Leading indicators of systemic banking crises: Finland in a panel of EU countries

• Author: Patrizio Lainà,Juho Nyholm,Peter Sarlin
• DOI: http://doi.org/10.1016/j.rfe.2014.12.002
• Published date: 01 January 2015
• Date: 01 January 2015

Leading indicators of systemic banking crises: Finland in a panel of

EU countries☆

Patrizio Lainà

a

,JuhoNyholm

b

, Peter Sarlin

c,d,e,

⁎

a

Departmentof Political and EconomicStudies, University of Helsinki,Finland

b

HECER,Finland

c

Centerof Excellence SAFE at Goethe UniversityFrankfurt, Germany

d

Departmentof Economics, HankenSchool of Economics, Helsinki,Finland

e

RiskLabFinland at Arcada Universityof Applied Sciences, Helsinki,Finland

abstractarticle info

Articlehistory:

Received 2 June 2014

Receivedin revised form 21 November2014

Accepted19 December 2014

Availableonline 5 January 2015

JEL classification:

E440

F300

G010

G150

C430

Keywords:

Leading indic ators

Macro-financialindicators

Bankingcrisis

Signalextraction

Logit analysis

This paper investigates leading indicators of systemic banking crises in a panel of 11 EU countries, with a particular

focus on Finland. We use quarterlydata from 1980Q1 to 2013Q2, in order to create a large number of macro-

financial indicators, as well as theirvarious transformations. We make use of univariate signal extractionand

multivariate logitanalysis to assess what factors lead the occurrence of a cri sis and with what horizon the indicators

lead a crisis. We findthat loans-to-deposits and house price growth are the best leading indicators. Growth rates and

trend deviations of loan stock variables also yield useful signals of impending crises. While the optimal lead horizon

is three years, indicators generally perform well with lead times ranging from one to four years. We also tap into

unique long time-series of the Finnish economy to perform historical explorations into macro-financial

vulnerabilities.

© 2014 ElsevierInc. All rights reserved.

1. Introduction

This paper investigates macro-financial factors as leadingindicators

of systemic banking crisesin Europe, and particularly reflects over the

case of the Finnisheconomy. Our definitionof a systemicbanking crisis

implies simultaneous failures in the banking sector that significantly

impairs the capital of the banki ng system as a whole, which mostly

results in large economic effects an d government intervention. The

investigated questions in this pa per relate to what factors lead the

occurrence of a crisis and with what horizon the indicators lead a crisis.

The implementation of macroprudentialpolicies, particularlywhen

being of discretionary nature, may exhibit ch allenges in tackling the

vulnerability of the financialsystem to procyclicality.To this end, recent

legislative initiatives providea basis for the use of policy instruments.

Basel III, the EU's legislative acts CRD and CRR IV and the Finnish

Ministry of Finance (2012) all propose the implementation of macro-

prudential tools at the national level. These tools are designed for curbing

booms in household, especially real estate, sectors through controlling the

growth rate of private loan stocks. They are also meant to strengthenthe

banking sector by enhancing its loss absorbing capacity and by reducing

default probabilities and losses given default. Other tools such as counter-

cyclical capitalbuffers are intended for restraining booms in the wider

economy. Although some discretion and judgment will inevitably be

required, tying macroprudential instrument triggering to risk indicators

via simple rules aids in overcomingresistance to countercyclical measures

during booms (e.g., Agur & Sharma, 2013). Thus, before couplingrisk

indicators with precise policy instruments, an essential question is to

investigate how and provid e means for assessing whether ris ks are

concentrated in a particular sector or whether they extend to a

number of sectors.This paper studies indicatorsfor rule-based guiding

Reviewof Financial Economics 24 (2015)18–35

☆Thispaper comes with a supplementaryinteractivedashboard: http://vis.risklab.fi/#/

Laina. The paper has received useful commments duringpresentations and discussions

from members of the Financial Stabi lity and Statistics Department and the Monetary

Policy and ResearchDepartment at the Bank of Finland, particularlyEsa Jokivuolle, Simo

Kalatie, Karlo Kauko, T apio Korhonen, Kimmo Kosk inen, Helinä Laakkonen, P eter

Palmroos, Hanna Putkuri , Katja Taipalus, Jouni Timon en, Mervi Toivanen, Jukka

Vauhkonen,Jouko Vilmunen andMatti Virén, as well as threeanonymous referees.

⁎Correspondingauthorat: Goethe University,SAFE, Grüneburgplatz1, 60323 Frankfurt

am Main, Germany.

E-mailaddress: peter@risklab.fi(P.Sarlin).

Contents listsavailable at ScienceDirect

Review of Financial Economics

journal homepage: www.elsevier.com/locate/rfe

http://dx.doi.org/10.1016/j.rfe.2014.12.002

1058-3300/©2014 Elsevier Inc. All rightsreserved.

of the activation of countercycli cal capital buffers, loan-to-v alue

caps and risk weights, rather than overall discretion and judgment in

decisions orthe effects of these macroprudentialtools.

Macroprudential instruments have an ultimate aim of preventing and

mitigating the occurrence of financial crises. Yet, one key problem is that

the implementation takes time. To launch the tools, policymakers need

to be aware of risks and vulnerabilities building up at an early stage

(e.g., CRD IV specifies a 12-month implementation period). By focusing

on identifying underlying vulnerabilities and risks, this paper investigates

indicatorsthat function as early enough signalsof an impending crisis.

Another problem is that the implementation of these tools is costly,

whereas implementation is sensibleonly if it will prevent a crisis. This

motivatesfurther research on leadingindicators of financial crises,and

their specificspecification,including transformations and timehorizon,

as well as a balance between false alarms and missed crises. Eventually,

one should still note that analytical tools for early identification of risks

provide only guiding support, whereas direct early-warning signals are

an output of internal investigations and thorough scrutiny.

The previous literature has consistently foundexcessive growth in

credit aggregates and asset prices to lead bankingcrises. For instance,

the signal extraction approach is u sed by Kaminsky and Reinhart

(1999) to study the connection between financial and currencycrises

and by Alessi and Detken (2011) to investig ate predictors of asset

price booms with costly real economy consequences. Likewise, Borio

and Lowe (2002) havefound unusually rapid expansions in creditand

asset prices,particularly deviationfrom their long-termtrend, as useful

leading indicators of wide-s pread financial distress. Despite a large

number of studies on crisis determinants, only a few of th em have a

pure focus on European economies. Accordingly, the traditionall iterature

focuses on leading indicators in emerging markets (e.g., Frankel & Rose,

1996; Kaminsky, Lizondo, & Reinhart, 1998) or both developed and

developing economies (e.g., Demirgüç-Kunt & Detragiache, 1998).

While some studiesonly include Europe as an aggregate (e.g.,Lo Duca

& Peltonen, 2013; Sarlin & Peltonen, 2013), those that include individual

European countries also include economiesfrom other continents. For in-

stance, Reinhart and Rogoff(2009),Alessi and Detken(2011),Babecký

et al. (2013) and Boissay, Collard, and Smets (2013) all focus on devel-

oped, mainly OECD, economies. Those studies thatfocus on distress in

Europe have a different scope and aim. For instance, Betz, Oprică,

Peltonen, and Peter (2014) and Männasoo and Mayes (2009) include

country-level indicators, but aim at predicting distress at the level of

banks in most European and Eastern European transition countries,

respectively, whereas Behn, Detken, Peltonen, and Schudel (2013)

perform an exercise similar to building an early-warning model, but use

it for setting countercyclical capital buffers. Accordingly, Behn et al.

(2013) focus mainly on the role of credit variables. Further, diverting

from assessing core Europe, they also include Central and Eastern

European transition or developing economies.

This paper assessesleading indicators of systemic banking crises in

Europe, with a particular focus on the Finnisheconomy. To enable and

support the analysis of Finland, we collect data on el even developed

European economies. Hence, rath er than taking a pan-European or

single-country perspective , we aim at collecting data on a possibly

homogeneousset of economies.While the sample economiesare partly

chosen based upon data availability, we deliberately exclude transition

economies,for which the trajectory of financialdevelopment has been

of different nature compared to the rest of Eu rope. The considered

macro-financial indicators cover a range of asset, credit an d macro

variables,following the previousliterature. For developedEU countries,

this enables us to studynot only patterns of pre-crisis, crisis and post-

crisis dynamics, but also to perform a structured analysis and ranking

of leading indicators of systemic banking crises and their optimal

signaling horizons . Beyond this, we also test the impac t of a number

of model specifications o n early-warning performance.

This paper contributes to the literature on banking crisis determinants

as follows. We find strongest evidence on loans-to-deposit and house

price growth as leading indicators of systemic banking crises. Loan stock

variables –mortgages,householdloansandprivateloans–also perform

well as leading indicators. The indicators show best performance with a

lead time of three years,but generally perform well with up to a four-

year lead time. This provides input to policymakers in control of

macroprudential tools, as indicators with a three-year lead time are

early enough to support macroprudential tools with long activation

times. Further, we also tap intounique long time-series of the Finnish

economy to perform historical explorations into macro-financial vulnera-

bilities. Beyond thecurrent global financial crisis, Finlandexperienced

three crises at the beginning of the 20th century, as well as a severe bank-

ing crisis in the 1990s, which was impacted by both a currency crisis and

the collapseof the Soviet Union. Using the estimateson panel data, we

correctlycall most of the Finnish crises since the beginning of the 20th

century. This paper also contributes to the technical derivation of early-

warning indicators and models. When assessing different model specifi-

cations, we find that differences between absolute and relative trend

deviations are only minor and that growth rates tend to be the most

prominent transformation. If trend deviations of ratios are used, we

propose to detrend GDP as a denominatorto support persistence with

respect to short-term variation in the real economy. Further,w e propose

the use of cumulative estimated probabilities of logit analysis over

the entire historical forecast horizon, in addition to only assessing

non-cumulative pr obabilities.

This paper is organized as follows. Section 2 provides a revie w of

indicators and method used in the literature, and presents those

used in this paper. Section 3 presents descriptive statistics through

measures of pre-crisis, crisis and post-crisis dynamics. In Section 4,

we present the signal extraction results and discuss the usefulness

of each indicator, whereafter we turn to an assessment of the indicators

by means of multivariate logit analysis. Before concluding, Section 5

presents long time series for the Finnish economy in light of our previous

findings. In addition, the indicators analyzed in this paper have been

included in a supplementary interactive dashboard: http://risklab.fi/

demo/lainaetal/.

2. Data and methods

This sectionbriefly reviewsprevious works on early warningindica-

tors and models, particularly with respect to useddata and estimation

methods. Next, we turn to a discussion of the col lected data for this

study and the methods that we use in this pa per to assess leading

indicators.

2.1. A review of indicators and methods

As abovenoted, a large number of studieshave assessed leadingand

early-warningindicators of bankingand financial crises overall.Herein,

we briefly review previous works on earl y warning indicators and

models, in order to supportthe subsequent choice of data and estima-

tion methods. We have reviewed a large number of recent works on

early-warningindicatorsand models, andassessed successfulindicators

in termsof broad categoriesof indicators.For instance, creditaggregates

include mortgages, household loans, corporate loans and total loa ns,

among others,whereas asset pricesinclude equityindices, house prices

and other property prices, as wellas their various transformations.

Table 1 shows the performance (or significance) of proposed

indicators in terms of broad indicator categories. It highlights the

significance of indicators rel ated to credit aggregates a nd asset prices,

but also the lack of a directconsensus in the used indicators and their

performance. This might be a consequence of variations in the analyzed

economies, types of crises and time spans. Thus, it highlights the impor-

tanceof a study focusing on a homogeneous set of economies, on a

specific type of crisis and on the rec ent experience of turmoil.

Startingfrom credit variables,Table 1 shows thatcredit-related indi-

catorshave been included in allstudies and most have alsofound one or

19P. Lainàet al./ Reviewof Financial Economics24 (2015) 18–35