Municipal Bond Liquidity and Default Risk

• DOI: http://doi.org/10.1111/jofi.12511
• Author: MICHAEL SCHWERT
• Published date: 01 August 2017
• Date: 01 August 2017

THE JOURNAL OF FINANCE •VOL. LXXII, NO. 4 •AUGUST 2017

Municipal Bond Liquidity and Default Risk

MICHAEL SCHWERT∗

ABSTRACT

This paper examines the pricing of municipal bonds. I use three distinct, comple-

mentary approaches to decompose municipal bond spreads into default and liquidity

components, and find that default risk accounts for 74% to 84% of the average spread

after adjusting for tax-exempt status. The first approach estimates the liquidity com-

ponent using transaction data, the second measures the default component with credit

default swap data, and the third is a quasi-natural experiment that estimates changes

in default risk around pre-refunding events. The price of default risk is high given

the rare incidence of municipal default and implies a high risk premium.

SINCE THE FINANCIAL CRISIS, the poor condition of local government finances has

captured the attention of academics, legislators, and the popular press. Ris-

ing retirement and healthcare costs are straining state and local government

budgets. The post-crisis spate of defaults by cities and counties, including the

bankruptcy filing by Detroit in 2013, represents the most significant default

episode in the municipal bond market since the Great Depression.1Based on

recent events, it seems that municipal default risk is on the rise. However,

a longer view reveals that defaults are extremely rare in the municipal bond

market. The historical five-year cumulative default rate for all municipal bonds

is 0.08% from 1970 to 2014, and the corresponding rate for general obligation

bonds is 0.01% (Moody’s (2015)).

The academic literature on municipal bonds focuses instead on tax effects

and illiquidity as important issues in this market. Most municipal bonds are

exempt from federal and state taxes, so the pricing of their cash flows relative to

Treasuries has drawn interest (Green (1993), Ang, Bhansali, and Xing (2010),

Longstaff (2011)). Trading is done over-the-counter and the tax exemption at-

tracts retail investors as the primary clientele, resulting in high transaction

∗Fisher College of Business, The Ohio State University. This paper is based on a chapter of my

PhD dissertation at Stanford GSB. I thank Michael Roberts (the Editor), an anonymous Associate

Editor,three referees, Andrew Ang, Jonathan Berk, Antje Berndt, Darrell Duffie, Peter Feldhutter,

Marco Giacoletti, Lorenz Kueng, Robert Novy-Marx, Josh Rauh, Bill Schwert, Ilya Strebulaev,Ren´

e

Stulz, Edward Watts, Toni Whited, and seminar participants at the LBS Trans-Atlantic Doctoral

Conference and Ohio State for helpful comments and suggestions. I thank the Municipal Securities

Rulemaking Board for providing the municipal bond transaction data and Francis Longstaff for

sharing his marginal tax rate estimates. I have read the Journal of Finance’s disclosure policy and

have no conflicts of interest to disclose.

1For a review of the history of municipal defaults, see Spiotto, Acker, and Appleby (2012)and

Ang and Longstaff (2013).

DOI: 10.1111/jofi.12511

1683

1684 The Journal of Finance R

costs due to search frictions and dealer market power (Green, Hollifield, and

Schurhoff (2007b), Green, Li, and Schurhoff (2010), Schultz (2012), Li and

Schurhoff (2014)). Novy-Marx and Rauh (2011a,2011b) draw attention to the

dire condition of public pensions, but little work has been done on the pricing of

default risk in the municipal bond market. In this paper I address the following

question: what are the relative contributions of default risk and liquidity to the

pricing of municipal bonds?

After adjusting for the tax exemption, I find that default risk accounts for

74% to 84% of the average municipal bond spread over the period 1998 to

2015. The average default spread of 101 bps and the extremely low historical

default rate imply a default risk premium that is an order of magnitude larger

than estimates of the risk premium in the corporate bond market (Berndt

et al. (2005)). Investors in general obligation bonds receive between 78 bps and

126 bps of taxable-equivalent compensation per 1 bp of expected default loss.

To arrive at the above conclusions, I use three distinct approaches to de-

composing municipal bond spreads. The first approach follows from Dick-

Nielsen, Feldhutter, and Lando (2012) and estimates the liquidity component

of the spread using the transaction data.2The second approach follows from

Longstaff, Mithal, and Neis (2005) and measures the default component with

credit default swap (CDS) spreads. The third approach is a quasi-natural ex-

periment that studies the reaction of spreads to pre-refunding, an event that

renders the bond risk-free until its call date (Fischer (1983), Chalmers (1998)).

These complementary approaches lead to similar findings and mitigate con-

cerns that any shortcomings associated with each approach are driving the

results.

In this paper, I show that default risk plays an outsized role relative to the

observed rate of default due to a high risk premium. I find that the role of

liquidity is not as large as one might infer from the literature on transaction

costs in the municipal bond market. Intuitively,the liquidity spread depends on

the expected cost of trading and the expected trading intensity, or need to sell

the bond. The typical investor in this market is a buy-and-hold retail investor,

so trading intensity is low and the liquidity discount is small.3

These findings have policy implications that merit further exploration. My

results imply that if policy makers wish to reduce the cost of financing state

and local government investment, then efforts to reduce transaction costs and

improve liquidity will not have a significant impact on borrowing costs. In con-

trast, reducing the default risk premium to the level of the risk premium in the

corporate bond market would dramatically reduce municipal bond yields. One

potential source of this risk premium is the tax exemption, which could operate

through two channels. First, the tax exemption is unattractive to institutional

2Marlowe (2013) also uses the Dick-Nielsen, Feldhutter,and Lando (2012) approach to estimate

the liquidity spreads of individual municipal bonds, but he focuses on the impact of the monoline

insurer downgrade on liquidity spreads.

3At the end of 2015, $1.6 trillion of the $3.7 trillion bonds outstanding were owned directly

by households, with another $975 billion held by mutual funds on behalf of households (Federal

Reserve Statistical Release Z.1, Table L.212).

Municipal Bond Liquidity and Default Risk 1685

investors with low marginal tax rates, which limits the supply of risk-bearing

capital in the market. Second, most states exempt only local bonds from state

income tax, leading to home bias and imperfect risk-sharing (Babina et al.

(2015)).

This paper relates to the literature on the financial condition of state and

local governments. Novy-Marx and Rauh (2011a,2011b) show which states

and local governments are in the worst financial condition due to pension un-

derfunding, but there is little academic work on the bond market’s assessment

of state and local government default risk. I find that Detroit (pre-bankruptcy),

Chicago, and Illinois have significantly higher default spreads than their peers,

consistent with Novy-Marx and Rauh’s (2011a,2011b) results on pension un-

derfunding. However, even the least creditworthy states and large cities trade

at spreads similar to marginally investment grade corporate issuers, with Illi-

nois priced like a BBB-rated corporate issuer and Detroit and Chicago priced

close to BB-rated corporates.

Closely related to my work is a contemporaneous paper by Ang, Bhansali,

and Xing (2014), who reach very different conclusions. They find that liquidity

accounts for 74% of the average municipal bond spread. Their decomposition re-

lies on computing synthetic risk-free municipal bond yields using an estimated

zero-coupon curve from pre-refunded bonds. Importantly, they use separate

pools of bonds that have been and have not been pre-refunded, with no controls

for omitted issuer or bond-specific factors. My quasi-natural experiment with

pre-refunded bonds, which focuses on within-bond changes in spreads and con-

trols for bond and time-specific unobservables, shows that the pre-refunding

event reduces the bond spread by at least 41%, assuming no offsetting effect

from reduced liquidity.4

In another paper that uses pre-refunded bonds to control for noncredit fac-

tors, Novy-Marx and Rauh (2012) study the effect of state pension investment

losses on state bond yields. In contrast to Ang, Bhansali, and Xing (2014), these

authors compare bonds from the same issuer that have and have not been pre-

refunded, rather than using a marketwide pool of pre-refunded bonds. The

results in Novy-Marx and Rauh (2012) are consistent with my finding that

default risk is the main driver of municipal bond spreads. They find that, be-

tween September 30 and December 31, 2008, the average non-pre-refunded

bond spread increased by 116 bps, while the average pre-refunded spread in-

creased by only 27 bps. Their triple-difference regression estimates imply that

48% of the differential spread increase for the average issuer can be attributed

to the deterioration in state pension funding.

Wang, Wu, and Zhang (2008) also find that liquidity plays a secondary role

to default risk in municipal bond pricing. They decompose municipal bond

4Ang, Bhansali, and Xing (2014) show that liquidity is significantly reduced after pre-refunding

and impose a liquidity correction of 15 bps to 50 bps to compare pre-refunded and non-pre-refunded

bond spreads. The average spread prior to pre-refunding is 211 bps, so using the liquidity adjust-

ment from Ang, Bhansali, and Xing (2014) would increase the estimated contribution of default

risk in my quasi-natural experiment to between 48% and 65%.