advertising, corporate bond yields, and liquidity · advertising, corporate bond yields, and...
TRANSCRIPT
Advertising, Corporate Bond Yields, and Liquidity
by
Ali Nejadmalayeri1, Manohar Singh2, Ike Mathur3
1 Spears School of Business, Oklahoma State University, Tulsa, OK, 74106, Tel: +1-918-594-8399, Fax:
+1-918-594-8281. Email: [email protected] (Corresponding author). 2 Great Valley School of Graduate and Professional Studies, Pennsylvania State University, Malvern, PA 19355. Email: [email protected]. 3 Department of Finance, Southern Illinois University, Carbondale, IL 62901-4626. Email: [email protected].
1
Advertising, Corporate Bond Yields, and Liquidity
Abstract
Research suggests that product market advertising can enhance a firm’s visibility, thus attracting investors and improving stock liquidity and returns. Does advertising for improving visibility then increase the firm value? To answer this, we examine how advertising affects corporate bonds. Unlike stockholders, bondholders may view advertising skeptically. Without proven effectiveness in improving revenues, large pre-interest, advertising expenditures can be seen as detrimental to meeting debt service obligations. We find that although greater advertising improves bond liquidity, it does not lower credit spreads. However, ineffective advertising widens credit spreads and reduces bond liquidity. When costly, improving visibility then might even destroy value. JEL classification: M37; G30; G32 Keywords: Corporate bonds; Credit spreads; Liquidity; Advertising
2
“E*Trade Group Inc. may have sponsored the Super Bowl’s half-time show, but
apparently not too many convertible bond investors were watching. Despite a weeklong
road show, investors balked at the terms originally proposed by E*Trade and its
underwriter FleetBoston Financial Corp.’s Robertson Stephens for the $500 million
offering. By the time the firm sat down to discuss final pricing yesterday evening, it
appeared as though E*Trade would be paying a higher interest rate. The company also
had to offer more generous conversion terms to convince investors to subscribe for the
full deal….”
The Wall Street Journal, Tuesday, February 2, 2000
The E*trade anecdote questions the conventional wisdom that “familiarity breeds
investments.”1 This intuition about familiarity, of course, stems from Merton’s (1987) seminal
work showing that incomplete information is reflected in a firm’s stock return. Low-visibility
firms with small investor bases then carry large incomplete information premiums.2 Visibility-
enhancing activities such as advertising campaigns can improve returns by attracting new
investors.3 Grullon, Kanatas, and Weston (2004) show that by promoting its own name, brands,
and products, a firm attracts a larger pool of institutional investors and thus enhances its stock
1Huberman (2001) coins the phrase “familiarity breeds investments.” He shows that Regional Bell Operating Companies’ shareholders tend to live in the areas that these companies serve. He sees this as “compelling evidence that people invest in the familiar.” Odean (1999) proposes that investors manage the daunting task of choosing among thousands of possible stock purchases by narrowing their search to stocks that recently caught their attention. Barber and Odean (2008) indeed find evidence in support this proposition. 2A large body of evidence supports this notion. For instance, Grinblatt and Keloharju (2001) and Huberman (2001) find that investors tend to concentrate their portfolio strategies on local firms. Coval and Moskowitz (1999) find that portfolio managers tend to invest in locally headquartered firms. 3Evidence also suggests that large public dissemination of previously known information by a third party can also lead to price appreciation. Huberman and Regev (2001) show that subsequent to being featured in the Sunday edition of the New York Times for a cancer drug discovery, EntreMed experienced unpredicted appreciation in stock price. This was, however, barely “news” since this breakthrough in cancer drug research was reported in Nature more than five months earlier.
3
liquidity. Chemmanur and Yan (2008) find that advertising increases the levels of trading
volume and analyst coverage, and hence improves contemporary stock returns.4
This viewpoint, however, rests on the premise that the positive externality of advertising in
terms of increased visibility among investors outweighs the net cost of carrying out advertising
campaigns. This premise finds support in equity markets where potential investors may view
advertising costs as posing relatively trivial risk compared to higher potential returns due to
increased market liquidity of their equity holdings. In stock markets characterized by high
trading frequencies5 and an emphasis on capital gains,6 returns are intimately linked with market
liquidity. Increased liquidity brought about by advertising campaigns would then be deemed as a
significant positive externality by equity investors. Moreover, as residual claimants, stock
investors may see advertising as a real option with significant upside potential. Possible gains
made in product markets (Farquhar 1989; Keller 2002; Thomas, Shane, and Weigelt 1998) can
indeed make advertising a valuable option for stockholders over and above the implementation
costs.
4A growing body of literature in marketing (e.g., Joshi and Hanssens 2004; Cheng and Chen 1997; Chauvin and Hirschey 1993; Rao, Agarwal, and Dahlhoff 2004; among others) also suggests that advertising positively relates to corporate value. Keller (2002) attributes brand equity related augmented cash flows to customer loyalty, increased marketing efficiency, brand extensions, and higher profit margins. Similarly, while Farquhar (1989) suggests that advertising-related cash flow augmentations may be traceable to price premiums, Boulding, Lee, and Staelin (1994) relate increased cash flows to capturing greater market share. 5Both institutional and individual stockholders engage in high-frequency trading. Barber and Odean (2000) find that while the average U.S. household’s portfolio has a 75% annual turnover, it underperforms the market by 2% per year. Barber, Lee, Liu, and Odean (2009) find that in Taiwan during the period of 1995–1999, individual investors account for more than 85% of trading volume and trading value despite the fact that their average trade size was about one-fourth of the average institutional investors’ trade. Rubin and Smith (2009) show that larger institutional ownership in dividend paying stocks leads to greater volatility. Sias, Starks, and Titman (2006) also find that institutional trading tends to be informationally motivated and has high frequency. Using a sophisticated method of matching 13-F institutional holdings to tick-by-tick trading data, Campbell, Ramadorai, and Schwartz (2009) show that institutional trading resembles those of momentum traders. 6Fama and French (2001) report that while more than 66% of firms were paying dividends in 1978, by 1999 only 20% of firms paid dividends. Grullon and Michaely (2002) further report that based on COMPUSTAT data, the percentage of aggregate share repurchase expenditures to earnings grew from barely 5% in 1980 to 42% in 2000. Using a very long time series of stock prices in the U.S. and U.K. for the period of 1927 to 1992, Goetzman and Jorion (1995) report a dividend yield and a capital gain of, respectively, 4.67% and 7.41% for the U.S. and a dividend yield and a capital gain of, respectively, 5.28% and 8.94% for the U.K.
4
In contrast with equity holders, corporate bond investors do not have any residual claim and
tend to hold instruments for a long time to guarantee yields.7 Bondholders’ returns are thus
highly sensitive to interest payments and the borrower’s ability to make good on contractual
payments becomes the first order of concern.8 Any activity that could deteriorate the financial
health and stability of the borrower’s income stream would be deemed detrimental despite the
potential positive impact on liquidity. Since advertising expenditures occur before debt service,
they limit the firm’s ability to meet debt obligations, thereby increasing default probability and
corporate bond yields. Moreover, if advertising campaigns are seen as ineffective in improving
future cash flows, they are most likely seen as risky gambles with significant potential for wealth
transfer from bondholders to stockholders (Myers 1977).
Given debt investors’ distinct preferences compared to equity investors, corporate
bondholders’ assessment of advertising offers an interesting setting to measure how benefits and
costs of increased visibility are valued by investors. In this paper, we intend to achieve this by
examining the impacts of advertising visibility vis-à-vis advertising efficacy on both corporate
bonds’ credit spreads and liquidity. If bondholders see advertising as unproductive utilization of
resources --unless a firm has a record of effective advertising campaigns-- then advertising
visibility will have minimal impacts on both corporate bonds credit spread and liquidity.
Effective advertising, however, is expected to have a meaningful impact on credit spreads and
liquidity of corporate bonds. Following Grullon et al. (2004), we assume that the size of the
7 Sarig and Warga (1989) show that after the initial offering phase, the corporate bond market becomes quite illiquid mainly because the buyers tend to hold the instruments for a long time. 8 Elton, et al. (2001) state that corporate bonds’ credit spreads reflect three important factors: expected default, personal taxes, and a systematic risk premium. For instance, for an A-rated, 10-year bond, they find that about 18% and 47% of the credit spreads is due to default risk and taxes, respectively. Of the remaining spread, as much as 85% is explained by Fama-French risk factors, leaving little room for all other factors, including liquidity. Interestingly, they also find that the proportional contribution of default risk to the credit spread grows exponentially with credit rating. For instance, for a 10-year, BBB-rated bond, the percentage contribution of the default risk rises to 41%.
5
advertising budget serves as a reasonable proxy for the positive externality of advertising
campaigns in terms of increased investor familiarity (exposure).9 We measure effectiveness of
advertising in terms of its historical average impact on future revenue. Our analysis shows that
while increased visibility narrows credit spreads --albeit marginally-- ineffective advertising
significantly and pronouncedly widens the credit spreads. Interestingly, we find that much like
equities (see Grullon, et al. 2004), larger scale of advertising increases liquidity of corporate
bonds. However, advertising efficacy remains a more significant determinant of corporate bond
liquidity. Companies with a history of ineffective advertising (i.e., advertising campaigns not
succeeded by larger sales) experience significantly lower liquidity in their bonds.
Our analysis thus extends recent studies on the impact of familiarity on asset returns (Coval
and Moskowitz 1999; Grinblatt and Keloharju 2001; Huberman 2001; Grullon, et al. 2004;
Chemmanur and Yan 2008). Building on a recent work by Grullon et al. (2004), we show that
unlike equity, for debt securities (where costs of achieving increased visibility through
advertising may be significant), the impact of advertising scale in improving liquidity is limited
at best. Interestingly, gains made from increased advertising (in terms of lower credit spreads or
greater liquidity) are not enough to offset the losses due to increased spread/lower liquidity if
advertising is ineffective. Firms cannot simply reduce their borrowing cost by ramping up their
advertising. In fact, without clear evidence of past success in utilizing advertising expenditures
to improve sales, any increase in advertising expenditures will result in unfavorable borrowing
conditions. Gains resulting from better stock and bond liquidity should be taken with a grain of
salt because a priori the net effect of increasing advertising on the total firm value is unclear.
9 Grullon, et al. (2004) use advertising expenditures as a proxy for the degree of visibility. Citing Bagwell’s (2001) survey of the economics of advertising, they argue that although “… advertising is presumably aimed at increasing the firm’s market share in the product market, at the minimum it should make the firm’s name and products better known to both consumers and investors…” (p. 440). Indeed, Grullon et al. (2004) find results consistent with the idea that more advertising can increase the advertising firm’s stock market liquidity and institutional ownership.
6
These results support the notion that actions purely targeted to seize investors’ attention can
have negative consequences. “Attention is a scarce resource” (Barber and Odean 2008, p. 785)
and thus among “many alternatives, options that attract attention are more likely to be
considered, hence more likely to be chosen.” As Barber and Odean (2008) point out, attention
leads to optimal choice when salient attributes of the chosen alternative are critical to the
investor’s utility. Attempts to subvert the attention to suboptimal choices, which offer no
relevancy to investor’s utility, may not be welcomed. Our results confirm that bondholders
significantly consider these salient attributes (i.e., borrower’s ability to derive real value from
advertising) in their investment decision. Advertising aimed at grabbing bond investors’
attention thus needs to be substantiated by real business and marketing prowess.
Our findings also shed additional light on the economics of advertising. Economic theorists
contend that advertising provides valuable product market information that can help the firm to
signal quality more credibly and compete more effectively (Stigler 1961; Telser 1964; Milgrom
and Roberts 1986; Bagwell and Ramey 1994; Chemmanur and Yan 2009). With the exception
of Grullon et al. (2004), most evidence is limited to the product markets (Thomas et al. 1998).
While our results further confirm that product market advertising can affect value through a
capital market channel, they also highlight the fact that the inherent signaling facet of advertising
must be credible if advertising is to be a net value-adding proposition. Absent a verifiable
positive product market effect, advertising’s impact on total firm value exclusively through a
capital market channel remains limited at best.
Our paper also extends recent efforts in the study of credit spreads (Elton et al. 2001; Collin-
Dufresne, Goldstein, and Martin 2001) in that we identify advertising, particularly advertising
effectiveness, as an additional factor that affects credit spreads. We also contribute to the
7
growing literature on corporate bond liquidity (Longstaff, Mithal, and Neis 2005; Chen,
Lesmond, and Wei 2007) by identifying advertising as a determinant. Since liquidity risk is
pertinent to credit spreads, it is not surprising that advertising is a common determinant for both
credit spreads and corporate bonds’ liquidity. However, our results indicate that there are also
unique channels through which adverting affects corporate bonds’ credit spread and liquidity
individually.
The rest of this paper is organized as follows. Section 1 describes sample selection and the
data. In Sections 2 and 3, we discuss our empirical methodologies and present the results. In
Section 4, we offer some robustness checks, and finally in Section 5 we present our conclusions.
1. Sample Selection, Variable Description, and Summary Statistics
1.1 Sample selection
To construct our sample of potential corporate bonds, we start with all bonds issued by U.S.
firms that can be identified in the Fixed Income Database (FISD), as provided via WRDS, for the
period of 1994 to 2006. Our main focus is on bond transaction as reported by FISD.10 As is the
convention of previous research, we ensure that payout characteristics of bonds in our sample are
similar; hence we exclude all bonds with option-like features such as callability, putability,
convertibility, and sinking fund provisions. Additionally, we exclude zero-coupon and floating-
rate bonds. We also delete bonds without ratings by either Standard & Poors (S&P) or Moody’s.
Similar to previous bond pricing studies (see, e.g., Collin-Dufresne et al. 2001; Eom, Helwege,
and Huang 2004), we exclude regulated industries (i.e., financial services and utilities).
10Other recent studies by, for example, Elton et al. (2001); Eom, Helwege, and Huang (2004); Gebhardt, Hvidkjaer, and Swaminathan (2005); and Guntay and Hackbarth (2007) also rely on the Fixed Income Database.
8
Next, we merge the data with Treasury term structure information from the Board of
Governors of the Federal Reserve. We then find the average characteristic of each transaction
per firm per month. We merge our data with data from monthly CRSP and OptionMetrics. We
use monthly CRSP to obtain stock prices, stock return volatility, and market volatility. We use
OptionMetrics to obtain the probability of return jump implied by the S&P500 Index options.
We only keep those firms that have valid transaction month-end’s stock price and rolling two-
year return standard deviations. We use the COMPUSTAT annual database to obtain accounting
information such as leverage, interest coverage, quick ratio, profitability, earnings volatility, and
earnings management (accruals). We require our firms to have valid accounting measures in the
year prior to transaction. Some of accounting characteristics, however, are multi-year averages.
In general, for a firm to be considered, accounting information must be available for three years
prior to transactions. To avoid biases due to outliers, all of our accounting characteristics are
winsorized at the 2% level (i.e., observations are trimmed at the 1% level at both tails). After
merging our transaction data with COMPUSTAT and deleting firms that do not have valid
advertising expenses for either the current or previous years, we have a final sample of 27,792
firm-month observations.
1.2 Variable definitions
1.2.1 Credit spread
For our purpose, the credit spread is computed as the difference between the corporate bond
yield and the fitted yield on an otherwise equivalent Treasury bond. Following Duffee (1998),
Collin-Dufresne et al. (2001), and Guntay and Hackbarth (2007), we use a linear interpolation
scheme for the Treasury yield rates reported by the Federal Reserve Board of Governors (H.15
release of the Federal Reserve System) for maturities 1, 2, 3, 5, 7, 10, 20, and 30 years to
9
approximate the entire yield curve. Since yields on only the aforementioned bonds are available
from the Fed, we use interpolation to arrive at Treasury yield corresponding to the corporate
bonds in our sample. We then define the credit spread (YLDSPRD) as the difference between
the reported yield-to-maturity of the corporate bond and the corresponding Treasury yield.11
1.2.2 Bond liquidity
Recent work indicates that liquidity is a priced risk in corporate bonds’ yields (Chen et al.
2007; Covitz and Downing 2007). Following recent studies (Chen et al. 2007; Covitz and
Downing 2007), we use a host of proxies for bond liquidity. Following Covitz and Downing
(2007) and Guntay and Hackbarth (2007), we use contemporaneous and forward measures of
three proxies of bond liquidity: fractional trading months (LIQ), dollar trading volume
(LogDVolm), and trading volume (LogTVolm). The fractional trading months (LIQ) is a bond-
level proxy for liquidity that counts the number of months a bond has a market quote during the
past 12 months. To obtain liquidity, LIQ, we then divide this count by 12 to standardize this
measure to the unit interval. The dollar trading volume (LogDVolm) is the natural log of the
cumulative dollar value of trading over the past 12 months. The trading volume (LogTVolm) is
the natural log of the cumulative number of bonds traded over the past 12 months.
1.2.3 Advertising visibility and efficacy
We assess the influence of advertising on cost of debt by studying two distinct dimensions -
--scale and effectiveness-- of advertising campaigns. First, as in Grullon et al. (2004), we
assume that the scale of advertisement captures the total visibility impact of advertising. The
basic intuition here is that “… while such advertising is presumably aimed at increasing firm’s
market share in the product market, at a minimum it should make the firm’s name and products
11Although other more sophisticated methods can be used to find the fitted Treasury yield curve, Elton et al. (2001) note that these different proxies yield qualitatively similar results. As a result, we use simple interpolated fitted Treasury yields for the analysis pursued in the paper.
10
better known to both consumers and investors” (Grullen et al. 2004, p. 440). Following Grullon
et al. (2004), we proxy the visibility impact of advertising by its scale measured as the natural
log of a firm’s advertising expenditures. We use two variations of visibility measure: the natural
log of the current fiscal year’s advertising expenditures (LOGADV0), and the natural log of the
sum of the current and previous fiscal year’s advertising expenditures (LOGADV1). To control
for the industry effects, we normalize our variables as follows:
}0LOGADV{min}0LOGADV{max
}0LOGADV{min0LOGADVLOGADV0
,SICdigit2,SICdigit2
,SICdigit2,
,tiitii
tiiti
tin∈∈
∈
−
−=
and
}1LOGADV{min}1LOGADV{max
}1LOGADV{min1LOGADV1LOGADV
,SICdigit2,SICdigit2
,SICdigit2,
,tiitii
tiiti
tin∈∈
∈
−
−= .
Verma (1980) shows that from a price theoretic approach --where consumers maximize their
utilities net of information costs and advertising helps to reduce these costs-- the sales-
advertising relationship is highly nonlinear and complex. When firms face uncertain sales, such
nonlinearities are further complicated by the firm’s risk aversion (Nguyen 1987). Linear
approximations of such sales-advertising relationships take the form of time series regressions
whereby sales levels are related to both past sales and past advertising levels. Intuitively then,
we can define advertising effectiveness in terms of the degree to which past levels of advertising
outlays determine current levels of sales. For instance, Depken and Wilson (2004) use time
series regression coefficient estimates to arrive at the imputed value of advertising elasticity of
magazine subscriptions. However, given the limited number of time series observations on firm-
level advertising expenses in the COMPUSTAT database, estimating such time series models
can prove to be difficult. We thus use two variables that closely mimic the notion of elasticity
11
while allowing us to circumvent the data limitations. First, we define advertising ineffectiveness
(AvgA2SL) as the 10-year average of the ratio of lagged advertising to current sales. That is, for
the firm i at time t:
∑= −
−−=9
0 ,
1,, Sales
Adv101AvgA2SL
j jti
jtiti .
This inverse modeling of advertising efficacy (i.e., elasticity to sales) is necessitated by
several advertising expenditure observations being zero. The interpretation of the measure is that
a higher value of AvgA2SL reflects greater ineffectiveness of advertising outlays of the firm in
question. We expect a positive relation between this measure and credit spreads for a firm. One
shortcoming of the above measure is that it does not reflect sensitivity of changes in sales to the
corresponding changes in advertising. Moreover, it does not allow for possible nonlinearities.
To allow for possible nonlinearity and to capture advertising-sales sensitivity, we define an
alternative measure of advertising ineffectiveness (eA2S5) as the five-year correlation between
lagged log advertising expenditure and current log sales, or for firm i at time t:
21
4
0
4
0
24
051
24
015
11
4
0
4
051
4
015
11
)Saleslog()Saleslog()Advlog()Advlog(
)Saleslog()Saleslog()Advlog()Advlog(51
eA2S5
⎟⎟⎟
⎠
⎞
⎜⎜⎜
⎝
⎛
⎥⎥⎦
⎤
⎢⎢⎣
⎡−
⎥⎥⎦
⎤
⎢⎢⎣
⎡−
⎥⎥⎦
⎤
⎢⎢⎣
⎡−
⎥⎥⎦
⎤
⎢⎢⎣
⎡−
=
∑ ∑ ∑∑
∑ ∑∑
= = =−−
=−−−−
= =−−
=−−−−
k k jjtkt
jjtkt
k jjtkt
jjtkt
t .
This measure is similar to the coefficient estimate of a time series regression that relates log
sales to log lagged advertising. To account for industry variations, we normalize our
effectiveness variables as follows:
}AvgA2SL{min}AvgA2SL{max
}AvgA2SL{minAvgA2SLnAvgA2SL
,SICdigit2,SICdigit2
,SICdigit2,
,tiitii
tiiti
ti
∈∈
∈
−
−=
12
and
}eA2S5{min}eA2S5{max
}eA2S5{mineA2S5neA2S5
,SICdigit2,SICdigit2
,SICdigit2,
,tiitii
tiiti
ti
∈∈
∈
−
−= .
1.2.4 Control variables
We include several control variables to ensure that the known determinants of credit spreads
and bond liquidity do not confound the influence of our test variables. The choice of control
variables follows Elton et al. (2001), Collin-Dufresne et al. (2001), Campbell and Taksler (2003),
Chen et al. (2007), and Guntay and Hackbarth (2007). Given that firms with higher default
probability and/or lower expected recovery rates are expected to have higher credit spreads, we
control for several macroeconomic, bond-specific, and firm-specific proxies for common default
and recovery risk factors. Table 1 provides the list of all variables with brief descriptions. The
main control variables are defined as follows.
1. Credit rating. As in Collin-Dufresne et al. (2001) and Chen et al. (2007), we control for
credit rating (CRD) as a possible determinant of credit spreads. We utilize COMPUSTAT’s
numerical equivalent of an average of Moody’s and S&P’s credit rating. As measured, a
higher numeric value reflects greater risk and is expected to relate positively to credit
spreads.
2. Risk-free rate. In structural models of credit risk, a rise in the spot rate effectively reduces
the likelihood of default (Leland 1994; Longstaff and Schwartz 1995). Previous empirical
studies (Duffee 1998; Chen et al. 2007) indicate that credit spreads tend to fall when
Treasury yields rise. To control for variations in the risk free rate, we include the one-year
Treasury bill yield (LEVEL) in our credit spread model.
13
3. Treasury term structure. Litterman and Scheinkman (1991) show that the term structures of
Treasury rates have explanatory power in predicting both the interest rate movements and
macroeconomic growth. Ju and Ou-Yang (2006) show that as the yield curve becomes
steeper, credit spreads widen. We control for the term structure influence on credit spreads
by including the difference in the yield on ten-year and two-year constant maturity Treasury
instruments (SLOPE).
4. Treasury liquidity. As in Chen et al. (2007), we use the spread between the three-month
Euro-dollar rate and the three-month Treasury bill yield (EUROD) to capture the Treasury
bonds’ “crowding out” adverse liquidity effect.
5. Years-to-maturity. Merton (1974) shows that credit spreads and maturity are nonlinearly
related. Helwege and Turner (1999), however, find that on average, the term structure of
credit spreads is upward-sloping. We include the natural log of the maturity of a bond
(LogMAT) to control for the credit spread of term structure.
6. Volatility. Structural models also predict that the volatility of firm value is positively related
to credit spreads (see Leland 1994; Longstaff and Schwartz 1995; and Acharya and
Carpenter 2002). We utilize equity volatility (RETVOL) to control for volatility in firm
value. Specifically, we define RETVOL as the annualized standard deviation of the firm’s
monthly stock returns over the preceding 24 months.
7. Age. Time since the issuance of a bond has been shown to relate positively to credit spreads
(see Warga 1992; Perraudin and Taylor 2004; Yu 2005). Generally speaking, the older a
bond becomes, the less often it will transact leading to a higher spread. To control for the
effects of the age of a bond, we include log of bond age (LogAGE) defined as the log of the
number of years elapsed between the settlement date and the issuance date.
14
8. Leverage. Default risk is directly related to the amount of debt outstanding. Following
Chen et al. (2007), we use the ratio of a firm’s book value of total liabilities to its market
value of equity (TD2Cap).
9. Earning volatility. Earnings volatility poses significant risk for debt holders and hence is
reflected in credit spreads. We control for historical earnings volatility (VOLEARN)
measured as the five-year standard deviation of ratio of earnings before interest, taxes,
depreciation, and amortization (EBITDA) to assets.
10. Profitability. Firms with higher operating income can meet debt service obligations more
easily and hence are less likely to default in the near future. As in Guntay and Hackbarth
(2007), we use the ratio of earnings before tax and depreciation divided by book value of
total assets to control for profitability as a determinant of credit spread.
11. Quick ratio. A firm’s ability to meet debt obligations is directly related to its liquid assets.
We use the ratio of cash and receivables to total assets as a measure of asset liquidity
influencing credit spreads.
12. Interest coverage. The ability to meet periodic debt service is the first test in determining
whether a borrower is at default. Following Chen et al. (2007), we measure the incremental
influence of the pre-tax coverage using four censored variables constructed per the
procedure outlined in Blume, Lim, and MacKinlay (1998).12
In our analysis of bond liquidity, we use the following control variables.
12 These four censored variables reflect whether a firm’s five-year interest coverage is lower than 5%, 10%, 20% or not. All censored variables are initially set to zero. For firms with the five-year average interest coverage ratio of less than 5%, the first censored variable takes on the value of average interest coverage. For firms with the five-year average interest coverage ratio of more than 5% but less than 10%, the second censored variable takes on the value of average interest coverage minus 5%. For firms with the five-year average interest coverage ratio of more than 10% but less than 20%, the third censored variable takes on the value of average interest coverage minus 10%. For all other firms, the fourth censored variable takes on the value of average interest coverage minus 20%.
15
13. Firm size. Chen et al. (2007) show that larger firms’ bonds tend to have better implied
liquidity and smaller bid-ask spreads. We use the log of firm size (SIZE) --defined as the
sum of its market value of equity and book value of debt-- as a control. Since larger firms
tend to have larger equity trading volume, we also use the natural log of the firm’s past 12
months cumulative trading volume (LogEVolm).
14. Yield volatility. Chen et al. (2007) show that corporate bonds with more volatile yields tend
to have less liquidity. To control for the influence of yield volatility on bond liquidity, we
include the rolling 12-month standard deviation of the bond’s yield-to-maturity (YldVol) in
our liquidity model.
1.3 Summary statistics
Table 1 provides summary descriptive statistics for the variables analyzed. The mean credit
spread for our sample bonds is 2.191%. The mean three-month T-bill yield (LEVEL) is 3.545%,
and the mean difference between 30-year T-bond and three-month T-bill yields (SLOPE) is
1.102%. Duffee (1998) shows that on average credit spreads are between 0.67% and 1.42%,
centered at 1.01% for medium-term A-rated bonds. Elton et al. (2001) show that credit spreads
of industrial firms range from 0.392% to 1.349% over the period of 1987 to 1996. They also
report that the corresponding Treasury yields range from 5.265% to 8.382%. Firms in our
sample have larger credit spreads than those documented previously. The difference may be
reflective of the unique time frame of our sample. Our sample extends over the period between
2000 and 2004 when the collapse of corporate icons like Enron led to historically high credit
spreads. At the same time, the Federal Reserve’s efforts to combat economic slowdown caused
Treasury yields to fall to historically low levels. Firms in our sample have a mean size of about
$17 billion USD and generate an average EBITDA return on assets of around 14.5%. The mean
16
long-term debt ratio for our sample firms is about 37%, which is comparable to those reported in
previous studies. Overall, firms in our sample are large and profitable industrial firms with
relatively low leverage. Our sample average bonds have 11.408 years-to-maturity and are 3.605
years old. Comparable to the samples in previous empirical studies (see Collin-Dufresne et al.
2001; Guntay and Hackbarth 2007), our sample bonds on average trade one month per year.
[Insert Table 1 here.]
2. Impact of Advertising Effectiveness and Visibility on Credit Spreads
2.1 Univariate results
To provide preliminary insights into the relationship between credit spreads and advertising,
we plot the credit spread of each bond versus the issuing firm’s advertising visibility and
effectiveness in Figure 1. Panel A depicts the relation between credit spreads and advertising
visibility, nLOGADV0, while panel B shows the distribution of credit spreads across two levels
of the advertising effectiveness, nA2SL. The Figure shows that large and effective advertisers
have lower average credit spreads and tighter credit spread distributions.
[Insert Figure 1 here.]
Table 2 provides a comparison of credit spreads across ratings, maturities, firm sizes, and
leverage levels. Larger scale advertisers, i.e., firms with greater visibility, have consistently
smaller spreads across various maturity, size, and leverage categories. The relationship between
advertising scale and spread, however, is not consistent in sign or significance across ratings.
Interestingly, highly levered, relatively smaller firms and firms with shorter-term bonds seem to
gain more from the increased visibility. For instance, large-scale advertising firms in the highest
leverage categories have an average credit spread that is less than half of their counterparts in the
small-scale advertising group. With respect to advertising effectiveness in general, effective
17
advertisers have statistically significantly lower credit spreads. The gap between the effective
and ineffective advertiser firms’ spreads widens almost monotonically with deteriorating credit
quality. Further, effective advertisers with longer-term debt, larger firm size, and lower leverage,
in particular, have smaller spreads. For instance, while on average effective advertisers’ credit
spreads are lower by almost 10 basis points, large effective firms have a 42 basis point credit
spread advantage over their ineffective counterparts. In sum, our univariate results suggest that
greater visibility and more effective advertising may narrow credit spreads.
[Insert Table 2 here.]
2.2 Multivariate results
To analyze the relationship between credit spread and advertising scale and effectiveness in a
multivariate regression framework, we use a reduced form panel regression model13 described
below:
CSPRDb,i,t =α + β1 VISIBILITYi,t + β2 INEFFICACYi,t + ?b,i,t Xb,i,t + εb,i,t (1)
where the dependent variable (CSPRDb,i,t) is the credit spread on bond b of firm i at time t;
VISIBILITYi,t is one of our measures of advertising visibility for firm i at time t; and
INEFFECTIVENESSi,t is one of our measures of advertising ineffectiveness for firm i at time t.
The explanatory variables in Xb,i,t are controls for macroeconomic conditions, bond-level
characteristics, and firm-level attributes. These variables include credit rating (CRD), Treasury
bill yield (LEVEL), Treasury term spread (SLOPE), Euro dollar/Treasury bill yield spread
(EUROD), the natural log of bond’s age (LogAGE), the natural log of years-to-maturity
(LogMAT), stock return’s volatility (RETVOL), bond’s liquidity (LIQ), ratio of total debt to
13Current empirical models of credit spreads sidestep direct estimation of any structural equation. Duffee (1998), Elton et al. (2001), Collin-Dufresne et al. (2001), Longstaff et al. (2005), Yu (2005), and most recently Chen et al. (2007) are a few among a growing list of empirical works on credit spread that have utilized the reduced-form empirical modeling of credit spreads to avoid the rigidity and complexity of structural models.
18
capital (TD2Cap), long-term debt to assets ratio (LTDB), earnings volatility (EARNVOL), quick
ratio (QUIK), EBITDA to assets ratio (ROA), and four interest coverage dummies per Blume et
al. (1998) (INTD1, INTD2, INTD3, and INTD4).
We report the results in Table 3. Supportive of our main hypothesis and consistent with the
univariate results, the results indicate that credit spreads increase with advertising
ineffectiveness. The regression coefficients corresponding to both ineffectiveness measures
(nAvgA2SL and neA2S5) are positive and statistically significant at better than 10% level of
significance. Surprisingly though, our measures of visibility (nLOGADV0 and nLOGADV1) do
not seem to statistically significantly relate to credit spreads. All our regression models use
robust (White 1980, heteroskedasticity adjusted) standard errors corrected for correlation across
multiple observations for a given firm. All models have reasonably high explanatory power as
indicated by model R-squares that exceed 58% and are comparable to recent research on credit
spreads (see, e.g., Guntay and Hackbarth 2007; Klock, Mansi, and Maxwell 2005).
[Insert Table 3 here.]
The coefficient on normalized average lagged advertising to sales (nAvgA2SL) is
significantly positive at the 10% level. Depending on the specification, this coefficient lies
between 3.332 to 3.751, indicating that for every percentage increase in ineffectiveness, credit
spreads increase by approximately 33 basis points. Similarly, the coefficients on the normalized
five-year correlation between sales and lagged advertising (neA2S5) is also positive and
statistically significant at better than 1%. The magnitude of the coefficient being between 0.463
and 0.579 indicates that for every percentage point increase in ineffectiveness, credit spreads
increase by 46 to 58 basis points. As mentioned earlier, the scale of advertising as a measure of
visibility, does not seem to have a statistically significant impact on credit spreads. These results
19
suggest that our univariate results may be confounded by the effects of other determinants of
credit spreads and perhaps visibility, per se, is not an important determinant of credit spreads.
The apparent insensitivity of credit spreads to visibility can be due to a number of factors.
Sarig and Warga (1989) show that except for the on-the-run corporate bonds, the rest of the bond
market is quite illiquid. This may be rationalized in terms of the fact that a significant proportion
of the corporate bonds is held as a part of institutional portfolios where risk management
guidelines may require long-term investment horizons. Since holders of these corporate bonds
may have little incentive to trade them, marginal improvements in the market liquidity caused by
increased advertising may not be material.
All of the control variables, regardless of model specification, behave in a fashion consistent
with the evidence reported in previous research. For example, as in Duffee (1998), we find both
the Treasury bill yield (LEVEL) and the Treasury term spread (SLOPE) to be negatively related
to credit spreads across all empirical specifications. Firms with better credit quality (CRD) have
smaller credit spreads and greater liquidity. More profitable (as measured by ROA) firms and
firms with better asset liquidity (as measured by QUIK) have smaller credit spreads, while more
levered firms (as measured by LTD and TD2Cap) have wider credit spreads. Firms with greater
equity risk (as measured by RETVOL and EARNVOL) have larger spreads, whereas firms with
more equity trading volume have better bond credit spreads. Lastly, longer maturity (LogMAT)
and older bonds (LogAGE) lead to wider credit spreads.
2.3 Advertising and credit spreads across sub-samples
A concern with our previous analysis may be that the effect of advertising is confounded by
the inherent nonlinearity of the term structure of credit spreads. The extant structural models
suggest that the term structure of credit spreads may be nonlinearly related to a firm’s credit
20
quality, debt maturity, firm size, and leverage levels. Merton (1974) shows that the shape of the
credit spread curve changes with changes in a firm’s leverage and earnings volatility. Empirical
evidence (see Duffee 1998; and Yu 2005) also indicates that credit spreads and credit quality are
linked nonlinearly. To control for these nonlinearities, following Collin-Dufresne et al. (2001)
among others, we estimate out baseline regression model separately for firms sorted on credit
rating, bond maturity, firm size, and leverage terciles.
The results are reported in Tables 4 and 5. For brevity, we present coefficient estimates of
our test variables only. In Table 4 we report coefficient estimates of our two proxies each for
visibility and effectiveness and for the three categories of credit ratings and maturities. The
coefficient estimates of scale of advertising remain insignificant except for the low credit quality
firms. In fact, for these firms, higher advertising scale associates with higher credit spread. The
coefficient estimates of the ineffectiveness measure come out as generally positive. They,
however, are statistically significantly only for the low credit quality firms. In terms of
economic significance, estimates in Panels C and D suggest that a one percentage point increase
in correlation between sales and advertising (increasing effectiveness) can lead up to a 127 basis
point drop in credit spreads for BB or lower rated firms. Further, advertising ineffectiveness
appears to be associated with higher credit spreads in all four models (Panels A through D) for
short-term maturity bonds and in two of the four models in other maturity categories.
Interestingly, the adverse impact of advertising ineffectiveness on credit spreads is most
significant in mid-term debt maturity transactions.
[Insert Tables 4 and 5 here.]
In Table 5 we report coefficient estimates of our two proxies each for the visibility and
ineffectiveness and for the three firm size and firm leverage categories. Results in Table 5
21
indicate that, in general, scale of advertising is not of material significance for credit spreads
except perhaps for the smallest firms where it actually increases the spreads (Panel A) and in the
most highly levered firms where it may help reduce the spreads (Panel D). While in the former
case, a one percentage point increase in advertising effectiveness may lead to a 62 basis point
decrease in credit spread, in the latter case, a similar increase in effectiveness can result in a 125
basis point reduction in credit spreads. Advertising effectiveness does not seem to have
significant influence on credit spreads for large firms or for firms with low leverage.
3. Impacts of Advertising Efficacy and Visibility on Bond Liquidity
3.1 Univariate results
We first analyze the relation between bond liquidity and advertising characteristics within a
univariate framework. Figure 2 plots bond liquidity (LIQ) versus a firm’s advertising scale and
effectiveness. Panel A depicts the relation between liquidity and advertising scale for the below
and above median advertising scale groups. Panel B shows the distribution of liquidity across
two levels of the advertising effectiveness. The figure indicates that large-scale and effective
advertiser firms have bonds that are more liquid.
[Insert Figure 2 here.]
Table 6 provides a comparison of two measures of liquidity --proportion of trading months in
the past 12-month period, LIQ, and the natural log of the total volume traded in the past 12
months, LogTVolm-- across ratings, maturities, firm sizes, and leverage levels. For the whole
sample, larger advertisers as well as more effective advertisers have greater bond liquidity, i.e.,
more visible firms have greater liquidity. Panel A indicates that pattern generally holds across
various credit rating categories except for the lowest credit rating classes. The Panel B results
indicate that irrespective of maturity categories, large-scale and more effective advertisers have
22
more liquid bonds. In Panel C we report that irrespective of firm size, effective advertisers have
greater liquidity in their bonds. Further, except for the firms in the bottom third firm size
categories, the greater scale of advertising relates positively to bond liquidity. Finally, in Panel
D we report that except for the most levered tercile, bond liquidity is greater for more effective
advertiser firms. Similarly, irrespective of leverage levels, a larger advertising scale relates to
greater liquidity of its bonds. In sum, these univariate results suggest that greater visibility
related to large-scale advertising and better advertising effectiveness may contribute to greater
liquidity of a firm’s bonds.
[Insert Table 6 here.]
3.2 Multivariate results
Extending our univariate analysis, we relate advertising characteristics to bond liquidity in a
multivariate regression framework. Following recent empirical studies (Covitz and Downing
2007; Chen et al. 2007), we estimate the following reduced-form credit risk model:
Liquidityb,i,t = δ + λ1 VISIBILITYi,t + λ2 INEFFECTIVENESSi,t + ?b,i,t Zb,i,t + ς b,i,t (2)
where the dependent variable (Liquidityb,i,t) is the liquidity of a corporate bond b of firm i at time
t; VISIBILITYi,t is the normalized natural log of current advertising expenditures (nLOGADV0)
for firm i at time t; and INEFFECTIVENESSi,t is the five-year correlation between lagged
advertising and sales (neA2S5) for firm i at time t. Zb,i,t is a vector of control variables for
corporate bond b of firm i at time t. The explanatory variables in Zb,i,t control for macroeconomic
conditions, bond-level characteristics, and firm-level attributes. These variables include credit
rating (CRD), the natural log of a bond’s age (years after issuance) (LogAGE), the natural log of
years-to-maturity (LogMAT), firm size (SIZE), trading volume (LogEVolm) of a firm’s stock,
and bond yield volatility (YldVOL).
23
The results of our multivariate regression analysis presented in Table 7 are largely consistent
with our main hypothesis and previously reported univariate results. Bond liquidity is increasing
in advertising scale and decreasing in ineffectiveness across all six measures of liquidity. The
regression coefficients on visibility (nLOGADV0) and ineffectiveness (neA2S5) are,
respectively, positive and negative and statistically significant at the better than 10% level. All
our regressions use robust standard errors corrected for correlation across multiple observations
for a given firm. All models have explanatory power comparable to that reported by recent
research on credit spreads (see, e.g., Covitz and Downing 2007; Chen et al. 2007).
[Insert Tables 7 and 8 here.]
The coefficients on the visibility measure, normalized natural log of the current advertising
expenditure (nLOGADV0), are positive and statistically significant at better than 1%. For the
models with proportional trading months (LIQ and FLIQ) as measures of liquidity, the
magnitude of the coefficient is between 0.033 to 0.028, indicating that for every percentage point
increase in the advertising scale, the average bond trading frequency of one month per year will
increase by an additional 2 weeks. For the models with log trading volume (LogDVolm and
FLogDVolm) as measures of liquidity, the magnitude of the coefficient is between 2.128 to
1.796, indicating that for every percentage point increase in advertising scale, there will an 8 fold
increase in average bond trading. For the models with log trading volume (LogTVolm and
FLogTVolm) as measures of liquidity, the magnitude of the coefficient is between 0.244 to
0.240, indicating that for every percentage point increase in advertising scale, the average bond
will trade by an additional 1.2 times the original number of bonds in every year.
The coefficients on the ineffectiveness measure, namely the normalized five-year correlation
between sales and lagged advertising (neA2S5), are negative and statistically significant at better
24
than 1%. For the models with proportional trading months (LIQ and FLIQ) as measures of
liquidity, the magnitude of the coefficient is between -0.045 to -0.039, indicating that for every
percentage point decrease in ineffectiveness, the average bond trading frequency of one month
per year will increase by an additional 3 weeks. For the models with log trading volume
(LogDVolm and FLogDVolm) as measures of liquidity, the magnitude of the coefficient is
between -2.253 to -1.968, indicating that for every percentage point decrease in ineffectiveness,
the average bond will trade by an additional 9 times the original number of bonds in every year.
For the models with log trading volume (LogTVolm and FLogTVolm) as measures of
liquidity, the magnitude of the coefficient is between -0.382 to -0.388, indicating that for every
percentage point decrease in ineffectiveness, the average bond will trade by an additional 1.4
times the original number of bonds in every year. An important point to note here is that over
and above the positive visibility impact of greater scale of advertising, the ineffectiveness of
advertising carries a significant adverse influence on bond liquidity. That is, bondholders, while
possibly swayed by advertising-generated visibility, have material concerns about the
ineffectiveness of advertising campaigns. Specifically, greater ineffectiveness relates to lower
liquidity in bonds and hence may contribute to higher credit spreads.
In terms of control variables, it appears that lower credit quality bonds are more liquid as are
bonds of larger firms and firms with greater equity trading volume. While higher yield volatility
associates positively with bond liquidity, as expected, bonds are more liquid at either end of their
lifecycles.
3.3 Advertising and liquidity across sub-samples
25
To control for possible nonlinearities, we follow the extant literature (e.g., Chen et al. 2007)
and estimate our liquidity regression models separately for firm terciles based on credit rating,
bond maturity, firm size, and leverage.
In Table 8, Panels A through C, we report results using three different measures of liquidity.
In all three panels, while the scale of advertising helps liquidity only for the lower credit rating
firms, the adverse impact of ineffective advertising is pervasive in all credit quality categories.
Further, we report that irrespective of debt maturity structure, while advertising scale helps
improve liquidity, ineffective advertising significantly relates to lower liquidity across all three
maturity structures. In addition, the economic significance of the negative impact of advertising
ineffectiveness on bond liquidity seems to increase with debt maturity.
In Table 9 the results of a similar analysis for subgroups of firms based on firm size and
leverage terciles are reported. The results do not indicate specific patterns in the variation of
coefficient estimates across size or leverage categories. In general, but with varying degrees, the
results are consistent with the full sample results in that while advertising scale proxying for
visibility seems to improve the liquidity of bonds, ineffective advertising carries a significant
negative impact on liquidity.
[Insert Table 9 here.]
4. Robustness Analyses
4.1 Model specifications
To check the robustness of our results, we first estimate our credit spread and liquidity
models accounting for the year, industry, and firm fixed effects. Next, we estimate regressions
with Newey-West standard errors. Third, we adopt the Fama-Macbeth approach by estimating
cross-sectional regressions for each month and estimating average coefficients. Finally, we
26
estimate cross-sectional regressions based on time-series averages of variables averaged at the
bond-level. The robustness test results are reported in Table 10 for credit spreads and Table 11
for liquidity.
The Table 10 results indicate that while the effects of advertising scale are weak at best,
ineffectiveness of advertising strongly and consistently affects credit spreads adversely. Thus,
for bond investors, effectiveness of advertising is a material consideration. In general,
bondholders require a greater return on their investment in firms with a history of ineffective
advertising. The coefficient estimates for the control variables are consistent with the ones
reported for the basic model in Table 3.
[Insert Table 10 here.]
The Table 11 results, consistent with the previously reported results in Table 7, indicate that,
as expected, greater advertising scale associates with greater liquidity of bonds. We interpret
these results as supporting the hypothesis that larger scale of advertising helps improve visibility
among bond investors and hence may result in greater liquidity of bonds. At the same time,
consistently significant negative coefficient of advertising ineffectiveness suggests that bonds of
the firms engaged in ineffective advertising have lower liquidity and hence larger credit spreads.
The coefficient estimates for our control variables are similar to the ones reported earlier in the
paper.
[Insert Table 11 here.]
4.2 Simultaneity of credit spreads and bond liquidity
Bond liquidity measures may contain information about credit quality and thus may affect
credit spreads through the credit risk channel. Chen et al. (2007, p. 135) argue that “… Under
the assumption that much of the liquidity costs are due to adverse selection under asymmetric
27
information … asymmetric information on its credit quality … is the main reason for adverse
selection costs. … [H]igher liquidity costs could [then] mean lower credit quality, which could
lead, in turn, to higher yield spreads…”
To control for the potential endogeneity problems arising from the possibility that advertising
may be simultaneously determining credit spread and liquidity, we employ the following
simultaneous equation model of credit spreads and bond liquidity:
tibtibtib
tititibtib
,,,,,,
,2,1,,,,XΦ
ENESSINEFFECTIVVISIBILITYLiquidityCSPRDε′+′+
β′+β′+η′+α′=
tibtibtib
tititibtib S,,,,,,
,2,1,,,,ZΘ
ENESINEFFECTIVVISIBILITYCSPRDLiquidityς′+′+
λ′+λ′+μ+δ′=
The results are presented in Table 12. The estimates indicate that after controlling for
potential endogeneity problems, while advertising visibility only marginally affects credit
spreads, it significantly contributes to greater liquidity. Further, consistent with the previous
results and our main hypothesis, we find that advertising ineffectiveness associates with larger
credit spreads and lower bond liquidity. The results are robust across various measures of
liquidity. The control variable estimates are broadly consistent with the previously reported
results.
[Insert Table 12 here.]
5. Conclusion
The extant literature suggests that since product market advertising brings about consumer
and investor familiarity, advertising can lead to lower equity risk premiums. Although the
existing studies support the notion that more familiarity breeds more equity investment, the
evidence with respect to bonds is limited at best. We argue that for asset classes such as
corporate bonds, for which enhancing familiarity may be significantly costly, the net impact on
28
the risk premium is unclear. We show that for corporate bonds, while greater advertising scale
marginally improves credit spreads and liquidity, the product market efficacy of the advertising
campaigns may play a more crucial role. Firms that cannot otherwise demonstrate a notable
ability to improve sales through advertising may face higher borrowing costs that may far exceed
any benefits that they may derive from increased familiarity. In essence, investors treat the
embedded signaling value of any corporate action aimed at improving visibility as distinct from
pure visibility enhancement. Evidence suggests that corporations cannot hope to beneficially
attract investors’ attention through advertising without credibly substantiating the positive real
cash flows and value effects of advertising outlays.
In view of our results from bond markets, we venture to predict that in instances where
visibility enhancement carries material costs for the claimants, the value of increased visibility
can be dwarfed by the sheer magnitude of the costs. For instance, in a cross-section of firms,
dividend paying firms with a significant commitment to cash disbursement should gain less from
advertising than otherwise comparable non-dividend paying firms. Similarly, firms with large
bank loans and concentrated, private debt may suffer a net loss to their overall value if increased
visibility does not lead to real and tangible cash flows improvements. Further research indeed
could shed light on the veracity of these predictions.
29
References
Acharya, V., and J. Carpenter. 2002. Corporate Bond Valuation and Hedging with Stochastic Interest Rates and Endogenous Bankruptcy. Review of Financial Studies 15:1355–1383. Bagwell, K. 2001. The Economic Analysis of Advertising, in The Economics of Advertising K. Bagwell (ed.), Edward Elgar Press, Cheltenham, UK.
Bagwell, K., and G. Ramey. 1994. Advertising and Coordination. Review of Economic Studies 61:153–177. Barber, B.M., and T. Odean. 2000. Trading Is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors. Journal of Finance 55:773–806.
Barber, B.M., and T. Odean. 2008. All That Glitters: The Effect of Attention and News on the Buying Behavior of Individual and Institutional Investors. Review of Financial Studies 21:785–818.
Barber, B.M., Y.T. Lee, Y.J. Liu, and T. Odean. 2009. Just How Much Do Individual Investors Lose by Trading? Review of Financial Studies 22:609–632.
Blume, M., F. Lim, and C. MacKinlay. 1998. The declining credit quality of U.S. corporate debt: Myth or reality, Journal of Finance 53:1389–1413.
Boulding, W., E. Lee, and R. Staelin. 1994. Mastering the Mix: Do Advertising, Promotion, and Sales Force Activities Lead to Differentiation? Journal of Marketing Research 31:159–172.
Campbell, J.Y.; T. Ramadorai, and A. Schwartz. Caught on Tape: Institutional Trading, Stock Returns, and Earnings Announcements. Journal of Financial Economics, 92, 66-91.
Campbell, J.Y., and G. Taksler. 2003. Equity volatility and corporate bond yields. Journal of Finance 58:2321–2349. Chauvin, K.W., and M. Hirschey. 1993. Advertising, R&D Expenditures, and the Market Value of the Firm. Financial Management 22:128–140. Chemmanur, T., and A. Yan. 2008. Advertising, Attention, and Stock Returns. Unpublished working paper. Boston College.
Chemmanur, T., and A. Yan. 2009. Product Market Advertising and New Equity Issues. Journal of Financial Economics, 92, 40-65.
Cheng, A.C.S., and C.J.P. Chen. 1997. Firm Valuation of Advertising Expense: An Investigation of Scaler Effects. Managerial Finance 23:41–62. Chen, L., D.A. Lesmond, and J. Wei. 2007. Corporate Yield Spreads and Bond Liquidity. Journal of Finance 57:119–149.
30
Collin-Dufresne P., R. Goldstein, and S. Martin. 2001. The Determinants of Credit Spread Changes. Journal of Finance 56:2177–2207. Coval, J.D., and T.J. Moskowitz. 1999. Home Bias at Home: Local Equity Preference in Domestic Portfolios. Journal of Finance 54:2045–2073. Covitz, D., and C. Downing. 2007. Liquidity or Credit Risk? The Determinants of Very Short-Term Corporate Yield Spreads. Journal of Finance 62:2303-2328. Depken, C.A., and D.P. Wilson. 2004. Is Advertising a Good or a Bad? Evidence from U.S. Magazine Subscriptions. Journal of Business 77(2):S61–S80. Duffee, G. 1998. The Relationship Between Treasury Yields and Corporate Bond Yields Spreads. Journal of Finance 53:2225–2241. Elton E., M. Gruber, D. Agrawal, and C. Mann. 2001. Explaining the Rate Spread on Corporate Bonds. Journal of Finance 56:247–277. Eom, Y., J. Helwege, and J. Huang. 2004. Structural Models of Corporate Bond Pricing: An Empirical Analysis. Review of Financial Studies 17, 499-544. Fama, E.F., and K.R. French. 2001. Disappearing Dividends: Changing Firm Characteristics or Lower Propensity to Pay? Journal of Financial Economics, 60, 3-43. Farquhar, P.H. 1989. Managing Brand Equity. Marketing Research 1:24–33. Gebhardt, W.R.; S. Hvidkjaer, and B. Swaminathan. 2005. The Cross-Section of Expected Corporate Bond Returns: Betas or Characteristics? Journal of Financial Economics, 75:85-114. Goetzmann, W.N., and P. Jorion. 1995. A Longer Look at Dividend Yields. Journal of Business, 68:483-508. Grinblatt, M., and M. Keloharju. 2001. What Makes Investors Trade? Journal of Finance 56:589–616. Grullon, G., G. Kanatas, and J.P. Weston. 2004. Advertising, Breadth of Ownership, and Liquidity. Review of Financial Studies 17:439–461. Grullon, G., and R. Michaely. 2002. Dividends, Share Repurchases, and the Substitution Hypothesis. Journal of Finance 57:1649-1684. Guntay, L., and D. Hackbarth. 2007. Corporate Bond Credit Spreads and Forecast Dispersion. working paper, Indiana University. Huberman, G. 2001. Familiarity Breeds Investment. Review of Financial Studies 14:659–680.
31
Huberman, G., and T. Regev. 2001. Contagious Speculation and a Cure for Cancer: A Nonevent that Made Stock Prices Soar. Journal of Finance 56:387-396. Helwege, J., and C. Turner. 1999. The Slope of the Yield Curve for Speculative-Grade Issuers. Journal of Finance 54:1869–1884. Joshi, A., and D.M. Hanssens. 2004. Advertising Spending and Market Capitalization. Working Paper, University of California at Los Angeles, Anderson School of Management. Ju, Nengjiu; Hui, Ou-Yang. 2006. Capital Structure, Debt Maturity, and Stochastic Interest Rates. Journal of Business 79:2469-2502. Keller, K.L. 2002. Strategic Brand Management: Building, Measuring, and Managing Brand Equity. Paramus, NJ: Prentice Hall. Klock, M.S., S.A. Mansi, and W.F. Maxwell. 2005. Does Corporate Governance Matter to Bondholders? Journal of Financial and Quantitative Analysis 40:693–719. Leland, H.E., 1994, Corporate Debt Value, Bond Covenants, and Optimal Capital Structure. Journal of Finance 49:1213–1252. Litterman, R., and J. Scheinkman. 1991. Common Factors Affecting Bond Returns. Journal of Fixed Income 1:1–18. Longstaff, F., S. Mithal, and E. Neis. 2005. Corporate Yield Spreads: Default Risk or Liquidity? New Evidence from the Credit-Default Swap Market. Journal of Finance 60:2213–2253. Longstaff, F., and E. Schwartz. 1995. A Simple Approach to Valuing Risky Fixed and Floating Rate Debt. Journal of Finance 50:789–819. Merton, R.C., 1974. On the Pricing of Corporate Debt: The Risk Structure of Interest Rates. Journal of Finance 29:449-470. Merton, R.C. 1987. A Simple Model of Capital Market Equilibrium with Incomplete Information. Journal of Finance 42:483-510 Milgrom, R., and J. Roberts. 1986. Price and Advertising Signals of Product Quality. Journal of Political Economy 94:796–821. Myers, S.C. 1977. Determinants of Corporate Borrowing. Journal of Financial Economics 5: 147-175. Nguyen, D. 1987. Advertising, Random Sales Response, and Brand Competition: Some Theoretical and Econometric Implications. Journal of Business 60:259–279. Odean, T. 1999. Do Investors Trade Too Much? American Economic Review 12:79–98.
32
Perraudin, W., and A.P. Taylor. 2004. On the Consistency of Ratings and Bond Market Yields. Journal of Banking and Finance 28:2769-2788. Rao, V.R., M.K. Agarwal, and D. Dahlhoff. 2004. How Is Manifest Branding Strategy Related to the Intangible Value of a Corporation? Journal of Marketing 68:126–142. Rubin, A., and D.R. Smith. 2009. Institutional Ownership, Volatility and Dividends. Journal of Banking and Finance 33:627–639. Sarig, O., and A. Warga. 1989. Bond Price Data and Bond Market Liquidity. Journal of Financial and Quantitative Analysis 24:367–378.
Sias, R.W., L.T. Starks, and S. Titman. 2006. Changes in Institutional Ownership and Stock Returns: Assessment and Methodology. Journal of Business 79:2869–2910. Stigler, G.J. 1961. The Economics of Information. Journal of Political Economy 69:213–225. Telser, L.G. 1964. Advertising and Competition. Journal of Political Economy 72:537–562. Thomas, L., S. Shane, and K. Weigelt. 1998. An Empirical Examination of Advertising as a Signal of Product Quality. Journal of Economic Behavior & Organization 37:415–430. Verma, V.K. 1980 A Price Theoretic Approach to the Specification and Estimation of the Sales-Advertising Function. Journal of Business 53:115–137. Warga. A. 1992. Bond Returns, Liquidity, and Missing Data. Journal of Financial and Quantitative Analysis, 27, 605-617 White, H. 1980. A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity. Econometrica 48:817–838. Yu, F. 2005. Accounting Transparency and the Term Structure of Credit Spreads. Journal of Financial Economics 75:53–84.
33
Table 1 Variable description and sample statistics Variable Description Mean Median Std. Dev. CSPRD Credit spread (%) 2.191 1.398 2.265 CRD Numerical rating 3.843 4.000 1.194 LIQ Fraction of months in past twelve months with bond trading 0.113 0.000 0.196 LogDVolm Log of rolling 12 month bond dollar volume 7.315 0.000 8.159 DVolm The rolling 12 month bond dollar volume ($M) 1.503 1.000 3.495 LogTVolm Log of rolling 12 month bond trading volume 0.829 0.000 1.271 TVolm The rolling 12 month bond trading volume (‘thousands) 2.291 1.000 3.564 LogEVolm Log of rolling 12 month equity volume 5.720 5.943 1.497 EVolm The rolling 12 month equity volume (‘millions) 3.049 3.811 0.047 AGE Number of years past issuance 3.605 2.838 3.197 MAT Years to maturity 11.408 8.000 10.863 LEVEL One-year Treasury bill’s yield (%) 3.545 3.450 1.790 SLOPE Difference between 10-yr. and 2-yr. Treasury bonds’ yields (%) 1.102 0.810 0.932 EUROD Difference between LIBOR and 3-month Treasury bill yield (%) 0.258 0.200 0.193 SIZE Log of equity market value plus debt book value 9.733 9.703 1.492 MValue Equity market value plus debt book value ($B) 16.865 16.366 4.446 LTDB Long-term debt to total assets 0.370 0.353 0.162 EARNVOL 5-year volatility of EBITDA to assets 0.034 0.022 0.047 ROA 5-year average of net income to assets 0.145 0.146 0.064 QUIK Cash plus receivables by current liabilities 1.961 0.739 3.179 INTCOV EBITDA to interest expense 8.945 6.196 10.141 TD2Cap Total debt to market value of equity 3.085 0.895 7.363 RETVOL 2-year volatility of monthly stock returns (%) 10.029 8.950 5.106 MKTVOL 2-year volatility of monthly market returns (%) 4.321 4.949 1.316 JUMP Probability of jump as described in Collin-DuFrense et al. 0.348 0.329 0.124 VIX Average monthly VIX index 20.244 19.532 6.618 YldVOL Firm average of rolling 12 month yield volatility 0.011 0.000 0.087 LOGADV0 Log of the current year’s advertising 8.684 8.955 1.642 ADV0 The current year’s advertising ($M) 5.907 7.746 5.166 nLogADV0 Normalized LOGADV0 0.781 0.831 0.230 LOGADV1 Log of the sum of the current and last year’s advertising 9.224 9.537 1.699 ADV1 The sum of the current and last year’s advertising ($M) 10.138 13.863 5.468 nLogADV1 Normalized LOGADV1 0.802 0.839 0.197 eA2S5 The 5-year correlation of annual change in sales and annual
change in advertising -0.149 -0.226 0.462
neA2S5 Normalized 5-year correlation of annual change in sales and annual change in advertising
0.192 0.000 0.273
AVGA2SL The ten-year average of lagged advertising to current sales 0.023 0.015 0.023 nAvgA2SL Normalized AvgA2SL 0.005 0.000 0.019 This table reports mean, medina and standard deviation of variables in our sample. Our sample consists of 69,769 coupon-paying, plain-vanilla corporate bonds of nonfinancial firms. The corporate bond data are from the Mer-gent’s FISD database. The sample period covers the years 1994 through 2006. The data for the term structure of interest rates are from the Board of Governors of the Federal Reserve. All accounting data are from annual COM-PUSTAT. Stock price and returns are from CRSP. S&P 500 option data are from OptionMetrics. The overall market volatility index is from Chicago Board of Option Exchange (CBOE).
34
Table 2 Credit spreads by categories
Effective
Advertisers Ineffective Advertisers
Small Advertisers
Large Advertisers
NOBS CSPRD NOBS CSPRD Diff. NOBS CSPRD NOBS CSPRD Diff. All Firms 13,955 2.142 13,896 2.241 0.099a 13,937 2.848 13,914 1.533 -1.314a Panel A. Credit Rating: AAA, AA+, AA, AA- 1,540 0.668 1,434 0.824 0.156a 464 0.627 2,510 0.765 0.138a A+, A, A- 4,147 0.957 5,036 1.147 0.190a 3,419 1.067 5,764 1.058 -0.008 BBB+, BBB, BBB- 4,256 1.763 4,417 2.093 0.330a 4,303 2.004 4,370 1.860 -0.145a BB+, BB, BB- 2,026 3.206 1,822 3.857 0.651 2,861 3.474 987 3.630 0.155b B+, B, B- 1,708 5.022 912 6.099 1.077a 2,370 5.400 250 5.374 -0.026 CCC+ and less 278 8.319 275 8.506 0.187 520 8.438 33 8.003 -0.435 Panel B. Maturity: Short-term (7 ≤ yrs.) 6,981 2.237 6,281 2.302 0.065 7,304 3.048 5,958 1.311 -1.737a Medium-term (7 < yrs. ≤ 12) 4,015 2.288 3,064 2.262 -0.026 4,133 2.911 2,946 1.386 -1.525a Long-term (yrs. > 12) 2,959 1.719 4,551 2.142 0.423a 2,500 2.157 5,010 1.884 -0.272a Panel C. Firm Size: Small Firms (Bottom 33%) 4,171 3.588 2,645 3.743 0.155 6,353 3.675 463 3.274 -0.401a Medium Firms (Mid 33%) 4,325 2.009 4,811 2.388 0.380 a 5,163 2.518 3,973 1.806 -0.712a Large Firms (Top 33%) 5,459 1.142 6,440 1.513 0.371a 2,421 1.379 9,478 1.334 -0.045 Panel D. Leverage: Low (Bottom 33%) 4,973 1.483 3,820 1.600 0.117a 4,613 1.802 4,180 1.238 -0.564a Medium (Mid 33%) 4,666 1.860 5,081 2.099 0.239a 4,488 2.486 5,259 1.557 -0.930a High (Top 33%) 4,316 3.205 4,995 2.875 -0.331a 4,836 4.181 4,475 1.782 -2.398a This table reports mean credit spreads across credit ratings, maturities, firm sizes, and leverage ratios. Our sample consists of 27,792 coupon-paying, plain-vanilla corporate bonds of nonfinancial firms. The corporate bond data are from the Mergent’s FISD database. The sample period covers the years 1994 through 2006. The data for the term structure of interest rates are from the Board of Governors of the Federal Reserve. All accounting data are from annual COMPUSTAT. Stock price and returns are from CRSP. S&P 500 option data are from OptionMetrics. The overall market volatility index is from the Chicago Board of Option Exchange (CBOE). a, b, c denote credit spread differences that are significant at the 1%, 5%, and 10%, respectively.
35
Table 3 Impact of advertising visibility and efficacy on credit spreads Model (1) Model (2) Model (3) Model (4) Model (5) Model (6) Model (7) Model (8) Constant -2.248*** -2.229*** -2.228*** -2.146*** -1.441*** -1.435*** -1.370*** -1.281*** (-5.89) (-5.86) (-5.99) (-5.77) (-3.16) (-3.10) (-2.88) (-2.60) nLOGADV0 0.079 0.036 -0.074 -0.125 (0.49) (0.23) (-0.49) (-0.83) nLOGADV1 0.074 -0.025 -0.087 -0.211 (0.40) (-0.13) (-0.52) (-1.18) nAvgA2SL 3.334* 3.332* 3.727* 3.751** (1.81) (1.83) (1.96) (1.97) neA2S5 0.463*** 0.470*** 0.561*** 0.579*** (2.78) (2.77) (3.32) (3.33) CRD 0.296*** 0.296*** 0.298*** 0.297*** 0.264*** 0.265*** 0.265*** 0.263*** (13.21) (13.22) (13.28) (13.30) (12.19) (12.12) (11.95) (11.84) LEVEL -0.167*** -0.171*** -0.178*** -0.183*** -0.188*** -0.187*** -0.201*** -0.203*** (-3.42) (-3.51) (-3.44) (-3.55) (-3.73) (-3.83) (-3.82) (-3.94) SLOPE -0.258*** -0.263*** -0.283*** -0.291*** -0.269*** -0.266*** -0.298*** -0.298*** (-2.73) (-2.81) (-3.02) (-3.12) (-3.03) (-3.09) (-3.38) (-3.48) EUROD -0.185 -0.184 -0.159 -0.155 -0.157 -0.158 -0.124 -0.121 (-1.52) (-1.51) (-1.31) (-1.28) (-1.29) (-1.29) (-1.05) (-1.02) LogAGE 0.170*** 0.170*** 0.160*** 0.160*** 0.160*** 0.160*** 0.147*** 0.147*** (4.47) (4.46) (4.49) (4.47) (5.23) (5.21) (5.33) (5.32) LogMAT 0.212*** 0.212*** 0.202*** 0.203*** 0.176*** 0.176*** 0.164*** 0.164*** (5.08) (5.11) (4.85) (4.88) (4.49) (4.50) (4.08) (4.11) RETVOL 0.165*** 0.165*** 0.166*** 0.166*** 0.150*** 0.150*** 0.150*** 0.150*** (10.74) (10.80) (10.66) (10.73) (11.57) (11.63) (11.42) (11.49) LIQ -0.436*** -0.435*** -0.404*** -0.399*** -0.514*** -0.514*** -0.483*** -0.477*** (-4.05) (-4.06) (-3.95) (-3.95) (-4.56) (-4.57) (-4.52) (-4.51) TD2Cap 0.047** 0.047** 0.046** 0.046** (2.21) (2.21) (2.23) (2.23)
36
LTDB 0.603 0.602 0.670* 0.674* (1.47) (1.47) (1.66) (1.67) EARNVOL 1.150 1.139 1.370* 1.339* (1.48) (1.47) (1.76) (1.73) QUIK -0.041* -0.041* -0.041* -0.041* (-1.76) (-1.75) (-1.78) (-1.76) ROA -2.649** -2.655** -2.960** -2.977** (-2.01) (-2.01) (-2.22) (-2.24) INTD1 -0.015 -0.015 -0.017 -0.018 (-0.40) (-0.41) (-0.47) (-0.48) INTD2 0.015 0.015 0.016 0.015 (0.43) (0.43) (0.45) (0.44) INTD3 0.037 0.037 0.042* 0.042* (1.60) (1.60) (1.78) (1.77) INTD4 -0.000 -0.000 -0.000 -0.000 (-0.24) (-0.24) (-0.18) (-0.17) N. Obs. 27,792 27,792 27,792 27,792 27,792 27,792 27,792 27,792 Adj. RSQ 0.5849 0.5849 0.5871 0.5870 0.6145 0.6145 0.6178 0.6179 This table reports results of the regression model of credit spread using different measures of advertising visibility and ineffectiveness as test variables. LogAGE and LogMAT are natural logarithms of a bond’s age and maturity. INTD1, INTD2, INTD3, and INTD4 are censored interest coverage ratios per Blume et al. (1998). All other variables are defined in Table 1. Robust (heteroskadasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
37
Table 4 Impact of advertising on credit spreads across credit ratings and maturities
AAA – AA
Rated A – BBB
Rated BB – C Rated
Short-term Debt
Mid-term Debt
Long-term Debt
N. Obs. 2,969 17,825 6,998 13,241 7,053 7,498 Panel A. nLOGADV0 0.041 -0.128 0.525* -0.177 0.086 0.115 (0.70) (-1.01) (1.68) (-0.86) (0.51) (0.69) nAvgA2SL -0.132 4.710 8.774* 4.242** 2.483 3.842 (-0.43) (1.45) (1.83) (2.46) (1.07) (1.14) Adj. RSQ 0.3152 0.4000 0.5250 0.6305 0.6495 0.5811 Panel B. nLOGADV1 0.069 -0.155 0.618* -0.178 0.075 0.129 (1.07) (-1.10) (1.68) (-0.74) (0.38) (0.71) nAvgA2SL -0.138 4.793 8.598* 4.262** 2.483 3.821 (-0.46) (1.46) (1.82) (2.46) (1.06) (1.14) Adj. RSQ 0.3154 0.4001 0.5249 0.6305 0.6495 0.5811 Panel C. nLOGADV0 0.065 -0.105 0.239 -0.212 0.032 0.066 (1.15) (-0.72) (0.73) (-1.02) (0.20) (0.36) neA2S5 -0.088 0.281** 1.262*** 0.549** 0.717*** 0.380** (-1.25) (2.04) (3.39) (2.54) (3.27) (2.10) Adj. RSQ 0.3168 0.3999 0.5339 0.6324 0.6561 0.5837 Panel D. nLOGADV1 0.110* -0.151 0.129 -0.281 -0.081 0.036 (1.79) (-0.84) (0.33) (-1.11) (-0.42) (0.17) neA2S5 -0.097 0.292** 1.276*** 0.567** 0.731*** 0.382** (-1.36) (2.03) (3.38) (2.56) (3.26) (2.06) Adj. RSQ 0.3173 0.4001 0.5336 0.6324 0.6561 0.5837 This table reports results of the sub-sample regression models of credit spread using normalized advertising visi-bility and ineffectiveness as test variables. A bond is denoted as short-term, mid-term, or long-term if its maturity is, respectively, less than 7 years, between 7 and 12 years, or more than 12 years. All variables are defined in Table 1. Robust (heteroskadasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
38
Table 5 Impact of advertising on credit spreads across firm sizes and leverages
Small-Cap
Firms Mid-Cap
Firms Large-Cap
Firms Low
Leverage Medium Leverage
High Leverage
N. Obs. 6,794 9,119 11,879 8,777 9,729 9,286 Panel A. nLOGADV0 0.480** -0.113 0.083 -0.097 0.027 -0.175 (1.99) (-0.34) (0.53) (-0.47) (0.12) (-0.70) nAvgA2SL 5.145* 10.634* -0.127 1.418 6.144** -0.308 (1.75) (1.96) (-0.10) (0.86) (2.10) (-0.13) Adj. RSQ 0.5976 0.6577 0.5279 0.4922 0.5455 0.6607 Panel B. nLOGADV1 0.337 0.086 0.176 -0.114 0.074 -0.175 (1.26) (0.27) (0.87) (-0.51) (0.28) (-0.61) nAvgA2SL 5.023* 10.202* -0.216 1.436 6.086** -0.306 (1.70) (1.91) (-0.17) (0.86) (2.08) (-0.13) Adj. RSQ 0.5966 0.6576 0.5282 0.4922 0.5455 0.6606 Panel C. nLOGADV0 0.385 0.042 0.022 -0.116 -0.002 -0.273 (1.53) (0.11) (0.14) (-0.59) (-0.01) (-1.12) neA2S5 0.600** 0.460** 0.476 0.088 0.735*** 1.185*** (2.05) (2.21) (1.62) (0.50) (2.96) (2.65) Adj. RSQ 0.5995 0.6546 0.5349 0.4922 0.5507 0.6689 Panel D. nLOGADV1 0.185 0.174 0.024 -0.151 0.004 -0.475* (0.65) (0.43) (0.10) (-0.75) (0.01) (-1.74) neA2S5 0.622** 0.442** 0.474 0.097 0.735*** 1.247*** (2.08) (2.03) (1.55) (0.56) (2.92) (2.75) Adj. RSQ 0.5987 0.6547 0.5349 0.4923 0.5507 0.6695 This table reports results of the sub-sample regression models of credit spread using normalized advertising visi-bility and ineffectiveness as test variables. A firm is denoted as low-, mid-, or high-leverage if the ratio of its long-term debt to total assets is, respectively, in the bottom, middle, or top terciles of the COMPUSTAT universe. A firm is denoted as small-, mid-, or large-cap if the sum of its market value equity plus book value of debt is, respectively, in the bottom, middle, or top terciles of the COMPUSTAT universe. All variables are defined in Table 1. Robust (heteroskadasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in paren-theses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
39
Table 6 Bond liquidity by categories
Effective
Advertisers Ineffective Advertisers
Small Advertisers
Large Advertisers
LIQ LogTVolm LIQ LogTVolm LIQ LogTVolm LIQ LogTVolm All Firms 0.125 0.914 0.100 0.745 a, a 0.091 0.674 0.135 0.986 a, a Panel A. Credit Rating: AAA, AA+, AA, AA- 0.163 1.127 0.105 0.811 a, a 0.101 0.708 0.141 1.024 b, a A+, A, A- 0.122 0.890 0.094 0.658 a, a 0.082 0.615 0.121 0.850 a, a BBB+, BBB, BBB- 0.140 1.026 0.115 0.885 a, a 0.095 0.696 0.159 1.209 a, a BB+, BB, BB- 0.104 0.782 0.099 0.709 –, a 0.102 0.747 0.103 0.748 –, – B+, B, B- 0.098 0.746 0.075 0.576 a, a 0.089 0.681 0.099 0.743 –, – CCC+ and less 0.045 0.367 0.059 0.517 –, a 0.049 0.403 0.091 1.054 a, a Panel B. Maturity: Short-term (7 ≤yrs.) 0.130 0.925 0.102 0.749 a, a 0.091 0.665 0.148 1.058 a, a Medium-term (7 < yrs. ≤ 12) 0.140 1.077 0.121 0.943 a, a 0.111 0.848 0.161 1.259 a, a Long-term (yrs. > 12) 0.094 0.666 0.084 0.605 a, a 0.056 0.409 0.104 0.739 a, a Panel C. Firm Size: Small Firms (Bottom 33%) 0.085 0.615 0.067 0.522 b, a 0.079 0.577 0.072 0.601 –, – Medium Firms (Mid 33%) 0.110 0.810 0.090 0.638 a, a 0.090 0.660 0.111 0.796 a, a Large Firms (Top 33%) 0.167 1.225 0.122 0.916 a, a 0.124 0.955 0.148 1.084 a, a Panel D. Leverage: Low (Bottom 33%) 0.140 1.012 0.082 0.596 c, b 0.076 0.543 0.158 1.150 a, a Medium (Mid 33%) 0.133 0.984 0.110 0.827 a, a 0.111 0.834 0.130 0.960 a, a High (Top 33%) 0.099 0.726 0.105 0.775 a, a 0.087 0.649 0.119 0.863 a, a This table reports mean of two measures of bond liquidity (LIQ and LogTVolm) across credit ratings, maturities, firm sizes, and leverage categories. Our sample consists of 27,792 coupon-paying, plain-vanilla corporate bonds of nonfinancial firms. LIQ is the number of trading months in past 12 months. LogTVolm is total number of bonds traded in past 12 months. The corporate bond data are from the Mergent’s FISD database. The sample period covers the years 1994 through 2006. The data for the term structure of interest rates are from the Board of Governors of Federal Reserve. All accounting data are from annual COMPUSTAT. Stock price and returns are from CRSP. S&P 500 option data are from OptionMetrics. The overall market volatility index is from the Chicago Board of Option Exchange (CBOE). a, b, c denote credit spread differences that are significant at the 1%, 5%, and 10%, respectively.
40
Table 7 Impact of advertising on bond liquidity LIQ FLIQ LogDVolm F LogDVolm LogTVolm FLogTVolm Constant -0.142*** -0.114*** -2.587 -1.894 -1.108*** -1.115*** (-3.01) (-2.71) (-1.63) (-1.14) (-3.22) (-2.94) nLOGADV0 0.033** 0.028* 2.128*** 1.796*** 0.244** 0.240** (2.49) (1.94) (4.13) (3.29) (2.47) (2.00) neA2S5 -0.045*** -0.039*** -2.253*** -1.968*** -0.382*** -0.388*** (-3.01) (-2.76) (-4.15) (-3.53) (-3.65) (-3.27) CRD 0.003* 0.002 0.108* 0.077 0.025** 0.024* (1.77) (1.25) (1.91) (1.33) (2.13) (1.84) SIZE 0.017*** 0.017*** 0.665*** 0.674*** 0.126*** 0.155*** (3.21) (3.53) (3.62) (3.62) (3.12) (3.46) LogEVolm 0.015*** 0.013*** 0.495*** 0.563*** 0.116*** 0.125*** (3.84) (3.42) (3.19) (3.70) (3.43) (3.33) YldVOL 0.186*** 0.030** 6.067*** 1.860*** 1.007*** 0.322*** (3.37) (2.40) (3.70) (3.43) (3.76) (3.14) LogAGE -0.013*** -0.034*** -0.829*** -1.428*** -0.136*** -0.303*** (-6.39) (-13.59) (-9.18) (-16.63) (-8.55) (-15.28) LogMAT -0.016*** -0.011*** -0.542*** -0.523*** -0.109*** -0.107*** (-4.30) (-3.21) (-4.15) (-3.79) (-4.25) (-3.65) N. Obs. 27,797 27,797 27,797 27,797 27,797 27,797 Adj. RSQ 0.0646 0.1049 0.0669 0.0923 0.0915 0.1333 This table reports results of the sub-sample regression models of corporate bond liquidity using normalized adver-tising visibility and effectiveness. LIQ, LogDVolm, and LogTVolm are proxies of liquidity and, respectively, are defined as the number of trading month in past 12 month, the dollar amount of trading in past 12 month, and the total number of bonds traded in past 12 months. FLIQ, FLogDVolm, and FLogTVolm are one-year lead equiva-lents of the aforementioned liquidity measures. LogAGE and LogMAT are natural logarithms of the bond’s age and maturity. All other variables are defined in Table 1. Robust (heteroskadasticity, autocorrelation, and firm clus-tering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
41
Table 8 Impact of advertising on bond liquidity across credit ratings and maturities
AAA – AA
Rated A – BBB
Rated BB – C Rated
Short-term Debt
Mid-term Debt
Long-term Debt
N. Obs. 2,969 17,831 6,997 13,243 7,051 7,503 Panel A. Dependent Variable is LIQ nLOGADV0 -0.002 0.019 0.040** 0.036** 0.035** 0.036** (-0.09) (1.01) (2.51) (2.07) (2.02) (2.44) neA2S5 -0.051** -0.037* -0.029** -0.033* -0.039* -0.061*** (-2.54) (-1.76) (-1.97) (-1.79) (-1.75) (-4.36) Adj. RSQ 0.1016 0.0702 0.0681 0.0655 0.0715 0.0787 Panel B. Dependent Variable is LogDVolm nLOGADV0 0.439 1.328* 3.029*** 2.622*** 1.919*** 2.027*** (0.41) (1.96) (4.56) (3.93) (2.71) (2.84) neA2S5 -1.922** -2.165*** -1.544** -1.796*** -2.157*** -2.814*** (-2.03) (-2.99) (-2.33) (-2.99) (-2.68) (-4.07) Adj. RSQ 0.1066 0.0776 0.0499 0.0668 0.0660 0.0765 Panel C. Dependent Variable is LogTVolm nLOGADV0 0.048 0.094 0.389*** 0.266** 0.317** 0.197* (0.32) (0.70) (3.24) (2.20) (2.55) (1.71) neA2S5 -0.380** -0.347** -0.238* -0.291** -0.356** -0.482*** (-2.47) (-2.53) (-1.93) (-2.31) (-2.31) (-4.65) Adj. RSQ 0.1477 0.1070 0.0684 0.0918 0.1030 0.1070 This table reports results of the sub-sample regression models of corporate bond liquidity using normalized adver-tising visibility and effectiveness as test variables. LIQ, LogDVolm, and LogTVolm are proxies of liquidity and, respectively, are defined as the number of trading month in past 12 month, the dollar amount of trading in past 12 month, and the total number of bonds traded in past 12 months. A bond is denoted as short-term, mid-term, and long-term, if its maturity is, respectively, less than 7 years, between 7 and 12 years, or more than 12 years. All variables are defined in Table 1. Robust (heteroskedasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
42
Table 9 Impact of advertising on bond liquidity across firm sizes and leverages
Small-Cap
Firms Mid-Cap
Firms Large-Cap
Firms Low
Leverage Medium Leverage
High Leverage
N. Obs. 6,794 9,121 11,882 8,778 9,732 9,287 Panel A. Dependent Variable is LIQ nLOGADV0 0.033*** 0.027 0.032 0.035* 0.012 0.053*** (2.61) (1.05) (1.18) (1.79) (0.56) (3.26) neA2S5 -0.028* -0.014 -0.067*** -0.043** -0.059** -0.025 (-1.78) (-0.61) (-2.65) (-2.34) (-2.07) (-1.01) Adj. RSQ 0.0503 0.0647 0.0510 0.1014 0.0732 0.0393 Panel B. Dependent Variable is LogDVolm nLOGADV0 2.383*** 1.762* 1.852* 2.447*** 1.364* 2.600*** (4.13) (1.68) (1.95) (3.34) (1.68) (3.69) neA2S5 -1.622** -1.547* -2.641*** -2.086*** -2.728*** -1.675* (-2.33) (-1.87) (-3.14) (-3.02) (-2.67) (-1.93) Adj. RSQ 0.0401 0.0606 0.0613 0.1098 0.0694 0.0413 Panel C. Dependent Variable is LogTVolm nLOGADV0 0.310*** 0.193 0.209 0.290** 0.023 0.403*** (3.55) (1.08) (1.06) (2.14) (0.14) (3.44) neA2S5 -0.217* -0.179 -0.550*** -0.407*** -0.443** -0.221 (-1.94) (-1.13) (-3.18) (-3.43) (-2.17) (-1.29) Adj. RSQ 0.0544 0.0877 0.0838 0.1485 0.1032 0.0530 This table reports results of the regression models of corporate bond liquidity using normalized advertising visibility and effectiveness as test variables. LIQ, LogDVolm, and LogTVolm are proxies of liquidity and, respectively, are defined as the number of trading month in past 12 month, the dollar amount of trading in past 12 month, and the total number of bonds traded in past 12 months. A firm is denoted as low-, mid-, or high-leverage if the ratio of its long-term debt to total assets is, respectively, in the bottom, middle, or top terciles of the COMPUSTAT universe. A firm is denoted as small-, mid-, or large-cap if the sum of its market value equity plus book value of debt is, respectively, in the bottom, middle, or top terciles of the COMPUSTAT universe. All variables are defined in Table 1. Robust (heteroskedasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
43
Table 10 Robustness regressions for credit spreads and advertising
Year Fixed
Effects
Year & Industry
Fixed Effects
Firm, Year & Industry
Fixed Effects
Newey-West Standard
Errors
Fama-McBeth
Regression
Cross-Sectional
Regression Constant -0.177 0.258 -0.836 -1.370*** -8.265 -0.315 (-0.41) (0.52) (-1.14) (-12.19) (-1.01) (-0.57) nLOGADV0 -0.164 -0.097 -0.002 -0.125*** 3.799 0.050 (-0.99) (-0.58) (-0.01) (-2.88) (0.93) (0.32) neA2S5 0.453*** 0.411** 0.353 0.561*** 0.434** 0.814*** (2.60) (2.49) (1.62) (13.44) (2.24) (5.06) CRD 0.289*** 0.289*** 0.271*** 0.265*** 0.222*** 0.299*** (12.31) (12.53) (7.00) (52.39) (5.40) (21.18) LEVEL -0.407*** -0.409*** -0.428*** -0.201*** 0.628 -0.301*** (-9.18) (-9.34) (-10.64) (-16.49) (0.59) (-4.48) SLOPE -0.751*** -0.759*** -0.756*** -0.298*** 1.210 -0.346** (-9.09) (-9.30) (-10.64) (-11.87) (1.20) (-2.36) EUROD 0.404*** 0.397*** 0.518*** -0.124* -2.096 -0.860** (5.28) (5.29) (7.54) (-1.90) (-0.95) (-2.02) LogAGE 0.146*** 0.148*** 0.100*** 0.147*** 0.124*** 0.256*** (5.19) (5.84) (7.34) (19.48) (9.43) (9.68) LogMAT 0.164*** 0.152*** 0.164*** 0.164*** 0.249*** 0.042 (4.06) (3.57) (5.49) (13.21) (5.27) (0.94) RETVOL 0.124*** 0.127*** 0.097*** 0.150*** 0.228* 0.131*** (8.42) (8.68) (5.00) (35.61) (1.85) (16.00) LIQ -0.421*** -0.423*** -0.127** -0.483*** 0.185 -2.425*** (-3.73) (-3.95) (-2.31) (-11.84) (0.30) (-6.54) TD2CAP 0.047** 0.045** 0.130*** 0.046*** 0.052*** 0.059*** (2.21) (2.24) (2.80) (16.89) (8.42) (13.23) LTDB 0.514 0.401 1.345 0.670*** -0.189 0.811*** (1.28) (0.91) (1.42) (6.89) (-0.38) (3.40) EARNVOL 1.255 1.152 0.816 1.370*** 1.648*** 1.242** (1.62) (1.55) (0.74) (6.50) (4.15) (2.00) ROA -0.035 -0.050 -0.013 -0.041*** -0.066 -0.058*** (-1.47) (-1.37) (-0.27) (-10.60) (-1.58) (-4.99) QUIK -2.774** -2.857** -4.707*** -2.960*** -1.287* -5.781*** (-2.17) (-2.49) (-3.63) (-12.68) (-1.71) (-9.21) INTD1 -0.024 -0.028 0.015 -0.017** 0.106 -0.037 (-0.68) (-0.82) (0.49) (-2.05) (0.89) (-1.29) INTD2 -0.007 -0.011 -0.047* 0.016** 0.488 0.040 (-0.21) (-0.31) (-1.75) (2.49) (1.02) (1.04) INTD3 0.035 0.030 0.010 0.042*** 0.192 0.122*** (1.25) (1.10) (0.46) (9.55) (0.99) (4.36) INTD4 -0.000 0.000 0.000 -0.000 -0.012 0.001** (-0.47) (0.09) (1.58) (-0.25) (-0.75) (2.42) Year Dummies Yes Yes Yes - - - Industry Dummies - Yes Yes - - - Firm Dummies - - Yes - - -
44
N. Obs. 27,792 27,792 27,792 27,792 27,792 1,829 Adj. RSQ 0.6367 0.6405 0.7653 0.6178 0.7232 0.7113 This table reports results of the robustness regression models of credit spread using normalized advertising visibility and effectiveness as test variables. In these regressions, the impact of year, industry, firm, and bond fixed effects are controlled for using a series of dummy variables. The panel regression results with Newey-West t-statistics are also reported. The cross-sectional regressions results based on the time-series averages of 27,792 bonds for 1,829 firms are also reported. For brevity, the coefficients on year, industry, firm, and bond dummy variables are not reported. LogAGE and LogMAT are natural logarithms of the bond’s age and maturity. INTD1, INTD2, INTD3, and INTD4 are censored interest coverage ratios per Blume et al. (1998). All other variables are defined in Table 1. Robust (heteroskedasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
45
Table 11 Robustness regressions for liquidity and advertising
Industry Fixed
Effects
Industry & Firm
Fixed Effects
Firm, Year & Industry
Fixed Effects
Newey-West Standard
Errors
Fama-McBeth
Regression
Cross-Sectional
Regression Constant -0.190*** 0.088 0.193*** -0.142*** -0.088*** -0.065*** (-3.59) (1.18) (3.33) (-9.89) (-5.31) (-2.67) nLOGADV0 0.034*** 0.023** 0.016 0.033*** 0.018** 0.051*** (2.96) (2.20) (1.60) (6.78) (2.06) (5.16) neA2S5 -0.038*** -0.024** -0.026** -0.045*** -0.032*** -0.053*** (-2.70) (-2.36) (-2.44) (-9.78) (-6.81) (-5.24) CRD 0.004** 0.001 -0.003 0.003*** 0.000 -0.000 (2.15) (0.37) (-1.57) (6.43) (0.33) (-0.22) SIZE 0.021*** -0.016* -0.016** 0.017*** 0.020*** 0.012*** (3.91) (-1.88) (-2.56) (11.45) (12.61) (5.11) LogEVolm 0.014*** 0.031*** 0.005 0.015*** 0.003** 0.005*** (3.86) (5.74) (1.09) (12.41) (2.25) (2.69) YldVOL 0.184*** 0.163*** 0.162*** 0.186*** 0.321*** 0.210*** (3.43) (3.25) (3.25) (4.03) (6.99) (4.96) LogAGE -0.012*** -0.008*** -0.007*** -0.013*** -0.016*** -0.009*** (-6.54) (-3.00) (-2.93) (-14.53) (-7.89) (-5.83) LogMAT -0.015*** -0.007* -0.006 -0.016*** -0.012*** -0.011*** (-4.04) (-1.70) (-1.52) (-9.92) (-7.46) (-3.75) Industry Dummies Yes Yes Yes - - - Firm Dummies - Yes Yes - - - Year Dummies - - Yes - - - N. Obs. 27,797 27,797 27,797 27,797 27,797 1,827 Adj. RSQ 0.0715 0.1505 0.1609 0.0646 0.1438 0.1475 This table reports results of the robustness regression models of corporate bond liquidity using normalized advertising visibility and effectiveness as test variables. The proxy for liquidity is LIQ, which is the fraction of trading months in the past 12 months. In these regressions, the impact of year, industry, firm, and bond fixed effects are controlled for using a series of dummy variables. The panel regression results with Newey-West t-statistics are also reported. The cross-sectional regressions results based on the time-series averages of 1,662 bonds are also reported. For brevity, the coefficients on year, industry, firm, and bond dummy variables are not reported. LogAGE and LogMAT are natural logarithms of bond’s age and maturity. All other variables are defined in Table 1. Robust (heteroskedasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
46
Table 12 Simultaneous equation results for the impact of advertising on credit spreads and liquidity
Model (1) Model (2) Model (3) CSPRD LIQ CSPRD LogDVolm CSPRD LogTVolm
Constant -1.062*** -0.143*** -0.928*** -2.888*** -1.000*** -1.056*** (-7.27) (-10.16) (-5.48) (-4.93) (-6.71) (-11.67) CSPRD 0.003** 0.106* 0.031*** (2.14) (1.80) (3.41) LIQ -2.194*** (-7.74) LogDVolm -0.061*** (-7.40) LogTVolm -0.311*** (-7.81) nLOGADV0 0.030 0.034*** 0.096* 2.190*** 0.035 0.247*** (0.65) (5.74) (1.88) (8.93) (0.75) (6.53) neA2S5 0.477*** -0.046*** 0.434*** -2.273*** 0.459*** -0.395*** (13.75) (-10.57) (11.55) (-12.57) (12.99) (-14.19) CRD 0.256*** 0.002** 0.256*** 0.079** 0.255*** 0.013*** (64.57) (2.34) (62.87) (2.54) (65.01) (2.72) LEVEL -0.220*** -0.215*** -0.230*** (-13.81) (-12.94) (-13.91) SLOPE -0.326*** -0.324*** -0.335*** (-12.05) (-11.73) (-12.21) EURO -0.159*** -0.162*** -0.172*** (-2.60) (-2.63) (-2.80) SIZE 0.020*** 0.812*** 0.140*** (15.12) (14.87) (16.47) LogEVolm 0.011*** 0.316*** 0.097*** (10.88) (7.37) (14.46) YldVOL 0.182*** 5.924*** 1.005*** (14.15) (11.20) (12.13) LogAGE 0.124*** -0.014*** 0.102*** -0.844*** 0.112*** -0.140*** (16.15) (-14.67) (10.78) (-21.65) (13.35) (-23.40) LogMAT 0.141*** -0.017*** 0.143*** -0.573*** 0.144*** -0.115*** (11.94) (-11.45) (11.96) (-9.20) (12.47) (-12.04) RETVOL 0.154*** 0.154*** 0.155*** (69.96) (69.23) (70.49) TD2CAP 0.046*** 0.047*** 0.046*** (35.44) (35.66) (35.53) LTDB 0.734*** 0.740*** 0.732*** (11.29) (11.36) (11.25) EARNVOL 1.440*** 1.413*** 1.490*** (7.64) (7.51) (7.81) ROA -0.040*** -0.039*** -0.041*** (-12.79) (-12.53) (-13.00) QUIK -3.100*** -3.032*** -3.118*** (-17.25) (-17.28) (-17.86)
47
INTD1 -0.019*** -0.018*** -0.019*** (-3.00) (-2.84) (-2.97) INTD2 0.017** 0.015** 0.017** (2.25) (2.07) (2.26) INTD3 0.042*** 0.042*** 0.042*** (8.08) (7.99) (8.00) INTD4 -0.000 -0.000 -0.000 (-0.03) (-0.02) (-0.07) NOBS 27,788 27,788 27,788 27,788 27,788 27,788 Adj. RSQ 0.5986 0.0619 0.5872 0.0647 0.6038 0.0873 This table reports results of a simultaneous system of equations estimation for credit spreads and liquidity using normalized advertising visibility and ineffectiveness as test variables. LIQ, LogDVolm, and LogTVolm are proxies of liquidity and, respectively, are defined as the number of trading month in past 12 month, the dollar amount of trading in past 12 month, and the total number of bonds traded in past 12 months. LogAGE and LogMAT are natural logarithms of bond’s age and maturity. INTD1, INTD2, INTD3, and INTD4 are censored interest coverage ratios per Blume et al. (1998). All other variables are defined in Table 1. Robust (heteroskedasticity, autocorrelation, and firm clustering corrected) t-statistics are reported in parentheses. Coefficients that are statistically different from zero at the 1%, 5%, and 10% levels are marked with ***, **, and *, respectively.
48
Panel A
Panel B
Figure 1. Credit spreads and advertising visibility and effectiveness Panel A plots credit spreads and advertising visibility (above and below median of nLOGADV0). Smaller advisers are denoted by zero where as large advertisers are denoted by one. Panel B plots credit spreads and advertising effectiveness (above and below median of nA2SL). Effective advertisers are denoted by zero where as ineffectiveness advertisers are denoted by one. The 25th – 75th percentiles are limits of the gray box with the median shown by a line in the middle. The next biggest and smallest observations are shown using bars. The outliers are shown with diamonds.
49
Panel A
Panel B
Figure 2. Bond liquidity and advertising visibility and effectiveness Panel A plots bond liquidity and advertising visibility (above and below median of nLOGADV0). Smaller advertisers are denoted by zero where as large advertisers are denoted by one. Panel B plots bond liquidity and advertising effectiveness (above and below median of nA2SL). Effective advertisers are denoted by zero where as ineffectiveness advertisers are denoted by one. The 25th – 75th percentiles are limits of the gray box with the median shown by a line in the middle. The next biggest and smallest observations are shown using bars. The outliers are shown with diamonds.