Why does a small stretch of northern Italy produce most of the world's high-end ceramic tiles? Why is nearly every surgical instrument in Pakistan made in Sialkot, a single mid-sized city? And why have countless government-built industrial parks across the developing world sat empty for decades, despite generous subsidies?

These questions point to one of the most durable puzzles in development economics. Manufacturing rarely spreads evenly across geography. It concentrates, sometimes ferociously, in specific places that develop deep pools of expertise, dense supplier networks, and tacit knowledge that competitors elsewhere struggle to replicate.

For developing countries trying to industrialize, understanding cluster dynamics is not a peripheral concern. It is central to whether industrial policy succeeds or produces expensive monuments to good intentions. The evidence suggests that clusters cannot be conjured into existence by proclamation, but they can be nurtured, accelerated, and sometimes redirected. The challenge is knowing the difference between supporting agglomeration and pretending to create it.

Alfred Marshall identified the core logic more than a century ago, and modern research has largely confirmed his intuition. When related firms concentrate geographically, three distinct advantages emerge that isolated producers cannot access. Each operates through different channels, but they reinforce one another over time.

The first is a specialized labor pool. In a cluster, workers with industry-specific skills accumulate through training, migration, and on-the-job learning. Firms can hire quickly when demand rises and shed workers when it falls, knowing replacements exist. Workers benefit from thicker markets that reduce the risk of specializing. This dynamic explains why Bangalore's software engineers or Shenzhen's hardware technicians command premiums their skills alone would not justify elsewhere.

The second is supplier depth. Manufacturing requires countless intermediate inputs, from precision components to specialized services like tooling and repair. Where clusters mature, suppliers proliferate and compete, driving down costs and enabling small firms to focus on their core competencies. A shoe manufacturer in Wenzhou can source a specific type of sole in hours; a competitor in an isolated location might wait weeks and pay several times more.

The third, and perhaps most important, is knowledge spillover. Ideas leak. Workers change jobs, engineers meet at trade fairs, suppliers share problems across clients. This tacit knowledge, difficult to codify and impossible to license, is precisely what separates productive clusters from ordinary industrial zones. It is also why proximity still matters in an age of digital communication.

Takeaway

Productivity is not just a property of firms; it is a property of places. The same worker, the same machine, the same investment produces different results depending on what surrounds it.

Most successful clusters were not planned. Silicon Valley grew from Stanford's engineering department and a few defense contracts. The garment districts of Bangladesh coalesced around a handful of pioneering entrepreneurs who trained workers who later founded their own firms. Sialkot's surgical instrument cluster traces back to colonial-era demand and generations of accumulated craft knowledge.

This organic emergence contains an uncomfortable lesson for policymakers. The specific mix of history, personalities, initial conditions, and lucky breaks that produce a cluster is nearly impossible to replicate on demand. Governments across Africa, Latin America, and South Asia have built industrial parks with tax holidays, subsidized land, and modern infrastructure, only to watch firms operate as isolated islands rather than integrated ecosystems.

The failures share a pattern. Planners typically focus on physical infrastructure and financial incentives while neglecting the softer conditions that make clusters work: skilled labor pipelines, functioning input markets, competent local suppliers, and mechanisms for knowledge to circulate. A factory dropped into a green-field zone with none of these supports remains an enclave, dependent on imported inputs and expatriate management, generating employment but little transformation.

Yet dismissing all cluster policy as futile goes too far. Ethiopia's Hawassa Industrial Park, China's township and village enterprises, and Vietnam's electronics zones all show that intentional policy can accelerate cluster formation when it works with rather than against underlying comparative advantages. The successful cases share common features: they built on existing capabilities, encouraged linkages with local firms, and treated the cluster as a system rather than a collection of tenants.

Takeaway

You cannot decree an ecosystem into existence, but you can create conditions where one might grow. The difference between gardening and manufacturing matters here.

The most effective cluster policies share a common orientation: they identify emerging concentrations of activity and remove specific bottlenecks rather than attempt to build clusters from scratch. This shifts the government's role from architect to gardener, watching what grows and helping it along.

Coordination is often the binding constraint. Individual firms may benefit from better testing laboratories, shared logistics infrastructure, or specialized training institutes, but no single firm can justify the investment. Public provision or public-private partnerships to fill these gaps have produced strong returns in places like Taiwan's Hsinchu Science Park and India's Tirupur knitwear cluster. The intervention is targeted, technical, and responsive to firms' actual demands.

Trade policy also matters more than is often recognized. Clusters that produce for export face constant competitive pressure that forces upgrading, while those insulated from global markets often stagnate. Successful policies have combined selective protection during infancy with clear expectations of export orientation, avoiding the trap of permanent import substitution. Discipline matters as much as support.

Perhaps most importantly, effective cluster policy embraces experimentation and failure. Not every cluster attempt will succeed, and the goal is not to pick winners but to enable enough experiments that some will flourish. This requires patient capital, honest evaluation, and the political maturity to phase out programs that underperform. It is closer to venture capital than industrial planning, and it demands institutions capable of learning.

Takeaway

Good industrial policy is less about vision than about diagnosis. The right question is not what should we build, but what is trying to grow, and what is holding it back.

Industrial clusters represent one of the few areas in development economics where geography, institutions, and firm-level dynamics visibly intersect. Ignoring them means missing much of what actually drives structural transformation.

The lessons for policymakers are humbler than the grand industrial strategies of earlier eras but potentially more useful. Look for existing concentrations of activity. Understand what constrains them. Provide the coordinating investments that individual firms cannot. Maintain export discipline. Accept that some experiments will fail.

Development is ultimately about building productive capabilities, and capabilities live in ecosystems as much as in firms. Countries that grasp this tend to grow. Those that keep building empty industrial parks tend to wonder why the factories never quite become industries.