2. Assessing institutional investment in infrastructure

An overview of how much is invested, through which financial instruments, and in which sectors, is an essential starting point for the discussion on how to accelerate and shift investment in new green infrastructure assets (OECD, 2018[1]). Granular data allows a targeted examination of how different types of institutional investors invest, or could invest, in infrastructure (as defined in Box 2.1). It is also fundamental to identify investment instruments and policy levers that can be used to transition to greener alternatives at the pace and scale needed.

This chapter presents results of an empirical mapping of current1 infrastructure holdings by institutional investors domiciled in OECD and G20 countries. For the purposes of this report, the term institutional investor includes pension funds, insurance companies, sovereign wealth funds2 and asset managers.

The quantitative mapping undertaken for this report contributes to wider efforts to plug investment information gaps. To that end, the empirical analysis below makes best efforts to overcome current data gaps. It should be noted that data unavailability on certain parameters, most notably bond ownership, use of proceeds and by extension structured debt products, continues to be a constraint. In the event more data becomes available, the estimates presented here may be revised retroactively, e.g. on the online data explorer accompanying the report or in possible future iterations of this empirical exercise.

This chapter begins with a brief overview of the underlying methodology for the mapping (details can be found in Annex 2.A), followed by a presentation of key findings and their implications. Building on the detailed presentation of data in this chapter, Chapter 3 discusses pathways and levers to scale-up green infrastructure investment.

Data for the mapping exercise is sourced from multiple commercial databases. Commercial data is supplemented by primary data3 collection and econometric techniques to fill gaps. The main databases used for data gathering are Thomson-Reuters (2020[2]), Preqin (2020[3]) and IJGlobal (2019[4]). The aggregation avoids double-counting and other overlaps by collecting data at a disaggregated level, at which distinctions are easily made, and then aggregating to the level presented here. For details on data treatment, the merging of databases as well as statistical and econometric techniques used to fill data gaps, see Annex 2.A.

The main results of the quantitative mapping are presented as Sankey charts. These charts provide a snapshot of current infrastructure investment by institutional investors. In other words, the charts below provide information on the current stock of investments, i.e. infrastructure holdings, and should not be misread as flows (i.e. time series data). At the core this exercise has the ambition to develop a baseline and inform the development of a forward view on how to shift and expedite the flow of institutional capital to support global climate and development objectives.

The empirical mapping also includes, separately from other infrastructure investments, institutional investment in shares (stocks) of companies engaged in infrastructure development and/or operation. These securities of companies deriving revenues from infrastructure development, management and/or operation are often considered listed infrastructure. It is important to note, however, that stock prices of such corporations are determined by factors beyond the cash flow from physical assets, for instance market contagion. Further, investing in corporate stocks through secondary markets does not channel capital to the investee company4. As a result, an investment in an infrastructure-related corporate’s stock has limited effect on new asset creation5, which is the main focus of this report. Accordingly, direct stock holdings are presented and treated separately (see Figure 2.2).

As previously mentioned, unavailability of ownership and use-of-proceeds data for bonds is a constraint for the empirical mapping. Given the over-the-counter6 (OTC) nature of the bond market, information available in commercial databases on the amount of investment made through corporate and other bonds is limited. The empirical mapping undertaken for this report therefore includes limited information on bond investments. Though it is difficult to precisely estimate the amount of institutional capital channelled towards new asset creation through corporate bonds (balance sheet financing), it should be noted that 80-90% of project-level debt finance is provided by commercial banks rather than institutional investors. In-depth interviews conducted for this report confirm this view. Therefore, while bonds play a role in institutional investment in infrastructure, given the prevalence of project finance and the focus of this report on new assets, missing bond information likely does not greatly reduce the comprehensiveness of the mapping exercise. Nonetheless, in the event more information is available, results presented hereafter may be updated retroactively in future iterations of this report and accompanying online sources.

Figures 2.1-2.14 illustrate sectoral splits, debt/equity preference, alignment with environmental goals and geographical splits. The taxonomy of investment channels is based on OECD (2015[6]). The figures aim to present comparable values (see Annex 2.A on comparability) and to provide, collectively, a composite picture of institutional investment in infrastructure.

From left to right, the following Sankey charts track the origin and destination of institutional investment in infrastructure. Nodes on the far left show the sources of investment, categorised by type of institutional investor, while nodes on the far right show the final destination of capital by sector. Figures 2.1-2.4 show investment in seven broad sectors, while the Sankey charts following those drill down to the sub-sectoral level. It is important to note that information regarding assets invested in by certain infrastructure funds is unavailable. The amount allocated to such assets is ascribed to the category ‘unknown’ and excluded from all sectoral figures. For definitions of activities included in each sector, please refer to Annex 2.B. In the interest of readability, the Sankey charts do not differentiate between investment in physical assets and corporations.

Nodes located between the far left and far right present financial instruments through which investment is channelled. Except in the case of direct equity and debt, the ownership relation between the assets (grouped by sectors on the far right) and the investors (far left) is indirect. It is an analytical construction to highlight the diversity of the investment landscape. In other words, investment in assets should be read as attributed to rather than made by the investors. Depending on the financial instrument used, investors may be economic owners of the assets, legal owners or both. For instance, investors of unlisted funds7 (limited partners) are economic owners of the fund’s assets but not owners on record. All attributions in this report are made to reflect economic ownership. Annex 2.A covers details of this attribution and underlying estimation.

Figure 2.3 presents investment in green infrastructure only. Given the absence of a globally accepted definition of green infrastructure, this report undertakes a comparative analysis of select sustainable finance taxonomies, green bond definitions and guidelines from OECD and G20 countries (see Annex 2.B for details and methodology). The objective of the comparative exercise is to identify the lowest common denominator in terms of sectors accepted across all analysed sources to be green. Certain sectors are unequivocally considered green, for instance solar or wind, while certain others are unequivocally not, for example fossil fuels. However, there are infrastructure sectors for which climate and other environmental implications are not quite as clear, for instance roads. For the purpose of this analysis only the sectors accepted as green by all or most of the reference sources considered are included. Figure 2.3 reflects this. It must be noted that according to some reference sources, certain sectors are considered green only if the assets meet a prescribed emission or other threshold. The absence of asset-level emission data makes it difficult to apply this conditionality to the dataset. For the purpose of analysis in this report, therefore, relevant assets are deemed to meet the prescribed thresholds. In other words, the investment amount ascribed to green assets in this report presents an upper bound or the most optimistic attribution.

Figures 2.1 and 2.2 together provide a snapshot of the current institutional holdings in infrastructure - a total of USD 3.34 trillion. As shown in Figure 2.1, USD 1.04 trillion is allocated through all instruments (other than listed stocks). Unlisted funds are the dominant conduit of these infrastructure investments, with USD 380 billion (ca. 37%) in invested assets. USD 173 billion is currently held in direct project equity, with USD 26 billion in direct project debt. Investment through securitised structures including REITS, YieldCos8, MLPs and INVITS9 together represent 43% of current institutional investment. As shown separately in Figure 2.2., ca. USD 2.3 trillion is directly held in listed stocks of companies developing, managing and/or operating infrastructure assets (listed infrastructure). As discussed above, as stock investments do not channel capital to the investee company and therefore cannot direct capital to new investments, the remaining analysis, including in Chapter 3, does not take these investments into account.

Figure 2.3 presents current holdings of institutional investment in green infrastructure, which in total amounts to USD 314 billion. This equals 30% of all institutional investment in infrastructure (excluding investment in corporate stocks). Approximately 49% of all investment in green infrastructure is channelled through YieldCos (USD 155 billion). Unlisted funds and direct project equity follow YieldCos, with USD 93 billion and USD 44 billion, respectively. It is important to note that while unlisted funds account for 37% of investment in Figure 2.1, only 31% of their capital is currently allocated to green assets. This suggests that there is considerable potential upscale green infrastructure investment through unlisted funds. To contrast, 97% of all investment held through YieldCos is allocated to green infrastructure.

The exposures provided by the variety of instruments can be broadly characterised either as exposure to financial assets or exposure to real assets (the latter having stronger linkages to the real economy). The most direct exposures to real assets are provided by direct investment at the project level, unlisted funds and securitised structures like YieldCos, INVITs and REITs.

Figure 2.3 narrows the mapping to the present holdings of institutional investment in green infrastructure. The role of direct investments, unlisted funds and securitisation is even more pronounced in the investment landscape for green infrastructure assets, where they account for almost all investments. In the context of accelerating and shifting institutional capital towards green infrastructure, these three instruments merit further investigation. Chapter 3 focuses on direct investment, unlisted funds and securitised vehicles and discusses the potential of these instruments for scalability.

In terms of origin of investment, it is essential to recognise differences between asset owners (pension funds, insurance companies and sovereign wealth funds) and asset managers. Activities of asset owners and managers are driven by different considerations and different incentives.

Among asset owners, pension funds account for over 71% of the investment depicted in Figure 2.1 (i.e. excluding investments in listed corporations). Over 90% of these pension fund investments are made through direct equity and unlisted funds, in comparison to small holdings in YieldCos and INVITs. These allocations suggest long-term capital appreciation as a major driver of infrastructure investment, and an illiquidity preference possibly incentivised by the illiquidity premium. Data on pension fund commitments in unlisted funds shows a shift in recent years towards riskier strategies, e.g. value added (see Box 1.3 above for a categorisation and explanation of strategies). This is line with the overall market drift towards riskier infrastructure strategies (UBS, 2019[7]). Increasing risk appetite of pension funds is unsurprising given persistent low yields on traditional assets: Non-core strategies provide comparatively high returns and greater opportunities for capital appreciation. One implication of this trend going forward is potential increment in institutional capital available for construction stage projects.

Of the USD 371 billion currently invested by pension funds in infrastructure, 25% is allocated to green assets. A closer look at transaction data reveals that annual direct equity investment by pension funds in ‘non-green’10 assets consistently exceeds that in green. This is driven by investment in natural resources infrastructure and in buildings (social infrastructure such as hospitals). The lack of emission data for all buildings included in the mapping, makes it difficult to distinguish the share of green buildings. However, given that only a small share of the global stock of buildings is green, it is safe to consider the amount directed towards buildings as investment in non-green assets. On the debt side, the lion’s share of direct debt is extended to renewable energy projects. Debt investment in fossil fuel projects stood at 50% of the debt extended to renewables in 2019 (mainly by private pension funds).

The bulk (75%) of infrastructure investment by pension funds is channelled through unlisted funds. Based on capital committed in funds with vintages11 after 2010 and taking a fund lifespan of 15 years, at least USD 40 billion12 can be considered unavailable for shifting to greener investments. This demonstrates the importance of the choices made and instruments used by long-term investors. The illiquidity and financial lock-in of their investments in non-green infrastructure leads to lock-in of higher emissions and a delayed opportunity to shift to green infrastructure. Disaggregation to the regional level in section on cross-regional investments below provides further insights into the investment behaviour of pension funds.

In contrast to pension funds, insurance companies appear to have relatively modest investment holdings in infrastructure. This is explained by different investment preferences of life and general insurers. Infrastructure allocations are primarily made by life insurers on account of their long-term liabilities. General insurers typically underwrite infrastructure instead of investing long-term given their need for short-term liquidity. The infrastructure holdings mapped for this report therefore include investment by life insurers and represent a subset of total insurance assets in OECD and G20 countries. At USD 101 billion in infrastructure assets, investment by insurance companies is ca. 10% of all infrastructure investment. Like pension funds, insurance companies’ infrastructure investment also appears to be guided by long-term capital appreciation—with 81% of current investment established through unlisted funds and direct equity provision. About 38% of total insurance company investment is allocated to green assets. Direct equity investment by insurance companies in green assets has been on a steady upward trajectory in recent years, chiefly due to investments in wind and solar projects. In 2018, direct debt investment by insurance companies in renewables far exceeded that in fossil fuels (Preqin, 2020[3]). In 2019 however, the data shows a reversal i.e. a higher share of direct debt provision to fossil fuel projects. Commitments by insurance companies in unlisted funds with vintages13 after 2010 suggest at least USD 12.5 billion14 locked in non-green assets (versus USD 6 billion in green assets).

Sovereign wealth funds (SWFs) appear to play a limited role in the infrastructure investment landscape. The absence of disclosure around portfolios of SWFs may be one explanatory factor. A second factor is that, in OECD and G20 countries, SWFs have significantly less combined AUM (USD 3.6 trillion) than e.g. pension funds (USD 33 trillion), in part simply because not every country has an SWF. Based on the current mapping, SWFs (like pension funds and insurance companies) seem to invest in infrastructure assets for long-term capital appreciation and possibly illiquidity premium. The share of SWF investment in non-green assets through unlisted funds has been increasing in recent years--driven by investment in fossil fuel projects and natural resources infrastructure. Green investment through unlisted funds is led by wind and solar. There is a recent trend of countries creating SWFs to mobilise capital towards specific policy objectives. SWFs of this nature may be capitalised by national Governments as well as SWFs of other countries, as in the case of NIIF in India. Strictly speaking such SWFs have a different nature than commercial financial entities and are better placed within the context of public sector de-risking.

Listed instruments dominate infrastructure investment by asset managers. Read together, Figures 2.1 and 2.2 show that over 90% of infrastructure investment, i.e. ca. USD 2.4 trillion, by asset managers is allocated to stocks (USD 2 trillion). This is followed by units of YieldCos (USD 153 billion), MLPs (USD 71 billion), mutual funds (USD 5.1 billion), ETFs (USD 3.5 billion), REITS (USD 201 billion) and INVITS (USD 2 billion). Besides the asset owners covered in this report (pension funds, insurance companies and SWFs), asset managers invest on behalf of an array of other clients as well (e.g. retail investors). These other clients, unlike asset owners, have a low illiquidity tolerance and risk appetite. Stocks and units of securitised vehicles offer the benefits of liquidity and stable distributions. Apart from stocks and units of securitised vehicles, asset managers invest ca. USD 77 billion through unlisted funds – this excludes equity participation (as a limited partner) by asset managers in their own funds (funds where the asset manager is the general partner).

Asset managers hold ca. 56% of total institutional investor holdings of green infrastructure (Figure 2.3). This is largely due to investment in YieldCos, which account for 49% of institutional investment in green infrastructure and are a major investment conduit for renewables investment. The value of YieldCos is driven, at least in part, by factors other than the underlying assets, for instance market contagion. Therefore, this major vehicle for institutional investment in green infrastructure can be considered to have significant but not exclusive exposure to underlying infrastructure assets.

Direct infrastructure debt comprises a small portion of the investment landscape. Compared to equity, infrastructure debt is a relatively new asset type for institutional investors. Underlying investment data exhibits a rising interest in infrastructure debt in recent years, both through direct transactions as well as commitments to funds pursuing an infrastructure debt strategy. With persistently low yields on corporate and sovereign bonds, infrastructure debt presents an attractive fixed-income alternative to institutional investors. Government and investment-grade bond yields are expected to be further compressed in the aftermath of the COVID-19 pandemic-- infrastructure debt stands to profit from this trend. In-depth interviews conducted for this report support this view. Among the direct debt transactions tracked for this report, investment in green assets far exceeds that in non-green assets. This is driven by debt extended to renewable energy projects. In terms of asset owners, insurance companies are the most active in this space.

As mentioned previously, data on bond ownership is opaque. Bond investment tracked in Figure 2.1 amounts to USD 0.5 billion. This value must be read as a lower bound of institutional infrastructure investment through bonds in light of data limitations. With this caveat in mind, a look at the sectors invested in, reveals a diverse use of bonds. Bonds can be an effective instrument to raise capital for infrastructure from investors looking for predictable income-generating assets - this includes investor types other than institutional. Some jurisdictions have bond products dedicated to infrastructure, for instance infrastructure debentures in Brazil and infrastructure bonds in India. Based on discussions with experts, green bonds and other labelled fixed-income products have to date not delivered significant financing for infrastructure projects. The most direct means are through green project bonds but that to date has accounted for only a small fraction of the market.

As Figure 2.1 further shows, institutional infrastructure investment channelled through exchange-traded funds (ETFs)15 and mutual funds dedicated to infrastructure amount to USD 2.7 billion and USD 5.2 billion, respectively. Note that these numbers must be read as lower bounds as well. Ownership data for ETFs and mutual funds is often incomplete, thereby preventing the mapping from including institutional investments through funds with uncertain and unknown ownership.

As shown in Figure 2.4, the lion’s share of the investment through unlisted funds and direct equity and debt is directed towards physical assets. 22% of the money channelled through unlisted funds is allocated to unknown sources, and 7% is allocated to corporates. These include renewable energy IPPs, private companies that operate and/or manage infrastructure.

Most of the current positions through unlisted funds and direct project-level equity investment are established through secondary stage investment, i.e. acquisition of operational projects (Figure 2.5). Risk profile of projects is the most elevated during construction phase. However, once projects are operational, project risk is lowered and becomes more palatable to institutional investors. While this preference for operational projects is a longstanding trend, primary stage investment activity by institutional investors has increased in recent years, as in-depth interviews confirm. Construction stage projects with their higher risk-adjusted returns offer an attractive avenue to investors searching for higher yields. As exhibited in Figure 2.5 below, the share of direct debt investment allocated to primary stage opportunities almost equals debt extended to secondary stage projects. Declining yields in the bond market and rising risk appetite of institutional investors (as evidenced above) augur well for increased construction stage credit provision by institutional investors.

A more granular look at the sectoral level provides additional insights into the current investment landscape. Figures 2.6-2.12 show the different sectors and differences in instruments used between sectors. Like in figures 2.1 and 2.3, corporate stocks are excluded.

Figure 2.6 shows an overview of institutional investment in energy infrastructure. Of all infrastructure sectors, energy accounts for the largest investment holdings with USD 488 billion. Asset managers hold energy assets worth USD 263 billion, USD 159 billion is held by pension funds, USD 48 billion by insurance companies and USD 18 billion by SWFs.

Notably, the largest sub-sector in institutional energy holdings (excluding listed stocks) is renewables. For more information, see the breakdown of renewables investments in Figure 2.7. Since much of fossil fuel-based energy infrastructure is held by corporations, much of the institutional investment in this sub-sector is held through shares of these corporations (see Figure 2.2).

Other than renewables, Figure 2.6 shows investment in fossil fuel-based energy infrastructure. Fossil fuel-based energy infrastructure includes, among others, coal, gas, oil power plants, heating as well as natural resource infrastructure (for instance pipelines and storage facilities for oil and gas). Further, smaller categories are nuclear energy and energy efficiency16. The utilities sub-category consists mostly of power utilities that could not be categorised further due to lack of data.

A look at the instruments reveals the centrality of YieldCos, but also highlights the role of MLPs particularly for the natural resources infrastructure. Note that the energy sector is the only sector in which MLPs are used, as fossil-based energy projects are the only eligible projects. With USD 73 billion, MLPs account for 15% of institutional investment in fossil fuel based infrastructure. Notwithstanding the currently exclusive association between MLPs and fossil fuel based infrastructure, the potential of the MLP structure to channel large sums of capital towards real assets (physical assets) is noteworthy. The role of securitised vehicles in scaling-up investment in green infrastructure is discussed in Chapter 3.

Figure 2.7 shows a breakdown of institutional investment in renewable energy infrastructure. Diversified renewables portfolios are the largest category at USD 157 billion, followed by wind (USD 60 billion) and solar (USD 27 billion). Almost all investment in diversified renewable portfolios is held through YieldCos. Lack of data regarding fair values of constituent assets prevents splitting the renewables portfolio category further. However, details (in quarterly reports of YieldCos) regarding installed capacity of portfolios indicate that most of the underlying assets are wind and solar power plants.

As figures 2.1 and 2.2 show, stocks and YieldCos constitute the bulk of total investment by asset managers in infrastructure. As mentioned previously, asset managers invest on behalf of a variety of clients besides institutional investors (e.g. retail investors and high net-worth individuals). These clients have different risk-return preferences (e.g. lower tolerance for illiquidity), to which YieldCos are well-suited; they provide liquid access to physical assets like renewable energy projects. See Chapter 3 for more details.

Figure 2.8 shows the breakdown of institutional investment in infrastructure (USD 130 billion). Of the total amount tracked, only 16% (USD 21 billion) is presently allocated to green infrastructure. The largest single sub-sector is roads with USD 42 billion. Roads (which include toll roads, bridges, tunnels and highways), airports, ports etc. are core infrastructure assets. Such assets generally offer steady revenue streams, often through concessions or availability payments17. Well-established project finance structures exist for transport infrastructure in most jurisdictions analysed in this report. This can be seen in the role of direct project equity and unlisted funds, which are used for ca. 45% and ca. 47% of total transportation infrastructure investment, respectively.

Transport infrastructure provides essential services. Historically, revenues from transport Infrastructure have been stable, as revenues from concessions or availability payments are generally predictable, following broader economic activity trends. However, they may not be immune to large economic shocks. Assets with merchant risk18 can be particularly susceptible to demand shocks such as the one caused by the COVID-19 public health emergency. The demand shock resulting from pandemic-control shutdown measures caused some investors to devalue some transport assets in their portfolios, notably shares of airport operators. To shore-up the attractiveness of critical and green infrastructure like rapid transit systems, proposals have been made to implement public de-risking measures covering revenue shortfall during such exceptional demand shocks.

The use of securitised products for transport infrastructure is relatively modest compared to the energy and telecommunications sectors. However, INVITs are a noteworthy recent addition to the transport investment landscape. Investment through INVITs already stand at USD 1.5 billion i.e. around a third compared to more establish securitised vehicles like REITs (USD 4.6 billion). This is driven by a rising interest in monetising operational assets, both by the public and private sector, to free construction stage equity in certain markets (for instance India).

The prevalence of pension funds is unsurprising given the alignment between their long-dated liabilities and the long-term predictable revenues from transport assets.

Figure 2.9 shows the breakdown of institutional investment in telecommunications infrastructure (USD 186 billion in total). Asset managers hold the largest share among institutional investors (USD 139 billion) primarily through REITs (ca. 96%). Wireless communication infrastructure including telecom towers is the largest recipient of institutional investment. Telecom towers have been relatively unaffected by the COVID-19 crisis. Going forward, the sector is expected to see higher capital allocation by institutional investors. Internet-related infrastructure (fibre optic cables, data centres etc.) is also expected to receive higher institutional investment on the back of expected demand growth and resilience exhibited during the pandemic (Infrastructure Investor, 2020[8]).

The lion’s share of investment in social infrastructure is channelled through unlisted funds and REITs. As shown by Figure 2.10, REITs and unlisted funds account for 84% of total investment. Healthcare is the largest sub-sector at USD 53 billion, i.e. 46% of all institutional investment in social infrastructure. Given that social infrastructure mostly comprises of buildings, the use of REITs is unsurprising. The need for healthcare, education and other social assets is critical to deliver on global climate and development commitments. Efforts to ramp up healthcare infrastructure are seen in government spending plans announced in the wake of the COVID-19 pandemic. REITs offer a scalable means to channel more capital towards developing crucial social infrastructure. REITs are well established vehicles and are often considered to be a traditional rather than an alternative asset. Given longstanding industry experience and comfort with the instrument, scaling-up social infrastructure investment is, in a certain sense, easier than for other kinds of infrastructure. While this report does not address green buildings, as real estate is an asset class separate from infrastructure, it is important to acknowledge the important role of REITs for green buildings. Coupled with the establishment and strengthening of green building codes, targeted use of REITs is a promising means for scaling-up investment in green buildings, delivering sustainable urban centres and achieving significant emission reductions. Targeted use of REITs coupled with green building codes can deliver sustainable urban centres and propel our economies on a low-emissions trajectory.

Institutional investment in water supply infrastructure accounts for a mere 1.6% of all investment holdings mapped in this report (excluding listed stocks). As shown by Figure 2.11, only USD 17 billion19 is presently invested in water supply -related assets, with the bulk emanating from pension funds (USD 12 billion). The investment landscape of the sector is also much less diverse in terms of instruments and vehicles used to channel private capital.

Modest levels of private investment in water supply infrastructure can be explained by some structural aspects of the sector. In most jurisdictions, water supply services, including treatment and distribution, are owned and financed by public authorities rather than private investors. Further, the water sector generally has a poor record of cost recovery, with tariffs often too low to fully cover operational and maintenance costs, and rarely covering capital costs (OECD, 2018[9]). Many jurisdictions lack an independent regulator for tariff setting and concerns regarding affordability often keep tariffs below cost reflective levels. Given the essential nature of water supply services, operators typically cannot disrupt services in the case of non-payment. These factors limit the attractiveness of the sector’s risk-return profile for private investors compared to other infrastructure sectors.

The UK water sector is a notable exception, as water supply services in England and Wales were privatised in 1989 (Ofwat, 2020[10]). Water supply and sanitation infrastructure assets are privately owned and managed. The sector has an independent economic regulator, OFWAT, which oversees tariff setting and capital investment planning of water operators. According to the investment data tracked for this report, 56% of the assets included in figure 2.11 are located in the UK, held by domestic and international institutional investors.

In principle, water infrastructure could offer predictable long-term cash-flows that align well with long-dated liabilities. Steady revenues derived from long-lived assets based on inelastic demand, such as for water supply services, treatment and production of bulk water (e.g. from non-conventional sources, such as desalination), align well with the long-dated liabilities of institutional investors. A stronger enabling environment for investment, with cost-reflective tariffs, independent economic regulation and ring-fenced revenue streams for operators would contribute to a more attractive risk-return profile .

Figure 2.12 shows that institutional investors invest in waste-related and circular economy infrastructure mainly through unlisted funds, with 87% (USD 4 billion) of their holdings through this instrument. Waste-related infrastructure mostly consists of waste management infrastructure which contains sub-categories such as infrastructure for circular economy, and cannot with the current data be disaggregated into smaller categories. Other categories are sewage treatment and sewage utilities20 They are both exclusively held through unlisted funds. While this finding may be a result of the small number of investments, it may also reflect a need for special expertise. Unlisted funds may be more likely to acquire the expertise necessary for these types of investment than other instruments

Figures 2.13-2.16 below present cross border holdings by institutional investors from OECD and G20 countries, categorised by region of the investor’s domicile. The Figure exhibits cross border investment in real assets through unlisted funds, direct equity and debt as well as INVITs where participation in the initial set-up and placement of the vehicle is known. Investment holdings through REITs YieldCos and MLPs are excluded due to lack of clarity on which positions were established during the initial placement and which positions were established through the secondary market. This distinction is observed given this report’s focus on the real economy impact of institutional investment.

Each pair of chord diagrams in Figures 2.13-2.16 presents outbound investment (to all regions including the investor’s region of domicile) in all infrastructure and green infrastructure (i.e. a subset of all infrastructure). With the exception of investors from the Middle East and Europe, institutional investors allocate the majority of their capital to assets located in their region of domicile. This propensity is even stronger, and without exception, for green infrastructure investment—the lion’s share of green infrastructure investment by institutional investors is channelled within their regions of domicile.

Among Asian investors, SWFs and insurance companies are most active in infrastructure investment, led by the Chinese SWF and insurance companies from South Korea. Among Asian pension funds, investment activity by South Korean pension funds far exceeds that by others in the region.

European pension funds are the most active investors in their region—led by funds from the United Kingdom, Netherlands and Denmark. Pension funds from the United Kingdom and Denmark also lead capital allocation to green infrastructure. Among insurance companies, German insurers hold the largest amount in green infrastructure assets, followed by companies from Denmark. In general, European institutional investors exhibit a preference towards mature markets.

In Oceania, pension funds are the most active investors in infrastructure, followed by asset managers. Investors from the region also exhibit a preference towards assets located in mature markets.

South American investors demonstrate the strongest inward preference. Investment activity is led by Brazilian pension funds with all capital allocated to assets within South America.

Among institutional investors in the Middle East21, SWFs have the highest amount allocated to infrastructure with a clear preference for assets located in mature markets. The entire amount is attributable to investment by the SWF of Saudi Arabia. This is followed by pension funds and insurance companies from Israel. Insurance companies domiciled in Israel lead the region’s investment in green infrastructure with bulk of the capital allocated to assets in Middle East and Europe.

Among North American investors, pension funds are the most active investors in infrastructure – led by pension funds from Canada. They are followed by insurance companies domiciled in the United States. North American investors also demonstrate a preference towards mature markets.

The chord diagram for Africa is comprised entirely of South African investors. Pension funds lead investment in infrastructure overall with a strong African preference. The majority of green infrastructure investment originates from insurance companies who also exhibit a domestic preference.

Figure 2.17 provides an overview of cross-border investment amounts.

These findings on cross-border investments highlight that institutional capital exhibits a strong regional preference. The cross-border investments that do take place are primarily targeted at assets located in mature markets. This highlights the critical role of domestic policy frameworks and an investment-grade enabling environment to attract and scale-up institutional investment. Chapter 3 discusses this in greater detail.

A persistent low yield environment is increasingly prompting institutional investors to look to alternatives to obtain higher returns. While infrastructure assets presently account for only a small portion of investable institutional AUM, they offer avenues for higher returns as well as income. Empirical mapping undertaken for this report suggests that infrastructure allocations of pension funds, insurance companies and SWFs are geared at long-term capital appreciation and opportunities to earn an illiquidity premium.

The mapping shows asset managers’ preference for liquid assets. This highlights the potential of securitised structures such as YieldCos, INVITs and infrastructure REITs to scale-up real economy infrastructure investment. Of the institutional investors under study, asset managers have the largest holdings of green infrastructure assets owing to their investments in REITs and YieldCos.

Unlisted funds, direct project-level equity/debt and securitised products are important instruments to upscale green infrastructure investment. Further, data tracked22 for this report points towards a rising risk appetite among investors, particularly pension funds, that bodes well for scaling-up primary stage investment going forward. Direct infrastructure debt is a growing asset type and can offer an attractive alternative to low yielding bonds as well as an increasing source of credit for new assets.

Institutional investors’ choices of financial instruments for infrastructure investment can have important implications for the low-carbon transition. Money channelled towards non-green assets through instruments with low liquidity and lock-in periods, like unlisted funds, can lock-in long-term emissions.

Institutional investors demonstrate a preference towards assets located within their region of domicile. This propensity is more pronounced in case of green infrastructure. Data also shows a clear tendency of cross border investment majorly when assets are located in mature markets. This speaks to the importance of a conducive policy environment to attract and scale-up institutional investment in infrastructure.

While this report doesn’t analyse in detail the impact of COVID-19 on the infrastructure sector, there are early signs that the pandemic might have accelerated an already changing paradigm vis-à-vis sectoral preferences. Coverage around industry sentiment and priorities suggest that telecommunication, in particular data centres and internet-related infrastructure, is poised to receive larger allocations. Another category that might receive increased investor attention is social infrastructure. REITs can be especially useful to scale-up capital allocation towards healthcare and education assets. Additionally, infrastructure spending will form an essential pillar in government efforts around the world to fuel economic activity. This stands to add to the momentum in the private sector and create an opportunity to build green infrastructure that can avoid long-term emission lock-in and ensure public health and wellbeing.

Data for the Sankey charts in chapter 2 are the result of merging multiple databases containing infrastructure investment data. Merging these databases poses several definitional and technical challenges, most notably challenges regarding data gaps as well as diverging or overlapping definitions of actors and sectors. The following describes how the analysis underlying the Sankey charts of chapter 2 tackles these challenges.

All investment data used in this report derives from the infrastructure database of Preqin (2020[3]), the listed securities and listed funds EIKON database of Thomson-Reuters (Thomson-Reuters, 2020[2]) and the infrastructure deal database of IJGlobal (IJGlobal, 2019[4]). Note that despite the overlapping scopes of the above-mentioned databases, there is no overlap or double-counting in the aggregated data.

With the aim of comparability, Figures 2.1-2.XYZ were aggregated in a manner that accounts for differences in investment valuation. For example, while stock investment data is directly attributable to an investor, investments made through unlisted funds have to be attributed based on commitments to funds and based on information of these funds’ asset deals.

Institutional investment data suffers from quality and availability gaps. Data gaps are mainly due to general lack of disclosure on the type of business transactions included in this report. Availability gaps may also be due to the data gathering processes of underlying commercial data. To provide a reasonable attribution and an aggregate picture of investments, these gaps must be plugged with estimations.

To develop a composite view of global infrastructure investments, this report employs statistical techniques to estimate investment values where gaps were found. Since the nature of data gaps differs between, and sometimes even within databases, estimation methods differ as well. The statistical techniques used for this report aim to leverage the provided information as effectively as possible to develop representative estimates.

Wherever possible, observed investment data is used. Any unobserved values are replaced through prediction-based approaches. When prediction is infeasible or does not lead to robust results, estimations rely on averaging over peer-groups of the observations in question. The following sections provide details on the prediction, averaging and aggregation methods employed and discusses how investment values are attributed to investors and sectors.

For investments made through unlisted funds, the estimated and observed data is used to construct an indirect ownership relationship between investors and infrastructure assets. Note that investors in a fund are not the owners on record of the invested assets and the returns for a fund’s investor are based on the portfolio of the fund’s underlying assets.

The bulk of the observations for unlisted funds are sourced from Preqin (2020[3]), containing open-ended and closed infrastructure funds, participating in relevant infrastructure transactions. In preparation for estimations, all past owners are excluded. This is to ensure that the aggregated results only reflect current investment.

The total commitments of institutional investors in all funds in the database amount to 3533. Of these, 1318 commitments are observed and 2215 commitments are estimated. The total number of all deals executed by unlisted funds in Preqin (2020[3]) and additional23 transactions by relevant unlisted infrastructure funds added from IJGlobal (2019[4]) amounts to 1766. Of this, transaction amounts for 857 deals are observed and 909 are estimated.

Although individual deals cannot directly be attributed to the investors of a fund, investments of a fund can be attributed to investors of that fund according to how much the single investors committed to the funds in question.

Figure 2A.1 shows the attribution of current fair value of an investor’s investment through an unlisted fund based on that investor’s commitment in the fund. Guided by the pro-rata distribution principle underpinning limited partnership structures, the commitment values can be used as weights of the fund’s residual value to estimate the fair value of an investor’s investment in that fund.

In line with this approach, in step one shown in Figure 2A.1, all unobserved commitment values are estimated using econometric techniques. Following the pro-rata distribution principle, based on the called24 percentage information and the RVPI25 (residual value to paid-in), step two calculates the residual value-equivalent of an investor’s commitments.

A similar approach is employed for the transactions side to construct an indirect ownership relationship. To do so, unobserved values of transactions by all observed infrastructure funds are estimated in step three. Deal values are used as weights and applied to the residual value of the fund to calculate the sectorial allocation of the fund in step four. This means that the deal value based weights are also applied to the fair value of an investor’s investment to develop an investor-sector-region observation.

Note that direct attribution of several commitments to sectors is impossible since data on the fund that links them is missing. Instead, the calculation assumes that the average shares found for the commitments that can be attributed is representative for the commitments with missing fund data as well. The calculation therefore attributes the residual value-equivalents according to these average sector shares in commitments of the rest of the sample.

The econometric technique used to estimate and predict unobserved values is described in the following paragraphs. Figure 2A.2 presents an overview of the hierarchy of estimation methodology followed to ensure a consistent use of the best method applicable. For the definition of sectors and merging of sector-definitions, see Annex 2.B.

Commitment values estimated in step one above, are based on information on the fund, its AUM26 and its investors. Based on these regression results, predicted values are filled in where no commitment value is observed. The adjusted R² of 0.89 confirms that the model is well adjusted. An F-test confirms the significance of the model as a whole. A Breusch-Pagan test confirms absence of heteroscedasticity and comparing Akaike information criterion values confirms the choice of the model over alternatives that were run as robustness checks. Further, comparisons of in-sample predictions with observed values show that even outlying values are never more than twice the observed value, pointing towards reasonable accuracy of the predictions.

In cases where an out of sample prediction of commitments is not possible due to missing data, the missing commitment values are replaced by averaged commitment values. Averages are calculated on the closest peer-group of observed commitments, and if data is missing, averages are calculated based on a less directly comparable peer-group. The closest peer group for calculated commitment averages is a group of commitments with the same industry, strategy, country and inception year of the fund. These categories are gradually relaxed to less comparable peer-groups if missing data could not be filled in.

Deal values estimated in step three depend on available information about the deal as well as asset information, as well as information on the fund investing in the deal27. Based on regression results, predicted values are filled in where no deal value is observed. The adjusted R² of 0.75 confirms that the model is well adjusted. As for the commitment regression, the F-test, Breusch-Pagan test, the Akaike information criterion and comparisons of in-sample predictions with observed values all confirm modelling choices.

Additional investments through funds are available in the IJGlobal (2019) database and are included in the unlisted fund estimations. Inclusion of the deal values in the deal estimations is straightforward as the information is available (as is the case Preqin data). As all funds involved in the deals are also in the Preqin database, the IJGlobal deals could be attributed through these funds.

Data on direct project-level investments by institutional investors is sourced from IJGlobal (2019), Preqin (2020) and (OECD, 2018). This information on direct investment is merged to arrive at the overall direct investments by institutional investors, using manual merging and OpenRefine to avoid double-counting of investments. As in the case of unlisted funds, careful attention is paid to exclude past owners of assets.

The merged data provides information on 953 observed transactions with equity participation by an institutional investor. Due to missing values, equity investment are estimated for a portion of these transaction. To estimate the unobserved equity value, first a regression is run using information about the investors and the asset28. Next, gearing ratios29 are applied to arrive at equity portions of deals, and percentage stakes acquired by investors are applied to arrive at the absolute value of direct institutional equity investment.

The merged data also provides information on 168 observed transactions involving debt provision by an institutional investor. Of these, 4% of the unobserved debt investment values are straightforwardly calculated based on observed information. For the remaining data gaps of 79% of the investments, values are estimated. An in-sample comparison reveals that the average of total observed debt investment share for a deal is a good approximation of the observed USD debt shares. Consequently 18% of the remaining missing values are replaced based on these averages. Missing data for the remaining 61% of observed deals are replaced by averages of investments in a peer-group based on asset, deal and investor characteristics, assuming representativeness of these groups. One final observed debt investment is dropped since no useful data for estimation was observed for this investment.

Investment data for publicly listed infrastructure funds and stocks is retrieved from Thomson-Reuters EIKON (Refinitiv, 2020).

For listed stocks of corporations, the EIKON data provides a list of investors and the percentage shares of investments in these companies. These shares are then multiplied by the market capitalisation as on last trading day of February 2020. All non-USD values are converted to USD equivalents using an average of the 2019 exchange rates from the EIKON database. These values combined with the investor information provide the investor-company-level information on investments, including the investment value. Further, EIKON provides a sector-classification, which is transferred into the classification presented in Annex 1.B.

For listed infrastructure funds the analysis starts by filtering all funds tagged as infrastructure in the Lipper funds database of EIKON. The available funds include listed mutual funds, INVITs and ETFs. Out of these 2000 funds, useful data exists for only 148 funds. The analysis is based on these 148 funds only since no useful information on the other funds is available to estimate their size as well as holdings or ownership composition. Fund holdings typically are equity shares (e.g. stocks), fixed income instruments (e.g. bonds) and cash. Rather than include all investments by these funds as infrastructure, the analysis includes only those fund holdings matching the infrastructure definition outlined in chapter 2 (see discussion in Box 2.1). Data on YieldCos and REITs has been treated similarly. Where possible, desk research is used to supplement EIKON data to increase comprehensiveness for the instruments. This is especially true for INVITs where most of the data is collected through desk research.

Note that overlap is avoided between institutional investors holdings through listed funds and direct institutional investor holdings in corporations. Since the direct holdings do not include holdings of listed fund shares, the funds’ holdings are only included through the listed fund holdings. So while an institutional investor may hold shares of a corporation directly as well as through listed infrastructure funds, these are cumulative holdings rather than double-counted.

For all listed items, observed ownership and holdings are noticeably incomplete as they do not add up to 100% of shares. As is the nature of publicly traded data, information on details is largely available, but not always complete. This would indicate that the aggregates presented in chapter 2 are only a lower bound. However, typically data for large transactions and for large investors is systematically better tracked than for small investors or transactions. The analysis can reasonably assume that institutional investments in the stock market belong in these categories. Therefore the aggregates of chapter 2 for listed stocks should be a reasonable estimate of the actual value of institutional holdings of listed infrastructure stocks. For listed funds the same applies, with the exception of funds without data, for which the analysis has to stay agnostic.

Table 2.B.1 below provides an overview of the activities, sub-sectors and sectors included in this report. The following table has been developed based on the classification found in Preqin (2020[3]), Thomson-Reuters (2020[2]) and IJGlobal (2019[4]). All infrastructure assets and corporate entities included in the dataset developed through the empirical mapping fall into one or more of the following activities.

The OECD defines infrastructure as “the system of public works in a country, state or region, including roads, utility lines and public buildings”. This includes electricity generation, transmission and distribution assets. Table 2.C.1 below lists infrastructure-relevant sectors and technologies that qualify as ‘green’ under select sustainable finance taxonomies, green bond standards and/or guidelines (analysed resources) in select OECD and G20 jurisdictions. This exercise aims to highlight the lowest common denominator to develop a working definition of ‘green infrastructure’ for the sole purpose of the mapping in this report. To identify the lowest common denominator, all infrastructure-related sectors in the analysed resources are mapped alongside each other. The sectors that are accepted as green by all or most of the analysed resources are designated green for the purpose of this report. It must be noted that some analysed resources prescribe emissions or other thresholds for assets belonging to certain sectors while others don’t. For instance passenger rail is unequivocally green according to the standards and definitions in Japan and China but maybe considered green as per the EU taxonomy only if the asset in question meets a stipulated threshold. Given the absence of granular emission-level data, it is difficult to overlay such a conditionality on the assets in this report’s dataset. Therefore in the interest of facilitating analysis, wherever applicable, all assets in this report’s dataset are assumed to meet the prescribed thresholds. In the Table below, sectors that are unequivocally green are indicated as dark green, sectors subject to a stipulated threshold are marked as light green. White or blank cells indicate absence of coverage.

With the exception of the Climate Bonds Initiative (CBI) taxonomy (which is a market-based taxonomy as distinct from an official, government-established taxonomy or definition), the taxonomies and standards/guidelines compared below do not explicitly exclude sectors. The taxonomies and standards/guidelines assessed only indicate sectors, and projects/activities therein, that qualify as ‘green’. The extreme right column indicates the sectors in which institutional investment has been observed in the empirical mapping.

Table 2.C.2 below compares the definition/meaning of the term ‘green’ under relevant standards/guidelines/principles prescribed by the competent authority* in select OECD and G20 jurisdictions. The objective is to highlight common elements to arrive at a working assumption for the meaning of ‘green’ for the purposes of this report.


[11] Amenc, N., F. Blanc-Brude and A. Chreng (2017), “The rise of “fake infra”: The unregulated growth of listed infrastructure and the dangers it poses to the future of infrastructure investing”.

[4] IJGlobal (2019), IJ Global Infrastructure Transactions Database.

[13] IJGlobal (2017), IJGlobal - Transaction Data, https://ijglobal.com/data/search-transactions (accessed on 4 December 2017).

[8] Infrastructure Investor (2020), How digital infrastructure became ‘mission-critical’, https://www.infrastructureinvestor.com/how-digital-infrastructure-became-mission-critical/.

[5] OECD (2020), OECD Glossary of Statistical Terms - Infrastructure Definition.

[9] OECD (2018), Financing water Investing in sustainable growth.

[1] OECD (2018), Innovation, Standardization and Data Collection for Long Term Investment: OECD Workshop on Data Collection for Long-term Investment – November 2018 -Summary record, https://www.oecd.org/daf/fin/private-pensions/OECD-Workshop-on-Data-Collection-Summary-2018.pdf.

[6] OECD (2015), Mapping Channels to Mobilise Institutional Investment in Sustainable Energy, Green Finance and Investment, OECD Publishing, Paris, https://dx.doi.org/10.1787/9789264224582-en.

[10] Ofwat (2020), Water sector overview, https://www.ofwat.gov.uk/regulated-companies/ofwat-industry-overview/ (accessed on 21 September 2020).

[3] Preqin (2020), Alternative Assets Data, Solutions and Insights.

[12] Röttgers, D., A. Tandon and C. Kaminker (2018), “OECD Progress Update on Approaches to Mobilising Institutional Investment for Sustainable Infrastructure”, OECD Environment Working Papers, No. 138, OECD Publishing, Paris, https://dx.doi.org/10.1787/45426991-en.

[2] Thomson-Reuters (2020), EIKON.

[7] UBS (2019), Top infrastructure trends for 2020 | UBS Global topics, https://www.ubs.com/global/en/asset-management/insights/asset-class-research/real-assets/2019/top-infrastructure-trends-for-2020.html (accessed on 17 June 2020).


← 1. As on end of February 2020

← 2. Note that this report does not include Strategic Investment Funds (SIFs) due to lack of data. They would be a relevant addition as a vehicle for institutional investments, including those of Sovereign Wealth Funds (SWFs).

← 3. Note that data on listed infrastructure investments was downloaded in late February 2020 and therefore before the COVID-19 crisis fully hit the stock markets. Data was not updated to post-COVID-19 for two reasons. First, an update of listed data would inevitably have happened during rather than after the crisis, i.e. it would be influenced by the crisis, but at the time of writing it would not have been possible to say to what extent. Second, as other data, e.g. unlisted funds data, is updated only periodically, an update of only the listed investment data would have been inconsistent.

← 4. Investments at the time of an initial public offering could be an exception here, since capital raised may be used for new assets.

← 5. While participation in primary issuances may provide investment for new asset creation, stock investments in the secondary market do not provide additional capital to the company concerned. Therefore an investment in an infrastructure company’s stock does not cause a direct change in the real economy (except in the case of a primary issuance, i.e. a initial public offering). While these secondary market activities may provide incentives to engage in the primary activity of setting up corporations, this indirect effect is beyond the remit of this report.

← 6. Where transaction data is more limited than on public exchanges

← 7. Unlisted funds pool capital from multiple investors. Funds are typically structured as limited partnerships with an asset/fund manager (party raising capital) as the general partner and investors (including institutional investors) in the fund as limited partners. Funds have a fixed lifespan which may be extend by the consent of limited partners. During the investment period, limited partners are entitled to cash flow which may either be distributed or reinvested. Distributions are typically paid on a pro rata basis.

← 8. Unless stated otherwise, the term YieldCo in this report refers generally to the legal structure that enables securitising illiquid physical assets, and not to any particular vehicle or strategy in existence in the market either presently or at any time in the past.

← 9. Infrastructure Investment Trusts (INVITs) and master limited partnerships (MLPs), like YieldCos, combine access to infrastructure cash-flow with liquidity. INVITs and infrastructure REITs are publicly traded trusts that own and operate infrastructure assets. MLPs are particular to the United States. They are pass through vehicles for tax purposes and are for the use in infrastructure restricted by law to activities related to natural resource exploitation.

← 10. ‘Non-green assets’ excludes the following infrastructure sectors for which climate and other environmental implications are not quite as clear: telecommunications infrastructure, roads, bridges, tunnels, highways.

← 11. Vintages of some funds in the underlying data are uncertain and have been excluded from the calculation for the sake of precision. However, when the amount held through funds with uncertain vintages is factored in, the estimate of capital locked in non-green assets rises by USD 100 billion.

← 12. ‘Non-green assets’ excludes the following infrastructure sectors for which climate and other environmental implications are not quite as clear: telecommunications infrastructure, roads, bridges, tunnels, highways.

← 13. Vintages of some funds in the underlying data are uncertain and have been excluded from the calculation for the sake of precision. However, when the amount held through funds with uncertain vintages are factored in, the estimate of capital locked in non-green assets rises by USD 20.5 billion.

← 14. ‘Non-green assets’ excludes the following infrastructure sectors for which climate and other environmental implications are not quite as clear: telecommunications infrastructure, roads, bridges, tunnels, highways.

← 15. Exchange Traded Funds (ETFs) are a mix between open-ended and closed-end funds. Like closed-end funds, units of ETFs trade on public exchanges. However, like open-ended funds, ETFs are always open for new subscription i.e. new units are created and the fund size expands based on new demand. Redemption by investors leads to contraction of the fund size.

← 16. Note here that energy efficiency included here does not include energy efficient real estate.

← 17. An availability payment is a contractual payment, as part of an offtake agreement, usually by the public sector in PPP formats.

← 18. Economic dispersion in revenues due to variations in end-user demand.

← 19. The difference between the aggregate of the far left and far right side of Figure 2.11 are due to rounding.

← 20. Note that since these services are water-related, some of these investments may be captured in the water utilities category of Figure 2.11.

← 21. Note here that due to the choice of country scope of this report, some other SWFs from the Middle East are not reflected.

← 22. While the data gathered for this report represents the current stock (holdings) of investment and not flows (i.e. time series data), evolving risk appetite of investors can be ascertained from the vintages of unlisted funds. Capital commitments by asset owners, vintages and strategies of funds together suggest a trend.

← 23. Note that to avoid overlap, each single deal added from the IJGlobal database is manually checked against deals from the Preqin database.

← 24. Note that missing data on called percentage values for observed funds was replaced by the average called percentage. While not exact, this approach is reasonable given the narrow distribution of called percentage values around the average.

← 25. Note that missing data on RVPI is replaced by averaging over observed RVPI values of gradually widening peer-groups of funds. Factors used to identify peer-groups include the size of the investor in terms of AUM, the year of the fund as well as the country, strategy and core industry target of the fund in question.

← 26. Other information included in the regression underlying the prediction are country of origin of the investor, the investor type (asset managers, private pension funds, public pension funds, insurance companies, sovereign wealth funds, investment companies and funds of funds), other funds invested in by the investor, as well as fixed effects of the investor and fund.

← 27. Note that investments recorded in Preqin (2020) in currencies other than USD were converted to USD using OECD National Accounts (2020) data.

← 28. Information included in the regression underlying the prediction are country of origin of the investor, the investor type as well as the country, year and industry of the investment.

← 29. Note that data gaps for gearing ratios and acquired stakes are filled using averaging of the observed values by peer-groups. Similar to the averaging procedure for missing values estimated for private equity data, the peer-group categories are gradually relaxed if there is no relevant peer-group over which to average.

Metadata, Legal and Rights

This document, as well as any data and map included herein, are without prejudice to the status of or sovereignty over any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or area. Extracts from publications may be subject to additional disclaimers, which are set out in the complete version of the publication, available at the link provided.

© OECD 2020

The use of this work, whether digital or print, is governed by the Terms and Conditions to be found at http://www.oecd.org/termsandconditions.