The Private Production of Safe Assets

• Published date: 01 April 2021
• Author: GUILLAUME VUILLEMEY,CHRISTOPHE PÉRIGNON,MARCIN KACPERCZYK
• Date: 01 April 2021
• DOI: http://doi.org/10.1111/jofi.12997

THE JOURNAL OF FINANCE •VOL. LXXVI, NO. 2 •APRIL 2021

The Private Production of Safe Assets

MARCIN KACPERCZYK, CHRISTOPHE PÉRIGNON,

and GUILLAUME VUILLEMEY

ABSTRACT

Using high-frequency, granular panel data on short-term debt securities issued in

Europe, we study the existence, empirical boundaries, and fragility of private assets’

safety. We show that only securities with the shortest maturities, issued by banks

(certif‌icates of deposit, or CDs), benef‌it from a safety premium. The supply of such

CDs responds positively to excess safety demand. During periods of stress, this re-

lation vanishes for all issuers of private securities, even though their aggregate vol-

umes do not collapse. Other dimensions of heterogeneity, including issuers’ balance

sheets or their domicile countries’ f‌iscal capacity, are less relevant for private safety.

ASAFE ASSET IS AN ASSET THAT IS IMMUNE to adverse selection concerns and

thus can be valued without expensive and prolonged analysis (Gorton (2017)).

Such an asset, for example, a Treasury, has money-like attributes and can

serve as a store of value (Nagel (2016)). Traditionally, government debt has

been the dominant type of safe asset (Krishnamurthy and Vissing-Jorgensen

(2012)). Recent decades, however, have been characterized by their apparent

shortage. One potential explanation is that due to booming savings in emerg-

ing economies, the demand for public safe assets increased faster than its sup-

ply, pushing real interest rates to historically low levels, and creating global

imbalances (Caballero (2006), Caballero, Farhi, and Gourinchas (2016,2017),

Caballero and Farhi (2018)).

Kacperczyk is at Imperial College London. Pérignon and Vuillemey are at HEC Paris.We thank

two anonymous referees; Stefan Nagel (the Editor); Gary Gorton; Robin Greenwood; Raj Iyer;

Arvind Krishnamurthy; Stefan Lewellen; Patrick McCabe; Alan Moreira; Enrico Perotti; Alexi

Savov; Jeremy Stein; Amir Suf‌i; Adi Sunderam; Anjan Thakor; David Thesmar; Emily Williams;

Marius Zoican; and seminar participants at the 2018 EFA and 2019 AFA meetings, ACPR, the

Federal Reserve Board, FederalReserve/Maryland short-term funding markets conference, Goethe

University Frankfurt, HEC Paris, Imperial College, NBER Summer Institute, University of Ams-

terdam, University of Mannheim, and Washington University for useful comments; and Laurent

Clerc and the Banque de France for providing data on certif‌icates of deposit and commercial paper.

We thank Robin Greenwood for sharing code to estimate T-bill safety premia. Pérignon and Vuille-

mey thank the ACPR Chair on Regulation and Systemic Risk and the French National Research

Agency (F-STAR ANR-17-CE26-0007-01 and ECODEC ANR-11-LABX-0047) for supporting their

research. Kacperczyk acknowledges research support from European Research Council Consolida-

tor Grant 682156. We have read The Journal of Finance disclosure policy and haveno conf‌licts of

interest to disclose.

Correspondence: Guillaume Vuillemey, HEC Paris, Department of Finance, 1 rue de la Libéra-

tion, Jouy-en-Josas, 78351 France; e-mail: vuillemey@hec.fr.

DOI: 10.1111/jof‌i.12997

© 2020 the American Finance Association

495

496 The Journal of Finance®

To cater to investors’ safety demand, private f‌inancial institutions can issue

assets with safety attributes similar to those of Treasuries (Bernanke et al.

(2011), Gennaioli, Shleifer, and Vishny (2013), Sunderam (2015)). While the

private production of safe assets has been documented in aggregate samples

(Sunderam (2015)), our understanding of the microlevel characteristics of pri-

vate safety is limited. First, the same issuer can issue debt securities with

different maturities. In theory, assets with shorter maturities are more likely

to be perceived as safe. We do not know, however, whether safety concentrates

among very short-term assets only, or whether it extends to longer maturities.

Second, the ability of the private sector to issue safe assets has been described

as fragile and ineff‌icient (Stein (2012)), but no prior study has investigated

the ability of the private sector to issue safe assets during both stable periods

and stress periods. Relatedly, we do not know whether all issuers stop catering

to safety demand during stress periods or whether time-series patterns in the

data arise from selection mechanisms. Third, it is not clear whether all f‌inan-

cial institutions, or only a subset of them, cater to safety demand. Similarly,

among the securities issued by f‌inancial institutions, we do not know which

ones are considered truly safe. Finally, given that prior empirical evidence

is based largely on a single country, the United States, one cannot conclude

whether the production of private safe assets is a local or global phenomenon.

Addressing these questions is important to distinguish between theories of safe

assets and to identify possible fragilities in the f‌inancial system.

In this paper, we shed light on these questions using high-frequency data

on 1.36 million euro-denominated certif‌icates of deposit (CDs) issued by com-

mercial banks from several countries. Our main data set covers most of the

European euro-denominated short-term private debt and Treasury bill (T-bill)

markets between 2008 and 2014. All of these assets are reasonable safe asset

candidates, as they have short maturities (up to one year) and are issued by

borrowers with high credit quality.

To summarize our empirical f‌indings, among the f‌ive sources of heterogene-

ity we consider, we f‌ind that three are of f‌irst-order relevance: asset maturity,

market conditions (stress vs. not), and the type of issuer (bank vs. others). The

two other dimensions, namely, the issuer’s balance sheet characteristics and

the country in which it is located, play only a minor role.

In our f‌irst test, we study which maturities of assets issued by private insti-

tutions are considered safe by investors. We show that only short-term CDs,

with maturities less than or equal to one week, carry a safety premium:their

interest rates are below the risk-free rate, by 8 bps per year on average (as com-

pared to 15 bps for one-month T-bills). The premium, which captures the non-

pecuniary benef‌its associated with holding a safe asset (Krishnamurthy and

Vissing-Jorgensen (2012), Sunderam (2015)), is economically large, as the av-

erage risk-free rate over our sample period is 40 bps. The fact that only short-

term CDs are considered safe—while longer-term CDs carry a risk premium—

is consistent with the view that assets with shorter maturities are less

sensitive to the arrival of new information. We additionally show that the term

structure of CD interest rates exhibits a change in slope precisely at the point

497

at which safety disappears, consistent with the view that investors clearly dis-

tinguish between safe and nonsafe assets (Gorton and Pennacchi (1990)).

In our second set of tests, we show that the supply of private assets responds

positively to measures of excess safety demand. The main identif‌ication chal-

lenge we face is to isolate the role of demand and supply forces. In contrast with

previous work that tends to focus on aggregate data, our granular data allow

us to include a variety of f‌ixed effects. Our baseline specif‌ication is inspired by

Khwaja and Mian (2008). Specif‌ically, we estimate the response of CD supply

when T-bill issuance changes after including issuer-date f‌ixed effects. The f‌ixed

effects fully capture total CD demand at the issuer-date level. The only source

of variation that remains is supply forces across various CD maturities. We

f‌ind that when the aggregate supply of T-bills in low-risk countries decreases

by 1%, the quantity of short-term CDs, def‌ined as CDs with maturity less than

one week, increases by about 4%. Given that these contracts are precisely the

ones that benef‌it from signif‌icant safety premia, this result is consistent with

the shortage of publicly issued safe assets creating demand for privately issued

assets with similar safety attributes. Notably, this substitution effect does not

hold for longer-term CDs, which do not benef‌it from a safety premium.

Third, we study the fragility of private safety production by exploiting both

the time-series and the cross-sectional dimensions of our data. Theoretically,

if investors seek information-insensitive assets, then spikes in uncertainty or

negative public news can imply that some assets become information-sensitive.

If this is the case, adverse selection reappears and these assets are no longer

appealing to safety-seeking investors (Dang, Gorton, and Holmström (2012),

Moreira and Savov (2017)). Consistent with this idea, we show that while T-

bills enjoy a safety premium at all times, an average CD loses its safety status

during periods of market stress, irrespective of its maturity. The main decline

in the private safety premium occurs in 2008, at the height of the f‌inancial cri-

sis. In terms of quantities issued, we f‌ind that the substitution effect between

the issuance of public and private safe assets disappears when market stress

is high. During low-stress periods, in contrast, the magnitude of the substitu-

tion effect is higher: a 1% decrease in T-bill issuance is associated with a 7%

increase in short-term CD issuance. Notably, while the substitution between

public and private assets drops to zero in high-VIX period, CD market volume

does not collapse. Why the aggregate CD market did not collapse is beyond

the scope of this paper. It could be due to policy interventions, the regulatory

framework, or other motives. Whatever the reason, the time-series variation

in our results is in line with a change in investors’ perceptions: CDs continue

to be issued, but investors no longer perceive them as substitutes for T-bills in

periods of stress.

Fourth, we use cross-sectional issuer characteristics to distinguish between

theories of safe asset fragility. If investors have public access to issuer-level

information, they may still consider issuers with high asset quality as safe

(Moreira and Savov (2017)). Alternatively, if they only consider information

about the asset class as a whole, then all issuers may cease being perceived

as safe regardless of their balance sheet characteristics (Dang, Gorton, and