Technology and Development

Technology improvement is central to development. In developing countries, this often means adopting and adapting technologies from the frontier rather than inventing new ones. But access to frontier technologies is not enough. New technologies are adopted when they fit local conditions, when complementary inputs are available, when adoption is profitable at the relevant scale, and when decision-makers have the incentives and information needed to use them. The role of policy is therefore not simply to promote ‘technology’ in general. It is to identify where the process of adoption, adaptation, or innovation breaks down, and to choose policies that relax these constraints.

A key policy lesson from this review is that there is no single lever for technology adoption. The relevant policy depends on why adoption is low. If modern equipment cannot be used because power, roads, or logistics are unreliable, infrastructure is the relevant margin. If firms cannot upgrade because they lack access to better inputs or buyers, trade and supplier policies matter. If the missing input is technical expertise, engineering capacity, or management, education and training, become central. If adoption requires suppliers, workers, buyers, and infrastructure providers to move together, government action can help coordinate expectations and investments. If available technologies do not fit local conditions, adaptation or local innovation may be needed. The same policy has limited effects when another constraint binds, and can be costly when it supports technologies that are not yet productive in the local economy.

Policy should therefore start from a diagnosis. Which technologies could plausibly be adopted next? Are they well suited to local conditions, or do they require adaptation? Is adoption held back by missing infrastructure, finance, skills, buyers, suppliers, or standards? Does adoption require many actors to move together? Are firms failing to experiment because returns are uncertain, risks are large, or managers and workers have weak incentives to change how production is organised? The answers to these questions differ across countries, sectors, and technologies. Answering them is the first step in deciding whether policy should focus on infrastructure, education, trade, finance, coordination, experimentation, or local innovation.

The evidence reviewed here points to five general lessons for policy.

1. Focus on feasible technologies

Policy should start from technologies that are actually relevant for the local economy. For some technologies, the main problem is adoption: the technology exists and fits local conditions, but is not used at scale. For others, the problem is adaptation: the technology must be modified for local ecology, factor endowments, skills, infrastructure, demand, or regulation. In some cases, especially in agriculture, health, climate adaptation, and some services, the available technology menu may be poorly suited to local needs. Then policy may need to support local innovation rather than only subsidise adoption.

2. Ask whether infrastructure or another input is the bottleneck

Many technologies fail to diffuse because the inputs needed to use them are missing. Machines require reliable electricity, transport, maintenance, and logistics. Digital tools require connectivity, devices, trust, and enough users. Quality upgrading may require storage, standards, testing, and access to buyers. When these inputs are missing, governments can raise the return to adoption by improving infrastructure, reliability, procurement, regulation, standards, or access to input and output markets.

But infrastructure is usually necessary rather than sufficient. Roads, electricity, ports, and broadband have large effects when they remove a real bottleneck or help coordinate other investments. They have smaller effects when firms still lack finance, workers, inputs, buyers, or the managerial capacity to respond. The policy question is therefore not simply whether to build infrastructure, but whether infrastructure is the margin that prevents adoption.

3. Shift education policy toward the skills required by modern technologies

While basic education is essential, it is often not enough for the adoption of modern technology. As production technologies become more complex, countries need more vocational training, technical skills, college-educated workers, engineers, managers, and workers who can move across firms and sectors. The evidence is clear that modern technologies are more demanding in this sense: they require workers and managers who can operate machinery, solve problems, adapt codified knowledge to firm-specific conditions, and reorganise production.

Education policy should be linked more directly to the technology frontier that firms can plausibly reach. At early stages, the binding constraint may still be literacy and numeracy. As countries develop, the policy focus has to shift increasingly toward technical schools, colleges, engineering capacity, management training, and the allocation of skilled workers to firms that can use them.

4. Governments need to coordinate different actors when adoption requires many moving together

Some technologies are not profitable for any one actor to adopt alone. A firm may not upgrade if suppliers, buyers, workers, logistics providers, or utilities do not also adjust. A digital payment system has little value if few customers and merchants use it. Infrastructure and private adoption can also be mutually stalled: firms may not invest when electricity and logistics are unreliable, while governments or utilities may hesitate to expand capacity when demand is low.

In these settings, government action can help coordinate expectations and investments. This may involve standards development, clustering policies, shared facilities, or support for supplier relationships. The case is strongest when the complementarity is clear and support is tied to observable upgrading, such as investment, adoption, exports, training, or productivity improvements. Otherwise, policy can end up protecting activities that remain unproductive.

5. Build learning into policy

There is no universal recipe for technology adoption. Constraints differ across sectors, regions, firms, and stages of development. They also change over time as firms become more capable, sectors evolve, and technologies become more complex. Policymakers should therefore compare their setting to similar contexts, test interventions at meaningful scale, measure adoption and productivity responses, and update policy when the binding constraint changes.

This is especially important because adoption failures are not always about low returns. Firms may underexperiment because returns are uncertain, downside risk is costly, providers are not trusted, workers bear some of the losses, or digital adoption makes activity more visible to owners, regulators, or tax authorities. In these cases, broad subsidies may be less effective than policies that reduce risk, improve information, support experimentation, or change incentives inside firms.

The overall message is simple. Technology adoption requires more than access to technology. It requires technologies that fit local conditions, complementary inputs that make them productive, coordination when many actors must move together, and incentives that make adoption worthwhile. Policy can facilitate this process, but only if it starts from the constraint that prevents adoption at scale.