The credit quality of many infrastructure issuers rated by S&P Global Ratings rapidly deteriorated in the spring of 2020 as the world plunged into recession. When it became clear in the first quarter that COVID-19 was quickly spreading outside of China, financial conditions sharply tightened and economic activity abruptly slowed as governments around the world instituted stay-at-home orders, mandated business closures, and imposed travel restrictions.

The sharp contraction in economic activity and tightened financial conditions damaged credit quality in most corporate sectors along with subsectors of infrastructure where issuers were exposed to volume risk. The number of infrastructure downgrades doubled from a year earlier to 214 in 2020, and there were just 31 upgrades. The number of upgrades minus downgrades (-183) was the lowest since 2002 (-183), and upgrades minus downgrades as a proportion of issue ratings at the start of the year (-13%) hit the lowest since 2003 (-20%).

The deterioration in credit quality among infrastructure issuers led to more issue defaults in 2020 than in 2019; defaults more than tripled to 10 (from three), and nine of these were from issues rated speculative-grade at the beginning of the year, pushing the speculative-grade issue default rate to 2.69% (from 0.30% in 2019) (see chart 1). Nonetheless, infrastructure ratings performance generally held steady amid the 2020 recession, with most defaulting issues rated 'B-' or lower at the beginning of the year and no investment-grade issue defaults.

Of the 10 infrastructure issue defaults in 2020, nine were rated and one was not rated when 2020 began (this issue was first rated 'CCC' in August 2020). Nine of the defaults were from U.S. issuers (including tax havens), and one was from an issuer based in Argentina. Eight of the defaults were from corporate infrastructure issuers and two were from project finance issuers. By subsector, oil and gas led issue defaults (six), followed by power (two), utilities (one), and transportation (one).

This study covers infrastructure ratings in utilities, oil and gas (mostly midstream issuers), transportation, power, social infrastructure, and other industries. For the purposes of this study, our definition of infrastructure does not include U.S. public finance or international public finance rated infrastructure issues.

Table 1

Regional Default Trends

The U.S. leads all regions in speculative-grade ratings ('BB+' or lower) (see chart 2). Speculative-grade infrastructure ratings in the U.S. rapidly increased from 2003 to 2004, then steadily increased until peaking in 2018 (see chart 3).

Of the 144 infrastructure issue defaults since 1981, 103 have been from issuers based in the U.S., with defaults outside the U.S. exceeding U.S. defaults just five times annually in that span. The larger number of speculative-grade issues in the U.S. has contributed to the higher number of U.S. infrastructure defaults.

Subsector Default Trends

By subsector, utilities account for the most infrastructure defaults since 1981 (47%), followed by power (24%) and oil and gas (12%) (see chart 4). Utilities (59%), oil and gas (13%), and power (8%) combined account for over 80% of annual rated issues in that period.

In examining the rating distribution by subsector over time, power defaults appear overrepresented, accounting for 24% of defaults since 1981 while accounting for just 8% of annual infrastructure ratings during that time (see chart 5). However, over half of annual power subsector issue ratings in that time have been speculative-grade, and the subsector accounts for 21% of annual speculative-grade infrastructure issue ratings since 1981, which helps explain the high proportion of defaults. Merchant power projects are characterized by high market risks (low barriers to entry; merchant pricing risk; unregulated, aggressive leverage; and lower-than-forecast demand), although this is not the case for contracted power. Merchant power issuers also face regulatory risk, particularly with respect to green mandates.

Conversely, utility defaults appear underrepresented, accounting for 47% of defaults since 1981 while accounting for 59% of annual infrastructure ratings. However, nearly 90% of annual utility issue ratings since 1981 have been investment-grade ('BBB-' or higher), and while the subsector accounts for 59% of issue ratings, it accounts for just 34% of annual speculative-grade infrastructure issue ratings since 1981.

Defaults from oil and gas, transportation, social infrastructure, and other categories are more proportional to their percentages of ratings.

Growth in rated issues from oil and gas, transportation, power, social infrastructure, and other subsectors has outpaced that of utilities since 1997. Utilities accounted for at least 80% of rated infrastructure issues annually through 1996 but currently account for a much lower proportion (43%), though it is still the largest subsector. The growth in ratings in subsectors other than utilities coincides with outsize growth in speculative-grade ratings, particularly in the oil and gas and power subsectors.

Corporate infrastructure generally has lower average default rates than project finance (see chart 6). However, the corporate infrastructure default rate was slightly higher than project finance's in 2020. The average default rate for corporate infrastructure since 1981 is 0.38%, lower than 0.73% for project finance.

Historical Default Cycles

In general, infrastructure's relative peak default rates aligned with those of nonfinancial corporates during the 2000 recession. However, the sector made it through the global financial crisis relatively unscathed, with a peak default rate under 1%, compared with just under 6% for nonfinancial corporates (see chart 8).

On a trailing-12-month basis, the overall infrastructure default rate was higher than the nonfinancial default rate only from January 2005 through December 2005. This was largely because of the Dec. 21, 2005, bankruptcy filing of Calpine Corp. (a corporate infrastructure power developer and energy merchant company), which covered nine defaulting instruments in our dataset.

The lower infrastructure default rate in most periods in part reflects a consistently lower proportion of issues rated in either the 'B' or 'CCC'/'C' categories in infrastructure compared with nonfinancial corporates. At the beginning of 2021, 43% of nonfinancial corporates were rated in either the 'B' or 'CCC'/'C' categories, compared with just 10% of infrastructure issues.

Table 2

Table 3

Table 4

Among defaulting issues, there is generally a positive correlation between initial ratings and time to default (see chart 9). The shortest time to default in our dataset was for GenOn Energy Inc., which defaulted in June 2017, just 26 days after it initially was rated 'CCC+'. Conversely, the longest time to default was for Public Service Co. of New Hampshire, which defaulted in January 1988, nearly 26 years after receiving an initial rating of 'A'.

We define the marginal default rate as the percentage of rated projects that default every 12 months. On average, the infrastructure 'BBB' category marginal default rate holds steady over 10 years, the infrastructure 'BB' category marginal default rate increases gradually to a peak in year four before sharply declining, and the infrastructure 'B' category marginal default rate peaks in year two before sharply declining (see chart 10).

Speculative-grade ratings tend to have high marginal default rates in earlier years before declining, while investment-grade ratings tend to have persistently low marginal default rates. Infrastructure default rates tend to be lower, but marginal default trends are similar to those for nonfinancial corporates.

Historical Rating Action Cycles

As in most sectors, rating movements in infrastructure have most often been negative historically (see chart 11). Downgrades outnumbered upgrades 3,327 to 2,549 from 1981 through 2020 and outnumbered upgrades in 27 of those 40 years. The number of annual downgrades peaked in 2002 at 222, with 2020 close behind at 214. Upgrades peaked in 2007 at 170, with the second-highest total coming in 2017 at 142.

All subsectors within infrastructure had more downgrades than upgrades in 2020 (see chart 12). Transportation had both the lowest number of upgrades minus downgrades and the lowest proportion of upgrades to total rating actions, followed by oil and gas. Transportation issuers such as toll roads and airports felt the effects of drops in volume, while oil and gas issuers faced pressure from the sudden drop in energy demand. These two subsectors and the utilities subsector combined accounted for about three-quarters of infrastructure downgrades in 2020.

The period for infrastructure with the highest number of upgrades minus downgrades was from 2005 to 2008, which followed the period with the lowest number of upgrades minus downgrades, from 1998 to 2004 (see chart 13).

Upgrades minus downgrades in 2020 matched the low from 2002, but credit deterioration was spread more evenly across subsectors in 2020, whereas in 2002, the utilities subsector accounted for the bulk of downgrades.

Table 5

Table 6

Rating Transitions

Transition tables provide detail on the speed with which infrastructure ratings change. Table 8 shows average one-, three-, and five-year rating transition rates from 1981 to 2020. Each cell indicates the weighted average percentage of ratings maintaining the same rating category at the end of each time horizon. The rows indicate the starting rating, and the columns show the rating category after each designated period.

Note the large percentages of withdrawn ratings (not rated, or NR) as time increases. The high frequency of ratings withdrawn for redemption, maturity, or any other reason makes longer time horizons of up to 20 years impractical for infrastructure. This is also a byproduct of our analysis being built on issue ratings; issues tend to be paid off and thus move to not rated. This stands in contrast to our nonfinancial corporate ratings, which are issuer credit ratings and typically remain active much longer than any underlying issue ratings.

The highest values in the transition matrix are along the diagonal, indicating that as expected, most ratings do not move in a given period. However, by reading to the left of the diagonal and the right of the diagonal, one can determine the ratings that have transitioned to higher (left) and lower (right) categories.

In one year, 3%-9% of investment-grade ratings move to lower rating categories; in three years, 6%-20% of investment-grade ratings move to lower rating categories; and in five years, 7%-24% of investment-grade ratings move lower. The 'A' category is typically the most stable, and 'AA' category ratings are the most likely to move to a lower category at each time horizon. The smaller sample sizes of issuers in the 'AAA' and 'AA' categories likely contribute to this.

From 1981 to 2004, the proportion of downgrades from speculative-grade issuers was close to the proportion of speculative-grade ratings, accounting for about 14% of downgrades and 12% of ratings on average. However, as the proportion of speculative-grade infrastructure ratings has grown, there's been an even larger increase in the proportion of downgrades from speculative-grade issuers: From 2005 to 2009, speculative-grade downgrades accounted for about one-quarter of annual downgrades, with about one-fifth of infrastructure issuers rated speculative-grade on average. But from 2010 to 2020, speculative-grade downgrades accounted for about two-fifths of annual downgrades and one-quarter of ratings on average.

Table 7

The high percentage of infrastructure ratings that have been withdrawn, as indicated in the NR column, obscures the data and our conclusions to some extent. For example, after five years, 36% of infrastructure ratings in the 'BBB' category have been withdrawn, compared with 23% of nonfinancial corporate ratings in the same category. There is no indication of whether the removed ratings would have remained in the same category, moved higher, or moved lower. This results in less precise comparisons between infrastructure and nonfinancial corporate ratings.

Table 8

Focusing on the 'BBB' category, which is generally the largest infrastructure rating category, allows for comparison between infrastructure subsectors. All subsectors except utilities exhibit greater negative rating movement than positive rating movement in the 'BBB' category. Social infrastructure ratings are the most likely to remain in the 'BBB' category from one to five years, in part due to the greater revenue stability conferred by availability payments, which are more common in this subsector. Power infrastructure ratings exhibit the highest rate of downward movement at all horizons.

Table 9

Gini Coefficients

The Gini coefficient--a summary statistic of the Lorenz curve--is a measure of the relative ability of ratings to differentiate default risk. Infrastructure Lorenz curves vary in their similarity to ideal curves, and Gini coefficients fall within a wide range. Over one year, three years, and five years, the coefficients for infrastructure as a whole--at 79%, 69%, and 64%, respectively--show a strong relationship between lower ratings and the propensity to default.

The points along the Lorenz curve represent different ratings, starting with issues at the 'CCC+' and lower ratings on the far left (see chart 14). The first point indicates about 33% of transitions to default were at 'CCC+' and lower ratings. At higher ratings, the number of defaults decreases, yielding points that continue to climb through the remainder of the speculative-grade scale before turning sharply to the right at the investment-grade rating categories, reflecting the low proportion of defaults from investment-grade issues. Table 11 shows the coordinates of the Lorenz curve shown in chart 14.

Table 10

However, this relationship is not the same for all subsectors within infrastructure (see table 12). To conduct a comparison, a three-year horizon for the Gini coefficient is used because it is the longest period in which a sizable number of issues remain outstanding (either prior to redemption or due to another cause for rating withdrawals).

Social infrastructure has a three-year coefficient of nearly 100% (extremely high), while the transportation, other, and utilities subsectors have high coefficients (85%, 79%, and 66%, respectively) and the oil and gas and power subsectors have low coefficients (55% and 49%, respectively), indicating a high frequency of defaults among higher-rated issues.

While it is not as extreme, there is some regional variation as well. Canada has an extremely high coefficient (nearly 100%), while Europe, the Middle East, and Africa (EMEA), Asia-Pacific, the U.S., and Latin America have high coefficients (at 68%, 66%, 65%, and 64%, respectively).

Table 11

Recovery Rates

Ultimate recovery rates for infrastructure debt instruments have been broadly stronger than for nonfinancial corporates, since the recoveries of infrastructure corporates lift the average. Many of the infrastructure corporates and projects that achieved the highest recoveries exhibited high asset quality and low industry risk. Furthermore, regulated utilities specifically tend to show above-average recovery rates (in particular, for senior unsecured bonds), and this subsector accounts for a notable share of defaulted infrastructure debt instruments.

Recovery rates for infrastructure (including bank debt and bonds from corporate infrastructure and project finance) average 80%, which is higher than the 60% average for nonfinancial corporates (see chart 15).

Within infrastructure, corporate infrastructure (excluding project finance) displays an average recovery of 84% (from 316 instruments), and more than half of the defaulted instruments recovered par or greater. This average recovery for corporate infrastructure is notably higher than the average recoveries from project finance (60% from 59 instruments) and from nonfinancial corporates (60% from over 4,000 instruments).

Average recoveries from project finance instruments are considerably lower than for corporate infrastructure and are nearer to the average recoveries of nonfinancial corporates. However, the average recovery from project finance is based on a much smaller sample of 59 instruments, and the distribution of these recoveries is more varied than the distribution within corporate infrastructure. The projects with the lowest recoveries typically exhibited high market risk or encountered technology issues.

Our dataset of recoveries for infrastructure corporate and project finance includes defaulted instruments that were resolved through a debt restructuring, a distressed exchange, or a liquidation and sale of assets. To calculate recovery rates on defaulted projects, we relied on public documents, court filings, and company reports. Because we used public information, we were unable to calculate recovery rates on all defaulted projects. Most of the recovery rates we were able to calculate were from legal jurisdictions, such as the U.S., that published an abundance of post-default data. These recoveries are calculated on a nominal basis and have not been discounted for the time between default and emergence.

In addition to calculating recovery rates as an average of instrument recoveries, weighted by the number of instruments, we also looked at recoveries on a dollar-weighted basis. For the dollar-weighted averages, we calculated the nominal sum of debt recovered from the pool of defaulted instruments, divided by the total of defaulted debt in the sample for that instrument type.

Dollar-weighted recoveries for infrastructure tend to be lower, on average, than the instrument-weighted averages. However, the 71% dollar-weighted average recovery for the infrastructure sector overall remains higher than the 56% dollar-weighted average for nonfinancial corporates.

This dataset includes 68 defaulted obligors, affecting 375 debt instruments. While the dataset includes obligors from the emerging and developed markets, nearly 92% of the instruments included in the recovery dataset are from U.S.-based obligors. These include entities that defaulted from 1988 to 2020. The count of instruments includes senior debt, senior unsecured debt, and subordinated debt. Many obligors have defaulted on a variety of instruments, and some have defaulted more than once, causing the number of observed instrument defaults to far exceed the number of defaulters.

Our ongoing enhancement of the database used to generate this study periodically includes the addition of defaulted instruments from prior years, where sufficient information on the ultimate recovery value has become subsequently available. For instance, since our prior study, we've included the recoveries of debt instruments from Pacific Gas & Electric Co., following both its 2001 and 2019 defaults.

Recovery By Instrument Type

By debt type, senior secured and subordinated infrastructure debt shows average recovery rates that are very close to the average recoveries of nonfinancial corporates. By contrast, senior unsecured debt from infrastructure displays a markedly higher average recovery (near 85%) than that of nonfinancial corporates (close to 52%) (see chart 16).

Senior secured debt (which includes bank debt along with senior secured bonds) from the infrastructure sector shows an average recovery of 77.5%, which is modestly lower than 79% for nonfinancial corporates. Meanwhile, subordinated debt from the infrastructure sector averages 31%, near the 32% average recovery of subordinated debt from the nonfinancial sectors.

In our dataset, average recoveries of senior unsecured debt are also higher than the average recoveries of senior secured bonds. While the pool of senior secured debt is split nearly evenly between project finance and infrastructure corporates, the senior unsecured debt of infrastructure corporates is largely from regulated utilities. This concentration lifts the average recovery of senior unsecured debt. With its regulatory support and ability to pass through costs to consumers, the utilities subsector has the highest average recovery rate (near 98%) and the highest dollar-weighted recovery rate (73%) (see chart 17).

Utilities tend to have higher recoveries than the other infrastructure industries for several reasons. With essential assets and as providers of essential services, such as power, water, and gas, utilities often have little or no practical substitute. This supports a business model shielded by regulation, and issuers' ability to add debt is often limited by regulation or secured bond indentures. These characteristics support the very low industry risk for the regulated utilities subsector, which is the only corporate subsector to have this assessment from S&P Global Ratings.

Recovery By Default Type

Recovery rates can also vary with the type of default that the infrastructure entity undergoes. In nonbankruptcy restructurings, which include distressed exchanges, along with defaults that are cured within or outside of the grace period, the recovery rates for unsecured and subordinated debt tend to be higher, on average, than they would be through a bankruptcy restructuring.

The average recovery across all instruments is higher, at 83%, following a nonbankruptcy restructuring than the 80% average following a bankruptcy (see table 13). By debt type, the difference stems from senior unsecured and subordinated debt, which have an average recovery that is modestly higher (by 4.5 percentage points) following a nonbankruptcy restructuring. Senior secured debt experienced an average 78% recovery following a bankruptcy, near the 77% average recovery following a nonbankruptcy restructuring.

Table 12

A higher share of instruments from infrastructure and project finance have experienced the highest recoveries--par or greater--than have instruments from nonfinancial corporates. Overall, 53% of defaulted infrastructure debt instruments recovered par or greater, compared with 28.5% of nonfinancial corporate debt instruments (see chart 18).

At the low end of the distribution, just 8% of defaulted infrastructure debt instruments recovered 20% or less, compared with about 25% of nonfinancial corporate debt instruments. However, the sample size of infrastructure recoveries remains considerably smaller than that of nonfinancial corporates.

The distribution of project finance recoveries tends to vary more widely than that of corporate infrastructure. Nearly as many project finance instruments recovered par or greater as experienced recoveries at the low end of 20% or less (see chart 19).

Among other projects with higher recoveries, many tend to benefit from characteristics similar to those of regulated utilities. For instance, infrastructure project financings hold essential assets, have stable cash flows, and exhibit low industry risk. In many cases, low industry risk may be supported by long-term offtake contracts, such as power purchase agreements or availability-based payments under a concession, operations and maintenance contracts, or fuel-supply contracts.

The distribution of recoveries within corporate infrastructure tends to be more highly concentrated at the high end of the recovery range. The majority of defaulted debt instruments (about 58%) from corporate infrastructure experienced recoveries of par or greater, while less than 5% of instruments recovered 20% or less (see chart 20).

Appendix I: Definitions And Methodology

A central question in creating this study was how to determine which entities to include in the dataset. Infrastructure investment is broadly defined as investment exposure (debt or equity) to real assets vital to a country's economic development and prosperity that provide essential services for the orderly operations of an economy, such as transportation networks, health and education facilities, communication networks, and water and energy distribution systems. Key risks are in design, permitting, construction, operations, volume, and maintenance, with debt repaid predominantly by cash generated from the assets themselves.

Characteristics of infrastructure entities vary, and the variety of risks managed by infrastructure investments is broad (such as greenfield/brownfield, contracted/regulated/merchant, and availability/demand risk). Characteristics sought by investors in infrastructure include longevity, strong competitive advantages, high barriers to entry and monopolistic characteristics, low volatility, differentiation between return and volatility, and inflation-linked or predictable cash flows.

Our third annual infrastructure default and rating transition study continues the analysis of the inaugural study, as well as expands on and supersedes project finance studies published by S&P Global Ratings in previous years. The data herein encompasses project finance ratings from past reports and adds ratings from other segments within infrastructure. The advantage of this report over its predecessors is its broader perspective and larger dataset, which allow for more conclusions and increase the robustness of the findings while enabling comparisons between infrastructure issuer types and subsectors.

Our long-term infrastructure default and rating transition study uses long-term ratings on individual project finance debt issues. For projects with multiple tranches of debt, we used a single rating for all pari passu tranches to mitigate the risk of overweighting such projects in the sample set. For projects with senior and subordinated debt with different ratings, we counted each class separately and collapsed multiple tranches of pari passu debt into a single rating.

An infrastructure rating reflects S&P Global Ratings' opinion of the obligor's ability and willingness to meet its financial commitments on time, and it generally indicates our opinion of the likelihood of default regarding the specific rated issuance. Unlike issuer credit ratings, project finance ratings consider the value of collateral and recovery in determining the rating.

We analyzed the rating histories of corporate infrastructure and project finance credits that S&P Global Ratings first rated between Dec. 31, 1980, and Dec. 31, 2019. We categorized these ratings into one of five subsectors: utilities, power, oil and gas, social infrastructure, or transportation, with other obligors grouped in a separate category.

The analysis excludes ratings based on the guarantee of another entity and obligations wrapped by a monoline insurer, unless the obligations bear an S&P Underlying Rating (SPUR), in which case we included the SPUR. Corporate issuers, structured finance vehicles, public-sector issuers, and sovereign issuers are the subjects of separate default and transition studies, so we have excluded them from this study. The infrastructure data makes up a sufficiently large set, although individual subsectors are small, so that a few defaults or rating transitions can have a sizable effect on portfolio performance within those groups.

Our ongoing enhancement of the database used to generate this study could lead to outcomes that differ to some degree from those reported in previous studies. However, this poses no continuity problem because each study reports statistics back to Dec. 31, 1980. Therefore, each annual default study is self-contained and effectively supersedes all previous versions.

Withdrawn ratings

When a rating is withdrawn, it is changed to not rated (NR). Issue ratings are withdrawn when an obligation is repaid or at the request of the issuer. S&P Global Ratings also withdraws ratings when sufficient information to maintain a rating is unavailable. If we withdrew a rating after the issue defaulted, we counted it as a default.

Definition of default

A default is assumed to take place on the date S&P Global Ratings revised the rating to 'D', which could occur when a payment on the issue is missed, a distressed exchange offer is completed, or the issuer filed for or was forced into bankruptcy. When an issue defaults, it is not uncommon for S&P Global Ratings to subsequently withdraw the 'D' rating.

For the purposes of this study, if an issue defaults, we end its rating history at 'D'. For those instances when an issue defaulted and was subsequently upgraded, we entered the rating history after the default into the calculations as a new rating.

Many practitioners use statistics from this default study to estimate "probability of default" and "probability of rating transition." It is important to note that S&P Global Ratings' credit ratings do not imply a specific probability of default.

Static pool methodology

We conduct our default studies on the basis of groupings called static pools. For the purposes of this study, we form static pools by grouping issues by rating category at the beginning of each year, quarter, or month that the database covers. Each static pool is followed from that point forward. All issues included in the study are assigned to one or more static pools. When an issue defaults, we assign that default back to all of the static pools to which the issue belonged.

We use the static pool methodology to avoid certain pitfalls in estimating default rates. This is to ensure that default rates account for rating migration and to allow for default rates to be calculated across multiperiod time horizons.

The pools are static in the sense that their membership remains constant over time. Each static pool can be interpreted as a buy-and-hold portfolio. Because errors, if any, are corrected by every new update and because the criteria for inclusion or exclusion of issues in the default study are subject to minor revisions as time goes by, it is not possible to compare static pools across different studies. Therefore, every update revises results to the same starting date of Jan. 1, 1981, to avoid continuity problems.

Because static pools only include issues with active ratings as of the beginning date of a given pool, we exclude those with withdrawn ratings, as well as those that have defaulted, from subsequent static pools. If the rating on an issue is withdrawn after the start date of a particular static pool and the issue subsequently defaults, we will include it in that static pool as a default and categorize it in the rating category of which it was a member at that time.

For instance, the 1983 static pool consists of all issues rated as of 12:01 a.m. on Jan. 1, 1983. Adding those issues first rated in 1983 to the surviving members of the 1983 static pool forms the 1984 static pool. All rating changes that took place are reflected in the newly formed 1984 static pool through the ratings on these issues as of 12:01 a.m. on Jan. 1, 1984. We used the same method to form static pools for 1985-2020.

From Jan. 1, 1981, to Dec. 31, 2020, a total of 776 first-time-rated project finance issues were added to form new static pools, while we excluded 40 defaults and 423 issues with last ratings of NR (after they were included in the initial 776 in prior static pools).

Consider the following example: An issue is originally rated 'BB' in mid-1996 and is downgraded to 'B' in 1998. This is followed by a rating withdrawal in 2000 and a default in 2003. We would include this hypothetical issue in the 1997 and 1998 pools with the 'BB' rating, which was the rating at the beginning of those years. Likewise, it would be included in the 1999 and 2000 pools with the 'B' rating. It would not be part of the 1996 pool because it was not rated as of the first day of that year, and it would not be included in any pool after the last day of 2000 because the rating had been withdrawn by then. Yet each of the four pools in which this company was included (1997-2000) would record its 2003 default at the appropriate time horizon.

Transition matrices

Transition rates compare issue ratings at the beginning of a period with the ratings at the end of the period. To compute one-year rating transition rates by rating category, we compared the rating on each entity at the end of a particular year with the rating at the beginning of the same year. An issue that remained rated for more than one year was counted as many times as the number of years it was rated.

For instance, an issue continually rated from the middle of 1984 to the middle of 1991 would appear in the seven consecutive one-year transition matrices from 1985 to 1991. If the rating on the issue was withdrawn in the middle of 1991, it would be included in the column representing transitions to NR in the 1991 transition matrix. Similarly, if it defaulted in the middle of 1991, it would be included in the column representing transitions to 'D' in the 1991 one-year transition matrix.

All 1982 static pool members still rated on Jan. 1, 2021, had 39 one-year transitions, while companies first rated on Dec. 31, 2019, had only one. Each one-year transition matrix displays all rating movements between letter categories from the beginning of the year through year-end. For each rating listed in the matrix's leftmost column, there are nine ratios listed in the rows, corresponding to the ratings from 'AAA' to not rated.

The only ratings considered in these calculations are those on entities at the beginning of each static pool and those at the end. All rating changes that occur in between are ignored. For example, if an issue was rated 'A' on Jan. 1, 2014, and was downgraded to 'BBB' in the middle of the year and then upgraded to 'A' later in the year (with no other subsequent rating changes), this instrument would be included only in the percentage of issues that began the year as 'A' and that ended the year as 'A'. This also applies to transition matrices that span longer time horizons. If an issue defaults or if the rating is withdrawn in the middle of the year, then it would be considered rated 'D' or not rated as of Dec. 31 of that particular year.

Multiyear transitions

Multiyear transitions were also calculated for periods of two up to five years. In this case, we compared the rating at the beginning of the multiyear period with the rating at the end. For example, three-year transition matrices were the result of comparing ratings at the beginning of the years 1981-2018 with the ratings at the end of the years 1983-2020.

Otherwise, the methodology was identical to that used for single-year transitions. We calculated average transition matrices on the basis of the multiyear matrices just described. These average matrices are a true summary, the ratios of which represent the historical incidence of the ratings listed in the first column changing to the ones listed in the top row over the course of the multiyear period. Transition matrices that present averages over multiple time horizons are also calculated as issue-weighted averages.

Rating modifiers

We use rating modifiers (plus and minus signs) to calculate upgrade and downgrade percentages, as well as the magnitude of rating changes, throughout this study. However, some transition tables may use full rating categories for practical reasons: The use of a rating category suggests that transitions to 'AA-' from 'AA', for example, or to 'BBB-' from 'BBB+' are not considered to be rating transitions because the rating remained within the rating category.

Issue-weighted statistics

Averages that appear in this study are calculated based on the number of issues, rather than the dollar amounts, affected by defaults or rating changes. Although dollar amounts provide information about the portion of the market affected by defaults or rating changes, we believe issue-weighted averages are more useful measures of the performance of ratings.

Average cumulative default rate calculation

The cumulative default rates in this study average the experience of all static pools by first calculating marginal default rates for each possible time horizon and for each static pool, weight-averaging the marginal default rates conditional on survival (survivors being nondefaulters), and accumulating the average conditional marginal default rates.

We calculated conditional default rates by dividing the number of issues in a static pool that default at a specific time horizon by the number of issues that survived (did not default) to that point in time. Weights are based on the number of issues in each static pool. Cumulative default rates are one minus the product of the proportion of survivors (nondefaulters).

Utilities

Utilities make up the largest group of ratings within infrastructure, with 549 issues, all of which are rated under corporate finance. These ratings are applied using our corporate ratings methodology and generally have above-average credit quality. These entities are also corporations that offer essential or near-essential infrastructure products, commodities, or services with little or no practical substitute (mainly electricity, water, and gas)--a business model that is shielded from competition (naturally, by law, by shadow regulation, or by government policies and oversight) and subject to comprehensive regulation by a regulatory body or implicit oversight of its rates (sometimes referred to as tariffs), service quality, and terms of service.

The 'BBB' category has been the largest for utility ratings since 2000 and currently represents 45% of ratings, but the 'A' category is nearly as large, representing 39% of ratings (see chart 21). Just 12% of issues are rated speculative-grade--tied for the lowest with the "other" category among subsectors.

Oil and gas

The oil and gas subsector has the second-highest number of ratings, at 216. Of these, 191 are in corporate infrastructure and 25 are in project finance. These numbers are further split between midstream oil and gas pipelines and refining, the latter being more exposed to earnings volatility.

The 'BBB' category is the largest for oil and gas, representing 45% of ratings (see chart 22). Since 2012, about half of ratings have been speculative-grade on average, and currently 50% are speculative-grade--the second-highest proportion among infrastructure subsectors.

Transportation

Transportation includes road, rail, bridge, port, airport, and tunnel projects. The subsector has 164 ratings, 89 of which are in corporate groups and 75 in project finance. Some transportation entities are frequently exposed to competitive pressures from other modes of transit. For instance, a toll road could be affected by neighboring roads or other modes of transportation that could draw drivers and their tolls from the toll road. Even in the absence of competition, transportation often correlates directly with the local economy because of increases in utilization and economic growth.

The largest rating category for transportation ratings has been 'BBB' since 2003, and it currently represents 41% of ratings (see chart 23). About 27% of transportation ratings are speculative-grade.

Power

Power, which includes electricity generation and transmission power projects, has 160 ratings. These ratings are primarily for fossil fuel power plants, although renewable energy plants are also included. Power plants range from small, aging coal plants to newer, highly efficient natural gas projects.

Included within power are eight ratings for project developers. We define a project developer as a corporate entity, typically structured as a holding company, that owns, develops, and operates assets mainly in the energy and infrastructure sectors. Typically, a developer conducts its business chiefly through operating subsidiaries.

The operating subsidiaries are generally financed with nonrecourse debt, either on a corporate finance or project finance basis. Developers have little or no operations of their own and rely on distributions--such as dividend and fee income received from a diversified portfolio--to service their debt and other expenses. Operating assets are mainly in the energy, infrastructure, utility, and transportation industries. Common examples include power plants, toll roads, electric utilities, drill ships, and ports.

The 'BBB' category is the largest for power, representing 39% of ratings, but the 'BB' category is nearly as large, representing 37% of ratings (see chart 24). Since 2002, between half and three-fifths of ratings have been speculative-grade; currently, 54% of issues are rated speculative-grade--the most among infrastructure subsectors.

Social infrastructure

Social infrastructure has 92 ratings, all but one for project finance transactions. Financings in this area cover:

• Primary and secondary schools and tertiary teaching facilities;
• Health care projects, from large and relatively complex regional hospitals to smaller local primary care and psychiatric facilities;
• Social housing, military barracks, and student accommodations; and
• Detention centers (prisons) and judicial facilities.

The 'BBB' category is the largest for social infrastructure, representing 50% of ratings (see chart 25). About 18% of ratings are speculative-grade--the third lowest among subsectors.

Other ratings

Finally, this study includes 93 other ratings. This number of uncategorized ratings has fluctuated over time, with the highest count in 2008 at 165. These ratings are primarily project finance ratings that don't fit into one of the categories above or ratings on international public finance entities that finance infrastructure but are rated with different criteria. Such entities include social housing developers and ports.

The 'A' category is the largest for the other category, representing 52% of ratings (see chart 26). About 12% of ratings are speculative-grade--tied for the lowest among subsectors.

Table 13

Appendix II: Additional Exhibits

Table 14

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

Table 22

Table 23

Table 24

Table 25

Table 26

Table 27

Table 28

Table 29

Table 30

Table 31