Endogenous debt maturity and rollover risk

• DOI: http://doi.org/10.1111/fima.12250
• Author: Marco Macchiavelli,Emanuele Brancati
• Date: 01 March 2020
• Published date: 01 March 2020

DOI: 10.1111/fima.12250

ORIGINAL ARTICLE

Endogenous debt maturity and rollover risk

Emanuele Brancati1Marco Macchiavelli2

1LUISS Guido Carli, Rome, Italy

2FederalReserve Board, Washington, DC

Correspondence

MarcoMacchiavelli, Federal Reserve Board,

Washington,DC.

Email:marco.macchiavelli@frb.gov

Abstract

We empirically study the nature of rolloverrisk and show how banks

manage it. Having to roll over debt does not lead to higher default

risk per se. Only banks that lose significant access to new funding

while having to roll over debt displayhigher default risk. We identify

a factor that determines this buildup of risk: specifically, debt matu-

rity shortening (forcing debt to be more frequently rolled over) and

reduced access to new funding are both driven by marketpessimism

about banks’ future performance. We also provide evidence consis-

tent with dynamic coordination risk.

1INTRODUCTION

Rollover risk arises when preexisting debt obligations become due, and the resulting liquidity needs are potentially

unmet. In this paper,we empirically document the dynamic nature of rollover risk and show what determines the banks'

ability to manage such risk. Our findings suggest that, when investors lose faith in the future performance of a bank,

they become less willing to be locked in long-term obligations and thus force the bank to shorten its debt maturity.

At the same time, as the expected performance of a bank worsens, it becomes less likely to issue new debt. The more

frequent liquidity needs originating from the shorter debt maturity structure, together with the lower likelihood of

raising new funds, ultimately result in greater default risk.

Gorton, Metrick, and Xie (2015) argue that the buildup of fragility and the associated maturity shortening that

started in August 2007 eventuallyculminated with the default of Lehman.1We provide bank-level evidence consistent

with their narrative and propose an explanationfor those dynamics. Our findings regarding the link between maturity

shortening and default risk are also consistent with He and Milbradt (2015).

As a starting point, we show that the bank's need to roll over expiring debt per se has a negligible effect on default

risk; expiring debt has a detrimental effect only when the bank is expected to perform poorly in the future. Figure 1

shows this nonlinearity: the marginal effect of expiring debt on the credit default swap (CDS) spread—a measure of

default risk—is positive and significant only when marketexpectations about the bank's future return on assets (ROA)

are subdued. Motivated by the following analysis, we interpret such subdued expectations as an indication that the

bank is less able to raise new funds. Therefore, it is the combination of expiring debt and the impaired ability to raise

new funds that leads to greater default risk.

c

2018 Financial Management Association International

Financial Management. 2020;49:69–90. wileyonlinelibrary.com/journal/fima 69

70 BRANCATIAND MACCHIAVELLI

FIGURE 1 Expiring precrisis (Sep 2007)

Notes: Marginal effect of a unitary increase in Expiring/Totalassets on banks'CDS spread for different values of

𝔼t(ROAt+1Y)—from estimates in Table2. Expiring is the amount of debt expiring in month tissued before the crisis. The

black arrows highlight the regions of significance at the 5% level. Q1, Q2, and Q3 report the values of 𝔼t(ROAt+1Y)at

the first, second, and third quartiles

Second, we investigatethe effect of market expectations on debt issuance. We document that, indeed, worse expec-

tations about the bank's future ROA result in debt issuance being less likely,in lower amounts, and at shorter maturi-

ties. Short-term debt is usually seen as a cause of risk. We argue that it is instead a symptom—not a cause—of risk.

Specifically,we show that maturity endogenously shortens in response to worsening market expectations about bank

performance. Worsening expectations also lead to less debt issuance, which, compounded with the maturity shorten-

ing, leads to higher default risk.2

Next, we document new features of dynamic liquidity risk management: subdued market expectations hinder the

bank's ability not only to roll over currently maturing debt, but also to build precautionary liquidity buffers for future

rollover needs. Specifically, the need to roll over debt in the near future while markets expect the bank to perform

poorly limits its current ability to issue debt, which is consistent with dynamic coordination risk, as modeled in He and

Xiong (2012). Finally,debt issuance does not respond to market expectations when the bank under consideration has

no sizable amount of debt coming due within the next few months.

The empirical analysis presents severalidentification challenges, such as dealing with the endogeneity of both debt

maturity and market expectations, as well as isolating funding-supply from funding-demand factors. We mainly take

twoapproaches to overcome endogeneity issues. First, in the “backward-looking” regressions that document the effect

of expiringdebt on default risk, we identify exogenous variations in liquidity needs by using the amount of expiring debt

that was issued before the 2007 financial crisis, similar to the approaches in Almeida, Campello, Laranjeira, and Weis-

benner (2009) and Benmelech and Dvir (2013). This strategy, if anything, underestimates the true effect of expiring

debt on default risk, as discussed in more details later on.

Second, whenever we use marketexpectations as a regressor, we treat it as an endogenous variable, and we rely on

an instrumental variable approach that exploits past forecast errors to isolate exogenousvariation in current market

expectations, as in Brancati and Macchiavelli (2018). Since the instrumenting strategyfor market forecasts captures

the way investors (funding suppliers) revisetheir expectations in light of past forecast errors, we are not likely to cap-

ture factors related to the demand for funding.

Nonetheless, in order to more directly account for credit demand factors, we also control for Tobin'sQ,

a proxy for banks'investment opportunities, which should influence the demand for funding. Moreover, the

hypothesis that subdued expectations indicate lower funding demand is at odds with the result that expir-

ing debt increases default risk for banks expected to perform poorly: if these banks are simply not asking for

new funding they should not experience higher default risk. The latter is only consistent with lower funding

supply.