Market reactions to the credit risk announcements





by

Jungsoon Shin

January 24
th
2008







A dissertation submitted to the Faculty of the Graduate School of the

State University of New York at Buffalo

in partial fulfillment of the requirements for the degree of





Doctor of Philosophy

Department of Finance & Managerial Economics














UMI Number: 3291588
3291588
2008
UMI Microform
Copyright
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, MI 48106-1346
by ProQuest Information and Learning Company.
Dedicated to
EunDong Shin and KyungBun Park, Parents

And Misook Choi, Wife
For their consistent care and support.








































ii

Acknowledgements



Completion of dissertation marks an end of my academic training and the beginning of my
professional life. Reflecting on my experience of doctoral years, I am deeply indebted to Dr.
Kee H. Chung for serving as my dissertation chair. Without his guidance and encouragement, I
would not have come to completion of doctoral study.
I would like to thank two other committee members, Dr. Joseph P. Ogden and Dr. Kenneth
A. Kim. Dr. Ogden’s constructive comments have been useful. As a coordinator in doctoral
program, Dr. Kim provided me with invaluable advices. I am also grateful to Dr. Michael S.
Rozeff. He has taught me two doctoral seminars, instilling deep academic insight.
I am honored to discuss academic issues with finance faculty in the class and in the
seminar presentations. Being with them has meant a great deal to me.













iii
Table of contents
Dedication ii
Acknowledgement iii
Abstract vi

Essay 1: Credit ratings and market liquidities
1 Introduction 1
2 Related literatures 6
3 Data and descriptive statistics 8
4 Measurement of variables 10
4.1 Liquidity and other market microstructure variables for corporate bond market 10
4.2 Liquidity and other market microstructure variables for stock market 11
5 Empirical results 13
5.1 Pre-credit and post-credit rating announcement tests 13
5.2 Unconditional tests using the event study methods 16
5.3 Conditional tests using the event study methods 19
5.4 Cross sectional regression analysis 21
5.4.1 Regression results for corporate bonds 21
5.4.2 Regression results for stocks 23
6 Transaction cost determinants 24

6.1 Transaction cost determinants in bond market 25
6.2 Transaction cost determinants in stock market 26
7 Conclusions 27

iv
Essay 2: The effect of credit announcements on bond and stock prices
1 Introduction 51
2 Credit rating process and hypotheses development 56
2.1 Credit rating and watchlists 56
2.2 Credit rating agencies practices and objectives 57
2.3 Testable hypotheses 58
2.3.1 Information contents between reviews and rating actions 58
2.3.2 Market impact on categorical credit rating levels 58
2.3.3 Bond market versus stock market 59
2.3.4 Conflict of interests hypothesis 60
3 Data and descriptive statistics 60
4 Measurement of abnormal bond and stock returns 63
4.1 Measurement of abnormal bond returns 63
4.2 Measurement of abnormal stock returns 65
5 Unconditional (Unconditional) market impacts on the bond and stock markets 66
5.1 Unconditional tests of market impacts on bond and stock markets 66
5.2 Conditional tests of market impacts on the bond and stock markets 68
6 Regression analyses of abnormal returns around watchlists dates 71
7 Cross sectional analysis of anticipation hypothesis 73
8 Conclusions 74




v


vi
Abstract

Essay 1:
Using the extensive datasets, we analyze the effect of credit rating announcements on the
bond and stock liquidities. We show that the average transaction costs increases (decreases)
after downgrade (upgrade) credit rating news events. We interpret these results as evidence
that liquidities improve or deteriorate following upgrade or downgrade announcement. We
also find more pronounced phenomenon among the lower credit rated securities. That is,
securities with poorer credit quality respond more strongly to the news announcements.
Lastly, we report that the watchlists announcement convey more information contents than
real rating changes.



Essay 2:

Using the traditional event study methodology, we examine the effect of Moody’s credit
rating announcements on the bond and stock prices. We show that the cumulative abnormal
bond and stock returns are positive (negative) and significant for the upgrade (downgrade)
credit rating news events. We also find more prominent results among the lower credit rated
bonds and stocks. For the highest quality group of bonds and stocks, we find meager and
insignificant reaction to the news announcements. We also report that the rating reviews
announcement release has more information contents than following rating changes.



Essay 1: Credit ratings and market liquidities


1. Introduction
Liquidity is the ability to trade large quantities quickly, at low cost, when investors want
to trade. It is one of the most important characteristics in well-functioning markets. Conceptually,
transaction costs are closely related to liquidity because transaction costs are low in a liquid
market. Market microstructure studies have generated numerous papers concentrating on stock
markets regarding these issues (e.g., Benston and Hagerman, 1974; Demsetz, 1968; Glosten and
Milgrom, 1985; Huang and Stoll, 1996; Kaul and Nimalendran, 1991; Roll, 1984; Stoll, 1978,
1989, 2000). On the other hand, just a few studies offer analytical and experimental evidence
regarding the bond transaction costs. Whereas Chakravarty and Sarkar (2003) find that the bid–
ask spread is lower for higher credit rating bonds, Schultz (2001) reports that the liquidity pattern
is not associated with credit risk. Edwards, Harris, and Piwowar (2007) find secondary corporate
bond transaction costs increase with credit risk. However, none of these studies consider the
effect of credit rating announcements on the liquidity of either the stock or the bond market.
1

The primary cause for the paucity of research on the fixed income markets is that reliable data
were not available until recently. Not surprisingly, some prior works have relied on weekly or
monthly data to study the corporate bond market.
2

In this study, we examine the short-run performance of both stock and corporate bond
market transaction costs with different credit risk profiles when there are credit news
announcements. Specifically, we address following questions:
1. How do credit rating announcements such as watchlists and rating changes affect

1
Liquidity, transaction cost, and bid–ask spread are used interchangeably in this paper.
2
Katz (1974) investigates monthly yield changes, Grier and Katz (1976) examine monthly bond returns,
and Hite and Warga (1997) also study monthly abnormal bond returns.



1

liquidities? Specifically, are upgrade (downgrade) news events associated with increased
(decreased) liquidity changes? If so, which announcement between watchlists or rating
changes has more impact on the liquidities?
2. Are liquidity changes similar across the credit rating level? For example, will highly rated
bonds respond to rating announcements in a similar way as poorly rated bonds?
3. Is there any structural difference between the bond and stock market in terms of liquidity
changes?
4. Do asymmetric reactions exist between downgrade and upgrade news announcements?
Harris and Piwowar (2006) and Edwards et al. (2007) note that bonds with lower credit
risk are more expensive to trade, which is consistent with adverse selection risk theory in the
market microstructure literature. Odders-White and Ready (2006), who report similar findings in
the stock market, show that firms with greater risk of private shocks—and, therefore, higher
levels of equity–market adverse selection—have lower credit ratings. Higher adverse selection
risk on lower rated bonds is consistent with the general notion of the corporate bond market.
Institutions such as insurance companies, mutual funds, and public pension funds are the main
investors for corporate bonds. According to regulations, they are required to buy and hold quality
bonds, and, thus, do not actively trade high-yield bonds. However, compared with these
institutional investors, relatively less regulated hedge funds actively trade speculative bonds.
Therefore, information-based trading is more likely to occur on lower rated bonds.
Chakravarty and Sarkar (2003) argue that based on the inventory model in a market
microstructure, dealers’ costs of adjusting inventory should be incorporated into the bid–ask
spread, implying that transaction costs increase with the risk of the security.
3
Therefore, to the

3

Refer to Garman (1976), and Chakravarty and Sarkar (2003) for the inventory hypothesis.

2

extent that transaction costs increase with credit risk, we expect that the bid–ask spread will
increase (decrease) with a downgrade (upgrade) credit rating announcement. In other words,
liquidity will improve (deteriorate) with an upgrade (downgrade) credit news event. Furthermore,
following similar logic, more pronounced spread changes will be detected on speculative-rated
bonds compared with high-quality bonds because informed trading is more likely to occur on
lower rated bonds, causing dealer inventory imbalance.
The stock market has strived to improve transparency. To enhance the ability of market
participants to observe information regarding pretrade order flow and posttrade price and
quantity, several rules have been made or changed (Boehmer, Saar, and Yu, 2005; Chung and
Zhao, 2007).

Although transparency in the corporate bond market improved following the
introduction of the Trade Reporting and Compliance Engine (TRACE) system in 2002, it is not
yet comparable to stock market. In addition, the bond market is less liquid than the stock market
in terms of trading activity. If we assume that market transparency and liquidity are important
factors in market efficiency, we can conclude that the corporate bond market is less efficient than
the stock market.
4
Based on these arguments, we conjecture that information regarding credit
rating announcements will be more efficiently incorporated into the stock market than the bond
market, so that liquidity transitions will be better detected in the stock market.
We separate the short-term effects of two related announcements regarding credit rating:
watchlists and rating changes. Specifically, watchlists, sometimes referred to as rating reviews,
are the statements made by a full rating committee prior to the actual rating change
announcements.
5

For stability purposes, the rating agencies tend to make subsequent rating

4
Transparency is related to the market efficiency. Bloomfield and O’Hara (1999) show that trade
transparency increases information efficiency.
5
Moody’s uses watchlists and rating reviews interchangeably.

3

changes in the same direction as the watchlists events.
6
In other words, their rating system
management practices attempt to limit rating reversals. Thus, rating reviews appear to contain
more information content than rating changes news. In addition, agencies are likely to issue
watchlists when actual rating changes are likely in the near future. That is, putting ratings under
review sends a strong signal to the market that actual rating changes are impending. By the
nature of credit rating procedures, watchlists announcements are more likely to be unexpected
and contain more information; hence, they have less preannouncement effect than actual rating
changes.
To examine the effect of credit news on liquidities, we compare the daily percentage
spreads before and after rating announcements. Our results show that for the corporate bond
market, the mean daily percentage spread declines (rises) after upgrades (downgrades) rating
news announcements. As expected, we find the effect on high-yield bonds is more pronounced
than on investment grade bonds. Specifically, bonds graded Caa and below show much stronger
rating announcement effects, and bonds graded A and above do not follow expected patterns.
7

Consistent with our conjecture, the statistical magnitudes based on watchlists dates are greater
than those based on rating changes dates.

We further analyze announcement effects using the three-day cumulative abnormal
spread (CAS) surrounding both credit watchlists and rating changes dates. Consistent with our
before and after test results, we find that downgrade (upgrade) announcements based on rating
reviews and rating changes dates are associated with positive (negative) mean CASs. We find
similar results when we partition watchlists conditioned on future rating changes. Based on

6
Keenan, Fons, and Carty (1998) show that 76.44% of ratings placed on review for upgrades are, in fact,
upgraded and 66.77% of downgrade rating reviews are actually downgraded

7
Bonds rated A and above include Aaa, Aa1, Aa2, Aa3, A1, A2, and A3, and bonds rated C and below
include Caa1, Caa2, Caa3, Ca, and C.

4

watch start dates, downgrade (upgrade) news announcements, followed by rating downgrades
(upgrades), are associated with positive (negative) mean CASs.
To check whether our results are driven by other factors, we perform cross-sectional
regressions of the CASs onto credit rating levels, multinotch, and fallen angel/rising star by
controlling issue characteristics such as coupon rate, face value, time to maturity, price, trade size,
and issuer size.
8
Our results indicate that observed changes in bid–ask spreads cannot be
attributed to changes in issue attributes.
9
We apply the same tests to the stock market and obtain
the similar results. In addition, the magnitudes of rating announcement effects are greater for the
stock market than corporate bond market in the cross sectional analysis.
Finally, we run a cross-sectional regression for the determinants of transaction costs.

Consistent with prior studies, we find that transaction costs are positively associated with coupon
rate and inverse price and negatively associated with maturity, trade size, and maturity.
10
We also
include two market microstructure variables—volatility and number of trades—that prior studies
have not considered. As expected, volatility is significantly and positively related and number of
trades is negatively related to the transaction costs.
This study makes several contributions to the literature. First, we examine the effect of
credit rating announcements on liquidity in terms of transaction costs. As far as we know, this
study is the first to model and test the behavior of bond and stock liquidities around credit rating
announcements. Second, we report the results of both the corporate bond market and the stock
market. By comparing both markets, we confirm that the stock market is more liquid and

8
For issuer characteristics, we control for leverage, trading volume, number of trades, volatility, and
issuer size.
9
However, statistical significance is mainly observed in the case of downgrade announcements.
10
Chakravarty and Sarkar (2003), Harris and Piwowar (2006), and Edwards et al. (2007) also analyze
transaction costs determinants using issue-specific variables and other bond complexities. Bond-
specific variables include coupon rate, maturity, and bond age and bond complexities include sinking
funds, callable bonds, float rate bonds, variable rates, and so on.


5

efficient. Our results support Edwards et al.’s (2007) assessment that further actions are needed
to increase bond market transparency. Third, only a few studies, using limited samples, have
presented evidence on rating reviews. However, drawing from Moody’s extensive and historic

database, we show that rating reviews convey information to the market participants in greater
magnitude than real rating changes.
The remainder of this paper is organized as follows. Section 2 discusses the related
literature, Section 3 describes data sources and presents descriptive statistics, and Section 4
explains the measurement of variables. Section 5 discusses the empirical results, Section 6
examines transaction cost determinants, and Section 7 concludes.

2. Related literature
Several studies have analyzed bond market transaction costs. Chen, Lesmond, and Wei
(2002) extend Lesmond, Ogden and Trzcinka’s (1999) indirect estimate of equity transaction
costs to corporate bonds. They hypothesize that measured prices will reflect new information
only if the information value of the informed marginal trader exceeds the total liquidity costs.
They assume that a zero-return day (including a nontrading day) is observed when the true price
changed by less than the transaction costs. Using this assumption and applying a two-factor
return generating model, Chen et al. develop an estimate of transaction costs. However, their
method only uses information from the last transaction on each day and treats all transactions as
if they occurred at or near the end of each trading day.
Schultz (2001), who proposes a regression-based approach to retain a much larger
percentage of his corporate bond observations, finds that credit ratings are not associated with
spread. However, his results, which are inconsistent with our findings, may be due to the noise

6

created in his data from including only month-end bid quotes. Chakravarty and Sarkar (2003)
compare trading costs in the U.S. corporate, municipal, and Treasury bond markets. They report
that the bid–ask spread is not statistically different between the corporate and Treasury bonds but
is higher for municipal bonds. They use Capital Access International (CAI) data, which contain
transaction information by institutions. However, because only institutional trades are reported,
the CAI data do not represent the entirety of the corporate bond market. Hence, their sample
induces selection bias. Hong and Warga (2004) estimate effective spreads for municipal bonds

with publicly available Municipal Securities Rulemaking Board transaction data, but their study
is limited to the municipal bond market, excluding the corporate bond market. Harris and
Piwowar (2006) also examine secondary trading costs in the U.S. municipal bond market and
propose new methods for measuring transaction costs. They find that transaction costs are related
to credit rating and time to expected maturity in their cross-sectional analysis.
With the introduction of the National Association of Securities Dealers’ comprehensive
TRACE database, a few researchers began to study transaction costs of corporate bonds in terms
of transparency. Edwards et al. (2007) examine the corporate bond transaction costs in terms of
bond market transparency by extending Harris and Piwowar’s (2006) transaction cost estimation
methods. They find that trading costs are lower for transparent bonds than for similar opaque
bonds and that these costs fall when a bond’s prices become transparent. They interpret these
results as evidence that transparency improves liquidity in corporate bond markets.
Bessembinder, Maxwell, and Venkataraman (2006) perform a similar study. Using a
sample of institutional trades in corporate bonds before and after initiation of the TRACE
reporting system, they develop a model of the effect of public transaction reporting on trade
execution. They find that trade execution costs fell approximately 50% (20%) for bonds eligible

7

(ineligible) for TRACE transaction reporting. Goldstein, Hotchkiss, and Sirri (2007) report
similar results in their examination of the impact of increased posttrade transparency on market
liquidity in a controlled setting. Including only 120 BBB-rated corporate bonds from July 2002
to February 2003 in their sample, they find that, except for the largest trade size group, spreads
decrease for bonds whose prices become transparent; however, they do not find a significant
change in spreads for very thinly traded bonds. Thus, they conclude that, depending on trade size,
increased transparency has a neutral or positive effect on liquidity. Using the TRACE database,
most studies focus on transparency issues in terms of transaction cost reduction.

3. Data and descriptive statistics


For the bond market liquidities, we obtain credit ratings and watchlist data from Moody’s
Default Risk Service Database. This database provides access to Moody’s complete proprietary
default database, featuring data on rating reviews and actions at the issuer and issue levels. It also
provides access to credit histories for over 10,000 issuers and more than 200,000 individual debt
securities and includes relevant information such as face amount, time to maturity, debt class,
market type, and coupon rate of each rated debt issue. In addition to the ratings history, the
database includes watchlist events, which can be understood as an interim review by credit rating
agencies toward future rating changes. Since September 30, 1991, Moody’s has assigned one of
the following three categories to the issue under review: possible upgrade (UPG), possible
downgrade (DNG), and uncertain (UNC). Watchlists are usually followed by official credit
rating changes later; thus, actual rating changes are more or less expected for the bonds with
precedent watchlist information.
The corporate bond transaction data are acquired from TRACE, which was established on
July 1, 2002, to disseminate corporate bond prices for all bonds traded in the over-the-counter

8

market. The stated purpose of creating this system is to improve the transparency of the
corporate bond market. TRACE’s unique and comprehensive data set offers an unparalleled
opportunity for bond transaction-related study. The final, merged database includes 2,853
individual watchlist events during the sample period from July 2002 to April 2006.
Panel A of Table 1 describes issue characteristics of 6,426 samples according to letter
credit ratings. We identify 11,686 securities over our sample period; after filtering, 6,426 bonds
with prices both before and after the announcements dates remain.
11
The results show that
coupon rates are negatively associated with bond ratings; that is, For the highest credit rating
group of A and above, the coupon rate is 5.4707% whereas for the lowest category of Caa and
below, the coupon rate is 8.2587%. We find a similar result in a comparison of the investment
(5.6690%) and speculative (6.5785%) credit rating categories; that is, bonds with higher ratings

are issued with lower coupon rates, which is consistent with Edwards et al.’s (2007) result.
Because coupon rate could be a proxy for credit quality, the lower quality bonds pay higher
coupon rates to attract investors to invest in risky bonds. Time to maturity (avg. = 12 years)
12
is
calculated by subtracting bond issue date from maturity date. To examine debt seniority, we
classify issues into three categories: secured, subordinated, and others.
13
If Moody’s debt
description includes a phrase equivalent to “subordinated,” we classify it as subordinated. If a
phrase similar to “secured” or “senior” appears, the bond is classified as secured. The remainder
is classified as others. The proportion of subordinated bonds increases as credit rating
deteriorates. Only 4.55% of bonds rated A and above are subordinated whereas 18.22% of bonds
rated C and below are subordinated. Finally, we provide the nationality of the bond, which is

11
We only pick up samples that are detected around the announcement events dates.
12
Instead of years, we report days in Table 1.
13
Moody’s look-up data file defines 16 subgroups such as senior secured, junior subordinated, senior
unsecured, and so forth.


9

defined as market in which debt was issued. The results show that bonds with high credit ratings
are more likely to be issued in United States rather than Europe or other regions.

[TABLE 1 ABOUT HERE]


For the stock market liquidities, we obtain the data from the NYSE’s Trade and Quote
database. For the issuer characteristics, we merge the Trade and Quote database with the
Compustat and CRSP databases. Panel B of Table 1 describes issuer characteristics such as firm
size, Tobin’s q, leverage, market capitalization, and return on assets. Some variables show linear
function of credit ratings. As expected, firm size and market capitalization are positively related
to the credit ratings, and leverage is negatively associated with credit ratings. These results are
logical in that firms with poor credit ratings are often small companies with heavy debt loads.
For example, the leverage is approximately 24% for the investment grade bonds, compared with
35% for speculative grade bonds.

4. Measurement of variables
4.1 Liquidity and other market microstructure variables for corporate bond market
We measure bond price by the mean daily transaction price, return volatility by the
standard deviation of daily returns calculated from daily closing transaction price, number of
trades by the average daily number of transactions, trade size by the average daily dollar
transaction size, and quantity by par value volume of the reported trade.
As a proxy of bond market liquidity, the percentage spread is calculated by the following
formula:
14


14
By definition, the bond spread calculation is subject to a bond with at least one buy and one sell

10


Percentage bond spread
it

= ((
N
t1
1
=
/N)ΣA
it
– ( /M)ΣB
it
)/P
it
, (1)
M
t=1
1

where A
it
is the price of transaction at the ask price for bond i, B
it
is the price of transaction at the
bid price for bond i, N is the number of transactions that occurred at the ask price per day, M is
the number of transactions that occurred at the bid price per day, and P
it
is the mean daily bond
price, which is the daily mean of A
it
and B
it
.

15
In other words, we calculate percentage spread as
the difference between mean daily ask price and mean daily bid price divided by mean daily
bond price. In addition, we filter the sample to minimize data errors: We exclude (a) the bid–ask
quotes if the spread is greater than $10 or less than zero; (b) trade price P
t
if |(P
t
– P
t
– 1)/P
t–1
| >
0.50; (c) ask price A
t
if |(A
t
– A
t–1
)/A
t–1
| > 0.50; and (d) bid price B
t
if |(B
t
– B
t–1
)/B
t–1
| > 0.50.

Table 2, Panel A, shows that transaction costs are an increasing function of credit risk,
which is consistent with Chakravarty and Sarkar (2003) and Edwards et al.(2007). Specifically,
the average transaction cost for grade A bonds is 1.63%; for Baa1 and above, 2.24%, for B1 and
above, 2.27%; and C, 3.00%. In overall, the average transaction cost is 2.10%. As Edwards et al.
note, higher rated bonds cost less to trade than lower rated bonds. Number of trades, volatility,
and trade size are positively associated with credit risks.

[TABLE 2 ABOUT HERE]

4.2 Liquidity and other market microstructure variables for stock market

transaction in a day.
15
TRACE data in the Wharton database does not provide information on whether a specific transaction is
initiated by the seller or the buyer; therefore, it is difficult to estimate ask or bid price for the specific
transaction. Employing the tick test suggested by Lee and Ready (1991), we classify a trade as buyer
initiated if the trade price is above the previous price. Correspondingly, when the trade price is below
the previous price, we classify it as seller initiated.

11

We measure share price by the mean daily quote midpoints and return volatility by the
standard deviation of daily returns calculated from daily closing quote midpoints. We measure
number of trades by the average daily number of transactions and trade size by the average dollar
transaction size. For the depth measurement, we divide the sum of bid size and ask size in two.
We calculate percentage dollar spreads using the following formula:

Quoted percentage spread
it
= (A

it
– B
it
)/M
it
, (2)

where A
it
is the posted ask price for stock I, B
it
is the posted bid price for stock i, and M
it
is the
mean of A
it
and B
it
. We follow traditional filtering rules to minimize the errors by excluding (a)
bid–ask quotes if the spread is greater than $5 or less than zero; (b) before-the-open and after-
the-close quotes; (c) trade price P
t
if |(P
t
– P
t
– 1)/P
t–1
| > 0.10; (d) ask quote A
t

if |(A
t
– A
t–1
)/A
t–1
|
> 0.10; (e) bid quote B
t
if |(B
t
– B
t–1
)/B
t–1
| > 0.10; and (f) trades if the price or volume is less than
or equal to zero. Table 2, Panel B also shows select attributes of our sample. Transaction costs
are an increasing function of credit risk. Transaction cost is approximately 2.5 times greater for
speculative grade than for investment grade bonds. If we compare the highest and lowest
categories, we find that the transactions cost for stocks graded A and above is 0.26% and for
stocks graded C and below is 1.79%, which is six times greater. When we compare the stock and
bond markets, we find that transaction costs are lower in the stock market. Unlike the corporate
bond market, the number of trades and trade size are negatively associated with credit risks.
These differences are attributed to higher trading frequency of high-yield bonds. However, in the
stock market, good quality firms are actively traded in terms of both the number and volume.


12

5. Empirical results

5.1 Precredit and postcredit rating announcements tests
In this section, we provide graphical and analytical results for the credit rating
announcements effects on the transaction costs of corporate bonds and stocks. Figure 1 displays
the average bond spreads, ranging from –90 days to 90 days around the watchlist announcement
dates. We can easily see how bond spreads behave around the event day. In response to
downgrade rating reviews, the bond spread begins to increase sharply about three weeks prior to
the event and then becomes stable following the event date. However, in the case of upgrade
news events, the spread declines very gradually during the 90 days before and after the event day.
Asymmetric responses of spreads patterns are clearly observed because spread movement is
more pronounced in the case of a downgrade event. Figure 2, which plots the spread patterns in
stock market, shows similar results to Figure 1. Liquidity improves with upgrades and
deteriorates with downgrades around the watch news announcement day. Figure 3 and 4 show
the bond spread behaviors of both investment and noninvestment groups around the watchlist
dates. For the downgrade watchlist announcement, transaction costs increase for both investment
and speculative category. However, for the upgrade rating review news event, they do not seem
to decrease in a significant manner. Figures 5 and 6 show the stock spread patterns of both
investment and speculative grade. To the possible downgrade (upgrade) watchlists, speculative-
grade stocks react with increased (decreased) spreads movements. However, investment-grade
stocks do not seem to respond to the news noticeably.

[FIGURES 1–6 ABOUT HERE]


13

Panels A and B of Table 3, which provide the mean percentage spreads for corporate
bonds during the pre- and postperiods of rating announcements, show whether the differences in
mean spreads between two periods are statistically significant. Based on rating changes dates, the
mean effective percentage spreads decline by 0.0002 (1.5% decline) for upgrade rating changes.
On the other hand, the same spreads rise by 0.0007 (4% rise) for downgrade rating changes.

Therefore, the changes in spreads are statistically significant and support our hypothesis. Much
stronger effects are found in downgrade announcements.
16
One noticeable point is that most of
changes in spread occur in speculative-grade ratings. In response to the investment grades for
downgrade rating changes, the spread change is 0.0002 whereas the corresponding figure in
high-yield grades (i.e., noninvestment grades) is 0.0020, which is 10 times greater than that of
investment grades. Specifically, in C grade and below category, the spread change is 0.0035.
17
In
contrast, it is –0.0013 for A grade and above; notwithstanding weak magnitude, this sign is
against our expectation. When we redo the same test based on watch beginning dates, we find the
similar results. The mean spreads decline by 0.0005 (3.6% decline) for the upgrade rating
changes and rise by 0.0016 (9% rise) for the downgrade rating changes. Asymmetric effects are
found in the changes in spreads, which are statistically significant. To compare with rating
changes, we perform the same tests for watch beginning dates and find that the magnitude is
much greater. For example, the decline is 1.5% (4%) versus 3.6% (9%) for the upgrade
(downgrade) rating announcements, which is consistent with our hypothesis. Because watchlists
events are unexpected announcements, they have greater announcement effect than rating

16
A large amount of literature has explored the relation between bond prices and changes in credit ratings.
One of the principal findings is an asymmetric response to the rating changes. That is, positive ratings
give rise to smaller, less significant price movements (Hand, Holthausen and Leftwich, 1992;
Holthausen and Leftwich, 1986; Goh and Ederington, 1993;.Pinches and Singleton, 1978).
17
We classify our sample into four categories. Grade A and above include Aaa, Aa1, Aa2, Aa3, A1, A2,
and A3; grade Baa includes Baa1, Baa2, and Baa3; grades Ba and B include Ba1, Ba2, Ba3, B1, B2,
and B3; and grade Caa and below includes Caa1, Caa2, Caa3, Ca, and C.



14

changes news.

[TABLE 3 ABOUT HERE]

Panels C and D of Table 3 show the mean percentage spreads for the stock market during
the pre- and postperiods of rating announcements. Based on rating changes dates, spreads decline
by 0.0003 (11% decline) for the upgrade rating change and rise by 0.0014 (21% rise) for the
downgrade rating changes, which is statistically significant at the l% level. When we perform the
test with watch beginning dates, we obtain a 15% rise (19% decline) for the downgrade
(upgrade) rating reviews. These values of decline and rise are much greater in the stock sample
than in the corporate bond sample. We conjecture that because corporate bonds are less
frequently traded, less transparent, and less liquid, spreads do not respond to the credit risk news
as efficiently as the stock market.
18

Overall, our results suggest that corporate bonds and stocks experience a significant
increase in transaction costs around upgrade announcement dates and decrease around
downgrade news events. We interpret these results as evidence that liquidity improves after a
positive credit rating. In contrast, liquidity deteriorates following a negative credit rating
announcement. These results for corporate bonds are not surprising because prior research has
found that higher rated bonds cost less to trade than lower rated bonds (Chakravarty and Sarkar,
2003; Edwards et al., 2007; Hong and Warga, 2000). However, more interesting, we obtain the
similar results for the stock market because prior studies have not investigated the relation
between credit risk and transaction costs. Furthermore, we find more pronounced

18
Chen, Wang, and Wu (xxx) report that the corporate bond market is less informationally efficient than

the stock market, based on various specifications of the volatility–volume model.

15

increases/decreases in spreads in lower graded bonds as well as stocks, especially in the high
probability of default risk rating bonds such as Caa1, Caa2, Caa3, Ca, and C. In addition,
consistent with our conjecture, the magnitude of spreads’ reactions to the credit ratings news is
greater for watch announcement events than real rating changes and for stocks than for corporate
bonds.

5.2 Unconditional tests using the event study methods
To access the magnitude and statistical significance of spread changes surrounding the
rating announcement dates, the standardized abnormal percentage spread of security i on day t
(SAS
i,t
) during the event period is calculated as

SAS
i,t
= (SPREAD
i,t
–û
i
)/ŝ
i
, (3)


where û
i

and ŝ
i
are the sample mean and sample standard deviation, respectively, of the spread of
security
i
during the sample period. The average abnormal spread on day t, (AAS
t
), is calculated
by averaging the standardized abnormal spread across all securities, AAS
t
= Σ
i
SAS
i,t
/N, where N
is the number of securities. Then, the CAS, (CAS
τ
) = Σ
τ
AAS
t
, where Σ
τ
denotes the summation
over t = υ through τ, where υ and τ are, respectively, the beginning and ending day of each CAS
calculation. To assess the statistical significance of the CAS, we calculate both the t-statistic and
z-statistic with their respective p-value. We obtain the z-statistic using the following formula:
19



Z
τ
= CAS
τ
/[N * (υ – τ + 1)]
1/2
.

(4)

19
See Patell (1976) and Chung and Charoenwong (1998) for a detailed description.

16


[TABLE 4 ABOUT HERE]

Panels A, B, C, and D of Table 4 show mean of three-day CASs, t-statistics, and z-
statistics for corporate bonds for each downgrade (upgrade) event based on both watchlist and
rating changes dates. The mean CAS is 1.2475 for the possible downgrade and –0.5600 for the
possible upgrade around the rating review dates. An asymmetric reaction of CAS is also detected
because the magnitude of CAS to the possible downgrade watchlists is about two times greater
than to the possible upgrade watchlists. However, both are statistically different from zero. We
discover that stocks with lower credit ratings have larger variations of liquidities.
20
The mean
CAS of stocks in the group rated Caa and below for the possible downgrade watchlists is 2.5871,
which is the greatest value among all of the groups. However, in the grade A and above category,
the sign—to say nothing of magnitude—is unexpected. That is, for the possible downgrade and

upgrade watchlists, liquidity improves. The mean CAS of the grade C group is –1.6670
compared with –0.2420 in grade A group. However, we find this result puzzling because the
magnitude and statistical significance of the average CAS are greater for investment grade
category. However, we find a rationale when we take a closer look at the relation between CAS
and credit risk.
We find the mean CAS of bonds graded from Baa3 to Baa1 in the investment credit
rating group is much greater than that from B3 to Ba1 in the speculative rating category,
indicating that the average CAS is not strictly an increasing function of credit risk across all the
categories. We note the fallen angel (rising star) effect when credit ratings change from

20
In a study of credit risk and return relation, Norden and Weber (2004) separate the sample into firms
with rating above and below the median and find stronger negative effects for downgrade when the
rating is low.

17

investment (speculative) grade to speculative (investment) grade. We expect that rising
(dropping) to the investment (speculative) grade barrier is associated with a stronger market
reaction. For the case of fallen angel, the mean CAS is 2.4293—about double the size of the
overall mean CAS of 1.2475—and statistically significant at the 1% level. We obtain similar
result for the rising star, with a mean CAS of –0.5230, which is twice as the size of the overall
test result of –0.2620, thus supporting our conjecture. When we perform the same test based on
rating changes dates, we obtain similar results with a smaller mean value of CAS. Rather than
upgrade changes, we have a more pronounced mean value of CAS for the downgrade changes.
For the fallen angel and rising star effect, we report meager statistical results based on rating
changes dates, indicating that information regarding fallen angel or rising star may be
incorporated into watchlists announcement events in advance. Overall, we also observe that
bonds with lower credit risk dominate the announcement effects.
Panel E, F, G, and H of Table 4 present mean of three-day CASs, t-statistics, and z-

statistics in the stock market for each downgrade (upgrade) based on rating review and rating
changes dates. The overall mean CAS is 3.9576 (Panel G) for the possible downgrade and –
1.2010 (Panel H) for the possible upgrade around the rating review dates. Compared with the
bond market, the magnitude of test results in the stock market is much greater. An asymmetric
reaction of CAS is also detected in stocks as well as bonds. We discover that stocks with lower
credit ratings have larger variations of liquidities, notwithstanding some mixed results. For
example, C grade stocks have the largest mean CAS (7.7227) for possible downgrade watchlists
(see Panel G). Stocks with the second largest mean CAS for possible downgrade watchlists are in
the grade A group (5.3923), following by stocks graded from B3 to Ba1 (3.9161) and Baa3 to
Baa1 (2.3669).We interpret these results as evidence that liquidity improves for the possible

18