Only one USPF bond rated by S&P Global Ratings defaulted in 2018. This is the lowest number since 1997, and 2018 was just the fifth year since 1986 with one or no defaults. The overall default rate was 0.004%, down from 0.085% in 2017, when there were 20 defaults. The total number of ratings at the start of 2018 was slightly lower than in 2017. This is the first time the total number of USPF ratings declined year over year, with 23,500 at the start of 2018 compared with 23,578 at the start of 2017 (including housing). The total number of 'CCC'/'C' ratings--which often precede a default--was 27 (25 non-housing and two housing) at the beginning of 2018, down from 34 (32 non-housing and two housing) in 2017. As of Jan. 1, 2019, there were 31 'CCC'/'C' ratings (14 non-housing and 17 housing).

S&P Global Ratings' forecast for U.S. GDP growth is 2.3% in 2019, lower than 2.9% in 2018. Slower economic growth could have a negative effect on municipal ratings--or at least mitigate the trend of upward movement from other factors. Upgrades increased to 1,427 in 2018, driven by criteria updates, which accounted for 695 of them. Downgrades fell to 684 from 728, with the most commonly cited reason being deteriorating issuer financial conditions. The ratio of upgrades to downgrades was 2.09 in 2018, up from 1.28 in 2017. However, excluding upgrades and downgrades based on changes to criteria, the ratio is almost identical (1.29 in 2018 against 1.28 in 2017).

Overall Rating Movement Was Positive, But Sector Movement Was Mixed

The largest sector, local governments, had the vast majority of USPF upgrades and downgrades (see Table 1). Looking at the smaller sectors, the results overall were mixed, with housing, higher education, health care, and charter schools having more downgrades than upgrades. There were 84 state government upgrades and 60 downgrades that stemmed from the two general obligation (GO) upgrades and downgrades.

Table 1

Data constraints prevent this study from distinguishing local government from state ratings in most instances. These two sectors are combined into two rating types: tax secured and appropriation. The tax-secured category covers state or local government GOs, while appropriation debt is subject to annual appropriations by issuers and is typically rated one notch lower than the GO rating.

Most tax-secured and appropriation ratings are for local government, so local government rating movement generally determines the direction of changes for these two rating types. Since 2015, S&P Global Ratings has categorized every tax-secured and appropriation rating action as either state or local government in its quarterly rating reports. Local government rating movement has been consistently positive, whereas state ratings were lowered more frequently in 2016 and 2017. The general pattern of upgrades surpassing downgrades for the tax-secured and appropriation sectors is largely due to the raising of local government ratings.

The high number of state rating actions deserves further explanation. S&P Global Ratings maintains GO ratings on each of the 50 states, plus some U.S. territories. When one of these ratings change, frequently many other ratings move at the same time, affecting dozens or even up to about 100 ratings. In addition, there are some special ratings at the state level that are not associated with the GO. In 2018, there were upgrades to the state GO ratings of Michigan and Minnesota and downgrades to Connecticut and Kentucky. These are indicated in the second row of the above table.

There were 478 more USPF rating actions (522 more upgrades and 44 fewer downgrades) in 2018 than in 2017. This was the sixth-highest number of upgrades since 1986 and the seventh-highest number of downgrades. The 1,427 combined housing and non-housing upgrades in 2018 were the third-highest upgrade total since 2013, with 1,498 upgrades that year and another 2,419 in 2014. Both of these years also had a high number of rating changes from criteria in local government, the largest USPF sector. Downgrades have fallen in each of the past two years. In 2011, S&P Global Ratings lowered more than 1,000 housing ratings that were tied to the sovereign rating on the U.S., which was lowered to 'AA+' from 'AAA'.

Rating Transitions In 2018

USPF rating transitions in 2018 reflected stronger movement among ratings in the 'A' category or lower (see Table 2). Ratings in the two highest rating categories were more likely to slip to lower categories than to be raised due to having very little or no room for improvement. Nevertheless, just 1.77% of ratings in the 'AAA' and 1.14% of ratings in the 'AA' categories were lowered in 2018. We withdrew 3.29% of 'AAA' ratings and 4.77% of 'AA' ratings.

An average of 93.2% of USPF ratings in the three highest categories ended 2018 in the same category or higher. 'BB' rated entities were more likely to be upgraded than downgraded, while 'B' rated bonds were just as likely to be upgraded than downgraded. Unusually, there was extremely low default activity, even among the lower rating categories, with only one USPF rating that started the year at 'AA' defaulting in 2018.

Table 2

Criteria Changes Caused The Most Upgrades In 2018

Our quarterly reports found that criteria changes were the most common reason for upgrades in 2018, with 695. The most commonly attributed cause of downgrades in 2018 was finances, with 371 (see Chart 1). Criteria in local government was the main driver of upgrades for criteria. There was an unusually high number of criteria revisions in USPF in 2018, seven in total, three focused on local government. The other four criteria updates were two in utilities, and one each in health care and transportation. The significance of criteria changes was greatest in local government, which represents the majority of all USPF ratings. Please refer to the list at the end of this study for the criteria that triggered significant rating actions. The second most commonly cited cause of upgrades in 2018 was finances, with 490, and the second most common reason for downgrades was criteria changes, with 119.

The numbers of rating actions in this section do not match those throughout most of the report because the data here are based on quarterly activity as opposed to the entire year. See "Annual versus quarterly rating movement" in Appendix I for a description of the methodology used for tracking rating actions on a quarterly basis.

Rating Distribution Improves

Over the past two years, the number of ratings in some sectors and categories has changed noticeably. For example, 'AAA' in the appropriations sector increased by 45% (to 42 from 23) between 2016 and 2018. On the other end, 'CCC'/'C' housing ratings increased to 17 from two over the same timeframe. In general, Table 3 illustrates the improvement in ratings over the last two years.

Table 3

Historical Stability Of USPF Ratings

For the years covered in this study (Jan. 1, 1986 - Jan. 1, 2019), unenhanced debt (debt obligations not supported by financial guarantees, structuring techniques, multiple-party features, or other external credit support) rated by S&P Global Ratings generally has demonstrated significant credit stability throughout a wide variety of events, including various economic cycles, federal government mandates, and tax reform measures. The data suggest the following:

• There is a correlation between the rating and the likelihood of default, whereby higher-rated issues have a lower rate of default and vice versa.
• Higher-rated issues generally take longer to default than lower-rated issues, and a series of incremental rating changes typically precedes a default.
• Higher rating levels correspond to greater rating stability and vice versa.
• Rating distributions throughout the sector remained relatively stable despite economic stress, including the Great Recession and the slow recovery thereafter.

Because housing sector debt ratings reflect the credit quality of a mix of both public-sector entities and external credit enhancement providers, we present conclusions from this sector in separate tables. The following USPF debt types outside of housing are included in this study:

• GO
• Lease/appropriation/moral obligation
• Special tax (sales, gas, etc.)
• Special district
• Water and sewer revenue
• Public power
• Airports
• Ports
• Toll roads and bridges
• Parking
• Various types of bond pools
• Transit
• Public and private higher education
• Auxiliary higher education debt
• Independent schools
• Nonprofit entities and organizations
• Charter schools
• Hospitals (stand-alone and systems)
• Continuing care
• Physicians' practices

Our analysis of housing ratings in the text and tables includes all the types of municipal housing on which S&P Global Ratings maintains ratings. These include the following:

• GO for housing finance agencies and social housing providers, including public housing authorities and community development financing institutions
• Single-family whole loans
• Single-family U.S. government-supported mortgage-backed securities (MBS)
• Multifamily loans for affordable housing with U.S. government support
• Multifamily loans for affordable housing without an explicit or implied U.S. guarantee
• Military housing
• Mobile home parks
• Section 8 multifamily housing
• Capital fund securitization for public housing authorities

One Default In 2018

In 2018, one rated USPF bond defaulted, down from 20 in 2017, 15 in 2016, 12 in 2015, and eight in 2014 (see Chart 2 and 3). The default rate of USPF nonhousing ratings in 2018 was 0.005%, the lowest non-zero default rate in our records. The lone default--Platte County, Mo.--started 2018 at 'AA-' before being downgraded to 'A' in May, 'B-' in September, 'CC' in November, and then defaulting in December. The journey from 'AA-' to default began with insufficient sales tax revenue and developer contributions, followed by a lawsuit filed by the county to renegotiate the terms of the debt. The board of commissioners met and determined not to pay while the litigation was pending, ensuring a default.

The one non-housing default in 2018 is down from 18 the previous year. In 2017, there were two housing defaults, and there were none in 2018. For a detailed list of all defaults since 1986, see Tables 10 and 11.

The average one-year default rate for non-housing USPF from 1986 to 2018 is 0.029% (see Chart 4). The highest default rate, 0.084%, was from 2017, when there were 18 defaults. The average for the past five years is almost double the historical average, with 0.048% of non-housing ratings defaulting. Looking at just investment-grade ratings, the default rates are extremely small. The highest investment-grade default rate was in 2008, when there were three investment-grade defaults for a rate of 0.024%. Over the last five years, the average has been lower at 0.002% compared with the historical average of 0.004%.

Regarding speculative grade non-housing USPF defaults, the historical average is 2%, with the highest annual default rate of 6.54% from 14 defaults in 2013. The five year average of defaults is above the historical average, with 2.98% average annual default rates.

Among housing ratings, the historical average of defaults is 0.07%. The highest annual default rate was in 1991 with 18 defaults (a default rate of 0.63%). The average over the last five years is lower than the historical average, with an average annual default rate of 0.05%. The average annual default rate for investment-grade housing ratings is very low at 0.01%, with the highest annual investment-grade default rate in 2017 off of only two defaults (a rate of 0.09%). The population of speculative-grade housing debt is very small, with only 47 ratings at the start of 2018. The historical average of annual default rates for speculative-grade housing is 4.16%, with the highest default rate also from 1991, when 18 out of 80 ratings defaulted (a default rate of 22.5%).

Default Rates By Sector

Default rates vary among USPF sectors, though they are all generally low (see Table 3). The highest average annual default rate is in the health care sector (0.12%), followed by charter schools at 0.1% and transportation at 0.6%. Appropriations, higher education, and utilities all average 0.03% annual default rates, and tax secured has the lowest average annual default rate at 0.01%.

Table 4 summarizes the 2018 default rates and the mean and median default rates from 1986-2018 for each sector in USPF. Appropriations had the only default, which was in 2008, and a default rate of 0.02%. For more details on default rates by sector, see Tables 13-23.

Table 4

Transition Rates Remain Highly Stable

This study's transition analysis reveals the degree to which ratings change over time. In the transition tables, the intersection of a row and column of the same rating shows the percentage of that rating that started and ended the period at that rating. If ratings never changed, 100% would appear along the diagonal in the table. Movement toward the left shows the percentage of ratings that were raised, and movement to the right indicates ratings that were lowered, defaulted, or withdrawn. (This section provides an overview of rating transitions in USPF. For more details, see Tables 24-28, 41, and 43-44.)

Rating transition tables provide insight into the general rating trends of USPF and each sector by showing how different rating categories emerged at the end of a given timeframe. To determine the percentage of ratings that change in a given period, find the rating at the beginning of the period in the left column. The values in that row show the percentages of ratings as labeled in each column at the end of the period. For example, just 1.77% of 'AAA' ratings in USPF were lowered in 2018, and all of those that changed ended the year in the 'AA' category (see Table 5). At the other end of the scale, 'CCC'/'C' category ratings were raised to the 'BB' category 3.7% of the time, and another 3.7% were raised to the 'B' category. As discussed previously, 2018 is an outlier for defaults. There were no defaults among the lowest rated obligors. However, historically, 20.15% of 'CCC'/'C' rated debt defaults by the end of the year.

The total percentages of ratings that were raised or lowered into other rating categories can be computed by adding the values to the left or right, respectively, of the intersection of the row and column for each rating. For instance, 9.27% of 'BBB' ratings were raised (0.99% to the 'AA' category and 8.28% to the 'A' category), and 2.72% were lowered (2.19% to the 'BB' category, 0.26% to the 'B' category, and 0.26% to the 'CCC'/'C' category).

Table 5

Ratings generally moved higher in all rating categories except the two highest, where upgrades are difficult or impossible to achieve, and the lowest, for which there is usually a strong possibility of default (see Table 6). Due to the extremely low default activity in 2018, there were no downward rating movements among the lowest ratings.

Table 6

Stability has been a hallmark of USPF ratings since 1986--even more so when not including housing. At the 'AAA' level, for example, 97.1% of ratings remain at 'AAA' one year later, on average (see Tables 7 and 27). Some ratings in the 'AA' category shift down slightly in one year, but 93.7% remain in that category or move higher one year later. Meanwhile, 'A', 'BBB', and 'BB' category ratings tend to move up more often than down within one year. Significant proportions of defaults don't appear until the 'B' and lower rating categories.

Table 7

Examining rating transitions by modifier (the plus [+] or minus [-] after the rating) provides more evidence of upward rating movement (see Table 8). Ratings in the 'AA' category are not as likely to be raised to 'AAA' as they are to move to a lower rating category, but 'AA' ratings are raised more frequently than they are lowered (see Table 8). For all 'AA' ratings outside of housing, 3.65% are raised to 'AA+', while just 1.62% are lowered to 'AA-'.

Table 8

For housing bonds, movement is less favorable but still positive, with 2.2% of 'AA' ratings raised to 'AA+' and 1.95% lowered to 'AA-' (see Table 9). On average, 5.17% of 'A' ratings (excluding housing) are raised to 'A+' compared with 1.64% that are lowered to 'A-'. For 'A' housing ratings, an average of 3.03% are raised to 'A+', and 1.53% are lowered to 'A-'.

Table 9

The Number Of Defaults Remains Low While The Rating Universe Expands

Excluding housing, USPF default counts among bonds rated by S&P Global Ratings from 1986-2018 vary from none in nine of the years to a high of 18 in 2017. Of the 123 total nonhousing defaults in the study, 117 were rated speculative grade in the year in which the default occurred (see Tables 10 and 13). In the housing sector, there were 72 total defaults over the years covered by this study, ranging from zero in eight of the years to 18 in 1991 (see Table 23).

Although defaults are infrequent among rated USPF bonds, certain triggers can lead to multiple defaults. Typically, defaults are telegraphed by a series of downgrades before the defaults. Detroit; Stockton, Calif.; and Puerto Rico all experienced multiple downgrades before defaulting on a total of 44 issues from 2012-2018.

In rare cases, housing defaults may unexpectedly occur among higher-rated bonds due to idiosyncratic factors. This was the case in 1991, when 11 bonds with investment-grade ratings ('BBB-' or higher) at the beginning of the year defaulted due to the failure of a common investment agreement provider, Executive Life Insurance Co.

The municipal market has tended to be self-selecting; municipal issuers with lower credit quality have generally not requested ratings. Correspondingly, the municipalities rated by S&P Global Ratings generally are more creditworthy and less likely to default. Nonetheless, some securities issued by municipalities can still bear meaningful default risk.

While the initial credit quality of a municipal issue might be strong, that strength can decline over time. Of the 123 defaulted issues excluding housing, 78 (63% of the total) were initially rated 'BBB+' or lower. However, the proportion of nonhousing ratings at 'BBB+' or lower was 8.34% as of Dec. 31, 2018, down slightly from 8.46% at the end of 2017.

The Small Number Of Ratings Limits Analysis

As expected, default rates generally are higher for lower-rated obligations, and vice versa, but we observed a less-precise relationship when examining default rates by rating modifier. This is not surprising because the relatively small number of defaults and the small number of credits at some rating levels allow a single event to affect average default rates more than it would if there were a larger number of credits.

Examining default rates by rating category and by year further clouds the interpretation of default rates. No USPF defaults outside of housing have occurred at the 'AAA' level, and just one 'AAA' rated housing bond has ever defaulted. There are currently more than 1,100 'AAA' ratings in all of USPF. Significant default volatility becomes evident as we examine smaller subsets of the data.

Relative Ratings Performance

In addition to examining the absolute performance of ratings, we examine their relative performance. One technique we use to measure relative performance is the Gini coefficient. The Gini coefficient--a summary statistic of the Lorenz curve--is a measure of the relative ability of ratings to differentiate risk. We believe that comparing Lorenz curves and Gini coefficients is useful. Lorenz curves for USPF--both excluding and including housing--closely resemble the ideal Lorenz curves (see Chart 6 below and Charts 17-23). For more detail regarding the derivation and construction of Lorenz curves and Gini coefficients, see "Comparing transition rates with default rates" in Appendix I.

Chart 6 shows the one-year Lorenz curve for USPF excluding housing. The points along the Lorenz curve represent different ratings, starting with the 'CCC' and lower category on the far left. The first point indicates that the one-year transition to default was 66% among 'CCC' and lower ratings. As the ratings rise, 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 in the 'BBB' category. As the curve rises to investment-grade ratings, the vertical distance representing defaults changes little, while the proportion of ratings increases dramatically.

The Lorenz curve for USPF shows an extremely strong relationship between ratings at the beginning of a year and the default experience of the group of ratings. The left side of the Lorenz curve travels vertically toward 100% before slightly turning at 'CCC'/'C' at 66%, meaning that 66% of defaults within a one-year time horizon were for bonds that began the year in the 'CCC' rating category or lower. Meanwhile, 87.7% of defaults occurred among bonds rated speculative grade (below 'BBB-') at the start of the year (see Table 10). Just over 1% of defaults occurred among bonds rated 'A' or higher, even though they made up 46% of USPF ratings.

The table below further illustrates the information in the Lorenz graph. The lowest value in the left column shows the cumulative proportion of defaults at the 'CCC'/'C' category, or 66%, and the number to the right is the cumulative proportion of the rated universe, a minuscule 0.11%. This means that 66% of the defaults came from just 0.11% of the ratings, all of which were in the 'CCC' or lower category. Moving up the table, one can observe that once ratings reach 'A', representing 46% of all ratings, 99% of defaults have occurred at lower ratings.

Table 10

Looking at the Gini ratio for 2018 by itself, the ratio would be negative. This was due to there having been only one default in 2018, and that one happened to be investment grade. There are only a handful of instances where there was only one or no defaults in a given year.

Sector Overviews

The following section summarizes the rating distributions for all sectors within USPF. We will also cover the utility, housing, higher education, health care, transportation, and charter school sectors in separate reports later this year. The number of state GO ratings, at 50, is too small to provide enough data points for a transition study, while local government ratings are too diverse to cover as one rating type.

State and local government

State and local government ratings are combined into two rating types: tax-secured and appropriation. As mentioned earlier, data constraints prevent this study from distinguishing local government from state ratings in most instances. There are 17,058 tax-secured and appropriation ratings. A state or local government could have many ratings, which leads to the high number. We limited this discussion to the 50 GO ratings for states, and as a proxy for local government, we identified trends for the combined total of tax-secured and appropriation ratings. While state ratings are included in these groups, the overwhelming number of tax-secured and appropriation ratings are in local government.

State government.  The state sector has a very high ratings distribution, with a median rating of at least 'AA' since 1986 and at times as high as 'AA+'. Because the ratings in the sector are few in number and clustered in the two highest rating categories, we included the rating modifiers to better show the ratings distribution over time. In addition, because the small number of ratings precludes the use of transition tables, we instead used Chart 7 to show rating movement for each year.

Rating quality for the state sector reached peaks during and after the financial crisis: The number of 'AA' ratings reached 23 by the beginning of 2008, and the number of 'AAA' ratings increased to 16 at the start of 2016 from nine in 2008. However, in 2016, state bonds experienced a significant number of downgrades. Over 2016 and 2017, 15 GO ratings ended the year lower than they started. Meanwhile, just two bonds were rated higher at the end of the year than at the start. The only similar period with negative rating activity was during the early 1990s. However, in 2018, one state--Minnesota--was upgraded to 'AAA', the first such upgrade since Tennessee in 2016.

There were four rating actions on state GO bonds in 2018 (see Chart 8). The two upgrades were Minnesota and Michigan, upgraded to 'AAA' and 'AA' respectively, and the two downgrades were Kentucky and Connecticut, both to 'A' from 'A+'.

Local government.  Chart 9 shows all tax-secured and appropriation ratings, which are primarily from local government, although many are state government ratings. For the purpose of illustrating trends, combining the ratings is sufficient. Local government ratings make up approximately 75% of all USPF ratings. They are generally in the 'A' or 'AA' categories. The number of ratings held steady at about 4,000 in the early 1990s and then rose steadily to 10,000 in 2009. A spike in the number of ratings occurred in 2010 and continued to the current high point.

The percentage of high-investment-grade ratings ('AA-' or higher) has typically increased, but it fell in 11 of the years since 1986. Four of the down years were during or after recessions. In 2013, the proportion of high-investment-grade ratings declined to a four-year low of 42%, but that rate surged to a new high two years later because of revised criteria for local government ratings, which applied to more than 4,000 credits. S&P Global Ratings raised 41% of these ratings and lowered 4%. According to the most up-to-date numbers, through the start of 2019, 50% of these ratings were high-investment-grade. There were 1,176 upgrades in 2018 among state and local government ratings, almost double the 672 from 2017. Downgrades also fell slightly to 413 from 495. The ratio of upgrades to downgrades rose to 2.85 from 1.36 in 2017.

Utility

USPF utility ratings have consistently improved since 1986, with a peak of 51% in the high-investment-grade category in 2012 compared with 17% in 1986. The proportion of high-investment-grade utilities has been constant at 49% from the start of 2017 to the start of 2019. There were 85 upgrades in 2018, down from 110 in 2017, and downgrades fell slightly to 56 from 60. The ratio of upgrades to downgrades fell to 1.52 from 1.83 in 2017.

The number of ratings in the sector has grown as well. There were fewer than 800 USPF utility ratings until 2000, but total ratings surpassed 1,000 in 2002. By 2011, there were 1,667 ratings, more than double the 2000 total. The sector has added an average of more than 86 ratings each year since 2010, so by the start of 2019, there were 2,386 ratings. The number of ratings in the 'AA' category has grown the most over the past eight years; there were 982 'AA' category ratings at the start of 2019, up from 400 in 2008.

Housing

Despite high ratings, at times the USPF housing sector experiences more downgrades than upgrades. These rating changes can emerge from unexpected triggers. For example, criteria revisions on uninvested assets in 2010 led to a drop of about 500 'AAA' ratings, which were the benchmark at that time. The following year, S&P Global Ratings lowered the sovereign rating on the U.S. to 'AA+' from 'AAA', leading to the lowering of more than 1,300 'AAA' ratings. The number of 'AAA' housing ratings has climbed to 232 at the start of 2019 from 202 at the start of 2017.

There were 58 housing upgrades in 2018, up from 36 in 2017, and downgrades more than doubled to 117 from 49. The ratio of upgrades to downgrades fell to 0.5 from 0.73.

Higher education

The distribution of ratings for higher education and nonprofit organizations has remained stable since 1986. The most common rating category in the sector is 'A', while the percentage of ratings in the 'AA' category or higher has not fluctuated significantly from 30%. The percentage of 'BBB' category ratings has moved within a narrow band of 20%-25%.

Upgrades in higher education fell to nine from 19 in 2017, and downgrades fell to 31 from 41. The ratio of upgrades to downgrades fell to 0.29 from 0.46.

Health care

As of Jan. 1, 2019, S&P Global Ratings maintained 461 health care ratings in USPF, the fewest ratings in the sector since 1986, when the number of ratings was at its high of 858. The tally later declined to 559 in 1997, due in part to the widespread use of bond insurance, particularly in the health care sector, before the development of S&P underlying ratings (SPURs). The number of USPF health care ratings reached 704 in 2009. But since then, it has fallen by an average of nearly 25 each year, reflecting the industry consolidation and health care system development that have been in process for many years.

At least 26% of not-for-profit health care ratings have been 'BBB' or lower every year since 1986, and the proportion reached as high as 51% in 2004. Currently, 30% of ratings are below the 'A' category. There was a steady increase in 'BBB' or lower ratings from 1986-2006. The growth in 'BBB' category ratings in the late 1990s reflected the growth of SPURs as bond insurers increasingly targeted the 'BBB' category. The overall growth in lower-rated providers also reflected the growth of the senior living sector and the extended impact of the Federal Balanced Budget Act of 1997, which sharply lowered federal reimbursement for Medicare.

There were 36 health care upgrades in 2018, up slightly from the 31 in 2017, and there were 37 downgrades, down slightly from 40. The ratio of upgrades to downgrades rose to 0.97 from 0.78.

Transportation

Although ratings in the USPF transportation sector tend to be lower than in other areas of U.S. municipal finance, the sector is among the most stable. Transportation ratings are less likely than other USPF ratings to be raised, reflecting economic and competitive pressures.

Most USPF transportation ratings are in the 'A' category, while only 22% are in the 'AA' or 'AAA' categories, compared with about 46% of USPF ratings overall, excluding housing. Because they tend to be lower rated, transportation issues are more susceptible to default than other municipal bonds, although the number of defaults has been small over time.

There were 49 upgrades in 2018, up from 19 in 2017, and four downgrades, down from five. The ratio of upgrades to downgrades rose greatly to 12.25 from 3.8.

Charter schools

Charter schools are a fast-growing segment of USPF, with 282 ratings on Jan. 1, 2019, up from two in 2000. Charter schools are also the lowest-rated group within USPF, with no ratings ever above the 'BBB' category without support from another entity. Ratings have trended down since 2003, when all unsupported ratings were in the 'BBB' category.

By January 2019, just 44% of the ratings were in the 'BBB' category, while the majority were speculative grade. Defaults in the sector have been rare, though this might be a function of the shorter tenure of the bond ratings, the majority of which were assigned after 2012. Since 2013, four charter schools have defaulted, most recently at the start of 2017.

There were 14 upgrades in 2018, down from 18 in 2017, and 26 downgrades, down from 38. The ratio of upgrades to downgrades rose to 0.54 from 0.47.

Appendix I: Methodology And Definitions

This long-term U.S. public finance default and rating transition study uses an internal database of long-term local currency credit ratings. With the exception of housing bonds, the analysis is limited to unenhanced debt (debt obligations not supported by financial guarantees, structuring techniques, multiple-party features, or other external credit support). Within this study, the conclusions for USPF overall and the aggregated tables present data on unenhanced debt and exclude structured and housing debt. Because housing sector debt ratings reflect the credit quality of a mix of both public-sector entities and external credit enhancement providers, we present conclusions from this sector in separate tables.

S&P Global Fixed Income Research's 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, 1985. Therefore, each annual default study is self-contained and effectively supersedes all previous versions.

For this study, there are no changes in methodology or data from the "2017 Annual U.S. Public Finance Default Study And Rating Transitions," May 8, 2018. However, in this version, we have added explanations and descriptions to enhance readers' understanding of the main technical concepts.

Issuers included in this study

This study analyzes the rating histories of 37,889 USPF obligations that S&P Global Ratings rated as of Dec. 31, 1985, or that were first rated between that date and Dec. 31, 2018. These include obligations categorized as GOs, lease/appropriation/moral obligations, special tax (sales, gas, etc.), special districts, water and sewer revenue, public power, airports, ports, toll roads and bridges, parking, various types of bond pools, transit, public and private higher education, auxiliary higher education, independent schools, hospitals (stand-alone and systems), continuing care, and physicians' practices in the U.S. with long-term local currency ratings.

In terms of housing, the study includes 10,448 obligations tracked over the same period. These include obligations categorized as GOs, single-family whole loans, single-family multiple MBS, multifamily Fannie Mae, multifamily Freddie Mac, multifamily Ginnie Mae, affordable housing, Federal Housing Administration-insured loans, Section 8 housing, military housing, and capital fund securitizations.

Withdrawn ratings

S&P Global Ratings most often 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.

Annual versus quarterly rating movement

In this study, we quantify rating changes in two different ways. In most instances, the data on rating changes refer to ratings on the first day of the year and the difference, should any exist, from one year to another. For example, a rating of 'A' on Jan. 1 that changed to 'A+' in April and to 'AA-' in September would be counted as one rating change for that year.

However, in the sections of this report that discuss the causes of rating movement in 2017, each upgrade and downgrade is counted. In this way, if a rating changed twice in one year, the cause of each rating action is recorded. The individual rating changes in 2015, 2016, and 2017 in Table 1 and Chart 1 are for each upgrade and downgrade. This results in more instances of rating movement, though general trends are common for each method.

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.

S&P Global Ratings assigns a 'D' rating when it 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 believes 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 is 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 S&P Global Ratings revised the rating 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 S&P Global Ratings to subsequently withdraw the 'D' rating. For the purposes of this study, if an issuer defaults, we end the associated bond'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 this issuer (or, in the case of a selective default, just the bonds) into the database as a new set of bonds. The rating history on the bonds after the default event is included in all calculations as entirely separate from the experience leading up to the earlier default.

Calculations

Static pool methodology.  S&P Global Fixed Income Research conducts its default studies on the basis of groupings called static pools. For the purposes of this study, we form static pools by grouping bonds 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 bonds included in the study are assigned to one or more static pools. When an issuer defaults, we assign that default back to all of the static pools to which the issuer belonged.

S&P Global Fixed Income Research uses 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 multiple-period time horizons. Some methods for calculating default and rating transition rates might charge defaults only against 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 new study revises results back to the same starting date of Jan. 1, 1986, so as to avoid continuity problems.

Ratings that have been withdrawn--that is, revised to 'NR' (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 a subsequent default occurs, we will include that series 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 1986 static pool consists of all debt rated as of 12:01 a.m. on Jan. 1, 1986. Adding debt first rated in 1986 to the outstanding debt of the 1986 static pool forms the 1987 static pool. All rating changes that took place as of 12:01 a.m. on Jan. 1, 1987, are reflected in the newly formed 1987 static pool. We used the same method to form static pools for 1988-2017.

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 calculation.  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 32 years the study covers.

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 a more useful measure 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 (see Tables 29-38). We calculated conditional default rates by dividing the number of issuers in a static pool that default 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, as shown in Table 37, the weighted-average first-year default rate for entities rated in the 'B' category for all 32 pools was 2.15%, meaning that an average of 97.85% made payments in accordance with their terms for the first year. Similarly, the second- and third-year conditional marginal averages were 2.61% for the first 30 pools (97.39% of those parity obligations that did not default in the first year did not default in the second year) and 2.46% for the first 29 pools (97.54% of those entities that did not default by the second year did not default in the third year either), respectively. Multiplying 97.85% by 97.39% results in a 95.30% nondefault rate at the end of the second year, or a two-year cumulative average default rate of 4.70%. Multiplying 95.30% by 97.59% results in a 92.96% nondefault rate at the end of the third year, or a three-year 7.04% cumulative average default rate. (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.)

Transition analysis

Transition rates compare issuer ratings at the beginning of a period with 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 continually 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 'NR' in the 1991 transition matrix. Similarly, if it experienced default in the middle of 1991, it would be included in the column representing transitions to 'D' in the 1991 one-year transition matrix.

All 1986 static pool members still rated on Jan. 1, 2018, had 32 one-year transitions, while debt first rated on Jan. 1, 2017, had only one. Tables 43 and 44 display the summaries of one-year transitions for USPF obligations not including housing and for only housing, respectively. 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 'D', plus an entry for 'NR'.

Multiyear transitions.  Multiyear transitions were also calculated for periods of two up 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, three-year transition matrices were the result of comparing ratings at the beginning of the years 1986-2015 with the ratings at the end of the years 1988-2017. 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.

Initial-to-last transitions 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 rating "states" inhabited by USPF ratings during their lifetimes.

Gini coefficient calculation

The Gini coefficient, developed by Corrado Gini, is a summary statistic of the Lorenz curve, which shows visually 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 the Lorenz curve 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 the 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 Charts 6 and 16-23, and is derived by dividing area B by the total of 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, such as one year or three years; for a single period, such as the year ended 2016; or by aggregating a series of one-year periods. To calculate the one-year Gini for 2016, 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 level and the proportion of defaulters at each rating, based on the ratings at the beginning of the period. The aggregate Gini for 1986-2017 combines data for each of the 32 one-year periods. The same parity obligation may be counted multiple times in the aggregate Gini coefficient.

Time sample

This study limits the reporting of default rates to the 15-year time horizon. However, the data were gathered for 32 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' credit 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.

Appendix II: Additional Exhibits

Default summaries

Table 11

Table 12

Table 13

Table 14

Table 15

Table 16

Table 17

Table 18

Table 19

Table 20

Table 21

Table 22

Table 23

Transition tables

Table 24

Table 25

Table 26

Table 27

Table 28

Lorenz curves

Cumulative default rates

Table 29  |  View Expanded Table

Table 30  |  View Expanded Table

Table 31  |  View Expanded Table

Table 32  |  View Expanded Table

Static pool default rates

Table 33

Table 34

Table 35

Table 36

Table 37

Table 38

Recent statistics

Table 39

Table 40

Table 41

Table 42

Table 43

Table 44

Related Criteria And Research

Related Criteria

• Priority-Lien Tax Revenue Debt, Oct. 22, 2018
• U.S. Municipal Retail Electric And Gas Utilities: Methodology And Assumptions, Sept. 27, 2018
• Special Assessment Debt, April 2, 2018
• U.S. And Canadian Not-For-Profit Acute Care Health Care Organizations, March 19, 2018
• U.S. And Canadian Not-For-Profit Transportation Infrastructure Enterprises: Methodologies And Assumptions, March 12, 2018
• Solid Waste System Financings, Jan. 29, 2018
• Issue Credit Ratings Linked To U.S. Public Finance Obligors' Creditworthiness, Jan. 22, 2018

Related Research

• Criteria Revisions Led To Favorable Rating Trends In U.S. Public Finance In 2018, April 4, 2019
• U.S. GDP Growth Hits A Soft Patch--Not Quicksand, April 4, 2019
• U.S. Municipal Issuers Keep Their Footing On An Uneven Road Into 2019, Jan. 10, 2019
• 2017 Annual U.S. Public Finance Not-For-Profit Health Care Default Study And Rating Transitions, Dec. 12, 2018
• 2017 U.S. Public Finance Utility Default Study And Rating Transitions, Dec. 12, 2018
• 2017 U.S. Public Finance Transportation Default Study And Rating Transitions, Nov. 19, 2018
• 2017 Annual U.S. Public Finance Housing Default Study And Rating Transitions, Sept. 27, 2018
• 2017 Annual U.S. Public Finance Higher Education And Nonprofit Organization Default Study And Rating Transitions, July 26, 2018
• 2017 Annual U.S. Public Finance Charter School Default Study And Rating Transitions, June 12, 2018
• 2017 Annual U.S. Public Finance Default Study And Rating Transitions, May 8, 2018