Taiwan Ratings' 2006 Default And Rating Transition Study

2007/03/21

Analysts:

Chun Huang, CFA Serene Hsieh, CPA Susan Chu Daisuke Fukutomi

Coverage
Taiwan Ratings Corp.'s default and rating transition study examines the track record of ratings rated by Taiwan Ratings since it first assigned credit ratings in 1998. The study shows that the rating performance has been broadly similar to Standard & Poor's Ratings Services' global experience, however, rating movements have been more volatile, especially at lower rating levels. The study primarily measures rating movements over time and provides a quantitative measure of rating performance, implying Taiwan Ratings' historical ability to predict the likelihood of default.

This report covers 208 solicited issuer credit ratings rated by Taiwan Ratings between 1998 and 2006, inclusive. The study analyzes the rating performance of Taiwan-based obligors, which include industrials, utilities, insurance companies, financial holding companies, banks, securities firms, and other financial institutions. The study includes nonconfidentially and confidentially rated entities, as well as those whose ratings were withdrawn after being assigned.

Key Findings

• The credit quality of rated obligors in Taiwan was relatively stable in 2006, with M&A as the major driver of change. M&A are expected to continue to drive credit profile changes in 2007, but consolidation activity, especially for financial services companies, is likely to be muted. The credit quality of firms not directly involved in M&A is expected to remain relatively stable.
• Although the default experience in Taiwan's rated pool is likely to develop over a few more business cycles, the study has supported a positive correlation between rating levels and default probability. A higher rating implies a lower historic default rate and vice versa.
• Taiwan Ratings' rating transition over 1998-2006 broadly mirrors Standard & Poor's global study over 1981-2006, which reveals that higher rated issuers exhibit higher stability than their lower rated counterparts. However, Taiwan Ratings' rating transition tends to be faster and more volatile, especially in the lower rating categories because of the evolving credit profiles of Taiwan issuers in recent years and the statistical limitations of a smaller sample and shorter rating history.

Credit Trends In Taiwan

The credit quality of rated obligors was relatively stable in 2006, reflecting the island's generally stable economic conditions. M&A or alliances accounted for slightly over half of the rating changes on financial services companies, which made up about 67% of the rated sample.

The relatively stable conditions in 2006 represent a slight change from the rapid improvement in credit quality seen in 2002-2005. During the 2002-2005 period, many financial institutions that were severely hit in the previous credit cycle improved their risk management mechanisms and engaged in consolidation. Many formed or joined financial holding company groups. The credit quality of weaker financial institutions accordingly rose on the back of the reallocation of resources among group members.

As a result of this factor, as well as the growing number of new credits rated 'twA' or above (see table 1), the majority of issuers in Taiwan continued to be rated 'twA' or above at the end of 2006. The share of 'twA' or above ratings inched up to about 81%, compared with a significantly lower 49% at the end of 2003 (see chart 1).

Overall credit conditions for both industrial companies and financial institutions are expected to be stable in 2007. Taiwan Ratings expects rating movements to be largely driven by M&A and strategic alliances in the financial services industry. However, the pace is expected to be slower than in previous years.

Ratings Are A Good Indicator Of Default Probability

Despite a smaller sample and shorter time frame compared with Standard & Poor's Global Corporate Default Study of over 12,000 obligors and stretching over a period of more than 20 years, this study does support a positive correlation between rating levels and default probability. Higher ratings have a lower probability of default (see table 2).

The only issuer that defaulted in 2006 was Taitung Business Bank. The bank was placed under regulatory supervision after failing to recapitalize. For the purpose of this study, financial institutions that are placed under regulatory supervision are classified as having defaulted. Placing a financial institution under regulatory supervision, or 'twR', does not necessarily indicate a default event, but the regulator may have the power to favor one class of obligations over others or pay some obligations and not others.

This brings the total number of defaults in Taiwan's entire nine-year pool to four, consisting of three financial institutions and one industrial company. All four defaulters were rated at 'twBBB' or below.

The three stressed financial institutions were eventually placed under regulatory supervision, which accordingly resulted in a high level of protection for most of their financial obligations. This highlights the regulator's strong commitment to maintaining a stable domestic financial system. The default ratio in 2007 will be impacted by the placement of two further banks--Enterprise Bank of Hualien and Chinese Bank--under regulatory supervision at the beginning of the year. Prior to regulatory intervention, these two banks were rated at 'twBB' and 'twBBB-', respectively. The outlook on both entities was negative.

In light of Taiwan's increasing deregulation and market liberalization, Taiwan Ratings expects to see the development of more mature financial and capital markets. The transition to greater market discipline may result in defaults by fundamentally weaker issuers. Defaults are a by-product of market efficiency, as seen in the developed economies of North America and Europe.

The study's findings are partly qualified by Taiwan Ratings' issuer rating pool being statistically smaller, set over a shorter period, and relatively undiversified. For a small sample, a single default may significantly alter the results, as in the case of the 'twB' category in table 2. Furthermore, about half of the issuer ratings have been assigned by Taiwan Ratings for less than five years, which is the average length of time for an issuer to default, according to Standard & Poor's global study. A large part of the rating pool is not yet well seasoned. Rating appropriateness is best determined after a few business cycles and set on a sizeable as well as diversified pool of issuer ratings.

The scarcity of defaults can be attributed to credit ratings in Taiwan still being in a development phase and the preponderance of financial services entities in the ratings pool. In Standard & Poor's experience, issuers with stronger credit profiles tend to solicit ratings in the early years when credit ratings are introduced into a market. This tendency reduces the likelihood of default occurring during the early years of a study. Furthermore, about 67% of Taiwan Ratings' rating pool consists of financial services entities (see table 3), which are overseen by the government. Regulatory oversight tends to ensure that, barring major systemic shocks, financial institutions exhibit a lower risk of default than would otherwise be the case. The default rate of Standard & Poor's global rated pool of financial institutions and insurance companies not surprisingly is lower on average than for industrials (see table 4).

Taiwan's Rating Transitions Largely Mirror Global And Regional Rating Behavior
The results of Taiwan Ratings' transition study largely mirror Standard & Poor's global and regional observations that higher rated issuers tend to exhibit less ratings volatility than their lower rated counterparts. For instance, table 5 shows that the probability that a Taiwan issuer rated 'twAA' at the beginning of a year will still be rated 'twAA' at the end of the year is 93.3%, whereas the probability that an issuer rated 'twBB' at the beginning of a year will be rated 'twBB' at the end of the year is only 61.6%. The probability that a global issuer rated 'AA' will retain this rating after one year is 87.3%, whereas the probability that a global issuer rated 'BB' will retain this rating after one year is only 75.7%. The same relationship holds true among the Asia ex-Japan rated issuers.

Taiwan Ratings' higher rated categories appear extremely stable, however, its lower rated categories registered faster rating movements than Standard & Poor's global and regional rated pools, as a result of its issuer rating pool being statistically small and set over a short period, as well as the volatility inherent in smaller and/or weaker financial institutions. Caution must be used in interpreting the higher stability rates associated with the 'twCCC/CC' rating category relative to the 'twB' rating category in light of the extremely small sample size. Also, the relatively high number of withdrawals diluted the rating stability at the 'twB' rating category.

Taiwan Ratings' study has a smaller number of issuers (208 entities as of the end of 2006) and a shorter study period (1999-2006) in comparison with the global (12,293 entities; 1981-2006) and the Asia ex-Japan (582 entities; 1988-2006) studies. Statistically, a smaller number of observations and a shorter time period may potentially contribute to wider deviations of estimation from the findings of the larger number of observations. The higher volatility in Taiwan Ratings' lower rating categories also reflects the characteristics of domestic ratings, which have finer gradations and thus make Taiwan Ratings' ratings more sensitive to changes in the credit profiles of rated issuers.

Long-term rating transition relationships (see table 6) remain true even when entities that are not rated at some point during their rating history are removed from consideration.

Global observations reveal that lower rated credits generally move toward lower categories, however, lower rated credits (or even most categories) in Taiwan Ratings' pool tended to move up rather than down. This is mainly a result of a large-scale rating adjustment in December 2004, which was made to reflect fundamental improvements beyond Taiwan Ratings' expectations at smaller/weaker financial institutions.

The ratings in Taiwan ratings' pool exhibited higher volatility than the ratings in Standard & Poor's global pool. The frequency of annual rating changes in Taiwan ranged between 12% and 72% in 1999-2006 and averaged 36% in the period (see table 7). This contrasts with the annual rating changes in Standard & Poor's global pool, which ranged between 24% and 34% in 1997-2006 and averaged 28% in the period (see table 8). Standard & Poor's Asia ex-Japan study also highlighted higher rating volatility, which was offset to some extent by its larger pool and longer length of observation (see table 9). The high ratio of annual rating changes in Taiwan Ratings' pool also reflects the large-scale rating adjustment conducted in December 2004, which also resulted in a lower downgrade-to-upgrade ratio for Taiwan Ratings' pool that averaged at 0.3 times in 1999-2006 in comparison with 1.6 times for Standard & Poor's regional and global pools.

Default Implications Of Ratings Transition

As the rating performance of Taiwan Ratings' scale continues to develop, including the size of the rated pool and the length of rating history, the default and rating transition is likely to move toward Standard & Poor's global study after its issuer pool undergoes testing through future business cycles. However, there remains a major difference in the implicit default risk between Standard & Poor's global scale and Taiwan Ratings' scale, which is primarily positioned as the national scale and does not reflect sovereign risk (Taiwan is rated AA-/Negative/A-1+ by Standard & Poor's Rating Services).

Based on Standard & Poor's historic observations, cumulative default rates may be projected into the future, based on the assumption that the rating transition rates are stable. The one-year default rate column in table 10 is equivalent to the D (default) column in table 11, while the three-year default rate column in table 10 is about the level of the D (default) column in table 11. The slight difference in results between the two tables mainly stems from the different static pools used to calculate transition to default and cumulative average default rates. Cumulative average default rates are the summary of all static pools, while the number of pools used in the average transition rate is limited by the transition's time horizon.

Appendix: Default Methodology And Definitions

This long-term corporate default and rating transition study uses the CreditPro? 7.51 database of long-term issuer credit ratings. An issuer credit rating reflects Taiwan Ratings' opinion of a company's overall capacity to pay its obligations (that is, its fundamental creditworthiness). This opinion focuses on the obligor's ability and willingness to meet its financial commitments on a timely basis, and it generally indicates the likelihood of default regarding all financial obligations of the firm. It is not necessary for a company to have rated debt in order to be assigned an issuer credit rating.

Although a company's senior secured debt (particularly debt with strong covenants) may occasionally be rated higher than the issuer credit rating on the company, specific issues are typically rated as high as or lower than the issuer rating, depending on their relative priority within the company's debt structure. If they are lower rated entities, issuer credit ratings are generally two notches higher than subordinated debt ratings. Otherwise, they are generally one notch higher. Therefore, though a 'twBB+' issuer credit rating is generally paired with a 'twBB-' subordinated debt rating, a 'twAA' issuer credit rating usually corresponds to a 'twAA-' subordinated rating.

Standard & Poor's ongoing enhancement of the CreditPro? database used to generate this study may 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, 1998. Therefore, each annual default study is self-contained and effectively supersedes all previous versions.

Issuers included in this study

The study analyzed the rating histories of 208 companies that were rated by Taiwan Ratings as of Dec. 31, 1998, or that were first rated between that date and Dec. 31, 2006. These companies include industrials, utilities, insurance companies, financial holding companies, banks, securities firms, and other financial institutions in Taiwan with long-term credit ratings. The global data presented in this report refers to Standard & Poor's ratings histories of all 12,293 long-term rated issuers from Dec. 31, 1980 to Dec. 31, 2006. And the Asia ex-Japan data refers to Standard & Poor's ratings histories of 582 long-term rated issuers from Dec. 31, 1992 to Dec. 31, 2006. The study includes nonconfidentially and confidentially rated entities as well as those whose ratings were withdrawn after initial assignment. The analysis excludes public information (pi) ratings and ratings based on the guarantee of another company. Structured finance vehicles, public-sector issuers, and sovereign issuers are the subject of separate default and transition studies and are also excluded from this study.

Subsidiaries whose debt is fully guaranteed by a parent or whose default risk is considered identical to that of their parents were excluded. The latter are companies whose obligations are not legally guaranteed by a parent but whose operating or financing activities are so inextricably entwined with those of the parent that it would be impossible to imagine the default of one and not the other. At times, however, some of these subsidiaries might not yet have been covered by a parent's guarantee, or the relationship that combines the default risk of parent and child might have come to an end, or might not have begun. Such subsidiaries were included for the period during which they carried a distinct and separate risk of default.

Definition of default

A default is recorded on the first occurrence of a payment default on any financial obligation, rated or unrated, other than a financial obligation subject to a bona fide commercial dispute; an exception occurs when an interest payment missed on the due date is made within the grace period. Preferred stock is not considered a financial obligation; thus, a missed preferred stock dividend is not normally equated with default. Distressed exchanges, on the other hand, are considered defaults whenever the debt holders are coerced into accepting substitute instruments with lower coupons, longer maturities, or any other diminished financial terms.

Issue ratings are usually lowered to 'D' following a company's default on the corresponding obligation. In addition, 'SD' is used whenever Taiwan Ratings believes that an obligor that has selectively defaulted on a specific issue or class of obligations will continue to meet its payment obligations on other issues or classes of obligations in a timely matter. 'twR' indicates that an obligor is under regulatory supervision owing to its financial condition. This does not necessarily indicate a default event, but the regulator may have the power to favor one class of obligations over others or pay some obligations and not others. 'D', 'SD', and 'twR' issuer ratings are deemed defaults for purposes of this study. A default is assumed to take place on the earliest of: the date Taiwan Ratings revised the ratings to D', 'SD', or 'twR'; the date a debt payment was missed; the date a distressed exchange offer was announced; or the date the debtor filed or was forced into bankruptcy.

Static pool methodology

Taiwan Ratings conducts its default studies on the basis of groupings called static pools. Static pools are formed by grouping issuers by rating category at the beginning of each year covered by the study. Each static pool is followed from that point forward. All companies included in the study are assigned to one or more static pools. When an issuer defaults, that default is assigned back to all of the static pools to which the issuer belonged.

Taiwan Ratings uses the static pool methodology to avoid certain pitfalls in estimating default rates, to ensure that default rates account for rating migration, and to allow for default rates to be calculated across multi-period time horizons. Some methods for calculating default and rating transition rates might charge defaults against only the initial rating on the issuer--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--this method may yield comparatively low default rates during periods of high rating activity, as they ignore prior years' default activity.

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 companies 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 update revises results back to the same starting date of Dec. 31, 1998, so as to avoid continuity problems.

Entities that have had ratings withdrawn--that is, revised to N.R.--are surveilled with the aim of capturing a potential default. These companies, as well as those that have defaulted, are excluded from subsequent static pools.

For instance, the 1999 static pool consists of all companies rated as of 12:01 a.m. Jan. 1, 1999. Adding those companies first rated in 1999 to the surviving members of the 1999 static pool forms the 2000 static pool. All rating changes that took place are reflected in the newly formed 2000 static pool. This same method was used to form static pools for 2001 through 2006.

Consider the following example: An issuer is originally rated 'twBB' in mid-1998 and is downgraded to 'twB' in 2000. This is followed by a rating withdrawal (N.R.) in 2002 and a default ('D') in 2005. This hypothetical company would be included in the 1999 and 2000 pools with the 'twBB' rating, which it was rated at at the beginning of those years; likewise, it would be included in the 2001 and 2002 pools with the 'twB' rating. It would not be part of the 1998 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 2002 because the rating had been withdrawn by then. Yet each of the four pools in which this company was included (1999-2002) would record its 2005 default at the appropriate time horizon.

Ratings are withdrawn when an entity's entire debt is paid off or when the program or programs rated are terminated and the relevant debt extinguished. They may also occur as a result of M&A. Others are withdrawn because of a lack of cooperation, particularly when a company is experiencing financial difficulties and refuses to provide all the information needed to continue surveillance on the ratings.

Default rate calculation

Annual default rates were calculated for each static pool: first in units, and later as percentages with respect to the number of issuers in each rating category. Finally, these percentages were combined to obtain cumulative default rates for the nine years covered by the study (see table 2).

Issuer-weighted default rates

Averages that appear in this study are calculated 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 statistical performance of ratings.

Many practitioners use statistics from this default study and CreditPro? to estimate the probability of default and the probability of rating transition. It is important to note that Taiwan Ratings' ratings do not imply a specific probability of default; however, Taiwan Ratings' historical default rates are frequently used to estimate these characteristics. When estimating the probability of default, issuer-weighted statistics have less variance than dollar-weighted statistics and are therefore preferable.

Cumulative average default rate calculation

Cumulative default rates that average the experience of all static pools were derived by calculating marginal default rates, conditional on survival (survivors being nondefaulters) for each possible time horizon and for each static pool, weight averaging the conditional marginal default rates, and accumulating the average conditional marginal default rates (see table 10). Conditional default rates are calculated by dividing the number of issuers in a static pool that default at a specific time horizon by the number of issuers that survived (did not default) to that point in time. Weights are based on the number of issuers 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 'B' rated companies for all 26 pools was 4.99%, meaning that an average of 95.01% of obligors survived for a period of one year or longer. Similarly, the second- and third-year conditional marginal averages were 6.24% for the first 25 pools (93.76% of those companies that did not default in the first year survived the second year) and 5.59% for the first 24 pools (94.41% of those companies that did not default by the second year survived the third year). Multiplying 95.01% by 93.76% results in a survival rate of 89.08% to the end of the second year, or a two-year cumulative average default rate of 10.92%. Multiplying 89.08% by 94.41% results in an 84.10% survival rate to the end of the third year, or a three-year cumulative average default rate of 15.90%.

N.R.-removed default rates

A slightly different method is used to obtain N.R.-removed default rates. These are obtained by omitting those issuers that had ratings withdrawn. The N.R.-removal replicates the default rate that a buy-and-hold portfolio would experience if the portfolio were reallocated among the non-N.R. members of the portfolio each time the rating on a company is withdrawn. The numerators and denominators of the default rates decrease gradually as companies merge, leave the public fixed-income markets, or request the ratings on them be withdrawn. These rates are, in general, greater than those of the conventional default rate calculation, but the overall behavior of the default rates is quite similar. That is, the higher the rating, the lower the default likelihood.

The N.R.-removed default rate calculation may unduly inflate default rates as shown by the following example. Suppose that there were 10 issuers in a static pool, nine of which became N.R. over a 10-year time span for benign reasons such as mergers or retiring of debt. If, in the 10th year, the one company that was still rated was to default, the N.R.-adjusted default rate would be 100% for the 10-year time horizon. In order for the conventional default rate to reach 100%, all nine of the N.R. issuers would need to default after the ratings on them were withdrawn. Although the N.R.-removed default rate likely overstates the risk of default, it is included in this study because some investors use it as a conservative estimate of average default rates.

Time sample

This update limits the reporting of default rates to the selected time horizon; however, the data was gathered for nine years and all calculations are based on the rating experience of that period. The maturities of most obligations are much shorter than the selected time horizon. 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 increases its effect on averages.

Default patterns share broad similarities across all static pools, suggesting that Standard & Poor's rating standards have been consistent over time. Adverse business 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.

Transition analysis

Transition rates compare issuer ratings at the beginning of a time period with ratings at the end of the period. To compute one-year rating transition rates by rating category, the rating on each entity at the end of a particular year was compared with the rating at the beginning of the same year. An issuer that remained rated for more than one year was counted as many times as the number of years it was rated. For instance, an issuer continually rated from the middle of 1998 to the middle of 2003 would appear in the four consecutive one-year transition matrices from 1999 to 2002. All 1999 static pool members still rated on Dec. 31, 2005, had 7 one-year transitions, while companies first rated between Jan. 1, 2006, and Dec. 31, 2006 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 left-most column, there are nine ratios listed in the rows, corresponding to the ratings from 'twAAA' to 'D,' plus an entry for N.R.

Practical application of transition rates

Rating transition rates are useful to investors and credit professionals for whom rating stability is important. For instance, investors restricted by law or inclination to invest in top-grade bonds would want to assess the likelihood that Taiwan Ratings' analysts will continue to assign top ratings to their investments. Conversely, investors buying high-yield bonds in hopes of profiting from a rating upgrade would be able to gauge that expectation realistically.

The credit community might also use rating transition information, in part, to determine maturity exposure limits or to measure credit risk in the context of the value-at-risk models. Assuming that the rating transition rates are stable and follow a first-order Markov process, cumulative default rates could be projected for any number of years into the future. Rating transition matrices could also be constructed to produce stressed default rates. Such matrices are often used in the area of credit risk measurement. In addition, multiyear transition matrices are valuable tools that can be used to forecast future rating distributions and may be better suited for certain applications than are one-year transition matrices.

N.R.-removed transition rates

The difference between a N.R.-removed table and one that is not is that the former is based on pools that have been gradually pared down by dropping those obligors whose ratings have been withdrawn (set to 'N.R.'). The number of withdrawn ratings grows particularly large in the case of speculative-grade ratings categories after just a few years. Little is known about 'N.R.' obligors except that there is no public record of a default. Indeed, default might be unlikely for those obligors whose debt has been extinguished.

Comparing transition rates with default rates

Rating transition rates may be compared with the marginal and cumulative default rates described in the previous section. For example, note that the one-year default rate column of table 2 is equivalent to column 'D' of the average one-year transition matrix found in table 5. Cumulative average default rates are the summary of all static pools from 1998 through 2006 while the number of pools used in the average transition rate is limited by the transition's time horizon.