Upload
dinhdang
View
216
Download
2
Embed Size (px)
Citation preview
Distance Still Matters: Evidence from Municipal Bond Underwriting
Alexander W. Butler* Rice University
University of South Florida
Comments welcome.
May 21, 2003
* Department of Finance, College of Business Administration, BSN3403, University of South Florida, Tampa, FL 33620-5500. Email: [email protected]. Acknowledgements: I thank Utpal Bhattacharya, Gustavo Grullon, George Oldfield, David T. Robinson, Kim Rodgers, Chip Ryan, and James Weston for helpful suggestions and Kristy Freeman and Lindsay Altazin for superb research assistance. Seminar participants at Rice University, Southwest Texas State University, Baylor University, University of Kentucky, University of New Orleans, and University of Texas at San Antonio made valuable comments. I am especially grateful for substantial input from Lee Ann Butler and others at the Louisiana Department of the Treasury. Any errors are my own.
Distance Still Matters: Evidence from Municipal Bond Underwriting
Abstract
We provide evidence of “soft” information production in investment
banking. Using 2191 municipal bond offerings from 1997-2001, we find
that “local” investment banks have substantial comparative and absolute
advantages over non-local counterparts – locals charge lower fees and sell
bonds at lower yields. Local investment banks’ strongest comparative
advantage is at underwriting bonds with higher credit risk and those bonds
not rated by rating agencies. Our interpretation is that high-risk bonds and
non-rated bonds are more difficult to evaluate and market, and investment
banks with a local presence are better able to assess “soft” information and
place difficult bond issues.
JEL Codes: G24, G28, D80 Keywords: Municipal bonds, investment bank fees, local underwriters, soft information
1
1. Introduction
How do financial intermediaries gather and process information? This question is at the
heart of the academic literature on why commercial banks are “special” (see Bhattacharya and
Thakor [1993] for a review). As technology has advanced, commercial banks have come to rely
more upon “hard” information—information that can be put on paper or stored electronically,
and hence transferred to others—rather than “soft” information—information that is more subtle
and arises through familiarity and relationship lending (Petersen and Rajan [2002] and Berger, et
al. [2002]). As a result, lending arrangements between commercial banks and borrowers have
evolved from strict ex ante screening and costly ex post monitoring, to more frequent ex post
monitoring and quick intervention (Petersen and Rajan [2002]). Indeed, the physical distance
between commercial banks and their borrowers increased steadily and dramatically from 1973 to
1993 and commercial banks and borrowers are communicating “less and less in person.”
Like commercial banks, investment banks are also in the business of gathering and
processing information. However, a critical difference between investment banks and
commercial banks is that investment banks generally do not retain a direct financial exposure to
the issuers they serve. Thus, due to the nature of underwriting services, investment banks have
little or no reason or opportunity for ex post monitoring because they do not maintain a position
in the securities they sell, and so must rely almost solely upon ex ante screening and evaluation.
Investors in new issues rely upon investment bank underwriters to evaluate and synthesize
information in both its “soft” and “hard” forms and convey this information to the market. This
paper examines the ability of investment banks to evaluate and exploit information in the
municipal bond underwriting arena.
2
Municipal bond underwriting is a part of the investment banking business where personal
relationships can be as important as ability.1 Because the location of a municipal issuer is well-
defined, whereas a corporate issuer might have numerous and geographically disperse offices or
branches, municipal issuers provide a useful laboratory to study whether location of a financial
intermediary matters.
Are local investment banks, i.e., those with an on-going presence in a particular state,
“better” underwriters for local bond issues? Local underwriters may have better knowledge of
soft information about the municipalities offering the bonds. Daily exposure to local news
stories, first-hand knowledge of the local economy, and personal relationships with key people at
the issuing body would all give local underwriters an advantage at evaluating soft information
over underwriters without local connections. Similarly, local underwriters may have better
knowledge of local investors who might be interested in buying the bonds. Personal knowledge
of and relationships with local money managers, bankers, institutional investors could all assist
local underwriters in placement of the securities. These comparative advantages over
underwriters without a local presence would allow the local investment banks to be more
competitive, and charge lower fees and get better prices (and hence better interest rates) on the
bonds they sell, ceteris paribus.
Political connections can also be very valuable (see, e.g., Fisman [2001]), and local
underwriters could have political connections in the state that better enable them to win
underwriting bids. If so, local investment banks would try to capture economic rents, thereby
charging higher fees, ceteris paribus. Exploiting personal connections with governmental
officials or existing relationships with issuing bodies could improve a local investment bank’s
probability of winning an offering and/or their ability to charge higher fees. This might be a 1 Financial relationships have historically been very important as well. See Section 2.C below.
3
result of blatant corruption, or could be as innocuous as states giving preference to underwriters
with an “on-going commitment to the state.”
Which effect is more important? That is, are local underwriters more capable and adept
at assembling soft information? Or can local underwriters exploit their connections and
geography to the detriment of municipal bond issuers? The answer could vary from state to
state, and whether local underwriters are “better” is, ultimately, an empirical question. We
address this question using a sample of 2191 municipal bond offerings from 41 states and the
District of Columbia during the period 1997-2001. We find that local underwriters charge lower
fees on average than their “non-local” counterparts and place municipal bonds at lower yields
than their non-local counterparts. There is substantial variation state-by-state, though. For
instance, local underwriters in Louisiana charge around 140 basis points less in gross spreads on
average than non-local underwriters. In contrast, local underwriters in West Virginia charge
around 160 basis points more on average. Similarly, local underwriters in Louisiana and Ohio
sell bonds at yields about 200 basis points lower than non-locals, and in Washington and
Colorado locals sell bonds at yields about 100 basis points higher than non-locals.
The benefits of having a local underwriter are evident even after controlling for other
factors. Both gross spreads and yields are statistically significantly lower for bonds underwritten
by locals than non-locals. Interestingly, these benefits are most acute for very opaque issuers.
The data indicate that local investment banks have a strong comparative advantage at
underwriting munis with higher credit risk. Higher credit risk bonds command higher
investment bank fees—on average, about 5.7 to 6.8 basis points per incremental “notch” in credit
rating. Local investment banks, though, charge about 3.5 to 4.9 basis points less in gross spreads
per credit rating notch than non-local underwriters after controlling for other factors, which
4
represents about two thirds of the effect that ratings have on gross spreads. In similar fashion,
bonds that are not rated command higher fees as well (about 73 basis points, over half the
average gross spread in our sample). It is not surprising that gross spreads are higher for these
bonds—they are likely more difficult to place because they lack the external certification that a
bond rating agency provides. Local investment banks are able to place these “difficult” bonds at
significantly lower fees than their non-local counterparts. We interpret these results as
suggesting that relatively high-risk offerings are more difficult to evaluate and market.
Investment banks with a local presence are better able to evaluate and/or market “difficult” bond
issues than underwriters without a local presence.
This “local” effect has an even larger effect on the yields on the bonds. Local
underwriters have a strong comparative advantage at placing difficult high credit risk and non-
rated bonds. Though yields are related to ratings—yields on bonds in our sample increase an
average of 10.5 to 14.4 basis points per rating notch—local investment banks get significantly
better yields for high risk and non-rated issues. The presence of a local investment bank
decreases the yield at issue by about 7.1 to 10.2 basis points per rating notch, which represents
about 70% of the effect ratings have on yields. Similarly, while non-rated bonds have yields
about 200 basis points higher than rated bonds, local investment banks are able to cut this “no
rating penalty” by over one third, or 76 basis points. We find similar results when examining
what municipal bond market participants refer to as the “all-in costs”– that is, the sum of the
investment banking spread and yield. Local underwriters significantly lower all-in costs and
significantly reduce the cost-increasing impact of high-risk ratings and non-rated bonds.
We interpret these results as suggesting that relatively high-risk offerings are more
difficult to evaluate and market and investment banks with a local presence are better able to
5
evaluate and/or market “difficult” bond issues than underwriters without a local presence. Our
paper fits most closely with the literatures on the importance of commercial bank location
(Almazan [2002], Hauswald and Marquez [2002]) and how commercial banks evaluate soft and
hard information (Stein [2002], Petersen and Rajan [2002], Berger et al. [2002]). Although
commercial banks may be able to exploit proximity by charging higher rates to nearby borrowers
(Degryse and Ongena [2002]), we find precisely the opposite for investment banks. We attribute
this to investment banks’ inability to informationally capture and exploit client firms as
commercial banks sometimes can through relationship banking (see Sharpe [1990], Rajan
[1992], and Boot [2000]). Thus, in contrast to Petersen and Rajan [2002], our paper provides
evidence that “distance matters” for financial intermediation and soft information production
when ex post monitoring is not feasible.
This paper is among the first to examine investment banks’ ability to evaluate soft
information and to document that investment bank location matters.2 On average, investment
banks with operations in the same state as municipal bond issuers provide services at lower costs
and sell bonds at lower yields, particularly for bond issues that are especially difficult to value
and place with investors. This evidence indicates that local investment banks have an absolute
advantage at providing underwriting services.
Our findings suggest that investment banking firms with active municipal bond
underwriting departments should find it useful to open branch offices in several states to better
evaluate local issuers. Many states, naturally, will agree with this because they want more
companies and employers in their states, as well as because they want lower costs for their bond
2 Malloy [2002] shows that geographically proximate security analysts provide better earnings forecasts. Coval and Moskowitz [2001] show that local mutual fund managers earn abnormal returns investing on local stocks. Their findings are broadly consistent with those presented in this paper – that proximity enhances information production. Neither paper addresses the underwriting role of investment banks.
6
issues. Moreover, many states evaluate underwriters’ “commitment to the state” when deciding
among investment banks vying for deals, and so investment bankers should (and perhaps do)
take this into account.
The remainder of this paper is organized as follows. Section 2 gives a brief description
of municipal bonds, a generalization of the underwriting process and the selection of
underwriters, and a discussion of “pay to play” practices in the municipal bond industry. Section
3 describes the data we employ. Section 4 follows, and presents our results, and the final section
provides concluding remarks.
2. Municipal bonds and underwriting
A. Underwriting and the investment banking process
For a typical “firm commitment” security issuance, investment banks purchase directly
from the issuer all the securities that are to be sold to investors.3 The investment banks pay the
issuer less than the price at which they sell the securities to the public. The difference between
the price at which the securities are bought from the issuer and sold to the public is referred to as
the gross spread. This gross spread must compensate for all the services the investment banking
group provides. The gross spread is generally broken down into three components – the
management fee, the selling concession, and the underwriting fee.4
3 See Ellis, Michaely, and O’Hara [1999] for a detailed description of the institutional details of the investment banking process for initial public offerings of stock. 4 The role of investment banks in financial transactions has drawn substantial attention recently. One focal point of this interest has been the fees that investment banks receive for their services, especially in initial public offerings (IPOs) of equity. See Chen and Ritter [2000], Hansen [2001], Ljungqvist, Jenkinson, and Wilhelm [2001], Torstila [2001], and Butler and Huang [2003].
7
The management fee is the compensation the lead underwriter(s), or bookrunners
receive.5 The selling concession is the amount per share received by investment bankers in the
selling syndicate – the group of investment bankers assembled by the lead underwriter to market
and sell securities to the public.6 The underwriting fee is the amount per security paid to the
investment banks in the underwriting syndicate; this is the compensation they receive for bearing
the risk that the securities will not be sold. The selling syndicate’s responsibility is to line up
investors to buy the securities being offered in the firm commitment offering. The underwriting
syndicate’s responsibility is to take ownership of the securities from the issuer and hold those
securities during the short time period prior to their being placed with the investors.
B. Municipal bonds
The municipal bond (muni) market is inherently different than other new issues markets.
Rather than a corporation issuing securities, a state or local government is, at least indirectly, the
issuer. The bonds typically mature in one to forty years and fund public projects such as roads,
bridges, buildings, airports, and utilities. Every state has statutes that require “open meetings” or
other disclosure of the terms of municipal bond offerings and essentially all deal terms for
negotiated muni offerings become public record before the bonds are actually issued, including
the investment banking fees. This feature facilitates the communication of pricing information
among underwriters, which could in turn facilitate collusion among them. Furthermore, the
Securities and Exchange Commission (SEC) has little power to directly regulate municipal bond
issuers (Beckett 1997).
5 Throughout, we use the terms “bookrunner”, “manager”, and “lead investment bank” interchangeably; “underwriter” can refer to the bookrunner or a non-managing member of the underwriting syndicate. Context will make the meaning clear. 6 In municipal bond offerings, the government body issuing the bonds may choose the selling group and/or underwriting syndicate in whole or in part.
8
Munis generally fall into one of two categories: general obligation (“GO”) bonds and
revenue bonds. GO bonds are backed by the full faith and credit of the issuing entity and thereby
guaranteed. There is usually a limit set on the amount of general obligation indebtedness an
entity can issue at any one time. This limit is often referred to as the debt limit or debt cap.
Revenue bonds do not carry the same guarantee as do GO bonds and are not typically limited by
debt cap statutes. While GO bonds are usually paid from ad valorem revenues such as the
general tax pool, revenue bonds are funded from specific fees, taxes, or assessments on the item
they are supporting.7 GO bonds therefore carry lower interest rates because of the full faith and
credit guarantee while revenue bonds have higher rates since their repayment is dependent upon
the success or failure of the project they support.
Before issuing either type of bond, the issuing entity, with the help of its financial
advisor, must evaluate a few basic questions: how much money is needed to finance the project,
what debt capacity is available, and what financial institutions and advisors will be used. This
paper focuses on the issuer’s choice of investment bankers. But, investment bankers are not the
only professionals involved in a muni issue. Attorneys are also hired to issue opinion letters on
disclosure requirements, the legality of the issue, and tax-exempt status questions. The lawyers
are needed to calm and appease prospective buyers, but according to one state treasurer, the
underwriters are the “quarterbacks” of any bond issue.8
The two most prevalent means of selecting investment bankers are through competitive
bidding and negotiated contract. Competitive bidding is accomplished by soliciting and
receiving sealed bids. The governmental unit receives sealed bids, opens them at a public
7 For example, revenue bonds used to fund a toll road might be repaid using the tolls collected on that road. 8 Source: Louisiana State Treasurer John Neely Kennedy.
9
hearing and reads aloud the deal terms submitted by each potential underwriter. Using this
process, contracts are awarded on the basis of lowest bid received.
A governmental entity wishing to select an underwriter through negotiation first issues a
Request for Proposals (“RFP”) or similar solicitation. Potential underwriters submit written
proposals that are “graded” by the staff of the governmental unit. There may be oral
presentations and question and answer sessions after the grading or the government may award
the contract on the basis of the proposals alone.
C. Pay to Play
Historically, municipal bond underwriters have been notorious for bid rigging, bribery,
insider trading, and other illegal activities (Mitchell and Vogel [1993]). Though recent
regulatory scrutiny has effectively eradicated corruption in the industry, personal and financial
relationships between bond underwriters and politicians were at one time a critical dimension of
competition among rival investment banks. In order to get lucrative underwriting contracts,
investment banks would routinely make substantial campaign and other political contributions to
politicians who would allocate underwriting business for their municipality or state. This
widespread practice became known as “pay-to-play,” and these contributions became known as a
normal cost of doing business in the municipal underwriting industry.
Intense scrutiny of the municipal bond market and pay-to-play practices began in 1993,
shortly after Arthur Levitt became the SEC chairman. The SEC brought nineteen municipal
securities enforcement cases in the three years immediately following Levitt’s appointment.
Reform imminent, the municipal bond underwriting industry voluntarily agreed to eliminate the
pay-to-play political contributions. The initial draft of the self-regulatory plan was written in
1993 under the direction of Frank Zarb, then chairman of Primerica, the parent company of
10
Smith Barney Shearson, a major municipal bond underwriting firm at the time (Fuerbringer
[1993]). In April 1994, the SEC established a rule that investment houses making political
contributions could not sell bonds from that city/state for two years (Bradsher [1994]).9 The
SEC’s rule had its intended effect. Pay-to-play is no longer prevalent, but nonetheless,
municipal bond underwriting is still a relationship-intensive part of the investment banking
business. To avoid confounding effects, we examine municipal bond issues that came to market
subsequent to the pay-to-play era.
3. Data
We obtain information for taxable municipal bond offerings that postdate the pay-to-play
era (specifically, we use the period 1997-2001) from Securities Data Company (SDC). Our
initial sample contains 2,283 bonds which have data for both issue size and the gross spread paid
to investment banks. While taxable munis are a small proportion of the total municipal issues
market, they provide a more attractive laboratory than their non-taxable counterparts. First,
relatively low disclosure requirements reduce data availability for non-taxable munis. Second,
the ability of investment banks to place non-taxable munis would depend heavily upon state-by-
state tax brackets and rates. Using taxable municipals circumvents both these problems.
Each of our bonds has a lead investment bank, or bookrunner, and many have an
underwriting syndicate of other investment banks that help sell the bonds. For each investment
bank in the sample, we hand collect information on company headquarters and the principal
locations of business. We determine these locations primarily from official state websites listing
9 A suit was subsequently brought by William B. Blount, chairman of the Democratic Party in Alabama and municipal banker at Blount Parrish Roton (a Montgomery, Alabama investment bank) that the SEC’s stifling of pay-to-play was a violation of first and tenth amendment rights (Wayne [1994]). The suit was not successful (Gasparino [1998]).
11
companies that have principal business offices in each respective state, and secondarily from the
websites of the Bond Market Association, Underwriters and Financial Advisor Resources, and
Virtual Finance Library. Location information was not available for Nebraska, New Hampshire,
Oklahoma, or Virginia, which had 13, 4, 49, and 41 bonds in the initial sample, respectively. As
we wish to examine municipal bond offerings state-by-state, we further eliminate from our
sample those states that had 5 or fewer bond issues during our sample period. This data filter
removes Delaware, Hawaii, South Dakota, Vermont, and Wyoming, which after our initial
screen had 2, 1, 2, 1, and 0 bonds, respectively. Our final sample has 2191 bonds from 41 states
and the District of Columbia. Summary statistics and variables for the sample are discussed in
the next section.
In our sample, one or more bond rating agencies rate 1724 of the bonds. Bond rating
agencies, particularly Moody’s, have been criticized for their unsolicited ratings of municipal
bonds. Moody’s is well known for assigning unsolicited ratings to mortgage-backed, asset-
backed, and municipal bonds; those ratings were substantially lower than ratings solicited from
competing agencies.10 For example, when the Jefferson County, Colorado School District issued
general obligation bonds in October 1992, they hired Standard and Poor’s to rate the bonds.
S&P assigned a rating of AA. Moody’s, without having been hired, assigned an unsolicited
rating several notches below the S&P rating. To mitigate potential biases, we use Standard and
Poor’s ratings where they are available. If Standard and Poor’s ratings are unavailable we use
Moody’s ratings, and if neither is available we treat the bond as non-rated. Following Cantor
and Packer (1997), we classify the agencies’ ratings to a numeric scale, assigning a value of 1 to
the highest rated bonds (Aaa or AAA), a value of 2 to the next-highest credit quality rating (Aa1
10 For discussion of solicited versus unsolicited ratings in the U.S. see Butler and Rodgers [2002] and Woolley, Schroeder, and Yang [1996].
12
or AA+), and so on. Thus, higher numerical bond ratings can be interpreted as denoting higher
credit risk.
Following Megginson and Weiss [1991], we measure investment bank reputation as
market share by year. Market share is calculated as the total gross proceeds of municipal bond
offerings an investment bank manages in a year divided by the total gross proceeds of all
municipal bond issuances in the year. Traditional measures of investment bank reputation may
not be relevant in the municipal bond market. Issuers of municipal bonds (state and local
governments) may base their decision of which investment banks to hire on factors such as a
particular bank having an office and employing residents of the state, rather than the bank’s
ability to value, market, and place the bonds. In that case, the decision of which bank to hire
may be independent of any reputation, ability, or market share the investment bank enjoys.
4. Results
A. Market Concentration
The market for underwriting municipal bonds is fragmented. Table 1 shows concentration
measures for municipal bond underwriting, and for comparison, investment bank market share
for other new issues markets – seasoned equity offerings, convertible bond issues, and initial
public offerings. Market share among investment banks is substantially less concentrated in the
municipal bond market. This is consistent with investment banks being unable to reap rents from
reputation, and hence not attempting to build market share in the municipal bond market. The
top 10 investment banks have only 73% market share for municipal bond offerings, whereas that
concentration is 86%, 87%, and 98%, respectively for IPOs, SEOs, and convertible bonds (see
Table 1). There are 189 unique bookrunners in our sample. There are 55 investment banks that
13
are the lead manager for only one offering in our sample, and 75 that are lead manager for less
than three. The most prominent investment banks by number of issues for which they are the
lead manager are Goldman Sachs (116 offerings) and William R. Hough (105 offerings).
Goldman Sachs is a national “bulge bracket” investment bank that is routinely at the top of
underwriting league tables (Johnson and Miller [1988]). Hough specializes in fixed-income
securities, including municipal bonds. They are headquartered in Florida and have major offices
in Maryland, Texas, Arizona, South Carolina, and Ohio.
<Insert Table 1 here>
B. Sample Characteristics
Table 2 presents descriptive statistics for our sample. The average issue size in our
sample is $16.8 million. Minnesota has the smallest bond issues, on average, at $2.98 million,
and New Jersey has the largest at $61.0 million. Approximately 82.5% of the bonds in our
sample have local bookrunners. Rhode Island and Florida have the largest percentage of local
bookrunners with 100% and 97.2%, respectively. Nevada has the smallest at 15.8%. Texas has
the most local members of the underwriting syndicates – 97.1% of the investment banks
involved in the underwriting syndicates are local, compared to 81.4% of the entire sample.
Nevada again has the smallest percentage of local syndicate members at 32.2%.
<Insert Table 2 here>
The average percentage spread in our sample is 1.16%. This is substantially lower than
the spreads for equity issuances, which are typically 7.00% for initial public offerings and around
5-6% for seasoned equity offerings. The highest average percent spread for a state is 2.46%
(West Virginia) and the lowest average percent spread is 0.55% (District of Columbia).11
11 One possible reason for these relatively high spreads in West Virginia is that they also have the highest risk bonds, on average.
14
For several observations, yield and/or bond rating data are unavailable. There are 1680
observations with yield data and 1724 observations with bond rating data. The average yield in
our sample is 6.7%, and the average rating is 2.35, which corresponds to slightly better than Aa2
or AA rated bonds. Several issuing states have average ratings in our sample of 1.00, or Aaa –
Louisiana, New Mexico, and Utah, as well as the District of Columbia. West Virginia has the
worst average bond ratings in our sample—5.5, or, halfway between A1 and A2. Note that the
average ratings presented in the tables are the ratings for the bonds that are issued, not the states’
ratings. For instance, Table 2 reports that the average rating for bonds issued by municipalities in
Louisiana is Aaa, but the state’s credit rating is A2.12 The ratings can be different because
municipalities are issuing the bonds, not the state, and credit risk for the entities can be different.
Also, many bonds are insured; insured bonds have a rating that reflects their insurer’s
creditworthiness (generally AAA) rather than the municipality’s creditworthiness. While most
bonds in our sample are rated by one or more rating agencies, many are not. Approximately
79% of the bonds in our sample are rated. State-by-state variation is large—many states issue
only rated bonds, but the proportion of non-rated bonds can range as high as 40% (Indiana and
Iowa) to 70% (West Virginia).
C. Spreads and Differences for Local and Non-Local Underwriters
The average spread in our overall sample is 1.160%. Unlike equity markets, there
appears to be no substantial clustering of spreads at any particular level (Chen and Ritter [2000]).
Figure 1 is a scatterplot of gross spread and the natural logarithm of issue size in millions, and
shows the substantial variation in gross spreads and lack of substantial clustering at any
particular spread.
12 The Louisiana state credit rating is A2 (Moody’s; A for Standard and Poor’s and Fitch) as of January 2001. The information is available in documents on the Louisiana State Legislature web site (www.legis.state.la.us).
15
<Insert Figure 1 here>
Local bookrunners are more prevalent than those without a local presence. Note that, for
instance, William R. Hough, while headquartered in Florida, has offices in Maryland, Texas,
Arizona, South Carolina, and Ohio, and so would be considered “local” for bond issues in any of
those states. There are 1810 local bookrunners and 381 non-local bookrunners in our sample. In
only seven states (Connecticut, District of Columbia, Michigan, Mississippi, Nevada, North
Dakota, and South Carolina) are there fewer municipal bond issues managed by local
bookrunners than by non-local underwriters.
Table 3 presents mean spreads yields, issue size, and ratings by state and by local/non-
local bookrunners. The overall mean spread charged by non-local bookrunners is 1.340%,
whereas the overall mean spread charged by local bookrunners is 1.121%. The difference in
means is statistically significantly different from zero (t = 4.23). The difference between local
and non-local spreads on a state-by-state basis can be quite large, however. For those states with
more than two non-local underwriters, the largest average non-local spreads occur in Kansas,
Louisiana, and New Jersey (3.31%, 2.37%, and 2.30%, respectively), and the smallest average
non-local spreads occur in Colorado and Connecticut (0.39% and 0.47%, respectively). For
those states with more than two local underwriters, the largest average local spreads occur in
Alaska, Kansas, and West Virginia (1.57%, 1.56%, and 2.77%, respectively), and the smallest
average local spreads occur in Kentucky and Maine (0.79% and 0.73%, respectively).
<Insert Table 3 here>
The overall mean yield on bonds in our sample that are underwritten by non-local
bookrunners is 6.91%, whereas the overall mean yield for bonds underwritten by local
bookrunners is 6.67%. The difference in means is statistically significantly different from zero (t
16
= 2.81). The overall mean size of bonds underwritten by non-local bookrunners is $25.1 million,
versus $32.5 million for bond underwritten by locals. The difference is statistically
indistinguishable from zero. The average numerical bond rating is 2.61 for non-local
bookrunners and 2.30 for locals. The difference of means is statistically different than zero (t =
2.29).
D. Spreads and Yields by Categories
Table 4 presents descriptive statistics for yield, spread, total cost (i.e., yield plus spread –
sometimes referred to as the “all-in cost” by underwriters and local governments), and
percentage locally underwritten for several categories: by rating, by proximity of underwriter, by
year, and several sub-categories. Surprisingly, yield and spread are higher for AAA or Aaa rated
bonds than bonds with worse ratings, underscoring the importance of controlling for various
determinants of yields and spreads. Non-rated bonds have the highest costs (average 9.62% total
cost) and a low proportion of local underwriters (78.5%). Most bonds (1442) are underwritten
by locals and are rated.
<Insert Table 4 here>
E. Correlations
Table 5 presents pairwise correlations between variables. Issue proceeds (Size) and the
log of proceeds are related to percent spread and dollar spread. Consistent with other studies
(e.g., Chen and Ritter [2000], Hansen [2001]), investment banking fees decrease as a percentage
of offer size, although the total dollar fees increase with offer size. Size is also related to bond
rating – larger issues are associated with better ratings. This is consistent with issuers taking
advantage of strong credit ratings to issue large bond issues, and less credit-worthy issuers being
17
unable to issue larger bond offerings. Higher reputation underwriters tend to be used for larger
issues, and larger issues tend to have longer maturities.
<Insert Table 5 here>
The percent gross spread is strongly related to the bond yield, bond rating, and
bookrunner reputation (market share). Consistent with the findings of Jewell and Livingston
[1998], investment banks charge higher spreads to underwrite less credit-worthy bonds. Further,
higher reputation investment banks tend to underwrite safer bonds. Interestingly, higher
reputation investment banks charge lower spreads (on average) than their lower reputation
counterparts. This is consistent with a fragmented market where firms are unable to earn rents
through reputation building.
Where there are local bookrunners, the composition of the entire underwriting syndicate
tends to be local as well; the correlation between the Local dummy variable for a local
bookrunner and the In-state variable, which denotes the percentage of the underwriting syndicate
that is local, is over 90%. All the regressions described below were also done with the In-state
variable replacing Local. As one might expect, all the results were qualitatively quite similar,
and so are not reported.
F. The Effect of Local Investment Banks on Gross Spreads, Yields and Total Costs
We hypothesize that local investment banks have an advantage over non-local investment
banks at providing underwriting services. Because gross spread is related to several
characteristics of the bonds and the underwriters as described above, it is important to control for
these in order to test our hypothesis. To test our hypothesis, we regress gross spread on variables
to capture the effects of local underwriters and on control variables. Table 6 presents our
regression results.
18
<Insert Table 6 here>
We regress gross spread and yield to maturity on several independent variables. In all
our regressions, the percentage of variation that is explained by our model is 34 to 40%. Our first
two regression models regress gross spread on Ln(Size), underwriter reputation as proxied by a
log transformation of investment bank market share, Bond Rating, a dummy variable for Non-
rated bonds, years to maturity, the local bookrunner dummy, dummy variables for each year in
the sample (omitting 2001), and a constant term. In the second regression we also include
dummy variables for each state that has at least five local underwriters and at least five non-local
underwriters. Because many states have only one non-local underwriter (and one state, Rhode
Island, has none), including dummy variables for these states would be capturing not only any
state-specific effects, but also the effects of local underwriters in general. Requiring both several
local and non-local underwriters mitigates this problem. Twenty-four states in our sample meet
this requirement.13
As expected, percent spreads decline with the size of the bond issue, reflecting economies
of scale to larger bond issues. Investment bank market share is also negatively related to gross
spreads. Bonds with higher numerical ratings—that is, with more credit risk—and bonds that are
not rated are associated with higher spreads. For each incremental rating notch, gross spreads
increase by 2.70 to 2.85 basis points. The absence of a rating increases fees about 53 basis
points. These findings are consistent with underwriters having more difficulty placing higher
risk bonds, and requiring more compensation for the additional work. The presence of a local
underwriter, after controlling for other determinants of fees, statistically significantly decreases
gross spreads, which we interpret as evidence of an advantage that local bookrunners enjoy over
13 Results are qualitatively similar when we use less restrictive screens for state-specific dummies.
19
non-locals. This advantage may manifest itself as having better networks of potential investors,
or as a superior ability to evaluate municipal bond offerings and certify them to investors.
The second two regression models in Table 6 regress yield to maturity on the same
variables—Ln(Size), underwriter reputation as proxied by a log transformation of investment
bank market share, Bond Rating, a dummy variable for Non-rated bonds, years to maturity, the
local bookrunner dummy, dummy variables for each year in the sample (omitting 2001), and a
constant term. In the second of these two regressions we also include state-specific dummy
variables as described above. Results indicate that yields decline with the size of the bond issue,
and that investment bank market share is unrelated to yields. Bonds with higher numerical
ratings—that is, with more credit risk—and bonds that are not rated are associated with
significantly higher yields. For each incremental rating notch, yields increase by 4.5 to 6 basis
points. The absence of a rating increases yields by 139 to 145 basis points. These findings are
consistent with investors requiring a premium to hold higher risk bonds and bonds without a
rating. The presence of a local underwriter, after controlling for other determinants of yields,
statistically significantly decreases yields.
The last two regression models use total costs (that is, the sum of the investment banking
spread and the yield to maturity on the bonds) as the dependent variable. Consistent with the
regressions that use spreads and yields individually as dependent variables, these results indicate
that total costs decline with the size of the bond issue, that bonds with more credit risk and bonds
that are not rated are associated with significantly higher yields. For each incremental rating
notch, total costs increase by 7.8 to 9.5 basis points. The absence of a rating increases total costs
by 210 to 217 basis points. The presence of a local underwriter, after controlling for other
determinants of costs, statistically significantly decreases total costs by 27 to 30 basis points.
20
We interpret this as further evidence of the advantage that local investment banks have over non-
locals.
G. The Marginal Effect of Local Underwriters on Gross Spreads
This section and the next contain our main results. In the regressions described above,
we use a dummy variable for bonds underwritten by local investment banks to capture the impact
that investment bank “locality” has on fees and yields. Those regressions indicate that local
investment banks have an absolute advantage over non-local counterparts. For which bonds is
this advantage the strongest? It is this question that this section and the next two shed light upon.
In Table 7, we revisit the regressions described above, but add two additional variables.
We regress gross spreads on the log of issue size, investment bank market share, numerical bond
rating (as described above), a dummy variable for non-rated bonds, the years to the bond’s
maturity, a local bookrunner dummy, and controls for each year and state-specific dummies. We
add two variables to this model specification—variables that capture the interaction between
local investment banks and bond rating, and the interaction between local investment banks and
non-rated bonds. By controlling for the total effect that local investment banks have
(through the Local dummy), the interpretation of the coefficients on the interaction terms are the
marginal effects of a local bookrunner.
<Insert Table 7 here>
The results indicate that local investment banks have the strongest advantage at
underwriting extremely difficult to place and informationally opaque bonds. Though higher
credit risk bonds are associated with higher investment bank fees, this credit risk “penalty” is
much lower for bonds underwritten by local investment banks. While on average fees increase
about 5.7 to 6.8 basis points per incremental “notch” in credit rating, local investment banks
21
charge about 3.5 to 4.9 basis points less in gross spreads per credit rating notch than non-local
underwriters. This difference represents approximately two thirds of the incremental effect that
ratings have on gross spreads and is statistically significant.
Similarly, bonds that are not rated also have higher fees. Non-rated bonds have fees
more than 70 basis points higher, on average, than rated bonds. Again, local investment banks
impose a much smaller “penalty” for underwriting non-rated bonds—about one third less than
their non-local counterparts. This difference is statistically significant.
Note that when including the two interaction terms, Local X Rating and Local X Non-
Rated, the coefficient on the Local dummy becomes weakly positive. This is because the
strongest effect that local investment banks have on lowering fees (holding other things constant)
is in the non-rated bonds and high credit risk bonds. Thus, it is the interaction terms that absorb
the fee-reducing impact of local investment banks and it is the relatively high-risk municipal
bond offerings for which local investment banks have a comparative advantage. Conversely,
when there is very little asymmetry of information and bonds are informationally transparent,
local underwriters have no comparative advantage. Bonds that are least informationally sensitive
are those rated AAA (a numerical rating of 1). Consider the coefficients for Local, Local X
Rating, and Local X Non-rated in Table 7. The sum effect of these three factors on fees for a
bond rated AAA and underwritten by a local investment bank is 0.0927 [the Local dummy
effect] + (-0.0488 * 1) [the Local X Rating effect] + 0 [the Local X Non-rated effect] = 0.048.
Thus, the net effect of a local bookrunner on fees for the most informationally transparent bond
is weakly positive. Though the magnitude is positive, we cannot reject that the sum of the
coefficients is different from zero for any of the regression specifications (test statistics not
reported). Overall, our interpretation of the regression results presented in Table 7 is that local
22
investment banks are better able than non-local counterparts to evaluate and market the bond
issues that are otherwise difficult to place.
H. The Marginal Effect of Local Underwriters on Yields
This “local” effect has an even larger effect on the yields on the bonds. In Table 8, we
parallel the regressions described in the previous section, but the dependent variable is the yield
on the bonds. Thus, we regress yields on the log of issue size, investment bank market share,
numerical bond rating (as described above), a dummy variable for non-rated bonds, the years to
the bond’s maturity, a local bookrunner dummy, and controls for each year and state-specific
dummies. As above, we include the two interaction terms, Local X Rating and Local X Non-
Rated, so that we can determine the marginal effects of a local bookrunner.
<Insert Table 8 here>
As in the Table 7 regressions, when we include the two interaction terms the coefficient
on the Local dummy becomes weakly positive.14 The strongest effect that local investment
banks have on lowering yields (holding other things constant) is in the non-rated bonds and high
credit risk bonds. We find that though yields increase, on average, 10.5 to 14.4 basis points per
credit rating notch, local investment banks get significantly better yields for high risk bonds.
Local bookrunners, as opposed to non-locals, decrease the yield at issue by about 7.1 to 10.2
basis points per rating notch. This represents about 70% of the effect ratings have on yields.
Non-rated bonds have yields about 200 basis points higher than rated bonds. Local
investment banks, though, are able to reduce this “no rating penalty” by about 76 basis points, or
approximately one third. As with the regressions where we investigate the determinants of gross
14 The interpretation of this result is the same as the interpretation of the analogous result described in the previous section. The net effect of a local bookrunner on yields—that is, the sum of the coefficients on Local and Local X Rated--is about a 9 basis point increase. However, we cannot reject that this increase is statistically different from zero for any of the regression models (test statistics not reported).
23
spreads, we interpret these results as suggesting that local investment banks are better able to
evaluate and/or market difficult bond issues, and this manifests itself in lower yields for the
issuer.
I. The Marginal Effect of Local Underwriters on Total Costs
The “local” effect on total costs is even sharper than that on yields and spreads
separately. In Table 9, we regress total costs on the same independent variables as in the
previous two tables. As above, we include the two interaction terms, Local X Rating and Local X
Non-Rated, so that we can determine the marginal effects of a local bookrunner.
<Insert Table 9 here>
Local investment banks have a strong impact on lowering total costs (holding other
things constant) for non-rated bonds and high credit risk bonds. We find that though total costs
increase about 20 basis points per credit rating notch on average, local investment banks
significantly reduce total costs for high risk bonds. Local bookrunners, as opposed to non-locals,
decrease the total cost by about 13.7 to 16.3 basis points per rating notch. This represents about
70% of the effect ratings have on total costs.
Non-rated bonds have total costs about 300 basis points higher than rated bonds. Local
investment banks reduce this “no rating penalty” by about 105 basis points, or approximately one
third. We interpret these results as suggesting that local investment banks are better able to
evaluate and/or market difficult bond issues, and this manifests itself in lower costs for the
issuer.
24
5. Conclusion
The impact that local investment banks have on municipal bond costs is non-trivial.
Despite the bad reputation municipal bond underwriters earned in the pay-to-play era, we find
that investment banks with local connections have substantial comparative and absolute
advantages over non-local counterparts. Using 2191 municipal bond offerings from 1997-2001,
we find that investment banks with an in-state presence charge lower fees and sell bonds at lower
yields than non-local underwriters. This local advantage is especially strong for the underwriting
of high-risk and non-rated municipal bonds. One interpretation of these results is that risky and
non-rated bonds are more difficult to evaluate and market, and investment banks with a local
presence are better able to assess soft information and can better handle “difficult” bond issues
than non-local counterparts.
An implication of our findings is that municipal bond underwriting investment banks
might find it useful to open branch offices in several states to better evaluate local issuers and to
establish long-term connections with local banks, insurance companies, and others who are
regular purchasers of municipal bonds. Indeed, this is consistent with our finding that over 80%
of municipal bonds are underwritten by an investment bank with a local office. Furthermore,
because many states evaluate underwriters’ commitment to the state when selecting among
investment banks vying for deals, a local presence might further benefit investment bankers
looking for deal flow.
Our paper also provides indirect evidence on the distinctiveness of relationship lending
by commercial banks. Other researchers have shown that commercial banks are able to exploit
informationally captured borrowers by charging higher rates. In contrast, we find that local
investment banks charge lower fees and sell bonds at lower yields than non-local counterparts.
25
That investment banks are unable to informationally capture client firms points to the importance
of on-going ex post monitoring that commercial banks can pursue.
26
References Almazan, Andres, 2002, A Model of Competition in Banking: Bank Capital vs Expertise, Journal of Financial Intermediation 11, 87-121. Altinkiliç, Oya and Robert S. Hansen, 2000, Are There Economies Of Scale In Underwriting Fees? Evidence Of Rising External Financing Costs, Review of Financial Studies 13, 191-218. Beckett, Paul, 1997, SEC Hits Barrier to Muni-Bond Reform, Wall Street Journal, 5/16/1997, A13. Berger, Allen N., Nathan H. Miller, Mitchell A. Petersen, Raghuram G. Rajan, and Jeremy C. Stein, 2002, Does Function Follow Organizational Form? Evidence From the Lending Practices of Large and Small Banks, NBER working paper. Bhattacharya, S., and A. Thakor, 1993, Contemporary Banking Theory, Journal of Financial Intermediation 3, 2-50. Bradsher, Keith, 1994, SEC Curbs Donations by Bond Dealers, New York Times, 4/7/1994, D1. Boot, Arnoud W. A., 2000, Relationship Banking: What Do We Know?, Journal of Financial Intermediation 9, 7-25. Butler, Alexander W. and Pinghsun Huang, 2003, On the Uniformity of Investment Banking Spreads: The Seven Percent Solution is not Unique, Journal of Multinational Financial Management 13, 265-272. Butler, Alexander W. and Kimberly J. Rodgers, 2002, Relationship Rating: How Do Bond Rating Agencies Process Information?, Rice University and Penn State University Working Paper. Cantor, Richard and Frank Packer, 1997, Differences of opinion and selection bias in the credit rating industry, Journal of Banking and Finance 21, 1395-1417. Chen, H.C. and Jay Ritter, 2000, The 7 Percent Solution, Journal of Finance 55, 1105-1132. Coval, Joshua D., and Tobias J. Moskowitz, 2001, The Geography of Investment: Informed Trading and Asset Prices, Journal of Political Economy 109, 811–841. Degryse, Hans and Steven Ongena, 2002, Distance, Lending Relationships, and Competition, working paper. Ellis, Katrina, Roni Michaely, and Maureen O’Hara, 1999, A Guide to the Initial Public Offering Process, Corporate Finance Review 3, 14-18. Fisman, Raymond, 2001, Estimating the Value of Political Connections, American Economic Review, 91(4), 1095-1102. Fuerbringer, Jonathan, 1993, Muni Industry Trying to Fix Itself, New York Times, 10/11/1993, D5. Gasparino, Charles, 1998, ‘Pay to Play’ Getting New SEC Review, Wall Street Journal, 12/17/1998, C1. Hansen, Robert, 2001, Do Investment Banks Compete in IPOs? The Advent of the “7% Plus Contract”, Journal of Financial Economics 59, 313-346.
27
Hauswald, Robert and Robert Marquez, 2002, Competition and Strategic Information Acquisition in Credit Markets, University of Maryland working paper. Jewell, Jeff and Miles Livingston, 1998, Split Ratings, Bond Yields, and Underwriter Spreads, Journal of Financial Research 21 (2), 185-204. Johnson, James M. and Robert E. Miller, 1988, Investment Banker Prestige and the Underpricing of Initial Public Offerings, Financial Management 17 (2), 19-29. Light, Larry and Leah Spiro, 1993, Behind Closed Doors: The Private Bids for Public Funds, Business Week, 5/24/1993, 123. Ljungqvist, Alexander, Tim Jenkinson, and William Wilhelm, 2001, Global Integration in Primary Equity Markets: The Role of U.S. Banks and U.S. Investors, forthcoming Review of Financial Studies. Malloy, Christopher, 2002, The Geography of Equity Analysis, University of Chicago working paper. Megginson, William L. and Kathleen H. Weiss, 1991, Venture Capital Certification in Initial Public Offerings, Journal of Finance 46, 879-903. Mitchell, Constance and Thomas T. Vogel, Jr., Illegal Payments Mar the Muni Market, Wall Street Journal, 5/5/1993, C1. Petersen, Mitchell A. and Raghuram G. Rajan, 2002, Does Distance Still Matter? The Information Revolution in Small Business Lending, Journal of Finance, forthcoming. Rajan, Raghuram G., 1992, Insiders and Outsiders: The Choice between Informed and Arm’s-Length Debt, Journal of Finance 47, 1367-1400. Sharpe, Steven, 1990, Asymmetric Information, Bank Lending and Implicit Contracts: A Stylized Model of Customer Relationships, Journal of Finance 45, 1069-1087. Stein, Jeremy, 2002, Information Production and Capital Allocation: Decentralized versus Hierarchical Firms, Journal of Finance 57, 1891-1921. Torstila, Sami, 2001, The Clustering of IPO Gross Spreads: International Evidence, Journal of Financial and Quantitative Analysis, forthcoming. Wayne, Leslie, 1994, Tests of Rights for Municipal Bankers, New York Times, 12/9/1994, D1. White, Halbert, 1980, A Heteroskedasticity-Consistent Covariance Estimator and a Direct Test for Heteroskedasticity, Econometrica 48, 817–838. Woolley, Suzanne, Michael Schroeder, and Catherine Yang, 1996, Now it’s Moody’s Turn for a Review: The Justice Department is Probing Possible Antitrust Violations, Business Week, p. 116, April 8, 1996.
28
Figure 1 – Underwriter fees vs. Issue Size
The sample includes 2170 taxable municipal bond offerings from 1997-2001. Six observations have a spread greater than 6% and are omitted from the graph for presentation purposes.
0
1
2
3
4
5
6
-2.5 0 2.5 5 7.5
Ln(Issue Size $mm)
Gro
ss S
prea
d %
29
Table 1 – Investment bank concentration ratios This table presents summary information for offerings of taxable municipal bonds, initial public offerings, seasoned equity offerings, and convertible bonds. Aggregate Proceeds and Number of Issues reflect the total dollar amount of offering proceeds in billions and the total number of security offerings over the period 1997-2001, respectively. Concentration measures reflect the market share of the top 5, 10, and 25 lead bookrunners for each type of deal over the same time period. All data come from Securities Data Company.
Aggregate Proceeds ($bb)
Number of Issues
5-firm concentration
10-firm concentration
25-firm concentration
Municipal Bonds 63.0 4208 58.2% 73.2% 87.2%
Initial Public Offerings 237.3 2067 69.6% 86.4% 96.2%
Seasoned Equity Offerings 403.7 2551 66.6% 87.2% 97.7%
Convertible Bonds 82.9 177 85.9% 98.0% 100.0%
30
Table 2 – Descriptive Statistics by State Each row shows the mean value for each variable for each state. Local denotes a dummy variable that takes a value of 1 if the lead bookrunner for an offering has an office in the state. Instate denotes the proportion of the underwriting syndicate, including the lead bookrunner, that has a local presence. Spread denotes the investment banking gross spread for the issues, expressed in percentage (%). Size reflects the proceeds of the bond issues expressed in millions of dollars. YTM denotes the yield to maturity for the bonds in percent. Maturity reflects the life of a bond in years. Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc. Non-rated refers to the fraction of the bonds that are not rated by Moody’s or Standard and Poor’s. State Max Obs Local Instate Spread % Size YTM % Maturity Rating Non-Rated Alabama 40 0.850 0.844 1.287 11.5 7.066 12.8 1.83 0.275 Alaska 5 0.800 0.732 1.729 53.0 7.297 23.6 2.00 0.000 Arizona 34 0.647 0.726 1.380 66.3 7.124 13.6 3.28 0.147 Arkansas 27 0.519 0.668 1.019 8.8 6.751 16.6 1.74 0.296 California 252 0.810 0.800 1.045 23.8 6.831 20.9 2.14 0.155 Colorado 103 0.961 0.937 1.082 18.6 6.636 18.1 2.35 0.136 Connecticut 33 0.455 0.513 0.754 85.5 6.429 8.9 2.39 0.152 D. of Columbia 6 0.333 0.333 0.549 46.2 6.562 24.0 1.00 0.000 Florida 142 0.972 0.965 1.084 6.2 7.093 16.7 1.55 0.134 Georgia 62 0.742 0.750 1.228 12.4 6.919 12.1 2.76 0.210 Idaho 27 0.926 0.926 1.064 5.9 6.885 11.4 1.36 0.185 Illinois 122 0.820 0.784 1.208 29.2 6.653 14.8 1.89 0.279 Indiana 44 0.864 0.763 0.993 186.8 6.274 15.2 1.92 0.409 Iowa 87 0.494 0.486 1.572 28.3 6.532 4.7 4.40 0.402 Kansas 37 0.892 0.869 1.753 4.5 6.792 11.7 3.04 0.324 Kentucky 22 0.682 0.722 1.014 102.4 8.020 13.0 1.63 0.273 Louisiana 35 0.886 0.821 1.129 7.9 6.675 19.6 1.00 0.143 Maine 8 0.875 0.820 1.076 12.7 6.501 9.1 2.57 0.125 Maryland 23 0.870 0.870 1.142 7.1 6.576 16.8 2.50 0.217 Massachusetts 20 0.900 0.799 0.987 19.6 7.278 14.5 2.35 0.000 Michigan 45 0.444 0.470 1.189 21.3 6.852 14.8 2.16 0.311 Minnesota 93 0.957 0.953 1.477 24.4 6.831 8.2 3.84 0.387 Mississippi 30 0.467 0.528 1.290 13.4 6.751 11.2 2.00 0.333 Missouri 62 0.984 0.940 1.106 21.8 6.741 19.6 1.15 0.226 Montana 6 0.833 0.883 1.102 24.3 5.700 18.5 3.83 0.000 Nevada 19 0.158 0.322 0.815 17.4 6.359 14.1 1.24 0.105 New Jersey 58 0.914 0.886 1.229 94.5 7.086 14.6 2.33 0.207 New Mexico 33 0.697 0.598 1.147 6.3 6.680 22.6 1.00 0.121 New York 108 0.954 0.951 0.780 34.5 6.470 15.2 4.15 0.157 North Carolina 19 0.789 0.690 1.024 20.1 6.285 9.6 2.94 0.053 North Dakota 7 0.429 0.429 1.141 4.3 6.475 8.6 4.14 0.000 Ohio 44 0.773 0.787 1.215 237.8 6.620 11.2 1.57 0.318 Oregon 40 0.825 0.871 1.154 18.8 6.752 13.6 2.85 0.350 Pennsylvania 87 0.862 0.836 0.941 26.6 6.122 12.8 1.95 0.057 Rhode Island 7 1.000 0.879 0.721 8.8 6.585 14.7 1.57 0.000 South Carolina 15 0.467 0.640 1.066 55.4 7.497 18.1 2.00 0.133 Tennessee 48 0.813 0.830 1.132 10.9 6.652 10.8 2.16 0.333 Texas 160 0.969 0.971 1.361 17.7 6.892 13.6 2.12 0.194 Utah 26 0.962 0.964 0.870 15.0 6.513 23.2 1.00 0.077 Washington 63 0.952 0.829 1.062 7.0 6.161 19.6 2.81 0.063 West Virginia 14 0.786 0.807 2.430 3.9 6.859 7.9 5.50 0.714 Wisconsin 78 0.846 0.777 1.236 10.4 6.128 11.0 3.35 0.308 All 2191 0.826 0.814 1.160 31.2 6.711 14.8 2.35 0.213
31
Table 3 – Means by Local/Non-Local Bookrunner and by State Each row shows the mean value for each variable separately for bonds underwritten by local bookrunners and non-local bookrunners for each state. Local denotes the lead bookrunner for an offering has an office in the state. Spread denotes the investment banking gross spread for the issues, expressed in percentage (%). YTM denotes the yield to maturity for the bonds in percent. Size reflects the proceeds of the bond issues expressed in millions of dollars. Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc. Test statistics for difference of means (for the full sample) are reported in the bottom row. *** and ** denote statistical significance at 1%, 5%, and 10% levels, respectively.
Spread % YTM % Size Rating State Local Non-local Local Non-local Local Non-local Local Non-local Alabama 1.372 0.805 7.115 6.850 12.71 4.32 1.72 2.50 Alaska 1.569 2.368 7.320 7.250 62.85 13.60 1.00 6.00 Arizona 1.438 1.273 7.139 7.102 100.60 3.54 3.32 3.20 Arkansas 1.128 0.903 6.781 6.716 3.11 14.98 2.57 1.25 California 1.086 0.871 6.975 6.256 27.97 5.93 2.09 2.32 Colorado 1.110 0.389 6.685 5.657 15.29 101.08 2.35 2.25 Connecticut 1.092 0.472 6.804 6.080 36.45 126.46 2.08 2.63 D. of Columbia 0.613 0.518 6.160 6.763 59.85 39.33 1.00 1.00 Florida 1.073 1.465 7.069 8.563 6.30 2.08 1.52 3.00 Georgia 1.007 1.861 6.655 7.484 16.46 0.73 2.78 2.67 Idaho 1.025 1.556 6.829 7.525 6.34 1.00 1.38 1.00 Illinois 1.145 1.495 6.420 7.637 33.38 10.38 1.68 2.87 Indiana 0.926 1.418 6.075 7.519 215.56 4.62 1.96 1.50 Iowa 1.372 1.768 6.675 6.413 9.24 46.89 3.50 5.07 Kansas 1.564 3.312 6.790 6.806 4.31 5.78 2.96 4.00 Kentucky 0.788 1.500 8.289 7.281 16.46 286.64 1.67 1.50 Louisiana 0.969 2.369 6.473 8.633 8.89 0.35 1.00 1.00 Maine 0.730 3.500 6.043 9.250 13.30 8.40 2.57 N/A Maryland 1.204 0.730 6.758 5.485 6.59 10.30 2.80 1.00 Massachusetts 0.960 1.236 7.223 8.000 21.63 1.00 2.17 4.00 Michigan 0.885 1.432 6.313 7.307 26.50 17.07 2.07 2.25 Minnesota 1.482 1.370 6.857 6.313 25.47 1.40 3.80 6.00 Mississippi 1.304 1.279 6.705 6.789 18.56 8.85 2.33 1.86 Missouri 1.076 2.950 6.741 6.750 22.15 0.20 1.15 N/A Montana 1.133 0.950 5.700 N/A 14.10 75.00 2.80 9.00 Nevada 0.642 0.847 5.154 6.618 94.33 3.02 1.00 1.29 New Jersey 1.128 2.300 7.023 7.684 103.32 0.64 2.05 6.33 New Mexico 0.810 1.923 6.365 7.545 8.21 2.00 1.00 1.00 New York 0.740 1.600 6.367 7.864 36.06 2.82 4.15 4.33 North Carolina 1.077 0.823 6.417 5.847 21.97 12.90 3.36 1.50 North Dakota 0.793 1.403 5.800 7.150 7.80 1.75 3.00 5.00 Ohio 1.154 1.425 6.108 8.044 306.22 5.03 1.71 1.00 Oregon 1.185 1.008 6.656 7.308 21.83 4.37 2.81 3.00 Pennsylvania 0.916 1.096 6.109 6.244 30.43 2.67 2.03 1.45 Rhode Island N/A 0.721 N/A 6.585 N/A 8.81 N/A 1.57 South Carolina 1.031 1.097 6.373 8.060 38.87 69.93 2.86 1.00 Tennessee 0.990 1.750 6.437 7.300 11.12 10.18 2.14 2.25 Texas 1.346 1.820 6.859 7.720 18.17 2.06 2.13 1.00 Utah 0.862 1.056 6.470 7.375 15.53 1.10 1.00 1.00 Washington 1.038 1.536 6.184 5.080 6.19 23.87 2.91 1.00 West Virginia 2.770 1.184 6.923 6.538 2.74 8.23 6.00 5.00 Wisconsin 1.176 1.570 5.987 6.829 10.70 8.95 3.36 3.33 All 1.123 1.336 6.669 6.907 32.51 25.09 2.30 2.61 t-test for difference of means 4.23*** 2.81*** -0.71 2.29**
32
Table 4 – Descriptive Statistics by Categories Each row shows the mean value for each variable for each category. Local denotes a dummy variable that takes a value of 1 if the lead bookrunner for an offering has an office in the state; % Local denotes the percentage of local bookrunners in each category. Spread denotes the investment banking gross spread for the issues, expressed in percent. Yield denotes the yield to maturity for the bonds in percent. Total cost denotes the sum of yield and spread where both data are available. Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc. Non-rated refers to the fraction of the bonds that are not rated by Moody’s or Standard and Poor’s.
Category Yield Spread Total Cost % Local Max
Obs Rating = 1 6.482 0.973 7.496 85.0% 1092 Rating = 2, 3, or 4 6.292 0.808 7.129 83.9% 342 Rating = 5, 6, or 7 6.431 1.283 7.803 78.2% 238 Rating = 8 or higher 6.964 1.395 8.295 80.4% 51 Non-rated bond 7.688 1.692 9.621 78.5% 465 Local underwriter 6.667 1.107 7.841 1807 Non-local underwriter 6.906 1.318 8.348 381 Issued in 1997 6.777 1.237 8.034 83.8% 346 Issued in 1998 6.365 1.237 7.652 80.6% 480 Issued in 1999 6.863 1.188 8.130 80.8% 556 Issued in 2000 7.677 1.048 8.819 84.2% 406 Issued in 2001 5.982 0.986 7.071 84.9% 384 Local underwriter and rated 6.457 0.971 7.466 1442 Local underwriter and non-rated 7.556 1.644 9.426 365 Non-local underwriter and rated 6.477 1.122 7.674 281 Non-local underwriter and non-rated 8.147 1.868 10.300 100
33
Table 5 – Correlations Among Variables Data are for the entire sample of 2191 municipal bond issues, but the number of observations for each variable varies with data availability. Correlations among variables and the p-values for those correlations are shown (below). Size denotes the proceeds of a bond issue, and Ln(Size) denotes the natural logarithm of Size. Local is a dummy variable that takes a value of 1 if the lead manager is “local” (i.e., has an office in that state) and zero otherwise. Instate is the percentage of the underwriting syndicate that is “local.” Spread % is the investment bank gross spread as a percentage of the proceeds. Yield is the yield to maturity on the bonds. Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc. Maturity reflects the life of a bond in years. Market share is the percent proportion of all proceeds a bookrunner managed in a year. Ln(Mkt Share) is the natural log of ((market share * 10) + 1).
Ln(Size) Size Local Instate Spread % Yield Rating Maturity Ln(Mkt Share)
Size 0.4115 0.0000 Local 0.1106 0.0153 0.0000 0.4752 Instate 0.1001 0.0142 0.9011 0.0000 0.5056 0.0000 Spread % -0.4113 -0.0625 -0.0901 -0.0754 0.0000 0.0034 0.0000 0.0004 Yield -0.1658 -0.0210 -0.0683 -0.0567 0.2905 0.0000 0.3880 0.0051 0.0199 0.0000 Bond Rating -0.1193 -0.0569 -0.0551 -0.0495 0.1275 0.0290 0.0000 0.0181 0.0221 0.0400 0.0000 0.2887 Maturity 0.4438 0.1297 0.0802 0.0675 -0.2320 0.1128 -0.1669 0.0000 0.0000 0.0002 0.0017 0.0000 0.0000 0.0000 Ln(Mkt Share) 0.4498 0.0420 0.1045 0.0907 -0.3391 -0.0744 -0.1430 0.3313 0.0000 0.0496 0.0000 0.0000 0.0000 0.0023 0.0000 0.0000 Market Share 0.3395 0.0723 0.1085 0.0806 -0.2071 -0.0250 -0.0877 0.2275 0.7327 0.0000 0.0007 0.0000 0.0002 0.0000 0.3051 0.0003 0.0000 0.0000
34
Table 6 – Total Effects of Local Underwriters on Gross Spreads, Yields, and Total Costs Data are for the sample of municipal bond issues with data availability (2162 observations for regressions where gross spread is the dependent variable, 1676 observations where yield to maturity or total cost is the dependent variable). Gross spread denotes the investment banking gross spread for expressed in percent. Yield denotes the yield to maturity in percent. Total cost denotes the sum of yield and gross spread where both data are available. Ln(Size) denotes the natural logarithm of the proceeds of the bond issue. Non-rated is a dummy variable that takes a value of one if the bond is not rated and zero otherwise. Bond Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc., and a value of zero if the bond is not rated. Note that assigning non-rated bonds a value of zero is arbitrary, and any value other than numerical ratings assigned to our bonds would produce identical results. Maturity reflects the life of a bond in years. Ln(Market Share) is the natural log of ((market share * 10) + 1), where market share is the percent proportion of all proceeds a bookrunner managed in a year. Local takes a value of 1 if the lead manager is “local” (i.e., has an office in that state) and zero otherwise. Robust standard errors (see White 1980) are used to calculate p-values, which appear which appear in parentheses below the coefficient estimates.
Dependent = Gross Spread
Gross Spread Yield Yield Total
Costs Total Costs
Ln(Size) -0.1374 -0.1314 -0.1107 -0.1061 -0.2414 -0.2283 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Ln(Market Share) -0.0697 -0.0689 0.0247 0.0167 -0.0378 -0.0443 (0.000) (0.000) (0.195) (0.403) (0.109) (0.070) Bond Rating 0.0270 0.0285 0.0460 0.0600 0.0788 0.0950 (0.002) (0.001) (0.001) (0.000) (0.000) (0.000) Non-rated dummy 0.5250 0.5347 1.3918 1.4525 2.0996 2.1688 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Maturity 0.0003 -0.0001 0.0321 0.0311 0.0362 0.0351 (0.820) (0.932) (0.000) (0.000) (0.000) (0.000) Local Bookrunner -0.0648 -0.0696 -0.1543 -0.2009 -0.2712 -0.3018 (0.089) (0.086) (0.034) (0.012) (0.003) (0.003) Year dummies Yes Yes Yes Yes Yes Yes State dummies No Yesa No Yesa No Yesa Constant Yes Yes Yes Yes Yes Yes Adjusted R-squared 0.3780 0.3446 0.4032 0.3507 0.4066 0.4379 Observations 2162 2162 1676 1676 1676 1676
a Dummy variables for each state with at least five local and five non-local bookrunners are included (24 states).
35
Table 7 – Marginal Effects of Local Underwriters on Gross Spreads Data are for the sample of 2162 municipal bond issues with data availability. The dependent variable is the percent gross spread. Ln(Size) denotes the natural logarithm of the proceeds of the bond issue. Non-rated is a dummy variable that takes a value of one if the bond is not rated and zero otherwise. The variable denoted Bond Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc., and a value of zero if the bond is not rated. Note that assigning non-rated bonds a value of zero is arbitrary, and any value other than numerical ratings assigned to our bonds would produce identical results. Maturity reflects the life of a bond in years. Ln(Market Share) is the natural log of ((market share * 10) + 1), where market share is the percent proportion of all proceeds a bookrunner managed in a year. Local denotes a dummy variable that takes a value of 1 if the lead manager is “local” (i.e., has an office in that state) and zero otherwise. Robust standard errors (see White 1980) are used to calculate p-values, which appear in parentheses below the coefficient estimates.
Dependent = Gross Spread ( I ) ( II ) ( III ) Ln(Size) -0.1377 -0.1320 -0.1349 (0.000) (0.000) (0.000) Ln(Market Share) -0.0705 -0.0695 -0.0670 (0.000) (0.000) (0.000) Bond Rating 0.0671 0.0576 0.0681 (0.000) (0.000) (0.000) Non-rated dummy 0.7352 0.7074 0.7443 (0.000) (0.000) (0.000) Maturity 0.0004 -0.0001 0.0002 (0.735) (0.953) (0.877) Local Bookrunner 0.0927 0.0513 0.0893 (0.078) (0.366) (0.090) Local X Bond Rating -0.0488 -0.0353 -0.0445 (0.006) (0.051) (0.013) Local X Non-rated -0.2576 -0.2118 -0.2513 (0.033) (0.084) (0.037) Year dummies Yes Yes Yes State dummies No Yesa Yesb
Constant Yes Yes Yes Adjusted R-squared 0.3477 0.3641 0.3502 Observations 2162 2162 2162
a Dummy variables for each state with at least five local and five non-local bookrunners are included (24 states). b Dummy variable for New York only is included.
36
Table 8 – Marginal Effects of Local Underwriters on Yields Data are for the sample of 1676 municipal bond issues with data availability. The dependent variable is the percent yield to maturity for the bonds. Ln(Size) denotes the natural logarithm of the proceeds of the bond issue. Non-rated is a dummy variable that takes a value of one if the bond is not rated and zero otherwise. The variable denoted Bond Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc., and a value of zero if the bond is not rated. Note that assigning non-rated bonds a value of zero is arbitrary, and any value other than numerical ratings assigned to our bonds would produce identical results. Maturity reflects the life of a bond in years. Ln(Market Share) is the natural log of ((market share * 10) + 1), where market share is the percent proportion of all proceeds a bookrunner managed in a year. Local denotes a dummy variable that takes a value of 1 if the lead manager is “local” (i.e., has an office in that state) and zero otherwise. Yield is the yield to maturity on the bonds. Robust standard errors (see White 1980) are used to calculate p-values, which appear which appear in parentheses below the coefficient estimates.
Dependent = Yield ( I ) ( II ) ( III ) Ln(Size) -0.1103 -0.1075 -0.1104 (0.000) (0.000) (0.000) Ln(Market Share) 0.0221 0.0149 0.0220 (0.249) (0.455) (0.255) Bond Rating 0.1047 0.1440 0.1047 (0.000) (0.000) (0.000) Non-rated dummy 2.004 2.0746 2.003 (0.000) (0.000) (0.000) Maturity 0.0320 0.0309 0.0320 (0.000) (0.000) (0.000) Local Bookrunner 0.1667 0.1826 0.1668 (0.106) (0.109) (0.107) Local X Bond Rating -0.0712 -0.1021 -0.0713 (0.018) (0.002) (0.019) Local X Non-rated -0.7647 -0.7702 -0.7649 (0.001) (0.001) (0.001) Year dummies Yes Yes Yes State dummies No Yesa Yesb
Constant Yes Yes Yes Adjusted R-squared 0.3575 0.395 0.3575 Observations 1676 1676 1676
a Dummy variables for each state with at least five local and five non-local bookrunners are included (24 states). b Dummy variable for New York only is included.
37
Table 9 – Marginal Effects of Local Underwriters on Total Costs Data are for the sample of 1676 municipal bond issues with data availability. The dependent variable is Total Cost, the sum of the percent gross spread and the percent yield to maturity. Ln(Size) denotes the natural logarithm of the proceeds of the bond issue. Non-rated is a dummy variable that takes a value of one if the bond is not rated and zero otherwise. The variable denoted Bond Rating reflects a bond rating agency’s credit assessment – a numerical score of 1 denotes a AAA rating, 2 denotes a AA+ or Aa1 rating, etc., and a value of zero if the bond is not rated. Note that assigning non-rated bonds a value of zero is arbitrary, and any value other than numerical ratings assigned to our bonds would produce identical results. Maturity reflects the life of a bond in years. Ln(Market Share) is the natural log of ((market share * 10) + 1), where market share is the percent proportion of all proceeds a bookrunner managed in a year. Local denotes a dummy variable that takes a value of 1 if the lead manager is “local” (i.e., has an office in that state) and zero otherwise. Robust standard errors (see White 1980) are used to calculate p-values, which appear in parentheses below the coefficient estimates.
Dependent = Total Cost ( I ) ( II ) ( III ) Ln(Size) -0.2413 -0.2307 -0.2379 (0.000) (0.000) (0.000) Ln(Market Share) -0.0405 -0.0462 -0.0380 (0.085) (0.058) (0.110) Bond Rating 0.1942 0.2289 0.1948 (0.000) (0.000) (0.000) Non-rated dummy 2.9668 3.0108 2.9751 (0.000) (0.000) (0.000) Maturity 0.0363 0.0350 0.0359 (0.000) (0.000) (0.000) Local Bookrunner 0.2568 0.2616 0.2533 (0.035) (0.057) (0.039) Local X Bond Rating -0.1408 -0.1631 -0.1375 (0.000) (0.000) (0.000) Local X Non-rated -1.0767 -1.0386 -1.0724 (0.000) (0.000) (0.000) Year dummies Yes Yes Yes State dummies No Yesa Yesb
Constant Yes Yes Yes Adjusted R-squared 0.4149 0.4459 0.4152 Observations 1676 1676 1676
a Dummy variables for each state with at least five local and five non-local bookrunners are included (24 states). b Dummy variable for New York only is included.