In this paper we examine how well CreditWatch is used by credit rating agencies to balance two conflicting goals: rating timeliness and rating stability. Examining equity market reactions around CreditWatch events in 2002-2005, we find evidence that while CreditWatch has improved rating timeliness, its intended purpose has not been completely achieved. Equity prices start to change days before companies are listed on CreditWatch and abnormal equity returns of firms prior to being listed on CreditWatch are effective predictors of the ultimate change in ratings. The findings in the study suggest that in the pursuit of rating stability, rating agencies may have sacrificed rating timeliness.
One of the most surprising events during the 2007-2008 financial crisis was the filing of Chapter 11 bankruptcy by Lehman Brothers, an investment bank with a 158-year history, on Monday, September 15th, 2008. The news of Lehman’s collapse shook financial markets world-wide, including a drop in the Dow Jones Industrial Average of more than 500 points. The collapse of Lehman took many investors by surprise because as recently as September 12th, the previous Friday, Lehman Brothers’ bonds were rated “A”, an investment grade.
Though unique in its impact on global markets, the precipitous fall of Lehman from investment grade status directly to bankruptcy is not unprecedented. Several other high-profile bankrupt companies, including Enron, also maintained investment-grade ratings until just days before the bankruptcy was announced. Are rating agencies too slow in adjusting ratings?
Indeed, one major criticism of credit rating agencies is the lack of timeliness in making rating changes.1 Studies in the finance literature have shown that credit rating changes are anticipated by the equity market (Norden and Weber [
The challenges faced by rating agencies in conveying timely default information are multidimensional. In addition to the criticism for not being able to obtain timely information, for not using the best rating methodologies, and for not expending maximum effort (Goldstein, Kaminsky, and Reinhart [
Another reason for the slow reaction may be related to an argument put forth by rating agencies that markets expect stable ratings. Ratings are often used by investors and regulators as guidance for portfolio governance.3 Frequent changes in ratings may force portfolio managers to trade securities more frequently, thereby increasing transaction costs. Frequent rating changes may also force portfolio managers to sell securities at lower prices (when they are downgraded) and to purchase at higher prices (when they are upgraded) more frequently and thus suffering more losses. Consequently, rating agencies tend to meet the market expectations by making rating changes only when a reversal in ratings in the near future is unlikely (Cantor [
Rating agencies could have made rating changes as soon as they observe changes in bond issuers’ credit risks, regardless of whether the changes in the credit risks are long-term or temporary. By doing so, however, rating agencies may have made rating changes too hastily and therefore have to make a reversal in rating changes to correct the previous overreactions. To avoid such a mistake, rating agencies would rather make gradual, step- wise, and sequential small changes in ratings, instead of making sudden, abrupt, and large rating changes and then making subsequent reverse rating changes to revise the previous changes. In balancing the two conflicting tasks, rating agencies choose the former over the latter. In other words, they would rather appear to be “slow” than to be “wrong”.
Some studies (e.g. Loffler [
Credit rating agencies have not been insensitive to the criticism. One specific action by Standard and Poor’s (S&P’s) was the creation in a service known as CreditWatch, which was first offered in November 1981. CreditWatch provides information to investors about potential changes in default risk prior to an actual change in rating. One major purpose of CreditWatch is to ease the tension between the market expectation of rating stability and the market demand for rating timeliness (Altman and Rijken [
When a company is listed on CreditWatch, it is typically listed with either a positive or a negative potential.5 In a listing with positive potential, the rating of the company will usually be eventually upgraded or affirmed (i.e. unchanged), and the rating is rarely downgraded. Similarly, in a listing with negative potential, the rating of the company will usually be eventually downgraded or affirmed, and the rating is rarely upgraded. Once the rating is changed or affirmed, the listed company is delisted (removed) from the CreditWatch list. Unlike credit rating changes in which rating agencies convey the default risk to the market in one action (i.e. the rating change), the publication of CreditWatch conveys the default information to the market through two sequential actions—first through listing, and then through delisting. The listing conveys information about the direction of the rating change, and the delisting reveals the magnitude of the actual rating change. Although listing on CreditWatch can lead to a bond rating change, only a small fraction of all actual rating changes are preceded by a listing on CreditWatch.
In this study, we examine the response of equity prices of firms listed and delisted from CreditWatch to determine if CreditWatch improves the timeliness of rating changes. Despite its intended purpose of informing investors of a potential rating change in a timely fashion, we find that CreditWatch does not completely achieve this goal. We report three major findings in support of this conclusion. First, we find equity markets experience substantial positive (negative) reaction to the listing of companies with positive (negative) potential on CreditWatch prior to the actual date of listing. Furthermore, the magnitude of pre-listing abnormal returns is closely correlated to the magnitude of rating changes the follow the listing date. These results suggest that rating agencies are still slow in conveying information about the change in default risk to the market. Second, equity markets exhibit significant abnormal returns on the CreditWatch listing day, but exhibit little reaction to the removal of a company from CreditWatch. The results indicate that the impact of rating agency’s announcement has moved forward from rating changes to CreditWatch listings that precede the rating changes. The results suggest that CreditWatch listing improved the timeliness of rating functions. Collectively, the findings in the study suggest that while CreditWatch mitigates the problem of the lack of timeliness by bringing forward its impact from the subsequent rating change to the CreditWatch listing, it does not completely resolve the timeliness issue caused by rating stability policy. Third, to further confirm the conclusion, we also use the abnormal returns surrounding the CreditWatch listing date to predict the magnitude of subsequent rating changes and find that the abnormal returns surrounding the CreditWatch listing date indeed serve a good predictor for the rating changes subsequent to the CreditWatch listing. Collectively, our findings suggest that rating agencies may sacrifice timeliness for the sake of stability and that even CreditWatch, which is designed to mitigate the disadvantage caused by stable rating policies, is not a completely effective instrument.
The remainder of the paper is organized as follows. Section 2 discusses the data and the methodology. Section 3 presents the empirical analyses and results, and Section 4 concludes.
Our sample construction begins with firms placed on CreditWatch between January 2002 and December 2005. We hand-collect the following data for firms at the time of listing: 1) company name, 2) listing date, 3) existing S&P senior debt rating, and 4) listing potential. From this group, we exclude all firms with insufficient data from the Center in Research and Security Prices (CRSP) to compute abnormal returns surrounding the listing date. We also exclude firms for which definitive information about the action taken by S&P regarding the firms’ rating upon delisting is unavailable. The final sample consists of 604 observations, with 101 listed with “positive” potential, 503 listed with “negative” potential.6 The sample composition is consistent with similar studies in the literature (e.g. Dichev and Piotroski [
Following Morgan [
Financial data is obtained from Compustat for the year preceding the date of listing. The sample sizes are reduced slightly because of missing data in Compustat. Statistically significant differences exist regarding the size of the companies (measured by total assets) and the cash ratio. Specifically, firms listed with negative potential are larger in size and have lower cash ratios compared to firms listed with positive potential. Firms listed with positive potential tend to remain on the CreditWatch list longer than those listed with negative potential, but the difference is not statistically significant.
To capture the reaction of the equity market, we employ an event study methodology by computing daily abnormal returns (AR) and cumulative abnormal returns (CAR) of the companies in event windows surrounding the listing and delisting dates. For robustness, we consider three estimation procedures—the market model, market adjusted return model, and the Fama-French (1992) model—to calculate the abnormal returns. The market index is the CRSP value-weighted index and the daily Fama-French factors are also obtained from CRSP. The estimation period is the 200 trading days ending 61 trading days prior to the event date.
Our goal in this paper is evaluating the ability of CreditWatch to convey information to markets in a timely fashion. Accordingly, we examine the equity market’s reaction to both the listing on and delisting from CreditWatch. We also examine to what extent the rating action that occurs upon delisting (i.e. an affirmed rating or a change in rating) is reflected in the abnormal returns.
To assess whether CreditWatch listings reflect the changes of the listed companies’ financial positions in a timely fashion, we calculate the daily AR on days surrounding the date of listing and the CAR for several event windows. For robustness, we use three different return generating models, as described in the previous section. Separate results of the mean values for listings with positive and negative potential are presented in
Analysis of the daily AR shows a significant positive (negative) reaction by the market on the day of a listing (Day 0) with positive (negative) potential. Taken alone, this finding suggests that a listing on CreditWatch provides the market with new information. However, the magnitude of the reaction on the day of listing is often substantially smaller than the CAR present in the days prior to the listing date, suggesting the listing on CreditWatch is somewhat delayed. This trend is particularly pronounced for those firms listed with negative potential. For example, the market adjusted model abnormal return on Day 0 is −2.84%, but CAR (−30, −1) is nearly three times as large at −8.20%.
Although the abnormal returns are both statistically and economically significant on the listing day, we cannot conclude whether the return on the listing day was due to the announcement of the CreditWatch list, or it is part of the continuing adjustment process that may have started as early as 60 days before the listing. Even if the significant abnormal return on the listing day is entirely related to the announcement of the listing, the results still indicate the placement of firms on CreditWatch may not be timely enough.7 Regardless of the specification for computing abnormal returns (the market model, market adjusted return model, or the Fama-French model), the results in
. Equity market reaction to CreditWatch listing.
AR | Firms with Positive Potentials | Firms with Negative Potentials | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
Day | MKT Model | MADJ Model | FF Model | MKT Model | MADJ Model | FF Model | ||||||
−7 | 0.08% | 0.26% | 0.11% | −0.28% | ** | −0.31% | ** | −0.26% | ||||
−6 | −0.07% | −0.01% | −0.11% | −0.32% | *** | −0.38% | *** | −0.32% | *** | |||
−5 | 0.15% | 0.39% | 0.09% | −0.76% | *** | −0.77% | *** | −0.69% | ||||
−4 | 0.34% | * | 0.43% | ** | 0.36% | −0.24% | * | −0.32% | ** | −0.24% | ||
−3 | 0.78% | *** | 0.78% | *** | 0.70% | ** | −0.21% | * | −0.24% | * | −0.21% | *** |
−2 | 0.20% | ** | 0.34% | *** | 0.11% | ** | −0.75% | *** | −0.82% | *** | −0.68% | ** |
−1 | −0.10% | 0.13% | −0.12% | −1.41% | *** | −1.46% | *** | −1.40% | *** | |||
0 | 2.71% | *** | 2.79% | *** | 2.67% | ** | −2.83% | *** | −2.84% | *** | −2.86% | *** |
1 | 0.18% | 0.27% | 0.26% | * | −1.14% | *** | −1.18% | *** | −1.11% | *** | ||
2 | 0.01% | 0.09% | 0.03% | −0.24% | ** | −0.26% | ** | −0.27% | ** | |||
3 | −0.05% | 0.03% | 0.01% | 0.25% | * | 0.20% | 0.28% | |||||
CAR | Firms with Positive Potentials | Firms with Negative Potentials | ||||||||||
Days | MKT Model | MADJ Model | FF Model | MKT Model | MADJ Model | FF Model | ||||||
(−60, −1) | 3.78% | *** | 10.01% | *** | 2.89% | ** | −7.11% | *** | −10.82% | *** | −7.08% | *** |
(−30, −1) | 1.64% | ** | 4.85% | *** | 1.37% | * | −6.23% | *** | −8.20% | *** | −5.95% | *** |
(−7, −1) | 1.39% | *** | 2.32% | *** | 1.16% | −3.97% | *** | −4.32% | *** | −3.80% | *** | |
(0, 0) | 2.71% | *** | 2.79% | *** | 2.67% | ** | −2.83% | *** | −2.84% | *** | −2.86% | *** |
(+1, +3) | 0.13% | 0.40% | 0.29% | −1.13% | *** | −1.24% | *** | −1.10% | ** | |||
N | 101 | 101 | 101 | 503 | 503 | 503 |
This table presents the mean values of the daily abnormal returns (AR) and cumulative abnormal returns (CAR) over various event windows surrounding the listing day on CreditWatch. Results are presented separately based on the expected rating change potential (positive or negative) at the time of listing. The AR and CAR are calculated using three models as described in Section 1: the market model (MKT), the market adjusted return model (MADJ), and the Fama-French 3-factor model (FF). *, **, *** denotes statistically different from zero based on a Patell z-test at the 10%, 5% and 1% levels, respectively.
and CAR, and the results (unreported) confirm the findings of the analysis using the value of the means.
We next examine the equity market reaction surrounding the listing date categorized by the delisting action. We classify the listed companies into four categories based on the magnitude of the actual change in rating that occurs on the delisting day. The four categories are: no rating change (i.e. rating being affirmed), a small rating change (changed by one notch), a medium rating change (changed by two notches), and a large rating change (changed by three notches or more). The subcategories are further separated for companies listed with positive and negative potential, creating a total of eight possible categories. The mean daily AR and CAR over various event windows surrounding the listing date for each category are presented in
From
For companies listed with negative potentials (which is a much larger sample compared to the listings with
. Equity market reaction to CreditWatch listing categorized by delisting action.
AR | Positive | Negative | ||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Day | Affirmed | Up 1 Notches | Up 2 Notches | Up 3+ Notches | Affirmed | Down 1 Notches | Down 2 Notches | Down 3+ Notches | ||||||||
−7 | 0.66% | 0.14% | −0.17% | −0.76% | −0.07% | −0.25% | * | −0.88% | ** | −0.21% | ||||||
−6 | −0.32% | −0.12% | −0.55% | 0.77% | −0.29% | ** | −0.12% | −1.53% | *** | 0.39% | ||||||
−5 | −0.06% | 0.09% | 0.90% | * | 0.16% | −0.13% | −0.86% | *** | −1.61% | *** | −1.18% | * | ||||
−4 | 0.02% | 0.01% | 2.42% | *** | 0.35% | −0.02% | −0.46% | *** | −0.03% | −0.23% | ||||||
−3 | 0.55% | 0.34% | * | 3.91% | *** | 0.31% | 0.14% | −0.29% | ** | −0.34% | −1.13% | |||||
−2 | −0.24% | 0.00% | −0.42% | 1.75% | *** | −0.54% | *** | −0.17% | ** | −1.13% | *** | −4.17% | *** | |||
−1 | 0.19% | −0.68% | * | 1.41% | *** | 0.49% | −0.87% | *** | −1.74% | *** | −1.05% | *** | −3.20% | *** | ||
0 | 4.84% | *** | 0.07% | ** | 4.24% | *** | 8.94% | *** | −0.65% | ** | −3.74% | *** | −2.89% | *** | −8.63% | *** |
1 | 0.48% | 0.17% | −0.11% | 0.00% | −0.78% | *** | −0.80% | *** | −1.43% | *** | −3.95% | *** | ||||
2 | −0.14% | −0.18% | 1.34% | ** | 0.04% | −0.29% | −0.45% | *** | −0.30% | 0.95% | ||||||
3 | −0.77% | 0.22% | 0.23% | −0.30% | 0.07% | 0.03% | 0.53% | * | 1.51% | ** | ||||||
CAR | Positive | Negative | ||||||||||||||
Day | Affirmed | Up 1 Notches | Up 2 Notches | Up 3+ Notches | Affirmed | Down 1 Notches | Down 2 Notches | Down 3+ Notches | ||||||||
(−60, −1) | 7.53% | * | −0.48% | 11.02% | *** | 6.36% | * | −1.83% | *** | −7.68% | *** | −12.66% | *** | −23.69% | *** | |
(−30, −1) | 5.53% | −1.94% | 9.84% | *** | 3.01% | −2.08% | *** | −7.09% | *** | −10.70% | *** | −18.20% | *** | |||
(−7, −1) | 0.80% | −0.23% | 7.51% | *** | 3.07% | ** | −1.78% | *** | −3.88% | *** | −6.58% | *** | −9.72% | *** | ||
(0, 0) | 4.84% | *** | 0.07% | ** | 4.24% | *** | 8.94% | *** | −0.65% | ** | −3.74% | *** | −2.89% | *** | −8.63% | *** |
(+1, +3) | −0.44% | 0.21% | 1.46% | −0.26% | −0.99% | *** | −1.22% | *** | −1.19% | ** | −1.50% | *** | ||||
N | 20 | 55 | 11 | 14 | 193 | 199 | 68 | 38 |
This table presents the mean values of daily abnormal returns (AR) and cumulative abnormal returns (CAR) over various event windows surrounding the listing date on CreditWatch categorized by the change in rating that occurs when the firm is delisted. The AR and CAR are computed using the market model as described in Section 1. The sample size is reduced because firms delisted with rating changes in an opposite direction of the initial listing are excluded from analysis. Separate results are presented for firms listed with positive and negative potential. Within the categories of positive (negative) potential, results are further classified by the magnitude of delisting action: affirmed with no change in rating, up (down) by one notch, up (down) by two notches, or up (down) by three or more notches. Abnormal returns generated by both the market adjusted return model and Fama-French 3-factor model produce qualitatively similar results and are not reported. *, **, *** denotes statistically different from zero based on a Patell z-test at the 10%, 5% and 1% levels, respectively.
positive potential), the evidence is much stronger. The magnitude of the CAR prior to the listing, regardless of the event window, exhibits a consistent monotonic trend. This is strong evidence that equity markets have anticipated prior to the listing date not only the CreditWatch listing potential (positive or negative), but also the change in rating at delisting. This point is well-illustrated in
Having provided evidence that the magnitude of the rating change at delisting is reflected in equity prices prior to the listing announcement, we next examine the information content of the delisting event by computing the AR and CAR surrounding the date of delisting.
. We take care to ensure that none of the pre-delisting windows overlap with post-listing windows. The results in Table 3 are noticeably different from those presented in Table 2 for the listing date, as there is very little market reaction on the day of the de
. Equity market reaction to CreditWatch delisting.
AR | Firms Listed with Positive Potential | Firms Listed with Negative Potential | ||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Day | Affirmed | Up 1 Notches | Up 2 Notches | Up 3+ Notches | Affirmed | Down 1 Notches | Down 2 Notches | Down 3+ Notches | ||||||||
−7 | −3.31% | *** | −0.58% | 0.23% | −1.53% | 0.01% | −0.37% | * | −1.49% | *** | 4.95% | *** | ||||
−6 | 1.04% | ** | 0.18% | * | −0.49% | −1.66% | * | 0.22% | −0.42% | −2.07% | *** | 1.05% | *** | |||
−5 | −0.93% | *** | −1.01% | ** | −0.53% | −0.98% | −0.15% | −0.13% | 0.43% | −1.22% | ** | |||||
−4 | −0.34% | 0.63% | 0.24% | −0.55% | 0.19% | 0.59% | * | 0.33% | 0.84% | |||||||
−3 | 0.10% | −0.42% | −0.38% | −0.43% | 0.11% | 0.57% | * | −0.85% | ** | 0.57% | ** | |||||
−2 | −0.20% | 0.29% | 0.18% | −1.84% | 0.26% | −0.01% | 0.01% | −2.98% | *** | |||||||
−1 | −0.03% | −0.16% | −1.48% | * | 1.18% | 0.04% | 0.35% | * | 0.15% | −0.87% | ||||||
0 | −0.55% | −0.62% | ** | 0.65% | −0.36% | 0.28% | −0.07% | 1.04% | *** | −1.49% | *** | |||||
1 | 0.56% | 0.50% | * | 0.64% | 0.39% | 0.10% | 0.16% | −0.69% | ** | 0.94% | ** | |||||
2 | −0.26% | 1.37% | * | −1.72% | ** | −0.46% | 0.13% | 0.22% | −0.24% | 0.12% | ||||||
3 | 0.07% | 0.17% | −0.19% | −0.42% | 0.22% | 0.47% | ** | −1.26% | ** | 2.29% | *** | |||||
CAR | Firms Listed with Positive Potential | Firms Listed with Negative Potential | ||||||||||||||
Day | Affirmed | Up 1 Notches | Up 2 Notches | Up 3+ Notches | Affirmed | Down 1 Notches | Down 2 Notches | Down 3+ Notches | ||||||||
(−7, −1) | −3.66% | *** | −1.07% | −2.23% | −5.82% | * | 0.68% | * | 0.59% | −3.50% | ** | 2.35% | ||||
(0, 0) | −0.55% | −0.62% | ** | 0.65% | −0.36% | 0.28% | −0.07% | 1.04% | ** | −1.49% | *** | |||||
(+1, +3) | 0.37% | 2.05% | * | −1.27% | −0.49% | 0.44% | 0.85% | * | −2.16% | ** | 3.35% | *** | ||||
N | 19 | 49 | 8 | 6 | 191 | 196 | 63 | 32 |
This table presents the mean values of daily abnormal returns (AR) and cumulative abnormal returns (CAR) over various event windows surrounding the date of delisting from CreditWatch. The AR and CAR are computed using the market model as described in Section 1. The sample size is reduced because firms delisted with rating changes in an opposite direction of the initial listing are excluded from analysis. Separate results are presented for firms listed with positive and negative potential. Within the categories of positive (negative) potential, results are further classified by the magnitude of delisting action: affirmed with no change in rating, up (down) by one notch, up (down) by two notches, or up (down) by three or more notches. Abnormal returns generated by both the market adjusted return model and Fama-French 3-factor model produce qualitatively similar results and are not reported. *, **, *** denotes statistically different from zero based on a standard t-test at the 10%, 5% and 1% levels, respectively.
listing announcement (Day 0). This is true regardless of whether the firm was listed with positive or negative potential, and also irrespective of the magnitude of the rating change upon delisting. The results suggest that the announcement of delisting (in which the actual rating changes are made) contains limited information. Collectively, the results in
We have shown that CreditWatch service improved the information release timing and moved forward the impact on equity prices from rating change day to the CreditWatch listing day. However, we have also shown the improvement in rating timeliness is not complete and the timeliness issue is not fully resolved because equity markets experience significant reactions in the days prior to a firm’s listing day.
To further confirm that rating agencies are still issuing default risk information timely enough, we now examine whether the pre-listing equity market reaction is an effective predictor of the eventual change in rating upon delisting. If it is, then the CreditWatch is too slow in reflecting the changes in the firms’ financial positions.
While the results in
Model 1 contains only the CARs from the listing period as independent variables.8 The CARs for both the pre-listing (LIST_CAR (−7, −1)) and listing date (LIST_CAR (0, 0)) are negative and statistically significant, and an F-test for the equality of coefficients shows that the magnitude of the listing date CAR is significantly greater. This suggests that the announcement day contains significantly more information than the pre-listing period. However, the inclusion of additional control variables eliminates this statistical difference.
Model 2 adds control variables for the time between the listing and delisting dates (SPAN) and the initial numerical rating at the time of listing (LIST_RATING). We also include dummy variables to control for proximity to the threshold between investment-grade and speculative-grade (junk) status. A bond rating of BB+ and below is considered junk status. Prior studies (e.g. Brister, Kennedy, and Liu [
Model 4 adds basic financial characteristics of the listed firms as control variables. We include the natural log of assets, LN (ASSETS), as measure of size, ROA as a measure of profitability, DEBT_RATIO as a measure of leverage, and CASH_RATIO as a measure of liquidity, but none of the variables have a statistically significant impact on the magnitude of the rating change.
As a robustness check, Models 5 and 6 repeat the variable structure of Model 4, but separate the observations into categories of positive and negative potentials, respectively. Separating the sample into these two categories results in a truncation of the dependent variable (i.e. the change in rating), so we estimate Models 5 and 6 using a Tobit procedure. The results for both Models 5 and 6 are consistent with the findings in Models 1 through 4. The pre-listing and listing day CARs are statistically related to the change in rating upon delisting.
. Predicting the magnitude of rating change at delisting.
Dependent Variable: Rating Change | Both Positive and Negative | Positive | Negative | ||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
(1) OLS | (2) OLS | (3) OLS | (4) OLS | (5) Tobit | (6) Tobit | ||||||||||
Intercept | 0.426 | *** | 0.991 | *** | 0.993 | *** | 1.618 | *** | 1.905 | *** | −0.33 | ||||
LIST_CAR (−7, −1) | −1.604 | *** | −1.644 | *** | −1.863 | *** | −1.515 | *** | −1.345 | *** | −1.248 | ** | |||
LIST_CAR (0, 0) | −3.006 | *** | −2.456 | *** | −2.445 | *** | −2.151 | *** | −2.076 | *** | −1.987 | *** | |||
LIST_CAR (+1, +3) | 0.177 | 0.284 | 0.282 | 0.278 | 0.171 | −0.511 | |||||||||
SPAN | −0.002 | *** | −0.002 | *** | −0.002 | *** | −0.002 | *** | −0.001 | ||||||
LIST_RATING | −0.03 | ** | −0.031 | ** | −0.035 | * | −0.048 | ** | −0.032 | ||||||
NEAR_JUNK | 0.253 | ** | 0.294 | ** | 0.31 | ** | 0.383 | *** | |||||||
NEAR_INVESTMENT | −1.806 | *** | −1.837 | *** | −1.848 | *** | −2.944 | *** | |||||||
LIST_CAR (−7, −1)* NEAR_JUNK | 1.66 | * | 2.066 | ** | 1.271 | ||||||||||
LIST_CAR (−7, −1)* NEAR_INVESTMENT | −3.033 | −2.827 | −1.746 | ||||||||||||
LN(ASSETS) | −0.013 | −0.009 | −0.002 | ||||||||||||
ROA | −0.305 | −0.312 | −0.025 | ||||||||||||
DEBT_RATIO | −0.103 | −0.245 | −0.044 | ||||||||||||
CASH_RATIO | 0.276 | −0.08 | 0.051 | ||||||||||||
Year Dummy | No | No | No | Yes | Yes | Yes | |||||||||
R-Square | 0.087 | 0.194 | 0.209 | 0.217 | 0.47 | 0.340 | |||||||||
F-Value | 19.93 | *** | 21.58 | *** | 17.24 | *** | 11.1 | *** | |||||||
N | 598 | 598 | 598 | 583 | 583 | 583 | |||||||||
Equality of Coefficients (F-Test) | Difference | Difference | Difference | Difference | Difference | Difference | |||||||||
H0: LIST_CAR (−7, −1) = LIST_CAR (0, 0) | 1.402 | *** | 0.812 | 0.582 | 0.636 | 0.731 | 0.739 | ||||||||
This table presents results of regressing rating changes at delisting on the CAR surrounding the listing date and other control variables. The dependent variable in all models is the numerical change in rating that occurs upon delisting from CreditWatch. An upgrade (downgrade) is reflected by a negative (positive) number. LIST_CAR (T1, T2) is the CAR computed using the market model from days T1 to T2 relative to the day listed on CreditWatch. SPAN is the number of days between listing and delisting. LIST_RATING is the numerical rating at the time of listing. NEAR_JUNK is a dummy variable equal to one for negatively listed firms with an initial rating of BBB+, BBB, or BBB−. NEAR_ INVESTMENT is a dummy variable equal to one for positively listed firms with an initial rating of BB+, BB, or BB−. Financial variables are from the fiscal year ending prior to the listing date: LN (ASSETS) is the natural log of total assets, ROA is net income divided by total assets, DEBT_RATIO is total liabilities divided by total assets, CASH_RATIO is cash and equivalents divided by total assets. ***, **, * denotes statistical significance at the 1%, 5%, and 10% levels, respectively.
To ascertain the economic significance of our results, consider the coefficient for LIST_CAR (−7, −1) in Model 4 of −1.515. A one percentage point decline in the pre-listing CAR is associated with a rating downgrade of about 0.015 notches. At first glance this may seem trivial, but the mean pre-listing CARs from
Studies in the literature (e.g. Brister, Kennedy, and Liu, 1994; Jorion and Zhang, 2007) have shown that for the same magnitude of downgrade (e.g. downgrade by one notch), the impact on a low-grade bond (e.g. from B+ to B) tends to be greater than a high-grade bond (e.g. from A+ to A) because low-grade bonds are closer to bankruptcy and they are more scrutinized by investors and regulators. As a robustness check of our results, we examine whether the same principles holds true for the CAR surrounding the listing on CreditWatch. We choose the largest category with similar rating changes, the one notch downgrades, and construct three sub-categories based on the initial rating: high ratings (A+), medium ratings (BBB), and low ratings (B+).
The results of the analysis are presented in
We examine whether rating agencies are sacrificing timeliness by pursuing rating stability by investigating the response of the equity market to the listing and delisting of firms on S&P’s CreditWatch, a service whose intended purpose is to improve the timeliness of information about changes in credit ratings. We find that while it made some improvement on the intended purpose, CreditWatch is still not completely effective at achieving this goal.
. Reaction to one-notch downgrade based on initial bond quality.
AR | Downgrade-by-One-Notch | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
Day | Downgrade from A+ to A | Downgrade from BBB to BBB− | Downgrade from B+ to B | ||||||||
−7 | 0.18% | −0.58% | * | −1.00% | |||||||
−6 | −0.89% | ** | −0.74% | −1.23% | |||||||
−5 | 0.23% | −0.67% | −3.21% | ** | |||||||
−4 | −0.46% | −0.91% | ** | −1.84% | ** | ||||||
−3 | −0.68% | −0.64% | −1.38% | ||||||||
−2 | 0.00% | −0.28% | 2.96% | * | |||||||
−1 | −0.99% | −1.92% | *** | −2.30% | *** | ||||||
0 | −2.67% | *** | −0.28% | −8.69% | *** | ||||||
1 | −0.35% | 0.26% | −3.17% | *** | |||||||
2 | −0.38% | −0.81% | ** | 2.54% | ** | ||||||
3 | −0.85% | * | 0.33% | 4.54% | *** | ||||||
CAR | Downgrade-by-One-Notch | ||||||||||
Day | Downgrade from A+ to A | Downgrade from BBB to BBB− | Downgrade from B+ to B | ||||||||
(−60, −1) | −2.19% | −8.97% | *** | −19.52% | *** | ||||||
(−30, −1) | −3.73% | −7.91% | *** | −13.37% | ** | ||||||
(−7, −1) | −2.62% | * | −5.74% | *** | −8.00% | *** | |||||
(0, 0) | −2.67% | *** | −0.28% | −8.69% | *** | ||||||
(+1, +3) | −1.57% | ** | −0.22% | 3.91% | |||||||
N. | 10 | 23 | 12 | ||||||||
This table examines the effect of bond quality on the magnitude of daily abnormal returns (AR) and cumulative abnormal returns (CAR) over various event windows surrounding the listing date. Abnormal returns are computed using the market model (MKT) as described in Section 1. Downgraded by one notch samples are decomposed into different subgroups according to their original listing position. Variables are defined same as before. *, **, *** denotes statistically different from zero based on a standard t-test at the 10%, 5% and 1% levels, respectively.
We report three empirical results that support our conclusion. First, we find that equity markets experience substantial positive (negative) abnormal returns for companies listed with positive (negative) potential on CreditWatch prior to the listing date. The pre-listing abnormal returns not only reflect the direction of potential rating changes, but also the magnitude of actual rating changes on the delisting date. These results suggest that rating agencies are still slow in conveying information to the market. Second, equity markets exhibit significant abnormal returns on the CreditWatch listing day, but little reaction to the delisting of a company from CreditWatch. The results suggest that the impact of rating agency’s announcement has moved forward from rating change day to CreditWatch listings day. The results suggest that CreditWatch listing improved the timeliness of rating functions but has not completely resolved the issue of timeliness. Third, to further confirm the conclusion, we find that the pre-listing abnormal returns in equity markets are good predictors of both the direction and the magnitude of the eventual change in credit. This is especially true for those firms listed with negative potential, which is by far the most common listing type. Collectively, our findings suggest that rating agencies may sacrifice timeliness for the sake of stability and that even CreditWatch, which is designed to mitigate the disadvantage caused by rating stability, is not a completely effective instrument.
If an advance notice service such as CreditWatch is already substantially anticipated by the market and too slow in conveying information, credit rating agencies may need to reconsider whether a policy of issuing stable ratings is too costly. In order to repair the reputational damage suffered during the 2007-2008 financial crisis, credit rating agencies may have to develop more effective measures to convey changes in default probability to the market in a timely manner.
We are grateful for the helpful comments made by Kathy Fogel, Tomas Jandik, Wayne Lee, Alexey Malakhov, and Tim Yeager. The remaining errors, if any, are solely ours.