Influence of Rating Announcements and Their Characteristics on
Abnormal Liquidity in Corporate Debt Market
Pilar Abad a,**, Antonio Díaz b and M. Dolores Robles c
a
Universitat de Barcelona, Diagonal 690, 08034 Barcelona, Spain
Universidad de Castilla-La Mancha, Plaza de la Universidad 1, 02071 Albacete, Spain
c
Universidad Complutense de Madrid, Campus de Somosaguas, 28223 Pozuelo de Alarcón, Madrid, Spain
b
Abstract:
The influence of rating announcements on corporate debt market liquidity has been ignored for a
long time. Based on an event study, this article examines the effects of the announcements of
actual rating changes, outlook notices, and CreditWatch placements provided by Moody’s,
Standard and Poor’s and Fitch on abnormal liquidity in the Spanish corporate debt market. Also,
by means of cross-section regressions, we establish what factors determine the sign and intensity
of the liquidity reactions. The presented results indicate that factors related to the characteristics
of the rating announcement, the issuing company and the economic environment are relevant in
light of several hypotheses.
Keywords: Rating agencies, rating changes, liquidity
JEL Classification: G12, G14, C34.
* The information provided by AIAF, Fitch and Moody’s is appreciated. Any errors are solely the responsibility of the authors. This work has
been funded by the Ministry of Science and Technology (SEJ2005-03753 and SEJ2006-14354), the Castilla-La Mancha Council of Communities
(FEDER PAI-05-074) and the Region of Madrid (UCM940063).
** Departament d' Econometria, Estadística i Economia Espanyola. Facultat de Ciències Econòmiques i Empresarials. Universitat de Barcelona.
Diagonal, 690. 08034, Barcelona. Tel: +34 93 4035732.
E.mail addresses: pabad@ub.edu (P. Abad), antonio.diaz@uclm.es (A. Díaz), and mdrobles@ccee.ucm.es (M.D. Robles).
Influence of Rating Announcements and Their Characteristics on
Abnormal Liquidity in Corporate Debt Market
Abstract:
The influence of rating announcements on corporate debt market liquidity has been ignored for a
long time. Based on an event study, this article examines the effects of the announcements of
actual rating changes, outlook notices, and CreditWatch placements provided by Moody’s,
Standard and Poor’s and Fitch on abnormal liquidity in the Spanish corporate debt market. Also,
by means of cross-section regressions, we establish what factors determine the sign and intensity
of the liquidity reactions. The presented results indicate that factors related to the characteristics
of the rating announcement, the issuing company and the economic environment are relevant in
light of several hypotheses.
Keywords: Rating agencies, rating changes, liquidity
JEL Classification: G12, G14, C34.
1. Introduction
In this study, we examine the impact on the liquidity of the Spanish corporate debt market
of the announcements related to corporate debt rating: actual rating changes, outlook notices (or
medium-term rating trends), and CreditWatch placement (warnings of a possible short-term
rating change), made by the three largest international agencies: Moody’s, Standard and Poor’s
and Fitch. We also identify the determinants of abnormal liquidity by considering the
peculiarities of the change in question, the issuer and the economic environment.
Many authors present evidence of the informative content of rating announcements. Most
of them have focused on analyzing the effects of those announced changes on the stock prices
(e.g., Hand et al., 1992; Elayan et al., 2001 or, for the Spanish market, Abad & Robles, 2006).
Some others analyzed these effects on corporate debt prices (e.g., Kliger & Sarig, 2000, Steiner
& Heinke, 2001; in the European case, Gropp & Richards, 2001, Dallocchio et al., 2006; or, in
the Spanish case, Abad et al., 2007). In the above mentioned sources, we can find hypotheses
regarding the effects of rating change announcements that postulate the expected performance of
stock and corporate debt prices, as well as possible determinant factors. However, none of them
addresses the expected liquidity performance.
Ratings and rating changes can result in a specific market dynamic that could not only
affect prices, but could also directly influence the market liquidity. This dynamic is probably
caused by the way in which the investors use the rating, as well as by its actual informative
content. For example, the proliferation of “rating triggers” in the management of portfolios based
on rating changes could force operators to increase their sales transactions, and could even cause
a liquidity crisis.
In spite of the importance of the impact of rating actions on debt liquidity, there is
practically no theory or empirical research devoted to that question. To our knowledge, the first
work on that topic is the paper by Abad et al. (2007), where different liquidity measures in the
Spanish corporate debt market, and the effects of rating announcements on returns and yield
spreads were analyzed.
Based on the work of Abad et al. (2007), in this study we analyze how liquidity responds
to rating change announcements. We also formulate different hypotheses that link the potential
effect to different characteristics of the issue, such as sector, size, etc.; rating change
characteristics, such as the type of rating action in question, or if that action is expected by the
1
market, etc.; and economic environment characteristics. Finally, we define the possible
explanatory factors of the liquidity response under these hypotheses.
To carry out the analysis, we define a wide range of abnormal liquidity measures based on
trading volumes, trading frequency and market share. We analyze a sample of daily corporate
bond and commercial paper notes data from 1993 to 2004. This database of Spanish corporate
fixed income assets contains information about the trading volume per transaction, making it
possible to develop trading activity measures. First, we perform an event study to determine if the
rating changes generate significant abnormal liquidity, and then we analyze the effects of the
determinants by mean of a cross-section regression analysis.
The next section describes the evolution of the Spanish corporate debt market liquidity,
categorizing it by market segments and comparing it to the evolution of the public debt market.
Section 3 formulates the hypotheses addressing the reaction of liquidity to rating announcements.
Section 4 presents the liquidity measures analyzed. Section 5 describes the database and shows
the results of the empirical analysis. The main conclusions are summarized in section 6.
2. Liquidity and Trading on the Spanish Corporate Debt Market
This section introduces the liquidity of the leading and almost only Spanish corporate debt
market, AIAF (AIAF Mercado de Renta Fija - Fixed Income Market),1 describing its evolution
across the sample of selected years. At present, AIAF is one of the leading European corporate
fixed income markets. In fact, in 2006, it was the second largest European market in mortgagebacked securities2 and the first largest in covered bonds (“Cédulas Hipotecarias”).3
The current situation in Spanish market is the result of a rapid evolution that began at the
end of the last century. Until that moment, the primary market was weak and the secondary
market was very narrow and shallow. Fixed income financing suffered from the strong
competition of bank financing – with similar interest rates but without the loan issue costs,
excessive bureaucracy, and even fiscal obstacles – and the fierce competition of the public sector
1
Corporate fixed income assets are also listed through the electronic trading system of the Spanish stock exchanges
and through three of the four Spanish stock exchange markets (Barcelona, Bilbao and Valencia).
2
3
ESF Securitisation Data Report, Spring 2007.
Spanish mortgage certificates or Cédulas Hipotecarias are equivalent to German Pfandbriefe and French
Obligations Fonciéres.
2
in placing the issues. As for the secondary market, the public debt, with enormous outstanding
balances, attractive returns, high liquidity, wide variety of terms and better taxation, together with
a scant interest in national mutual funds for corporate fixed income, and the competition between
different trading environments with different regulations, were all factors that determined the lack
of liquidity.
Fortunately, the reforms that accompanied the launch of the Euro affected the Spanish
corporate fixed income market. First of all, measures intended to simplify transactions and reduce
the issue and admission costs were adopted.4 Second, the clearing and settlement systems were
harmonized, and AIAF was integrated into the Mercados Financieros holding (MF), together
with MEFF and SENAF,5 and later on, into Bolsa y Mercados Españoles (BME).6 Third, the
European Central Bank allowed the use of private high quality issues in monetary policy
operations, which gave a considerable push to simultaneous trading operations in AIAF, to obtain
liquidity. Fourth, the fiscal legislation enacted in 1999, for the first time levied the same tax on
returns on public and private fixed income in the corporate income tax.
In addition to these reforms, other factors have contributed to a greater interest in
corporate fixed income. First, there was a progressive reduction of the public deficit required by
the Maastricht treaty, which assumes a lowered pressure of Treasury debt on the market, and
then, in more recent years, there was a boom in securitisation.7 With the development of
securitisation bonds and the extension of future loan rights in 1998, it is possible to endow
financial entities with liquidity by releasing capital and providing securities to investment funds.
4
These measures were also introduced in the rest of the euro zone countries. Thus, Santos & Tsatsaronis (2002)
observe a significant reduction in investment commissions and assurance of new bond issues, whereas Melnik &
Nissim (2005) observe the same for the rest of issue costs.
5
MEFF is the Spanish market in financial futures and options and SENAF is the main electronic trading platform for
Spanish public debt.
6
Bolsas y Mercados Españoles (BME) is the company that integrates all the securities markets and financial systems
in Spain. The parent group comprises the Madrid, Barcelona, Bilbao and Valencia stock exchanges, MF Mercados
Financieros, Iberclear and BME Consulting.
7
Mortgage loans generate fixed assets for which a solvency rate of 8% should be maintained. Securitisation permits
the transference of these assets to a fund, which for acquisition thereof issues bonds backed by collection rights.
With the operation, the financial entity transforms its fixed assets into liquidity, which moreover is exempt from
the solvency ratio obligation.
3
The influence of all these factors can be seen in Tables 1 and 2, which show the
outstanding balances and the trading volumes of the AIAF market, together with the Spanish
public debt market and the fixed income stock market. Since 1997, there has been less issuing
pressure by the State in the case of Treasury bills – the outstanding balance in 2006 was not even
half of that in 1997 – and there was a growth in the government bond segment of 75% during that
period. In comparison, the AIAF amount outstanding in 2006 was sixteen times higher than in
1997, with spectacular growth figures in commercial paper notes, mortgage cedulas and
securitisation bonds.
[Insert Table 1]
The evolution of transacted volumes was similar in the case of outstanding balances
(Table 2). Since 1997, trading in the short-term segment has increased by 10664.0%, and in the
medium- and long-term segment, excluding securitization, by 840.8%, whereas the increase in
the Spanish public debt market was 72.6%.
[Insert Table 2]
In any event, Table 2 reveals a traditional problem of corporate fixed income, i.e. the lack
of liquidity. Except in the commercial paper segment, a certain weakness of the secondary market
is observed if the trading volume is accounted for. Thus, whereas the accumulated trading
volume of Government bonds throughout 2006 was more than ten times the figure of their
amount outstanding at the end of the year, the trading of bonds, notes and cedulas in AIAF did
not reach 59% of their outstanding balance in the same year. Of this figure, only 39.4% were
operations to maturity.
Another factor to be considered is the weight of retailers in trading operations. These
transactions account for 55% of the total number of operations to maturity undertaken in AIAF,
although their percentage is barely 1% over the market total, and 2% over that transacted to
maturity.8
3. Hypotheses Addressing the Reaction of Liquidity to Rating Announcements
8
Trading to remaining maturity is distributed among companies (11%, primarily in the commercial paper segment),
financial institutions (18%, focused on the medium- and long-term segment), insurance and collective investment
institutions (20%, they accrue 48% of commercial paper trading), non-resident (43%), public sector (1%), and
market members (5%). Source: “AIAF Market activity 2005” annual report.
4
In the literature there are different theories regarding the expected effects of rating
changes on prices in both the stock market and the corporate debt market. Nevertheless, these
hypotheses do not specify the expected impact on liquidity. One of the main theories is based on
the “information content hypothesis”. According to that hypothesis, the rating agencies handle
confidential information on the rated companies, and therefore rating revisions include new
information for the market. This new information is rapidly incorporated into prices.
Our hypothesis states that this incorporation of new information into prices is
accompanied by increased market activity. Therefore, increased activity can be observed around
the date of the rating announcement and, the greater the jump in categories indicated by the rating
change, the greater the expected reaction. According to this hypothesis, if the changes do not
include new information, i.e. they are anticipated by the market, no effect on prices or on
liquidity is expected.
In the corporate debt market, the earliest works that studied the effects on prices
postulated by this hypothesis concluded that rating announcements had no informational contents
(Weinstein, 1974; Wakeman, 1978; Zaima & McCarthy, 1988), whereas more recent studies
presented evidence in favor of this hypothesis (Ingram et al., 1983; Hand et al., 1992).
We could also consider the informative content of rating changes starting with market
microstructure models. According to these models, the response of the trading activity after new
information arrives is related to the existence of asymmetric information among the informed and
uninformed agents and market makers (e.g., see Balduzzi et al., 2001). Informed investors can
anticipate such information by increasing their activity before the announcement, trying to
manipulate uninformed investors. After the news is released, the volatility temporarily increases
as the new information is included into prices, and after that point, the trading activity decreases.
Thus, the final result observed after rating actions is a lower level of market activity.
As Howe (1995) indicates, a standard criticism of rating agencies is that they have a
medium-long term outlook and thus try to give stability to their ratings, by avoiding changes in
the rating due to short-term fluctuations. Löffler (2004) and Altman & Rijken (2006) indicated
that ratings are the result of using a “through-the-cycle” methodology that allows them to obtain
stability in the ratings and disregard short-term fluctuations in default risk. In similar terms,
Fledelius et al. (2004) and Löffler (2005) asserted that when the rating of a company is changed,
the agency wants to be quite sure that this adjustment is stable and is not going to be reversed
5
shortly after (“reversal aversion”). They observe that rating changes are statistically
interdependent and predictable using borrower fundamentals. This predictability is based on
rating bounce avoidance.
Rating stability is desirable to prevent procyclical effects. For example, an immediate
rating downgrade as a response to changes in current creditworthiness could induce investors to
liquidate their positions hurriedly, which could ultimately result in a credit crunch. Another
argument for rating stability is that it maintains the reputation of the agencies. Rating reversals
within a short period have a negative impact on an agency's reputation, even when the reversals
reflect true changes in creditworthiness.
The relative slowness of the rating agencies actions could lead investors, who are more
interested in knowing the real credit risk position at all times, to seek the information they need
from other sources. In this respect, rating changes are discounted prior to the announcement date
and, if this is the case, then they do not cause any increase at all in market activity.
In addition to rating changes the agencies make other rating-related decisions, e.g. outlook
notices or rating reviews (CreditWatch placement). These two processes differ from each other.
The placing of a security on the CreditWatch list occurs on special occasions such as changes in
regulation, unexpected changes in management, merger announcements, etc., and it indicates that
the security’s rating is under review and that the future change in its rating is likely to occur in a
short period of time. Outlooks indicate the credit worthiness trend in a medium-term time frame
and therefore can be considered as a refinement of the assigned rating. Altman & Rijken (2007)
found that bond ratings provide a better prediction of the default risk when they are adjusted for
outlooks. In this respect, announcements in outlook reports or of placement on the CreditWatch
list can provide useful information to the market and could cause more activity after that
announcement than would a rating change announcement. For instance, Steiner & Heinke (2001)
and Hull et. al. (2004) confirmed the impact on prices of the placement on the CreditWatch list.
Another hypothesis mentioned in the literature is that agency behavior could reflect a
“moral hazard risk problem” (Steiner & Heinke, 2001) that would undermine the reliability of
their ratings. As indicated by Steiner & Heinke, agencies may systematically overrate the issuers
in order to achieve a greater market share or maintain leadership. Covitz & Harrison (2003)
investigated whether credit rating agencies are biased and favor issuer interests at the expense of
investor interests. Issuers pay for ratings and also choose the rating agency to produce the ratings.
6
Almost all rating agency revenue comes from rating fees. One mechanism for acting in the
interest of issuers is to delay rating downgrades. Covitz & Harrison observed that the bond
market anticipates rating changes, but they find no evidence consistent with rating agencies
acting in the interests of issuers. Boot et al. (2006) provide a theoretical model based on an
implicit contract between the credit rating agency and the firm that should prevent further
downgrades. A rating agency initiates a monitoring regime through the credit watch procedure
and an issuer implicitly promises to undertake specific actions to mitigate the possible
deterioration of its rating.
“Moral hazard risk problem” could result from the inferior quality of ratings, diminishing
the reliability of rating agencies on the market. In that case, the rating changes would have no
effect because the market would not value them. However, competition between agencies could
help to mitigate this “problem,” by making the ratings assigned by different agencies equally
reliable. As for the price reaction, Holthausen & Leftwich (1986), Hite & Warga (1997) and
Steiner & Heinke (2001) find no differences in stock prices in response to changes announced by
Standard & Poor’s and Moody’s, whereas Kish et al. (1999) indicate that Moody’s is less
credible than Standard & Poor’s.
On the other hand, credit ratings allow us to make a distinction between investment grade
debt and speculative grade debt. This distinction translates into restrictions of the investment
decisions made by certain market participants because they lead to “rating triggers” that force
people to get rid of speculative grade bonds. For example, pension fund guidelines often stipulate
that investments are only allowed in investment grade issues. Similarly, specific markets, such as
the Eurobond market, may simply require the presence of a particular minimum rating before
listing the debt issue. These restrictions effectively condition the investors’ decisions on the
observed rating and influence the effects caused by rating changes, even though the changes
contain no new information for the market. Thus downgrades that result in a change to
speculative grade would force sales of bonds by institutional investors and increase the market
activity, whereas upgrades do not necessarily imply more activity. In this respect, a significant
percentage of investors that trade on the AIAF are collective investment institutions, many of
them foreign,9 which may be subject to this type of contingency clause. The empirical evidence
of an impact on prices is inconclusive. Holthausen & Leftwich (1986), Cornell et al. (1989),
9
Primarily German and French investment and pension funds.
7
Hand et al. (1992), Schweitzer et al. (1992) and Matolcsy & Lianto (1995) found no significant
influence of the change to speculative grade on stock prices, whereas Wansley et al. (1992) and
Hite & Warga (1997) did find a significant impact.
Nevertheless, institutional investors want to keep their portfolio rebalancing as low as
possible and usually use a buy-and-hold strategy. This fact could diminish the impact of rating
changes on liquidity, even if the grade changes from investment to speculative. Moreover, it is
usually easy for institutional investors to obtain information on the companies in their portfolio
and, therefore, the rating changes have a limited influence on their investment decisions.
Many authors, e.g. Steiner & Heinke (2001) and Hull et al. (2004), point out significant
differences in the reactions of prices to upgrades and downgrades of creditworthiness. Regarding
liquidity, the response may differ in the two cases. The agencies may prefer to proceed slowly so
as not to make mistakes and to safeguard their reputation. However, the loss of reputation
associated with giving a good rating to a high-risk company is more serious than that resulting
from assigning a poor rating to a low-risk company, since the first error could mean economic
losses for investors. According to Holthausen & Leftwich (1986) and Ederington & Goh (1998),
this asymmetry means that the agencies allocate more resources to revealing negative than
positive information. In addition, the pressure on prices following downgrades is greater than in
the case of upgrades because the former will result in sales transactions on the market, whereas
the latter would not necessarily increase the number of purchase transactions. Therefore, a
hypothesis can be formulated stating that the impact on liquidity will be greater in the case of
downgrades than in upgrades.
Another hypothesis to be examined is related to the possible information differences of
rating, associated with differences in the regulations that affect different companies (Schweitzer
et al., 1992). For example, financial firms are subject to greater control by regulators and there
may be more market information available about them than about firms from other sectors. Thus
the impact of rating announcements on that type of firm is diminished. The same is true for
public enterprises. Since the debt of these enterprises is guaranteed by the State, even in the event
of a downgrade the market does not interpret this information, in terms of the likelihood of
default, in the same way as it would have done for corporations.
The characteristics of the company may also modify the impact of rating change
announcements on liquidity. For instance, downgrades that affect larger companies may be
8
appraised differently than downgrades of smaller companies, due to the too-big-to-fail paradigm.
The impact on liquidity may also differ depending on the company’s profitability or growth rate,
because rating actions cause smaller increases in transactions in the case of a highly profitable
company than a less profitable one. Gropp & Richards (2001) analyze the impact of rating
changes of banks in major European financial markets observing significant influence on stock
prices but not on bond prices. They try to explain this insensitivity by the relevant size of
companies included in the sample, and the low probability of default in the European banking
sector.
The economic environment is also relevant according to some authors. For example,
Dialynas & Edington (1992), suggest that, in periods of prosperity, investors may be less
concerned about security and will assume higher levels of risk. Thus, the next hypothesis would
state that the effects of rating announcements differ according to the current phase of the
economic cycle.
Finally, the peculiarities of the Spanish market and the analyzed data sample should be
taken into consideration. We have a small number of issuers and in general, we find infrequent
trading in many issues, especially in the case of bonds. However, the case of commercial paper
notes is different because they constitute a very active segment of the market.
4. Liquidity Measures
Liquidity is a hard concept to define. It refers to the ease with which an asset can be
traded in a short period of time without causing a significant impact on the price. Liquidity
affects the valuation since it is associated with transaction costs and, according to Lo et al.
(2004), these costs imply infrequent trading. Since investors cannot regularly cover their risk,
they demand an ex ante risk premium that results in lower prices of these assets. Thus, for
identical promised cash flows, the illiquid bonds will be traded less frequently at lower prices and
with higher yield spreads. Liquidity premium is the difference between the return required of a
liquid security and the one corresponding to an illiquid security.
The literature has analyzed different aspects of liquidity, almost exclusively for the U.S.
government debt market, based on a fairly high number of variables, some of which are relative
to the market and others, to specific characteristics of the assets such as bond liquidity proxies.
Most of this literature tries to determine whether the liquidity differences translate into liquidity
9
premiums. These analyses compare similar pairs of Treasury securities or real securities to
synthetic ones.10 There are fewer studies that analyze corporate debt, although recently several
papers on that topic have been published. For instance, see Houweling et al. (2005), Ericsson &
Renault (2006), and Chen et al. (2007).11
In terms of specific asset characteristics, Fisher (1959) uses the size of the issue as a
determinant of liquidity, given the existing relationship between the amount of outstanding bonds
and the ease of trading those bonds. Sarig & Warga (1989) and Amihud & Mendelson (1991)
suggest that the youngest bonds are traded more frequently. Therefore, investors demand higher
premiums for the older bonds, given the difficulty of converting them into cash before their
maturity. The latter remain in inactive portfolios of investors that match their liabilities with these
assets. Moreover, according to Kamara (1994), the inventories of market makers tend to diminish
with the age of the bond. Warga (1992), on the other hand, uses the status as an explanatory
variable of liquidity, classifying the references as “on-the-run” or “off-the-run”.12
However, a market-related variable – the bid-ask spread – is the most commonly used
variable to measure liquidity costs. Although there are many works that defend its virtues, a key
factor to be taken into account is that most databases for the U.S. market have this information
but they usually do not have data on transaction prices and trading volumes.
In this respect, Elton & Green (1998) and Alexander et al. (2000) suggest that the best
liquidity proxy is the trading volume. Fleming (2001) finds that the best result is provided by the
number of transactions, and Longstaff et al. (2005) stress the role of the amount outstanding.13
Houweling et al. (2005) compare nine liquidity proxies (issued amount, listed, euro, on-the-run,
10
For example, see Amihud & Mendelson (1991), Kamara (1994), Fleming (2001), Díaz & Navarro (2002a),
Strebulaev (2002), Alonso et al. (2004), Cherian et al. (2004), Babbel et al. (2004), Jordan & Kuipers (2005), and
Díaz et al. (2006).
11
Previous works include those of Gehr & Martell (1992), Shulman et al. (1993), Chakravarty & Sarkar (1999),
Alexander et al. (2000), Hong & Warga (2000), Collin-Dufresne et al. (2001), Díaz & Skinner (2001), Elton et al.
(2001), Schultz (2001), and Díaz & Navarro (2002b).
12
The “on-the-run” reference is the one most recently issued of all those that have the same issue term and, therefore,
concentrates the market trading activity. The issue system by tranches of Spanish bonds, according to Díaz et al.
(2006), means that their life cycle corresponds to three categories: pre-benchmark, benchmark and seasoned.
13
Other empirical studies use different explanatory variables such as interest rate volatility (Kamara, 1994), trading
frequency (Shulman et al., 1993), growth of investment funds (Fridson & Jónson, 1995), or the daily number and
dispersion of operations (Houweling et al., 2002).
10
age, missing prices, yield volatility, number of contributors and yield dispersion) for a sample of
Eurobonds and find very few differences between the different measures. Lesmond (2005) uses a
measure of the impact on prices proposed in the field of equity markets by Amihud (2002),
together with a variable estimated on the basis of the model of Lesmond et al. (1999). The latter
is based on the appearance of null returns, equivalent to the missing prices or “runs” of Sarig &
Warga, 1989.14 Lesmond (2005), as well we Chen et al. (2007), obtain better results using the
variable of Lesmond et al. (1999), as that variable is a substitute for the bid-ask spread.
Concerning the spread, they remark that it “is not always available for all bonds or for all time
periods. This is especially true for thinly traded bonds or more mature bonds”. For the Spanish
government debt market, Díaz et al. (2006) propose the current market share and the share
expected in the future as liquidity measures. Estimation of the expected future share requires that
the life cycle of the asset trading activity be very regular.
In spite of this wide range of liquidity proxies, many of the measures mentioned above do
not fit the analysis aiming at the examination of changes in an asset’s liquidity in response to a
rating change announcement. Measures based on asset characteristics cannot be applied because,
for example, the evolution of the bond age is completely independent from the issue rating. Even
though measures based on the performance of the issue on the market could appear more
appropriate for our analysis, there is no information available on many of them, including the
bid-ask spread.
Our study uses measures of the trading activity as liquidity proxies. Specifically, we
analyze the evolution of the trading volume and the frequency of trading. In addition, since the
life cycle of commercial paper notes is extremely regular, we include the expected market shared
proposed by Díaz et al. (2006) in the study of these assets. Based on these three proxies, we
obtain the measures of abnormal liquidity caused by the event.
The trading volume on day t2 – the first trading day after the rating announcement – is
obtained on the basis of a measure of the effective trading volume for each of the outstanding
14
The Amihud measure is the ratio between the daily return in absolute value and the trading volume corrected by the
number of trading days in the last year. Just as other trading-based measures, this requires the existence of at least
one transaction. In the case of the Lesmond et al. (1999) model, its estimation requires that a return generator
model be used, and that daily prices be available for a long period of time and with at least 20% of non-null
returns.
11
issues of issuer i. A calculation is then made of the logarithmic rate of change of the trading
volume between session t2 and the last session prior to t2 during which an asset of the issuer was
traded, t1. Specifically:
cVi ,(t2 −t1 ) = vi ,t2 − vi ,t1
(1)
where vi,t1 and vi,t2 are the measures of the logarithm of the trading volume of the outstanding
issues of issuer i on days t1 and t2, respectively.
The abnormal trading volume variable AVi ,(t2 −t1 ) is obtained by comparing the observed
rate of change cVi ,(t2 −t1 ) which would be expected in the absence of the event. The latter,
considered as benchmark, shows the “normal” or typical volume of the issuer:
AVi ,(t2 −t1 ) = cVi ,(t2 −t1 ) − E (cVi ,(t2 −t1 ) )
(2)
where E(cVi,(t2 −t1 ) ) is the expected or “normal” rate of change of the trading volume on average
between all the issues of issuer i.
Three alternative forms have been used to calculate the benchmark variable. First of all, it
is calculated as the logarithmic rate of change between t2 and t1 of the mean traded volume per
day of each issue in the last three months (MDV). The mean traded volume per day refers to the
average total traded volume for all the outstanding assets of the issuer in the last three months,
divided by the number of days on which each asset is traded during that period. The second
alternative is the logarithmic rate of change of the mean daily trading volume accumulated in the
last three months (MDTVA), and the third is the logarithmic rate of change of the mean daily
trading volume (MDTV). The latter measure is obtained by dividing the total trading volume by
the number of working days in the last three months, regardless of whether or not the issue has
been traded. Consequently, this latter measure is corrected if the asset is kept outstanding for less
than three months.
The second abnormal liquidity measure is based on comparing the mean trading
frequency to the different outstanding issues by the issuer on the first day of transactions after the
event and the day prior to the event. These frequencies are calculated for each issue as the ratio
between the number of trading days and the number of days on which trading could have taken
place, i.e., working days in a predefined window. Specifically, the measure of abnormal
frequency is calculated as:
AFi ,(t2 −t1 ) = f i ,t2 − f i ,t1
12
(3)
where fi ,t2 and fi ,t1 are the logarithms of the two relative frequencies mentioned above. We
consider different sizes of the windows in which fi ,t2 and fi ,t1 are calculated.
For the first
windows, in which we measure the frequency after the event, we select 1- and 2-week and 1- and
2-month windows. For the second windows, in which we measure the frequency before the event,
we consider 1-, 2- and 3-month windows, and the entire period since issue. In this way, we are
able to see what happens to abnormal liquidity as the date of the rating change announcement
approaches and, furthermore, if the impact is only observed in the market immediately after the
event or it is more long-lasting.
As remarked previously, the trading activity performance of commercial paper notes over
their short life cycle is very regular. In this respect, the issuing activity of new commercial paper
notes by large corporations is constant. Institutional investors trade these instruments very
actively after issue. After a few days, trading ceases almost completely. The regularity in the life
cycle is also observed by Díaz et al. (2006) for Spanish government bonds.15 Although the
evolution of market share of both groups of assets differs widely, it follows a uniform pattern in
both cases. For government bonds, it presents an initial spike with a subsequent exponential drop.
For commercial paper, it shows a sharp initial drop and a much smoother exponential decline
afterwards. This performance results from the fact that institutional investors take up positions on
the recently issued commercial paper note and subsequently its trading loses appeal and it
becomes residual. Thus, the average market share of a commercial paper note in its first week of
life is 33.4% and it drops to 2.7% the following week.
The model proposed by Diaz et al. (2006) is used to study the behavior of the weekly
market share of each issue as a smooth, non-linear function of its age.16 They prefer the market
share over the trading volume per issue because the former variable is expressed in relative terms
with respect to the total volume traded on the market during the week and, therefore, it eliminates
possible data trends and possible volume fluctuations between weeks without relevance for asset
15
These assets are issued in successive tranches until they reach a certain amount outstanding. Thus, the amount
outstanding of the recently issued bond is reduced in comparison to that of its predecessor and its liquidity is lower.
The market share of this asset will continue to grow up to a maximum amount, after which a rapid decline will
begin as a result of the appearance of a new issue that replaces it as the market benchmark. Trading of this
seasoned bond is practically residual.
16
This model is inspired by actuarial methods used to model human mortality (see Heligman & Pollard, 1980).
13
liquidity. They define the market share of asset i during week t as the ratio between the nominal
volume traded per asset and the total volume traded per all the outstanding issues. The market
share permits to compare the different degrees of liquidity between issues and to monitor the
evolution of the liquidity of an issue throughout its lifetime.
In our case, the original expression of Díaz et al. (2006) is adapted to the performance of
the commercial paper market share. Thus, after adjusting several variants of the original equation:
MS it = β 1 exp( Ageit − β 2 ) 3 + β 4 ⋅ β 5
β
Ageit
+ u it
(4)
we see that the best adjustment is attained with the following expression:
MS it = β1
( Ageit − β 2 )β 3
+ β4 ⋅ β5
Ageit
+ u it
(5)
where β i , i = 1,...,5 are the parameters to be estimated, and uit is an error term i.i.d. with zero
mean and constant variance.
Equation (5) is estimated from the weekly market share of each commercial paper note
traded in the sampling period of 1998 to 2004.17 We considered all the outstanding issues for
each day of the week, regardless of whether or not they were traded. In other words, when a
commercial paper note was not traded during a session, its market share was zero and it was
taken into account to calculate the mean market share of all the commercial paper notes of the
same age.
Figure 1 shows the result of the estimation.18 These estimated market shares are the ones
we use as the benchmark market shares that a commercial paper note should achieve as a
function of its age and regardless of whether a rating event occurs. Thus, we define abnormal
market share as the difference between the rate of market share change observed around the
rating event and that expected of a typical commercial paper note of the same age:
AMSi ,( t2 −t1 ) = cMSi ,(t2 −t1 ) − E (cMSi ,( t2 −t1 ) )
(6)
where cMSi ,(t2 −t1 ) = msi ,t2 − msi ,t1 , msi ,τ is the market share in logarithms during week τ = t1 , t2 ,
and E(.) indicates the expected value according to model (5).
[Insert Figure 1]
17
The model was also estimated on the basis of effective volume as a dependent variable. The results are similar to
those obtained using the market share.
18
All the details have been omitted to save space, but they are available on request form the authors.
14
5. Empirical Results
5.1. Sample Data and Descriptive Analysis
The sample analyzed in this paper consisted of rating action announcements of Fitch,
Standard and Poor’s and Moody’s from June 1993 to December 2004. Part of this information
was provided by Fitch and Moody’s. The “Hemeroteca de El País” [“El País” newspaper library]
was also used to obtain information on the announcements of Standard and Poor’s. The original
sample was composed of 349 rating announcements, including rating changes, outlook changes,
and CreditWatch placement.
To compute the abnormal liquidity measures we used daily observations of trading
volume and transactions carried out on all commercial paper notes and bonds traded on the
secondary corporate debt market, AIAF. The database of bonds begins in 1993, whereas the
commercial paper database begins in 1998, and both end in 2004. For each reference, AIAF
provides daily information on the number of transactions and the nominal and effective
transaction volumes. We excluded from the sample issues with special characteristics, such as
floating interest rate issues, convertible bonds, issues with tax incentives, etc.
From this sample, the issues of re-rated companies have been selected, and the cases that
lacked the minimum of liquidity around the announcement date have been excluded. The final
sample consisted of 158 rating action announcements that affected 1058 issues (271 bonds and
787 commercial paper notes).19 Table 3 shows the 158 events divided into six categories: rating
upgrades or downgrades, positive or negative outlooks, and CreditWatch placement for negative
or positive reasons. In all, the sample contained 109 rating announcements that affected the bond
market and 120 that affected the short-term market. Of these, 71 simultaneously affected both
markets.
[Insert Table 3]
Table 3 also shows the number of expected rating announcements. So as other authors, we
used the CreditWatch placement to distinguish between expected and unexpected rating changes.
When a rating announcement was preceded by a placing of a security on the CreditWatch list in
the same direction, it could be anticipated by the market and would not provide new information.
19
In many cases, the rating changes affect companies whose issues are not traded around the event. Other issues are
not traded in the secondary market because they are fully incorporated into the investors’ portfolios. Moreover,
some large issuers put their debt into circulation on other international markets.
15
In both segments of the market, we found that more than 50% of the events were expected.
Finally, of the 229 announcements, 143 involved a deterioration of creditworthiness and 86
involved improvements, which seemed to indicate a somewhat increased credit risk in the
Spanish corporate debt market during the period under study.
Table 4 shows the classification of rating announcements according to the number of
notches the debt shifts after the rating change. Although the three agencies used different symbols
to designate the different credit risk categories, equivalence between these symbols could be
determined. This allowed us to transform the ordinal scale applied by the agencies into a numeric
scale in which the highest values denoted greater probability of default.20 Table 4 also shows in
parentheses those changes that imply an entrance into speculative grade. As we can see, only two
rating downgrades in the case of bonds caused a drop to the speculative grade.
[Insert Table 4]
Table 5 shows the distribution of rating announcements according to the agency. Forty
nine percent of the rating actions were by Moody’s, 27% by Fitch, and the remaining 23% by
Standard and Poor’s. In addition, the percentages were similar when we considered the two
market segments separately.
[Insert Table 5]
Finally, Table 6 presents the classification of the issuers, distinguishing between issuers
in the bond market, issuers in the commercial paper market, and those that issue in both markets.
The issuers were also classified as belonging to the financial sector or to another economic
sector, and also according to their public or private ownership. In 68% of the rating
announcements the issuer was a savings bank or a commercial bank, whereas only 18% of the
announcements were related to public enterprises.
[Insert Table 6]
5.2. Estimation of Abnormal Liquidity
Almost all issuers affected by rating actions simultaneously maintain various issues on the
market, especially in the case of commercial paper notes. On most days, several references of
each issuer are traded on the secondary market. To avoid correlation in the cross section resulting
from the fact that the trading activity of the references issued by the same company may be
20
Rating Aaa of Moody’s and AAA of Fitch and S&P corresponds to 1 on the numeric scale, rating Aa1 and AA+ to
2, rating Aa2 and AA to 3, and so on.
16
highly correlated, we constructed portfolios with all the bonds on the one hand and all the
commercial paper notes on the other, before computing the liquidity measures. In this way, all the
outstanding references of each issuer were aggregated and weighted by the volume of issues
traded on the corresponding day in a portfolio, which was treated as an individual observation.
In the event analysis, we used two statistics to test the null hypothesis of inexistence of
abnormal performance due to the rating action announcement, i.e., zero mean abnormal liquidity:
a standard t-ratio and, to avoid the effects of non-normality, the Wilcoxon rank test.
The results are shown in Tables 7 and 8. The first one shows the mean abnormal liquidity
after upgrades and downgrades in the corporate bond portfolios. The results for the commercial
paper portfolios are shown in Table 8. Both tables show the results for the two proxies of
abnormal liquidity: the trading volume-based and the trading frequency-based measures.
[Insert Table 7]
As shown in the left panel of Table 7, the downgrade announcements imply a
significantly positive abnormal liquidity when trading frequency is used as liquidity proxy,
whereas no excess significant liquidity is observed when trading volume is used. It is interesting
to note that increased liquidity after the rating downgrade announcement is only observed in the
measures based on the shortest, less than one month, post-event windows. In addition, liquidity
increases more as the date of the announcement approaches, i.e., it is more clearly observed for
the shortest pre-event windows. The results are independent of the test used.
In the case of rating upgrade announcements (right panel, Table 7), again the excess
liquidity is significantly positive, and this result is robust for the way of calculating liquidity
(volume or frequency) and the test used (parametric or non-parametric). Just as in the case of
downgrades, for the frequency measures the effect is greater immediately after the announcement
date and it diminishes after that point.
Results for the short-term corporate debt market are shown in Table 8. In this case, the
market share is also analyzed as liquidity proxy. There are no effects on liquidity after the
downgrade announcements (Table 8, left panel). The mean abnormal liquidity is not significant
when it is measured via trading volume but a significant drop of the mean market share occurs,
whereas with the frequency measures a significant positive effect on liquidity is observed, except
in two of the fifteen analyzed windows. These apparently contradictory results should not come
as a surprise in view of the differences between the measures. While in the case of market share
17
the performance of this variable is compared on two specific days (before and after the event), in
the case of frequency the comparison concerns what happens in windows around the event.
Consequently, this result indicates that the abnormal market share decreases immediately after
the downgrade announcement, and at the same time the abnormal trading frequency increases in
windows subsequent to the announcement (the abnormal frequency being higher in the narrowest
windows).
[Insert Table 8]
In the case of effects on commercial paper notes liquidity after rating upgrades (right
panel, Table 8), significant effects are detected only with the frequency-based measure. The
effect is positive in all the windows, indicating an increase in abnormal frequency after the
upgrade. This increased trading activity diminishes as the post-event window is broadened,
indicating that its intensity decreases as time passes after the announcement.
In short, the data for both segments show a significant positive response of liquidity to
changes in both directions. An increased, more intense frequency is observed in the periods
closest to the announcement date. In the case of commercial paper notes, the downgrade
announcements cause a reduction of market share. These results are in tune with the informative
content hypothesis but counter other hypotheses that postulate asymmetric performance for credit
rating upgrade and downgrade announcements.
5.3. Determinants of Abnormal Liquidity
The purpose of this section is to analyze the determining factors of liquidity movements
as result of rating announcements. To do so, we estimate a multiple regression model in which
the variable to be explained is the measure of abnormal liquidity. The model is as follows:
ALi ,(t1 −t2 ) = β 0 + β1 AGi + β 2 EXPi + β 3 NOTCH i + β 4GRADi + β5 FIN i + β 6 PUBi +
+ β 7Wi + β8Oi + β9 IRVi + β10TBTFi + β11 EPi + β12 AGRi + ui
(7)
where ALi ,(t1 −t2 ) denotes each of the described abnormal liquidity measures in the event window.
In model (7), the explanatory variables help to verify the different hypotheses proposed in
Section 3. Thus, variable AG, which equals one if the announcement is by Moody’s and zero if it
is by S&P or Fitch, is used to verify the hypothesis of reliability of agencies and competition
between them. To test the informative content hypothesis, different variables are included: EXP,
which equals one if the announcement is preceded by a placement on the CreditWatch list in the
18
same direction, and zero otherwise,21 and NOTCH which indicates the number of notches the
debt rating changes. We also define GRAD; this variable equals one, if the announcement implies
a shift from investment grade to speculative grade, and zero otherwise. This latter variable also
allows us to test the hypothesis of pressure on prices associated with rating triggers. This variable
is only included in the models for downgrades in the case of bonds, since only in this case the
sample contains shifts from investment to speculative grade.
We included two variables to analyze the importance of regulations affecting the issuer:
FIN, which equals one if the announcement refers to a company from the financial sector, and
zero otherwise, and PUB, which equals one in the case of a public enterprise, and zero otherwise.
To test the hypothesis of long-term orientation of rating versus other agency actions, we
introduced W, which equals 1 if the announcement is a CreditWatch list placement/retirement,
and zero otherwise, and variable O, which equals one if the announcement is an outlook change,
and zero otherwise. If these rating actions included useful short-term information, their effect on
the model was positive.
To consider the effects of the economic cycle, the model included the one year Euribor
inter-annual rate of change (IRV).22 Using the hypothesis stating that investors were more
concerned with risk in periods of economic crisis, we expected a positive effect of this variable.
Finally, to analyze the effect of company-specific characteristics, three different variables
were used. One distinguishes between large and small companies. It is called TBTF and equals
one if the logarithm of the company asset is above the mean, and zero otherwise. The other two
are performance measures: the company asset growth rate (AGR) and the economic profitability
of the company (EP). 23, 24
21
Although this is the definition in the literature of what is considered an expected rating change, we also use other
definitions. Therefore, we construct a variable that is worth 1 if the announcement is preceded by an
announcement of a rating action in the same direction in the three preceding months (E3M).
22
We also alternatively consider the growth rate of the economy (GRE).
23
The economic profitability of the fiscal year has been calculated as the pre-tax results to total assets ratio.
24
We obtain the firm balance sheet information from different sources; for financial firms, it was provided by the
CECA (Spanish Confederation of Savings & Loans) and the AEB (Spanish Commercial Banking Association),
while for the remaining firms it was obtained from SABI database (Iberian Balance Sheet Analysis System).
19
In addition to these variables, we included two control variables: ISS which indicates the
number of issues that form the portfolio, and DAY which measures the number of days in the
event window, i.e., between t1 and t2.
Model (7) was estimated separately for the sample of downgrade and upgrade
announcements, for the sample of bonds and commercial paper, and for the different abnormal
liquidity measures provided in Section 4. This involved the estimation of 18 models in the case of
bonds and 19 in the case of commercial paper notes for the rating downgrades, and just as many
for the upgrades. We estimated the models by ordinary least-squares. In order to correct the
potential effects of heteroskedasticity in the variance-covariance matrix of the OLS estimator, we
used the White’s estimator of this matrix. Before estimating the models, we tested the existence
of significant correlations between the explanatory variables. The presence of multi-co-linearity
in the models was ruled out, since the highest correlation found did not exceed 0.45 in any case.
5.3.1. Results for Downgrades
Tables 9 to 12 show the estimation of model (7) results in the case of downgrades. The
first two tables show the results pertaining to bonds and the last two - the results related to
commercial paper notes.
The model for the estimation of abnormal liquidity of bonds calculated by trading volume
is shown in Table 9. As we can see, no significant effects are found in nearly all the variables
regardless of the trading volume measure used. Therefore, the results do not support any of the
proposed hypotheses. They are not surprising, however, since we have not found a significant
response of these abnormal-liquidity variables to the rating changes (see Table 7). We only
observed that the growth rate of the company has a significant negative effect in the case of
abnormal liquidity measured as the mean daily volume of trading (MDTV). This would suggest
that the faster the asset-growth of the issuer, the lower the additional trading volume associated
with the rating announcement. The effect of this variable on the models for MDTV and MDTVA
is also negative, although not significant. The number of issues in the portfolio also shows a
negative correlation with the trading volume, as the corresponding parameter is significant in the
case of MDTV and MDTVA.
[Insert Table 9]
20
When we estimated models for the trading frequency measures (Table 10), the results
were sharply different. In that case some significant factors were found, which seemed to depend
on the size of the window used to calculate the relative frequency before and after the
announcement.25 We observed that the explanatory capacity of the model was greater when the
post-event window was narrower.
[Insert Table 10]
In general, we found no response of the abnormal liquidity to the rating agency in
question, except in the model for measure 3m-1m, in which variable AG had a significant
negative effect at 10%. Just as in the case of the volume-based measures, the fact that the events
were expected did not lead to a differential effect. The number of notches that the rate jumped did
not provide any information either. These results were contrary to the informative content
hypothesis. The variable GRAD was only significant at 10% in the case of measure T-1w. The
sign of the effect was negative, contrary to what was expected, because the hypothesis of
restrictions on institutional investors implied more market activity after the shift from investment
to speculative grade. Nevertheless, it should be noted that, in the sample, there were only 2 such
grade shifts and both corresponded to public enterprises.
As for the variables used to test regulation hypothesis, the results were indeterminate. In
the models for the shorter post-event windows, we observed that being in the financial sector
caused significant effects. The effect was positive for the longer pre-event windows (T-2w and T1m) and negative for shorter pre-event windows (3m-1w and 2m-1w). In the case of variable
PUB, significant effects with a different sign were also observed for the shorter post-event
windows and, depending on the size of the pre-event windows. This result seems to indicate that
the liquidity increases much earlier than the date of the rating change. However, as that date
approaches, the effect becomes negative. This result would support the hypothesis of regulation,
since it suggests that if there is more information about the companies on the market, the impact
on abnormal liquidity is lower on the days nearer to the event.
With regard to the influence of the different rating actions, which are used to test whether
the market values them differently than the rating changes, we have observed the following. A
25
In the calculation of the liquidity measures, the sampling sizes change. This is because there is not enough
information in some cases to calculate the relative frequency in the pre- or post-event window under consideration,
especially in the bond sample.
21
placement on the CreditWatch list did not affect abnormal liquidity. However, a change of
outlook did have an effect. In all the models, except for those calculated with the longest preevent window, we observed that an outlook change had a significant positive effect. It seems that
in the bond segment, the information provided by these outlook changes is taken into account by
investors and it increases their trading activity.
On the other hand, the economic cycle provides information to the market, although only
in models estimated for short windows, in particular 2m-1w and 2m-2w. In these cases, abnormal
liquidity positively depends on the rate of variation of interest rates, which suggests that a
deterioration of economic conditions causes an increase in the trading activity after rating
downgrades. It seems that investors are prone to assume less risk in periods of recession.
In addition, the inherent characteristics of the issues also provide relevant information. In
particular, economic profitability and the growth rate of the company’s asset have a negative
effect on liquidity that is detected by the models using the shorter post-event windows. This
result indicates that the market reaction to rating downgrades is conditioned by the information
that investors have on the re-rated companies. The impact of downgrades is weaker on companies
with better performance, suggesting that investors use other information beside the rating
announcements.
Finally, no effects related to the firm size were observed, except for one abnormal
liquidity measure for which this effect was positive. This result contradicted the too-big-to-fail
hypothesis.
Tables 11 and 12 show the results for the rating downgrades in the case of commercial
paper notes. Certain effects were observed in the case of trading volume and market share
measures (see Table 11). In particular, in the case of mean volume per trading day, we observed
that the abnormal volume was lower for financial issuers than for the other issuers. The estimated
value for the rest of the trading volume measures also had a negative sign, although this variable
was not significant.
[Insert Table 11]
The results indicated that placement on the CreditWatch list and outlook changes
provided more information than the rating downgrades themselves, as the impact of variables W
and O was negative for the three volume measures and significant at 10% for measures MDV and
22
MDTV, respectively. In the case of the market share no significant effect was found, although for
variables FIN and W the estimated effects were negative and the p-values were relatively low.
Table 12 shows the results for rating downgrades in the commercial paper segment in the
case of trading frequency measures. Here, the explanatory capacity of the analyzed variables was
greater than in the case of bonds. The adjusted R-square of the models ranged from 0.412 to
0.165, and the model as a whole was statistically significant in all the cases.
[Insert Table 12]
Also in this case, the rating agency or the fact that the action was expected did not provide
relevant information in any of the cases. The number of notches that the rating shifted after the
announcement had a significant positive impact on most of the measures, and the corresponding
estimator was positive for all of them. This result supported the informative content theory,
because the higher the rating jump, the greater the effect on abnormal liquidity.
Being in the financial sector always has a negative effect on liquidity which is significant
in the case of measures 3m-1w and 2m-1w, and which has a relatively low associated p-value in
the case of 3m-2w and 2m-2w. Variable PUB also has a significant impact on the 6 models
corresponding to the 3- and 2-month pre-event windows and the less than one month post-event
windows. The impact is negative, indicating that the reaction of the commercial paper liquidity to
a downgrade is weaker in the case of public enterprises. Thus the results seem to support the
regulation theory.
The type of rating action does not seem to provide relevant information to explain the
abnormal frequency measures. A certain effect, however, was found in the case of the economic
cycle. For instance, in the case of measure T-2w, the interest rates rate of change had a significant
positive effect on liquidity. For the remaining measures, the estimator was positive and the pvalues were relatively low in 7 of the models. This result suggested a greater impact of rating
downgrades when economic conditions worsen.
We observed sharp changes in the performance of the firms. The higher the economic
profitability, the lower the abnormal frequency caused by a downgrade in commercial paper
notes. This relationship was clearly significant in all the models. At the same time, the asset
growth rate had a negative impact, although it was significant only in 8 models. Finally, the
company size also provided relevant information. For large companies, the abnormal frequency
23
after the event has always been lower than for medium-sized companies, which supports the toobig-to-fail hypothesis.
5.3.2. Results for Upgrades
Table 13 shows the estimation of model (7) for bonds when abnormal liquidity is trading
volume-based.26 Only three variables had significant impact and, although each one was only
significant for two trading volume measures, the signs were the same in the one for which the
variable was not significant. Specifically, we observed that the parameter associated with FIN
was significant and negative, indicating a lower increase in abnormal bond liquidity for financial
issuers. Just as for downgrades, this result supported the hypothesis of regulation effect. The
parameter associated with economic profitability (EP) was also significant and negative, showing
a lower impact for the more profitable companies. On the contrary, the parameter associated with
EXP was positive, which indicated that the excess liquidity in bonds after upgrades was greater
for the announcements expected than not expected by the market. This result contradicted the
informative content hypothesis, although it could be related to the loss of reputation hypothesis.
According to that hypothesis, agencies allocate more resources to revealing negative information
than positive. Consequently, after upgrades, investors do not pay attention to the placements on
the CreditWatch list, but rather seem to wait for confirmation of the change to make their
decisions.
[Insert Table 13]
Continuing with the long-term segment, Table 14 shows the results of the estimation of
model (7) using trading frequency as liquidity proxy. In two of the models, the constant was
significantly positive, as was to be expected in light of the results presented in Table 7. As we can
see, three variables that characterize the announcements have significant parameters: AG,
NOTCH and W. Variable AG has a significant positive parameter in some models. In this case,
the rating upgrades made by Moody’s increase the frequency of abnormal bond trading to a
higher extent than the other two agencies. A placement on the CreditWatch list (W) has a
significant negative impact when 2- or 3-month windows are used. This indicates that liquidity
26
In this case, the samples for some trading frequency measures were very small, especially in the case of bonds.
Therefore, it was not possible to simultaneously estimate the effect of all the variables under consideration. Thus, we
made an initial estimation by individually including the variables, and we selected those which showed a greater
correlation with the endogenous variable. This criterion was used in all the models for upgrades.
24
increases less when the announcement is based on CreditWatch lists. Therefore, the investors
perceive these announcements differently than those that involve a rating change. The parameter
of variable NOTCH is also significant and positive only in one of the fifteen estimated models.
This indicates that the higher the number of categories the rating shifts upward, the greater the
increase of abnormal liquidity.
[Insert Table 14]
On the other hand, the economy’s growth rate has a negative impact on some models. As
postulated by the economic environment hypothesis, investors attach more value to upgrades
when the general economic situation is worse. Just as when the liquidity is computed via volume,
the firm economic profitability has a negative impact on abnormal frequency. Finally, in the
models, large companies show a higher abnormal frequency after the upgrade than smaller ones.
In regard to the commercial paper notes, Table 15 shows the results for rating upgrades in
the case of the volume and market share measures. Regardless of the measure used, no significant
effects were found in any of the variables, except for AG in the case of MDV. This suggests that,
just as in the bond segment, when the announcement is made by Moody’s, the impact on
abnormal liquidity is different than when the announcement is made by the other agencies.
[Insert Table 15]
When we analyze the results for trading frequency measures in the short-term market (see
Table 16), we find a good number of variables that significantly affect liquidity. Moreover, the
variables and the signs of the effects reflect the situation in the bond market. Several models have
a significant positive constant term, indicating an increase in abnormal liquidity as a result of the
upgrade announcements, which supports the results of the event study.
[Insert Table 16]
Three characteristics of the announcement are relevant: the agency, if it is expected, and if
it is an outlook change. It seems that the announcements by Moody’s are more credible than those
by S&P and Fitch, since they have a stronger impact on liquidity. When an enhanced outlook
change is announced, liquidity increases to a smaller extent than when other announcements are
made. This indicates that the rating refinement contains different information for investors than a
change per se or a placement on the CreditWatch list. It seems that the actions of commercial
paper market investors do not support the hypothesis of Altman & Rijken (2007) that outlooks
provide a better adjustment of ratings in the prediction of the default risk. In accordance with the
25
informative content hypothesis, expected upgrades imply a lower increase in abnormal liquidity,
probably because the impact was anticipated due to previous announcements. In that case, the
expected upgrades are defined as those preceded by other announcements in the same direction in
the last three months.
In accordance with the regulation hypothesis, the impact on abnormal liquidity is weaker
when the announcement refers to financial entities subject to more regulations. Just as for bonds,
upgrades related to the most profitable issuing companies result in a lower liquidity increase in
the market in accordance with the baseline hypothesis.
Finally, the effects of upgrade announcements differ in the commercial paper notes
segment according to the cycle phase. Just as with the volume-based measures, worsening
economic conditions cause the impact of an upgrade to decrease. This could be related to the
agency loss of reputation hypothesis.
6. Conclusions
In our work we analyze the impact of credit rating agencies’ announcements of rating
changes, outlook changes and placement on the CreditWatch list on the liquidity of the Spanish
corporate debt market, and in particular on the liquidity of bonds and commercial paper notes.
Data from the Spanish corporate fixed income market allows us to perform this kind of analysis.
Specifically, our objective was to answer several questions: Do rating announcements have any
impact on the liquidity of the Spanish corporate debt market? And, if so, what are the
determinants of that effect? Are they the ones that could be expected in the light of the
reinterpreted hypotheses formulated by other authors to explain the impact on prices? Two
methodologies were used to answer these questions: event analysis and cross-section regressions.
Moreover, a set of 18 variables was identified to measure abnormal liquidity, three trading
volume-based and 15 trading frequency-based. For commercial paper notes, an additional
measure, based on market share, was used.
With regard to the first question, our findings indicate that both rating upgrade and
downgrade announcements cause a significant increase in abnormal liquidity, which is clearly
evident when trading frequency is used as a liquidity proxy. In accordance with the informative
content hypothesis, that evidence reveals that both types of announcements contain relevant
information for Spanish corporate debt market investors and they cause the same kind of
26
reaction: increases in trading activity of the securities by the companies targeted by the
announcement.
With regard to the study of the determinants of abnormal liquidity, the results were
consistent with the literature in several ways. First of all, as some authors argue, the stability of
ratings forces investors to seek additional information from other sources. For instance, Spanish
market investors combine the information contained in the announcements with the
characteristics of the issuer, its economic profitability, the growth it has experienced, and size.
Secondly, clear evidence is found in favor of the regulation hypothesis. In general, when the
announcement refers to financial firms, which are subject to greater regulation, there is a weaker
impact on trading frequency. The same is true for public sector enterprises, though in a less
explicit way. Thirdly, the impact of the announcements is not independent from the economic
cycle and the results support the proposed hypothesis, i.e., they are counter-cyclical in the case of
rating downgrades and cyclical in the case of upgrades. Fourthly, as it was postulated in the
informative content hypothesis, the higher the number of notches the rating shifts, the greater the
impact on trading frequency.
On the other hand, we found data that disagree with the most frequent position in
literature regarding the credibility of the rating agencies. In contrast to other markets, the Spanish
market grants greater credibility to the rating upgrades made by Moody’s, which may be related
to its higher relative weight, illustrated by almost 50% of the rating actions analyzed in our
sample, and to the fact that it has been operating for a longer time on the Spanish market.
In a final conclusion, these results provide new evidence that makes it possible to assess
the role of the rating agencies in the financial market. They help to understand the determinants
of the abnormal liquidity in the corporate bond market that follows a rating change. The results
suggest that the information contained in these changes is not complete, in the sense that
investors base their decisions also on other factors. This could be a general pattern in all
international markets or could indicate a specific Spanish situation. However, in order to answer
this question, data for other countries need to be generated.
27
References
Abad, P. & Robles, M.D., 2006: “Risk and returns around bond rating changes announcements in
Spanish stock market”, Journal of Business Finance and Accounting, 33(5-6), 885-908.
Abad, P.; Díaz, A. & Robles, M.D., 2007: “The Impact of Credit Rating Announcements on
Spanish Corporate Fixed Income Performance: Returns, Yields and Liquidity”, Working
Paper 316, FUNCAS, Fundación de Cajas de Ahorros, Madrid.
Alexander, G.; Edwards, A., & Ferri, M., 2000: “The determinants of trading volume of highyield corporate bonds”, Journal of Financial Markets, 3, 177-204
Altman, E. I. & Rijken, H. A., 2006: “A point-in-time perspective on through-the-cycle ratings”,
Financial Analysts Journal, 62(1), 54-70
Altman, E. I. & Rijken, H. A., 2007: “The added value of Rating Outlooks and Rating Reviews
to corporate bond ratings”, Financial Management Association meeting, Barcelona
Amihud, Y., & Mendelson, H., 1991: “Liquidity, maturity, and the yields on U.S. Treasury
securities”, Journal of Finance, 46(4), 1411-1425
Amihud, Y., 2002: “Illiquidity and stock returns: cross-section and time-series effects”, Journal
of Financial Markets, 5, 31–56
Babbel, D.F.; Merrill, C.B.; Meyer, M.F.; & de Villiers, M., 2004: “The Effect of Transaction
Size on Off-the-Run Treasury Prices”, Journal of Financial and Quantitative Analysis,
39(3), September
Balduzzi, P.; Elton, E.J. & Green, T.C., 2001: “Economic News and Bond Prices: Evidence From
The US Treasury Market”, Journal of Financial and Quantitative Analysis 36, 523-543
Boot, A.W.A.; Milbourn, T.T. & A. Schmeits, 2006: “Credit Ratings as Coordination
Mechanisms”, The Review of Financial Studies, 19 (1), 81-118
Chakravarty S. & Sarkar, A., 1999: “Liquidity in U.S. Fixed Income Markets: A Comparison of
the Bid-Ask Spread in Corporate, Government and Municipal Bond Markets”, Working
Paper, Purdue University and Federal Reserve Bank of New York
Chen, L.; Lesmond, D.A. & J. Wei, 2007: “Corporate Yield Spreads and Bond Liquidity”,
Journal of Finance, 62(1), 119-149
Cherian, J.A.,;Jacquier, E. & Jarrow, R.A., 2004: “A Model of the Convenience Yields in Onthe-run Treasuries”, Review of Derivatives Research, 7(2), 79-97
Cornell B.; Landsman W., & Shapiro A.C., 1989: “Cross-sectional regularities in the response of
stock prices to bond rating changes”, Journal of Accounting, Auditing and Finance 4,
460–479.
Collin-Dufresne, P.; Goldstein, R. S., & Martin, J. S., 2001: “The Determinants of Credit Spread
Changes”, Journal of Finance 56, 2177-2207
Covitz, D., & P. Harrison, 2003: ‘‘Testing Conflicts of Interest at Bond Rating Agencies with
Market Anticipation: Evidence that Reputation Incentives Dominate,’’ working paper,
Federal Reserve Board, Washington
Dallocchio, M.; Hubler, J.; Raimbourg, P.; & Salvi, A., 2006: “Do Upgradings and
Downgradings Convey Information? An Event Study of the French Bond Market”,
Economic Notes: Review of Banking, Finance and Monetary Economics, 35 (3), 293–317
Dialynas, C.P. & Edington, D.M., 1992: “Bond yield spreads: a post modern view”, Journal of
Portfolio Management, Fall, 68-75
Díaz, A. & Navarro, E., 2002a: “La prima de liquidez en la Deuda del Estado española” Revista
de Economía Aplicada, 10(29), 23-58
28
Díaz, A. & Navarro, E., 2002b: “Yield spread and term to maturity: default vs. liquidity”
European Financial Management, 8(4), 449-477
Díaz, A. & Skinner, F., 2001: “Estimating Corporate Yield Curves”, Journal of Fixed Income,
11(2), 95-103
Díaz, A.; Merrick, J. & Navarro, E., 2006: “Spanish Treasury Bond Market Liquidity and
Volatility Pre- and Post-European Monetary Union”, Journal of Banking and Finance,
30(4), 1309-1332
Ederington, L., & J. Goh, 1998: ‘‘Bond Rating Agencies and Stock Analysts: Who Knows What
and When?’’, Journal of Financial and Quantitative Analysis, 33, 569–585
Elayan, F.A.; Hsu, W.H. & Meyer, T.O., 2001: “The informational content of credit rating
announcement for share prices in a small markets”. Journal of Economics and Finance,
27, 337-356.
Elton, E.J., & Green, T.C., 1998: “Tax and liquidity effects in pricing government bonds”,
Journal of Finance, 53(5), 1533-1562
Elton, E.; Gruber, M.; Agrawal, D. & Mann, D., 2001: “Explaining the rate spread on corporate
bonds”, Journal of Finance, 56, 247–277
Ericsson, J. & Renault, O., 2006: “Liquidity and Credit Risk”, Journal of Finance, 61(5), 2219 –
2250
Fisher, L., 1959: “Determinants of risk premiums of corporate bonds”, Journal of Political
Economy, 67(3), 217-237
Fledelius, P.; Lando, D. & Nielsen, J.P., 2004: “Non-Parametric Analysis of Rating Transition
and Default Data”, Journal of Investment Management, 2 (2), Second Quarter
Fleming, M.J., 2001: “Measuring Treasury market liquidity”, Federal Reserve Bank of New York
Staff Reports 133, June.
Fridson, M.S. & Jónsson, J.G., 1995: “Spread versus treasuries and the riskiness of high-yield
bonds”, Journal of Fixed Income, 5(2), 79-88
Gehr, A.K. & T.F. Martell, 1992: “Pricing Efficiency in the Secondary Market for InvestmentGrade Corporate Bonds”, Journal of Fixed Income, 2(3), 24-38
Gropp, R. & Richards A., 2001: “Rating agency actions and the pricing of debt and equity of
European banks: What can we infer about private sector monitoring of bank soundness?”
Economic Notes by Banca Monte dei Paschi di Siena, 30(3), 373-398
Hand, J.; Holthausen, R. & Leftwich, R., 1992: “The effect of bond rating agency announcements
on bond and stock prices”, Journal of Finance, 57, 733-752
Heligman, L. & Pollard, J.H., 1980: “The age pattern of mortality”, Journal of the Institute of
Actuaries, 107, 49-80
Hite, G. & Warga, A., 1997: “The effect of bond rating changes on bond price performance”,
Financial Analysts Journal, 53, 35-51
Holthausen, R.W. & Leftwich, R.W. 1986: “The Effect of Bond Rating Changes on Common
Stock Prices”, Journal of Financial Economics, 17(1), 57-89
Hong, G. & Warga, A., 2000: “An empirical study of bond market transactions”, Financial
Analyst Journal, 56(2), 32-46
Howe, J. 1995: “Credit analysis for corporate bonds”, in F.J. Fabozzi y T.D. Fabozzi, Ed., The
Handbook of Fixed Income Securities, New York, Irwin Professional Publishing, 373-410
Houweling, P.; Mentink, A. & Vorst, T., 2002: “Is liquidity reflected in bond yields? Evidence
from the Euro corporate bond market”, Working Paper, Erasmus University, Rotterdam,
April.
29
Houweling, P.; Mentink, A. & Vorst, T. 2005: “Comparing possible proxies of corporate bond
liquidity”, Journal of Banking & Finance, 29(6), 1331-1358
Hull J.; Predescu M. & White, A. 2004: “The relationship between credit default swap spreads,
bonds yields, and credit rating announcements”, Journal of Banking & Finance, 28(11),
2789-2811
Ingram, R.; Brooks, L. & Coperland, R., 1983: “The information content of bond rating changes:
A note”, Journal of Finance, 38, 997-1003
Jordan, B.D. & Kuipers, D.R., 2005: “End-of-Day Pricing in the U.S. Treasury Market: A
Comparison of GovPX and the FRBNY”, Journal of Financial Research, Spring, 97-113
Kamara, A., 1994: “Liquidity, taxes, and short-term treasury yields”, Journal of Financial and
Quantitative Analysis, 29(3), 403-417
Kish, R.J.; Hogan, K.M., & Olson, G., 1999: “Does the market perceive a difference in rating
agencies?”, Quarterly Review of Economics and Finance, 39, 363-377
Kliger, D., & O. Sarig, 2000: ‘‘The Information Value of Bond Ratings’’, Journal of Finance,
55, 2879–2902
Lesmond, D.A., 2005: “Liquidity of emerging markets”, Journal of Financial Economics, 77,
411–452
Lesmond, D.A.; Ogden, J. & Trzcinka, C., 1999: “A new estimate of transaction costs”, Review
of Financial Studies, 12, 1113–1141
Lo, A.; Mamaysky, H. & Wang, J., 2004: “Asset prices and trading volume under fixed
transaction costs”, Journal of Political Economy, 112, 1054–1090
Löffler, G., 2004: “An anatomy of rating through the cycle”, Journal of Banking and Finance,
28, 695–720
Löffler, G. 2005: “Avoiding the rating bounce: Why rating agencies are slow to react to new
information”, Journal of Behavior & Organization, 56, 365–381
Longstaff, F.; Mithal, S. & Neis, E., 2005: “Corporate yield spreads: Default risk or liquidity?
New evidence from the credit-default swap market”, Journal of Finance, 60, 2213–2253
Matolcsy Z.P. & Lianto T., 1995: “The incremental information content of bond rating revisions:
the Australian evidence” Journal of Banking and Finance 19, 891–902.
Melnik, A. & Nissim, D., 2005: “The Effects of the EMU on the Primary Market for
International Bonds”, European Financial Management Association meetings, Milan
Santos, A. C. J. & Tsatsaronis, K., 2002: “The Cost of Barriers to Entry: Evidence from the
Market for Corporate Euro Bonds Underwriting”, Mimeo, Federal Reserve Bank of New
York, 1-47
Sarig, O. & Warga, A., 1989: “Bond price data and bond market liquidity”, Journal of Financial
and Quantitative Analysis, 24(3), 367-378
Schultz, P., 2001: “Corporate Bond Trading Costs: a Peek Behind the Curtain”, Journal of
Finance, 56(2), 677-698
Schweitzer R.; Szewczyk S.H. & Varma R., 1992: “Bond rating agencies and their role in bank
market discipline” Journal of Financial Services Research, 6(3), 249–263.
Shulman, J.; Bayless, M. & Price, K., 1993: “Marketability and default influences on the yield
premia of speculative-grade debt”, Financial Management, Autumn, 132-141
Steiner, M. & Heinke, V. G., 2001: “Event study concerning international bond price effects of
credit rating actions”, International Journal of Finance and Economics, 6, 139-157
Strebulaev, I. A., 2002: “Many Faces of Liquidity and Asset Pricing: Evidence from the U.S.
Treasury Securities Market”, Working Paper, London Business School, March.
30
Wakeman, L. M., 1978: “Bond rating agencies and the capital markets”, Working Paper,
University of Rochester, New York
Wansley J.W.; Glascock J.L. & Clauretie T.M., 1992: “Institutional bond pricing and information
arrival: the case of bond rating changes” Journal of Business Finance & Accounting
19(5), 733–750.
Warga, A., 1992: “Bond returns, liquidity, and missing data”, Journal of Financial and
Quantitative Analysis, 27(4), 605-617
Weinstein, M. J., 1977: “The effect of a rating change announcement on bond price” Journal of
Financial Economics, 5, 329-350
Zaima, J.K. & McCarthy, J., 1988: “The impact of bond rating changes on common stocks and
bonds: Tests of the wealth redistribution hypothesis”, Financial Review, 23 (4), 483-498
31
Table 1. Outstanding Balances
Treasury Govern.
Bills Bonds
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
71.8
59.8
53.1
44.7
35.6
35.8
38.8
37.0
33.3
31.3
159.9
182.0
212.2
232.1
245.6
255.6
254.8
265.0
273.8
277.9
SecuritiCommr.
sation Prefer. Total
Total Paper
MDPA Notes
Bonds Cedulas bonds shares AIAF
231.7
241.8
265.3
276.8
281.2
291.5
293.6
302.0
307.1
309.2
2.2
3.1
18.0
20.7
22.1
21.1
30.3
45.2
57.4
70.8
31.8
32.2
48.4
55.0
29.9
32.0
38.2
71.2
104.3
132.9
1.4
1.2
5.3
7.3
10.6
22.1
41.9
63.1
96.1
139.2
0.2
3.2
5.4
6.4
25.1
36.8
67.4
109.9
162.1
222.9
0.6
4.4
2.9
11.8
15.0
19.9
18.7
22.5
23.1
35.6
40.3
81.6
92.3
99.4
126.9
197.7
308.0
442.4
588.9
Fixed
income
market
22.0
19.9
20.5
18.3
10.6
12.6
14.5
14.5
16.3
17.1
Note: Figures in thousands of millions of euros.
Source: AIAF, Banco de España and CNMV.
Table 2. Trading Volumes
Commr.
SecuritiTreasury Govern.
Total Paper Bonds & sation Prefer. Total
Bills Bonds Strips MDPA Notes cedulas bonds shares AIAF
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
187.9
100.5
79.8
79.8
58.1
40.1
89.8
115.0
116.1
95.5
1534.6
1617.7
1752.4
1543.2
1968.3
2255.2
2144.6
2012.4
2213.9
2862.5
22.4
16.7
15.4
15.6
12.5
9.3
8.2
14.6
1722.5
1718.2
1854.6
1639.8
2041.8
2310.9
2246.9
2136.7
2338.2
2972.6
4.5
7.8
21.9
46.4
97.4
204.9
261.5
286.5
404.4
481.6
16.9
24.4
19.7
6.8
17.3
23.1
45.8
105.9
153.0
159.0
0.1
10.7
40.5
38.7
24.7
34.7
69.4
170.1
310.8
254.9
0.0
0.7
0.9
1.5
2.2
3.4
4.1
4.0
4.6
21.4
43.1
86.3
99.8
140.8
265.0
380.2
566.6
872.3
900.2
Fixed
income
market
54.2
53.2
44.9
40.7
57.5
69.8
74.8
82.8
93.7
94.0
Note: Figures in thousands of millions of euros. The fixed income market includes fixed income and public debt
transactions in the electronic fixed income system of the Spanish exchanges and Autonomous Region debt transactions
on the Barcelona, Bilbao and Valencia exchanges.
Source: AIAF, Banco de España and CNMV.
32
Figure 1. Weekly market share in the commercial paper notes segment
5 .0 0 %
4 .5 0 %
4 .0 0 %
Market Share
3 .5 0 %
3 .0 0 %
O b s e rve d M a rk e t S h a re
2 .5 0 %
E s tim a te d M a rk e t S h a re
2 .0 0 %
1 .5 0 %
1 .0 0 %
0 .5 0 %
0 .0 0 %
0
10
20
30
40
50
60
A g e (w e e k s )
Table 3. Distribution of Rating Announcements Analyzed
Downgrades
Of rating
Of outlook
CreditWatch list
Upgrades
Of rating
Of outlook
CreditWatch list
Total
Bonds
Commercial Paper
Notes
73
38 (25) [24]
12 [8]
23 [15]
36
17 (9) [12]
11 [8]
8 [4]
109 (34) [71]
70
36 (25) [24]
9 [8]
25 [15]
50
23 (12) [12]
20 [8]
7 [4]
120 (37) [71]
Total
143
74
21
48
86
40
31
15
229
Note: Expected announcements are in parentheses. Coincidences between segments are in brackets
33
Table 4. Distribution of Rating Grade Changes
Notches
1
2
3
4
5
6
10
Total
Bonds
Downgrades Upgrades
Commercial Paper Notes
Downgrades
Upgrades
29
38
2
1
12
19
2
3
30
34
4
2
13
32
1
3
1
2 (1)
1 (1)
73
36
70
50
Total
84
123
9
9
1
2
1
229
Note: Notches is the number of categories that the debt rating changes. In the case of outlook changes
and CreditWatch list inputs, it is considered that a shift of one grade has occurred. The changes that
imply entering or leaving the speculative grade are in parentheses.
Table 5. Distribution of Rating Agency
Moody’s
Standard & Poor’s
Fitch
Total
Bonds
Commercial Paper
Notes
52 (33)
27 (18)
30 (20)
109
59 (33)
28 (18)
33 (20)
120
Note: Coincidences between segments are in parentheses.
Table 6. Distribution of Debt Issuers
Financial Sector Public Enterprises
Bonds
Commercial Paper
Notes
Total
10 (8)
4 (1)
15 (10)
17 (8)
2 (1)
23 (10)
Note: Number of companies that issue simultaneously in both markets are in parentheses.
34
Table 7. Abnormal Liquidity in the Bond Market
Downgrades
Mean abnormal
T-ratio
liquidity %
Volume Measures
MDV
-0.087
MDTV
0.063
MDTVA
-0.148
Frequency Measures
T-1w
0.354
T-2w
0.187
T-1m
0.089
T-2m
-0.045
3m-1w
0.408
3m-2w
0.268
3m-1m
0.197
3m-2m
0.062
2m-1w
0.378
2m-2w
0.242
2m-1m
0.164
2m-2m
0.026
1m-1w
0.276
1m-2w
0.128
1m-1m
0.060
Wilcoxon
rank test
-0.450
(0.653)
0.712
(0.476)
-0.630
(0.529)
0.096
(0.924)
1.644
(0.100)
0.431
(0.667)
8.815**
(0.000)
4.550**
(0.000)
2.258**
(0.024)
-1.025
(0.305)
9.232**
(0.000)
6.396**
(0.000)
4.768*
(0.000)
1.452
(0.146)
7.839**
(0.000)
5.398*
(0.000)
3.781**
(0.000)
0.577
(0.564)
5.848**
(0.000)
2.894**
(0.004)
1.493
(0.135)
5.626**
(0.000)
3.748**
(0.000)
1.860*
(0.063)
0.875
(0.382)
5.752**
(0.000)
4.990**
(0.000)
3.891**
(0.000)
1.301
(0.193)
5.4291**
(0.000)
4.4857**
(0.000)
3.351**
(0.001)
0.449
(0.653)
4.391**
(0.000)
2.611**
(0.009)
1.423
(0.155)
Upgrades
Mean abnormal
Wilcoxon
T-ratio rank test
liquidity %
0.834
0.628
1.355
0.371
0.292
0.217
0.029
0.511
0.318
0.311
0.158
0.405
0.208
0.184
0.053
0.334
0.151
0.088
2.523**
(0.012)
2.129**
(0.033)
3.234**
(0.001)
2.001*
(0.045)
2.332**
(0.020)
2.843**
(0.005)
3.923** 2.951**
(0.000) (0.003)
3.082** 2.456**
(0.002) (0.014)
2.575** 2.613**
(0.010) (0.009)
0.413
0.036
(0.680) (0.971)
5.035** 3.551**
(0.000) (0.000)
3.442** 2.962**
(0.001) (0.003)
4.092** 3.077**
(0.000) (0.002)
2.531** 2.287**
(0.011) (0.022)
3.900** 3.0301**
(0.000) (0.002)
2.229** 1.5680
(0.026) (0.117)
2.278* 1.960**
(0.023) (0.050)
0.819
0.350
(0.413) (0.726)
3.122** 2.485**
(0.002) (0.013)
1.510
1.381
(0.131) (0.167)
1.183
1.216
(0.237) (0.224)
Note: MDV: total volume traded in the last three months divided by the number of days on which each asset is traded during that
period. MDTV: mean daily trading volume in the last three months. MDTVA: mean daily trading volume accumulated in the last
three months. Abnormal frequencies calculated as the difference of the logarithm of the mean trading frequency of a pre-event
window (PREW) with respect to a post-event window (POSTW): PREW-POSTW., where PREW= 1, 2 , 3 months and the entire
period since the issue (1M, 2M, 3M, T) and POSTW= 1 & 2 weeks and 1 & 2 months (1w, 2w, 1m, 2m). * and ** indicate
significance at least at 10% or at 5%, respectively. p-value is in parentheses.
35
Table 8. Abnormal Liquidity in the Commercial Paper Notes Market
Downgrades
Mean abnormal
T-ratio
liquidity %
Volume Measures
MDV
-0.177
MDTV
-0.455
MDTVA
-0.706
Market Share
MS
Wilcoxon
test
-0.623
(0.533)
-1.396
(0.163)
-1.561
(0.119)
0.175
(0.861)
0.921
(0.357)
1.551
(0.121)
-0.883
-1.886*
(0.059)
2.048**
(0.041)
Frequency Measures
T-1w
0.692
T-2w
0.467
T-1m
0.284
T-2m
0.140
3m-1w
0.509
3m-2w
0.270
3m-1m
0.118
3m-2m
-0.024
2m-1w
0.427
2m-2w
0.194
2m-1m
0.037
2m-2m
-0.107
1m-1w
0.270
1m-2w
-0.130
1m-1m
0.053
21.367**
(0.000)
14.128**
(0.000)
7.890**
(0.000)
3.536**
(0.000)
15.243**
(0.000)
8.177**
(0.000)
3.242**
(0.001)
-0.621
(0.535)
12.749**
(0.000)
5.590**
(0.000)
1.019
(0.308)
-2.780**
(0.005)
7.770**
(0.000)
-3.604**
(0.000)
1.550
(0.121)
6.843*
(0.000)
7.054**
(0.000)
6.101**
(0.000)
3.294**
(0.001)
6.539**
(0.000)
5.826**
(0.000)
2.808**
(0.005)
0.933
(0.351)
6.459**
(0.000)
4.817**
(0.000)
0.596
(0.551)
3.043**
(0.002)
5.686**
(0.000)
3.816**
(0.000)
1.806
(0.071)
Upgrades
Mean abnormal
Wilcoxon
T-ratio
liquidity %
test
0.025
0.105
(0.916)
-0.116
(0.908)
-0.787
(0.431)
0.193
(0.847)
0.303
(0.762)
0.666
(0.505)
-0.467
-1.006
(0.315)
0.811
(0.417)
0.701
14.42**
(0.000)
10.10**
(0.000)
6.861**
(0.000)
3.661**
(0.000)
14.15**
(0.000)
8.627**
(0.000)
3.424**
(0.001)
-0.347
(0.728)
10.48**
(0.000)
4.845**
(0.000)
1.854*
(0.064)
-1.288
(0.198)
7.674**
(0.000)
2.463**
(0.014)
-1.393
(0.164)
5.639**
(0.000)
5.867**
(0.000)
5.254**
(0.000)
3.051**
(0.002)
5.295**
(0.000)
4.922**
(0.000)
3.148**
(0.002)
0.692
(0.489)
3.4903**
(0.001)
3.353**
(0.001)
1.878*
(0.060)
1.942
(0.052)
5.252**
(0.000)
2.393**
(0.017)
2.218
(0.027)
-0.039
-0.402
0.509
0.349
0.190
0.556
0.400
0.139
-0.015
0.449
0.270
0.090
-0.064
0.292
0.106
-0.062
Note: See note in Table 7. MS is the abnormal market share computed as the difference between the rate of change of the market
share observed around the rating event and the one expected according to the model (5).
36
Table 9. Determinants of excess bond liquidity after rating downgrades: Volume measures
Constant
Moody’s (AG)
Expected (ESX)
No. of Notches Shifts (NOTCH)
Shift from investment to speculative (GRAD)
Financial Sector (FIN)
Public Enterprise (PUB)
CreditWatch List (W)
Outlook (O)
Interest rate variation (IRV)
Economic Profitability (EP)
Size (TBTF)
Asset growth rate (AGR)
No. of issues (ISS)
Window size (DAY)
Adjusted R-squared
F
F p_val
Obs
MDV
MDTV
MDTVA
-0.666
( 0.656)
0.394
( 0.402)
-0.035
( 0.960)
0.234
( 0.472)
0.061
( 0.953)
0.438
( 0.554)
-0.148
( 0.939)
0.427
( 0.610)
0.764
( 0.376)
-0.692
( 0.389)
5.528
( 0.683)
0.141
( 0.876)
-1.258*
( 0.088)
-0.215
( 0.308)
-0.002
( 0.662)
-0.038
0.824
( 0.641)
68
1.431
( 0.475)
0.324
( 0.534)
-0.281
( 0.754)
0.177
( 0.621)
0.258
( 0.849)
-0.241
( 0.813)
-1.662
( 0.459)
0.661
( 0.530)
0.611
( 0.553)
-1.687
( 0.132)
-5.292
( 0.728)
-0.678
( 0.581)
-0.782
( 0.450)
-0.565**
( 0.047)
-0.003
( 0.512)
-0.005
0.979
( 0.487)
68
1.504
( 0.458)
0.160
( 0.769)
-0.311
( 0.735)
0.207
( 0.561)
0.026
( 0.984)
-0.153
( 0.883)
-1.619
( 0.474)
0.879
( 0.431)
0.679
( 0.518)
-2.020
( 0.111)
-7.654
( 0.633)
-0.619
( 0.623)
-0.989
( 0.342)
-0.621**
( 0.032)
-0.003
( 0.570)
0.022
1.107
( 0.373)
68
Note: AG: dummy worth 1 if the announcement is from Moody’s, EXP: dummy equal to one if the announcement is preceded by a
CreditWatch list input/output in the same direction, NOTCH: number of notches that the debt rating changes, GRAD: dummy
equal to one when the announcement implies a shift from investment grade to speculative grade, FIN: dummy equal to one when
the announcement refers to a financial sector company, PUB: dummy equal to one when a public enterprise is involved, W:
dummy equal to one if the announcement is a CreditWatch list input/output, O: dummy equal to one if the announcement is a
change in outlook, IRV: is the rate of inter-annual variation of the Euribor at one year, GRE: is the growth rate of the economy
(GDP), TBTF: dummy equal to if the logarithm of the company asset is above the mean, AGR: is the growth rate of the company
asset, EP: is the economic profitability of the company, ISS: is the no. of issues that form the portfolio, DAY: is the number of
days in the window (t1, t2). MDV: total volume traded in the last three months divided by the number of days on which each asset
is traded during that period. MDTV: mean daily trading volume in the last three months. MDTVA: mean daily trading volume
accumulated in the last three months. Estimation by OLS with the White`s estimator of the variance-covariance matrix robust for
heteroscedasticity. * and ** indicate significance at least at 10% or at 5%, respectively. p-value is in parentheses.
37
Table 10. Determinants of excess bond liquidity after rating downgrades: frequency measures
Const.
AG
EXP
NOTCH
GRAD
FIN
PUB
W
O
IRV
EP
TBTF
AGR
ISS
DAY
Adj. R2
F
F p_val
Obs
T-1w
T-2w
T-1m
0.185
( 0.428)
-0.089
( 0.369)
0.038
( 0.765)
0.068
( 0.153)
-0.518*
( 0.095)
0.032
( 0.828)
0.491**
( 0.027)
0.014
( 0.941)
0.088
( 0.484)
-0.115
( 0.643)
-3.298
( 0.160)
0.175
( 0.218)
-0.468**
( 0.001)
0.033
( 0.311)
-0.001
( 0.205)
0.120
1.420
( 0.206)
44
-0.266
( 0.354)
-0.001
( 0.995)
-0.203
( 0.179)
0.059
( 0.413)
-0.519
( 0.301)
0.522**
( 0.002)
0.663**
( 0.050)
-0.046
( 0.803)
-0.071
( 0.638)
-0.113
( 0.525)
2.147
( 0.381)
0.351**
( 0.016)
-0.247
( 0.164)
0.007
( 0.852)
0.000
( 0.800)
0.152
1.600
( 0.131)
48
-0.029
( 0.913)
0.015
( 0.860)
-0.146
( 0.244)
-0.027
( 0.643)
0.039
( 0.933)
0.238*
( 0.063)
0.366*
( 0.050)
-0.165
( 0.324)
-0.055
( 0.702)
0.018
( 0.925)
3.474
( 0.151)
-0.069
( 0.575)
-0.196*
( 0.096)
0.040
( 0.220)
0.000
( 0.857)
0.158
1.748*
( 0.082)
57
T-2m
3m-1w
3m-2w 3m-1m 3m-2m
0.057
0.907**
0.419
( 0.853) ( 0.001) ( 0.164)
0.050 -0.164* -0.107
( 0.605) ( 0.089) ( 0.298)
-0.114
0.115
0.015
( 0.364) ( 0.347) ( 0.900)
-0.053
0.043
0.031
( 0.505) ( 0.385) ( 0.681)
-0.025
-0.198
-0.176
( 0.967) ( 0.546) ( 0.734)
0.052 -0.399**
0.024
( 0.753) ( 0.044) ( 0.922)
0.179
-0.208
0.007
( 0.451) ( 0.365) ( 0.982)
-0.138
0.099
0.128
( 0.447) ( 0.542) ( 0.450)
-0.105
0.351** 0.305**
( 0.483) ( 0.009) ( 0.014)
-0.238
0.198
0.215
( 0.298) ( 0.244) ( 0.194)
3.808 -10.778** -5.018**
( 0.118) ( 0.000) ( 0.080)
-0.176
-0.005
0.087
( 0.248) ( 0.974) ( 0.689)
-0.121 -0.786** -0.601**
( 0.531) ( 0.000) ( 0.014)
0.009
-0.007
-0.028
( 0.801) ( 0.835) ( 0.441)
0.001
0.000
0.001
( 0.335) ( 0.685) ( 0.245)
0.049
0.282
0.129
1.220
2.373**
1.562
( 0.294) ( 0.019) ( 0.135)
61
50
54
0.275
( 0.284)
-0.029
( 0.756)
0.018
( 0.872)
-0.017
( 0.743)
-0.100
( 0.802)
0.109
( 0.535)
0.181
( 0.399)
0.101
( 0.515)
0.264*
( 0.062)
0.148
( 0.332)
-2.182
( 0.407)
-0.056
( 0.767)
-0.218
( 0.227)
-0.010
( 0.756)
0.003
( 0.404)
0.114
1.504
( 0.153)
56
0.412
( 0.177)
-0.001
( 0.995)
0.082
( 0.534)
-0.039
( 0.558)
-0.156
( 0.753)
-0.089
( 0.653)
0.017
( 0.946)
0.096
( 0.528)
0.257*
( 0.083)
0.068
( 0.639)
-2.540
( 0.338)
-0.157
( 0.413)
-0.213
( 0.234)
-0.041
( 0.225)
0.009**
( 0.041)
0.020
1.083
( 0.398)
59
2m-1w
1.105**
( 0.000)
-0.103
( 0.346)
0.023
( 0.844)
0.024
( 0.718)
-0.137
( 0.763)
-0.573**
( 0.010)
-0.475**
( 0.054)
0.005
( 0.975)
0.345**
( 0.009)
0.307**
( 0.077)
-11.792**
( 0.001)
-0.156
( 0.344)
-0.775**
( 0.000)
0.010
( 0.772)
0.000
( 0.632)
0.344
2.833**
( 0.006)
50
2m-2w 2m-1m 2m-2m 1m-1w 1m-2w 1m-1m
0.624*
( 0.055)
-0.065
( 0.566)
-0.067
( 0.561)
0.026
( 0.770)
-0.199
( 0.742)
-0.177
( 0.509)
-0.259
( 0.442)
0.092
( 0.597)
0.318**
( 0.016)
0.289**
( 0.083)
-6.478**
( 0.042)
-0.044
( 0.850)
-0.610**
( 0.015)
-0.019
( 0.618)
0.001
( 0.605)
0.210
2.005**
( 0.044)
54
0.353
( 0.194)
0.041
( 0.694)
-0.053
( 0.618)
-0.028
( 0.675)
-0.071
( 0.890)
-0.055
( 0.759)
-0.065
( 0.773)
0.022
( 0.890)
0.259*
( 0.055)
0.191
( 0.207)
-2.414
( 0.393)
-0.197
( 0.316)
-0.124
( 0.526)
0.012
( 0.718)
-0.002
( 0.633)
0.144
1.658
( 0.104)
56
0.499
( 0.125)
0.057
( 0.615)
0.018
( 0.887)
-0.048
( 0.539)
-0.124
( 0.838)
-0.261
( 0.214)
-0.231
( 0.381)
0.008
( 0.962)
0.255*
( 0.095)
0.130
( 0.407)
-2.958
( 0.296)
-0.296
( 0.159)
-0.137
( 0.467)
-0.018
( 0.629)
0.004
( 0.369)
-0.004
0.984
( 0.485)
59
0.523
( 0.238)
-0.106
( 0.460)
0.032
( 0.788)
0.063
( 0.315)
-0.449
( 0.282)
-0.232
( 0.399)
-0.270
( 0.360)
0.124
( 0.557)
0.268**
( 0.029)
0.029
( 0.900)
-6.332
( 0.170)
0.000
( 0.998)
-0.251
( 0.353)
-0.007
( 0.898)
-0.004
( 0.754)
-0.078
0.798
( 0.663)
40
0.055
( 0.893)
-0.029
( 0.838)
-0.139
( 0.199)
0.056
( 0.537)
-0.462
( 0.465)
0.181
( 0.475)
-0.069
( 0.845)
0.167
( 0.354)
0.210*
( 0.100)
0.024
( 0.927)
-1.221
( 0.752)
0.156
( 0.169)
-0.061
( 0.838)
-0.035
( 0.440)
-0.003
( 0.824)
0.031
1.099
( 0.398)
44
0.216
( 0.427)
0.023
( 0.829)
0.014
( 0.899)
0.003
( 0.952)
-0.361
( 0.383)
-0.102
( 0.526)
-0.103
( 0.602)
0.173
( 0.272)
0.298**
( 0.050)
0.009
( 0.968)
-0.284
( 0.917)
-0.232
( 0.119)
0.068
( 0.653)
-0.016
( 0.671)
0.003
( 0.682)
-0.029
0.898
( 0.567)
52
Note: See note in Table 9. Abnormal frequencies calculated as the difference of the logarithm of the mean trading frequency of a pre-event window (PREW) with respect to a post-event window (POSTW):
PREW-POSTW., where PREW= 1, 2 , 3 months and the entire period since the issue (1m, 2m, 3m, T) and POSTW= 1 & 2 weeks and 1 & 2 months (1w, 2w, 1m, 2m).
38
Table 11. Determinants of excess commercial paper notes liquidity after rating
downgrades: volume and MS measures
Constant
Moody’s (AG)
Expected (EXP)
No. of notches shifts (NOTCHES)
Financial Sector (FIN)
Public Enterprise (PUB)
CreditWatch List (W)
Outlook (O)
Interest rate variation (IRV)
Economic Profitability (EP)
Size (TBTF)
Asset growth rate (AGR)
No. of issues (ISS)
Window size (DAY)
Adjusted R-squared
F
F p_val
Obs
MDV
MDTV
MDTVA
MS
3.238
( 0.255)
-0.372
( 0.735)
-1.413
( 0.312)
0.273
( 0.896)
-4.150*
( 0.053)
-1.393
( 0.508)
-4.468*
( 0.094)
-1.682
( 0.284)
2.704
( 0.180)
-15.350
( 0.564)
-0.794
( 0.574)
0.474
( 0.774)
0.115
( 0.172)
0.002
( 0.987)
0.039
1.193
( 0.312)
63
1.310
( 0.413)
-0.658
( 0.366)
-0.805
( 0.367)
0.255
( 0.817)
-0.369
( 0.730)
0.346
( 0.812)
-2.025
( 0.173)
-1.392*
( 0.064)
1.818
( 0.336)
12.572
( 0.512)
-0.615
( 0.499)
1.102
( 0.290)
0.034
( 0.498)
0.063
( 0.271)
-0.029
0.864
( 0.594)
63
1.893
( 0.287)
-0.987
( 0.254)
-0.582
( 0.558)
-0.314
( 0.812)
-1.818
( 0.211)
1.703
( 0.270)
-2.651
( 0.120)
-0.114
( 0.910)
0.901
( 0.580)
-9.084
( 0.700)
-0.328
( 0.739)
-0.203
( 0.857)
0.083
( 0.108)
-0.016
( 0.827)
-0.043
0.805
( 0.652)
63
2.155
( 0.474)
-0.398
( 0.715)
-0.551
( 0.703)
0.197
( 0.934)
-3.468
( 0.115)
-2.063
( 0.385)
-4.036
( 0.150)
-1.371
( 0.401)
1.523
( 0.444)
-7.988
( 0.766)
-0.697
( 0.637)
0.383
( 0.804)
0.105
( 0.236)
0.020
( 0.846)
0.017
1.083
( 0.395)
63
Note: See note in Table 9. MS is the abnormal market share computed as the difference between the rate of
growth of the market share around the rating event and the one expected according to the model (5).
39
Table 12. Determinants of excess commercial paper notes liquidity after rating downgrades: frequency measures
T-1w
Const
0.858**
( 0.000)
AG
-0.024
( 0.709)
EXP
0.009
( 0.914)
NOTCH 0.068
( 0.526)
FIN
-0.098
( 0.379)
PUB
0.114
( 0.485)
W
0.028
( 0.845)
O
0.000
( 0.998)
ITV
0.149
( 0.105)
EP
-5.678**
( 0.000)
TBTF
-0.019
( 0.781)
AGR -0.337**
( 0.010)
ISS
-0.007
( 0.233)
DAY -0.041**
( 0.035)
0.412
Adj. R2
F
3.959**
F p_val ( 0.000)
Obs
56
T-2w
T-1m
T-2m
0.698**
( 0.000)
-0.007
( 0.912)
-0.141
( 0.233)
0.118
( 0.329)
-0.026
( 0.833)
-0.097
( 0.669)
-0.091
( 0.576)
-0.164
( 0.218)
0.253**
( 0.037)
-5.426**
( 0.000)
-0.032
( 0.665)
-0.313**
( 0.023)
-0.001
( 0.834)
-0.028
( 0.112)
0.350
3.484**
( 0.001)
61
0.502**
( 0.027)
-0.013
( 0.855)
-0.170
( 0.231)
0.120
( 0.250)
-0.056
( 0.732)
-0.155
( 0.542)
-0.124
( 0.491)
-0.179
( 0.300)
0.268
( 0.102)
-4.649**
( 0.019)
-0.057
( 0.570)
-0.364**
( 0.011)
0.008
( 0.198)
-0.021
( 0.357)
0.233
2.400**
( 0.014)
61
0.279
( 0.119)
0.016
( 0.843)
-0.129
( 0.336)
0.148
( 0.156)
-0.083
( 0.599)
0.139
( 0.761)
-0.041
( 0.822)
-0.080
( 0.627)
0.276
( 0.131)
-4.811**
( 0.013)
-0.132
( 0.140)
-0.428**
( 0.008)
0.016**
( 0.037)
-0.017
( 0.453)
0.252
2.578**
( 0.009)
62
3m-1w 3m-2w 3m-1m 3m-2m 2m-1w 2m-2w 2m-1m 2m-2m 1m-1w 1m-2w 1m-1m
0.732**
( 0.000)
-0.078
( 0.274)
-0.068
( 0.520)
0.138
( 0.213)
-0.262*
( 0.051)
-0.475**
( 0.005)
0.077
( 0.612)
0.030
( 0.780)
0.057
( 0.560)
-5.101**
( 0.004)
-0.187**
( 0.033)
0.055
( 0.670)
-0.001
( 0.847)
-0.027*
( 0.075)
0.258
2.338**
( 0.022)
51
0.516**
( 0.009)
-0.044
( 0.474)
-0.163
( 0.235)
0.202**
( 0.047)
-0.217
( 0.150)
-0.389**
( 0.000)
0.002
( 0.990)
-0.106
( 0.485)
0.083
( 0.535)
-5.344**
( 0.001)
-0.158**
( 0.067)
0.041
( 0.702)
0.003
( 0.666)
-0.013
( 0.427)
0.206
2.077**
( 0.038)
55
0.326
( 0.117)
-0.009
( 0.911)
-0.140
( 0.274)
0.212**
( 0.004)
-0.035
( 0.825)
-0.333**
( 0.012)
-0.011
( 0.948)
-0.242
( 0.139)
0.225
( 0.162)
-5.106**
( 0.021)
-0.118
( 0.205)
-0.240*
( 0.061)
0.008
( 0.150)
0.003
( 0.879)
0.306
3.032**
( 0.003)
61
0.145
( 0.345)
0.028
( 0.726)
-0.121
( 0.256)
0.219**
( 0.009)
-0.040
( 0.769)
-0.050
( 0.873)
0.029
( 0.859)
-0.168
( 0.251)
0.230
( 0.161)
-5.214**
( 0.007)
-0.199**
( 0.015)
-0.289**
( 0.026)
0.015**
( 0.018)
0.005
( 0.785)
0.360
3.637**
( 0.001)
62
0.686**
( 0.000)
-0.105
( 0.107)
-0.079
( 0.400)
0.143
( 0.213)
-0.256*
( 0.059)
-0.413**
( 0.021)
0.091
( 0.546)
0.062
( 0.588)
0.068
( 0.493)
-4.720**
( 0.007)
-0.238**
( 0.007)
0.109
( 0.377)
0.000
( 0.950)
-0.018
( 0.262)
0.279
2.490**
( 0.015)
51
0.508**
( 0.006)
-0.055
( 0.425)
-0.174
( 0.169)
0.177
( 0.107)
-0.245
( 0.151)
-0.316**
( 0.011)
-0.035
( 0.831)
-0.041
( 0.796)
0.108
( 0.433)
-5.565**
( 0.002)
-0.178**
( 0.053)
0.062
( 0.616)
0.004
( 0.565)
-0.005
( 0.757)
0.178
1.898**
( 0.060)
55
0.317**
( 0.093)
0.010
( 0.897)
-0.145
( 0.219)
0.190**
( 0.010)
-0.055
( 0.708)
-0.248*
( 0.075)
-0.037
( 0.821)
-0.187
( 0.216)
0.271
( 0.120)
-5.335**
( 0.017)
-0.129
( 0.128)
-0.247*
( 0.053)
0.007
( 0.157)
0.015
( 0.399)
0.308
3.054**
( 0.002)
61
0.123
( 0.420)
0.044
( 0.579)
-0.120
( 0.247)
0.203**
( 0.014)
-0.067
( 0.615)
0.039
( 0.907)
0.017
( 0.918)
-0.105
( 0.453)
0.277
( 0.118)
-5.460**
( 0.004)
-0.209**
( 0.009)
-0.301**
( 0.023)
0.014**
( 0.015)
0.018
( 0.271)
0.361
3.656**
( 0.000)
62
0.503**
( 0.006)
-0.028
( 0.725)
0.044
( 0.707)
0.132*
( 0.100)
-0.030
( 0.825)
0.200
( 0.268)
0.092
( 0.601)
-0.093
( 0.455)
0.006
( 0.963)
-5.516**
( 0.011)
-0.116
( 0.166)
-0.084
( 0.529)
-0.005
( 0.455)
0.016
( 0.488)
0.207
2.107**
( 0.034)
56
0.342*
( 0.060)
-0.014
( 0.854)
-0.031
( 0.791)
0.143**
( 0.040)
-0.022
( 0.851)
-0.049
( 0.807)
-0.016
( 0.923)
-0.139
( 0.263)
0.105
( 0.492)
-6.100**
( 0.002)
-0.101
( 0.218)
-0.088
( 0.463)
-0.002
( 0.782)
0.015
( 0.423)
0.249
2.528**
( 0.010)
61
0.146
( 0.492)
-0.021
( 0.817)
-0.060
( 0.658)
0.144**
( 0.043)
-0.053
( 0.691)
-0.107
( 0.651)
-0.050
( 0.799)
-0.154
( 0.285)
0.120
( 0.533)
-5.324**
( 0.029)
-0.125
( 0.188)
-0.140
( 0.288)
0.008
( 0.133)
0.022
( 0.371)
0.165
1.913*
( 0.053)
61
Note: See note in Table 9. Abnormal frequencies calculated as the difference of the logarithm of the mean trading frequency of a pre-event window (PREW) with respect to a postevent window (POSTW): PREW-POSTW., where PREW= 1, 2 , 3 months and the entire period since the issue (1M, 2M, 3M, T) and POSTW= 1 & 2 weeks and 1 & 2 months
(1W, 2W, 1M, 2M).
40
Table 13. Determinants of excess bond liquidity after rating upgrades: volume
measures
Constant
Moody’s (AG)
Expected (EXP)
No. of grade shifts (GRAD)
Financial Sector (FIN)
CreditWatch List (W)
Outlook (O)
Economy Growth Rate (GRE)
Economic Profitability (EP)
Size (TBTF)
Adjusted R-squared
F
F p_val
Obs
MDV
MDTV
MDTVA
4.060
( 0.170)
0.047
( 0.964)
2.212**
( 0.041)
0.117
( 0.767)
-3.524
( 0.164)
1.404
( 0.368)
0.535
( 0.584)
0.027
( 0.921)
-69.313
( 0.195)
-0.856
( 0.450)
-0.029
0.904
( 0.539)
32
6.002
( 0.115)
-0.098
( 0.934)
2.953*
( 0.052)
0.545
( 0.290)
-5.591*
( 0.082)
2.803
( 0.143)
1.157
( 0.369)
0.217
( 0.487)
-118.677*
( 0.075)
-1.179
( 0.367)
0.075
1.278
( 0.303)
32
6.415*
( 0.081)
0.067
( 0.956)
2.215
( 0.114)
0.258
( 0.573)
-5.074*
( 0.100)
1.856
( 0.302)
0.367
( 0.754)
0.120
( 0.695)
-108.041*
( 0.086)
-1.242
( 0.332)
0.020
1.070
( 0.421)
32
Note: See note in Table 9.
41
Table 14. Determinants of excess bond liquidity after rating upgrades: frequency measures
T-1w
Const
-0.175
( 0.864)
AG
0.539
( 0.199)
EXP
0.068
( 0.924)
NOTCH 0.163
( 0.500)
FIN
0.373
( 0.503)
W
0.207
( 0.773)
O
-0.285
( 0.674)
GRE
-0.121
( 0.160)
EP
-6.100
( 0.571)
TBTF
0.383
( 0.246)
Adj. R2 -0.565
F
0.519
F p_val ( 0.804)
Obs
13
T-2w
T-1m
T-2m
-0.112
( 0.903)
0.401
( 0.239)
0.231
( 0.625)
0.152
( 0.317)
0.078
( 0.872)
0.122
( 0.827)
-0.075
( 0.900)
-0.089
( 0.132)
-9.886
( 0.264)
0.405*
( 0.068)
-0.168
0.777
( 0.651)
15
-0.350
( 0.660)
0.439*
( 0.093)
0.300
( 0.429)
0.212*
( 0.085)
-0.021
( 0.961)
0.468
( 0.308)
0.000
( 1.000)
-0.121**
( 0.008)
-2.993
( 0.726)
0.056
( 0.711)
0.250
1.593
( 0.276)
17
0.304
( 0.600)
0.036
( 0.861)
0.435
( 0.119)
0.099
( 0.290)
-0.543
( 0.227)
0.339
( 0.298)
0.039
( 0.873)
-0.048
( 0.245)
-8.563
( 0.309)
-0.012
( 0.950)
0.032
1.088
( 0.425)
25
3m-1w 3m-2w 3m-1m 3m-2m 2m-1w 2m-2w
2m-1m
0.924
( 0.193)
0.335
( 0.260)
0.003
( 0.992)
-0.028
( 0.748)
-0.026
( 0.961)
-0.167
( 0.612)
-0.258
( 0.272)
-0.065
( 0.367)
-6.625
( 0.363)
-0.356
( 0.379)
0.380
2.020
( 0.202)
16
0.463
0.496
0.733
1.044*
0.818*
( 0.498) ( 0.224) ( 0.225) ( 0.070) ( 0.052)
0.186
0.297* 0.741*
0.357
0.328**
( 0.459) ( 0.097) ( 0.068) ( 0.198) ( 0.044)
0.461
0.166
-0.386
-0.031
0.131
( 0.265) ( 0.565) ( 0.305) ( 0.921) ( 0.425)
-0.008
-0.049
-0.057
-0.112
-0.035
( 0.955) ( 0.577) ( 0.567) ( 0.457) ( 0.632)
-0.497 -0.560* -0.144 -0.651* -0.786**
( 0.321) ( 0.070) ( 0.700) ( 0.086) ( 0.031)
-0.026 -0.374* -0.386
-0.601
-0.180
( 0.948) ( 0.066) ( 0.270) ( 0.116) ( 0.282)
0.226
-0.011
-0.565
-0.217
-0.098
( 0.445) ( 0.951) ( 0.114) ( 0.412) ( 0.487)
-0.066 -0.064* -0.008
0.056
0.015
( 0.178) ( 0.090) ( 0.877) ( 0.304) ( 0.642)
-9.296 -10.908* -14.451 -20.59** -13.834**
( 0.298) ( 0.095) ( 0.177) ( 0.024) ( 0.012)
-0.067
0.223
0.042
0.093
-0.279
( 0.828) ( 0.211) ( 0.854) ( 0.464) ( 0.185)
0.229
0.221
0.629
0.562
0.681
1.462
1.662
3.265
2.857
4.317*
( 0.353) ( 0.203) ( 0.180) ( 0.162) ( 0.061)
15
22
13
14
15
1.021
( 0.192)
0.116
( 0.654)
0.161
( 0.665)
-0.085
( 0.500)
-0.475
( 0.407)
-0.396
( 0.300)
-0.016
( 0.958)
-0.045
( 0.616)
-11.712
( 0.171)
-0.119
( 0.794)
-0.170
0.741
( 0.670)
17
0.557
( 0.311)
0.205
( 0.415)
0.390
( 0.273)
-0.018
( 0.878)
-0.408
( 0.323)
-0.106
( 0.766)
0.130
( 0.644)
-0.077
( 0.116)
-9.053
( 0.109)
-0.010
( 0.969)
0.357
1.924
( 0.219)
16
0.521
( 0.161)
0.352*
( 0.095)
0.024
( 0.916)
-0.070
( 0.362)
-0.421
( 0.134)
-0.478**
( 0.025)
-0.124
( 0.555)
-0.075*
( 0.052)
-9.622*
( 0.064)
0.306*
( 0.070)
0.192
1.582
( 0.219)
23
0.783
( 0.371)
0.424
( 0.248)
-0.107
( 0.735)
-0.027
( 0.766)
0.053
( 0.938)
-0.147
( 0.695)
-0.325
( 0.110)
-0.039
( 0.642)
-6.064
( 0.499)
-0.436
( 0.374)
0.273
1.625
( 0.285)
16
0.932
( 0.326)
0.159
( 0.594)
0.087
( 0.811)
-0.099
( 0.464)
-0.456
( 0.526)
-0.428
( 0.303)
-0.049
( 0.870)
-0.012
( 0.905)
-11.428
( 0.300)
-0.193
( 0.714)
-0.349
0.540
( 0.808)
17
2m-2m 1m-1w
1m-2w
1m-1m
Note: See note in Table 9. Abnormal frequencies calculated as the difference of the logarithm of the mean trading frequency of a pre-event window (PREW) with respect to a postevent window (POSTW): PREW-POSTW., where PREW= 1, 2 , 3 months and the entire period since the issue (1m, 2m, 3m, T) and POSTW= 1 & 2 weeks and 1 & 2 months (1w,
2w, 1m, 2m).
42
Table 15. Determinants of excess commercial paper notes liquidity after
rating upgrades: volume and MS measures
Constant
Moody’s (AG)
Expected (E3M)
No. of notches shifts (GRAD)
Financial Sector (FIN)
CreditWatch List (W)
Outlook (O)
Interest rate variation (IRV)
Economic Profitability (EP)
Size (TBTF)
Adjusted R-squared
F
F p_val
Obs
MDV
MDTV
MDTVA
2.346
( 0.444)
1.710*
( 0.100)
0.488
( 0.735)
0.277
( 0.710)
-3.218
( 0.212)
-0.759
( 0.660)
-0.120
( 0.917)
-1.329
( 0.451)
-72.529
( 0.202)
-0.605
( 0.517)
-0.069
0.664
( 0.736)
48
2.631
( 0.366)
1.221
( 0.275)
1.508
( 0.312)
-0.033
( 0.967)
-3.166
( 0.212)
-1.795
( 0.328)
0.307
( 0.801)
-0.080
( 0.970)
-68.887
( 0.216)
-0.354
( 0.727)
-0.051
0.746
( 0.665)
48
0.711
( 0.705)
0.543
( 0.297)
0.173
( 0.832)
-0.573
( 0.280)
-0.322
( 0.844)
-0.902
( 0.235)
0.478
( 0.459)
0.835
( 0.388)
-22.661
( 0.489)
0.517
( 0.406)
0.034
1.185
( 0.332)
48
MS
2.457
( 0.397)
0.687
( 0.423)
-0.116
( 0.906)
-0.249
( 0.680)
-2.213
( 0.358)
-1.573
( 0.222)
0.294
( 0.781)
0.830
( 0.551)
-48.500
( 0.369)
0.163
( 0.839)
-0.085
0.589
( 0.798)
48
Note: See note in Table 9. E3M: dummy worth 1 if the announcement has been preceded by an
announcement in the same direction in the three previous months, MS is the abnormal market share
computed as the difference between the rate of growth of the market share observed around the rating
event and the one expected according to the model (5).
43
Table 16. Determinants of excess commercial paper notes liquidity after rating upgrades: frequency measures
T-1w
Const
0.484
( 0.163)
AG
0.107
( 0.396)
E3M
-0.082
( 0.572)
NOTCH -0.035
( 0.510)
FIN
0.237
( 0.329)
W
-0.104
( 0.560)
O
-0.111
( 0.481)
IRV
-0.197
( 0.268)
EP
4.660
( 0.484)
TBTF
0.045
( 0.713)
Adj. R2 -0.126
F
0.513
F p_val ( 0.853)
Obs
40
T-2w
T-1m
T-2m
0.534*
( 0.082)
0.229*
( 0.064)
-0.160
( 0.226)
0.018
( 0.784)
-0.050
( 0.800)
-0.149
( 0.364)
-0.162
( 0.324)
-0.370**
( 0.030)
-4.381
( 0.357)
0.095
( 0.460)
0.053
1.266
( 0.290)
44
0.193
( 0.607)
0.176
( 0.159)
-0.110
( 0.455)
-0.007
( 0.925)
0.167
( 0.495)
-0.163
( 0.429)
-0.213
( 0.192)
-0.392**
( 0.023)
2.319
( 0.717)
0.022
( 0.866)
0.007
1.032
( 0.435)
45
0.107
( 0.768)
0.225*
( 0.059)
-0.049
( 0.728)
-0.020
( 0.796)
0.104
( 0.630)
-0.253
( 0.215)
-0.249
( 0.117)
-0.421**
( 0.015)
-0.738
( 0.862)
0.099
( 0.443)
0.133
1.764
( 0.110)
46
3m-1w 3m-2w 3m-1m 3m-2m 2m-1w 2m-2w 2m-1m 2m-2m 1m-1w 1m-2w 1m-1m
0.353
( 0.209)
0.041
( 0.665)
0.082
( 0.376)
-0.007
( 0.860)
0.172
( 0.397)
0.055
( 0.726)
0.102
( 0.356)
-0.069
( 0.695)
0.169
( 0.970)
0.050
( 0.634)
-0.175
0.437
( 0.902)
35
0.278
( 0.198)
0.114
( 0.257)
0.083
( 0.294)
0.033
( 0.518)
0.011
( 0.930)
-0.012
( 0.933)
0.138
( 0.273)
-0.194
( 0.269)
-5.682*
( 0.064)
0.094
( 0.451)
0.035
1.158
( 0.356)
40
0.042
( 0.880)
0.158
( 0.119)
-0.056
( 0.613)
0.037
( 0.524)
0.041
( 0.831)
-0.077
( 0.625)
-0.203
( 0.118)
-0.297
( 0.107)
1.218
( 0.737)
0.023
( 0.818)
-0.008
0.963
( 0.486)
45
-0.035
( 0.918)
0.187*
( 0.072)
0.004
( 0.975)
0.023
( 0.722)
-0.015
( 0.950)
-0.168
( 0.348)
-0.224*
( 0.086)
-0.294*
( 0.093)
-1.821
( 0.667)
0.097
( 0.355)
0.077
1.416
( 0.218)
46
0.347
( 0.318)
-0.034
( 0.766)
-0.002
( 0.986)
0.016
( 0.758)
0.110
( 0.662)
0.032
( 0.864)
0.141
( 0.370)
-0.101
( 0.621)
-0.550
( 0.930)
0.086
( 0.589)
-0.220
0.319
( 0.961)
35
0.266
( 0.295)
0.039
( 0.753)
-0.014
( 0.884)
0.040
( 0.470)
-0.026
( 0.874)
-0.054
( 0.737)
0.172
( 0.296)
-0.219
( 0.266)
-5.413
( 0.216)
0.125
( 0.469)
-0.033
0.862
( 0.568)
40
0.063
( 0.844)
0.190
( 0.104)
-0.127
( 0.288)
0.031
( 0.607)
0.013
( 0.956)
-0.146
( 0.422)
-0.271*
( 0.068)
-0.330
( 0.101)
2.194
( 0.634)
-0.052
( 0.654)
0.001
1.004
( 0.455)
45
-0.011
( 0.976)
0.211*
( 0.073)
-0.067
( 0.564)
0.017
( 0.801)
-0.040
( 0.873)
-0.238
( 0.228)
-0.287*
( 0.056)
-0.313*
( 0.093)
-0.837
( 0.855)
0.021
( 0.859)
0.050
1.265
( 0.289)
46
0.799**
( 0.004)
0.103
( 0.250)
-0.043
( 0.683)
-0.011
( 0.796)
-0.402**
( 0.031)
-0.185
( 0.230)
-0.185
( 0.124)
-0.242
( 0.134)
-9.412**
( 0.025)
-0.046
( 0.613)
-0.052
0.786
( 0.631)
40
0.584**
( 0.044)
0.187*
( 0.068)
-0.175*
( 0.089)
0.019
( 0.676)
-0.428**
( 0.043)
-0.226
( 0.156)
-0.220*
( 0.083)
-0.314*
( 0.099)
-9.257
( 0.106)
-0.017
( 0.863)
0.078
1.402
( 0.226)
44
0.271
( 0.346)
0.130
( 0.252)
-0.128
( 0.247)
0.004
( 0.944)
-0.246
( 0.183)
-0.233
( 0.208)
-0.277*
( 0.050)
-0.334
( 0.111)
-3.445
( 0.302)
-0.079
( 0.463)
0.013
1.063
( 0.413)
45
Note: See note in Table 9. E3M: dummy worth 1 if the announcement has been preceded by an announcement in the same direction in the three previous months. Abnormal
frequencies calculated as the difference of the logarithm of the mean trading frequency of a pre-event window (PREW) with respect to a post-event window (POSTW): PREWPOSTW., where PREW= 1, 2 , 3 months and the entire period since the issue (1m, 2m, 3m, T) and POSTW= 1 & 2 weeks and 1 & 2 months (1w, 2w, 1m, 2m).
44