1 Do relationships matter in the corporate bond market? Abstract We examine the impact of information networks on the pricing of corporate bonds. Using the location of issuers, their bookrunners and institutional investors as a proxy for the quality of such networks, we find that issuers based in central, urban areas have significantly lower at-issue yield spreads compared to their remote, rural counterparts. Prior underwriting relationships as well as proximity between economic agents in the network result in a spread reduction, especially for remote and risky issuers. Our findings provide evidence of a new channel through which local information networks can impact firm value, namely the pricing of corporate bonds.
58
Embed
Do relationships matter in the corporate bond market? Abstractfmaconferences.org/.../Do_Relationships_Matter_in... · Do relationships matter in the corporate bond market? Abstract
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Do relationships matter in the corporate bond market?
Abstract
We examine the impact of information networks on the pricing of corporate bonds. Using the
location of issuers, their bookrunners and institutional investors as a proxy for the quality of such
networks, we find that issuers based in central, urban areas have significantly lower at-issue yield
spreads compared to their remote, rural counterparts. Prior underwriting relationships as well as
proximity between economic agents in the network result in a spread reduction, especially for
remote and risky issuers. Our findings provide evidence of a new channel through which local
information networks can impact firm value, namely the pricing of corporate bonds.
2
"We're not looking for just one transaction, we're looking for a relationship that is going to span many
years”. Chris Wood, head of high yield capital markets for SunTrust Robinson Humphrey, Bond
Bookrunning Group.
I. Introduction
The relationship between corporations, security underwriters and investors is a cornerstone of
financial markets. For instance, Burch et al. (2005), Binay et al. (2007), and Huang et al. (2008),
among others show that relationships such as this create unique networks of issuers and
institutional investors who tend to stay loyal to their main investment bank over the long term.
This suggests that, much like a social network, both issuers and investors could self-segment
through an affiliation with particular investment banks. Despite the importance of issuer-
underwriter-investor relationships, empirical research provides little evidence on the role of such
information networks in the pricing of securities. In this paper, we exploit the predominance of
institutional investors in the U.S. corporate new bond issues market to examine whether such
information networks have a direct effect on the pricing of securities in general, and bonds in
particular.1
Compared to equities, the strong institutional nature as well as lower liquidity and long term
investment horizon of bond investors, make corporate bonds an ideal laboratory to study the effect
of information networks on asset prices. The book-building process and general marketing efforts
during corporate bond issues create a segmented information flow directed at targeted groups of
long term investors. The reliance on such relationship specific knowledge becomes particularly
important for bondholders in the absence of formal road shows or frequent conversations with the
1 The corporate bond market is almost exclusively driven by institutional investors. For example, Bessembinder, Kahle, Maxwell and Xu (2009)
report that trades of $100,000 or greater, account for 96.7% of bond trading volume. The presence of institutions is even more pronounced in the
primary market, where individual investors typically do not have an opportunity to purchase a new issue until after the initial announcement.
3
issuing firm’s management during the book-building process.2 Underwriters also offer their clients
information intensive activities such as market making, analyst coverage, and advice on mergers
and acquisitions that often lead to repeat transactions between bond issuers and their lead
underwriter (Yasuda, 2005). Over time, these information sharing activities create suitable
conditions for both issuers and investors to form a relationship with their main investment bank.
To examine whether such relationships matter in the pricing of corporate bonds, we rely upon
the existing literature to construct measures of the strength of information networks between
issuers, underwriters and institutional investors. Because people are most likely to network and
share information with those that live or work nearby (see, e.g., Bayer et al., 2008 and Hong et al.,
2008), we consider the location of an issuing firm’s headquarters vis-à-vis its lead underwriter and
institutional investors as the main facet of the quality of information networks.
Specifically, we ask whether an issuer’s geographic location through the formation of local
information networks affects the pricing of debt. For example, decreased observability of remotely
headquartered companies and their distance from central locations that are characterized by large
concentrations of investment banks and institutional investors could create lower potential for
information sharing and networking opportunities through repeated interactions. As a result,
underwriters may find it more difficult to assess an issuer’s creditworthiness, and market and sell
its securities. In contrast, centrally located issuers are more likely to be closer to their lead
2 In their recent presentation to the SEC, the Credit Roundtable, a group of large fixed income institutional investors, describes the current practices
of corporate bond underwriting and distribution: “when new corporate bonds are issued, institutional investors’ ability to conduct proper diligence
is significantly diminished without conference calls, road show slides or presentations… From the time underwriters announce the new issue,
investors have as little as 15 minutes to evaluate prospectus terms (often which are not available), the issuer’s credit history, and pricing
expectations… For the large majority of new issues, there is no conference call with the management of the issuer… Covenant description may be
imprecise or confusing, with risk factors that are often generic and lack company specific details”. (The Credit Roundtable “Current Practices of
the Corporate Bond Underwriting & Distribution Process and Recommendations for Improvement”, presentation to the SEC, March 3 2009).
4
underwriter, who has better knowledge of and more repeat relationships with local institutional
investors that can be targeted during the book-building process. These local investment banks may
also have more intimate issuer-specific knowledge than remote investment banks through
familiarity with the local economy and personal relationships with the issuer’s management.
Indeed, Figure 1 shows a clear concentration of bond issuers, institutional investors and
bookrunners around large, central metropolitans. In this paper, we ask whether such clustering of
economic agents can benefit bond issuers through the formation of local information networks.
To determine the effect of location, our main proxy for the quality of information networks,
on the cross section of at-issue bond yield spreads, we start with a sample of US corporate bonds,
issued during 1998-2008. We follow Ivkovic and Weisbenner (2005) and Loughran and Schultz
(2005), among others, and aggregate issuers’ headquarters by metropolitan statistical areas
(MSAs), classifying firms as either Urban or Rural, based on the size and centrality of the
metropolitan where the firm is headquartered and its distance from major population and economic
activity clusters.
We find that firms headquartered in remote, sparsely populated areas (Rural) exhibit
significantly higher at-issue bond yield spreads than firms headquartered in central (Urban) areas,
especially for small and longer maturity issues, for which the spread differential can be as much
as 33 basis points. This is consistent with the notion that firms that are based in large, central areas
that are home to more investor networks offer issuers a comparative pricing advantage compared
to their remote, less observable counterparts.
We also find that the proximity of issuers to their lead underwriter (measured as the distance
between the headquarters of the issuer and that of their bookrunner) is highly beneficial to Rural
firms, firms issuing smaller size and longer maturity debt, and firms that issue less frequently.
Similarly, we find that higher levels of local institutional ownership with proximity to the issuer
5
and/or the lead underwriter, result in a significant reduction in at-issue yield spreads for Rural and
riskier issuers. These findings are consistent with the benefits of localized information networks
that give both investors and underwriters a comparative advantage in monitoring and assessing
soft information about local issuers (Butler, 2008).
Next we examine the impact of recent prior relationships between an issuer, its bookrunner
and institutional investors on the pricing of newly issued bonds. We find that the existence of a
recent prior underwriting relationship is highly beneficial for remote Rural issuers, as well as
issuers of riskier debt, resulting in a significant at-issue spread reduction compared to similar non-
relationship issues.
The existence of a statistically significant spread differential between bonds issued by Rural
and Urban companies is robust to a host of endogeneity and sensitivity tests, including an
instrumental variable, alternative measures of geographic location, sample selection criteria, firm
and industry controls, firm specific information and governance proxies. One of our most
convincing tests is based on a natural experiment using brokerage closures and mergers that
directly forced some firms to lose analyst coverage. Using a difference-in differences approach,
we show that while an exogenous loss of coverage results in an increase in the cost of debt for both
Rural and Urban issuers compared to their respective control groups that did not suffer coverage
loss, the spread increase is significantly more pronounced for Rural companies, indicating that
they are more adversely affected by this exogenous, forced loss of information intermediaries.
Finally, we examine whether our results hold for traded bonds. In complete markets, localized
networks of information should not affect bond pricing and thus, mispricing due to location should
be arbitraged away by sophisticated institutional bondholders in the secondary bond market.
However, if the distance between issuers and investors results in information friction and market
segmentation through the creation of localized information networks, we should continue to
6
document cross sectional variation in bond prices even for traded bonds. Using data on traded
bonds and controlling for firm, bond and macroeconomic characteristics, our results continue to
hold.
Our paper adds to several streams of research in the finance literature. First, we show that
networks play a role in the determination of asset prices. Our evidence is consistent with the work
of Asker and Ljungqvist (2010), Fernando et al. (2012), Grullon et al. (2012) and others, which
shows that investment banking relationships create issuer and/or investor networks. Our paper
provides evidence of a new channel of local networks through which the various investment bank
relationships can add value – the pricing of corporate bonds.
Second, our results are related to the literature on proximity investment (e.g., Coval and
Moskowitz, 1999, 2001). This literature has mostly focused on the equity side, showing that
investors tend to hold the stocks of firms located nearby because of information advantages,
familiarity, and/or local social interactions.3 We focus on the debt side and show that it has equally
important implications on firms’ cost of capital. This is particularly important given the size of the
corporate bond market, and the fact that over the last decade bond financing has become one of, if
not, the most important sources of external financing for US corporations.4
Finally, our paper is related to an extensive body of research about the benefits associated with
the agglomeration of economic agents across geographic or industry clusters that come from
3 Coval and Moskowitz (2001), Ivkovic and Weisbenner (2005), Ivkovic et al. (2008), and Colla and Mele (2010) conclude that local investors
have an informational advantage. Huberman (2001) shows that people tend to invest in the familiar. Grinblatt and Keloharju (2001), Hong et al.
(2004, 2005), and Brown et al. (2008) find that social interaction among investors is important for investment decisions. Recent research also
studies the implication of local bias for equity returns or prices. For example, Pirinsky and Wang (2006) show that the returns of firms located
within the same geographical vicinity co-move more strongly; Feng and Seasholes (2004) show correlated trading when investors are closer to the
headquarters; and Garcia and Norli (2012) find that locality is associated with higher stock returns.
4 As of March 2013 there was $9.2 trillion of US corporate debt outstanding (source: http://www.sifma.org/research/statistics.aspx).
learning, intellectual spillover, and access to informal “soft” information through interactions with
local professionals (see, e.g., Almazan et al., 2007; Christoffersen and Sarkissian, 2009; Ellison et
al., 2010; and Almazan et al., 2010). Our study adds to this literature by showing that local
economic clusters also play a role in the pricing of corporate bonds.
Overall, our findings suggest that the pricing of corporate bonds is strongly influenced not only
by firm and issue characteristics, but also by the business environment in the location where the
firm is headquartered. As such, they challenge the classical view of corporate finance that
bondholders have arm’s length transactions with firms that are only based on publically available
information (e.g., Myers, 1977; Myers and Majluf, 1984; and Dittmar and Thakor, 2007).
The remainder of the paper is organized as follows. In section II we describe the data, variables
and methodology. Section III presents the results. Section IV provides robustness tests and Section
V concludes.
II. Data
A. Sample and Methodology
Our data are from multiple sources and cover the 1998-2008 time-period. We construct the
information networks proxies using location and prior relationship measures. Our main location
measure is based on whether the issuer is headquartered in a large and central area (Urban) or in
a remote, sparsely populated one (Rural). To classify firms as Urban or Rural, we follow Coval
and Moskowitz, (1999) and Loughran and Schultz (2005), and others, and use a company’s
headquarters as a proxy for its location. Headquarters locations are from Compustat, SDC and
Hoover’s and for relocations we use various SEC filings, where applicable. We find the latitude
and longitude for each firm’s headquarters using the US Census Bureau’s Gazetteer city-state files,
8
and calculate the distance between the headquarters and the largest US metropolitan areas of at
least 1 million people as defined by the 2010 Census.
Following Loughran and Schultz (2005) and John, Knyazeva and Knyazeva (2011), among
others, a company is classified as an Urban firm if its headquarters MSA population size is 1
million or above. A company is classified as Rural if its headquarters is at least 250 kilometers
away from the nearest Urban firm in MSAs with population size under 1 million.
The main bond data source is the Securities Data Corporation (SDC) New Issues database. To
calculate bond yield spreads, we use the risk-free term structure of interest rates taken from
Bloomberg, including the monthly treasury benchmark yields for 2, 3, 5, 7, 10, and 30 year coupon
bonds. Following conventional sample selection criteria, we exclude firms headquartered and
incorporated outside the US and firms with asset size below $20 million. For debt issues to be
included in our analysis, data on the firm’s headquarters location, leverage, assets, proceeds and
par amount, at-issue yield, maturity date, and lead underwriter of the firm’s fixed coupon rate for
straight public debt securities must be available. After imposing the above selection criteria, our
final sample is comprised of 4,328 new issues made by 904 firms (794 Urban and 110 Rural), with
Rural and Urban companies making 409 and 3,919 issues, respectively. To calculate the distance
between an issuer’s headquarters and the location of its lead underwriter we obtain the name of
the lead underwriter(s) from SDC and hand collect information on its principal location of
business.
We obtain institutional bond investors data from Lipper’s eMAXX fixed income database
which provides information on their location and quarterly bond ownership. We only include firms
with complete information on par amount, coupon rate, issue and maturity date, address and US
location during our sample period, resulting in a sample size of 2,424 bond holdings by
institutional investors. We then calculate the distance between the location of each issuer and that
9
of its institutional investors, and the distance between the issuer’s bookrunner and each of its
institutional investors. Data on analyst following are from I/B/E/S, outside blockholders data is
from CDA Spectrum, and covenant provisions are from Fixed Income Securities (FISD). We also
obtain the Gompers et al. (2003) governance index (GIM Index) from Metrick’s website.
Secondary market bond trade data are from several sources. Transaction information such as
trade date, price, and underlying yield corresponding to all bond transactions between January 1,
2005 and December 31, 2008 are from the TRACE database. Share prices and accounting data, as
well as issuer headquarters location information are from Bloomberg and the historical location
files on Compustat. Treasury yields adjusted to constant maturities are from the Federal Reserve,
and LIBOR rates are from the British Bankers’ Association. These data are then merged to yield a
sample consisting of 2,215 domestic bonds issued by 754 firms that meet the criteria outlined
above. Finally, we supplement our traded bond dataset with data on 5-year CDS from Bloomberg
and obtain data on 298 firms with CDS data for 1,284 bonds for the January 2005- December 2008
time-period.
B. Description of Variables
The dependent variable is the at-issue yield spread (Spread), defined as the difference between
the yield to maturity on a coupon paying corporate bond and the yield to maturity on a coupon
paying government bond with the same maturity date. We use bond-specific, firm-specific and
demographic control variables in our analysis. Bond related measures include: credit rating
(Rating), issuance proceeds (Proceeds), time to maturity in years (Time to maturity), a dummy
variable to denote issues with restrictive covenants (Covenants), bond seniority (Senior) and
callable (Call) dummy variables, a dummy variable to denote high yield, non-investment grade
bonds with ratings below Baa3 (High yield issues), and a dummy variable to denote non-rated
bond issues (Non rated).
10
A firm’s credit rating is measured by Moody’s bond ratings at the issue date. Similar to Klock
et al. (2005) and others, we compute bond ratings using a linear conversion process in which Aaa
or AAA ratings are assigned a value of 1 and B3 or B- ratings are assigned a value of 16. For
example, a firm with an Aa1 Moody’s rating would receive a score of 2; a firm with Aa2 rating
would receive a score of 3; and so on. Thus higher numerical bond ratings denote higher credit
risk. All bond issuance related data are obtained from SDC.
Our firm-specific variables are primarily from Compustat and include headquarters location,
issue size (Size) measured as the natural log of the issue proceeds; firm leverage (Leverage),
calculated as the ratio of long-term debt to total assets; firm age (Age) defined as the difference in
years between the current year and the year of the company’s incorporation; and firm profitability
(ROA and market-to-book ratio). We calculate ROA as the ratio of earnings before interest, tax,
depreciation and amortization, divided by total assets. For the market-to-book ratio, we use the
end of the previous year’s CRSP market value of equity scaled by the prior fiscal year’s book value
(defined as book value of equity plus balance-sheet deferred taxes and investment credit minus the
book value of preferred stock).
To control for issuance frequency, we create a dummy variable for one time bond issuers
during the sample period (One time issuer). Following Fang (2005), we create a dummy variable
to denote prestigious lead bond underwriters in our sample (Prestigious underwriters). We also
control for the listing location of issuing firms by creating a dummy variable that is equal to 1 if
the firm is listed on Nasdaq, and 0 otherwise (Nasdaq).
To address the issue of information asymmetry we use the number of analysts following the
firm (Number of Analysts). Malloy (2005) and Bae et al. (2006), among others, show that the
information production role of analysts increases with their proximity to target firms. Analysts
coverage is defined as the number of analysts reporting current fiscal year annual earnings
11
estimates prior to the end of fiscal year t. Firms without I/B/E/S coverage are assigned a value of
zero analysts.
Better governed firms are expected to be less affected by distance from large cities and distance
from investors and underwriters who all serve a monitoring role. As such, we control for
institutional and governance related variables, namely, the percentage of outside blockholdings (%
outside blockholders) and the level of antitakeover defenses (GIM index). The presence of
institutional blockholders who own at least 5% of the firm’s stock is associated with the monitoring
of management (Shleifer and Vishny, 1997). To construct this variable, we match the percentage
of shares held by outside blockholders with our sample using data from Dlugosz et al. (2006).5 If
local institutions hold shares of Rural firms, these firms should face fewer governance problems
that would mitigate informational differences between Rural and Urban companies thereby,
reducing the likelihood of finding a significant location effect.
To account for factors associated with the issuer’s familiarity and visibility in the area, we
control for firm idiosyncratic risk. Hou and Moskowitz (2004) find that the idiosyncratic risk of
stocks is priced for small, less visible stocks, and Campbell and Taksler (2003) document that
idiosyncratic risk is positively associated with bond yield spreads. As such, firms that are more
visible are expected to have a lower cost of debt. Because firms headquartered in large
metropolitans are more visible and familiar to a larger set of investors than firms from rural areas,
by controlling for idiosyncratic risk we address the concern that our results could be driven by a
correlation between city centrality and firm familiarity. We follow Campbell and Taksler (2003)
and measure idiosyncratic risk as the standard deviation of the firm’s excess return over the market
5 Available at: http://www.som.yale.edu/faculty/am859/data.html
12
portfolio. We use the prior 180 trading days to calculate idiosyncratic risk which we multiply by
ten for scaling purposes.
In our analysis, our primary measure of location are firms that are headquartered in remote
rural areas (Rural) and firms headquartered in central, large cities (Urban), with the latter being
the control group. To account for bookrunner−issuer relationships, we measure the distance
between each issuer and its lead bookrunner, with a bookrunner defined as local if it is based less
than 250 kilometers away from the issuer’s headquarters, and non-local otherwise. Similarly, to
capture issuer- investor and investors- bookrunner local relationships, we calculate the distance
between the location of each issuer and that of its institutional investors, and the distance between
the issuer’s bookrunner and each of its institutional investors. Thus, for each bond issue, we define
local institutions as those based less than 250 kilometers away from the issuer or the issuer’s
bookrunner. Local (non-local) institutional ownership is then computed as the book value of the
issue held by local (non-local) institutional investors in the quarter of the issue divided by the total
book value of the bond issued. We define an indicator variable, High (Low) local institutional
ownership to denote whether the fraction of local institutional ownership is above (below) the
sample median for local institutional ownership in that year.
We also account for prior, recent relationship between an issuer and its bookrunner, and
between institutional investors and the bookrunner. We define Repeat relationship with same
bookrunner as a prior offering by the issuer that shares a lead underwriter with the current
offering.6 Because more recent offerings compared to distant past ones contribute more to an
underwriter’s knowledge of the issuer, we impose a five-year cutoff period between the current
6 If the previous offering had 2 lead underwriters then the current offering is a relationship offering if either of the previous leads is used for the
current offering. Similarly, if the current offering has 2 lead managers then the offering is a relationship offering if either of the lead managers was
used in a previous offering.
13
and past relationship offerings. Similarly, we calculate the fraction of holdings by institutions that
either purchased from the same issuer or the same bookrunner up to five years prior to the issue
and define High (Low) repeat institutional relationship to indicate whether the fraction of holdings
by institutional investors who have previously purchased bonds from the same issuer or
underwriter is above (below) the median for that year.
Finally, to conduct our secondary market robustness tests we use a set of variables that is based
on Elton et al. (2001), Campbell and Taksler (2003), and Chen et al. (2007), and others. These
include, accounting variables, bond specific and illiquidity measures, and several macroeconomic
variables. These variables are defined in Appendix I.
C. Summary Statistics
Table 1 presents the distribution of our sample across the two location groups, the MSA and
number of issues per MSA sorted by population size. Overall, there are 129 distinct MSAs (51
Urban and 78 Rural) in our sample ranging in population from 28,149 in Yazoo City, Mississippi
to about 18.9 million in New York City.
[Insert Table 1 about here]
A closer analysis of the data suggests that firms’ headquarters are clustered in a small number
of metropolitan areas. New York stands out as the dominant center, with about 10% of the bond
issues and 15% of the headquarters in our sample. Table 1, along with Figure 1, which displays
the distribution of public bond issuers, their bookrunners and institutional investors across the
continental US, show that most bond issuers are headquartered in or near the largest US cities.
Figure 1 also shows that bookrunners and institutional bond investors tend to cluster around the
same big cities. Rural areas are characterized by a significantly lower concentration of issuers,
investors and bookrunners and are also the most sparsely populated.
14
[Insert Figure 1 about here]
[Insert Table 2 about here]
Table 2 presents the number of debt issues and the average at-issue yield spreads by year and
location groups (Urban or Rural) over 1998-2008. It shows that bond issues have increased from
185 in 1998 to 375 in 2008 and that firms headquartered in Urban areas issue the highest volume
of bonds in each year, totaling over 90% of the overall sample size. This high issuance volume by
firms headquartered in Urban metropolitans is likely due to the fact that more than 40% of all
publicly traded companies are located in the top 20 most highly populated US counties. For each
year, Rural firms report the highest average at-issue yield spread with an overall average of
182.982 basis points. In comparison, Urban companies have an average at-issue yield spread of
140.267 basis points. This 43 points difference between mean at-issue yield spreads is significantly
different from zero at the 1% level.
To get additional insights on the effect of location on the cost of corporate debt, we examine
the characteristics of Rural and Urban firms in more detail. Table 3 provides summary statistics
for the variables used throughout the analysis. Panel A, which compares firm and institutional
characteristics across Urban and Rural firms, shows that Rural firms are significantly smaller than
firms headquartered elsewhere. Consistent with Loughran and Schultz (2005), we find that Rural
firms are also significantly more levered than Urban companies as shown by the leverage ratios
of 26.926% and 20.357%. Interestingly, we find that they issue debt somewhat less frequently
throughout the sample period (with 84.121% multiple Rural issuers compared to 92.482% multiple
Urban issuers). Note, however, that a firm’s outstanding debt also includes bank loans and other
types of debt which could help to explain the higher leverage of Rural firms. Later in Table 10, we
indeed show that Rural companies are more likely to rely on bank loans than firms headquartered
elsewhere.
15
[Insert Table 3 about here]
Consistent with Malloy (2005), Loughran and Schultz (2005) and others, we find that Rural
companies are more likely to be listed on Nasdaq, have higher idiosyncratic risk and less analyst
coverage compared to their Urban counterparts. Similar to Loughran and Schultz (2005), we find
that small, retail equity investors own a higher portion of the total shares outstanding in Rural
companies than Urban firms, whereas institutional investors tend to own a higher percentage of
shares in large city companies. This is consistent with the argument that institutional investors,
who are mainly concentrated in large metropolitans, are biased toward nearby, large city firms,
whereas the equity shares of remote rural companies are primarily held by local retail investors.
In contrast to to Rural issuers, Urban companies have a significantly higher proportion of their
bonds issued by local bookrunners and purchased by institutional investors who are based less than
250 kilometers away from their headquarters or the bookrunner’s office. To the extent that
geographic proximity reduces information asymmetry and facilitates the creation of social
networks, these local transactions could result in a comparative advantage for issuers based in
larger cities. Because both institutional investors and investment banks tend to be concentrated
around metropolitan areas, these findings are also consistent with Massa et al. (2009) that
institutional bondholders have a geographic bias in the ownership of local bonds which provides
for soft information advantages in the underwriting practices of investment banks (Butler, 2008).
In sum, the above findings suggest that rural firms are associated with lower visibility and
familiarity, greater information asymmetry and a lower concentration of both local bookrunners
and institutional ownership compared to their large-city counterparts. To the extent that the
distance from major concentrations of the investment community affect networking opportunities
and information sharing and are priced by bondholders, these factors would further exacerbate the
effect of an isolated rural location on the cost of debt.
16
Panel B of Table 3 displays the relationship between headquarters location and issue related
characteristics. The average at-issue yield spread for Rural firms is 182.982 basis points compared
to Urban firms, with 140.267 basis points, with a statistically significant spread differential at the
1% level. Consistent with a smaller average asset size, the average issue size of Rural firms is
significantly smaller than that of urban firms. Interestingly, Rural firms tend to have lower average
debt rating than firms based in larger, major metropolitans. This could explain why, on average,
issues of Rural companies tend to have more protective covenants compared to issues made by
other firms.
Panel A of Table 4 presents the identity and characteristics of the 20 metropolitans that host
the headquarters of the most costly debt issues, whereas Panel B presents the characteristics of the
20 metropolitans with the lowest at-issue yield spreads. Each MSA in Table 4 is a subset of at least
one local issuer. Consistent with our conjecture that remote rural location results in a higher cost
of debt, we find that the metropolitans in Panel A are associated with Rural areas, whereas the
metropolitans in Panel B represent firms headquartered in larger cities (Urban). Consistent with
the findings in Table 1, issuers are clustered in a small number of large metropolitan areas, with
the ten most highly populated metropolitans being the dominant centers.
[Insert Table 4 about here]
Industry composition also differs between Rural and Urban firms. Table 4 reports the number
of local issuers for each MSA by industry in the manufacturing (SIC 20-39), banking and finance
(SIC 60-67), transportation and utilities (SIC 40-49), mining (SIC 10-14) and business services
(SIC 73, 81, 87 and 89) industries. Taken together, these seven industries account for 83% of Rural
and 75% of Urban companies. Mining, transportation and utility companies account for a greater
proportion of Rural than Urban firms, and manufacturing, business services and banking and
finance account for a greater proportion of Urban issuers.
17
Panel B of Table 4 shows that Urban firms tend to cluster in the same sector of activity and
within proximity to other agglomerates of financial and business service firms, whereas as shown
in Panel A, Rural companies are more sparsely headquartered. These findings are consistent with
those in the extant urban economics literature which shows that urban firms’ headquarters tend to
cluster in areas with business services, other local headquarters and hiring pools. In contrast, rural
companies tend to locate in areas with a lower degree of regional product market competition (see,
e.g., Davis and Henderson, 2008; Fujita and Thisse, 2002).
Table 4 also sorts the sample by whether a firm’s headquarters is collocated in the same MSA
as its main production plant, research or service facility and therefore, derives benefits of intra-
firm monitoring such as lower communication costs with personnel. Panel A shows that while, in
general, Rural firms headquarters are collocated with their main plant, research or service facility,
Panel B shows that most Urban headquarters are not collocated with their major operational units.
These findings are consistent with the empirical literature on headquarters location (see, e.g.,
Duranton et al., 2004) which shows that for many regional companies, the headquarters
establishment is often collocated near an important production or service operation.
Bell et al. (2005) contend that firms are also attracted to locations with lower wage rates and
taxes and a higher quality of living. Table 4 shows that most Rural firms benefit from lower mean
wages and corporate tax rates compared to urban firms. Rural locations are also associated with
lower average commute times to work and lower median housing prices compared to Urban
metropolitans. Taken together, these findings, along with greater land availability and lower air
pollution suggest that Rural areas are characterized by lower congestion costs and cost of living.
Overall, Table 4 reflects some of the costs and benefits associated with headquarters location.
When faced with a tradeoff between labor costs, taxes, congestion costs, cost of living, intra-firm
communication and monitoring costs, supply of local services, local product market competition,
18
and communication with other local businesses, firms differ as to the preferred mix and seem to
balance their visibility and distance from potential investors in the overall location decision.
III. Empirical Results
So far, we have shown that remote and less visible areas are characterized by higher financing
costs. However, these costs could also differ across firm locations if there are systematic
differences in firm and bond characteristics. To address this issue, in this section we use regression
analysis to examine the extent to which differences in at-issue yield spreads of corporate bonds
can be attributable to a firm’s headquarters location. To minimize the effect of outliers, we
winsorized all accounting variables at the 1% level.
A. Information Networks and the Cost of Debt
We begin the analysis by examining the relationship between our headquarters location
proxy for the quality of information networks, denoted by whether an issuer is headquartered in
an Urban or Rural area, and at-issue yield spreads of corporate bonds, controlling for firm and
bond-specific measures.7 The general specification of the regressions is as follows: