formal change (see 9.3.1 on the work-to-rule).

Three — Goodhart's law. Reforms that work by measurement corrupt the measure (see 7.2.2). Target-driven reform produces gaming, creaming and displacement — the measured thing improves while the thing it was a proxy for does not (see 9.3.1, 9.1.2 on value-added).

Four — the constituency fights back. The people the arrangement advantaged have the resources and the motive to defend it, and they are frequently more organised and more effective than the diffuse beneficiaries of reform (see 8.8.1 on why redistribution is politically hard).

None of these means reform is futile. They mean reform that succeeds has to be designed against them — acting on multiple channels at once, changing practice as well as rule, avoiding proxy targets, and building or neutralising constituencies. Which is exactly what the successful cases in this Part had in common (see 9.1.3 on comprehensivisation; 8.8.1 on the countries that moved the distribution).

How institutions change

They do change. Five pathways, from the evidence of this Part.

One — critical junctures. A brief period in which the constraints loosen and several paths become genuinely possible, followed by a new lock-in (see 7.5.4). Wars, revolutions, depressions, pandemics, and the founding moments of new arrangements. The mid-century welfare settlement, the post-communist transition, the comprehensivisation reforms — each was a window that then closed. These are rare and largely exogenous — they happen to institutions.

Two — gradual change, which is where most change actually happens. The critical-juncture model implies long stability punctuated by rare upheaval, and that misses most of what occurs. A more useful typology identifies gradual mechanisms that transform institutions without any dramatic break.

Layering: new rules added alongside old ones, which gradually change the whole as the new grows and the old atrophies — private pensions layered onto public ones, private schooling alongside state.

Drift: the institution stays formally the same while the world around it changes, so its effect changes without any reform — a benefit not uprated with prices quietly erodes (see 8.6.2); an employment law written for stable jobs drifts as work fissures (see 9.3.3).

Conversion: the same rules redirected to new purposes by new actors — an institution captured and turned to ends it was not built for.

And displacement: the slow replacement of old arrangements by previously subordinate ones — the marginal becoming central.

These gradual mechanisms are how institutions mostly change, and they are easy to miss precisely because there is no dramatic moment — the fissured workplace, the diverging family, the secularising minority all happened this way, without a decisive event.

Three — endogenous contradiction. Institutions can undermine their own conditions (see 8.5.3 on capitalism and care). An arrangement that erodes what it depends on generates pressure for change from within.

Four — shifts in the environment that alter the returns. When the returns to the existing arrangement fall — because technology, demography or the economy shifts — the lock-in weakens, and alternatives that were uncompetitive become viable. The demographic transition changed the family this way (see 9.2.2); deindustrialisation changed class formation (see 8.2.3).

Five — agency, mobilisation and ideas. Institutions are defended and challenged by people, and organised challenge — movements, professions claiming jurisdiction, patients demanding recognition, activists disrupting a diagnosis — is a genuine source of change (see 9.4.2, 9.5.2, and Part 11 to come). The evenhandedness rule applies: this operates in every direction, and an account of institutional change that has agency only on the side the author favours is incomplete.

Two errors to avoid, one on each side.

Over-determinism: treating institutions as so locked-in that change is impossible, and reform as always futile. This is contradicted by the cases in this Part where institutions did change, sometimes fast, and where deliberate reform did move outcomes (comprehensivisation, the welfare settlements, the distributions of 8.8.1). The mechanisms of reproduction are strong; they are not absolute, and treating them as absolute is both wrong and politically convenient for those the arrangement advantages.

Over-voluntarism: treating institutions as easily redesigned by anyone with a good idea, and their persistence as mere inertia or failure of will. This is contradicted by every reform in this Part that rerouted, was absorbed, corrupted its measure, or was defeated by a constituency. Institutions are held by mechanisms that require no defender and defended by constituencies that do; wishing them away does not work.

The competent position holds both: institutions are durable, for specifiable reasons, and changeable, under specifiable conditions — and knowing the reasons and the conditions is the difference between reform that works and reform that disappoints.

Because a citizen who wants to change something needs to know what they are up against, and a citizen who is told nothing can change needs to know that is false.

The thread through this entire Part is that institutions produce outcomes nobody in particular intends and everybody in particular finds hard to change — the sorting school, the fissured workplace, the medicalised difficulty, the concentrated attention. These are not the results of anyone's decision; they are the results of arrangements held by mechanisms that operate behind the backs of the people enacting them (see 4.2.1, and 1.1.1 — this is the sociological imagination applied to the machinery of everyday life).

Four questions for any institution you want to understand or change.

Which of the five reproduction mechanisms holds it? Increasing returns, a constituency, legitimation, the taken-for-granted, complementarity. Usually several — and the reform has to address each.

Where would advantage relocate if you closed this channel? If the answer is "somewhere else", the single-channel reform will disappoint.

Is it changing already, gradually and unnoticed? Layering, drift, conversion and displacement are happening in every institution now, without a dramatic moment.

And what would have to loosen for it to change deliberately? A critical juncture, a shift in returns, a mobilised constituency, or a designed reform that acts on multiple mechanisms at once.

One closing observation that carries into the rest of the course. This Part has repeatedly found that institutions reproduce inequality without anyone deciding they should (see 8.3.3, and every Topic here). That is the most important single finding of the sociology of institutions, and it cuts in a specific direction.

It refutes the account in which unequal outcomes reflect individual choices freely made — because the choices are made inside arrangements that pattern them.

And it refutes the account in which unequal outcomes reflect a deliberate design by an identifiable villain — because there is usually no such designer.

What it leaves is harder than either: outcomes produced by durable arrangements, held by impersonal mechanisms, defended by interested constituencies, and changeable only by understanding all three. That is not a comfortable conclusion for anyone, which is a fair indication that it is close to right.

Every institution in this Part persists across generations, survives its staff, resists reform, and reproduces unintended outcomes — and yet institutions change, in patterned ways.

An institution is a stable pattern of rules, roles, expectations and practices that outlives its occupants, is simultaneously constraining and enabling, and is reproduced by nobody in particular.

The path-dependence illustrations are shakier than the mechanism — the keyboard claim is disputed — but the mechanism is solid, and the catchment school (capitalised into house prices, growing a resourced constituency) is the well-evidenced model.

Five reproduction mechanisms operate together: increasing returns and lock-in; the growth of a resourced constituency the arrangement itself produced; legitimation (the arrangement believed just requires no coercion); the taken-for-granted (reproduction requiring no belief, only habit and script); and complementarity (institutions holding each other in place, so piecemeal reform fails).

Reform disappoints for four documented reasons: advantage relocates; the informal system absorbs the formal change; Goodhart's law corrupts the measure; and the constituency fights back. Reform that succeeds is designed against all four — multiple channels, practice as well as rule, no proxy targets, constituencies built or neutralised.

Institutions change through five pathways: rare critical junctures (exogenous windows that close); gradual mechanisms — layering, drift, conversion, displacement — which is where most change happens and is easily missed for lacking a dramatic moment; endogenous contradiction; shifts in the environment that alter the returns and weaken lock-in; and agency, mobilisation and ideas, which operate in every direction.

Two errors: over-determinism (change impossible — contradicted by the cases that changed) and over-voluntarism (change easy — contradicted by every reform that rerouted or was absorbed). The competent position holds both: durable for specifiable reasons, changeable under specifiable conditions.

And the Part's central finding — institutions reproduce inequality without anyone deciding they should — refutes both the free-choice account and the deliberate-villain account, and leaves the harder truth of impersonal mechanisms, interested constituencies, and change that requires understanding all three.

Institution — a stable pattern of rules, roles, expectations and practices that outlives its occupants.

Path dependence — early choices constraining later ones through increasing returns; the mechanism is solid, the famous illustrations shakier.

Increasing returns / lock-in — the rising cost of switching that keeps an arrangement in place even against better alternatives.

Constituency — the interested party, frequently well-resourced, that an advantageous arrangement produces and that defends it.

Legitimation — the perception of an arrangement as right or natural, which removes the need for coercion.

Taken-for-grantedness — reproduction through habit and script requiring no belief.

Institutional complementarity — the fit between institutions that makes piecemeal change fail.

Critical juncture — a brief window of open possibility followed by a new lock-in.

Layering / drift / conversion / displacement — gradual mechanisms of change without a dramatic break.

Institutional drift — a stable institution whose effect changes because the environment does.

Over-determinism / over-voluntarism — the errors of treating institutions as unchangeable and as easily redesigned.

One — name the mechanisms. Take one institution you would change and identify which of the five hold it in place. There will usually be several.

Two — trace where advantage would go. For a reform you support, ask where the advantaged would reroute if it passed. Design the reform to close that route too.

Three — find the gradual change. Take one institution and look for layering, drift, conversion or displacement happening now, without a dramatic moment. It is usually there and usually unremarked.

Four — locate the constituency. For any arrangement that persists despite wide criticism, identify who benefits and how organised they are. That is why it persists.

Five — apply the closing finding. Take one unequal outcome from this Part. Write the free-choice account, the deliberate-villain account, and the impersonal-mechanism account. Decide which the evidence supports — and notice it is the least comfortable.