Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
1
Media Coverage and Debt Financing*
Agnes C.S. Cheng† Liangliang Jiang‡ Wei-Ling Song§
May 2019
Abstract
This study examines how media coverage influences firms’ debt structure and finds that media
coverage reduces firms’ reliance on bank loans but increases amounts of public bonds & notes
on their balance sheets. We document two channels that such effect is concentrated among
firms with less information asymmetry problems suggesting that the media complements other
information sources (information channel), and that such effect is mitigated among firms that
demand less external monitoring suggesting that the media substitutes for banks on firm
monitoring (governance channel). By studying news sentiment, we find that negative earnings
news significantly reduces the likelihood of new debt financing, particularly bank loans and
public bonds, but increases that of private placement. News sentiments predict subsequent
credit rating changes, which affect the terms of debt issuance for both private bank loans and
public bonds. We also show that when the markets are negative about firms’ future prospects,
firms report more negative discretionary accruals, which suggests that media coverage also has
accounting implication in affecting firm debt structure.
JEL classification: D82, L82, G21, G23, G32
Keywords: Media Coverage, Debt Structure and Choice, Corporate Governance, Information
free-riding
* We thank Vikas Agarwal, Tarun Chordia, Inder Khurana, Walid Saffar, and seminar and conference
participants at University of Texas – San Antonio, Shenzhen University, Tongji University, Zhejiang University,
and The Hong Kong Polytechnic University for their helpful comments. This research was substantially
supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China
(Project No. PolyU 13500117). † School of Accounting and Finance, Hong Kong Polytechnic University, Hung Hom, Hong Kong; Phone: 852-
27667771; Email: [email protected]. ‡ School of Accounting and Finance, Hong Kong Polytechnic University, Hung Hom, Hong Kong; Phone: 852-
27667033; Email: [email protected]. § E. J. Ourso College of Business, Louisiana State University, Baton Rouge, LA 70803, USA; Phone: 1-225-
578-6258; Email: [email protected].
2
1. Introduction
Debt financing is an important funding source of U.S. corporations. Data from the
Federal Reserve Board show that total bond issues by U.S. corporations have amounted to over
1.8 trillion dollars in 2017. At the end of 2017, banks also hold more than 2.1 trillion dollars
of commercial and industrial loans on their balance sheets. Public bonds and bank loans
represent two distinct but equally important sources of debt financing. While bank loans come
with close monitoring by their lenders, public bond market is known for its information free-
riding problem, thus less monitoring, due to diffusive bond holdings (Diamond, 1984).
However, the size of bond market suggests that the information free-riding problem has been
mitigated by some mechanisms, such as firms’ own reputation (Diamond, 1991), credit ratings,
and analyst following. In this paper, we investigate whether media coverage can be an
additional mechanism that drives firms’ debt financing choices. The findings can shed light on
the fundamental theories of debt structure and the extent to which the media serves as an
information provider and a governance component in the financial system.1
Media reporting nowadays has a profound effect on business and financial markets. Our
research question is motivated by the recent literature that documents the information role and
corporate governance role of the media in the financial market (e.g. Tetlock, 2007; Tetlock et
al., 2008; Dyck et al., 2008; Kothari et al., 2009; Bushee et al., 2010; Tetlock, 2015). Although
the literature on media coverage in corporate finance is still developing, a key finding is that
media coverage helps firms to raise capital. Nonetheless, current studies on media coverage
have focused extensively on the equity market (e.g. Tetlock, 2007; Tetlock, Saar-Tsechansky,
and Macskassy, 2008; Kothari, Li, and Short, 2009), and evidence on the debt market remains
1 A striking example of the governance effect of media is the cancellation of public bond issuance by Huawei
Technology Co. in April, 2018 after The Wall Street Journal reported that the firm was under investigation by
the Department of Justice despite the fact that investors had placed €2 billion of orders for a €500 million
offering. Although this type of outright cancellation of bond issuance is rare, it suggests the potential for media
outlets to serve as a gatekeeper of the capital markets. See “Huawei Scraps Bond Sale After U.S. Criminal Probe
Comes to Light”, The Wall Street Journal, April 26, 2018.
3
limited.2 Our research is built upon existing theoretical research that studies costs and benefits
between the public and private debt markets when the institutional differences are often
highlighted (e.g. Diamond, 1984, 1991; Fama, 1985; Rajan, 1992; Park, 2000). A widely held
view on the benefits of bank borrowing is that banks have superior access to private
information, more resources to process vast information, which enables them to provide
efficient and costly monitoring services that can mitigate asset substitution and
underinvestment problems. Compared to other debt holders, banks are more efficient and
effective monitors, are more flexible in renegotiation debt contracts, and better able to deal
with bankruptcy costs (Leland and Pyle 1977; Diamond, 1984; Gertner and Scharfstein, 1991;
Houston and James, 1996, Park, 2000; Denis and Mihov, 2003). Compared to banks, there is a
lack of incentive for public bondholders to monitor borrowers due to the diffusion of ownership
and free rider problems. Given the information role and the governance role that media may
play in the financial market, we hypothesize that firms with greater media coverage should
substitute away from more public debt financing and towards less bank loan borrowing.
To examine how media coverage affects firms’ debt financing, we employ RavenPack,
a leading provider of news analytics database, which covers all news articles and press releases
disseminated via Dow Jones Newswires. Specifically, we exclude corporate press released by
firms themselves. Our sample includes all the U.S. firms covered by Dow Jones Newswire and
have borrowed bank loans and/or issued bonds from 2001 to 2014. We find that firms with
more media coverage have significantly lower likelihood of having bank loans on their balance
sheets, but higher proportion of debt as bonds and notes. Our estimate suggests that an increase
of one standard deviation in total news coverage reduces bank loans usage by 9.6% of total
debt. In contrast, the usage of bonds and notes increases by 6.7% of total debt.
2 A few exceptional works are Jiang and Sun (2013), Gao, Wang, and Wu (2016), and Bushman, Williams, and
Wittenberg-Moerman (2017) and we will discuss them in Section 2.1.
4
In our analysis, we scale the total number of relevant news articles by total assets to
mitigate the concern that larger firms tend to have more coverage. To address the concern that
firms may try to attract public attention in order to issue bonds versus borrow bank loans,
besides looking at total news, we also conduct our analysis by restricting to certain categories
of news (e.g., those related to earnings, revenue, and dividends) that are considered more
difficult for firms to manipulate (Solomon, 2012) and we label these news as “earnings news”.
The findings are also robust if we use the instrumental variable (IV) estimation method for
identification, where we construct two variables, i.e. (1) the newspaper shutdown in local area
and (2) an “exogenous” growth opportunity index and use them as instruments for media
coverage. To mitigate the concern that news may become duplicate as news media keeps
disseminating the same information, in our robustness tests, we also use alternative media
coverage measures where we only consider new story within a 24-hour time window across all
news stories in a particular package. Finally, another potential problem with our sample is that
certain types of firms may have access to only one debt market (either bank loan market or
public bond market), we therefore restrict the sample to $100 million to $700 million sized-
firms that are more likely to choose between public bonds and bank loans. We find all of our
results remain intact.
To understand the potential channels through which the media impacts firms debt
financing choices, we also examine the effects of media coverage in different information
environments (information channel) and governance structures (governance channel). We use
firm size and whether a firm is included in a major stock index as proxies for more information
transparency. We hypothesize that if the media plays a role in substituting other information
sources or information dissemination role, then we should observe that media coverage has a
stronger impact among firms with less information availability, which we term as the
information substitution hypothesis. On the other hand, if the media mainly complements other
5
information sources and plays the information creation role, then the impact of media on the
reliance on bank debt will be stronger among firms with more information flows, which we
label as the information complementary hypothesis.
Consistent with the information complementary hypothesis, we find that the effects of
media coverage are concentrated among larger firms and firms that are members of major stock
indexes. Instead of providing more information for smaller firms and substituting for other
information sources, our findings indicate that the media complements to other information
sources. Media outlets pick out the newsworthy pieces and broadcast them widely. Such efforts
are by no means trivial because combining information through different sources can be time-
consuming and requires skills to pinpoint the relevant one. It also requires abundant public
information in the first place.
To examine how media coverage may affect firms’ debt financing through the corporate
governance channel, we employ two commonly used governance structures in the literature—
stock ownership concentration HHI index and whether a firm belongs to a relationship industry
as defined in Cremers, Nair, and Peyer (2008) and Bharath and Hertzel (2018). Existing studies
have shown that the existence of multiple large shareholders enhances external monitoring (e.g.
Maury and Pajuste, 2005). Therefore, firms with multiple large shareholders tend to have lower
external monitoring needs. Relationship industries refer to those have strong incentives to
establish and engage in long-term business relationships with their customers (Cremers, Nair,
and Peyer, 2008). As these firms have close ties with other stakeholders, more outside
monitoring may not be needed from the point of view of the equity holders, because these
stakeholders have an incentive and the expertise to monitor them. In other words, firms
belonging to “relationship” industries have less need for external monitoring. Our evidence
shows that the negative (positive) relationship between media coverage and bank loan (public
debt) is mitigated among firms with good governance, suggesting the governance role of the
6
media. Overall, our evidence is consistent with the view that the media plays a complementary
role as information provider compared to banks and also serves a governance role in
substituting banks’ monitoring function.
The research design in this study needs to address several empirical challenges. First,
it is difficult to separate the information contents, which can also be disseminated by other
information providers, such as credit rating agents and analysts mentioned above, from the
additional effects caused by the media.3 Therefore, we control for information provided by both
credit rating agencies and analysts as both have the obligations to provide timely information
prior to major corporate events. However, it is well known that both information providers are
subject to potential conflicts of interest due to the fees collected from corporate clients.4 For
this reason, there is indeed a role for media outlets to disseminate timely negative information.
By looking at negative news, we can also address the second empirical challenge— the reverse
causality problem. Firms and other compromised information providers have incentives to
promote favorable information. Therefore, positive media coverage can be associated with
outside financing activities but the causality runs from financing activities to media coverage.
In contrast, negative news is more likely to be disseminated by the media; thus, we can attribute
the findings from negative news to the effects of media coverage.5
The third empirical challenge is related to the fact that it is difficult to obtain detailed
debt usage information for each firm. Rauh and Sufi (2010) point out the problem as follows:
“While the debt financial footnotes typically list each individual debt issue, there is
often insufficient information in the footnotes alone to categorize the issue. For
3 Engelberg and Parsons (2011) solve the problem by examining local investors’ trading with access to different
local media coverage of the same firms’ earnings announcements. 4 See Bolton, Freixas, and Shapiro (2012) for the conflicts of interest of credit rating agencies and Michaely and
Kent (1999) for that of analysts. Mao and Song (2018) show that analysts who are subject to conflicts of interest
can significantly delay the release of negative information. Kothari, Li, and Short (2009) provide excellent
discussions on the different incentives for disclosures by the management, analysts, and business press. Ahern
and Sosyura (2014) also show that bidders tend to manage media coverage to increase stock prices prior to
merger announcements. 5 The media also has potential for conflicts of interest due to advertisement incomes (e.g., Gurun and Butler,
2012). However, this tends to occur among local media outlets. Our sample firms are more likely covered by
national press; therefore, this should not be a concern.
7
example, an issue labeled ‘9.5% notes due 2004’ could be medium-term notes, public
debt, term bank debt, or a private placement.”
Therefore, empirical work in this area typically relies on several databases to draw inferences
therefrom because each by itself is incomplete. We start with the debt structure information
provided by Capital IQ, from which we can categorize two major types of debt: (1) bank loans;
and (2) bonds and notes. However, there is no way that we can distinguish the quantities of
bonds and notes that are publicly issued from those that are privately placed using Capital IQ.
Nonetheless, this distinction is crucial for our purpose because privately placed debt is different
from publicly issued bonds. Private placement has less diffusive ownership, more covenants,
and provides higher yields due to its illiquidity.6 For this reason, we supplement our analysis
using the new issues databases provided by Thomson Reuters’s SDC Platinum. SDC
Platinum’s coverage of public bonds is rather comprehensive. However, its coverage of
privately placed debt is limited; for example, most of the interest spreads are missing because
it is privately negotiated.
To further understand how media coverage affects debt financing, we supplement our
analysis using new debt financing activities given different news sentiments. Debt structure,
which provides the composition of debt, is scaled by total debt, therefore, we don’t know
whether the size of debt pie, i.e., debt access, is affected by the media. We find that both the
level of media coverage and positive earnings news significantly increase the likelihood of
outside debt financing. However, negative earnings news significantly reduces debt access,
particularly bank loans and public bonds. Nevertheless, negative earnings news significantly
increases the likelihood of new privately placed debt. This finding is consistent with the
hypothesis on the governance role of the media and also consistent with the claim in Carey,
Post, and Sharpe (1998) that, in the private lending markets, banks tend to serve lower-risk
6 See Carey, Post, and Sharpe (1998) and Denis and Mihov (2003) for more discussions on the differences
between private debt and public bonds.
8
borrowers, while non-bank financial institutions serve higher-risk borrowers. Finally, we also
show that when the markets are negative about firms’ future prospects, firms report more
negative discretionary accruals. This result suggests that media coverage also has accounting
implication in affecting firm debt structure.
Our study contributes to the literature in several ways. First, we add to the growing
literature that documents the media’s effects in the financial markets. While a vast majority of
the studies, which we review in the second section, focus on the equity markets, our study
examines the debt markets. Although there are several papers examining debt instruments, their
emphasis is on the media’s information role whereas our results imply that the media also
serves an indispensable role in reducing the information free-riding problem and filling this
gap for bondholders.
Second, the analysis of debt financing contributes to our understanding of how firms
choose financing instruments. Existing empirical studies have shown that firms’ debt structure
is determined by growth opportunities, credit quality, the level of asymmetric information, and
accounting quality. 7 Firms’ choices of debt can also be affected by external governance
pressure, the divergence between ownership and control, and social capital. We add to the list
by showing that media coverage mitigates the information production incentive problem
among bondholders and significantly impacts firms’ usage of debt instruments.
Third, we contrast the roles of different information providers in the financial system
in the spirit of Kothari, Li, and Short (2009) who discuss and present evidence in the equity
market by comparing information disclosed by management, analysts, and business press. In
our study, we focus on credit rating agencies and media outlets. We find that firms with more
media coverage tend to have significantly better ratings. News sentiments also predict
subsequent rating changes, Tobin’s q, and earnings management. In addition, although both
7 See Section 2.2 for the list of references.
9
rating changes and news sentiments significantly affect pricing and some non-pricing terms of
debt, only news sentiments have impact on the choices of debt instruments. These findings
indicate that media coverage fills the void left by credit rating agencies.
The remainder of the paper proceeds as follows. Section 2 provides the literature review
and hypothesis development. Section 3 describes sample selection and the databases. Summary
statistics of key variables are also reported. Section 4 examines the relation between media
coverage and debt structure. Section 5 discusses the evidence of new debt financing activities.
Section 6 reports the analysis of earnings management. Finally, the concluding remarks are
provided in Section 7.
2. Literature Review and Hypotheses Development
2.1 Literature on the role of media in the financial markets
The important role played by the media in the financial markets has been largely
recognized. There is a growing body of literature that has examined the information role of the
media in the financial market (e.g. Tetlock, Saar-Tsechansky, and Macskassy, 2008; Kothari,
Li, and Short, 2009) and whether the media plays the role of information creation or
information dissemination in affecting investors (e.g. Bushee, Core, Guay, and Hamm, 2010;
Tetlock, 2011). Although the literature on media coverage in corporate finance is still
developing, a key finding is that media coverage helps firms to raise capital (Tetlock, 2015).
There are relatively fewer studies on the monitoring and corporate governance role of
the media. Existing research has shown that media can facilitate public monitoring through
insider trading disciplining or accounting fraud identification (e.g. Miller, 2006; Dai, Parwada,
and Zhang, 2015; Rogers, Skinner, and Zechman, 2016). Among them, studies by Dyck,
Volchkova, and Zingales (2008) and You, Zhang, and Zhang (2018) show that in Russia or
China the press plays the corporate governance role if firms are exposed to international media
10
or non-state-controlled media. Furthermore, current studies on media coverage have focused
extensively on the equity market (e.g. Tetlock, 2007; Liu, Sherman, and Zhang, 2014;
Hendershott, Livdan, and Schurhoff, 2015). Evidence on the debt markets remains limited.
There are several exceptions. Bushman, Williams, and Wittenberg-Moerman (2017) show that
media coverage encourages outside, less informed syndicate participants to originate loans by
reducing information asymmetry among those lenders. Gao, Wang, and Wu (2016) document
that media coverage reduces firms’ cost of issuing bonds because it increases investor
recognition. Jiang and Sun (2013) find that the illiquidity of corporate bond market is reduced
when public news arrives. All three papers focus on the information role of the media. Our
work differs from the above-mentioned research in at least three respects: (1) we study firms’
debt structure, new debt access, and new debt choices; (2) we go beyond the information role
to examine the corporate governance role of the media; and (3) we provide a complete picture
of the effect of media coverage in the debt markets by examining price and non-price terms of
bank loans, public bonds, and privately placed debt.
2.2 Literature on debt structure and choice
The size of debt financing is enormous in the U.S., with $3200 billion of new bank
loans borrowed and $1,500 billion of corporate bonds issued in 2015.8 It is well documented
that U.S. firms vary greatly in their debt structure (Rauh and Sufi, 2010). The different types
of debt instruments can influence firms’ capital structure decisions (Faulkender and Petersen,
2006), hence the return that firms earn for their shareholders.
A firm’s debt structure is critically important because it determines whether a firm can
survive during a financial crisis or recession. A large body of research has investigated the debt
structure of firms and particularly why different firms choose to use different types of debts.
8 Data source: Federal Reserve Bank of St. Louis and Securities Industry and Financial Markets Association.
11
Various financial theories have explained the benefits and costs of public debt versus private
debt (e.g. Diamond, 1984, 1991; Fama, 1985; Rajan, 1992; Park, 2000).
Empirical analyses also show that firms’ financial characteristics, such as growth
opportunities, credit quality, the level of asymmetric information, and accounting quality,
determine their debt structure (Houston and James, 1996; Denis and Mihov, 2003; Bharath,
Sunder, and Sunder, 2008; Li, Lin, and Zhan, 2019).9 Firms’ choice of debt type may also be
affected by external governance pressure (Bharath and Hertzel, 2018), the divergence between
ownership and control (Lin, Ma, Malatesta, and Xuan, 2013), state ownership (Boubakri and
Safffar, 2018) and social capital (Hasan, Hoi, Wu, and Zhang, 2017). However, research on the
impact of media coverage on debt structure has been relatively scarce. To our knowledge, our
study is the first to examine the governance role of the media in the debt markets.
2.3 Hypothesizing the information role and the governance role of media
The costly information-production models predict that bank debt is preferred over
public debt because the cost of producing information required for public debt financing is
higher than that for bank financing (Fama, 1985). Compared with public bondholders, banks
have at least two information advantages. First, banks have superior access to private
information that may not be known to the public (Fama, 1985). Second, compared to arm’s
length investors, banks use more resources to process vast amounts of information. A large
body of research has shown that by rebroadcasting the information, media draws attention to
certain stocks, lowers the searching costs, and improves market efficiency (e.g. Fang and Press,
2009; Fang, Peress, and Zheng, 2014). By examining the earnings announcement, Bushee,
Core, Guay, and Hamm (2010) argue that the business press increases information flow in the
market by creating new information by analyzing the implications of earning releases and
interpreting the management forecasts. Therefore, there are at least two mechanisms through
9 See Kale and Meneghetti (2011) for an extensive review of this topic.
12
which the media could influence the firms’ information environment: information
dissemination and information creation.10 When the press facilitates information dissemination
and/or creates new information, it reduces the level of information asymmetry between the
borrowing firm and lenders. To the extent that media alleviates the information asymmetry
problem, we predicts a negative (positive) relationship between media coverage and bank loan
borrowing (public debt financing).
The moral hazard model of the public/private debt choice is built upon the notion that:
(1) shareholders of levered firms need to be monitored because they have incentives to engage
in activities that damage the debtholders, i.e., the shareholder-creditor conflict; and (2) banks
are more efficient and effective monitors than public lenders. There is a lack of incentive for
monitoring on the part of bondholders due to diffused ownership and free-rider problems
(Diamond, 1984, 1991). Even if the bondholders are willing to monitor, as Houston and James
(1996) point out, it could still be inefficient for them to do so due to the duplication of efforts.
In contrast, a widely held view on one of the potential benefits of bank borrowing is that it
provides efficient and costly monitoring services that can mitigate agency problems, whereas
these services are not available from the widely dispersed investors in the public market.
Existing research has shown that media can alleviate agency problems through disciplining
insider trading or identifying accounting fraud (Miller, 2006; Dai, Parwada, and Zhang, 2015;
Rogers, Skinner, and Zechman, 2016). In particular, Dyck, Volchkova, and Zingales (2008)
show that the press plays a corporate governance role among Russian firms when these firms
are exposed to international media. Therefore, the governance role of media suggests a
negative (positive) relationship between media coverage and bank loan borrowing (public debt
financing).
10 Note that in our main analysis, we will not differentiate between the information dissemination and creation
effect of media on debt choice. Our argument is that either function of the media will alleviate the information
asymmetry between the borrower and the lender. Nonetheless, we will attempt to conduct additional tests to
examine whether original news as compared to stale news will make any difference to our results.
13
Taken together, we propose the following hypothesis: Holding everything else constant,
firms with greater media coverage rely less on bank loan borrowings and more on public bond
financing compared to other firms. It is worth to note that in our analysis, we do not try to
disentangle or separate the governance role from the information role of the media. Rather, we
are trying to document that media actually plays both the information role and the governance
role in affecting firms’ choice of debt financing.
3. Data, Sample Description, and Key Variables
The unit of our analysis is at the firm-year level. We obtain media coverage data from
RavenPack News Analytics (RP), a leading provider of news analytics data that cover all the
news articles and press releases disseminated via Dow Jones Newswires. It has over 22,000
online sources of financial news and opinion, and covers over 41,000 companies worldwide.
We use the RP database to determine the total number of news items for each firm over a one-
year period. As RP allows us to identify which publications lead the news cycle and how the
story developed in the media, throughout all of our analyses, we only use news items that are
not originated by firms themselves. Specifically, we use two media coverage measures as our
key measures. The first one is Total_News measured as the natural logarithm of the total
number of news items with a relevance score of 100 reported from RP scaled by the firm’s total
assets. Due to the concern that firms may try to attract public attention in order to issue bonds
rather than borrowing bank loans, we construct our second measure by restricting news items
to certain categories, that is, only those related to earnings, revenue, and dividends. These news
items are considered more difficult for firms to manipulate (Solomon, 2012). We use the
variable Earnings_News to denote this type of news and measure it as the natural logarithm of
the ratio of news items specializing in earnings evaluations with either positive or negative
sentiment * 100 to total news items for firm i in year t.
14
We construct firm-level debt structure variables using Capital IQ. Capital IQ aggregates
debt structure into seven categories: (1) total commercial paper, (2) total revolving credit, (3)
total term loans, (4) total senior bonds and notes, (5) total subordinated bonds and notes, (6)
capital leases (including current portion), and (7) other borrowings. For our purpose, we focus
on items (2)-(5) because these are related to the three types of newly issued debt instruments—
bank loans, public bonds, and privately placed debt—that will be analyzed in greater detail in
Section 5. We aggregate items (2) and (3) and calculate the first main dependent variable used
in this subsection, i.e., Bank_Loans/Total Debt, measured as the ratio of total bank loans to
total debt for each firm i in year t, where total debt refers the summation of items (1) to (7).
The second key dependent variable is the sum of items (4) and (5). In particular, we measure
it as Bond & Notes/Total Debt which is the ratio of senior and subordinated bonds and notes
over total debt for each firm i in year t. We also construct dummy versions of these two
variables to indicate whether each firm i has bank loans or bonds and notes on its balance sheet
in year t.
We also use the Capital IQ sample to match with the Compustat database to ensure that
our sample includes all firms that have private and public debt information and accounting data
available. Using Compustat, we construct a vector of time-varying firm characteristics,
including firm size (Size), firm profitability (ROA), Leverage, Market-to-Book ratio, Sales
Growth, and Cash Flow. We then combine these data with the RP database. The media
coverage measures and all the control variables are lagged by one year so that post-event news
coverage is not counted in our sample.
As RP data became available in 2001, we use a sample period ranging from 2001 to
2014. Our primary sample contains 36,002 firm-year observations involving 5,226 firms in the
U.S. The average (median) firm size is about $5,606 ($246) million U.S. dollars, and the mean
(median) of total news per firm per year is 180 (81) pieces. The average news items specializing
15
in earnings evaluations with either positive or negative sentiment to total news is around 13%
for each firm in each year. The average of Bank Loans/Total Debt ratio is about 0.43 and the
average of Bonds & Notes/Total Debt ratio is 0.41. In our primary sample, about 44% of firms
hold both bank loans and bonds and notes. The summary statistics of other variables can be
found in Table 1. The detailed definitions of all the key variables are presented in Appendix
Table 1.
To study the new debt issuance, we use the Securities Data Company (SDC) database
to obtain information on public bond issuance and private placement and Dealscan to obtain
information on bank loans. The new issues database of SDC Platinum provides detailed
financial information on newly issued securities. It covers more than 800,000 bond deals since
1970. Relying on the SDC database, we are able to construct two indicator variables,
New_Public_Bonds and New_Private_Placement, to indicate whether a firm has new
borrowings of public bonds or private bonds, where New_Public_Bonds
(New_Private_Placement) equals one if a firm has a new bond issuing (placed private debt) in
year t, and zero otherwise. We then construct the pricing and non-pricing terms of the newly
issued public bonds. In particular, we measure the price of public bonds, i.e.
Public_Bonds_Spread, as the natural logarithm of bond gross spread, where gross spread is the
interest spread in basis points and net of yield of treasury securities of comparable maturity.
For the non-pricing terms, we measure total issuance amount and the bond maturity, where
Public_Bonds_Amount is the natural logarithm of total amount of public bonds (in million $)
issued by firm i in year t, and Public_Bonds_Maturity is the natural logarithm of the bond
maturity in months. The pricing terms of private bond issuance are generally missing in SDC;
therefore, we are only able to capture the non-pricing terms of private bond issuance, i.e.
Private_Placed_Amount and Private_Placed_Maturity, which are measured similarly to public
bonds.
16
Using Dealscan, we examine the bank loan information at the facility level. For new
borrowing of loans, we construct an indicator variable New_Bank_Loans that equals one if a
firm has a new loan facility borrowing in year t, and zero otherwise. We measure loan pricing
(Bank_Loans_Spread) using the all-in-drawn spread, that is, the interest spread over LIBOR
plus associated loan origination fees. For bank loans’ non-pricing terms, we use the loan
amount in million US dollars as measures of loan size (Bank_Loans_Amount) and the time to
maturity in months as measures of loan maturity (Bank_Loans_Maturity). We take the natural
logarithm of each of these three variables in performing regression analysis.
To contrast the effects from other information providers in the financial system, we
control for earnings surprises and credit rating changes. We download sample firms’ monthly
S&P domestic long-term issuer credit ratings (SPLTICRM) from Compustat. The credit ratings
are transformed to numerical numbers by assigning 0 to no rating, 1 to rating D or below, and
so forth. The highest credit rating AAA has a value of 22. The change in credit rating
(Ratings_Change) is calculated using the rating at the end of year t minus that of year t-1.
Earnings Surprise is estimated using the Institutional Brokers’ Estimate System (I/B/E/S)
detail history database from Thomson Reuters. We first obtain the one-year ahead earnings
forecasts (EPS, Fiscal Year 1) by all analysts. We keep the most recent forecasts by each analyst
within one year prior to the earnings announcements. The consensus is estimated using the
mean value of these forecasts. An earnings surprise is calculated as actual earnings minus the
mean forecast scaled by the absolute value of mean forecast. Earnings forecast dispersion is
the standard deviation of these forecasts scaled by the absolute value of mean forecast.
4. Media Coverage, Debt Structure, and Credit Ratings
4.1 Media coverage and debt structure
17
We begin our analysis by examining the relation between media coverage and existing
debt structure on a firm’s balance sheet. Unlike new debt issuance, studying existing debt
structure provides a comprehensive picture of how firms use different debt instruments without
conditioning on the availability of new financing activities. Although the financing decisions
were made in the past, the fact that firms have not replaced them with other types of instruments
is an indication of continuing usage of a particular type of debt structure. Otherwise, firms
could refinance and modify the mix of debt structure. Therefore, this type of analysis provides
useful implications by showing the composition of debt structure that firms choose to adopt.
To examine the impact of media coverage on debt financing, we primarily use a panel
regression in which the unit of analysis is a firm-year observation and where we control for
both industry (𝜃𝑑) and year (𝜃𝑡) fixed effects. The year fixed effects control for all time-varying
influences that are common to all the firms each year, while the industry fixed effects condition
out all time-invariant industry characteristics. In particular, we estimate the following ordinary
least squares equation:
𝐷𝑒𝑏𝑡 𝐶ℎ𝑜𝑖𝑐𝑒𝑖𝑡 = 𝛽 ∙ 𝑀𝑒𝑑𝑖𝑎𝑖𝑡 + 𝛾′ ∙ 𝑋𝑖𝑡 + 𝜃𝑑 + 𝜃𝑡 + 𝜀𝑖𝑡. (1)
where 𝐷𝑒𝑏𝑡 𝐶ℎ𝑜𝑖𝑐𝑒𝑖𝑡 is one of the four measures of debt for firm i in year t (i.e.,
Dummy_Bank_Loans, Dummy_Bonds & Notes, Bank Loans/Total Debt, Bonds & Notes/Total
Debt). 𝑀𝑒𝑑𝑖𝑎𝑖𝑡 is one of the two media coverage measures, i.e. Total_News and
Earnings_News. Both variables measure the quantity of news available. The first one includes
all news items except for firms’ press releases scaled by a firm’s total assets. The second one
includes news items specializing in evaluations on firm earnings, revenue and dividends, with
either positive or negative sentiment over total news items. 𝑋𝑖𝑡 represents a vector of time-
varying firm traits: Size is the natural logarithm of the BHC’s total assets, ROA is the ratio of
net operating income over total assets, Leverage is the ratio of total debt over book value of
assets, Market_to_Book is the market value of equity plus book value of debt divided by book
18
value of total assets, Sales Growth is the percentage of annual growth in total sales, and Cash
Flow is the cash flows from operations deflated by average total assets. In seeking to assess the
impact of an intensification of media coverage on debt choice, we focus on estimating β. We
report heteroskedasticity-consistent standard errors that are clustered at the firm level.
Table 2 reports the estimation results on media’s impact on firm debt structure.
Columns (1) and (3) (columns (2) and (4)) show that total news (earnings news) negatively
affects bank loan borrowing, and by contrast, columns (5) and (7) (columns (6) and (8)) show
that total news (earnings news) positively affects public bond issuance. For example, in
columns (1) and (3), we find significantly negative coefficients on total news, which means
that firms with more news coverage have a lower tendency to use bank loans and a lower
proportion of bank loans on their balance sheets. To put this into economic magnitude, an
increase of one standard deviation in total news coverage, the estimate in column (3) indicates
a reduction in bank loan usage by 9.6% of total debt. This number is not only statistically
significant but also economically sizable. In contrast, the estimates in columns (5) and (7),
which analyze bonds and notes, show the opposite. For an increase of one standard deviation
in total news coverage, the usage of bonds and notes increases by 6.7% of total debt. Firms
with more media coverage do show better access to bonds and notes and rely less on bank
loans. In addition, when compare the control variables in Table 2, we find that firms of smaller
size tend to borrow bank loans and firms of larger size tend to issue corporate bonds, which is
consistent with the pecking order theory on debt issuance.
4.2 Instrumental variable estimation
Our baseline empirical results so far show a strong negative relationship between media
coverage and bank lending and a positive relationship between media coverage and bond and
note financing. However, reverse causality is a potential threat to reliable inferences in our
analysis because it may be possible that before making the debt issuance decision, firms choose
19
to attract media attention and disseminate more news. Also, despite including a series of firm
controls and industry and year fixed effects in the regressions, there are still concerns about the
omitted variable bias. If some unknown factors such as managerial characteristics are
correlated with both news coverage and debt financing choices, then our earlier results on the
impact of media coverage will be biased. Although it is extremely difficult to completely
eliminate endogeneity bias, in this section, we attempt to address the problem using the
instrumental variable approach.
The underlying assumption of the instrumental variables should be that they are highly
correlated with media coverage but do not directly affect firm decision on debt choice, except
through the media coverage channel. We use two instruments for media coverage. The first
one is newspaper shutdown, which is defined as the shutdown of metropolitan or local
newspapers in each firm’s headquarter state of each year. 11 We use this as an instrument for
firm’s media coverage as the shutdown of traditional media spread tool may induce exogenous
variation in news coverage for individual firms located in that area, meanwhile, local
newspapers closure is not expected to directly influence a firm’s choice of public versus private
debt. The second instrument is the state-level exogenous growth opportunity. Firms located in
a state with greater growth and investment opportunities are expected to be more appealing to
reporters, thus these opportunities should be highly correlated with individual firms’ media
coverage. However, state-level growth may be correlated with local variables such as firms’
debt choices; therefore, inspired by the work of Bekaert et al. (2005), we construct an
“exogenous” measure of a state growth opportunity by treating each state being composed of
a set of industries with each industry having time-varying growth opportunities, while
assuming that these growth prospects are reflected in the price to earnings (PE) ratios of a
11 These data are available at: https://en.wikipedia.org/wiki/List_of_defunct_newspapers_of_the_United_States
20
country’s (i.e. U.S.) industry portfolio.12 More specifically, we measure this variable using
𝐸𝑥𝑜𝑔𝑒𝑛𝑜𝑢𝑠 𝐺𝑟𝑜𝑤𝑡ℎ 𝑂𝑝𝑝𝑜𝑟𝑡𝑢𝑛𝑖𝑡𝑦𝑆 (∑ 𝑤𝑆𝐼 ∗ 𝑃𝐸𝐼𝐼 ) , where 𝑃𝐸𝐼 represents the industry PE
ratio in the U.S., and 𝑤𝑆𝐼 represents the state-level industry weight based on the industry’s
market capitalization. This measure has been proved to be valid and widely used in the finance
literature. We expect a positive association between the exogenous growth opportunity and
media coverage.
We apply a two-stage least squares (2SLS) estimation and present the results in Table
3. The first four columns in Table 3 report the second stage regressions where the dependent
variable in columns (1) and (2) is Bank Loans/Total Debt, and that in columns (3) and (4) is
Bonds and Notes/Total Debt. The key explanatory variables of media coverage are either
Total_News (Columns (1) and (3)) or Earnings_News (columns (2) and (4)). As shown by the
results, the coefficients on media coverage are still negative for Bank Loans/Total Debt and
positive for Bonds and Notes/Total Debt. These results not only continue to be statistically
significant, but also become slightly larger in terms of economic magnitude.
With regard to the validity of the instruments, we first present the first-stage results
from the IV regressions in columns (5)-(6) of Table 3. The positive coefficients on the measure
of the newspaper shutdown is interesting as it suggests that the closure of traditional media
causes firms overall media coverage to increase. This increase could be due to the entrance of
new media tools that fade out the traditional newspaper release. Meanwhile, the positive
coefficients on the measure of the state-level exogenous growth opportunity is consistent with
our expectation that firms’ media coverage is positively associated with the exogenous
economic growth opportunity at each state. Both instruments are significant at the 1% level in
12 The intuition behind is that a state’s PE ratio for a particular industry should be correlated to its country
counterpart given that all growth opportunities are competitively priced and absorbed by the U.S. capital
markets. Moreover, the weighting of industries within a particular state affects local GDP growth relative to the
country’s GDP growth (Bekaert et al., 2005; Bekaert et al., 2007).
21
the first-stage regressions, suggesting that we have powerful instruments that can easily pass
the relevance tests. The weak instrument tests further rejects the null hypothesis that our
instrument is irrelevant to the endogenous variable. Taken together, the IV results corroborate
our baseline findings that media coverage is negatively related to bank lending but positively
related to bond and note financing.
4.3 Sensitivity analysis
We conduct a series of sensitivity tests to verify the robustness of our results. Due to
space limit, we report these results in the Appendix tables. The first set of test is using novel
news as the measure of news coverage. Novel news is defined as a new story within a 24-hour
time window across all news stories in a particular package (Dow Jones, Web or PR Editions)
and with relevance score of 100 in RP. Due to the dissemination feature of news media, news
reports could be more duplicated over longer period of time. We use the fresher news to address
the concern that the media coverage effect is mainly due to the dissemination of the stale news
rather than creation of the new information. The results report in Appendix Table 2 are similar
to those using total news and earnings news.
Columns (1)-(4) of Appendix Table 3 report the second set of robustness tests in which
firms are restricted to those with total book value of equity between $100 million and $700
million U.S. dollars, because very large firms have good access to the bond markets and
routinely issue bonds through self-registrations or medium-term notes programs. Very small
firms may not have access to the bond market; therefore, there is no choice per se. By using
mid-sized firms, we confirm that our findings are not driven by including small and large firms.
In columns (5) and (6) of Appendix Table 3, we only use term loans to replace bank loans, i.e.
we do not include revolving credits, and the results continue to hold. Finally, we also remove
sample firms during the 2008 Financial Crisis and the following Great Recession (years 2008-
2010), and the findings remain robust. The results are not reported for brevity.
22
4.4 Channel tests
In this subsection, we further investigate the potential channels through which the
media plays a role in the debt markets. There are different layers of information asymmetry
problems. The most serious type is that there is no publicly available information, such as a
small start-up private company, while the other extreme is that there is too much information,
such as a well-known high-tech multinational firm. Companies with access to the bond markets
are likely to tilt towards the end of too much information. However, there is still a wide
spectrum of information availability among these companies. If the function of media coverage
is to resolve the information asymmetry problem in the “small company” sense, i.e. adding
information to boost information availability, we should observe that the effect of media
coverage is stronger among firms with more serious information asymmetry problems. We test
this as the information substitution hypothesis. However, if the information function of the
media is to separate the wheat from the chaff, then we will find stronger media effects among
firms with less information asymmetry problems, i.e., the information complementary
hypothesis. We use two proxies for the level of information flow: large firm size and whether
the firm is a member of a major stock index, where Large Size is defined as a dummy variable
that equals one if a firm’s total assets are above the sample median of that year, and zero
otherwise, and Stock Index Member is defined as a dummy variable that equals one if a firm is
a major stock index member, including Dow Jones, Nasdaq 100, or S&P 500 indexes, and zero
otherwise. We include the interaction term of our media coverage measure with each of the
proxies (Earnings_News * Large Size or Earnings_News * Stock Index Member) into equation
(1) to test our conjecture.
Table 4 reports the results. As shown, the estimated coefficients on the interaction terms
are negative for bank loan borrowing (columns (1) and (3)) and positive for bond issuance
(columns (2) and (4)). These results are unanimously consistent with the role of “separating
23
the wheat from the chaff” and the information complementary hypothesis. Using large firm
size as an example, the effect of earnings news is mainly driven by firms with above-median
total asset size in the sample. This information function is very different from that for small
firms. It also points out the limitations of the media that, if there is not enough publicly
available information in the first place (i.e. not much chaff) and if the news is not significant
or eye-catching (i.e. no wheat to pick), it is unlikely for the media to play an information role.
Besides the information role of media, we also focus on media’s role in facilitating
monitoring and corporate governance. A widely held view on one of the potential benefits of
bank borrowing is that it provides efficient and costly monitoring services that can mitigate
agency problems, whereas these services are not available from the widely dispersed investors
in the public market. Therefore, firms with poor governance problems or more monitoring
needs may choose bank debt. However, when the external governance pressure, such as the
media scrutiny is imposed, the monitoring needs of the bank will be reduced, and thus the
bank’s debt demands will be lowered (Bharath and Hertzel, 2018). In other words, although
firms with monitoring needs tend to rely on private debt, media coverage is expected to reduce
this reliance.
To further understand the governance channel through which the media can make a
difference, we examine the interactions between media coverage and two corporate governance
proxies—stock ownership concentration and whether a firm belongs to a relationship industry.
Existing studies have shown that the existence of multiple large shareholders enhances external
monitoring (e.g. Maury and Pajuste, 2005). Therefore, firms with multiple large shareholders
tend to have lower monitoring needs. We define High_Ownership_HHI as an indicator variable
that equals one if a firm’s Herfindahl-Hirschman index of its institutional ownership
concentration is above the sample median level of that year, and zero otherwise, and we include
the interaction term of High_Ownership_HHI and earnings news into our baseline model.
24
Our second governance proxy is whether a firm is in a relationship industry as defined
by Cremers, Nair, and Peyer (2008).13 The idea of using this variable to proxy for corporate
governance is that firms in such an industry have close ties with stakeholders, and more outside
monitoring may not be needed in the equity holders’ view because these stakeholders have an
incentive to monitor these firms. To test the conjecture on the corporate governance role of
media, we include the interaction term of our media coverage measure with a relationship
industry indicator into the baseline regression (i.e. Earnings_News * Relationship Industry).
As shown in Table 5, the estimates on the interaction terms between earnings news and high
ownership concentration (relationship lending industry) are positive on bank loan borrowing
(columns (1) and (3)) and negative on bond issuance (columns (2) and (4)).This result is
consistent with our corporate governance hypothesis that the negative (positive) relationship
between media coverage and bank loan (public debt) is mitigated among firms with good
corporate governance, suggesting the governance role played by the media.
4.5 Media coverage and borrowers’ credit ratings
The media is not the only information provider in the financial system. In this
subsection, we examine its relation with borrowers’ credit ratings. We focus on credit ratings
because they are important for debt instruments. Credit rating agencies are also major
information providers in the financial markets. As a preliminary step, in columns (1) and (2)
of Appendix Table 4, we show that firms with more news coverage have a significantly higher
chance of having an S&P domestic long-term issuer credit rating. The findings in columns (3)
and (4) indicate that firms with more news coverage also have significantly better credit ratings.
We then turn into news sentiments to examine whether the tone of the news that drives actions
because either positive or negative news should be the most relevant types of news for access
13 To be more specific, these industries include SIC codes of 15, 16, 17, 34, 35, 36, 37, 38, 39, 42, 47,
50, 51, 55, 60, 61, 62, 63, 64, 65, 67, 75, 76, and 87.
25
to the debt markets. We specifically focus on two types of news sentiments—positive and
negative earnings news, where positive (negative) earnings news is defined as news items
specializing in earnings evaluations with positive (negative) sentiment. Both measures are
expressed as ratios of total news. In Table 6, we examine the validity of information content
disseminated by the media and analyze whether news sentiments predict subsequent rating
changes and firm value. Results in Table 6 show that firms experiencing more positive
(negative) earnings news have significant increases (declines) in credit ratings and firm value.
These results validate the information content conveyed by our news measures. Otherwise, if
the news is stale, we should not observe systematic relations. To prevent the possibility that
news sentiments only transmit the same level of information as that transmitted by credit rating
changes, we control for it in the remaining tables.14 We also control for analyst earnings
forecast dispersion. By doing so, we can attribute the effects of news sentiments to media
coverage.
5. News Sentiments and New Debt Financing Activities
We start this section by revisiting the analysis of debt structure by replacing total news
or earnings news with news sentiments. The results are reported in Table 7. Interestingly, the
results we find in Table 2 are driven by negative news sentiment as none of the estimated
coefficients on positive earnings news are significantly different from zero. However, the
estimates on negative earnings news indicate that firms experiencing more negative news
coverage are related to significantly less bank loan usage but more bond and note issues.
It is possible that firms experiencing negative news sentiments have different
characteristics from firms with positive sentiments. Our results are confounded if any of these
14 For the same reason, we also control for earnings surprises to take care of additional earnings information
related to analysts. However, there are many missing values for this variable, so we do not report the results for
brevity. The results are available upon request.
26
characteristics also simultaneously contribute to the decision on debt choice. To address this
concern, we use propensity score matching approach. In particular, we compute news sentiment
by taking the difference between number of negative news and positive news over a year and
consider a firm as negative-news-firm if it is covered by more negative news than positive news
over the year. We then match negative-news-firms with non-negative-news-firms using
propensity scores and then estimate the average debt ratio between these two groups of firms.
The propensity scores are estimated via a probit model. The dependent variable is the dummy
variable for negative-news-firm and the independent variables include firm size, profitability,
sale growth, leverage, cash-to-asset ratio, market-to-book ratio, credit rating change, and
industry and year fixed effects. The matching estimator is Kernel-based matching techniques.
The results reported in Table 8 show that negative-news-firms consistently have lower bank
loans/total debt ratio and higher bonds & notes/ total debt ratio than similar non-negative-news
companies.
Since Capital IQ groups both privately placed debt and public bonds together, to
understand which one is driving the above results, we now turn to new issuance activities.
Before we examine each types of new borrowing, we analyze the level and sentiments of the
media coverage on debt access. Because Table 7 only shows the proportion of debt type relative
to total debt, we do not know whether the size of the pie itself is affected. In Table 9, we find
that more media coverage indeed enhances debt access in general, particularly the positive
earnings news. The likelihood of new debt financing is significantly reduced upon the release
of negative earning news.
In Table 10, we analyze new bank loans, public bonds, and privately placed debt. In
columns (2), (4), and (6), we examine negative news sentiments, and the results indicate that
firms with more negative news coverage significantly reduce the likelihood of borrowing new
bank loans and issuing new public bonds, but increase the likelihood of privately placing new
27
debt. The findings are consistent with the claim that the media plays a governance role in the
sense that it affects corporate behavior towards debt choices. When firms with public bond
access experiencing negative media coverage, they resort to private placement of debt rather
than bank loans. This finding is consistent with the points put forth in Carey, Post, and Sharpe
(1998). They show that, in the private lending markets, banks tend to serve lower-risk
borrowers, while non-bank financial institutions serve higher-risk borrowers. They argue that
this type of lending specialization is likely driven by regulations and capital requirements that
limit banks’ risk-taking ability. In addition, relationship banking means that banks tend to work
with firms when they are in financial distress, rather than forcing firms into liquidation.
Therefore, ex ante, banks only serve medium to low-risk firms.
Although lending terms have been examined in several papers, such as Bushman,
Williams, and Wittenberg-Moerman (2017), we also examine the effects of news sentiments
on price and non-price terms of new borrowings in Table 11 to complement research in this
area. Panels A, B, and C present the findings of bank loans, public bonds, and privately placed
debt, respectively. All estimated coefficients on news sentiments have expected signs,
particularly for interest spread and debt amount analyses. Despite controlling for rating changes
and earnings forecast dispersion, positive (negative) news significantly reduces (increases)
interest spread for both bank loans and public bonds. We do not analyze interest spread for
privately placed debt because the information is largely missing in SDC Platinum.
Improvements in ratings significantly reduce loan and bond interest spreads while earnings
forecast dispersion have no effects on both.
Firms with positive news coverage are able to raise significantly more money with bank
loans and public bonds, but there is no effect on privately placed debt. On the other hand, firms
experiencing negative news coverage raise significantly more capital through private
placement of debt and less with bank loans and public bonds. Finally, news sentiments only
28
affect bank loan maturity but have no effect on the other two types of debt instruments. Results
in Table 11 are largely consistent with extant literature in this area.
6. News Sentiment and Managers’ Post-Financing Earnings Management
Earnings management can be deliberately used by managers who focus on a desired
level of reported earnings. The choice of discretionary process may reflect the market sentiment
because news sentiment conveys investors’ belief about the companies’ cash flow and
investment risks that may not be available from accounting data. Ali and Gurun (2009)
document that optimism in the capital markets encourages managers to conduct more earnings
management. This is because the excess value generated by market optimism could allow
managers to report discretionary accounting earnings to their favorable outcomes. In the
pessimistic periods, however, managers are more conservative and thus may conduct
downward discretionary practices. In this section, we examine how the news sentiment affects
managers’ post-financing decisions on earnings management. In particular, we analyze
whether the positive or negative news affect managers’ report on discretionary accruals upward
or downward.
To examine earnings management, we look at discretionary accruals based on three
different models, that is, the Modified Jones Model (Jones, 1991; Dechow et al. 1995), the
models by McNichols (2002), and by Kothari et al. (2005). In the Modified Jones Model,
discretionary accruals are estimated using the residuals from the following regression:
𝑇𝐴𝑡
𝐴𝑇𝑡−1= 𝛼1 (
1
𝐴𝑇𝑡−1) + 𝛼2 (
∆𝑅𝐸𝑉𝑡−∆𝑅𝐸𝐶𝑡
𝐴𝑇𝑡−1) + 𝛼3 (
𝑃𝑃𝐸𝑡
𝐴𝑇𝑡−1) + 𝜀𝑡, (2)
where t is the hypothesized year of earnings management, ΔRECt is net receivables in year t
less net receivables in year t−1, ΔREVt is revenues in year t less revenues in year t−1, PPEt is
gross property plant and equipment at the end of year t, and ATt−1 is total assets at the end of
year t−1.
29
McNichols (2002)’s model is based on Dechow and Dichev (2002)’s model, where
discretionary accruals are estimated using residuals from the following regression:
𝑇𝐴𝑡
𝐴𝑇𝑡−1= 𝛼1 (
1
𝐴𝑇𝑡−1) + 𝛼2 (
𝐶𝐹𝑂𝑡−1
𝐴𝑇𝑡−2) + 𝛼3 (
𝐶𝐹𝑂𝑡
𝐴𝑇𝑡−1) + 𝛼4 (
𝐶𝐹𝑂𝑡+1
𝐴𝑇𝑡) + 𝛼5 (
∆𝑅𝐸𝑉𝑡
𝐴𝑇𝑡−1) + 𝛼6 (
𝑃𝑃𝐸𝑡
𝐴𝑇𝑡−1) + 𝜀𝑡,
(3)
Where CFO represents cash from operations and the other variables are defied the same as
before.
Kothari et al. (2005) estimate a model that is similar to the Jones and Modified Jones
Models, except that it is augmented to include firm profitability measure ROAt. This approach
is designed to provide a comparison of the effectiveness of performance matching versus
including a performance measure in the accruals regression. To put it formally, discretionary
accruals are estimated using residuals from the following regression:
𝑇𝐴𝑡
𝐴𝑇𝑡−1= 𝛼1 (
1
𝐴𝑇𝑡−1) + 𝛼2 (
∆𝑅𝐸𝑉𝑡
𝐴𝑇𝑡−1) + 𝛼3 (
𝑃𝑃𝐸𝑡
𝐴𝑇𝑡−1) + 𝛼4 (
𝑅𝑂𝐴𝑡
𝐴𝑇𝑡−1) + 𝜀𝑡, (4)
Using these three accrual-based models, we estimate the signed abnormal accruals and
test how news sentiments affect managerial discretionary accruals. The results are reported in
Table 12. The dependent variables in Table 12 are discretionary accruals based on the above
mentioned three models: modified Jones model (Dechow et al. 2005) (columns 1-2),
McNichols et al. (2002)’s model, and Kothari et al. (2005)’s model. All these discretionary
accruals are estimated one year after the new debt issuance. The explanatory variables used are
the same as in Table 10. From Table 12, we find that when the news is negative, that is, when
the markets are negative about firms’ future prospects, firms report more negative discretionary
accruals. This is consistent with our previous conjecture that managers may become more
conservative when the news coverage is pessimistic. However, we do not find that firms
manage their earnings upward after the debt financing when the news is positive. This could
because the new debt constrains managers from using their skills to opportunistically report
favorable earnings results.
30
7. Conclusions
By analyzing the effect of media coverage on firms’ debt financing, we find that media
coverage reduces firms’ reliance on bank loans but increases amounts of public bonds & notes
on their balance sheets. We document two channels that 1) such effect is concentrated among
firms with less information asymmetry problems suggesting that the media complements other
information sources (information channel), and 2) such effect is mitigated among firms with
better corporate governance suggesting that the media substitutes for banks on firm monitoring
(governance channel). By examining debt structure, this paper provides evidence that the media
serves both a complementary information role and a governance role in the debt markets.
We supplement the above analysis using new debt financing and find that higher level
of media coverage enhances new debt financing, particularly, positive earnings news.
However, negative earnings news significantly reduces new debt access, particularly public
bonds and bank loans, but increases the usage of private placement. It suggests that firms
experiencing negative news coverage refrain from issuing public bonds. Our results of news
sentiment on price and non-price terms of new borrowings corroborate this finding, i.e.
controlling for rating changes and earnings forecast dispersion, we find that positive (negative)
news significantly reduces (increases) interest spread for both bank loans and public bonds;
while there is no effect on privately placed debt with positive news coverage, firms with
negative news coverage increases capital significantly through private placement of debt.
We also investigate credit ratings and firm value changes post-media-coverage. We
show that news sentiments predict subsequent rating changes and firm values. It is worth to
note that although both news sentiments and rating changes affect borrowing terms, only news
sentiments have impacts on the choices of debt instruments. That is, firms experiencing more
positive (negative) earnings news have significant increases (declines) in credit ratings and
31
firm value. Finally, we analyze news sentiment impacts on managers’ report on discretionary
accruals and show that managers become more conservative in earnings reporting when the
news coverage is pessimistic, but there is no evidence that firms manage their earnings upward
with positive news after the debt financing. This result suggests that the new debt may constrain
managers to use their skills opportunistically to report favorable earnings results.
32
References
Ahern, K.R., Sosyura, D., 2014. Who writes the news? Corporate press releases during merger
negotiations. Journal of Finance 69, 241-291.
Ali, A., Gurun, U.G., 2009. Investor sentiment, accruals anomaly, and accruals management.
Journal of Accounting, Auditing & Finance 24, 415-431.
Bekaert, G., Harvey, C.R., Lundblad, C., 2005. Does financial liberalization spur growth?
Journal of Financial Economics 77, 3-55.
Bekaert, G., Harvey, C.R., Lundblad, C., Siegel, S., 2007. Global growth opportunities and
market integration. Journal of Finance 62 1081-1137.
Bharath, S., Hertzel, M., 2018. External governance and debt structure. Review of Financial
Studies, forthcoming.
Bharath, S., Sunder, J., Sunder, S.V., 2008. Accounting quality and debt contracting. The
Accounting Review 83, 1-28.
Bolton, P., Freixas, X., Shapiro, J., 2012. The credit ratings game. Journal of Finance 67, 85-
111.
Boubakri, N., Saffar, W., 2018. State ownership and debt choice: Evidence from privatization.
Journal of Financial and Quantitative Analysis, forthcoming.
Bushee, B.J., Core, J. E., Guay, W., Hamm, S.J.W., 2010. The role of the business press as an
information intermediary. Journal of Accounting Research 48, 1-20.
Bushman, R.M., Williams, C.D., Wittenberg-Moerman, R., 2017. The information role of the
media in private lending. Journal of Accounting Research 55, 115-152.
Carey, M., Post, M., Sharpe, S.A., 1998. Does corporate lending by banks and finance
companies differ? Evidence on specialization in private debt contracting. The Journal of
Finance 53, 845-878.
Cremers, M.K.J., Nair, V.B., Peyer, U., 2008. Takeover defenses and competition: The role of
stakeholders. Journal of Empirical Legal Studies 5, 791-818.
Dai, L., Parwada, J.T., and Zhang, B., 2015. The governance effect of the media’s news
dissemination role: Evidence from insider trading. Journal of Accounting Research 53,
331-366.
Dechow, P.M., Dichev, I.D., 2002. The quality of accruals and earnings: The role of accrual
estimation errors. The Accounting Review 77, 35-59.
Dechow, P.M., Sloan, R.G., Sweeney, A.P., 1995. Detecting earnings management. The
Accounting review 70, 193-225.
Denis, D.J., Mihov, V.T., 2003. The choice among bank debt, nonbank private debt, and public
debt: Evidence from new corporate borrowings. Journal of Financial Economics 70, 3-28.
Diamond, D., 1984. Financial intermediation and delegated monitoring. Review of Economic
Studies 51, 393–414.
Diamond, D., 1991. Monitoring and reputation: the choice between bank loans and directly
placed debt. Journal of Political Economy 99, 689–721.
33
Dyck, A., Volchkova, N., Zingales, L., 2008. The corporate governance role of the media:
Evidence from Russia. Journal of Finance 63, 1093-1135.
Engelberg, J., Parsons, C.A., 2011. The causal impact of media in financial markets, Journal
of Finance 66, 67-97.
Fama, E.F., 1985. What’s different about banks? Journal of Monetary Economics 15, 39-39.
Faulkender, M., Petersen, M.A., 2006. Does the source of capital affect capital structure?
Review of Financial Studies 19, 45-79.
Gao, H., Wang, J., Wu, C., 2016. Media coverage and cost of debt. Working paper.
Gertner, R., Scharfstein, D., 1991. A theory of workouts and the effects of reorganization law.
Journal of Finance 46, 1189-1222.
Goh, J.C., Ederington, L.H., 1993. Is a bond rating downgrade bad news, good news, or no
news for stockholders? Journal of Finance 48, 2001-2008.
Gurun, U.G., Butler, A.W., 2012. Don’t believe the hype: Local media slant, local advertising,
and firm value. Journal of Finance 67, 561-598.
Hasan, I., Hoi, C-K., Wu, Q., Zhang, H., 2017. Social capital and debt contracting: Evidence
from bank loans and public bonds. Journal of Financial and Quantitative Analysis 52,
1017-1047.
Hendershott, T., Livdan, D., Schurhoff, N., 2015. Are institutions informed about news?
Journal of Financial Economics 117, 249-287.
Houston, J., James, C., 1996. Bank information monopolies and the mix of private and public
debt claims. Journal of Finance 51, 1863-1889.
Jiang, H., Sun, Z., 2013. Understanding the illiquidity of corporate bonds: The arrival of public
news. Working paper, University of California, Irvine.
Jones, J.J., 1991. Earnings management during import relief investigations. Journal of
Accounting Research 29, 193-228.
Kale, J.R., Meneghetti, C., 2011. The choice between public and private debt: A survey. IIMB
Management Review 23, 5-14.
Kothari, S.P., Leone, A.J., Wasley, C.E., 2005. Performance matched discretionary accrual
measures. Journal of Accounting and Economics 39, 163-197.
Kothari, S.P., Li, X., Short, J.E., 2009. The effect of disclosures by management, analysts, and
business press on cost of capital, return volatility, and analyst forecasts: A study using
content analysis. The Accounting Review 84, 1639-1670.
Leland, H., Pyle, D., 1977. Information asymmetries, financial structure and financial
intermediaries. Journal of Finance 32, 371-387.
Li, X., Lin, C., Zhan, X., 2019. Does change in the information environment affect financing
choices? Management Science, forthcoming.
Lin, C., Ma, Y., Malatesta, P., Xuan, Y., 2013. Corporate ownership structure and the choice
between bank debt and public debt. Journal of Financial Economics 109, 517-534.
34
Liu, L.X., Sherman, A.E., Zhang, Y., 2014. The long-run role of the media: Evidence from
initial public offerings. Management Science 60, 1945-1964.
Michaely, R., Womack, K.L., 1999. Conflict of interest and credibility of underwriter analyst
recommendations. Review of Financial Studies 12, 653–686.
Miller, G.S., 2006. The press as a watchdog for accounting fraud. Journal of Accounting
Research 44, 1001-1033.
Mao, C.X., Song, W., 2018. Does reciprocity affect information production? Evidence from
syndication relationships in securities underwriting. Working Paper, Louisiana State
University.
Maury, B., Pajuste, A., 2005. Multiple large shareholders and firm value. Journal of Banking
and Finance 29, 1813-1934.
McNichols, M.F., 2002. Discussion of the quality of accruals and earnings: The role of accrual
estimation errors. The Accounting Review 77, 61-69.
Park, C., 2000. Monitoring and structure of debt contracts. Journal of Finance 55, 2157-2195.
Rajan, R., 1992. Insiders and outsiders: the choice between informed and arm's-length debt.
Journal of Finance 47, 1367–1400.
Rauh, J.D., Sufi, A., 2010. Capital structure and debt structure. Review of Financial Studies 23,
4242-4280.
Rogers, J.L., Skinner, D.J., Zechman, S.C., 2016. The role of the media in disseminating
insider-trading news. Review of Accounting Studies 21, 711-739.
Solomon, D.H., 2012. Selective publicity and stock prices. Journal of Finance 67, 599-637.
Stiglitz, J., Weiss, A., 1983. Incentive effects of terminations: Applications to credit and labor
markets. American Economic Review 73, 912-927.
Tetlock, P.C., 2007. Giving content to investor sentiment: The role of media in the stock market.
Journal of Finance 62, 1139-1168.
Tetlock, P.C., Saar-Tsechansky, M., Macskassy, S., 2008. More than words: Quantifying
language to measure firms’ fundamentals. Journal of Finance 63, 1437-1467.
Tetlock, P.C., 2011. All the news that’s fit to reprint: Do investors react to stale information?
Review of Financial Studies 24, 1481-1512.
Tetlock, P.C., 2015. The role of media in Finance, Handbook of Media Economics, vol. 1A,
(Chapter 18), Anderson, Waldfogel, and Stromberg (Eds.), Elsevier, Amsterdam, 701-721.
You, J., Zhang, B., Zhang, L., 2018. Who captures the power of the pen? Review of Financial
Studies 31, 43-96.
35
Figure 1. Media Coverage and Total Debt
This figure shows the trend of media coverage and total debt by year. Media coverage is the median
of the total number of news items covering each firm in each year, and total debt is the median of
the total amount of debt (in million $) borrowed by each firm in each year.
0
50
100
150
200
250
300
350
400
2000 2002 2004 2006 2008 2010 2012 2014
Year
Median Coverage and Total Debt
Total # of News Total Debt ($million)
36
Table 1. Summary Statistics
This table presents the summary statistics for the key variables used in the paper. The number of
observations is 36,002.
Variable Mean SD P25 Median P75
Total_News 6.07 3.04 3.70 5.39 8.65
Earnings_News 1.77 2.09 1.70 2.38 2.85
Positive Earnings News 0.34 2.78 0.15 1.56 2.23
Negative Earnings News 0.23 2.78 -0.25 1.31 2.12
Dummy_Bank_Loans 0.74 0.44 0 1 1
Dummy_Bonds & Notes 0.64 0.48 0 1 1
Bank Loans/Total Debt 0.43 0.4 0 0.33 0.88
Bonds & Notes/Total Debt 0.41 0.41 0 0.3 0.85
Size (in million $) 5606.32 48899.73 25.09 245.60 1420.81
Size (in log) 5.26 2.79 3.22 5.50 7.26
ROA 1.71 3.51 0.01 0.08 1.15
Sales Growth 0.09 0.2 -0.03 0.06 0.18
Leverage 0.88 1.30 0.13 0.31 0.82
Cash Flow 2.93 5.82 0.04 0.14 2.27
Market to Book 2.52 2.07 1.13 1.87 3.17
37
Table 2. Media Coverage and Debt Structure - Baseline
This table presents results of media coverage on debt structure. The sample includes all the U.S. firms that have either borrowed banks loans or
issued bonds or both during the sample period 2001-2014. The key media coverage variables are Total News (columns 1, 3, 5, and 7) and Earnings
News (columns 2, 4, 6 and 8), where Total News is equal to Ln(total number of news items for firm i in year t, scaled by firm i’s total assets);
Earnings News is equal to Ln(the ratio of news items specializing in earnings evaluations to total news for firm i in year t*100+0.01). The dependent
variables are Dummy_Bank_Loan (columns 1-2), Bank Loans/Total Debt (columns 3-4), Dummy_Bonds & Notes (columns 5-6), and Bonds &
Notes/Total Debt (columns 7-8). Detailed definitions of variables can be found in Appendix Table 1. Standard errors reported in parentheses are
heteroskedastic robust and clustered at the industry level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.
(1) (2) (3) (4) (5) (6) (7) (8)
Model Logit OLS Logit OLS
Dep Var Dummy_Bank_Loans Bank Loans/Total Debt Dummy_Bonds & Notes Bonds & Notes/Total Debt
Total_News -0.2796*** -0.0316*** 0.0812*** 0.0228*** (0.0243) (0.0032) (0.0235) (0.0030) Earnings_News -0.0278*** -0.0069*** 0.0253*** 0.0037***
(0.0078) (0.0010) (0.0069) (0.0009)
Size -0.2434*** -0.0588*** -0.0917*** -0.0701*** 0.5755*** 0.5163*** 0.0862*** 0.0708***
(0.0173) (0.0073) (0.0023) (0.0011) (0.0178) (0.0088) (0.0022) (0.0010)
ROA 0.0074 0.0107** 0.0071*** 0.0072*** -0.0494*** -0.0492*** -0.0039*** -0.0040***
(0.0054) (0.0054) (0.0008) (0.0008) (0.0054) (0.0054) (0.0008) (0.0008)
Sales Growth 0.2388*** 0.2002*** 0.0721*** 0.0670*** -0.3293*** -0.3103*** -0.0889*** -0.0855***
(0.0686) (0.0684) (0.0099) (0.0099) (0.0713) (0.0712) (0.0094) (0.0094)
Leverage 0.2496*** 0.2647*** 0.0005 0.0023 0.4399*** 0.4357*** 0.0460*** 0.0448***
(0.0168) (0.0167) (0.0023) (0.0023) (0.0183) (0.0182) (0.0022) (0.0022)
Cash Flow -0.0678*** -0.0703*** -0.0075*** -0.0078*** -0.0050 -0.0044 0.0052*** 0.0054***
(0.0032) (0.0032) (0.0005) (0.0005) (0.0032) (0.0031) (0.0004) (0.0004)
Market_to_Book -0.0494*** -0.0655*** -0.0058*** -0.0076*** -0.0418*** -0.0378*** 0.0015 0.0028***
(0.0065) (0.0064) (0.0010) (0.0010) (0.0075) (0.0073) (0.0010) (0.0010)
Industry fixed effects yes yes yes yes yes yes yes yes
Year fixed effects yes yes yes yes yes yes yes yes
N 35994 35994 36002 36002 35888 35888 36002 36002
R-sq 0.1187 0.1152 0.2079 0.2068 0.2517 0.2518 0.3154 0.3146
38
Table 3. Media Coverage and Debt Structure – IV Estimation
This table presents results of media coverage on debt structure using the instrumental variable (IV) approach. The sample includes all
the U.S. firms that have either borrowed banks loans or issued bonds or both during the sample period 2001-2014. Columns 1-4 present
second-stage results where the key explanatory variables of media coverage are Total News (columns 1 and 3) and Earnings News
(columns 2 and 4). The dependent variables are Bank Loans/Total Debt (columns 1-2) defined as the ratio of a firm’s bank loans to total
debt, and Bonds & Notes/Total Debt (columns 3-4) defined as the ratio of a firm’s public/private bonds and notes to total debt. Columns
5-6 present first-stage results where the instrumental variables used are (1) Newspaper Shutdown, defined as the shut-down of
metropolitan or local newspapers in each firm’s headquarter state in year t, and (2) State Growth Opportunity, defined as in Bakaert et
al. (2005) and equal to ∑ 𝑤𝑆𝐼𝑡 ∗ 𝑃𝐸𝐼𝑡 where each state has a set of industries 𝑤𝑆𝐼𝑡 with each industry having time-varying growth
opportunities while assuming that these growth prospects are reflected in the price to earnings (P/E) ratios of the country’s industry
portfolio 𝑃𝐸𝐼𝑡. Detailed definitions of all the other variables can be found in Appendix Table 1. Standard errors reported in parentheses
are heteroskedastic robust and clustered at the industry level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1%
levels, respectively.
(1) (2) (3) (4) (5) (6)
Second Stage Second Stage First Stage
Dep Var Bank Loans/Total Debt Bonds & Notes /Total Debt Total_News Earnings_News
Total_News -0.1075*** 0.0874**
(0.0385) (0.0373)
Earnings_News -0.0090** 0.0072**
(0.0036) (0.0030)
Size -0.1408*** -0.0698*** 0.1280*** 0.0704*** -0.8226*** 0.1264***
(0.0242) (0.0041) (0.0247) (0.0050) (0.0028) (0.0044)
ROA 0.0068*** 0.0073*** -0.0037* -0.0041** 0.0681*** 0.0230***
(0.0023) (0.0020) (0.0022) (0.0021) (0.0018) (0.0027)
Sales Growth 0.0804*** 0.0670*** -0.0959*** -0.0851*** -0.1585*** -0.3661***
(0.0170) (0.0189) (0.0218) (0.0215) (0.0256) (0.0453)
Leverage -0.0028 0.0024 0.0488*** 0.0446*** 0.2961*** 0.1632***
(0.0106) (0.0098) (0.0092) (0.0087) (0.0049) (0.0080)
Cash Flow -0.0069*** -0.0078*** 0.0046*** 0.0054*** 0.0659*** 0.0221***
39
(0.0017) (0.0015) (0.0013) (0.0013) (0.0011) (0.0017)
Market_to_Book -0.0012 -0.0073*** -0.0024 0.0026 0.0434*** 0.0128***
(0.0033) (0.0026) (0.0029) (0.0018) (0.0028) (0.0039)
Newspaper Shutdown 0.0687*** 0.0349**
(0.0104) (0.0153)
State Growth Opportunity 0.0593*** 0.8564***
(0.0064) (0.0049)
Weak Instruments Test
(Cragg-Donald Wald F
value) 228 3923
228
3923 - -
Industry fixed effects yes yes yes yes yes yes
Year fixed effects yes yes yes yes yes yes
N 35931 35931 35931 35931 35931 35931
R-sq 0.1954 0.2067 0.3073 0.3149 0.8998 0.2901
40
Table 4. Information Channel
This table presents results of media coverage on debt structure and media coverage interacting
with information transparency proxies. The sample includes all the U.S. firms that have either
borrowed bank loans or issued bonds or both during the sample period 2001-2014. The key media
coverage variable is Earnings News, which is equal to Ln(the ratio of news items specializing in
earnings evaluations with either positive or negative sentiment to total news for firm i in year
t*100+0.01), where total news is defined as total number of news items with a relevance score of
100 but excluding corporate press releases for firm i in year t. Information asymmetry proxies used
include Large Size and Stock Index Member with higher values indicating less information
asymmetry. Large Size is a dummy variable that equals one if a firm’s total assets are above the
sample median of that year, and zero otherwise. Stock Index Member is a dummy variable that
equals one if a firm is a major stock index member, including Dow Johns, Nasdaq 100, or S&P
500 indexes, and zero otherwise. Detailed definitions of all the other variables can be found in
Appendix Table 1. Standard errors reported in parentheses are heteroskedastic robust and clustered
at the industry level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels,
respectively.
(1) (2) (3) (4)
Dep Var
Bank
Loans/Total
Debt
Bonds &
Notes /Total
Debt
Bank
Loans/Total
Debt
Bonds &
Notes /Total
Debt
Earnings_News -0.0030 -0.0034 -0.0063*** 0.0030*
(0.0025) (0.0033) (0.0021) (0.0016)
Earnings_News * Large Size -0.0100*** 0.0194***
(0.0026) (0.0028)
Earnings_News * Stock Index
Member
-0.0250*** 0.0341***
(0.0069) (0.0088)
Stock Index Member -0.0764** 0.0663
(0.0354) (0.0403)
Firm controls yes yes yes yes
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 36002 36002 36002 36002
R-sq 0.2115 0.3228 0.2155 0.3248
41
Table 5. Corporate Governance Channel
This table presents results of media coverage on debt structure and media coverage interacting
with corporate governance proxies. The sample includes all the U.S. firms that have either
borrowed bank loans or issued bonds or both during the sample period 2001-2014. The key media
coverage variable is Earnings News, which is equal to Ln(the ratio of news items specializing in
earnings evaluations with either positive or negative sentiment to total news for firm i in year
t*100+0.01), where total news is defined as total number of news items with a relevance score of
100 but excluding corporate press releases for firm i in year t. Corporate governance proxies used
include High_Ownership_HHI and Relationship_Industry. High_Ownership_HHI is an indicator
variable that equals one if a firm’s Herfindahl-Hirschman index of its institutional ownership
concentration is above the sample median level of that year, and zero otherwise. Relationship
Industry is an indicator variable that equals one if a firm belongs to the following industries with
SIC codes of 15, 16, 17, 34, 35, 36, 37, 38, 39, 42, 47, 50, 51, 55, 60, 61, 62, 63, 64, 65, 67, 75,
76, and 87 (see Cremers, Nair, and Peyer, 2008). Detailed definitions of all the other variables can
be found in Appendix Table 1. Standard errors reported in parentheses are heteroskedastic robust
and clustered at the industry level. *, **, and *** indicate statistical significance at the 10%, 5%,
and 1% levels, respectively.
(1) (2) (3) (4)
Dep Var
Bank
Loans/Total
Debt
Bonds &
Notes /Total
Debt
Bank
Loans/Total
Debt
Bonds &
Notes /Total
Debt
Earnings_News -0.0140*** 0.0140*** -0.0121*** 0.0090***
(0.0035) (0.0031) (0.0029) (0.0027)
Earnings_News * High
Ownership_HHI 0.0103*** -0.0142***
(0.0029) (0.0034)
Earnings_News * Relationship
Industry 0.0077** -0.0077**
(0.0032) (0.0034)
Firm controls yes yes yes yes
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 35192 35192 36002 36002
R-sq 0.2126 0.3226 0.2071 0.3149
42
Table 6. The Information Content of News Sentiments
This table presents the results of news sentiments on credit rating changes and firm value, where
news sentiments are measured using Positive_Earnings_News (column 1) and
Negative_Earnings_News (column 2). Positive_Earnings_News (Negative_Earnings_News) is
defined as the ratio of news items specializing in earnings evaluations with positive (negative)
sentiment to total news for firm i in year t. For both measures, we multiple them by 100 before
taking the natural logarithm. The sample includes all the U.S. firms that have either borrowed bank
loans or issued bonds or both during the sample period 2001-2014. The dependent variable
Ratings_Change is defined as the credit rating score at the end of year t+1 minus that of year t.
The credit rating score is coded from 0 to 22 with 0 representing no rating, 1 representing D or
below, and 22 representing the highest S&P rating AAA. Tobin’s Q is used as a proxy of firm
value. Detailed definitions of all the other variables can be found in Appendix Table 1. Standard
errors reported in parentheses are heteroskedastic robust and clustered at the industry level. *, **,
and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.
(1) (2) (3) (4)
Dep Var Ratings_Change Tobin’s Q (t+1)
Positive Earnings News 0.0049*** 0.0305***
(0.0013) (0.0036)
Negative Earnings News -0.0067*** -0.0053**
(0.0015) (0.0025)
Size -0.0122*** -0.0088*** -0.0470*** -0.0306**
(0.0023) (0.0019) (0.0142) (0.0134)
ROA 0.0072*** 0.0068*** 0.0467*** 0.0473***
(0.0013) (0.0013) (0.0088) (0.0090)
Sales Growth 0.0725*** 0.0678*** 0.3124*** 0.3417***
(0.0243) (0.0231) (0.0351) (0.0368)
Leverage -0.0079* -0.0068 -0.0972*** -0.0971***
(0.0043) (0.0043) (0.0188) (0.0186)
Cash Flow -0.0008 -0.0007 0.0149*** 0.0154***
(0.0009) (0.0009) (0.0028) (0.0029)
Market_to_Book 0.0108*** 0.0112*** 0.1742*** 0.1782***
(0.0022) (0.0021) (0.0080) (0.0082)
Ratings_Change 0.0281*** 0.0303***
(0.0064) (0.0066)
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 36002 36002 35981 35981
R-sq 0.0174 0.0180 0.9579 0.9570
43
Table 7 News Sentiments and Debt Structure
This table presents the results of news sentiments on debt structure, where news sentiments are
measured using Positive_Earnings_News (columns 1 and 3) and Negative_Earnings_News
(columns 2 and 4). Positive_Earnings_News (Negative_Earnings_News) is defined as the ratio of
news items specializing in earnings evaluations with positive (negative) sentiment to total news
for firm i in year t. For both measures, we multiple them by 100 before taking the natural logarithm.
The sample includes all the U.S. firms that have either borrowed bank loans or issued bonds or
both during the sample period 2001-2014. Detailed definitions of all the other variables can be
found in Appendix Table 1. Standard errors reported in parentheses are heteroskedastic robust and
clustered at the industry level. *, **, and *** indicate statistical significance at the 10%, 5%, and
1% levels, respectively.
(1) (2) (3) (4)
Dep Var Bank Loans/Total Debt Bonds & Notes/Total Debt
Positive Earnings News 0.0001 -0.0003
(0.0012) (0.0014)
Negative Earnings News -0.0093*** 0.0061***
(0.0020) (0.0017)
Size -0.0715*** -0.0704*** 0.0718*** 0.0710***
(0.0043) (0.0039) (0.0047) (0.0049)
ROA 0.0073*** 0.0066*** -0.0043** -0.0038*
(0.0020) (0.0019) (0.0021) (0.0020)
Sales Growth 0.0681*** 0.0537*** -0.0859*** -0.0767***
(0.0195) (0.0197) (0.0228) (0.0225)
Leverage 0.0016 0.0032 0.0455*** 0.0445***
(0.0102) (0.0099) (0.0088) (0.0086)
Cash Flow -0.0079*** -0.0077*** 0.0055*** 0.0054***
(0.0015) (0.0015) (0.0013) (0.0013)
Market_to_Book -0.0078*** -0.0081*** 0.0029 0.0031
(0.0027) (0.0027) (0.0019) (0.0019)
Ratings_Change 0.0056 0.0028 -0.0031 -0.0013
(0.0043) (0.0043) (0.0042) (0.0043)
Forecast Dispersion 0.0097 0.0156* -0.0189** -0.0228***
(0.0092) (0.0086) (0.0073) (0.0074)
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 36002 36002 36002 36002
R-sq 0.2058 0.2091 0.3145 0.3159
44
Table 8. Media Coverage and Debt Structure – Propensity Score Matching Estimation
This table presents the relationship between media coverage and debt structure using propensity
score matching approach. The sample includes all the U.S. firms that have either borrowed bank
loans or issued bonds or both during the sample period 2001-2014. The propensity scores are
estimated via a probit model - the dependent variable is the dummy variable for negative-news-
firm and the independent variables include firm size, profitability, sale growth, leverage, cash-to-
asset ratio, market-to-book ratio, credit rating change, and industry and year fixed effects. The
matching estimator is Kernel-based matching techniques.
Variable Treated Controls Difference t-statistic
Bank Loans/Total Debt 0.428 0.461 -0.032 6.14
Bonds & Notes /Total Debt 0.413 0.396 0.017 3.24
45
Table 9 Media Coverage and Debt Access
This table presents the logit regression results of media coverage on debt access. The dependent
variable equals one if a firm has issued either bank loans, public bonds, or placed private debt in
year t, and 0 otherwise. The sample includes all the U.S. firms that have either borrowed bank
loans or issued bonds or both during the sample period 2001-2014. Detailed definitions of all the
other variables can be found in Appendix Table 1. Standard errors reported in parentheses are
heteroskedastic robust and clustered at the industry level. *, **, and *** indicate statistical
significance at the 10%, 5%, and 1% levels, respectively.
(1) (2) (3) (4)
Dep Var New Debt Issuance
Total News 0.2918***
(0.0556) Earnings News 0.0554**
(0.0231) Positive Earnings News 0.1008***
(0.0172)
Negative Earnings News -0.0309**
(0.0135)
Size 1.1719*** 0.9901*** 0.9602*** 0.9995*** (0.0413) (0.0213) (0.0220) (0.0214)
ROA 0.0847*** 0.0826*** 0.0792*** 0.0817*** (0.0132) (0.0133) (0.0132) (0.0133)
Sales Growth -0.3927** -0.3413** -0.4587*** -0.4072** (0.1572) (0.1580) (0.1605) (0.1602)
Leverage 0.3342*** 0.3172*** 0.3160*** 0.3208*** (0.0370) (0.0370) (0.0367) (0.0369)
Cash Flow -0.0289*** -0.0260*** -0.0267*** -0.0257*** (0.0075) (0.0075) (0.0075) (0.0075)
Market_to_Book 0.0292** 0.0432*** 0.0355** 0.0422***
(0.0144) (0.0142) (0.0143) (0.0142)
Ratings_Change 0.0129 0.0066 -0.0033 -0.0063 (0.0486) (0.0489) (0.0488) (0.0491)
Forecast Dispersion -0.1581 -0.1525 -0.1416 -0.1211
(0.1137) (0.1137) (0.1135) (0.1141)
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 15597 15597 15597 15597
R-sq 0.5479 0.5462 0.5477 0.5462
46
Table 10 News Sentiments and New Borrowings
This table presents the logit regression results of news sentiments on the choice of new borrowings. The sample includes all the U.S.
firms that have either borrowed new bank loans or issued new bonds or both during the sample period 2001-2014. The dependent
variables are New_Bank_Loans (columns 1-2), New_Public_Bonds (columns 3-4), or New_Private_Placement (columns 5-6), where
New_Bank_Loans is an indicator variable that equals one if a firm has a new bank loan borrowing in year t, and zero otherwise;
New_Public_Bonds is an indicator variable that equals one if a firm has a new bond issuing in year t, and zero otherwise; and
New_Private_Placement is an indicator variable that equals one if a firm has placed private debt in year t, and zero otherwise. Firm
controls include ROA, Sales Growth, Leverage, Cash Flow, and Market_to_Book. Detailed definitions of all the other variables can be
found in Appendix Table 1. Standard errors reported in parentheses are heteroskedastic robust and clustered at the industry level. *, **,
and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.
(1) (2) (3) (4) (5) (6)
Dep Var New Bank Loans New Public Bonds New Private Placement
Positive Earnings News 0.1021*** 0.0833*** 0.0883 (0.0168) (0.0269) (0.0646)
Negative Earnings News -0.0287** -0.0390** 0.1382**
(0.0132) (0.0190) (0.0621)
Size 0.8333*** 0.8722*** 0.8941*** 0.9189*** 0.3814*** 0.4146*** (0.0199) (0.0190) (0.0260) (0.0253) (0.0722) (0.0809)
Market_to_Book -0.0003 0.0063 0.0872*** 0.0910*** -0.0165 0.0006
(0.0140) (0.0139) (0.0162) (0.0161) (0.0725) (0.0657)
Ratings_Change -0.0045 0.0090 -0.1447 -0.1308 -0.0461 0.0051
(0.0926) (0.0934) (0.1329) (0.1335) (0.1648) (0.1606)
Earnings Surprise -0.0289 -0.0306 -0.0155 -0.0214 -0.3238 -0.4314
(0.0492) (0.0495) (0.0571) (0.0577) (0.6772) (0.6879)
Firm controls yes yes yes yes yes yes
Industry fixed effects yes yes yes yes yes yes
Year fixed effects yes yes yes yes yes yes
N 15597 15597 15597 15597 10781 10781
R-sq 0.4869 0.4849 0.4253 0.4248 0.1823 0.1861
47
Table 11 News Sentiments and Borrowing Price and Non-Price Terms
This table presents the results of news sentiments on borrowing price and non-price terms of bank loans, public bonds, and privately
placed debt in Panels A, B, and C, respectively. The sample is conditional on new debt financing activities available for analysis from
all the U.S. firms that have either borrowed bank loans or issued bonds or both during the sample period 2001-2014. Interest rate (spread)
is not available for privately placed debt. All the price and non-price terms are expressed in natural logarithm. Firm controls include
ROA and Leverage. Detailed definitions of all the other variables can be found in Appendix Table 1. Standard errors reported in
parentheses are heteroskedastic robust and clustered at the industry level. *, **, and *** indicate statistical significance at the 10%, 5%,
and 1% levels, respectively.
(1) (2) (3) (4) (5) (6)
Dep Var Bank_Loans_Spread Bank_Loans_Amount Bank_Loans_Maturity
Panel A: Bank Loans
Positive Earnings News -0.0194*** 0.0254*** 0.0046** (0.0021) (0.0044) (0.0021) Negative Earnings News 0.0246*** -0.0129*** -0.0038**
(0.0018) (0.0039) (0.0019)
Maturity 0.1052*** 0.1061*** 0.1539*** 0.1546*** (0.0073) (0.0073) (0.0147) (0.0147) Amount -0.1093*** -0.1094*** 0.0355*** 0.0356***
(0.0037) (0.0037) (0.0034) (0.0034)
Size -0.0933*** -0.1048*** 0.4875*** 0.4996*** -0.0295*** -0.0272*** (0.0035) (0.0034) (0.0059) (0.0056) (0.0033) (0.0032)
Ratings_Change -0.0237*** -0.0187*** -0.0058 -0.0076 0.0365*** 0.0358***
(0.0058) (0.0058) (0.0124) (0.0124) (0.0059) (0.0060)
Forecast Dispersion 0.0913*** 0.0722*** 0.0765*** 0.0878*** 0.0276** 0.0306**
(0.0125) (0.0125) (0.0266) (0.0268) (0.0128) (0.0129)
Loan purposes yes yes yes yes yes yes
Firm controls yes yes yes yes yes yes
Industry fixed effects yes yes yes yes yes yes
Year fixed effects yes yes yes yes yes yes
N 17807 17807 19940 19940 19940 19940
R-sq 0.5206 0.5231 0.4254 0.4247 0.2387 0.2386
48
Table 11 (Continued)
(1) (2) (3) (4) (5) (6)
Dep Var Public_Bond_Interest Spread Public_Bond_Amount Public_Bond_Maturity
Panel B: Public Bonds
Positive Earnings News -0.0821*** 0.1687*** -0.0192 (0.0083) (0.0147) (0.0117) Negative Earnings News 0.0449*** -0.0359*** 0.0001
(0.0060) (0.0106) (0.0081)
Maturity 0.2274*** 0.2288*** 0.1575*** 0.1560*** (0.0134) (0.0135) (0.0211) (0.0215) Amount 0.0582*** 0.0462*** 0.0965*** 0.0925***
(0.0096) (0.0095) (0.0129) (0.0127)
Size -0.2376*** -0.2485*** 0.2240*** 0.2408*** -0.1201*** -0.1205*** (0.0101) (0.0103) (0.0176) (0.0180) (0.0140) (0.0141)
Ratings_Change -0.0406*** -0.0321** -0.1020*** -0.1090*** 0.0148 0.0142
(0.0142) (0.0144) (0.0253) (0.0259) (0.0199) (0.0200)
Forecast Dispersion 0.3065*** 0.2983*** 0.1827*** 0.1560** -0.0502 -0.0433
(0.0399) (0.0404) (0.0689) (0.0704) (0.0539) (0.0543)
Firm controls yes yes yes yes yes yes
Industry fixed effects yes yes yes yes yes yes
Year fixed effects yes yes yes yes yes yes
N 3240 3240 3676 3676 3676 3676
R-sq 0.6015 0.5964 0.6045 0.5914 0.1773 0.1766
49
Table 11 (Continued)
(1) (2) (3) (4)
Dep Var Private_Placed_Amount Private_Placed_Maturity
Panel C: Private Placement
Positive Earnings News -0.0118 -0.0026 (0.0173) (0.0120) Negative Earnings News 0.0604*** -0.0011
(0.0204) (0.0145)
Maturity -0.1659** -0.1608** (0.0755) (0.0746) Amount -0.0801** -0.0794**
(0.0364) (0.0369)
Size 0.1840*** 0.1464*** 0.0228 0.0227 (0.0309) (0.0326) (0.0225) (0.0235)
Ratings_Change -0.1884* -0.2124** 0.1403* 0.1422*
(0.1073) (0.1062) (0.0745) (0.0747)
Forecast Dispersion 0.0811 0.0896 0.3190 0.3093
(0.3256) (0.3170) (0.2256) (0.2222)
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 429 429 429 429
R-sq 0.3485 0.3631 0.3051 0.3050
50
Table 12. News Sentiments and Earnings Management
This table presents the results of news sentiments on firms’ earnings management, where news sentiments are measured using
Positive_Earnings_News (column 1) and Negative_Earnings_News (column 2). Positive_Earnings_News (Negative_Earnings_News)
is defined as the ratio of news items specializing in earnings evaluations with positive (negative) sentiment to total news for firm i in
year t. For both measures, we multiple them by 100 before taking the natural logarithm. The sample includes all the U.S. firms that have
either borrowed bank loans or issued bonds or both during the sample period 2001-2014. The dependent variable are discretionary
accrual based on modified Jones model (Jones, 1991; Dechow et al. 1995), discretionary accrual based on McNichols et al. (2002)’s
model, and discretionary accrual based on Kothari et al. (2005)’s model. Firm controls include ROA, Sales Growth, Leverage, Cash
Flow, and Market_to_Book. Detailed definitions of all the other variables can be found in Appendix Table 1. Standard errors reported
in parentheses are heteroskedastic robust and clustered at the industry level. *, **, and *** indicate statistical significance at the 10%,
5%, and 1% levels, respectively.
(1) (2) (5) (6) (7) (8)
Dep Var EM-Modified Jones (t+1) EM – McNichols Model (t+1) EM – Kothari Model (t+1)
Positive Earnings News 0.0003 0.0006 0.0003
(0.0007) (0.0006) (0.0007)
Negative Earnings News -0.0028*** -0.0025*** -0.0022***
(0.0007) (0.0006) (0.0008)
Size 0.0073*** 0.0076*** 0.0039** 0.0044** 0.0076*** 0.0078***
(0.0024) (0.0022) (0.0018) (0.0017) (0.0026) (0.0024)
Market_to_Book -0.0014* -0.0015* -0.0010 -0.0010 -0.0016** -0.0017**
(0.0008) (0.0008) (0.0008) (0.0009) (0.0007) (0.0008)
Ratings_Change 0.0025 0.0021 0.0023 0.0020 0.0035 0.0032
(0.0035) (0.0035) (0.0033) (0.0032) (0.0035) (0.0034)
Firm controls yes yes yes yes yes yes
Industry fixed effects yes yes yes yes yes yes
Year fixed effects yes yes yes yes yes yes
N 23556 23556 22143 22143 23097 23097
R-sq 0.0515 0.0529 0.0440 0.0453 0.0462 0.0471
51
Appendix Table 1. Variable Definitions
This table presents variable definitions of the key variables used in the paper.
Variable Name Definition
Panel A: Media Coverage Variables
Total_News Ln(total number of news items with a relevance score of 100
but excluding corporate press releases for firm i in year t, scaled
by firm i’s total assets in billion $). The news package includes
Dow Jones, Web and PR Editions.
Earnings_News Ln(the ratio of news items specializing in earnings evaluations
with either positive or negative sentiment to total news for firm
i in year t*100+0.01), where total news is defined as total
number of news items with a relevance score of 100 but
excluding corporate press releases for firm i in year t.
Positive Earnings News Ln(the ratio of news items specializing in earnings evaluations
with positive sentiment to total news for firm i in year t
*100+0.01), where total news is defined as total number of
news items with a relevance score of 100 but excluding
corporate press releases for firm i in year t.
Negative Earnings News Ln(the ratio of news items specializing in earnings evaluations
with negative sentiment to total news for firm i in year t
*100+0.01), where total news is defined as total number of
news items with a relevance score of 100 but excluding
corporate press releases for firm i in year t.
Panel B: Debt Financing Variables
Dummy_Bank_Loans A dummy variable that equals one if a firm has bank loans
outstanding in year t, and zero otherwise.
Bank Loans/Total Debt The ratio of bank loans to total debt in year t, where total debt
include bank loans, bonds & notes, and other debt items, such
as leases.
Dummy_Bonds & Notes A dummy variable that equals one if a firm has public/private
bonds and notes outstanding in year t, and zero otherwise.
Bonds & Notes /Total Debt The ratio of public/private bonds and notes to total debt in year
t, where total debt include bank loans, bonds & notes, and other
debt items, such as leases.
New_Bank_Loans An indicator variable that equals one if a firm has borrowed
new bank loans in year t, and zero otherwise.
New_Public_Bonds An indicator variable that equals one if a firm has issued new
public bonds in year t, and zero otherwise.
52
New_Private_Placement An indicator variable that equals one if a firm has placed
private debt in year t, and zero otherwise.
New_Debt_Issuance An indicator variable that equals one if a firm has borrowed
new bank loans, or has issued new public bonds or placed new
private debt in year t, and zero otherwise.
Bank_Loans_Spread The natural logarithm of bank loan spread, where loan spread
is measured as the all-in-drawn spread above LIBOR.
Bank_Loans_Maturity The natural logarithm of bank loan maturity where maturity is
measured in months.
Bank_Loans_Amount The natural logarithm of bank loan amount where the amount
is measured in million U.S. dollars.
Public_Bonds_Interest
Spread
The natural logarithm of bond interest spread in basis points
and net of yield of treasury securities of comparable maturity.
Public_Bonds_Maturity The natural logarithm of public bond maturity where maturity
is measured in months.
Public_Bonds_Amount The natural logarithm of public bond amount where the amount
is measured in million U.S. dollars.
Private_Placed_Maturity The natural logarithm of privately placed debt maturity where
maturity is measured in months.
Private_Placed_Amount The natural logarithm of privately placed debt amount where
the amount is measured in million U.S. dollars.
Panel C: Firm Level Variables
Size The natural logarithm of total assets.
ROA Net operating income over total assets.
Leverage Total debt over total assets.
Market_to_Book Market value of equity plus book value of debt divided by book
value of total assets.
Sales Growth The percentage of annual growth in total sales.
Cash Flow Cash flows from operations deflated by total assets.
Dummy_Rating A dummy variable that equals one if a firm has an S&P credit
rating, and zero otherwise.
S&P_Rating Defined as Ln (1+ credit rating) where credit rating is coded
from 0 to 22 with 0 representing no rating, 1 representing D or
below, and 22 representing the highest S&P rating AAA.
Ratings_Change Credit rating score at the end of year t minus that at year t-1.
Credit rating score is coded from 0 to 22 with 0 representing
no rating, 1 representing D or below, and 22 representing the
highest S&P rating AAA.
Forecast Dispersion Defined as the standard deviation of analyst earnings forecasts
(one-year ahead EPS) issued by analysts scaled by the absolute
value of mean forecasts. We only include the most recent
forecast by each analyst within a year prior to the earnings
announcement.
Earnings Surprise Defined as actual earnings at the end of year t minus mean
forecasts (one-year ahead EPS) issued by analysts scaled by the
absolute value of mean forecasts. We only include the most
53
recent forecast by each analyst within a year prior to the
earnings announcement.
Panel D: Instrumental Variable and Channel Test Variables
State Growth Opportunity Defined as in Bakaert et al. (2005) and equal to ∑ 𝑤𝑆𝐼𝑡 ∗𝑃𝐸𝐼𝑡 where each state has a set of industries 𝑤𝑆𝐼𝑡 with each
industry having time-varying growth opportunities while
assuming that these growth prospects are reflected in the price
to earnings (P/E) ratios of the industry portfolio 𝑃𝐸𝐼𝑡.
Large Size A dummy variable that equals one if a firm’s total assets are
above the sample median of that year, and zero otherwise.
Stock Index Member A dummy variable that equals one if a firm is a major stock
index member, including Dow Jones, Nasdaq 100, or S&P 500
indexes, and zero otherwise.
High_Ownership_HHI An indicator variable that equals one if a firm’s Herfindahl-
Hirschman index of its institutional ownership concentration is
above the sample median level of that year, and zero otherwise.
Relationship Industry An indicator variable that equals one if a firm belongs to the
following industries with SIC codes of 15, 16, 17, 34, 35, 36,
37, 38, 39, 42, 47, 50, 51, 55, 60, 61, 62, 63, 64, 65, 67, 75, 76,
and 87 (see Cremers, Nair, and Peyer, 2008).
54
Appendix Table 2. Novel News and Debt Structure
This table presents the relationship between novel news and debt structure. The sample includes
all the U.S. firms that have either borrowed bank loans or issued bonds or both during the sample
period 2001-2014. The key explanatory variable Novel News is defined as a new story within a 24-
hour time window across all news stories in a particular package (Dow Jones, Web or PR Editions)
and with relevance score of 100 in RP. The dependent variables are Dummy_Bank_Loan (column
1), Bank Loans/Total Debt (column 2), Dummy_Bonds & Notes (column 3), and Bonds &
Notes/Total Debt (column 4). Detailed definitions of all the other variables can be found in
Appendix Table 1. Standard errors reported in parentheses are heteroskedastic robust and clustered
at the industry level. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels,
respectively.
(1) (2) (3) (4)
Dep Var
Dummy_Bank
_Loans
Bank Loans
/Total Debt
Dummy_Bonds
& Notes
Bonds & Notes
/Total Debt
Novel News -0.2731*** -0.0345*** 0.0616** 0.0220***
(0.0260) (0.0033) (0.0243) (0.0032)
Size -0.2461*** -0.0944*** 0.5635*** 0.0862***
(0.0190) (0.0025) (0.0186) (0.0023)
ROA 0.0122** 0.0076*** -0.0503*** -0.0043***
(0.0054) (0.0008) (0.0054) (0.0008)
Sales Growth 0.2326*** 0.0709*** -0.3242*** -0.0879***
(0.0687) (0.0099) (0.0712) (0.0094)
Leverage 0.2525*** 0.0006 0.4391*** 0.0458***
(0.0168) (0.0023) (0.0183) (0.0022)
Cash Flow -0.0678*** -0.0075*** -0.0048 0.0052***
(0.0032) (0.0005) (0.0032) (0.0004)
Market_to_Book -0.0501*** -0.0057*** -0.0407*** 0.0016
(0.0065) (0.0010) (0.0075) (0.0010)
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 35994 36002 35888 36002
R-sq 0.1181 0.2081 0.2516 0.3152
55
Appendix Table 3. Additional Robustness Tests
This table presents the results of media coverage on debt structure using an alternative sample (columns 1-4) or an alternative dependent
variable (columns 5-6) for sensitivity tests. The sample includes all the U.S. firms that have either borrowed bank loans or issued bonds
or both during the sample period 2001-2014, except in columns 1-4, the sample is restricted to firms with total book value of equity
between $100 million and $700 million U.S. dollars. The dependent variables are Bank Loans/Total Debt (columns 1-2) defined as the
ratio of a firm’s bank loans to total debt, Bonds & Notes/Total Debt (columns 3-4) defined as the ratio of a firm’s public/private bonds
and notes to total debt, and Term Loans/Total Debt (columns 5-6) defined as the ratio of a firm’s term loans to total debt. Firm controls
include ROA, Sales Growth, Leverage, Cash Flow, and Market_to_Book. Detailed definitions of all the other variables can be found in
Appendix Table 1. Standard errors reported in parentheses are heteroskedastic robust and clustered at the industry level. *, **, and ***
indicate statistical significance at the 10%, 5%, and 1% levels, respectively.
(1) (2) (3) (4) (5) (6)
Dep Var Bank Loans/Total Debt Bonds & Notes /Total Debt Term Loans/Total Debt
Total_News -0.0292** 0.0326*** -0.0203***
(0.0126) (0.0116) (0.0041)
Earnings_News -0.0054** 0.0048** -0.0064***
(0.0027) (0.0023) (0.0013)
Size -0.1204*** -0.0964*** 0.1436*** 0.1171*** -0.0523*** -0.0380***
(0.0108) (0.0099) (0.0199) (0.0151) (0.0064) (0.0059)
Firm controls yes yes yes yes yes yes
Restricted sample yes yes yes yes no no
Industry fixed effects yes yes yes yes yes yes
Year fixed effects yes yes yes yes yes yes
N 13733 13733 13733 13733 36002 36002
R-sq 0.1400 0.1390 0.2812 0.2797 0.2054 0.2055
56
Appendix Table 4. Media Coverage and Borrowers’ Ratings
This table presents the relation between media coverage and borrowers’ credit ratings. The sample
includes all the U.S. firms that have either borrowed bank loans or issued bonds or both during the
sample period 2001-2014. The dependent variables are Dummy_Rating (columns 1-2), which is
equal to 1 if a firm has an S&P credit rating, and 0 otherwise, and S&P_Rating (columns 3-4),
which is defined as Ln (1+ credit rating) where the credit rating is coded from 0 to 22 with 0
representing no rating, 1 representing D or below, and 22 representing the highest S&P rating
AAA. Detailed definitions of all the other variables can be found in Appendix Table 1. Standard
errors reported in parentheses are heteroskedastic robust and clustered at the industry level. *, **,
and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.
(1) (2) (3) (4)
Dep Var Dummy_Rating S&P_Rating
Total_News 0.3633*** 0.0970***
(0.0371) (0.0186)
Earnings_News 0.0828*** 0.0081*
(0.0119) (0.0044)
Size 1.8681*** 1.6083*** 0.4806*** 0.4168***
(0.0321) (0.0193) (0.0134) (0.0104)
ROA 0.0251*** 0.0243*** -0.0044 -0.0048
(0.0076) (0.0075) (0.0051) (0.0051)
Sales Growth -0.9493*** -0.8781*** -0.3142*** -0.3038***
(0.1023) (0.1028) (0.0435) (0.0431)
Leverage 0.6708*** 0.6576*** 0.1138*** 0.1091***
(0.0240) (0.0240) (0.0161) (0.0156)
Cash Flow 0.0038 0.0084* -0.0132*** -0.0124***
(0.0044) (0.0044) (0.0041) (0.0040)
Market_to_Book -0.0336*** -0.0162 0.0191*** 0.0248***
(0.0106) (0.0104) (0.0060) (0.0065)
Industry fixed effects yes yes yes yes
Year fixed effects yes yes yes yes
N 34984 34984 35018 35018
R-sq 0.5546 0.5530 0.5471 0.5452