An introduction
to industrial policy

While the definition of industrial policies is broad, not all government policies are industrial in nature. Many policies that seek to boost growth or employment – such as measures improving the business environment or the reduction of income tax or value added tax (VAT) rates – do not support one industry at the expense of another. Some “horizontal” policies – such as initiatives aimed at easing immigration requirements for highly skilled labour – nevertheless implicitly target a range of sectors and can thus be regarded as industrial policies. Many industrial policies target a narrow sector, such as the wind energy sector or the semiconductor sector, and these are often referred to as “vertical” policies.

Industrial policies were particularly popular in the aftermath of the Second World War, when government support for innovation and multilateral trade and finance arrangements were seen as the best way to speed up post-war reconstruction and raise living standards. At that point, industrial policies focused primarily on promoting sectors with significant spillovers to the rest of the economy. Industrial policies then fell out of favour in the 1970s and 1980s as new empirical evidence challenging their effectiveness emerged and the focus shifted to broader market-based strategies (see Box 1.1 for a brief overview of the history of industrial policies).

Industrial policies have recently become more popular again, partly because of a desire to address increasingly pressing market failures such as environmental degradation. The increased prominence of such measures reflects a realisation that markets and broad horizontal policies cannot always overcome important economic, social and environmental challenges, such as the need to speed up the transition to a green economy or ensure a guaranteed energy supply.

While industrial policies can be effective in overcoming coordination failures and promoting the creation and transfer of knowledge,2 they can be associated with high explicit fiscal costs. They may also result in high implicit costs – for example, in terms of distorting the efficient allocation of labour and capital in the market. The risk of capture by special interests is also high. Such market distortion can suppress innovation and drive up the prices of goods and services.3 Industrial policy instruments that are less distortive typically require greater administrative capacity and the capacity to raise fiscal revenue.

Given their mixed track record, it may be that the popularity of industrial policies is being driven primarily by domestic political economy considerations and rising geopolitical tensions. A succession of economic crises and growing awareness of the need to address environmental challenges have led to a desire for the state to play a larger role in the economy. Meanwhile, voters have tended to show a strong preference for subsidies over taxes.4 While industrial policies are more commonly seen in higher-income economies, they are also being deployed with growing frequency in economies with lower levels of administrative and fiscal capacity.

Increasingly, industrial policies are tending to target multiple objectives with no clear prioritisation. While these policies have traditionally tended to target economic growth and productivity, green objectives have been gaining in prominence, particularly in advanced economies5 – often in combination with a strategic desire to ensure the supply of critical materials and technology. Recent examples of industrial policies with such objectives include (i) the Creating Helpful Incentives to Produce Semiconductors (CHIPS) and Science Act and the Inflation Reduction Act (IRA) in the United States of America, (ii) the European Green Deal and the European Chips Act in the European Union (EU), and (iii) the Made in China 2025 plan (see Box 1.1). Regional development objectives have also become more important, particularly in the EBRD regions. Policymakers need to be mindful of these trends, articulating – publicly if possible, and in private at least – the key objective of each policy and building in evaluation mechanisms to ascertain whether a policy is on course to achieve its objectives or should be modified or abandoned.

In order to minimise distortion, policies can include competitive selection elements and specific end dates. While the percentage of policies with sunset clauses has risen, perhaps reflecting past experience with addiction to subsidies, policies discriminating in favour of domestic firms are widespread and the use of subsidies has increased. Where administrative capacity is low, policymakers could phase in policies, prioritise projects which fall within the remit of a single ministry and set up specialist units to oversee policy initiatives.

This chapter starts by documenting the increases seen in the number and scope of industrial policies, before examining the reasons for those trends and the benefits of industrial policies. It then surveys the changes seen in policy objectives and targeted sectors, before discussing the explicit and implicit costs of industrial policies, various features of policy design and the choice of instruments. The concluding section offers broad recommendations for the design of industrial policies with a view to maximising their benefits while minimising the associated risks and distortions, focusing on scenarios where administrative and fiscal capacity is limited.

China and the United States have implemented the largest numbers of industrial policies in the period since 2010, followed by Germany, Brazil, India, Italy, Japan, Russia, Canada, Spain, the United Kingdom and France. In the EBRD regions, the geographical spread of industrial policies is equally broad: Türkiye and Poland have implemented the most policies, followed by Greece, Hungary, Romania, Egypt, Czechia, Kazakhstan, Croatia and the Slovak Republic. (In general, economies with larger populations have implemented more industrial policies.)

The products that are targeted by each policy can be matched to data on exports and imports to get a sense of the scope of the various policies (with the decision to focus on exports and imports – rather than domestic production – being dictated by the availability of data).

In the EBRD regions, the percentage of exports that are affected by industrial policies has increased from around 10 per cent in 2010 to around 45 per cent in 2022 (see Chart 1.2), with similar patterns being observed for imports. These estimates represent upper bounds, since product matching is carried out using the first two digits of the Harmonized System code (referred to as “HS2 codes”) – a level at which wines, spirits and vinegar are grouped together, for example. Nevertheless, they point to a clear upward trend in the scope of industrial policies. This trend can also be seen in various other countries around the world. In 2019, for example, the United States introduced localisation into public procurement as part of its “Buy American” policies (see Box 1.3 on localisation rules). And in 2015 Germany introduced trade finance support for a range of goods, including wind turbines and vessels. In China, such policies often target vehicles and machinery. Some economies have small numbers of industrial policies, but they affect a substantial share of the economy. This is true, for instance, of the subsidies supporting the oil industry in Azerbaijan or the agricultural subsidies in Tunisia.

State aid for firms (including subsidies, grants and concessional finance) increased sharply during the Covid-19 crisis, rising from around 0.8 per cent of GDP to 1.5 per cent in EBRD economies in the EU and rising from 0.5 per cent to 2.2 per cent in Germany, Italy and the Netherlands, according to the European Commission’s State Aid Transparency Public Search tool. While state aid fell to 1.2 per cent of GDP in 2023 in those three advanced European economies, it averaged 1.6 per cent of GDP in EBRD economies in the EU.

While total spending on industrial policies may seem modest, it is comparable to government expenditure on education, which averaged 3.7 per cent of GDP globally in 2022 according to data from the United Nations Educational, Scientific and Cultural Organization (UNESCO).10 Meanwhile, the Marshall Plan payments which supported reconstruction in Europe after the Second World War were equivalent to approximately 2 per cent of US GDP and roughly the same share of the collective GDP of the recipient countries.11

The fact that higher-income economies tend to have more industrial policies may reflect the demands that such policies impose on administrative and fiscal capacity. Industrial policies often require deep knowledge of markets and the technology used by firms, regular collection and analysis of data, and other forms of technical expertise.12 Perhaps unsurprisingly, economies with higher levels of administrative capacity – as captured by the Varieties of Democracy (V-Dem) indicator assessing the rigour and impartiality of public administration – have implemented larger numbers of industrial policies (see Chart 1.4).

Industrial policies are also more common in economies with greater fiscal capacity, reflecting the high fiscal cost of those policies (see Chart 1.5). Here, fiscal capacity is measured using the V-Dem indicator assessing sources of fiscal revenue. This score is low for economies which are unable to raise revenue and reliant on external financing or ownership of assets (such as commodities), and higher for those where the state derives most of its revenue from the taxation of consumption, income/profits or capital. Even controlling for the logarithm of GDP per capita, economies with greater administrative and fiscal capacity tend to implement more industrial policies. These variables tend to have greater explanatory power than alternative measures of the quality of economic institutions, such as indicators capturing the rule of law or spending capacity.

Notwithstanding these correlations, industrial policies are also increasingly being implemented in environments where administrative and fiscal capacity constraints may be more binding.

While market failures call for some form of state intervention, interventions that do not involve using industrial policies may be less distortive or more efficient.13 For instance, environmental externalities could be addressed using carbon taxes, while coordination failures could be mitigated by bringing the various parties together, facilitating simultaneous investment commitments or introducing public guarantees. Indeed, industrial policies tend to be used more widely in economies with less financial depth, suggesting that access to market financing and government intervention may be substitutes.14 Various factors may have tilted the response to market failures in favour of industrial policy.

Political economy considerations are also increasing demand for industrial policy-type solutions. For instance, although well-designed taxes may be a more efficient means of addressing externalities, voters typically prefer subsidies to taxation, as the cost of subsidies in terms of future taxes is less salient. The results of the LiTS IV survey corroborate this.16 Indeed, studies suggest that industrial policies tend to be used more widely before elections and when economic conditions are weaker.17

Increasingly, that larger role for the state in the economy is co-existing with the presence of a handful of large and growing firms, both in emerging markets and in advanced economies.18 For instance, the combined revenue of the top 15 listed firms in emerging markets doubled as a share of GDP between 2005 and 2022, as discussed in Chapter 4. As many large firms in emerging markets share owners, the level of concentration may be far higher when ownership and personal connections are taken into account.19 At the same time, the 50 largest US firms accounted for 48 per cent of the total sales of the top 500 US firms in 2015, up from 41 per cent in 1999.20

That combination of a large role for the state in the economy and the presence of a handful of large firms may lead to increases in both demand for and the supply of industrial policies. Governments may decide that focusing policy on a few firms that are becoming increasingly important for the overall economy is an attractive option. The size of those firms may, in turn, enable them to successfully lobby governments for various subsidies. Indeed, industrial policies appear to be more common in country-years where the 15 largest listed firms account for a larger share of GDP when controlling for the logarithm of GDP per capita, population and public-sector employment as a percentage of total employment in the preceding year (based on data from Worldscope; see also Chapter 4).

Geopolitical fragmentation has become increasingly visible in global patterns of trade and greenfield investment.21 Foreign direct investment (FDI) between economies that can be seen as geopolitical rivals has declined rapidly as a share of total instances of FDI, especially since 2022, with “connector economies” (which have navigated geopolitical rivalries and benefited from geoeconomic fragmentation) accounting for a growing share (see Chart 1.7).22

Explicit responses to other economies’ industrial policies are common. In the course of 2022 and 2023, for instance, the European Commission published several legislative proposals in response to the United States’ CHIPS and Science Act and the IRA. The EU’s response to the IRA has mainly centred on the European Green Deal, which aims to support Europe’s green industries by simplifying the regulatory environment, increasing access to finance, improving skills and ensuring open trade. The EU has also responded directly to the IRA by relaxing state aid rules (extending the temporary loosening of state aid regulations that was adopted in response to Russia’s war on Ukraine). While the IRA is based primarily on tax relief, the EU’s approach relies on the direct disbursement of funds – which is, in turn, financed by an increase in debt. Similarly, in March 2023 Canada outlined a new industrial strategy called “A Made-in-Canada Plan”, aiming to attract new investment, create high quality jobs and support the green economy, seeking to match the incentives provided by the United States’ IRA. The Canadian government has indicated that this financial assistance will be revised if the incentives in the IRA are reduced or revoked.23 Also in response to the IRA, South Korea implemented new measures in early 2023 to support various segments of the electric vehicle supply chain (adjusting its own consumer tax credits for electric vehicles and fostering investment in technology and plants), as well as supporting South Korean battery makers.24

Total cross-border spillovers from industrial policies can significantly exceed the direct impact on immediate suppliers. For instance, a US$ 100 million increase in the production of electrical equipment in Germany, the United States or Japan is estimated to directly boost global production of various inputs by US$ 9 million, but increase production by US$ 39 million when the entire supply chain is considered. While most of these spillovers accrue to economies within Bloc 1 (defined here as North America, the EU and other economies that imposed sanctions on Russia in 2022), that increase in the production of electrical equipment in Bloc 1 is reliant on significant inputs from Bloc 2 (with those Bloc 2 inputs accounting, in value terms, for 13 per cent of direct inputs and 18 per cent of total inputs across the entire supply chain), as highlighted in last year’s Transition Report.26 The same is true of efforts to scale up production of electrical equipment in China or Russia, where the estimated reliance on inputs from economies in Bloc 1 is, if anything, even greater. This shows that “de-risking” the entire supply chain in the context of rising geopolitical tensions – which is, increasingly, a stated objective of industrial policies, as discussed in the next section – is a complex endeavour and can potentially be very costly.

Thus far, those differing objectives of industrial policies have received little attention in economic literature, beyond qualitative studies and research focusing on a few select economies. This section aims to fill that gap using a novel dataset. That dataset uses large language model (LLM) processing to codify the objectives of industrial policies on the basis of the descriptions available in the GTA database (see Box 1.2 for details).27

In this exercise, the objectives of industrial policies are grouped together in five clusters:

• growth and productivity, which includes policies that target innovation (by supporting startups, for example), foster exports and investment, and support investment in human capital and infrastructure (with a focus on the intensive margin, as opposed to the creation of jobs)
• creation of employment, which includes policies supporting small and medium-sized enterprises (SMEs), which tend to target employment rather than productivity (as discussed in Chapter 4)
• security of supply (ensuring energy or food security, for example) and protection of strategic industries such as aerospace, defence and semiconductors
• protection of the environment, which includes policies supporting the green transition and encouraging greener practices across the economy (in organic agriculture, for instance)
• regional development, which includes policies aimed at reducing intra-country disparities and supporting disadvantaged regions (for instance, “levelling-up”; see also Chapter 3).

Pursuing multiple objectives may dilute the effectiveness of policies. These clusters of objectives are not mutually exclusive, but one objective does not necessarily support another, so there are likely to be trade-offs between different objectives supporting the same sector. For instance, maximising value added and spillovers to the rest of the economy (through support for high-tech sectors, for example) may be different from maximising employment (through support for labour-intensive sectors such as agribusiness or tourism or support for SMEs, for instance). The latter may be a higher priority in economies where labour forces are growing rapidly. Similarly, a policy that seeks to accelerate the transition to a green economy may prioritise cheaper imports of products such as solar panels or electric vehicles if protecting domestic firms or jobs in those sectors will result in higher prices and slower adoption of green technologies. “Buy American” or “buy European” clauses may achieve geopolitical objectives, but at the cost of reduced efficiency in the short to medium term. Similarly, tariffs on solar panels may help to maintain public support for the green transition, but reduce the speed of that transition owing to the far higher cost of imported panels.28

In some cases, different policymakers target different objectives. In this kind of scenario, central coordination is necessary to prevent the objective of one ministry or region from conflicting with and working against the objective of another ministry or region.

Being more explicit about the aims of industrial policies – ideally focusing on a single objective, but at the very least acknowledging trade-offs between them and establishing a formal hierarchy of objectives – will make it easier for policymakers to acknowledge policies’ failures while taking credit for their successes.29

In fact, in the period 2010-22, 60 per cent of industrial policies in advanced economies and around 75 per cent of industrial policies in the EBRD regions and other emerging markets had multiple objectives, and multiple-objective policies have increased in recent years as a percentage of total policies (see Chart 1.9). More than 10 per cent of policies have three objectives or more, with such policies accounting for a growing share of total policies. As a result, the figure for the average number of objectives associated with an industrial policy has increased from 1.7 in 2010 to almost 2 in 2022.

These results echo the findings of Meckling and Strecker (2022), who found that 65 per cent of green innovation policies had at least one additional objective besides tackling climate change.30 As environmental policies often entail concentrated losses in the present and diffuse benefits in the future, governments often tie climate-related measures to job creation or other benefits as part of “green bargains”.

Industrial policies targeting regional development are more common in the EBRD regions than in advanced economies. Many of those policies support agribusiness, and they often target employment creation as well (see Box 1.4). Others (in Estonia, for instance) aim to promote FDI and the upgrading of value chains in the localities that have been most affected by the transition away from mining and carbon-intensive manufacturing.

Industrial policies targeting employment and SMEs saw a strong temporary increase during the Covid 19 crisis, with many governments implementing massive support programmes to stabilise the economy and reduce the social costs of unemployment and underemployment. Some of these policies also included measures seeking to address climate change.

Consistent with the examples above, industrial policies combining environmental objectives with a desire to ensure security of supply in strategic industries have gained in prominence in recent years, as have policies combining employment and regional development objectives (see Chart 1.12).

Where possible, policies should build in mechanisms allowing the evaluation of their success, based on the main objective of each policy. Such evaluations could enable policymakers to modify or abandon policies that fail to live up to expectations at an early stage. Building in policy evaluations upfront may help to alleviate the challenge of exiting unsuccessful policies. Indeed, past experience with industrial policies suggests that letting losers go may, in fact, be more difficult than picking winners.31

While discarding unsuccessful initiatives should, in principle, be a less demanding task than picking winners, political economy considerations may result in governments persisting with ineffective policies. For instance, support for infant industries may be maintained for longer than is necessary, with firms lobbying for the continuation of support rather than focusing on improvements in productivity. Human psychology may also lead to a reluctance to dispense with poorly performing policies, as people tend to be particularly averse to losing what they have and admitting failure.32 Recent research finds similar patterns for professional asset managers: they outperform the market considerably when picking stocks to invest in, but they perform poorly when it comes to exiting (performing worse than if they had chosen their exit points at random).33

Empirically, industrial policies appear to target a mixture of goods – both with and without comparative advantages (see Chart 1.13). Higher-income economies tend to focus more on existing advantages, perhaps because their more diverse skill base makes it easier to shift to producing and exporting similar products (with those economies being said to have a dense product space).35 For instance, the UK Innovation Strategy prioritises seven key technology families where the United Kingdom already has globally competitive research and development (R&D). In contrast, lower-income economies are more likely to target “aspirational” sectors.

The choice of instrument depends on the sector, the objective of the policy and the structure of the market, as well as the government’s administrative and fiscal capacity to deploy the instrument while minimising associated risks and distortions. Many of these instruments may implicitly or explicitly involve picking winners – firms or sectors that receive government largesse. This has direct fiscal costs (in terms of the current or future taxation that is needed to pay for the subsidies) and may, in turn, starve other firms of labour and capital. If subsidised firms have a lower return on labour and capital, this introduces significant distortions into the economy, negatively affecting the dynamism of business and productivity growth.

Some industrial policy instruments are particularly distortionary by nature. For instance, schemes involving government handouts – whether explicit (in the case of subsidies) or implicit (in the case of preferential treatment for procurement or export/import licensing) – are prone to capture by special interests, particularly when economic institutions are weaker. Economic distortion is greater when innovation grants are pocketed in the form of shareholder profits with little or no innovation taking place, and it is smaller (or absent entirely) when firms use those grants to achieve the maximum returns on capital. Even if corruption is not a concern, picking winners wisely still requires strong administrative capacity on the part of governments.

In general, bans (such as import/export bans), quotas and licensing requirements tend to be the most distortive (see IMF, 2024b), as they affect firms across the board with little room for adjustment by the affected firms. Perhaps reflecting this, as well as past experience with such policies, those instruments are used fairly infrequently (see Chart 1.15, where the respective bubbles are relatively small). However, those instruments are relatively easy to implement and have low direct fiscal costs. As a result, they tend to be used by economies with lower levels of bureaucratic quality and revenue-raising capacity.

In contrast, instruments such as trade finance, incentives to localise value added and localisation requirements in public procurement are associated with relatively high levels of administrative capacity. Accordingly, they are more common in advanced economies than in the EBRD regions (and tend, more generally, to be used more frequently in richer economies). Grants, meanwhile, are commonly implemented in economies with lower bureaucratic quality but tend to require relatively high levels of revenue-raising capacity – the most notable deviation from the diagonal in Chart 1.15, where the average administrative and revenue-raising capacities that are associated with policy instruments are otherwise broadly aligned.

In order to further examine the question of which instruments tend to be used most often in which economies, Chart 1.16 calculates, for each instrument, average control of corruption and government effectiveness scores on the basis of the World Bank’s Worldwide Governance Indicators (WGIs) for all countries that implement policies using the instrument in question. Control of corruption scores capture perceptions of the extent to which public power is exercised for private gain, while government effectiveness scores capture perceptions regarding the quality of public services, the quality of the civil service (and the extent of its independence from political pressure), the quality of the formulation and implementation of policies, and the credibility of the government’s commitment to such policies.37

Instruments such as trade finance and financial assistance abroad are typically used by economies with high scores for control of corruption and government effectiveness. Local content requirements and incentives, however, are typically used by economies with lower scores.

The instruments that are widely considered to be the most distortive tend to be used in economies with very low scores for the effectiveness of government and control of corruption (see the bottom left corner of Chart 1.16), where government policies in general tend to be distortive and prone to capture.

Market competition tests can improve accountability and leverage technology spillovers from integration.39 Investment promotion policies that seek to leverage FDI can be inexpensive, non-distortive and effective40 in facilitating knowledge transfer and re-shaping countries’ comparative advantages and export structures,41 including by boosting the complexity of exported products.42

Closed policies account for over 90 per cent of all industrial policies implemented between 2010 and 2022 (see Chart 1.17).44 The closed nature of most industrial policies increases the importance of strengthening competitive elements in the allocation of funding at domestic level – which, in turn, makes the implementation of those policies more challenging in terms of the administrative capacity that is required to maximise their benefits while limiting distortion.

Industrial policies in the EBRD regions and other emerging market economies tend to be broader in scope than their equivalents in advanced economies (see Chart 1.18). In the EBRD regions, 30 per cent of industrial policies target a single HS6 code, compared with around 40 per cent in advanced economies. Similarly, in the EBRD regions, over 10 per cent of industrial policies are very broad, targeting six HS2 codes or more, compared with less than 5 per cent of industrial policies in advanced economies.

The fact that firm-specific policies make up a larger percentage of total industrial policies in higher income economies is, in part, a reflection of the choice of instruments. Policy instruments such as trade finance, financial assistance in foreign markets and incentives to localise value added are more likely to target specific firms, and those instruments are more commonly used in advanced economies. Indeed, at a global level, more than 80 per cent of all industrial policies that use such instruments are firm-specific. At the same time, when using those particular instruments, higher income economies are also more likely to target specific firms than lower-income economies.

Higher incidence of firm-specific policies in a particular economy might also reflect a focus on global competition between economies, rather than competition within the economy in question.47 In many economies, a small number of “superstar” firms are now shaping their countries’ trade patterns and comparative advantages. For instance, Krka, a pharmaceutical company in Slovenia, accounts for around 3 per cent of the country’s exports and GDP. Similarly, at Nokia’s peak in the mid-2000s the firm accounted for 25 per cent of Finland’s total exports and 4 per cent of Finnish GDP.48 While firm specific policies are still used less widely in the EBRD regions than in advanced economies, they have become more common over time (see Chart 1.17).

Firm-specific policies in emerging markets often cite security of supply objectives (with such goals accounting for over 90 per cent of firm-specific policies in China and 60 per cent in Türkiye, for instance). In contrast, non-firm-specific industrial policies are more likely to have employment objectives, a trend that can be observed in advanced economies and emerging markets alike.

In the EBRD regions, Lithuania, Montenegro, Serbia and Slovenia all have programmes promoting collaboration between universities and the private sector in support of innovation, while Armenia and Mongolia have programmes focusing on marketing and branding.50 The EBRD’s Investment Councils – platforms for public-private dialogue where businesses and policymakers can come together to tackle investment climate-related challenges – can also be regarded as an example of “soft” policies. This type of approach can be particularly advantageous in situations where administrative capacity is low and the risk of capture by special interests is high.51

In addition to the establishment of built-in evaluation mechanisms to monitor progress towards policy objectives, incorporating sunset clauses (that is to say, automatic end dates) in policies at the outset may make it easier to phase policies out. Globally, around one-third of all industrial policies have an end date – one that has been announced publicly – according to information in the GTA database. End dates tend to be more common for (i) policies with a higher fiscal cost, such as financial grants and state aid, and (ii) policies targeting regional development (with 45 per cent of regional development policies having sunset clauses).

The incorporation of end dates has become more common in recent years, both in the EBRD regions and in advanced economies (see Chart 1.17). This development, which has coincided with the rise in more addictive instruments such as subsidies, is welcome and may, to some extent, indicate that countries are learning from decades of past experience with industrial policies. Nonetheless, the risk of policies being rolled over irrespective of their merits remains, even if the default option is for those policies to expire.

In the 19th century, governments increasingly prioritised infrastructure. The United Kingdom invested heavily in the expansion of railways and steamships and began liberalising trade, while the United States constructed railways, canals and telegraph networks.

In the late 19th century and the early 20th century, industrial policies evolved in order to strike a balance between state intervention and reliance on market forces. The United States, for example, implemented antitrust laws such as the Sherman and Clayton Acts to prevent monopolies and ensure competition, while the United Kingdom supported SMEs using tariffs, subsidies and tax breaks in order to promote economic growth. France, meanwhile, nationalised key industries and supported selected strategic sectors (including aviation), and Latin American countries such as Argentina, Brazil and Mexico used import substitution industrialisation (ISI) to reduce their dependence on foreign goods, deploying quotas, tariffs and subsidies in order to support their domestic industries.66

Industrial policies became particularly popular in the aftermath of the Second World War. At that time, there was a broad consensus that the provision of public goods and services, government support for innovation, and multilateral trade and finance arrangements were the key to speeding up post-war reconstruction and raising living standards.67 This policy consensus, which lasted around three decades, was also endorsed by developing economies, with notable examples including the success of industrial policies in the “miracle countries” of East Asia and the failure of import substitution in Latin America. Prominent examples of policies in East Asia and Latin America included support for the South Korean steel company POSCO and the Brazilian aircraft manufacturer Embraer.68 Emphasis was often placed on addressing coordination failures and promoting sectors that could supply essential inputs to other industries, thereby fostering complementary demand across the economy.69

In the 1970s, industrial policies started to lose their appeal, with mainstream economics stressing the distortion that arose from state intervention and documenting the failures of governments that had sought to rectify market failures. In 1986 US President Ronald Reagan famously remarked that the nine most terrifying words in the English language were “I’m from the government and I’m here to help”. Nevertheless, many countries continued to use industrial policy instruments, albeit often in a lower-key fashion.70

The late 1990s and early 2000s witnessed a resurgence in state intervention. Industrial policy evolved to address growing concerns about market failures (including in the area of environmental protection), reflecting broader recognition of the state’s role in fostering innovation and economic development.71 The succession of major economic crises over the last 15 years, coupled with increased awareness of environmental challenges and perceptions of rising inequality within economies, have further increased demands for state intervention.72

Rising geopolitical tensions have brought strategic industries and security of supply considerations into the spotlight.73 In the United States, the CHIPS and Science Act, which was adopted in 2022, aims to strengthen competitiveness, innovation and national security in the semiconductor sector and increase the numbers of people working in science, technology, engineering and maths (STEM) sectors by using tax credits to support investment in manufacturing, sectoral R&D funding and funding for education and skills.74 Meanwhile, the IRA aims to reshape the power sector by fostering the decarbonisation of the electricity generation and electric vehicle industries using production and investment tax credits for clean electricity and energy storage.75

In the European Union, major industrial policies adopted in recent years include (i) the New Industrial Strategy for Europe, which aims to support the green and digital transitions and reduce strategic dependence on imports, (ii) the European Chips Act, which is aimed at supporting semiconductor production in the European Union and reducing dependence on external producers, and (iii) the European Green Deal, which aims to make the EU climate-neutral by 2050 through initiatives that support renewable energy use, energy efficiency and sustainable agriculture, foster significant investment in clean technologies and create green jobs, while ensuring a “just transition” for all regions and industries.

The Made in China 2025 plan was the centrepiece of the industrial strategy that China launched in 2015, which sought to shift the economy towards innovation-driven production of higher-value products and services and reduce dependence on foreign suppliers in those sectors. That programme targeted 10 industries, including next-generation IT, high-end digital control machine tools and robotics, and electric power equipment.76

Juhász et al. (2023a) used a text-based approach to measure the number of industrial policies at a global level over time. Their algorithm categorised policies recorded in the GTA database over the period 2009-22 using supervised machine learning. By zooming in on the objectives of policies, the algorithm was able to categorise commercial policies as “industrial” or “non-industrial”.77 In the absence of information that unambiguously identified a policy as industrial, the policy in question was left unclassified. Thus, the number of policies classified as “industrial” by the algorithm is likely to constitute a lower-bound estimate of the total number of industrial policies.

Unclassified policies accounted for 43 per cent of all policies globally and 69 per cent of policies in emerging market economies (with figures as high as 100 per cent being recorded in some economies in Central Asia and the Caucasus). In Estonia, Germany, Latvia, the United Kingdom and the United States, by contrast, less than 20 per cent of policies were unclassified.

In order to examine industrial policies in the EBRD regions and other emerging markets, the Juhász et al. (2023a) dataset was expanded using ChatGPT and a finetuned prompt. In order to develop the prompt, stratified random samples of policies at the instrument-country-year level were coded manually for the United States and Kazakhstan. This manual coding was then compared with ChatGPT classifications obtained using a minimal prompt. It became apparent that the minimal prompt had resulted in ChatGPT repeatedly excluding certain types of policy, including policies in the transport, infrastructure and agriculture sectors, as well as policies used by various government agencies to support firms (such as guaranteed loans issued by the Export-Import Bank of the United States to support Boeing’s exports). The prompt was then adjusted to account for those discrepancies, resulting in the following prompt:

“You are an expert in industrial policy. You are very familiar with such policies, including but not limited to infrastructure, transportation, agriculture, manufacturing, etc. Specifically, if a policy clearly aims to support a domestic company’s export behaviour by granting loans to a foreign company (such as policies used by the Export-Import Bank of the United States to support US companies, and other similar policies used by similar government agencies in different countries), this policy should also be considered an industrial policy. Given the below policy text, is it an industrial policy? Please think step by step. Your answer should start with ‘Yes’ or ‘No’, and then the next paragraph should provide a concise explanation.”

As a result of this adjustment to the prompt, around 55 per cent of the policies that had previously been unclassified were coded as industrial policies. Thus, the ChatGPT classification now matched the manual coding for 94 per cent of the US policies and 85 per cent of the Kazakh policies. More generally, when looking at trends in terms of numbers of policies and distributions by instrument and by product targeted, the patterns in the expanded dataset were very similar to those in the Juhász et al. (2023a) dataset for advanced economies.

A number of entries were identical in terms of the policy instruments listed and concerned the same country and year, but had slight differences in their descriptions as regards the level of detail on objectives. In the Juhász et al. (2023a) dataset, those differences could result in policies being placed in different categories, but the same was not true of the ChatGPT classification. For example, there were two very similar descriptions of policies in the same country and year: “a ban on fuel exports” and “a ban on fuel exports, with the objective of protecting domestic producers”. In the Juhász et al. (2023a) dataset, the former was unclassified, but the latter (which included additional information on the objective of the policy) was classified as an industrial policy. In the expanded dataset, ChatGPT classified both as industrial policies.

The GTA database, which was the basis for the Juhász et al. (2023a) dataset, the expanded dataset used in this chapter and the dataset used by Evenett et al. (2024), is updated on a rolling basis when new information becomes available. Some policies are added when they are announced, while others are backfilled. Backfilling can mean that data for more recent years are less complete. For instance, at the time of writing, in 2024, authorities have only had two years to report policies that were implemented in 2022, but they have had 10 years to document state intervention that took place in 2014.78 The amount of backfilling varies by country. It is particularly common with China, reflecting the decentralised implementation of many policies.79 In order to make the flows of industrial policies comparable over time, analysis in this chapter generally follows the approach adopted in Juhász et al. (2023a) by applying a “same-year restriction” – that is to say, it only considers policies that are announced and included in the GTA database in the same calendar year.80

The expanded dataset also used ChatGPT to code the objectives of industrial policies on the basis of their descriptions in the GTA database. First, manual analysis of the descriptions of 65 industrial policies in various economies yielded 11 commonly used objectives.81 These were then grouped together in five main clusters: “growth and productivity” (which included measures to enhance competitiveness, investment promotion, export promotion and trade facilitation, infrastructure development, human capital development, and R&D and innovation), “employment” (which included employment creation and support for SMEs), “security of supply and strategic industries”, “environment” and “regional development”. ChatGPT was then asked to identify whether each policy pursued one of the 11 disaggregated objectives and provide a justification for its answer.

Localisation rules accounted for 15 per cent of all industrial policies in 2022, up from 9 per cent in 2010. Around 41 per cent of those rules relate to localisation in public procurement, 32 per cent provide incentives to localise value added, while 16 per cent are local content incentives. They are used in a wide variety of countries, from India to the United Kingdom and from Brazil to Germany, and cover a broad range of sectors. Examples include local sourcing requirements attached to grants for developing local broadband infrastructure in the United States, export rebates encouraging the use of local components in textiles and manufacturing in Egypt, local content requirements for subsoil operations in Kazakhstan and local operations requirements for the installation of 5G networks in Türkiye.

Like many other non-tariff measures, localisation rules are motivated by infant industry arguments asserting that less-established local producers in various value chains require state intervention in order to grow.82 They are more likely to target employment creation than other industrial policies (with 53 per cent of localisation policies having that as a stated objective, compared with 24 per cent for industrial policies in general). In turn, security of supply objectives are less common with localisation policies than with industrial policies in general (with those objectives being observed for only 31 per cent of localisation policies, compared with 42 per cent for industrial policies in general).

Localisation rules can help to deepen supply chains and upgrade technology. Such requirements may incentivise companies to reach out to existing or new local suppliers – and those suppliers, in turn, may be in a position to adopt the latest technologies, leveraging the scale of the new market open to them and benefiting from training provided by large off-takers. That was the case in Norway, for example, following the discovery of offshore oil and gas.83

However, such requirements can also increase the cost of domestic production and impose higher prices on consumers by shielding domestic producers from competition.84 If the right skills and incentives are not present, such requirements may create excess profits for firms supplying substandard products at inflated prices and limit imports of the latest technologies, thus undermining the development of the very industries that the local content requirements were intended to support.85 This can result in the multiplication of production facilities (which is not economically rational and entails higher production costs), the loss of jobs elsewhere in the economy, and lags in the introduction of new technologies and practices.

Localisation rules are likely to be less distortive where (i) domestic markets are large, allowing local producers to scale up quickly and subsequently expand abroad using knowledge they have acquired domestically, (ii) the goods and services that are sourced locally are produced competitively (so would probably be used even in the absence of such requirements), and (iii) goods and services are non-tradeable.

In order to minimise the distortion associated with local content requirements, it is important to provide an unambiguous definition of what constitutes local content and ensure that policy ambitions match the economy’s technical potential and resource capacity.86 As with other policies, it is important to monitor whether these rules are on course to achieve their stated objectives (in terms of boosting employment or increasing value added, for instance). Policy accountability could be further enhanced by creating a dedicated independent government authority responsible for monitoring local content and ensuring that eligible firms have equal opportunities to apply and compete for contracts with local content rules.

Calibrating such requirements and gradually phasing them out as local producers become internationally competitive relies on industry regulators being highly independent and highly professional. Thus, governments with significant administrative capacity have far more policy options than those with more limited capacity.87

Notable alternatives to conventional local content instruments include joint ventures with foreign firms, programmes supporting vocational training, and measures incentivising R&D.88

Industrial policies targeting agribusiness are more likely to have regional development objectives and employment objectives than industrial policies in general (with those objectives being observed for 46 per cent and 27 per cent of industrial policies targeting agribusiness respectively). The recent rise in industrial policies targeting agribusiness in the EBRD regions and other EMDEs has been driven by policies with environmental and regional development objectives (primarily relating to sustainable agriculture), while food security objectives have declined in prominence.

In the EBRD regions, industrial policies targeting agribusiness typically involve financial grants, subsidies, import tariffs and mechanisms aimed at stabilising the prices of agricultural commodities. In advanced economies, by contrast, trade finance, production subsidies and state loans are common features of such policies, while price stabilisation mechanisms are rare.

For instance, “Buy American” provisions introduced for purchases of final goods in the United States are estimated to have created up to 100,000 jobs, at an estimated cost of between US$ 111,000 and US$ 238,000 per job, with the cost per job rising over time.90 Under these policies – the origins of which date all the way back to 1933 – goods purchased by the US federal government must be manufactured in the United States, with local US content totalling at least 50 per cent, unless specific waiver conditions are met.

In contrast, the United Kingdom’s Regional Selective Assistance business support scheme, which offers investment subsidies to firms in depressed areas on the condition that they create or protect jobs, can be regarded as a relatively low-cost intervention, creating jobs at an estimated cost of US$ 6,300 per job.91 Increases in manufacturing employment under that scheme have been observed primarily for smaller firms with fewer than 150 employees and have been associated with lower levels of local unemployment.

Meanwhile, China’s support for its shipbuilding industry over the period 2016-23 (including both direct fiscal outlays and forgone revenue) is estimated to have totalled CNY 550 billion (equivalent to 1 per cent of its 2013 GDP). Entry subsidies (such as the option to purchase land at a discounted price) and production subsidies attracted many firms and increased China’s global market share by 40 per cent. Three-quarters of that increase is estimated to have come at the expense of producers in other countries, with the remaining quarter adding to the global supply of vessels. As a result of that support, the net profits of domestic producers rose by CNY 145 billion and consumers worldwide saved CNY 230 billion on their purchases of ships (relative to the amounts they had been willing to pay). The combination of those two effects fell short of the total value of subsidies, and industry profits failed to increase in the long run.92 Entry subsidies, which accounted for 60 per cent of total spending, attracted large numbers of inefficient producers and resulted in excess capacity.

Industrial policies can also create significant distortions in other industries, whether in upstream sectors that are supplying the target industry or in downstream sectors that are using its products. For instance, many economies have implemented industrial policies with the aim of boosting their steel industries, with examples including production subsidies in Argentina in the 1970s, government equity injections in Belgium from 1979 onwards, special export tax rebates in Brazil between 1977 and 1996, and debt forgiveness for steel producers in Germany in the 1990s. Quotas and high import tariffs are common across the board. While they are in place, such policies typically support local production of steel and associated employment; however, a study of such policies spanning 21 economies documents a sharp drop in the export competitiveness of local industries that use steel as a major production input, with larger adverse effects found in lower-income economies.93