Japanese Bank Productivity, 2007–2012: A Dynamic Network Approach

• Date: 01 October 2017
• Published date: 01 October 2017
• DOI: http://doi.org/10.1111/1468-0106.12199
• Author: Hirofumi Fukuyama,William L. Weber

JAPANESE BANK PRODUCTIVITY, 2007–2012: A

DYNAMIC NETWORK APPROACH

HIROFUMI FUKUYAMA*Fukuoka University

WILLIAM L. WEBER Southeast Missouri State University

Abstract. We estimate a dynamic network (DN) directional output distance function for 100 Japanese

banks operating during 2007–2012. Network production occurs in that deposits and other funds

raised are produced as intermediate products in stage 1 and those intermediate products are used

to generate a portfolio of assets in stage 2. The dynamic technology links production periods via

nonperforming loans (NPL) and carryover assets, which take the form of excess reserves. Carryover

assets expand the future production possibility set while NPL shrink future production possibilities.

We extend previous DN methods to measure the performance of three types of Japanese commer-

cial banks: city banks, regional banks and second regional banks. We test for and find differences

in the three bank technologies relative to a common technology. Such differences are likely due to

different institutional and regulatory structures. Unlike previous DN studies, we also allow for a

non-uniform abatement factor between previously-produced NPL and other inputs in stage 1 and

between performing loans and NPL in the current period. Measured productivity change is greater

when each bank faces their own group technology rather than the pooled technology consisting of

all bank types.

1. INTRODUCTION

The bursting of the bubble in Japanese real estate and stock market prices in

1991–1992 ushered in two decades of slow economic growth. While the

growth rate of Japan’s real GDP averaged 4.5% from 1970 to 1991, the aver-

age annual growth rate fell to only 0.9% during 1991–2012. Gross capital for-

mation fell from an annual average 4.4% during 1970–1991 to 0.9% from

1991 to 2012. The bursting of the bubble caused Japanese banks to be sad-

dled with high levels of nonperforming loans, and the banks were reluctant

to write off the loans and recapitalize. Various reforms have been adopted

in Japan to help resolve nonperforming loan problems and to recapitalize

banks (see Azad et al. 2014). However, even after these reforms, numerous

studies (Fukuyama and Weber (2015) provide a review) have found Japanese

banks to still be inefficient. Moreover, monetary policy tends to be more ef-

fective when banks are more efficient (Jonas and King 2008), so careful

modelling of the bank production process might yield insights that policy-

makers can use to enhance efficiency. Given the expected decline in Japan’s

population accompanied by population ageing, it is imperative that the

Japanese economy operate as efficiently as possible to mitigate the adverse

effects of demographic change (Yashiro 2001). In this paper, we employ the

*Address for correspondence: Hirofumi Fukuyama, Faculty of Commerce, Fukuoka University

8-19-1 Nanakuma, Jonan-Ku, Fukuoka 814-0180, Japan. E-mail: fukuyama@fukuoka-u.ac.jp.

Pacific Economic Review, 22: 4 (2017) pp. 649–676

doi: 10.1111/1468-0106.12199

© 2017 John Wiley & Sons Australia, Ltd

bs_bs_banner

dynamic network data envelopment analysis (DEA) method to examine the

performance of joint-stock commercial banks operating in Japan during fiscal

years 2006–2012. Japanese commercial banks are comprised of city banks, re-

gional banks and second regional banks. The three groups vary in average

bank size and have different regulatory and institutional structures. We test

for differences in performance as measured by a common technology versus

three different group technologies.

A series of papers by Fukuyama and Weber (2013, 2015, 2016) developed a

dynamic network (DN) DEA model to measure bank performance. These stud-

ies treated nonperforming loans (NPL) as an undesirable jointly produced by-

product of the loan production process. The models were estimated for Japanese

cooperative Shinkin banks and for joint-stock commercial banks. In addition,

the dynamic horizon was extended from three to many and risk-based capital

standards were included as constraints on the bank technology. The models

allowed productivity change to be decomposed into overall efficiency change,

which measures ‘catching up’to a given production frontier and technical

change, which measures the shift in the frontier.

When undesirable outputs are jointly produced along with desirable outputs

the researcher must specify an abatement factor when using DEA to estimate

the distance of a bank’s observed outputs to the production frontier. Fukuyama

and Weber (2013, 2015, 2016) assumed that all banks have a common abate-

ment factor. This specification has been implicit in most studies examining joint

production of desirable and undesirable outputs. However, Kuosmanen (2005)

showed that the assumption of common abatement across producers might

cause bias in performance measures.

In this paper we follow Kuosmanen (2005) and extend the DN model to

allow each bank to have a unique abatement factor for stage 1 and stage 2

of production. We measure bank performance for city banks, regional banks

and second regional banks, and compare the performance of these three bank

types with the estimates of cooperative Shinkin bank performance from Fuku-

yama and Weber (2016). NPL produced in one period become an undesirable

input in a subsequent period and constrain future production. The network

production technology consists of two stages. In the first stage banks use la-

bour, equity capital and physical capital to generate deposits. In the second

stage banks use deposits to produce a portfolio of interest-bearing assets com-

prised of securities investments and loans, with some of the loans becoming

nonperforming. In the subsequent period, NPL become an undesirable input

to the first stage of production. As more NPL are produced, the bank must

raise additional equity capital to continue generating deposits. In the second

stage of production the bank managers also choose whether to convert all de-

posits into loans and securities investments or carry over some deposits to the

subsequent period when economic conditions might make it easier to mini-

mize the jointly produced NPL. The dynamic aspect of our model develops

a performance measure that maximizes desirable outputs while simultaneously

minimizing undesirable outputs over multiple periods, rather than a single

period.

H. FUKUYAMA AND W. L. WEBER650

© 2017 John Wiley & Sons Australia, Ltd