Switching from Incurred to Expected Loan Loss Provisioning: Early Evidence

• Published date: 01 June 2021
• Author: GERMÁN LÓPEZ‐ESPINOSA,GAIZKA ORMAZABAL,YUKI SAKASAI
• Date: 01 June 2021
• DOI: http://doi.org/10.1111/1475-679X.12354

DOI: 10.1111/1475-679X.12354

Journal of Accounting Research

Vol. 59 No. 3 June 2021

Printed in U.S.A.

Switching from Incurred to

Expected Loan Loss Provisioning:

Early Evidence

GERMÁN LÓPEZ-ESPINOSA,∗GAIZKA ORMAZABAL,†

AND YUKI SAKASAI‡

Received 16 May 2020; accepted 15 January 2021

ABSTRACT

This paper provides early evidence on the effect of global regulation mandat-

ing a switch from loan loss provisioning (LLP) based on incurred credit losses

(ICLs) to LLP based on expected credit losses (ECLs). Using a sample of sys-

temically important banks from 74 countries, we find that ECL provisions are

more predictive of future bank risk than ICL provisions. Corroborating that

∗School of Economics, IESE Business School, Universidad de Navarra; †IESE Business

School, CEPR, and ECGI; ‡IESE Business School

Accepted by Rodrigo Verdi.We are grateful to the editor, the associate editor,and an anony-

mous referee. We also thank Eduardo Ávila Zaragoza (Head of Global Supervisors Relations

BBVA Group), Pietro Bonetti, Thomas Bourveau, Miguel Duro, Juan Carlos García Céspedes

(Director of Capital Measurement in BBVA Group), Gonzalo Gasós (Head of Banking Su-

pervision of the European Banking Federation), Urooj Kahn, Mohamed Janahi (discussant),

Fernando Peñalva, and Mario Quagliariello (Director of the Economic Analysis and Statistics

Department of the European Banking Authority), Srini Rangan, Rahul Vashishtha (discus-

sant), and participants of the 2nd IIMB (Indian Instituted of Management Bangalore) Ac-

counting Research Conference, and seminar participants at Warwick Business School. We also

thank Joakim Alderborn, Alessandro Maimone, Christopher Nance, and Claudia Serra for

excellent research assistance. Gaizka Ormazabal acknowledges financial contributions from

the Spanish Ministry of Economics, Industry and Competitiveness, grant ECO2015-63711-P,

and from the “Cátedra de Dirección de Instituciones Financieras y Gobierno Corporativo del

Grupo Santander.”Germán López-Espinosa acknowledges financial support from the Span-

ish Ministry of Economics, Industry and Competitiveness, Grant PID2019-105227RB-I00. An

online appendix to this paper can be downloaded at http://research.chicagobooth.edu/arc/

journal-of-accounting-research/online-supplements.

757

© 2021 The Chookaszian Accounting Research Center at the University of Chicago Booth School of

Business

758 g. lópez-espinosa, g. ormazabal, and y. sakasai

the switch to ECL provisioning results in more information to assess bank risk,

we also observe that the announcement of a larger first-time impact of the ac-

counting change elicits lower stock returns and higher changes in credit de-

fault swap spreads. Critically,these patterns are most pronounced when credit

conditions deteriorate. Additional analyses show that the higher information

content of the ECL model stems from the provisions for nondefaulted loans,

which did not exist under ICL. Our study contributes to the debate on the

effect of the ECL model on procyclicality, an especially pressing issue in the

context of the current pandemic.

JEL codes: G21, G38, M41

Keywords: expected credit losses; loan loss provision; bank accounting;

IFRS 9; CECL

1. Introduction

The 2007–2008 financial crisis generated the perception that the account-

ing for loan losses in force at that time exacerbated procyclicality by provid-

ing “too little, too late” provisioning (e.g., Gaston and Song [2014]; Bischof,

Laux, and Leuz [2019]).1As a consequence of this view, the traditional ap-

proach to measure impairment allowances based on incurred credit losses

(ICLs) is being gradually replaced by the alternative approach of building

provisions based on expected credit losses (ECLs). The new approach has

been embraced by IASB in the new standard IFRS 9, which is currently

being implemented around the world (the standard was enforced for the

first time in the financial statements corresponding to fiscal year 2018).

In the United States, FASB has also adopted an ECL approach in the Ac-

counting Standards Update (ASU) 2016-13, Financial Instruments—Credit

Losses (Topic 326).2This paper provides early evidence on the effect of

implementing the ECL approach on the information content of loan loss

provisions.

Although the switch from ICL to ECL provisioning is widely perceived

by regulators and market participants as a major change for banking

entities, there is still scant evidence on the economic consequences of

1More specifically, the argument runs as follows. At the beginning of the crisis, the provi-

sioned amounts (loan loss allowances) under the incurred loss model were not sufficient to

face the downturn. As such, banks had to create additional provisions, which resulted in lower

earnings (provisions are expensed in the profit and loss account) just at the time when the

entities could not recognize the interest income from nonperforming loans. The resulting

decrease in income reduced the banks’ regulatory capital during the economic downturn,

leading to severe funding and capital pressures, and eventually forcing many banks to delever-

age. A number of banks did so by selling noncore business lines and reducing the amount of

risk-weighted assets (RWAs). The reduction of RWA was done to a large extent by restricting

new lending, thereby generating negative externalities.

2Although both IFRS 9 and ASU 2016-13 are based on the ECL approach (i.e., provisioning

based on expected—rather than incurred—credit losses), the U.S. standard generally refers

to the ECL approach using the term “Current Expected Credit Loss” (CECL).

switching from incurred to expected loan loss 759

this rule change.3These consequences are likely first order, as loan loss

provisions (LLP) are the largest accruals in banks’ financial statements and

drive much of the variability in banks’ accruals (Beatty and Liao [2014]).

The relevance of LLP reporting is also highlighted by the post-crisis debate

on whether to include LLP in regulatory capital.4

To the extent that they are based on expected rather than on realized

losses, provisions estimated under the ECL approach should be more infor-

mative about banks’ risk than those under the ICL approach. Yet, the im-

plementation of ECL provisioning faces two challenges. First, embedding

forward-looking risk assessments in the measurement of the value of assets

is not an easy task; the ECL approach entails a nontrivial data collection

effort and requires expertise in the implementation of sophisticated risk

modeling. Second, compared to the ICL approach, the ECL approach re-

quires a higher degree of managerial judgment and discretion in the mod-

eling process (ESRB [2017]). As such, the ECL approach could result in

less informative LLP reporting due to measurement error and/or oppor-

tunism.

We study the informational effects of implementing the ECL approach

to loan loss provisioning by exploiting the recent worldwide implementa-

tion of IFRS 9. Using a comprehensive sample of systemically important

(SI) banks from around the world, we first collect information on the im-

plementation dates in each country where IFRS is mandatory. The control

group in our tests includes not only banks in countries where IFRS is not

mandatory, but also banks in countries where IFRS 9 has yet to be imple-

mented.

Our evidence is structured in three groups of analyses. In the first set of

tests, we find that LLP amounts reported under IFRS 9 are more predic-

tive of subsequent risk (measured as forward stock return volatility, average

absolute value of returns, tail risk, and subsequent changes in the amount

of nonperforming loans). This result is stronger in countries experiencing

3See, for example, the statements about ECL by the European Banking Authority

(EBA, 2018). This view is also shared by other prominent institutions. For example,

the American Banks Association described this regulatory change as “the most sweep-

ing change to bank accounting ever” (see https://www.aba.com/advocacy/our-issues/

cecl-implementation-challenges). Also, in a report issued in 2016, the Global Public

Policy Committee (which represents the six largest accounting networks), states: “For

many banks, the adoption of expected credit loss accounting will be the most momen-

tous accounting change they have experienced, even more significant than their tran-

sition to IFRS” (see https://www2.deloitte.com/bd/en/pages/financial-services/articles/

2016-gppc-the-implementation-of-ifrs9-impairment-requirements-by-banks.html).

4Advocates argued that letting loan loss reserves count would encourage banks to report

adequate and timely reserves, whereas opponents were concerned about a decline in the qual-

ity of regulatory capital. See Ng and Roychowdhury [2014] for a detailed discussion on this

debate. Based on their evidence, these authors conclude that the ability to add back loan

loss reserves to regulatory capital creates the illusion of nondeclining financial health, which

allows banks to avoid taking the actions necessary to reduce the risk of failure.