Financial crises and the asymmetric relation between returns on banks, risk factors, and other industry portfolio returns

• DOI: http://doi.org/10.1111/fire.12214
• Date: 01 February 2021
• Author: Kenneth Högholm,Seppo Pynnönen,Johan Knif,Gregory Koutmos
• Published date: 01 February 2021

DOI: 10.1111/fire.12214

ORIGINAL ARTICLE

Financial crises and the asymmetric relation

between returns on banks, risk factors, and other

industry portfolio returns

Kenneth Högholm1Johan Knif1Gregory Koutmos2

Seppo Pynnönen3

1Department of Finance and Statistics, Hanken

School of Economics, Helsinki, Finland

2Department of Finance, Charles F.Dolan School

of Business, Fairfield University,Fairfield,

Connecticut

3Department of Statistics, University of Vaasa,

Vaasa,Finland

Correspondence

GregoryKoutmos, Department of Finance,

CharlesF. Dolan School of Business, Fairfield

University,Fairfield, CT 06430.

Email:gkoutmos@fairfield.edu

Fundinginformation

HankenSupport Foundation

Abstract

We show that the relations between the returns on the banking

industry, risk factors, and other industries often are asymmetric.

Lagged banking industry returns seem to improve predictability

but the positive impact of a 1-month lag of the return on the

banking portfolio is much higher in the lower part of the return

distribution. However,after the Dodd-Frank Act in 2010, the cross-

autocorrelation with banks is changed and becomes negative in the

upper part of the distribution. Returns on banks also seem to lead

returns on five risk factors. This relation, however, is not robust

across the distribution.

KEYWORDS

banking industry, Dodd-FrankAct, financial crises

JEL CLASSIFICATIONS

G01, G10, G12, G21

1INTRODUCTION

This study focuses specifically on the role of the banking industry in the dynamic lead-lag relation to other industries.

The banking industry is sensitive to economic crises as well as to monetary policy changes and to changes in regula-

tions. Furthermore, as the banking industry by its liquidity providing nature is closely related to all other industries, it

is expected that the banking sector plays a major role in the dynamic interdependencies. On the one hand, the banks

are dependent on the performance of other industries. On the other hand, other industries are financially dependent

on the performance of the banking sector. Indeed, as Gorton and Winton (2002) show,banks account for nearly 25%

of external capital provided to firms. The purpose of the Dodd-FrankAct of July 2010 was to decrease the risk of the

effects of crises in the financial sector by enforcing transparency and accountability and place the regulation of the

financial industry in the hands of the government.

In times of crises, companies in different industries experience shortage of funding from internal untied equity and

the possibility for successful new share issues is limited. The dependence on the banking industry to provide funding is

Financial Review.2021;56:179–198. wileyonlinelibrary.com/journal/fire c

2019 The Eastern Finance Association 179

180 HÖGHOLM ET AL.

hence expected to be more pronounced during periods of crises. This would indicate a positive cross-autocorrelation.

During noncrises periods, the expectation is the opposite as companies haveuntied equity available and opportunities

for issuing new equity exists.As a consequence it might be harder for the banking industry to increase earnings on loan-

ing activities during noncrises periods. This would indicate a negative cross-autocorrelation with the banking industry.

The study follows and develops the approach of Högholm, Knif,and Koutmos (2014). Using monthly returns on 48

U.S. industry portfolios and six risk factors, the empirical results indicate a dynamic linkage between the returns on

the banking portfolio and other industry portfolios that appear to be asymmetric in two ways. First, a one-directional

causality relation running from the banking industry to several other industries is found but seldom the other way

around. Lagged banking industry returns seem to improve the predictability of returns for several industry portfolios.

Surprisingly,for many industry portfolios, returns on the banking industry portfolio can be regarded as exogenous and

Granger causes other industry returns. Second, in line with Chen (2007) and Baur,Dimpfl, and Jung (2012), the results

show asymmetry in the autocorrelation structure: positive in the lower part of the conditional return distribution and

negative in the upper part.

The empirical results further indicate that with the Dodd-Frank Act of August 2010, the dynamic link between

returns on the banking industry and other industry portfolios changes. Before the implementation, the Grangercausal-

ity from the banking industry was especially high. However, after the implementation of the Dodd-Frank Act, this

Granger causality almost disappears. The results for the conditional cross-autocorrelation with the banking industry

are similar. Before the implementation, this cross-autocorrelation is very high and positive, whereas it is lower and

becomes negative in the post-implementation subsample.

The results also suggest that the returns on the banking portfolio seem to improve predictability of the size, value,

momentum, and investmentfactors. This result is in line with the international empirical evidence in Baron, Verner, and

Xiong (2019) and also indicates that large bank equity declines tend to precede other crisis indicators. This suggests

thatbank losses are present at early stages of crises. However, the dynamic relationship found in our study is not robust

across the return distribution and can hence be regarded as endogenous.

Due to differences in exposures to risk factors, it is expected that in some market situations the returns on one

industry portfolio could lead or lag the returns on some other industry portfolios with different risk exposure char-

acteristics. However,the source of the empirically observed lead-lag effects is still a subject of debate. The literature

proposes five main explanations for this phenomenon: nonsynchronous trading (e.g., Lo & MacKinley, 1990a), time-

varying expected returns (e.g., Hameed, 1997), asymmetric information (e.g., Zebedee & Kasch-Haroutounian, 2009),

imperfect information (Chan, 1993), and slow diffusion of information (Lo & MacKinley,1990b; Merton, 1987).

Using a static model of multiple stocks where investors have access to limited information, Merton (1987) shows

that stocks with a smaller investor base are traded at greater discount due to limited risk sharing. Merton (1987) also

suggests that market segmentation and limited participation could be a reason for slowness of investorsin one market

to absorb information from another market. This argument is often called the gradual-information-diffusion hypoth-

esis. Hou (2007) finds that this slow diffusion of information is the leading cause of the lead-lag effect and that it is

predominantly an intra-industry phenomenon that is associated with firm size: big firms lead small firms. This explana-

tion is also provided byAyers and Freeman (2000) and thoroughly examined across industries by Cen, Chan, Dasgupta,

and Gao (2013). Anderson, Eom, Hahn, and Park (2013) find compelling evidencethat this partialprice adjustment is a

major source of the autocorrelation in returns.

Hong, Torous,and Valkanov (2007) also find that the gradual-information-diffusion hypothesis provides a key aux-

iliary explanationfor the lead-lag relations but might not bethe only one. Using monthly returns on 34 value-weighted

U.S. industry portfolios over the period from 1946 to 2001, they find that 14 industries are able to predict market

movements by 1 month. A few industries such as petroleum, metal, and financial could predict the market up to 2

months ahead. They also provide remarkably similar empirical evidence for the eight largest non-U.S.equity markets.

Their conclusion is that stock markets as a whole might react with a delay to fundamental information contained in

industry returns and that information diffuses only gradually across markets. On the other hand, Tse(2015) only doc-

uments one to seven industries out of 34 having predictiveability of stock market movements.