Professional Documents
Culture Documents
a r t i c l e
i n f o
Article history:
Received 8 September 2011
Accepted 15 March 2012
Available online 23 March 2012
JEL classication:
G15
G12
F36
a b s t r a c t
Global bonds are international securities traded and settled efciently in multiple markets. This paper
examines global bonds to evaluate the effects of multimarket trading on corporate bond liquidity and
pricing. The results show that global bonds are signicantly more liquid than similar-sized domestic
bonds of the same issuers, and their liquidity advantage is reected in higher market valuations. These
ndings support microstructure models that predict a positive relation between the number of potential
investors and liquidity in over-the-counter markets, and help explain the increasing use of global bonds
by corporate issuers.
2012 Elsevier B.V. All rights reserved.
Keywords:
Corporate bonds
Liquidity
International nancial markets
1. Introduction
Rather than issuing bonds in domestic markets, many of the
worlds largest corporations have recently started to issue global
bonds. Global bonds are offered simultaneously to investors in
the two major markets for dollar-denominated debt, namely the
US bond market and the overseas market known as the Eurobond
market. Unlike domestic bonds, global bonds are designed for multimarket trading. These bonds include features that facilitate their
trading, clearing, and settlement in the US bond market, the Eurobond market, as well as between markets. Global bonds have become the debt instrument of choice for large corporate issuers in
recent years, and corporate global bond issuance accounted for
more than 80% of the total US-dollar denominated corporate bond
issuance in 2009 (see Fig. 1).
Despite the increasing importance of global bonds, the effects of
multimarket trading on corporate bond liquidity and pricing are
not well understood. Miller and Puthenpurackal (2005) study global bond offerings from the issuers perspective, and nd that rms
can lower their cost of debt by issuing global rather than domestic
bonds. However, the benets of global bonds from the investors
perspective have not been previously documented. Not much is
known about the trading, liquidity, and pricing of global bonds in
secondary markets, and the effects of their multimarket trading
on liquidity. The present paper provides new evidence from secondary markets on the effects of multimarket trading on corporate
See, e.g. Longstaff et al. (2005), Chen et al. (2007), and Bao et al. (2011).
TRACE is the Trade Reporting and Compliance Engine, sponsored by the US
Financial Industry Regulatory Authority (FINRA).
2
100%
80%
60%
40%
20%
0%
98
99
00
01
02 03 04 05
Year of issuance
Global bonds
06
07
08
09
Fig. 1. Corporate issuance of global and US bonds (% of debt issued). The gure plots
the proportions of global and domestic issuance of US dollar-denominated
corporate debt from 1998 to 2009. Global bonds are designed for trading in the
US bond market and in the international market known as the Eurobond market,
whereas US bonds are designed for trading only in the US domestic bond market.
The gure includes all public issues of straight corporate debt denominated in US
dollars and registered in the US The source of the data is SDC.
2111
structure models that predict a positive relation between the number of potential investors and liquidity in over-the-counter markets (e.g. Dufe et al., 2005, 2007). Furthermore, the liquidity
advantage of global bonds is priced, and can explain the tendency
of large corporations to issue global rather than domestic bonds.
The remainder of the paper is organized as follows. Section 2
provides a brief description global bonds and their multimarket
trading. Section 3 develops testable hypothesis about the relation
between multimarket trading and corporate bond liquidity. Section 4 describes the data. Section 5 shows the empirical results,
including the analysis of transaction prices and liquidity. Section
6 provides concluding remarks.
2112
bonds and neither US nor foreign investors can gain a tax advantage by investing in global bonds.7 However, global bonds are sold
to investors in multiple markets and their features allow both US and
Eurobond investors to trade these bonds as if they were local securities. Global bonds are structured to clear through several clearing
systems, including DTC and Clearstream or (and) Euroclear. Consequently, US investors settle global bond transactions through DTC
as they would settle domestic bond transactions, and overseas investors settle global bond transactions through Clearstream or Euroclear
as they would settle Eurobond transactions. Cross-market trades in
global bonds are efciently settled between the depositories. Thus,
non-US investors are more likely to trade global bonds than US
domestic bonds, and global bonds have a larger investor base.
Although the rst global bonds were issued by the World Bank,
global bonds have recently become the debt instrument of choice
for many large corporate issuers. Fig. 1 compares the proportions
of global and domestic issuance of US dollar-denominated corporate bonds from 1998 to 2009. Included are all public issues of
straight corporate debt denominated in US dollars and registered
for public trading in the US. The gure reveals a large increase in
the issuance of global bonds compared to US domestic bonds over
time. Whereas global bonds accounted for less than 10% of the total
proceeds from publicly traded, US-dollar denominated corporate
bond issuance in 1998, they accounted for more than 80% in 2009.
3. Multimarket trading and liquidity
The most important channel through which multimarket trading can increase corporate bond value is liquidity. Global bonds
could be more liquid because they have a greater number of dealers and a wider investor base. For one thing, both the dealers in the
US market and the dealers in the Eurobond market stand ready to
buy and sell global bonds. The microstructure models of Demsetz
(1968) and Ho and Stoll (1983), among others, predict a positive
relation between the number of dealers making the market in a given security and its liquidity. A greater number of dealers leads to
more competitive dealer markets (Demsetz, 1968), and facilitates
inventory risk management through inter-dealer trading (Ho and
Stoll, 1983; Reiss and Werner, 1998). Lower inventory risk and
more intense competition among market makers, in turn, reduce
transaction costs.
Liquidity may also be greater for bonds with a wider investor
base. Since global bonds can be marketed to investors in both the
US market and the Eurobond market, their pool of potential investors is larger. Dufe et al. (2005, 2007) develop a model in which
transactions costs and liquidity in over-the-counter markets, such
as the corporate bond market, depend on the number of potential
investors. Their model predicts that illiquidity discounts are smaller if investors have access to multiple market makers, and the
number of qualied investors is larger. Thus, global bonds that
have an international investor base and a large network of dealers
should be more liquid than domestic issues.
In addition, global bonds are often placed with investors in multiple markets, allowing them to have a greater issue size. However,
prior research on the relation between bond issue size and its
liquidity is inconclusive. On the one hand, Longstaff et al. (2005),
Edwards et al. (2007), and Mahanti et al. (2008) nd a positive relation between issue size and liquidity. On the other hand, Crabbe
and Turner (1995) contend that large and small bonds issued by
the same borrowers are close substitutes. Chen et al. (2007) nd
little evidence of the importance of the outstanding principal
7
Global and domestic bonds held by US residents are taxed equally. Foreign
investors in either global or domestic bonds issued after 1984 are exempt from the US
withholding tax (see IRS Publication 515 (2011), Withholding of Tax on Nonresident
Aliens and Foreign Entities).
Table 1
Characteristics of global and domestic bond issuers. The table compares the
characteristics of 219 global bond issuers with the characteristics of 523 rms that
issue bonds only in the US domestic market. Presented are the averages across global
and domestic issuers, and the differences between the characteristics of global and
domestic issuers. The data are obtained from COMPUSTAT. Total assets and market
capitalization are measured at the end of each scal year. Leverage is measured as
short-term and long-term debt divided by end-of-year total assets. Book-to-market is
calculated as the book value of equity divided by market capitalization at the end of
each scal year. Foreign issuers are rms incorporated outside the United States,
nancial rms have SIC codes in the 6000s, and utilities in the 4900s.
**
Global
issuers
Domestic
issuers
Difference
127.92
41.26
0.30
0.55
22
26
21
29.97
10.08
0.33
0.67
20
25
21
97.95**
31.18**
0.04**
0.12**
2
1
0
No. of issuers
219
523
2113
Table 2
Summary statistics for the global and domestic bond sample. The table reports summary statistics for the sample of global and domestic bonds of global bond issuers. Information
on new bond issues, including bond type, total principal issued, and bond placements, is from the SDC data base. Total principal is the principal amount placed in all markets.
Foreign principal as % of total principal shows the percentage of the total principal issued that is initially placed with non-US investors. Bond characteristics and bond rating
history are from the FISD database. Trading yields and yield spreads are derived from secondary market transaction prices reported on TRACE. The numbers reported in the table
are the means (medians) of individual bond averages.
Sample mean
*
**
Sample median
Global
Domestic
Difference
Global
Domestic
Difference
1652.42
32.37
5.95
2.28
10.51
1.37
5.72
63
85
712.08
10.54
6.10
2.43
10.58
3.23
5.85
68
84
940.34**
21.83**
0.16*
0.15*
0.07
1.86**
0.13*
4
1
1197.25
39.44
5.79
1.87
7.82
0.86
5.73
100
100
462.05
0.00
5.93
2.05
7.05
2.66
5.77
100
100
735.20**
39.44**
0.14*
0.18**
0.77
1.80**
0.04
0
0
736
480
135
456
450
135
280**
296
480
135
99
450
135
197**
Inc. Overall, 2175 global and US bonds were issued by 219 global
bond issuers between January 1998 and December 2008.
The nal sample is comprised of a subset of bonds that are actively traded in US secondary markets. The source of secondary
market prices for global and US bonds is the Trade Reporting and
Compliance Engine, commonly referred to as TRACE. Since June
2002, all broker-dealers active in the US market have an obligation
to report transactions in publicly registered corporate bonds to
TRACE. Since global bonds are like US public bonds registered with
the SEC, they are subject to TRACE reporting and dissemination.
Eurobonds, in contrast, are not subject to TRACE reporting, and
therefore are not included in the sample. The information disseminated by TRACE includes bond CUSIP, transaction price, date, time,
and volume. I complement this data with bond descriptive information and bond ratings from the Fixed Income Securities Database (FISD).
The sample of secondary market transactions starts in 2003, the
rst full year of TRACE reporting. It is comprised of a subset of actively traded bonds. Specically, the analysis requires data on both
global and domestic bond transactions from the same day. Therefore, a global (domestic) bond is required to trade on the same
day as another domestic (global) bond of the same issuer to enter
the sample. Further, sample bonds are required to have at least
5 years to maturity at the time of trading. There are 930 such bonds
issued by 135 issuers; 480 are global bonds and 450 are US domestic bonds.
Several lters are applied to the sample trades. First, global
(domestic) bond trades must be matched by issuer and trading
day with domestic (global) bond trades to be included in the sample. Further, only institutional-sized trades of 100,000 dollars or
larger are examined. Retail-sized trades are discarded because
their transaction costs account for a non-negligible percentage of
the traded price (Edwards et al., 2007). Nevertheless, all trades
are considered when calculating measures of corporate bond
liquidity. In addition, the data are purged of trade reports that were
subject to cancellations or corrections, are missing key information, include commission in the price, or were entered by multiple
dealers.8 Next, I remove 3321 trades that appear to be outliers. The
outliers are identied on the basis of their relative distance from a
8
Dick-Nielsen (2009) argues that transactions with zero returns for consecutive
trades are the result of double reporting in interdealer trading. Failing to remove
these duplicates could bias the liquidity measures. Following Dick-Nielsen (2009), I
remove all reports with zero returns for consecutive trades.
9
Observations that are more than three standard deviations away from a
neighborhood of the 30 nearest valid observations are considered to be outliers.
The tests are not sensitive to the inclusion of these observations.
10
The daily yield curve for constant maturities of 2, 5, 10, and 30 years is obtained
from the US Department of Treasury. Linear interpolation is used for each month
within the intermediate maturities, and the yield curve is assumed to be at beyond
30 years.
11
The averages are rst calculated for each bond across all sample trades and then
across bonds.
2114
Table 3
Yield spreads between global and domestic bonds: paired t-tests. The table reports the differences in trading yields between matched pairs of global and domestic bonds. All
bonds are non-callable and matched by issuer. In addition, the bonds in Panel A mature within 2 years from one another. The bonds in Panel B have an issue size within 10% from
one another. The bonds in Panel C are matched by both maturity and size. The yield difference for each bond pair is calculated from all transactions that occur on the same days.
Shown is the number of matched pairs, the mean difference in trading yield spreads between global and domestic bonds across the matched pairs, the t-statistic for the mean
difference, the median difference in trading yield spreads, and the percentage of bond pairs for which the difference is negative.
No. of matched bond pairs
*
**
Mean difference
T-statistic
Median difference
0.22**
0.21**
0.61*
5.31
5.19
1.82
0.09**
0.09**
0.24*
73
75
64
0.13**
0.10**
0.92**
3.43
3.15
2.17
0.07**
0.07**
0.86**
62
63
62
0.20**
0.18**
1.02
3.44
3.47
1.63
0.08**
0.08**
0.86
69
73
62
Mean (median) differences marked with are signicant at the 10% levels according to the t-test (Wilcoxon signed rank test).
Mean (median) differences marked with are signicant at the 5% levels according to the t-test (Wilcoxon signed rank test).
bal bonds tend to have a lower age (1.37 years) than domestic
bonds (3.23 years) because rms are less likely to offer domestic
bonds after issuing global bonds (see Fig. 1).
5. Empirical analysis
5.1. Global and domestic bond yields
Prior research shows that global bond offerings are associated
with lower yields and borrowing costs than comparable domestic
offerings (Miller and Puthenpurackal, 2005). However, different offer yields of global and domestic bonds do not necessarily imply
that secondary market prices also differ. The prices of new corporate bond issues could also be affected by transitory factors such as
underpricing (Cai et al., 2007) or issuance price pressure (Newman
and Rierson, 2004).
To examine whether investors place higher valuations on global
bonds, this section compares the yields to maturity of global and
domestic bonds in secondary markets. When global and domestic
bonds of one issuer trade on the same day, their trading yields reect the same credit risk. However, the bonds may differ in their
size, maturity, or embedded call options. I use two methods to control for these characteristics. First, I create a matched sample of
bond pairs that have no call options and are similar in terms of
issue size and maturity. I conduct paired t-tests and Wilcoxon
signed rank tests to examine if these bonds trade at the same
yields. The second method is to estimate panel regressions that relate yield differentials to differences in bond characteristics.
Matched sample estimation has been widely used to test for differences in bond prices and transaction costs. I apply the matched
sample approach to test for differences in yields to maturity and
liquidity between global and domestic bonds. Specically, global
and domestic bonds that are non-callable and issued by the same
rms are matched into pairs by term to maturity and/or issue size.
For each pair, I calculate the average difference in trading yields
between the global and the domestic bond using only trades that
occur on the same days. I then compute the mean and median difference across all bond pairs and perform t-tests for the signicance of the mean difference. I also test for differences between
the matched pairs using the non-parametric Wilcoxon signed rank
test.
Panel A of Table 3 presents the test results for bonds matched
by issuer and term to maturity. Specically, the pairs are made
up of one global bond and one domestic bond that mature within
2 years from one another. There are 137 such bond pairs, made up
of 139 different bonds12 issued by 23 rms. As Panel A of Table 3
shows, global bonds trade on average at yields 22 basis points below
the yields on domestic bonds of the same issuers with similar maturities. The average difference in yields between all pairs of global and
domestic bonds is negative and statistically signicant at the 5% level using a paired t-test. The median yield difference of 9 basis
points is also statistically signicant using a Wilcoxon signed rank
test. The differences are negative for 73% of the bond pairs.
When the yields of investment grade bonds (132 pairs) and
speculative grade bonds (25 pairs)13 are examined separately, the
difference is greater for speculative grade bonds than for investment
grade bonds. The average difference between global and domestic
bonds is 21 basis points for investment grade bonds compared to
61 basis points for speculative grade bonds. Prior studies show that
illiquidity has a larger impact on the yields of speculative grade
bonds than on investment grade bonds (see, e.g. Longstaff et al.,
2005; Chen et al., 2007; Mahanti et al., 2008). Therefore, the nding
of larger yield differences for speculative grade bonds is consistent
with the liquidity explanation of the spread between global and
domestic bonds.
Panel B of Table 3 reports the results if bonds are matched by
issuer and issue size. The bond pairs are selected to have an issue
size within 10% from one another. There are 139 such pairs, made
up of 121 bonds issued by 17 rms. The average yield spread between the global and domestic bonds is 13 basis points, statistically signicant at the 5% level using a paired t-test or a Wilcoxon
signed rank test. The spread is larger for speculative grade bonds
than for investment grade bonds. Finally, in Panel C of Table 3,
bonds of the same issuers are matched by both time to maturity
and issue size. There are 59 possible bond pairs, made up of 78 different bonds issued by 16 rms. The mean yield spread between
global and domestic bonds is 20 basis points, and the median
spread is 8 basis points. The mean and median yield spread between global and domestic bonds is negative and signicant at
the 5% level.
Bond traders tend to regard the Treasury zero curve as the riskfree zero curve and measure a corporate bond yield spread as the
spread of the corporate bond yield over the yield on a similar gov-
12
One global (domestic) bond can match with two or more domestic (global) bonds.
Most speculative bonds are issued as investment grade bonds and enter the
speculative sample later due to downgrades so the same bond can be in both groups
at different times.
13
2115
Sp g
readijd
g b0 Con trols
g
b0 b1 Gl obal
ijd
ijd eijd ;
2
Table 4
Yield spreads between global and domestic bonds: panel regressions with issuer/day
xed effects. The table presents estimates from regressions of trading yield spreads on
an indicator variable that equals one for global bonds, the interaction of the global
bond dummy with the speculative grade indicator variable, and several control
variables. Maturity is the time to maturity and Age is the bond age, both measured in
years at the time of trading. Principal is the natural logarithm of the outstanding bond
principal. Call option is an indicator variable equal to one if the bond is callable.
Liquidity crisis is a dummy variable that equals one if the spread between the 3month LIBOR and the 3-month Treasury bill rate exceeds 2%. The regressions are
estimated using panel data transformed by subtracting issuer/day xed effects.
Robust t-statistics adjusted for clustering by bond are in parentheses.
Global bond
(2)
(3)
0.233**
(4.42)
0.104**
(5.59)
0.002**
(4.56)
0.029**
(2.95)
0.037
(0.70)
0.211**
(2.30)
0.109**
(4.08)
0.487**
(2.81)
0.094**
(6.07)
0.002**
(4.97)
0.036**
(3.81)
0.041
(0.77)
0.173*
(1.93)
0.088**
(3.05)
0.458**
(2.65)
0.094**
(6.15)
0.002**
(5.04)
0.034**
(3.66)
0.040
(0.78)
0.177**
(2.07)
0.378**
(3.36)
0.18
558,362
930
135
0.20
558,362
930
135
0.21
558,362
930
135
Speculative global
Maturity
Maturity squared
(1)
Age
Principal
Call option
Liquidity crisis global
Adj. R2
No. of trades
No. of bonds
No. of issuers
*
**
2116
et al., 2007; Edwards et al., 2007; Bao et al., 2011). Finally, I exclude
bonds issued by nancial rms and utilities. In all cases, I nd that
the subsample results are not substantively different from those
using the entire sample.
5.2. Measures of corporate bond liquidity
A major argument in favor of global bond issuance is greater
liquidity. Liquidity has multiple dimensions, and many alternative
measures of corporate bond liquidity have been proposed in the
literature.14 I measure liquidity in several ways, including trading
volume, trading frequency, zero volume days, price volatility, the
price impact of trades, and the autocovariance of daily returns.
The liquidity measures use data from TRACE for the US bond
market, complemented with additional information from TRAX
for the Eurobond market. The monthly trading volume in the US
bond market is computed using all trades on TRACE over a period
of 20 preceding trading days. For condentiality reasons, TRACE
does not disseminate the exact trading volume for transactions larger than the cap value of $5 million ($1 million for high yield
bonds). The actual transaction volume is therefore assumed to be
equal to the cap value. The monthly trading volume for the Eurobond market is obtained by multiplying the average daily volume
for the previous month provided by TRAX by 20. In contrast to
reporting to TRACE, which is mandatory for all broker-dealers registered in the US, only ICMA members have an obligation to report
their trades to TRAX. However, most Eurobond dealers are members of the ICMA. There is also very little overlap between ICMA
member rms and US bond dealers.15 Therefore, TRAX volume is
a reasonable proxy for trading activity in the Eurobond market.
I compute three measures of trading frequency from TRACE: the
monthly number of trades, the monthly number of large trades,
and the monthly number of zero-volume days counting the days
on which no large trades occur. Large trades are dened as transactions of 100,000 dollars or greater. Smaller, retail-sized trades
may not be indicative of greater liquidity (see, e.g. Edwards et al.,
2007). All the measures of trading frequency are counted over
the last 20 trading days preceding each observation.
Price volatility is related to liquidity through dealer inventory
risk (see, e.g. Stoll, 1978). I compute two volatility measures: the
monthly price range as a percentage of the average price, and the
coefcient of variation of the transaction price. The coefcient of
variation is the monthly standard deviation scaled by the average
price. Both measures are computed from intraday price changes
over the period of 20 trading days preceding an observation, and
only large trades are used in their calculation.
Further, I consider two return-based liquidity measures derived
from daily closing prices: the Amihud ratio and the Gamma measure (c). Let Pt denote the closing price on day t, adjusted for the
interest accrued since the last coupon date. Closing prices are obtained from the last large transaction on each day. For bonds trading on two consecutive business days, I calculate the daily
percentage return as rt = 100 (ln Pt ln Pt1). Both the Amihud
illiquidity and c are computed only for the subset of bonds with
15 or more daily return observations over a 20-day period. There
are 528 such bonds with 355,357 transactions, and the analysis
of return-based liquidity measures is conducted on this subsample.
Liquidity can be dened as the ability to buy or sell an asset in
large quantity quickly and without affecting the market price. To
measure illiquidity as the price impact of trading, I calculate the ratio proposed by Amihud (2002):
14
See, e.g., Houweling et al. (2005), Chen et al. (2007), and Bao et al. (2011).
The ICMA currently has about 400 members in 50 countries, and its TRAX system
covers the greater part of the Eurobond market. Only three ICMA member rms are
domiciled in the United States.
15
Amihud
Absrt
;
Vt
2117
Table 5
Summary statistics for liquidity measures. The table presents summary statistics for several transaction-based measures of corporate bond liquidity. Trading volume is the
monthly trading volume, reported separately for the US market and the overseas (Eurobond) market. Other liquidity measures are based on data from the US bond market. Large
trades are transactions of at least 100,000 dollars. Price range is scaled by the average price. Coefcient of variation is the standard deviation of prices scaled by the average price.
Amihud illiquidity is dened as the absolute value of daily returns divided by volume in millions of dollars. Gamma is the negative of the autocovariance in daily returns. Only
large trades are used to compute the number of zero volume days, price range, coefcient of variation, Amihud illiquidity, and Gamma. Gamma and Amihud illiquidity are
available for a subsample of 528 bonds.
Sample mean
**
Sample median
Global
Domestic
Difference
Global
Domestic
Difference
94.85
29.45
186.23
59.07
8.11
6.59
1.80
480
0.67
0.26
32.92
13.28
79.93
21.92
13.63
9.64
2.88
450
0.87
0.66
61.93**
16.17**
106.30**
37.15**
5.52**
3.05**
1.08**
60.37
11.78
104.36
40.23
7.57
4.93
1.34
480
0.59
0.09
13.45
0.63
29.76
10.05
15.00
6.02
1.92
450
0.74
0.29
46.92**
11.15**
74.60**
30.18**
7.43**
1.09**
0.58**
374
154
374
154
0.20**
0.40**
0.15**
0.20**
Table 6
Liquidity differences between global and domestic bonds: paired t-tests. The table reports the differences in liquidity between matched pairs of global and domestic bonds. All
bonds are non-callable, matched by issuer, and have an issue size within 10% from one another. The liquidity differences for each pair are calculated from all transactions that
occur on the same days. Shown is the number of matched bond pairs, the liquidity of the global and domestic bonds, the mean difference in liquidity between global and domestic
bonds across the matched pairs, and the t-statistic for the difference.
*
**
Liquidity measure
139
139
139
139
139
139
139
91
91
78.53
29.63
165.19
52.71
7.63
5.30
1.44
0.64
0.29
139
1011.30
Domestic bond
liquidity
T-statistic for
difference
58.88
15.64
137.94
42.68
9.47
6.06
1.82
0.78
0.46
19.65**
13.99**
27.25**
10.03**
1.84**
0.76*
0.38**
0.14**
0.18*
2.63
3.56
2.31
2.92
5.09
1.82
2.99
2.57
1.92
1008.35
2.95
0.91
domestic bonds matched by issuer and issue size. All sample bonds
are non-callable, matched by issuer, and have an issue size within
10% from one another. The sample is the same as that used in the
analysis of yields spreads in Panel B of Table 3. Although matching
by issue size typically reduces the liquidity differences between
global and domestic bonds (compare with Table 5), the differences
remain large and statistically signicant. For example, the monthly
trading volume of domestic bonds in the US bond market increases
to 59 million dollars after matching domestic bonds by size with
global bonds, but remains signicantly below that of global bonds
of 79 million dollars. Similarly, the number of zero volume days for
domestic bonds decreases to 9.5 after matching, which is signicantly more than the 7.6 zero volume days for global bonds. The
measures of price volatility and the return-based liquidity measures also indicate that global bonds are signicantly more liquid
in the matched sample analysis.
Alexander et al. (2000) and Edwards et al. (2007), among others, show that trading in corporate bonds is abnormally high in
the rst few months after issuance. Recently issued bonds could
therefore appear more liquid than seasoned bond issues. To address this concern, I exclude from the matched sample comparison
all newly issued bonds for 6 months after issuance. Table 7 reports
the results. Compared to Table 6, all bonds in Table 7 are less liquid, regardless of whether they are global or domestic bonds.
2118
Table 7
Liquidity differences between seasoned global and domestic bonds: paired t-tests. The table reports the differences in liquidity between matched pairs of seasoned global and
domestic bonds. The sample includes bonds seasoned for more than 6 months after issuance. All bonds are non-callable, matched by issuer, and have an issue size within 10%
from one another. The liquidity differences for each pair are calculated from all transactions that occur on the same days. Shown is the number of matched bond pairs, the
liquidity of the global and domestic bonds, the mean difference in liquidity between global and domestic bonds across the matched pairs, and the t-statistic for the difference.
*
**
Liquidity measure
Domestic bond
liquidity
125
125
125
125
125
125
125
63
63
61.36
22.54
122.74
46.23
8.51
7.25
2.01
0.77
0.36
44.46
7.45
93.30
33.98
10.42
8.02
2.35
1.06
0.55
125
990.37
985.20
T-statistic for
difference
16.90**
15.08**
29.44**
12.25**
1.91**
0.78
0.34*
0.29**
0.19*
2.86
6.97
4.54
4.21
4.91
1.32
1.87
3.26
1.89
5.16
0.99
Table 8
Liquidity differences between global and domestic bonds: panel regressions with issuer/day xed effects. The table presents estimates from regressions of several measures of
corporate bond liquidity on a dummy variable that equals one if the bond is global, and the interaction of the global bond dummy with the speculative grade indicator variable.
Also included are control variables for bond age, principal, time to maturity, and embedded call options. The dependent variables are the log-transformed (except for zero-volume
days) liquidity measures. The regressions are estimated on panel data transformed by subtracting issuer/day xed effects. Robust t-statistics adjusted for clustering by bond are in
parentheses.
Dependent variable
Global bond
Speculative global
Age
Principal
Maturity
Maturity squared
Callable bond
Adj. R2
No. of trades
No. of bonds
No. of issuers
*
**
US trading
volume
Overseas trading
volume
No. of
trades
No. of large
trades
Zero volume
days
Price
range
Coeff. of
variation
Amihud
illiquidity
Gamma
0.187**
(2.92)
0.23
(1.29)
0.138**
(8.56)
0.616**
(10.06)
0.013
(0.40)
0.001
(0.28)
0.098
(1.02)
0.502**
(2.49)
0.62
(1.28)
0.369**
(7.46)
1.468**
(8.29)
0.013
(0.14)
0.001
(0.19)
0.198
(0.73)
0.070
(1.39)
0.20
(1.07)
0.021
(1.27)
0.455**
(8.60)
0.029
(1.49)
0.001
(1.24)
0.246**
(2.79)
0.112**
(2.46)
0.04
(0.21)
0.080**
(5.76)
0.486**
(9.31)
0.012
(0.53)
0.001
(0.94)
0.155*
(1.91)
1.224**
(6.01)
0.13
(0.29)
0.192**
(3.18)
1.979**
(9.74)
0.140**
(2.66)
0.003*
(1.97)
0.400
(1.46)
0.003**
(3.60)
0.01
(0.40)
0.037**
(8.24)
0.002*
(1.72)
0.005**
(11.56)
0.001**
(9.48)
0.005**
(2.55)
0.020**
(3.55)
0.02**
(5.54)
0.011**
(3.17)
0.037**
(8.24)
0.055**
(20.34)
0.001**
(14.95)
0.031
(1.43)
0.071**
(2.31)
0.28**
(5.54)
0.039**
(5.26)
0.115**
(4.59)
0.029**
(2.29)
0.001**
(2.42)
0.045*
(1.69)
0.044**
(5.22)
0.12**
(6.97)
0.008**
(3.91)
0.004
(0.61)
0.009**
(2.70)
0.001**
(2.25)
0.015*
(1.87)
0.27
558,362
930
135
0.20
558,362
930
135
0.16
558,362
930
135
0.31
558,362
930
135
0.19
558,362
930
135
0.18
558,362
930
135
0.31
558,362
930
135
0.23
355,357
528
98
0.20
355,357
528
98
good proxy for bond liquidity since the majority of trades are retail-sized. The number of large trades, which likely better measures
liquidity, is positively related to global issuance. Additionally, the
number of zero volume days is signicantly smaller for global
bonds than for domestic bonds.
The return-based liquidity measures in Table 8 also indicate
that global securities are more liquid. Measures of price volatility,
the price range and the coefcient of variation, are lower for global
bonds. The Amihud illiquidity, which is a proxy for the price impact of trading, is signicantly lower for global bonds, in particular
if they are rated below investment grade. The measure of price
reversals (c) is strongly negatively related to the global bond dummy variable, and the effect is larger for speculative grade bonds. By
almost all measures, corporate bonds become less liquid as they
age and settle in investors portfolios, and larger issues are more liquid than smaller ones. However, differences in bond characteristics such as issue size, age, or maturity do not subsume the
signicance of global issuance for corporate bond liquidity.
2119
Global bond
Speculativeglobal
Maturity
Maturity squared
Age
Principal
Call option
(1)
(2)
(3)
0.135**
(3.04)
0.488**
(2.49)
0.106**
(5.26)
0.003**
(4.18)
0.053**
(3.76)
0.075
(1.01)
0.199*
(1.87)
0.087**
(2.16)
0.399**
(2.60)
0.025
(0.97)
0.001
(1.20)
0.037**
(3.09)
0.026
(0.37)
0.137
(1.55)
0.160**
(3.25)
0.428**
(6.18)
0.044**
(6.16)
1.569
(1.06)
1.168**
(4.12)
0.045
(1.25)
0.229
(1.55)
0.010
(0.44)
0.001
(0.72)
0.030**
(3.02)
0.001
(0.02)
0.123
(1.45)
0.064
(1.29)
0.405**
(5.73)
0.042**
(5.69)
1.702
(1.11)
1.028**
(3.73)
0.755**
(7.29)
0.891**
(4.09)
0.40
355,357
528
98
Trading volume
No. of trades
Zero volume days
Price range
Coefcient of variation
Amihud illiquidity
Gamma
Adj. R2
No. of trades
No. of bonds
No. of issuers
*
**
0.24
355,357
528
98
0.35
355,357
528
98
Table 10
The effect of taxes on the yield spreads between global and domestic bonds: Panel
regressions with issuer/day xed effects. The table presents estimates from regressions of trading yield spreads on an indicator variable that equals one for global
bonds, the interaction of the global bond dummy with the speculative grade indicator
variable, and several control variables including the coupon rate to account for the
effect of taxes. Maturity is the time to maturity and Age is the bond age, both
measured in years at the time of trading. Principal is the natural logarithm of the
outstanding bond principal. Call option is an indicator variable equal to one if the
bond is callable. Coupon is the coupon rate of the bond. Liquidity crisis is a dummy
variable that equals one if the spread between the 3-month LIBOR and the 3-month
Treasury bill rate exceeds 2%. The regressions are estimated using panel data
transformed by subtracting issuer/day xed effects. Robust t-statistics adjusted for
clustering by bond are in parentheses.
Global bond
(1)
(2)
(3)
0.224**
(4.30)
0.109**
(5.51)
0.002**
(4.56)
0.027**
(2.94)
0.045
(0.85)
0.212**
(2.29)
0.080**
(3.80)
0.102**
(4.05)
0.482**
(2.79)
0.091**
(5.95)
0.002**
(4.97)
0.035**
(3.83)
0.048
(0.92)
0.175*
(1.94)
0.077**
(3.59)
0.081**
(2.95)
0.452**
(2.62)
0.091**
(6.04)
0.002**
(5.04)
0.032**
(3.68)
0.047
(0.94)
0.178**
(2.08)
0.078**
(3.78)
0.382**
(3.40)
0.19
558,362
930
135
0.21
558,362
930
135
0.22
558,362
930
135
Speculative global
Maturity
Maturity squared
Age
Principal
Call option
Coupon
Liquidity crisis global
Adj. R2
No. of trades
No. of bonds
No. of issuers
*
**
2120
6. Conclusion
Large multinational corporations increasingly raise funds by
issuing global bonds. Global bonds resemble US domestic bonds,
but their distinctive features allow global bonds to be traded in
multiple markets. They are placed simultaneously with US and
overseas investors, and can be traded in the US bond market and
the Eurobond market, as well as between markets. However, the
effects of multimarket trading on corporate bond value are not well
understood. This paper examines how multimarket trading affects
corporate bond liquidity and prices in secondary markets.
The results conrm the hypothesis that multimarket trading
improves corporate bond liquidity. Compared to domestic bonds
issued by the same rms, global bonds are more liquid. They exhibit greater trading volumes in the US bond market and in the Eurobond market, trade more frequently, and their prices are less
volatile. Furthermore, the price impact of large trades is signicantly reduced for global bonds, and transitory price movements
that lead to serially correlated price changes are smaller. The
liquidity advantage of global bonds persists even after controlling
for their greater issue size and other characteristics, and it appears
to be related to their multimarket trading.
These ndings are consistent with microstructure models that
predict a positive relation between the number of potential investors and liquidity in over-the-counter markets. Dufe et al. (2005,
2007) develop a model in which transactions costs and liquidity in
over-the-counter markets, such as the corporate bond market, depend on the number of potential investors. Their model predicts
that illiquidity discounts are smaller if the number of qualied
investors is greater, and investors have access to multiple dealers.
Global bond offerings increase the pool of potential bondholders to
include investors overseas. In addition, liquidity in global bonds is
provided by both US bond dealers and Eurobond dealers. Thus, global bonds that are traded in multiple markets are more liquid than
US domestic bonds.
Another important nding is that the liquidity advantage of global bonds is priced. If two bonds of the same issuer, one global and
one domestic, trade on the same day, the yield on the global bond
is on average about 20 basis points lower. The yield difference is
greater for speculative grade bonds than for investment grade
bonds, and it increases during liquidity crises. In accordance with
the liquidity hypothesis, the spread between global and domestic
bonds is closely related to the differences in liquidity. In particular,
several trade-based liquidity measures explain a large part of the
yield difference between global and domestic bonds, and up to
16% of the cross-sectional variation in yield spreads between different bonds issued by the same rms.
Overall, the results show that investors value global bonds for
their greater liquidity. The liquidity advantage of global bonds
can account for the increasing popularity of global bonds in recent
years. The ndings also contribute to our understanding of the factors affecting corporate bond liquidity, and help explain prior evidence that global bond issues reduce the cost of debt.
Acknowledgements
I thank Charles Cao, Jaewon Choi, Stefano Corradin, Laura Field,
Pascal Francois, David Haushalter, Jean Helwege, Nancy R. Mahon,
Marco Rossi, and an anonymous referee for valuable comments. I
gratefully acknowledge nancial support from the Lamfalussy Research Fellowship of the European Central Bank. This research was
also supported in part by a Doctoral Research Award from the
Smeal College of Business.
Appendix A
See Table A1.
Table A1
The top 45 issuers of sample bonds.
Parent CUSIP
Corporate parent
Sample
trades
617446
25746U
842587
025537
060505
524908
209115
345370
149123
264399
590188
87612E
026874
00206R
370442
46625H
580135
78442P
035229
125581
225401
892331
92343V
949746
126408
438516
98385X
02209S
404280
494550
12189T
244199
319963
441815
45031U
49811T
079860
136375
172967
40414L
929903
002824
136385
263534
Morgan Stanley
4055
Dominion Resources Inc
3891
Southern Co
1095
American Electric Power Inc
602
Bank of America Corp
32,974
Lehman Brothers Holdings Inc
14,585
Consolidated Edison Inc
2087
Ford Motor Co
61,734
Caterpillar Inc
6894
Duke Energy Corp
3544
Merrill Lynch & Co Inc
24,042
Target Corp
15,516
AIG
9148
AT&T Inc
2127
General Motors Corp
127,326
JPMorgan Chase & Co
12,529
McDonalds Corp
4002
SLM Corp
12,767
Anheuser-Busch Cos Inc
2040
CIT Group Inc
10,871
Credit Suisse Group
13,273
Toyota Motor Corp
2237
Verizon Communications Inc
11,215
Wells Fargo, San Francisco, CA
3923
CSX Corp
1712
Honeywell International Inc
2108
XTO Energy Inc
3280
Altria Group Inc
17,015
HSBC
8013
Kinder Morgan Energy Partners
2477
Burlington Northern Santa Fe
1923
Deere & Co
3643
First Data Corp
4653
Household International Inc
646
iStar Financial Inc
1615
AIG Life Holdings (US) Inc
3941
BellSouth Corp
11,483
Canadian National Railway Co
331
Citigroup Inc
538
HCP Inc
647
Wachovia Corp, Charlotte, NC
3034
Abbott Laboratories
8217
Canadian Natural Resources Ltd
658
DuPont
8393
Sample
bonds
Credit
rating
25
23
21
19
19
19
17
17
16
16
16
16
15
15
15
14
14
13
12
12
12
12
12
12
11
11
11
10
10
10
9
9
9
9
9
8
8
8
8
8
8
7
7
7
Inv
Inv
Inv
Inv
Inv
Inv/Spec
Inv
Inv/Spec
Inv
Inv
Inv
Inv
Inv
Inv
Inv/Spec
Inv
Inv
Inv
Inv
Inv/Spec
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv
Inv/Spec
Inv
Inv
Inv
Inv
Inv/Spec
Inv
Inv
Inv
Inv
References
Alexander, G., Edwards, A., Ferri, M., 2000. The determinants of trading volume of
high-yield corporate bonds. Journal of Financial Markets 3, 177204.
Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series
effects. Journal of Financial Markets 5, 3156.
Bao, J., Pan, J., Wang, J., 2011. The illiquidity of corporate bonds. Journal of Finance
66, 911946.
Cai, N., Helwege, J., Warga, A., 2007. Underpricing in the corporate bond market.
Review of Financial Studies 20, 20212046.
Chen, L., Lesmond, D.A., Wei, J., 2007. Corporate yield spreads and bond liquidity.
Journal of Finance 62, 119149.
Crabbe, L.E., Turner, C.M., 1995. Does the liquidity of a debt issue increase with its
size? Evidence from the corporate bond and medium-term note markets.
Journal of Finance 50, 17191734.
Demsetz, H., 1968. The cost of transacting. Quarterly Journal of Economics 82, 3353.
Dick-Nielsen, J., 2009. Liquidity biases in TRACE. Journal of Fixed Income 19, 4355.
Dufe, D., Garleanu, N., Pedersen, L.H., 2005. Over-the-counter markets.
Econometrica 73, 18151835.
Dufe, D., Garleanu, N., Pedersen, L.H., 2007. Valuation in over-the-counter markets.
Review of Financial Studies 20, 18651900.
Edwards, A.K., Harris, L.E., Piwowar, M.S., 2007. Corporate bond market transaction
costs and transparency. Journal of Finance 62, 14211450.
Ho, T., Stoll, H.R., 1983. The dynamics of dealer markets under competition. Journal
of Finance 38, 10531074.
Houweling, P., Metnik, A., Vorst, T., 2005. Comparing possible proxies of corporate
bond liquidity. Journal of Banking and Finance 29, 13311356.
Hull, J., Predescu, M., White, A., 2004. The relationship between credit default swap
spreads, bond yields, and credit rating announcements. Journal of Banking and
Finance 28, 27892811.
2121
Kapur, D., Lewis, J.P., Webb, R., 1997. The World Bank: Its First Half Century.
Brookings Institution Press, New York.
Liu, S., Shi, J., Wang, J., Wu, C., 2007. How much of the corporate bond spread is due
to personal taxes? Journal of Financial Economics 85, 599636.
Longstaff, F.A., Mithal, S., Neis, E., 2005. Corporate yield spreads: default risk or
liquidity? New evidence from the credit default swap market. Journal of Finance
60, 22132253.
Mahanti, S., Nashikkar, A., Subrahmanyam, M., Chacko, G., Mallik, G., 2008. Latent
liquidity: a new measure of liquidity, with an application to corporate bonds.
Journal of Financial Economics 88, 273298.
Merton, R., 1974. On the pricing of corporate debt: the risk structure of interest
rates. Journal of Finance 29, 449470.
Miller, D.P., Puthenpurackal, J.J., 2005. Security fungibility and the cost of capital:
evidence from global bonds. Journal of Financial and Quantitative Analysis 40,
849872.
Newman, Y.S., Rierson, M.A., 2004. Illiquidity Spillovers: Theory and Evidence from
European Telecom Bond Issuance. Working Paper, Stanford University.
Reiss, P.C., Werner, I.M., 1998. Does risk sharing motivate interdealer trading?
Journal of Finance 53, 16571703.
Roll, R., 1984. A simple implicit measure of the effective bid-ask spread in an
efcient market. Journal of Finance 39, 11271139.
Schultz, P., 2001. Corporate bond trading costs: a peek behind the curtain. Journal of
Finance 56, 677698.
Stoll, H.R., 1978. The pricing of security dealer services: an empirical study of
NASDAQ stocks. Journal of Finance 33, 11531172.
Wood, P., 2008. Law and Practice of International Finance. Thomson Retuters Ltd.,
London.