Infrastructure finance and pension funds

The problem of intermediating long-term capital to infrastructure has been a subject of intense debate and policy engagement. Most of the discussion is confined to creating the supply of insurance and pension fund capital and developing deep and liquid bond markets. 

Instead, this post will examine how the existing supply and demand sides can be brought together in an institutional partnership to catalyse a new model for the flow of long-term capital to infrastructure sectors. It’ll do so in the context of pension funds. 

This idea should be seen alongside others on infrastructure finance that I have blogged about. This post lists a few stylized facts about the infrastructure financing market and lays down a comprehensive infrastructure financing strategy for India. I blogged here that India could take the lead in demonstrating how DFIs can work on both the demand and supply sides to de-risk infrastructure projects and crowd in long-term capital respectively. I blogged here on how to reduce the cost of capital for foreign investments in the infrastructure sector, here on how to revise the credit rating framework in general, and here on de-risking and lowering the cost of capital of bank loans and de-risk the use of guarantees to crowd-in bank financing. All of these ideas are consolidated in this long paper. 

Institutional investors like Pension Funds (PFs) can invest in infrastructure either directly (in infrastructure companies or specific project entities) or indirectly (through infrastructure funds). These investments can take the shape of listed and unlisted equity, or debt (subscribe to bonds issued by companies or projects), and can be in greenfield or brownfield projects. 

The OECD’s annual reports that analyse the investments made by large pension funds (LPFs) and Public Pension Reserve Funds (PPRFs) provide several important insights. This and this are the reports for 2022 and 2023 respectively. 

At the outset, it’s worth reminding that infrastructure forms around just 2-3% of the Assets Under Management (AUM) of pension funds globally. 

As seen above, infrastructure investments by LPFs are generally channelled through unlisted funds as equity. 

However, PFs in developing countries appear to prefer bonds and listed equity. The three Indian funds, among the largest in the world, are invested in infrastructure almost completely through bonds and have no unlisted equity exposure. It’s most likely that Indian PFs continue to invest in infrastructure almost completely through bonds. The Israeli and Latin American funds appear to prefer both bonds and unlisted equity. 

The sectoral exposures are almost completely in energy (conventional and renewable), transportation, and IT and communications, across countries. Social and urban utilities attract very little interest. 

All PF investments in infrastructure flow to brownfield assets. Greenfield investments by PFs are negligible. This is understandable given the significant construction risks associated with infrastructure assets. 

Most of PF investments are either domestic or confined to North America and Europe.

In terms of costs, there are five distinct approaches - internally managed (by an in-house investment team), private open-ended infrastructure funds, limited partners in closed-end funds with a fixed holding period, co-investment (with external management), and fund-of-funds. An analysis of the CEM database of 1132 defined benefit PFs over 1991-2018 period and covering $9.9 trillion AUM in 2018 (representing almost 25% of all global PF assets) shows the evolution of these approaches. 

It can be seen that PF’s investing as LPs in infrastructure funds has emerged to become the most common investment approach, rising from 15% in 2007 to about 41% among all PFs. It’s also seen that co-investment approach has been rising. 

The investment costs associated with each approach shows that internal and co-investments are the cheapest approach, while LPs in closed-end funds and fund-of-funds (these together form the standard unlisted infrastructure funds of today) are the most expensive. 

There’s an almost 150-200 basis points difference between the costs associated with internal and co-investment approaches and the traditional infrastructure funds route. This would apply not just to PFs, but all other institutional investors investing in infrastructure. It’s for this reason that the Canadian PFs have developed the institutional capacity for internal management of its investments and have generally avoided the infrastructure funds. 

This is an excellent primer on India’s pension funds landscape. India’s pension fund assets are just 10% of the GDP, far below that of Western countries where it exceeds the GDP. But the total amount of around Rs 22 lakh Crore (as of March 2022), is a rapidly growing corpus. The regulatory norms restrict investments in alternative assets like infrastructure to 5%, though the actual exposure in most funds is far less. This is far lower than up to 26% permitted in countries like the UK, US, and Canada.

In light of the above, some suggestions:

1. From the evidence presented above, there are a few important takeaways for PFs in India. One, only a tiny share of PF capital will flow into the infrastructure sector. Two, PFs should maximise internal or co-investments, and in particular avoid infrastructure funds. Three, PFs should avoid greenfield investments and confine their infrastructure exposure to operating assets whose revenue streams are well established. Four, in the initial years, PFs should focus on power generation and transmission, transportation (highways, railways, ports, and airports) and IT and communication sectors. Five, PFs should try to diversify away from passive investments in infrastructure bond offerings and strive to invest in infrastructure equity through appropriate structures. Six, foreign PFs are unlikely to invest in significant amounts outside the developed countries. 

2. India has a rapidly growing and massive pipeline of infrastructure projects that are attractive investment opportunities for its PFs. There’s a financial intermediation gap to channel PF capital to these infrastructure projects. Fortunately, there already experienced and credible institutional participants on both the supply and demand sides in India’s financial ecosystem who can bridge this intermediation gap. 

The country has a long history of Development Finance Institutions (DFIs) involved in infrastructure sector. Sectoral DFIs like REC, PFC, and HUDCO have been around for decades and have strong performance track-record. In recent years, DFIs like NIIF, IIFCL, and NaBFID have been created as multi-sectoral DFIs. All these DFIs should be mandated to take the lead in de-risking and catalysing a mutually beneficial and completely professional co-investment approach in infrastructure with major domestic PFs and other institutional investors like insurers. 

On the supply side, PFs like the EPFO, NPS, and APY and insurers like LIC and the four major public sector insurance companies, could be encouraged to take the lead in co-investing or investing in funds sponsored by the Indian DFIs. 

To start with as a confidence building measure, the DFIs should invite the PFs and insurers to invest in their current portfolio assets and in the less risky among their pipeline of investment opportunities. These initial investments can be used to build confidence, especially among the institutional investors, besides documenting and institutionalising processes and protcols at both PF and DFI sides in the due-diligence and processing of such co-investments or co-funding. 

3. The DFIs could, in turn, strike partnerships with banks who finance the vast majority of greenfield infrastructure assets and swap those loans into their portfolios, co-investing or co-funding with PFs/insurers. This would be an institutionalised take-out financing model which addresses multiple problems that currently bedevil the market for infrastructure finance - the higher cost bank loans could finance the risker but short-term loans of the construction phase; asset-liability mismatches faced by the banks can be bridged with the take-out of their loans by DFIs; the DFIs in turn would have access to a standard pipeline of institutional long-term capital from insurers and PFs. 

4. State-led initiatives to finance infrastructure have always been a feature of the infrastructure sector even in developed countries. DFIs have been central to the intermediation of long-term capital into the infrastructure sector across the world. 

Public policy engagement in the economy is becoming more active across the world, as seen in the surge in industrial policies especially in the developed world. In infrastructure itself, governments in several advanced countries have been trying to crowd-in long-term capital into infrastructure. A recent article in The Economist pointed to the rise of pension fund nationalism and wrote about government efforts across the world to channel pension funds to finance infrastructure investments.

Rachel Reeves, Britain’s new chancellor… on July 8th… said that she wants the country’s pension funds “to drive investment in homegrown businesses and deliver greater returns to pension savers”… Her predecessor, Jeremy Hunt, set the ball rolling by mandating that defined-contribution pension funds will have to disclose the scale of domestic investments by 2027. Other countries are joining in. Stephen Poloz, a former Governor of the Bank of Canada, is looking at how to increase pension fund investments in domestic assets on behalf of the Canadian government. Enrico Letta, a former Italian Prime Minister, recently urged in favour of an EU-wide auto-enrolment pension scheme that could be funnelled into green transport and energy infrastructure. 

5. This partnership will require long-term investment of the highest-level commitment and effort by institutional stakeholders at both the supply and demand sides. It must be supported with active co-ordination and facilitation by the respective Departments in the Ministry of Finance in the Government of India. This would require stable long-term leadership and 2-3 champions leading the cause for at least five years to lay a strong enough foundation to allow the model to take off.

If done effectively, catalysing a mutually beneficial co-investment/co-funding model between India’s DFIs and PFs/insurers can become a defining innovation in crowding in the rapidly growing pool of long-term capital to finance an equally fast-growing portfolio of infrastructure assets. It can be a model for pension funds in other countries grappling with the high costs of reliance on external fund managers like infrastructure funds but also don’t have the capabilities to manage the funds internally. 

6. This could be part of a basket of initiatives within India to shape the market landscape in infrastructure financing. Apart from DFI’s partnering with PFs and insurers to channel long-term capital, India could show the way with initiatives to de-risk bank lending to infrastructure, structure infrastructure credit guarantees through DFIs and bring together a coalition of banks to lend against them, revise the credit rating methodologies, and lower the cost of capital by revising capital adequacy ratios based on actual history of credit defaults in the sector. 

These reforms are also central to addressing the current hot-button issue of mobilising capital for climate change mitigation. India could position itself as the pioneer in plumbing reforms to attract long-term capital into climate finance. 

7. There’s already great interest among global alternative investment funds (AIFs) and intense associated lobbying to have PFs invest in infrastructure funds. But Indian PFs should avoid following the investment approach of relying on large global infrastructure funds. Instead, India should take a leaf out of its pioneering role in several areas of private investing in infrastructure, including its own PPP models in highways and power generation. As aforementioned, India should seek to create its own approach to PF investments in infrastructure. 

8. Finally, as you can see, the idea itself is simple and can be articulated, support mobilised, and got approved without much effort. The challenge is in its implementation - the commitment and effort to develop the model, pursue a few initial investments and de-risk it while also iterating to address the emerging flaws and problems with the model, monitoring and supervision to prevent the abuse of the model; and coordination and facilitation by the Government departments concerned are all critical ingredients and require long-drawn, painstaking, and detail-focused effort by a capable and committed team with visionary leadership. 

Weekend reading links

1. The new Labour government in UK wants to close down the "carried interest tax" loophole that Private Equity firms enjoy - it allows them to pay 28% instead of the regular 45% rate. Patrick Jenkins has a good article that examines the issue and proposes a compromise.

First, they make a principled point — that carried interest is not really income as reformers argue, but a genuine reward for executives, dubbed general partners in the industry, taking investment risk. If GPs invest alongside third-party investors — so-called limited partners (or LPs) — in a deal, any gain (or “carried interest”) they make should be treated as a capital gain because that is what it is. Second, the sector insists that implementing the policy as outlined would drive wealth creators and growth generators — key to Labour’s agenda of economic revival — out of the country... 

There are clear flaws in the industry’s arguments. Tax changes and differentials in this sector have not led to an exodus in the past. In 2017, Italy introduced a new regime, taxing carried interest at 26 per cent, instead of the 43 per cent of higher-rate income tax. Ireland taxes carried interest at barely half the UK rate. So far neither country has made huge inroads in attracting private equity executives... The more substantive point of principle is also moot. In many cases a private equity manager is not actually investing any of their own money, but is being gifted the “right to carry” by their employer, in much the same way as a banker might be gifted shares as part of a bonus (which is liable to income tax). There is no requirement to actually invest your own money in order to benefit from the carried interest tax break...

Reform is clearly needed, but with a spirit of pragmatic compromise. First, Reeves should follow through on her instinct that individuals must actually invest, say at a level equivalent to 1 per cent of the fund, as similar regimes in France and Italy already dictate. This would tighten the alignment between GPs and LPs, which is in everyone’s interest. Second, in order to qualify for carried interest taxation, the investment should genuinely be putting capital at risk. At present, CVC is one of very few firms where executives on a bad deal can actually forfeit money, even if the fund overall succeeds. Third, tax rates should be calibrated smartly. For cases where the threshold for real investment is met, a rate of, say, 33 per cent could be levied; if the threshold is not met, the rate would be 45 per cent. This would still be within the range of competitor jurisdictions, albeit towards the upper end. (France charges up to 34 per cent.)

2. The Skills Ministry appears to have bitten off more than it can chew with its target of having the country's top 500 companies providing internships to 10 million youth over the next five years. 

In the five years from 2019-20 to 2023-24, India added 4.47 million people to the salaried workforce, which stood at a little over 90 million in 2023-24, according to the Centre for Monitoring Indian Economy's consumer pyramids household survey. The salaried class includes managers, supervisors, white-collar professionals, clerks, industrial and non-industrial workers, and support staff. Offering internships to 10 million people in the next five years in India's top 500 companies will mean more than doubling the total employment generated across the salaried class in the past five years, or 11% of all the people employed under this category... According to Mint's study of annual reports of 94 companies on the BSE 100 index, there were 3.83 million employees at BSE 100 firms as of 31 March 2023. The data included both permanent and non-permanent employees, and excluded workers (in some cases, only permanent employees were considered due to unavailability of data)... 

As per the Union budget's announcement on Tuesday, the interns will be paid ₹5,000 per month along with a one-time assistance of ₹6,000. Companies will be expected to bear the training cost and 10% of the internship cost from their corporate social responsibility funds. The emphasis on skilling comes at a time when the Economic Survey for 2023-24 found that about one in two graduates straight out of college is not employable. “Estimates show that about 51.25% of the youth is deemed employable. In other words, about one in two are not yet readily employable, straight out of college," the Economic Survey, unveiled on 22 July, said. "However, it must be noted that the percentage has improved from around 34% to 51.3% in the last decade."

The scheme is part of five DBT schemes aimed at education, skilling and employment with an outlay of Rs 2 trillion over five years, to benefit 41 million youth. 

The difference of this scheme from the apprenticeship promotion efforts should be noted.

An apprenticeship is a structured system of training where individuals, known as apprentices, learn a trade or profession through a combination of on-the-job training and classroom instruction. Apprentices can be both graduates and non-graduates. Students who turn apprentices can also use the stipend to fund their education. The stipends depend on whether the candidate has been picked up under the NAPS (National Apprentice Promotion Scheme) or NATS (National Apprentice Training Scheme) programme. The former is meant for all trades and may take non-graduates, while the latter is largely for engineers and technical apprentices.

3. Nice snapshot of the budget numbers.  

This is the break-up of the capital expenditure proposals

I'm ambiguous on the Rs 1.5 trillion interest-free 50-year loans. State governments which had negligible or no liabilities to the central government over the years have become indebted to the central government through this scheme. It also means that the states are now ever more obliged under Article 293(3) of the constitution of India to the central government. 

And this on the important rural schemes

4. One area where India has trumped China is in its equity markets.

India formed just 6-7% of the MSCI ten years back. It has now risen to near 20%, whereas over the same time China has shrunk from over 40% to around 25%. But Indian stocks are clearly overvalued, trading at 24 times their expected earnings next year against 10 for China. 

5. Daron Acemoglu and James Robinson have an excellent article that articulates the need for the Democratic Party in the US to free itself from its current elite capture. 

When monarchies ruled... they had an elaborate justification for their legitimacy. In early modern England, it was the “divine right of kings.” In China, it was the “mandate of heaven.” It’s not just autocratic regimes that rely on such philosophies. The move toward greater popular participation also required legitimation and a new social contract. In England, that was articulated by philosophers such as John Locke, who provided the foundation of “popular sovereignty.”... That trust is largely lost. Center-left parties, which used to get a significant fraction of their votes from blue-collar workers and citizens without college degrees, now increasingly rely on votes (and money) from college graduates, professionals and managers.

That is all the more so in the United States, where the Democratic Party has gradually become associated with the preferences of the well-educated and urban voters. Democratic politicians often shy away from policies such as job guarantee programs, trade protection and stronger unions... Center-left parties need to lead the way in breaking this mold. This must start by the severing ties with tech billionaires, pharmaceutical giants and Wall Street tycoons. It is difficult to believe that a party that gets funding and ideas from the very wealthy will work hard for the well-being of the most disadvantaged. They must promote to leadership people with a background in manual work and from different educational paths. One visible and symbolic way of achieving this is to reserve a fraction of candidacies and leadership positions to individuals without a college degree. Similar strategies have been successfully used by Swedish social democrats and local governments in India... Campaign-finance reform would help, including public money for candidates that refuse support from big donors. There is also a case for introducing proportional representation voting, which can allow new parties to take up the mantle of working-class causes if the two major parties cannot get their act together. 

Note the point about severing ties with elite interests and promotion of people with a background in manual work and from different educational paths. In other words, return to truly representative governments.

6. The findings of a 3 year RCT study on unconditional income transfer among 3000 adults in Texas appears to pour cold water on UBI enthusiasts. 

The trial recruited 3,000 people in Texas and Illinois on the basis that they would be in a study receiving $50 a month or more for three years. Then a third of them were unexpectedly told they would instead receive $1,000 a month with no effect on any of their other income. The results definitively show that receiving more money provides a better life. Spending and saving rises... Time at work went down for both the recipients of the $1,000 and their partners, replaced by more leisure... The big question for the dynamic benefits of a universal income was what people would do with their additional time. Would they invest in their education, upskill, get better jobs or start businesses? The short answer was no. The findings ruled out “even small improvements” in the quality of employment and upskilling. The most that could be said was that the recipients spent some of their extra leisure time thinking about starting a business without actually doing it... Did universal support make recipients healthier than the control group? Again, the answer was no. Surveys and blood tests of recipients and the control group shows no improvement in physical health, and mental health improved only in the first year. There were more visits to medical facilities and more alcohol consumed, although also less problematic drinking.

7. Solar industry globally is facing a massive glut created by China that's driving down prices and destroying domestic solar industries in many countries, including in Europe. 

According to BloombergNEF, panel prices have plunged more than 60 per cent since July 2022. The scale of the damage inflicted has sparked calls for Brussels to protect European companies from what the industry says are state-subsidised Chinese products. Europe’s solar panel manufacturing capacity has collapsed by about half to 3 gigawatts since November as companies have failed, mothballed facilities or shifted production abroad, the European Solar Manufacturing Council estimates. In rough terms, a gigawatt can potentially supply electricity for 1mn homes. The hollowing out comes as the EU is banking on solar power playing a major role in the bloc meeting its target of generating 45 per cent of its energy from renewable sources by 2030.

The Chinese prices are so low that it requires very high tariff to level the playing field, besides constant surveillance to see whether the exports are being routed through third countries.

The Inflation Reduction Act.. has spurred almost $13bn of investment in solar manufacturing, more than six times the amount committed in the five years before the legislation... In May, it removed a tariff exemption for double-sided panels and lifted levies on Chinese imports of solar cells from 25 per cent to 50 per cent. Chinese companies now also face penalties if they are found to have dodged tariffs. US imports of Chinese polysilicon for solar panels had already been hit by a 2021 ban on products made or sourced from China’s Xinjiang because of concerns over the use of forced labour. Nevertheless, America’s solar power companies warn that the steps taken by the Biden administration this year will fail to provide enough protection... Chinese solar companies... dumping cells in south-east Asia, the source of the bulk of US imports. A solar panel manufactured in America using US-made cells costs 18.5 cents a watt, compared with 15.6 cents for a panel sourced in south-east Asia and just over 10 cents for one produced in China, according to estimates from BloombergNEF. 

Papers on extinction of vultures, business concentration, and industrial policy

This post will summarise a few recent papers.

1. Eyal Frank and Anant Sudarshan have a very good new paper that estimates the cost of the mass extinction of vultures in India caused by the introduction of veterinary medicine, diclofenac, that was passed to vultures via carcass. They use difference-in-difference strategy to compare districts with habitats highly suitable for vultures to those unsuitable, both before and after diclofenac use started. 

They write about the critical role of vultures in the Indian context

Vultures have long provided critical environmental sanitation services by quickly dispatching of carcasses in India, which was home to over 500 million livestock animals in 2019, the most in the world. However, beginning in the second half of the 1990s, and over the course of just a few years, the number of Indian vultures in the wild fell by over 95% following the introduction of diclofenac. Once numbering in the tens of millions, this decline is the fastest of a bird species in recorded history, and the largest in magnitude since the extinction of the passenger pigeon in the United States. As vultures died out, the scavenging services they provided disappeared too, and carrion were left in the open for long periods, likely leading to an increase in the population of rats and feral dogs, in the incidence of rabies, and in the transmission of pathogens and diseases such as anthrax to other scavengers, as well as to increased water pollution from carcass dumping and surface runoff.

They quantified the health costs

Our results suggest the functional extinction of vultures—efficient scavengers who removed carcasses from the environment—increased human mortality by over 4% because of a large negative shock to sanitation. We quantify damages at $69.4 billion per year. These results suggest high returns to conserving keystone species such as vultures.

2. Jaedo Choi, Andrei A. Levchenko, Dimitrije Ruzic, and Younghun Shim use firm-level historical data over 1972-2011 from South Korea to study the impact of firm concentration on the economy (also here). During this period of spectacular economic growth firm concentration rose sharply. 

This graphic captures the top 3 firms’ concentration ratio for four variables.

This measures the sales concentration (as a share of total manufacturing gross output) for the top 1, 5, and 10 manufacturing firms.

The share of the top 3 firms in each manufacturing sector in total manufacturing gross output increased from 10.1% to 28.5% between the 1970s and the 2010s. 

Their findings fly against the conventional wisdom and economic orthodoxy that business concentration is detrimental to economic competitiveness. 

To understand whether rising concentration contributed positively or negatively to South Korean real income, we build a quantitative heterogeneous firm small open economy model. Our framework accommodates a variety of potential causes and consequences of changing firm concentration: productivity, distortions, selection into exporting, scale economies, and oligopolistic and oligopsonistic market power in domestic goods and labor markets. The model is implemented directly on the firm-level data and inverted to recover the drivers of concentration. We find that most of the differential performance of the top firms is attributable to higher productivity growth rather than differential distortions… 

The top-3 firms experienced substantially higher TFP growth over this period. While they were 2.6 times more productive than other firms in 1972, they were 10.9 times more productive in 2011. They also experienced a faster increase in foreign demand. By contrast, the top-3’s relative labor and capital distortions fluctuated widely over this period but exhibited no long-run trend. The decomposition of the change in concentration shows that about 57% of the total increase from 1972 to 2011 is accounted for by sectoral reallocation – sectors with larger firms growing faster than sectors with smaller firms. The remaining 43% is driven by within-sector increases in the top-3 firm shares. Of that, about half is due to the churning of the set of the top-3 firms, indicating quite a bit of dynamism at the top of the firm size distribution over this period…

We compare the observed baseline economy to an alternative in which the top-3 firms instead exhibited the average sectoral growth rates of productivity, distortions, and export market access. Had the top-3 firms not experienced differential shocks, the top-3 concentration ratio change would have been one-fifth of that in the data, but 2011 real GDP would have been 15% lower, and the net present value of welfare over 1972-2011 would have been 4% lower. Most of the total effect is due to the top-3 firms productivity. Without the differential productivity growth of the top-3 firms, 2011 real GDP would have been 13.7% lower, and the NPV of welfare would have been 2.8% lower… Had Samsung Electronics and Hyundai Motors’ shocks evolved at the same rate as the regular firms’, real 2011 GDP would have been 6.4% and 0.35% lower, and welfare 1.0% and 0.5% lower, respectively. Thus, even a single large firm can exert a noticeable influence on the aggregate long-run outcomes… For a large majority of top-3 firms, idiosyncratic shocks contributed positively to both concentration and real GDP, echoing the aggregate results that the top-3 firms were a positive force. It appears that in South Korea, most – though not all – top firms were indeed superstars rather than supervillains.

We should be careful in interpreting the findings of this study. It informs that the South Korean economy experienced increasing business concentration across industries and it benefited the economy by increasing output. It does not say anything about why this happened or the dynamics and incentives that led to this outcome, much less that business concentration is desirable and this finding can be generalised across countries. 

In this context, it’s useful to revisit Joe Studwell’s description of the economic model pursued by North East Asian economies in their high-growth years. He writes that these countries pursued industrial policies that incentivised the development of manufacturing export capabilities and ensured global competitiveness by making companies compete with each other in exporting. The critical role of export competition is most often overlooked in the discussions on the East Asian economic miracle. The counterfactual on this is the South East Asian economies which too pursued similar industrial policies but without the disciplining force of export competition and ended up with vastly different and inferior outcomes. 

So perhaps the lesson from South Korea is that contrary to economic orthodoxy, there’s nothing inherently bad about business concentration. Instead, it can benefit overall economic competitiveness, but only if other policy elements accompany it. 

3. A working paper by Pinelopi K. Goldberg, Réka Juhász, Nathan J. Lane, Giulia Lo Forte, and Jeff Thurk examined industrial policy actions in the semiconductor industry. They contradict the conventional wisdom on the recent surge in the semiconductor industry as a deviation from the norm and instead find that governments have played a critical role in the growth of the industry. They argue that some subsidies (for new goods, green technologies, or sectors with inherent cross-border externalities) are beneficial and that while subsidies can appear to have limited immediate impact it can have significant long-term effects. 

Our model-based approach involves specifying a model of the semiconductor industry, estimating key demand and cost parameters using firm- or industry-level data, and then using the model structure to back out marginal costs for each product. In this framework, production subsidies are identified as a residual factor that reduces the marginal cost of all firms in a country, i.e., a factor that leads to lower marginal costs in a country relative to a benchmark country after accounting for all other relevant cost determinants…

First, government support has been critical for the semiconductor industry’s growth, particularly during its initial development phase. This support is evident across all major segments of the value chain, benefiting established leaders at the technology frontier, such as Korea and Taiwan, countries seeking to advance their industry, such as China and the U.S., and countries attempting to enter the market, such as India… Second, subsidies are the primary form of government support, manifesting as financial grants, state aid, tax incentives, loans and loan guarantees, and equity injections… These policies primarily target production improvement and research, development and innovation. Third, China has been a prominent user of subsidies. However, our estimates do not pinpoint China as an outlier in its subsidy use; rather, its level of support is comparable to other countries, when considering the size of its market… 

Fourth, cross-border technology transfer from more advanced to less advanced firms has been as crucial as state support for the industry’s development. This transfer has occurred through foreign direct investment (FDI), research collaborations, and technology licensing. Outside of the United States, our analysis found no instance where a domestic semiconductor industry developed without substantial foreign technology. This underscores the difficulty of developing the industry without foreign partners willing to share technology. It also explains why China has struggled to reach the technological frontier despite pursuing similar policies to other Asian economies that, as U.S. allies, had better access to foreign technology.

Fifth, policymakers aim to achieve several goals through their support of the semiconductor industry, including economic growth and development, international competitiveness, resilience, and national security. Implicit in these objectives is the belief in strong learning-by-doing effects in chip manufacturing, which lead to dynamic comparative advantage, economies of scale, and high industry concentration. In the presence of learning-by-doing, subsidies have a multiplier effect, amplifying and accelerating cost reductions as experience accumulates. Even when firms internalize the benefits of learning spillovers, so that private production is socially optimal, governments may still have valid reasons to support their domestic semiconductor industry. This support helps counter the natural tendency toward industry concentration, diversify the supply chain, and enhance its resilience. These objectives become even more critical when considering national security concerns, given that crucial segments of the semiconductor supply chain are concentrated in a few geopolitically critical countries. Learning spillovers across technologies and firms provides additional justification for subsidies.

Sixth… surprisingly, our model estimates suggest limited learning-by-doing at the firm-technology level—results which contrast with industry expectations and previous estimates… Importantly, however, our results indicate substantial within-firm spillovers across technologies, so that we estimate larger learning-by-doing at the firm level. Moreover, international spillovers across firms appear to be even larger… we hypothesize that they result from cross-country technology transfers and the close relationships between fabless firms and foundries, the latter of which may facilitate the global dissemination of knowledge. International spillovers imply potentially positive cross-border effects of subsidies. However, these effects depend on deliberate actions by market participants, meaning that firms in specific countries could be excluded. 

Iteration is critical to increase yield efficiencies and competitiveness in the semiconductor chip manufacturing industry.

Semiconductor manufacturing includes several intricate steps, such as wafer fabrication, lithography, etching, and doping, each demanding precise control. Given the complexities and microscopic scale of these processes, mastering them requires more than just theoretical knowledge; it necessitates practical, hands-on experience. The fast-paced nature of the industry means that there is a large opportunity for continuous process improvement via the accumulation of knowledge through hands-on experience in manufacturing operations.

Learning by doing can show up in several areas. First, there is scope to optimize operational processes. Regular involvement in fabrication processes helps engineers and technicians identify process inefficiencies and subtle variations, which are often missed in theoretical training. This practical experience is vital for developing more efficient manufacturing techniques, enhancing yield rates and product quality. Second, semiconductor manufacturing firms promote a culture of innovation in which production floor personnel are expected to experiment with new ideas in real-time. This hands-on adaptability is crucial in an industry characterized by rapid technological changes and shifting market demands. Third, semiconductor firms are constantly re-evaluating and improving worker skills. The advanced skills required to manage and optimize cutting-edge semiconductor manufacturing equipment are best developed through on-the-job learning, enhancing the technical proficiency of the workforce.

Nobody does such iteration better than TSMC, which itself was established by the Taiwanese government in 1987. The paper has a graphic that gives fascinating insights about TSMC’s business strategy.

As the world’s leading semiconductor foundry, TSMC mastered the art of quick iteration across its advanced technology nodes, notably its 7-nanometer and 5-nanometer processes. From Figure we observe that TSMC operates several line-widths at a given time and that the company quickly ramps-up production followed by a long period of reduced sales as a percent of total TSMC revenue. TSMC’s approach involves close collaboration with equipment suppliers and customers to speed up problem-solving and process optimization, resulting in high yield rates and enabling faster commercialization of new technologies. This focus on continuous improvement through hands-on problem solving has enabled TSMC’s market dominance and attractiveness to global clients like Apple and NVIDIA.

The paper describes how countries have pursued industrial policy

Through the 1960s, the US military bought most integrated circuits produced (72% of all integrated circuits produced in 1965). This major source of demand allowed producers to learn how to commercialize and mass produce a technology that was still in its infancy… large-scale government procurement meant that public agencies played an outsized role in shaping how the industry developed. As one example, the U.S. Army, Navy, and Air Force funded distinct approaches to solving the technical problem of soldering together more transistors before silicon integrated circuits emerged as the best solution. Importantly, the US military required that contractors share technology, a policy that promoted the flow of knowledge. This, together with rigorous antitrust policies pursued in the postwar period, allowed for the rapid dissemination of technology… 

In those economies where the technology gap was large, the state played a dominant and sometimes leading role in establishing the industry… Crucially, beyond making the industry more attractive to entrepreneurs by changing relative prices (through tariffs, subsidized loans, tax breaks etc.), each (East Asian) state was directly involved in the process of international technology acquisition, absorption, and diffusion. Put differently, developing a domestic industry seems to have required some infant industry promotion and, critically, also extensive international technology transfer facilitated by public agencies… 

As Japanese firms began entering semiconductors in the 1950s, MITI determined which firms received a technology license and, once approved, the government guaranteed that payments would be made… by coordinating technology licensing directly, the government pooled knowledge and lent the credibility of the Japanese government to firms that were unknown to the West… In Korea… beyond the usual incentive package for prioritized sectors, the government established an industrial estate for the production of semiconductors and computers, it protected the domestic market, and, most famously, it established the Electronics and Telecommunication Research Institute (ETRI)… ETRI… coordinated R&D efforts between public researchers and the private sector… the government accounted for the vast majority of the R&D spending in semiconductor technology through the 1980s, with private firms contributing a smaller share… through ETRI, the state pushed the chaebol, large conglomerates, to upgrade their production and move into more design-intensive, higher-return markets. In some cases, it did so by subsidizing and coordinating efforts by the private sector; in others, it undertook the basic research itself, working closely with the private sector… 

The Ministry of Communications (MOC) decided to develop indigenous switching technology, a hugely risky proposition that private firms would not have been willing to undertake on their own. The MOC however used its control of telecommunications public procurement to persuade foreign firms supplying the Korean market such as Ericsson to transfer some of their switching technology (a form of quid pro quo), and train ETRI personnel. ETRI then used this knowledge, as well as generous funding from the MOC, to design an electronic switching system that was better suited to developing country contexts with large rural populations… each of the three chaebol involved in semiconductors (Samsung, Goldstar and Hyundai) had multiple technology alliances for a variety of products with foreign firms (the authors list twenty-two distinct alliances for three years between 1983-1985). Moreover, each acquired US based firms “providing them with direct access to highly qualified scientists and engineers, advanced technologies and major markets”…

InTaiwan… through the newly founded Electronics Industry Research Center (reorganized later as Electronics Research and Service Organization, ESRO) the government set up a public pilot plant that developed technologies increasingly close to the frontier (in cooperation with RCA, a US-based firm). In this way, the government assumed the entirety of the risk of high-tech R&D. Over time, ESRO moved into all areas of the supply chain including computer- aided design (CAD) and IC mask making. Later, in the 1980s, it closed the technology gap further by moving into Very Large Scale Integration Technology (VLSI)... In the second stage, the government transferred the technology it had developed at ESRO to the private sector through licensing, spin-offs and technology diffusion policies. In the third stage, the government provided a variety of incentives to encourage the private sector to enter the industry. Like Korea, the government set up an industrial park and also offered a large range of generous fiscal incentives (cheap loans, tax breaks, accelerated depreciation of R&D equipment and low-cost land).

Some other interesting snippets

Today, fabless firms account for roughly 90% of all semiconductor firms and generate one-third of industry revenue… Taiwan accounts for approximately 59% of all outsourced wafers produced… U.S. foundries account for four percent of all third-party wafers produced.

4. In another paper, Réka Juhász and Nathan J. Lane examined the practical challenges (read the political economy) with the right policy choice and effective implementation of industrial policy. They discuss the political and capacity constraints associated with the pursuit of industrial policy. 

First, our framework implies that industrial policy should consider the realities imposed by the political world. A policy incongruent with these constraints is prone to government failure… In our view, government failure is not a necessary feature of industrial policy. Rather, it is endogenous and likely to emerge when industrial policies are chosen beyond a country’s political and capacity constraints. An implication is that we should be wary of unconditionally mimicking the precise policies pursued elsewhere… Successful industrial policies are ones that work within their political environment, and these particulars may vary.

Second, a political economy of industrial policy also reveals that the political challenges facing policymakers are not unique to industrial policy. We discuss how time inconsistency and political credibility may plague infant industry policy, yet they also plague much of monetary and fiscal policy, too… In some cases, the political constraints posed by second-best industrial policy are, in fact, not as steep as those faced by first-best policies… Third, working within the current political environment doesn’t mean governance constraints are immutable. On the contrary, many thoughtful industrial policies are designed with an eye to relaxing constraints. Our case study of climate policy illustrates that green industrial policy may, in fact, help relax the political constraints to future carbon pricing policies. 

Specifically, they write about the critical importance of investments in state capability to effectively pursue industrial policies. 

We argue that virtually every successful policy episode has involved substantial new investments in state capacity… South Korea and other postwar economies continually invested in bureaucratic capacity. Under General Park Chung Hee in South Korea, “[t]he developmental state was not a given, but a human artifact”, one cultivated by continual investment and political choices… Industrial policies themselves can be particularly capacity-intensive to administer; they often require deep knowledge of the markets and firms they interact with, regular data, technical expertise, and more… Bureaucratic autonomy, in particular, has been an essential feature of bureaucratic capacity in the world of industrial policy… Given the political temptations surrounding industrial policies, the autonomy bureaucracies have over policy has been vital for successful industrial policy… the pilot development agencies in East Asia evolved to have elite selection criteria, meritocratic promotion, and long, stable career paths. Bodies were thus staffed by highly trained civil servants and enabled longer-run policymaking. 

Emerging evidence in economics on bureaucracies shows how… depoliticizing bureaucratic selection can improve performance… Because industrial policies are complex, skill-intensive, and require careful consideration of appropriate instruments, there may be a case for delegating details of policy formulation to higher-capacity bodies. In postwar Japan, the pilot industrial policy agency, the Ministry of International Trade and Industry (MITI), practiced what Chalmers Johnson famously called “administrative guidance,” de facto power in shaping (and not simply implementing) the industrial policy of the 1950s and 1960s, which Johnson saw as consequential to policy success… When capacity-intensive industrial policies are lobbed onto the growing portfolio of bureaucracies, the problem is compounded…

Implementing industrial policy not only requires a high-quality bureaucracy, but one that continually interacts, negotiates, and exchanges information with industry and stakeholders more broadly. Industrial policy… is informed by and executed through continual interactions with market participants… It matters how connected or “embedded” bureaucracies are with the private sector… At its height, East Asian industrial policy was marked by webs of collaboration between bureaucratic agencies and the private sector… From South Korea’s monthly export promotion meetings to Japan’s use of deliberation councils, East Asian states purposefully cultivated embeddedness by institutionalizing interactions between firms and bureaucracy… Peru’s Mesas Ejecutivas (known as mesas or ME)… established in 2015, are regular, weekly private-public working groups dedicated to solving sector- specific policy… mesas helps identify market and coordination failures and, importantly, can triage and expedite solutions across bureaucracies. Bodies like mesas are notable in that they have a low fiscal footprint and, in fact, were an alternative to costly external consultations… 

Embedded agencies were instrumental to Ireland, Israel, and Taiwan entering dynamic IT markets, yet did so with wide institutional variation. Where the Taiwanese state was directly involved in the industrial R&D process (e.g., Industrial Technology Research Institution), Irish agencies took a more advisory and advocacy role (e.g., National Software Directorate)… embeddedness facilitates the flow of information between bureaucracy and industry. Doing so is essential given fundamental informational asymmetries between bureaucrats (principals) and the firms they interact with (agents)… Peter Evans famously called “embedded autonomy”, where both are required for industrial policy to succeed. Autonomy without embeddedness risks flying blind and constructing and deploying industrial policy in isolation from essential stakeholders. Embeddedness without autonomy risks incoherence and policies guided by private interests…

Thailand’s military-dominated governments pursued an inchoate form of import substitution industrialization (ISI)… Thailand pursued a contradictory mix of export promotion and ISI, or a type of “export-oriented protectionism”. South Korean export policies allowed de facto import liberalization for exporters. Thai policy did not; although exporters were given rebates from protectionist policy, they were insufficient and mismanaged. Where South Korean export policy allowed access to critical machinery and intermediate imports for export production, Thailand protected these goods without adequate relief for exporters, and even raised protection for capital goods through the decade… Where South Korea developed systems for scrutinizing export incentives in the 1960s, Thailand’s 1970 export strategy lacked such capacity, and poor administration created bottlenecks for producers… Thai political constraints made devaluation improbable, unlike postwar Taiwan and South Korea, whose politics allowed—or even compelled—them to pursue politically difficult devaluations before export promotion… Only in the 1980s did a coherent export-promotion policy emerge, promulgated by a new regime that seized upon a window of opportunity.

State capability weakness is an important reason for India’s struggles with the pursuit of industrial policy.