The international public finance (IPF) sector is divided into two main categories: local and regional governments (LRGs), like states, provinces, regions, cities, and towns; and non-LRGs, like non-U.S. universities, hospital systems, transportation systems, and housing providers. At the start of 2022, S&P Global Ratings rated 271 LRG issuers and 143 non-LRG issuers. The IPF sector is also divided regionally: Europe, the Middle East, and Africa (EMEA; 225), Canada (70), Asia-Pacific (69), and Latin America (59).

Overall credit quality improved in IPF last year. S&P Global Ratings raised 38 IPF ratings (37 LRG and one non-LRG) and lowered eight (three LRG and five non-LRG), compared with 34 upgrades and 18 downgrades in 2021 (see chart 1). There were no defaults among IPF issuers rated by S&P Global Ratings in 2022, down from two in 2021, marking the first year since 2017 with no defaults. Since 1975, the first year with a default among IPF issuers was 1999, and thereafter defaults have averaged just over two a year. All but one default has been among LRG issuers.

Rating Distributions

IPF ratings have increased to an all-time high of 425 at the start of 2022 from two in 1975 (see chart 2). The share of issuers with speculative-grade ratings was zero until 1996, rising to a high of 31% at the start of 2008, before falling steadily after the global financial crisis to the current level of 14% to start 2022. The number of 'AAA' rated issuers started the year at 30 and ended with 39 (eight upgrades to 'AAA' and one new rating).

Regionally, 'AAA' rated issues are concentrated in EMEA and Canada, with 9% and 21%, respectively, followed by 5% for Asia-Pacific (see chart 3). There are no 'AAA' rated IPF issuers in Latin America. Looking at investment- versus speculative-grade, Canada, EMEA, and Asia-Pacific are almost entirely investment-grade (100%, 96%, and 97%, respectively), whereas Latin America is 15% investment-grade.

Rating Changes

For IPF, upgrades rose to 8.9% of total ratings in 2022 from 8.1% in 2021, higher than the historical weighted average since 1975 of 7.9%. Downgrades fell to 1.9% from 3.8%, much lower than the historical weighted average of 6.7% (see table 1). There were no defaults in 2022, leading to a 0% default rate, down from 0.5% in 2021, compared to the historical weighted average of 0.3%. The stability rate (the proportion of ratings that were the same at the start and end of the year) rose slightly to 83.1% in 2022 from 82.9% in 2021, compared to the historical weighted average of 79.9%. The ratio of downgrades to upgrades fell to 0.2 to 1 in 2022 from 0.5 to 1 in 2021, lower than the historical weighted average of 0.85 downgrades to upgrade.

Table 1

The two main IPF sectors in 2022 diverged in credit quality. Of the 286 LRG ratings at the start of the year, 12.9% were upgraded, compared to 0.72% of the 139 non-LRGs, while 1.1% of LRGs were downgraded in 2022, compared to 3.6% of non-LRGs (see table 2 and chart 4).

Table 2

IPF credit quality improved across all regions. Canadian IPF improved dramatically, with 32.9% of the 70 ratings upgraded and zero downgrades. These upgrades followed the revision of S&P Global Ratings' institutional framework assessment for Canadian municipalities to extremely predictable and supportive, up from very predictable and well balanced (see "Various Rating Actions Taken On Canadian Municipal Governments On Improved Institutional Framework Assessment," June 1, 2022).

The largest region, EMEA, saw the smallest improvement in credit quality in 2022, with 3.5% upgrades and 2.6% downgrades. Offsetting that, EMEA also saw the highest stability rate, at 91.4%.

Rating Outlooks Imply Most Expected Weakness Ahead For EMEA

The population of IPF issuers with stable outlooks fell in 2022 to 330 from 350 in 2021, while positive outlooks increased to 31 from 21 and negative outlooks fell slightly to 50 from 52 (see chart 5). There were also two CreditWatches in 2022, one with positive implications and one with negative. Swan Housing Assn. Ltd. of the United Kingdom was placed on CreditWatch with positive implications on Dec. 1, before being upgraded to 'BBB+' from 'BB-' in February 2023 upon joining the Sanctuary Housing Group. Grand Delta Habitat SAS of France was placed on CreditWatch with negative implications on June 15, before being downgraded to 'A' from 'A+' in January 2023 upon merging with Vallis Habitat. Both issuers currently have stable outlooks.

Rating Transitions Demonstrate The Strength Of Our Rating System

Consistent with other S&P Global Ratings studies, IPF ratings over time show a negative correlation between higher ratings and incidents of default, and a positive correlation between higher ratings and rating stability. Over one year (see table 3), no investment-grade issuers have defaulted, and the higher the speculative-grade rating, the lower the default rate, starting with 0.8% for 'BB', 2.1% for 'B', and 12.6% for 'CCC/C'. For stability rates, the higher the rating, the more stable over a year on average, with 'AAA' ratings stable 94.6% of the time, 'AA' stable 93.3% of the time, and 'A' ratings stable 90.0% of the time.

Since most of the observed IPF ratings have been high investment-grade, as we descend the ratings, we see some idiosyncrasies due to smaller populations. For example, 81.8% of 'BBB' ratings were stable over one year, compared with 86.0% of 'BB' ratings. The population of 'BBB' and 'BB' ratings are currently the same, but a 'BB' rating has historically been much more common than 'BBB'.

As expected, over longer time horizons, the stability of ratings decreases. The stability of 'AAA' falls to 74.6% over five years, for example. For the longest time frame, which in this study is 15 years, the numbers can appear striking, with 39.0% of 'AAA' ratings stable on average. The static pools included in the 15-year transition matrices are 1975-2008.

Table 3

Table 4 looks at transitions over time for non-U.S. local and regional governments. The LRG population is about two-thirds of the IPF sector yet it represents all IPF defaults but one. The LRG population also contains a more diverse ratings distribution, with most IPF speculative-grade ratings found in LRG.

Table 4

For the most part, non-LRG ratings behave similarly to LRG ratings, with the major exception being the small population of speculative-grade ratings (maxing out at 8% of all non-LRG ratings in January 2008).

Table 5

Regional LRG populations vary greatly in size and rating distribution. Table 6 shows one-year transition rates for the five major regions we follow. The Canadian, EMEA, and Asia-Pacific regions are almost entirely investment-grade, while Latin America is almost entirely speculative-grade.

Table 6

Table 7 shows the geographic differences between the LRG and non-LRG populations, especially in Asia-Pacific and Latin America.

Table 7

No Defaults In 2022

There were no defaults among IPF issuers in 2022, down from two in 2021. Historically, as seen in table 8, defaults over all time horizons are concentrated in the lowest speculative-grade ratings. There have only been two defaults among issuers that ever held an investment-grade rating.

Table 8

There has only been one non-LRG default, when Monstransavto defaulted in 2008.

Table 9

There have been several instances of clustered defaults among IPF issuers, usually related to specific periods of financial instability in a country: Russia in 1999, Mexico in 2012, and Argentina in 2020.

Table 10

Gini Ratios And Lorenz Curves

In addition to examining the absolute performance of IPF ratings, we also assess their relative performance. One technique we use to measure relative performance is the Gini coefficient, a key measure of the relative ability of ratings to differentiate risk. Another is Lorenz curves, which are graphical representations of the proportionality of a distribution.

We believe comparing Lorenz curves and Gini coefficients is useful. Lorenz curves for IPF closely resemble the ideal Lorenz curves (see charts 6, 7 and 13-17). For more detail regarding the derivation and construction of Lorenz curves and Gini coefficients, see "Gini coefficient calculation" in the appendix.

Chart 6 shows the one-year weighted average Lorenz curve for LRGs. The points along the Lorenz curve represent different ratings, starting with the 'CCC' and lower categories on the far left. The first point indicates the obligations rated 'CCC' and lower represent 36.67% of defaults within one year. As the ratings rise, the cumulative percentage of defaults for each rating increases in the 'B' category and then reaches 100% at the 'BB' rating. From that point, there are no defaults at higher ratings. The 2021 one-year Gini coefficient was 0.9 (90%).

Table 11 provides the coordinates in the LRG Lorenz graph. The lowest value in the left column shows the cumulative proportion of defaults at the 'CCC' through 'C' categories, or 36.7%, and the number to the right is the cumulative proportion of the rated universe, 1.3%. This means 36.7% of the defaults came from 1.3 % of the ratings, all of which were in the 'CCC' or lower categories. Moving up the table, once ratings reach 'BB', representing 19.1% of all ratings, 100% of defaults have occurred.

Table 11

Table 12

LRG And Sovereign Ratings

Some LRG ratings are tied to sovereign ratings. LRG ratings may be capped at the sovereign rating, particularly if the LRG is not expected to maintain stronger credit characteristics than the sovereign in a stress scenario, if the issuer does not have an institutional framework that is predictable and stable and limits the risk of negative sovereign intervention, or if the issuer does not have the ability to mitigate intervention from the sovereign.

From the 'AAA' category to 'B', LRG ratings and sovereign ratings remain in those categories in almost the same proportions from one year to 15 years. However, LRG 'AAA' ratings are much more fluid than sovereign ratings at the highest level. After one year, 94.6% of LRG 'AAA' ratings maintain that rating, similar to 96.9% of sovereign ratings. By the 15th year, just 36.6% of 'AAA' LRGs maintain that rating, far lower than the 66.7% of sovereigns (see chart 8).

For 'B' category ratings, 71.7% of LRGs remain in that category after one year, compared with 87.1% of sovereigns. By the 10th year, just 9.9% of 'B' category LRGs maintain that rating, compared with 41.7% of sovereigns (see chart 9). The steadily decreasing percentage for LRGs is largely the result of ratings withdrawals.

Rating Distributions

There were eight upgrades to 'AAA' among LRG ratings in 2022 (see chart 10), five of which were Canadian: the cities of Hamilton and Guelph, and the counties of Essex (located in Ontario across the border from Detroit), Oxford (an agricultural county between London and Hamilton), and Wellington (which includes the city of Guelph). Two additional upgrades to 'AAA' were Swiss Cantons: Aargau (norther Switzerland, upgraded on management's commitment to tight fiscal policy, limited deficits, and strong economy) and Basel-Country (northern Switzerland, committed to fiscal discipline, diversified and solid economy).

There are five 'AAA' non-LRG issuers at the start of 2023, the same as last year (see chart 11). There was one fallen angel in 2022: the social housing issuer Swan Housing Assn. Ltd. There were no rising stars among non-LRG issuers.

Appendix I: Methodology And Definitions

This IPF default and rating transition study uses an internal database of long-term local currency issuer credit ratings. It covers LRGs and other IPF entities outside the U.S. and uses public and confidential ratings. For LRGs with national scale ratings only--mainly those in Mexico and Russia--we use the equivalent local currency rating in S&P Global Ratings' global scale. This study uses LRG issuer ratings and not the ratings on specific issues of debt by these governments. We do not make any adjustment for the size of the government or for the amount of debt issued by each government.

Our ongoing improvements to 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, 1974. Therefore, each annual default study is self-contained and effectively supersedes all previous versions. There are no changes in methodology in this study from the IPF default and rating transition study of 2021, which was published last year. The inclusion of non-LRG ratings within IPF began in the 2017 study and continues this year.

Issuers included in this study

This study analyzes the rating histories of IPF entities that S&P Global Ratings rated as of Dec. 31, 1974, or that were first rated between that date and Dec. 31, 2021. This includes defaults on an LRG's commercial debt, including bank and bond debt. We have included both public and confidential ratings when tallying the number of defaults. Even if a rating was withdrawn in a particular year and the entity went on to default on commercial debt during that year or in a subsequent year, we included the default in our statistics for that year.

For confidential ratings, we preserve anonymity by referring to the entities, typically LRGs, as "LRG 1," "LRG 2," etc. When ratings are withdrawn at the request of a particular LRG, we indicate this by the abbreviation "NR," signifying that the LRG is not rated.

Issuers with withdrawn ratings

S&P Global Ratings withdraws ratings when an entity's entire debt is paid off or when the program or programs rated are terminated and the relevant debt extinguished. Others are withdrawn because of a lack of cooperation, particularly when an entity fails to provide all the information needed to continue surveillance on the ratings, or at the entity's request.

Definition of default

An obligor rated 'D' (default) or 'SD' (selective default) is in payment default on one or more of its financial obligations (rated or unrated) unless S&P Global Ratings believes that such payments will be made within five business days in the absence of a stated grace period or within the earlier of the stated grace period or 30 calendar days. S&P Global Ratings also lowers a rating to 'D' when an issuer takes an action that could jeopardize payments on a financial obligation, including the rare instance of a public finance entity filing for bankruptcy.

A 'D' rating is assigned when S&P Global Ratings believes that the default will be a general default and that the obligor will fail to pay all or substantially all of its obligations as they come due. S&P Global Ratings assigns an 'SD' rating when it is believed that the obligor has selectively defaulted on a specific issue or class of obligations but will continue to meet its payment obligations on other issues or classes of obligations in a timely manner. A selective default includes the completion of a distressed exchange offer, whereby one or more financial obligations are either repurchased for an amount of cash or replaced by other instruments having a total value that is less than par.

We deem 'D' and 'SD' ratings to be defaults for the purposes of this study. A default is assumed to take place on the earliest of the date the rating was revised to 'D' or 'SD', the date a debt payment was missed, the date a distressed exchange offer was announced, or the date the debtor filed for or was forced into bankruptcy. When an issuer defaults, it is not uncommon for the 'D' rating to be subsequently withdrawn. For the purposes of this study, if an issuer defaults, we end the obligor's rating history at 'D'. If any defaulting entity reemerges from bankruptcy or otherwise restructures its defaulted debt instruments, thereby reestablishing regular, timely payment of all its debts, we reenter it into the database as a new obligor. The rating history on the obligor after the default event is included in all calculations as entirely separate from the experience leading up to its earlier 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 obligors 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 issuer credit ratings included in the study are assigned to one or more static pools. When an issuer defaults, we assign that default to all of the static pools to which the issuer belonged.

We use the static pool methodology to avoid certain pitfalls in estimating default rates, such as by ensuring that default rates account for rating migration and can be calculated across multiple-period time horizons. Some methods for calculating default and rating transition rates might charge defaults against only the initial rating, ignoring more recent rating changes that supply more current information. Other methods may calculate default rates using only the most recent year's default and rating data, which may yield comparatively low default rates during periods of high rating activity because they ignore prior years' default activity.

The pools are static in the sense that they remain 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 bonds 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 Dec. 31, 1974, so as to avoid continuity problems.

Ratings that have been withdrawn--that is, revised to not rated--are surveilled with the aim of capturing a potential default. Because static pools include only entities with active ratings as of the beginning date of a given pool, we exclude withdrawn ratings, as well as ratings on bonds that have defaulted, from subsequent static pools. If the rating on an entity's bond series is withdrawn after the start date of a particular static pool and there is a subsequent default, we will include that series in that static pool as a default and categorize it in the rating category it was a member of at that time.

For instance, the 1975 static pool consists of all debt rated as of 12:00:01 a.m. on Jan. 1, 1975. Adding debt first rated in 1975 to the outstanding debt of the 1975 static pool forms the 1976 static pool. All rating changes that took place as of 12:00:01 a.m. on Jan. 1, 1976, are reflected in the newly formed 1976 static pool. We used the same method to form static pools for 1977-2022.

Consider the following example: A parity debt rating of 'BB' is assigned in mid-1986 and is lowered to 'B' in 1988, followed by a default in 1993. This hypothetical rating would be included in the 1987 and 1988 pools as 'BB' and in the 1989-1993 pools as 'B'. All pools that include this obligation would capture its 1993 default. It would not be part of the 1986 pool because it was not rated as of the first day of that year. Yet each of the pools in which this debt was included (1987-1993) would record its 1993 default at the appropriate time horizon.

Default rate calculations

We calculated annual default rates for each static pool, first in units and later as percentages with respect to the number of issuers in each rating category. We combined these percentages to obtain cumulative default rates for the 47 years the study covers (see tables 13 and 15-17).

Issuer-weighted default rates.  All default rates that appear in this study are based on the number of issuers rather than the dollar amounts affected by defaults or rating changes. Although dollar amounts provide information about the portion of the market that is affected by defaults or rating changes, issuer-weighted averages are more useful measures of the performance of ratings. Many practitioners use statistics from this default study to estimate "probability of default" and "probability of rating transition." It is important to note that our ratings do not imply a specific probability of default.

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 then accumulating the average conditional marginal default rates. We calculated conditional default rates by dividing the number of issuers in a static pool that defaulted at a specific time horizon by the number of debt 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).

For instance, the weighted average first-year default rate for entities rated in the 'BB' category was 0.82%, meaning that an average of 99.2% made payments in accordance with their terms for the first year (see table 15). Similarly, the second- and third-year conditional marginal averages were 1.28% (98.6% of those parity obligations that did not default in the first year did not default in the second year) and 1.07% (99.1% of those entities that did not default by the second year did not default in the third year either), respectively. Multiplying 99.2% by 98.6% results in a 97.8% survival rate at the end of the second year, or a two-year cumulative average default rate of 2.18%. Multiplying 97.8% by 99.1% results in a 96.9% survival rate at the end of the third year, or a three-year cumulative average default rate of 3.1%. (The table is based on numbers calculated to four decimals. Calculations using numbers rounded to two decimals, as presented here, may yield slightly different results.)

Cumulative default rate calculation.   We use cumulative default rates to average the experience of all static pools in a select period by calculating marginal default rates for each possible time horizon and for each static pool. The marginal default rates are then turned into survival rates by subtracting the marginal default rates from one. To get the cumulative default rates for a horizon, the survival rates for all previous horizons are multiplied and then subtracted from one.

In the context of LRG ratings, we treat governments that selectively default as complete defaults. In this analysis, when the LRG emerges from selective default, we then treat it as a new entity altogether. At some point, the cumulative default rate stops increasing as the time horizon lengthens, because no additional defaults are captured by the longer horizon.

Cumulative default rates differ from transition-to-default rates in that defaults are counted only once, as opposed to averaging default observations, which we do when we calculate transition rates for periods longer than 24 months. In addition, we use all annual static pools in the period, not just the static pools that have enough years to complete the horizon.

Transition analysis

Transition rates compare issuer 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 parity obligation at the end of a particular year with the rating at the beginning of the same year. An obligation that remained rated for more than one year was counted as many times as the number of years it was rated. For instance, an obligation continuously rated from the middle of 1987 to the middle of 1991 would appear in the four consecutive one-year transition matrices from 1988-1991. If the rating on the obligation was withdrawn in the middle of 1991, it would be included in the column representing transitions to not rated 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 1975 static pool members still rated on Jan. 1, 2022, had 47 one-year transitions, while those first rated on Jan. 1, 2021, had only one. Longer time horizons (three or more years) have fewer transitions. Various tables in this report display the transitions for non-U.S. LRG ratings over several time horizons. Each transition matrix displays all rating movements between rating categories from the beginning of the period through the end.

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 entity rated 'A' on Jan. 1, 2021, 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 entity would be included only in the percentage of issuers that began the year as 'A' that ended the year as 'A'. This also applies to transition matrices that span longer time horizons. If an obligation defaults or the rating is withdrawn in the middle of the year, then we consider it to be either rated 'D' or not rated as of Dec. 31 of that particular year.

Multiyear transitions.   Multiyear transitions were also calculated for periods of two to 20 years. In this case, we compared the rating at the beginning of the multiyear period with the rating at the end. For example, five-year transition matrices were the result of comparing ratings at the beginning of the years 1974-2017 with the ratings at the end of the years 1979-2022. 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 issuer-weighted averages.

Comparing transition rates with default rates.   Rating transition rates may be compared with the marginal and cumulative default rates described in the previous sections. For example, the one-year default rate column of table 2 is equivalent to the one-year column in table 4.

Initial-to-last transition and default rates.   These transition rates compare debt ratings from the time of first rating to the last rating, regardless of the time elapsed in the interim. They provide a road map to all of the historically observed ratings held by non-U.S. LRGs during their lifetimes.

Gini coefficient calculations

The Gini coefficient, developed by Corrado Gini, is a summary statistic of the Lorenz curve and shows the accuracy of ratings' rank ordering. The Lorenz curve was developed by Max O. Lorenz as a graphical representation of the proportionality of a distribution. For this study, the Lorenz curve is built by plotting the cumulative proportion of parity obligations by rating category (from lowest to highest) with the cumulative proportion of defaulters by rating category. For example, if 'C' parity obligations represented 1% of the total parity obligations (x-axis) and 20% of the defaulters (y-axis), then the coordinate (1, 20) would be the first point on the curve.

To determine the relative performance represented by the Lorenz curve, we compare it with the random curve and the ideal curve. If the ratings' rank ordering only randomly approximated default risk, the Lorenz curve would fall along the diagonal, and its Gini coefficient would be zero.

If ratings were perfectly rank ordered so that all defaults occurred only among the lowest-rated obligations and all obligations with the lowest rating defaulted, the curve would be the ideal curve, and the Gini coefficient would be 1.

The Gini coefficient is a ratio of two areas, as illustrated in chart 17, and is derived by dividing area B by the total area A plus area B. In other words, the Gini coefficient captures the extent to which actual ratings accuracy diverges from the random scenario and approaches the ideal.

The Gini coefficient can be calculated for different lengths of time--for example, one year or three years; or for a single period, such as the year ended 2021; or by aggregating a series of one-year periods. To calculate the one-year Gini for 2022, we identified the parity obligation ratings at the beginning of the year and determined which did and did not default during the year. Then we calculated the proportion of parity obligations at each rating and the proportion of defaulters at each rating, based on the ratings at the beginning of the period. The aggregate Gini for 1975-2022 combines data for each of the 47 one-year periods. The same parity obligation may be counted multiple times in the aggregate Gini coefficient.

Time samples

This update limits the reporting of default rates to the 15-year time horizon. However, we gathered the data for 47 years, and all calculations are based on the rating experience of that period. In addition, average default statistics become less reliable at longer time horizons as the sample size becomes smaller and the cyclical nature of default rates has a bigger effect on averages.

Default patterns share broad similarities across all static pools, suggesting that S&P Global Ratings' rating standards have been consistent over time. Adverse conditions tend to coincide with default upswings for all pools. Speculative-grade issuers have been hit the hardest by these upswings, but investment-grade default rates also increase in stressful periods.

Table 13

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

Table 32

Table 33

Table 34

Table 35

Table 36

Related Research

• 2022 Annual Global Sovereign Default And Rating Transition Study, April 28, 2023