Every generation of reformers relearns the same lesson. One country adopts an institution that anchors prosperity or stability in another, often copying its formal text closely, and the institution fails to deliver. Independent central banks become instruments of the executive. Constitutional courts are packed or ignored. Anti-corruption agencies prosecute only the government's rivals. The usual explanations are insufficient political will, poor implementation, and cultural resistance. They are not wrong, but they are incomplete.

The deeper problem is that institutions rarely work as separate parts. They operate as configurations, in which the performance of any one element depends on the presence and quality of others. Economists formalised this as complementarity: two arrangements are complements when the returns to each rise with the presence of the other. Peter Hall and David Soskice built a typology of capitalist economies on this logic. Arend Lijphart's distinction between consensus and majoritarian democracy rests on a similar logic of design choices that reinforce one another.

Taking complementarity seriously changes how we should think about reform. It explains three things. Piecemeal change can leave societies worse off than either the old equilibrium or the intended new one. Borrowed institutions often become hollow shells. And the order in which reforms are introduced may matter as much as their content. What follows examines each of these in turn, drawing on evidence from post-communist transitions, Latin American constitutionalism, and East Asian development.

The basic logic is simple. In Paul Milgrom and John Roberts's treatment, institutions are complements when an increase in one raises the marginal return to the other. In Germany, long-term employment protection makes sense because patient, bank-based finance shields firms from short-term market pressure. That finance in turn makes sense because sector-wide wage bargaining and vocational training give firms reasons to invest in workers they expect to keep. Each element is rational only given the others.

Political institutions have the same structure. Lijphart's consensus model is a coherent package, not a list of independent features. Proportional representation produces multiparty legislatures. Multiparty legislatures produce coalition cabinets. Coalition government depends on norms of bargaining and on a civil service insulated enough to serve successive partisan combinations. Remove one element, for instance by pairing proportional representation with a powerful directly elected president, and the remaining pieces may begin working against each other.

That case is not hypothetical. Scott Mainwaring called presidentialism combined with a fragmented multiparty system a difficult combination. Presidents without legislative majorities face incentives to govern by decree or to build coalitions through patronage rather than policy programmes. Brazil eventually developed workable bargaining mechanisms under what scholars call coalitional presidentialism. The arrangement came at a considerable cost in transparency and has remained vulnerable to systemic corruption scandals.

The most important implication comes from the theory of the second best, formulated by Richard Lipsey and Kelvin Lancaster in 1956. When one condition of an optimal arrangement cannot be met, moving the remaining conditions closer to their ideal values does not necessarily improve outcomes and may make them worse. Applied to institutions, a partial reform that moves one element toward recognised best practice can disrupt the compensating adjustments that made the existing system work.

Russian privatisation in the 1990s shows the danger. Transferring state assets to private owners was meant to create constituencies for markets and the rule of law. But the country lacked functioning courts, securities regulation, and a state able to enforce contracts. Instead, privatisation concentrated assets among actors whose interests lay in preserving the weakness that had enriched them. The reform did not simply fall short. It entrenched an equilibrium hostile to its own completion.

Takeaway

An institution's value lies less in its design than in its fit. Before asking whether a reform is good, ask what it depends on, and what currently depends on the arrangement it would replace.

Legal scholars have debated transplants since Alan Watson argued in the 1970s that borrowing was the main engine of legal change. The empirical record suggests that how a country borrows matters enormously. Daniel Berkowitz, Katharina Pistor, and Jean-Francois Richard found that some countries developed their legal orders internally, or adapted imported law to local conditions and to a population familiar with its principles. These countries achieved markedly more effective legality than those that received law unadapted. The authors called this deficit the transplant effect.

The first recurring pattern is formal compliance without functional substance. Lant Pritchett, Michael Woolcock, and Matt Andrews call this isomorphic mimicry. States adopt the organisational forms that win legitimacy with donors and peers, such as procurement laws, ombudsmen, and audit agencies. They do not acquire the capability those forms are meant to express. The institution exists in statutes and organisational charts, but the behaviour it is supposed to produce does not.

The second pattern is the absence of supporting informal institutions. Douglass North emphasised that formal rules are embedded in norms, conventions, and enforcement practices that cannot be legislated. Many nineteenth-century Latin American constitutions drew heavily on the American model of separated powers. They lacked what sustained that model: dense networks of local self-government, established legislative traditions, and a settled elite consensus. Executive dominance and recurrent constitutional breakdown frequently followed.

The third pattern is capture by the very incentives the transplant was meant to discipline. An imported institution enters an existing distribution of power, and those who benefit from the status quo will try to redirect it. Anti-corruption commissions become weapons against opponents. Electoral bodies become agents of incumbents. Regulators are staffed through patronage. The form is not ignored but repurposed. That is often more damaging than outright rejection, because the institution now lends a veneer of legitimacy to the abuse.

Behind all three patterns lies a common error. Reformers treat the institution as the cause of the outcome in its country of origin. More accurately, it was one component of a configuration that produced the outcome jointly. The Westminster fusion of powers works in Britain partly because of unwritten conventions that restrain the executive. Exported to new states without those conventions, the same fusion has often enabled dominant-party rule.

Takeaway

Institutions are visible; the conditions that make them work mostly are not. What gets copied is usually the part that was easiest to see, rather than the part that was doing the work.

If complementarities are real, reform seems to face a dead end. No single element can be changed productively without the others, yet changing everything at once exceeds the capacity of any state. Comparative experience suggests three ways through. All three accept that the target configuration cannot be installed directly. Each step on the path must therefore be self-sustaining, both politically and institutionally.

The first is to work with existing complementarities rather than against them. China's reform era is the standard example. Yingyi Qian described its dual-track pricing system, township and village enterprises, and fiscal decentralisation as transitional institutions. These arrangements were far from textbook best practice. Yet they generated growth while compensating the actors who might otherwise have blocked change. Their efficiency came from fitting the existing system, not from resembling the destination.

The second is diagnostic targeting. Ricardo Hausmann, Dani Rodrik, and Andres Velasco proposed growth diagnostics because long lists of reforms implicitly assume the elements are independent. Their method identifies the binding constraint: the one whose relaxation yields the largest gains under current conditions. Reformers can then concentrate resources where complementarities amplify effort rather than cancel it out. The logic extends beyond economics. A judicial reform may matter less than the police and prosecutorial capacity that determine which cases reach court at all.

The third concerns the order in which state capacity and political accountability develop. Following Samuel Huntington, Fukuyama observes that where mass democratic competition came before a capable bureaucracy, as in the nineteenth-century United States or in Greece, parties tended to build support through patronage. This entrenched clientelism that later proved hard to dislodge. Where an autonomous bureaucracy came first, as in Prussia or Meiji Japan, later democratisation inherited a state less open to capture. The lesson is not that democracy should wait. It is that sequence shapes which coalitions form.

Each strategy implies a different conception of reform: iterative and problem-driven rather than led by a blueprint. Andrews, Pritchett, and Woolcock formalised this as problem-driven iterative adaptation. Reformers test local solutions, observe how they interact with existing arrangements, and adjust. The method gives up the clarity of a fixed model. In exchange, it discovers complementarities empirically, which is the only way they can be reliably known in a particular context.

Takeaway

The question is not what the ideal system looks like, but which next step is both an improvement and stable enough to make the following step possible. Good reform paths are built from viable intermediate states.

The complementarity perspective does not counsel fatalism. Institutional configurations do change, sometimes rapidly, and borrowed elements can take root and flourish. But it does require reformers to change their unit of analysis: the system, not the component.

Three propositions follow from the comparative record. First, partial reforms should be judged against the configuration they disrupt, not only against the ideal they approximate. Second, transplants succeed when the receiving society adapts them and develops its own demand for them, not when they are copied faithfully. Third, sequencing is itself a design choice, because each intermediate arrangement creates the coalitions that will shape what comes next.

This is Lijphart's insight extended beyond his typology of democracies. Institutions settle into recognisable patterns because their elements evolve together under shared pressures. Understanding those patterns explains why reform so often disappoints. It also points toward a more modest and more durable approach: changing systems in ways they can absorb.