INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely afiect reproduction.
In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright material had to be removed, a note will indicate the deletioiL
Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the back of the book.
Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.
UMI University Microfilms International A Bell & Howell Information Company 300 North Zeeb Road. Ann Arbor, Ml 48106-1346 USA 313/761-4700 800/521-0600
Order Number 9412036
A study of the banking production efficiency of the primary agricultural cooperatives in Korea: A cost function approach
Park, Seong Jae, Ph.D.
The Ohio State University, 1993
UMI 300 N. Zeeb Rd. Ann Arbor, MI 48106
A STUDY OF THE BANKING PRODUCTION EFFICIENCY OF THE PRIMARY AGRICULTURAL COOPERATIVES IN KOREA: A COST FUNCTION APPROACH
DISSERTATION
Presented in Partial Fulfillment of the Requirement for the Degree Doctor of Philosophy in the Graduate School of the Ohio State University
By
Seong Jae Park, B.A., M.A.
*****
The Ohio State University 1993
Dissertation Committee: Approved by
Richard L. Meyer, Ph.D. Adviser D. Lynn Foster, Ph.D. Department of Agricultural Economics and Rural Sociology Gary D. Schnitkey, Ph.D. To My Family ACKNOWLEDGMENTS
I am indebted to quite a number of persons for the completion of this study. I express my sincere appreciation to Dr. Richard L. Meyer, my principal advisor, for his guidance, support, and encouragement for this study as well as for my entire life in the
Ohio State University. Thanks also go to the other members of my committee, Drs. D.
Linn Foster and Gary D. Schnitkey, for their comments and suggestions that gave me insights into this research.
I extend my gratitude to the staffs of the primary agricultural cooperatives in
Choongnam and Kyounggi provinces, of the National Agricultural Cooperatives
Federation (NACF), and of the Korea Rural Economics Institute (KREI), for providing the data. Without their cooperation, this study would not have been conducted. I acknowledge with appreciation the financial aid received from the Agency for
International Development during my study in the Ohio State University.
I wish to express my gratitude for the encouragement and support of friends, fellow students, and faculty in the Department of Agricultural Economics and Rural
Sociology. I would like to thank, for proof reading and comments, Renee Drury and Diane
Hite. Thanks also go to Chang Kyu Seung, Jae Min Park, Sang Wook Kim, and Dr. Seung
II Na, for editing. I especially thank Dr. Dong Kyoon Kim for his great help, guidance, and computer assistance through the years in Columbus, Ohio. All the errors in this document.
111 however, are entirely mine and none of the above named persons are responsible for the errors.
Appreciation is extended to Mrs. Barbara Lee for her kindness and
encouragement, and to Dr. Cameron Thraen for econometric guidance. I wish to express
my sincere thanks to Dr. Chong Hyuk Suh and to Dr. Yang Boo Choi who encouraged me to study in this university and suggested the idea for this study.
In completion of this dissertation, special thanks go to my brothers, sisters, and
parents-in law for their encouragement and love during my stay in the U.S.. I deeply apologize to my family for my selfishness and express sincere appreciation and love to my wife, Seon Ok, and children. You Kyoung, So Jeong, and Jin Soo, for their patience,
sacrifice, and love during the years in Columbus.
IV VITA
August 25, 1953 ...... Bom in Cheonnam, Korea
1975 - 1977 ...... Military Service, Korea
1980...... B A , Agricultural Economics, Seoul
National University
1982...... M A , Agricultural Economics, Seoul
National University
1982 - 1988 ...... Researcher, Korea Rural Economics
Institute (KREI)
1988 - 1989...... Research Fellow (KREI)
1989 - 1993 ...... Research Associate, Department of
Agricultural Economics and Rural
Sociology, The Ohio State University
FIELD OF STUDY
Major Fields: Agricultural Economics and Rural Sociology
Rural Finance and Production Economics TABLE OF CONTENTS
DEDICATION...... ii
ACKNOWLEDGMENTS...... ni
VITA ...... V
LIST OF TABLES...... ix
LIST OF FIGURES...... xi
CHAPTER PAGE
I. INTRODUCTION...... 1
1.1 The Research Problem...... 4
1.2 The Objectives of Study ...... 7
1.3 The Hierarchy of Administrative Districts in Korea ...... 8
II A REVIEW OF BANKING PRODUCTION EFFICIENCY STUDIES:
THE COST FUNCTION APPROACH...... 11
2.1 Some Advantages of the Cost Function Approach ...... 13
2 .2 Studies of Banking Output Efficiency ...... 15
2.3 Functional Forms ...... 21
2.4 Production Frontier Approaches to Banking Production 23
2.5 The Objectives of Cooperatives ...... 28
2.6 Summary ...... 30
VI III KOREAN FINANCIAL MARKETS AND
AGRICULTURAL COOPERATIVES...... 32
3 .1 Economic Growth and Financial Development of Korea ...... 33
3.2 Formal Financial Market Structure and Banking by P A C ...... 46
3 .3 Agricultural Cooperatives and Banking Operation ...... 63
3.4 Summary...... 75
IV DATA AND DESCRIPTIVE ANALYSIS OF
BANKING COST STRUCTURE...... 77
4.1 Data and Sampling ...... 78
4.2 Operating Characteristics of Sampled PACs ...... 83
4.3 Descriptive Analysis of Banking Cost Structure ...... 95
4.4 Summaiy...... 106
V METHODS AND MODEL SPECIFICATION...... 108
5.1 Production Process and Cost Function ...... 109
5.2 Frontier Cost Function and Estimation Method ...... 114
5.3 Multiproduct Cost Concepts and Output Efficiency Measures ..119
5.4 Model Specification for the Frontier Cost Function ...... 136
5.5 Estimation Method of the Frontier Cost Function ...... 148
5.6 Variable Selection for the Analysis of
Input Inefficiency Sources ...... 151
VI EMPIRICAL RESULTS...... 155
6.1 The Mean Value of the Variables and Product Mix
by Deposit Size Class ...... 156
6.2 Model Specification T ests...... 160
v ii 6.3 Estimation Results of the Frontier Cost Function ...... 162
6.4 Tests for the Characteristics of Production Technology 166
6.5 Output Inefficiency ...... 168
6.6 Input Efficiency ...... 178
6.7 Summary ...... 182
VII CONCLUSION...... 185
7.1 Summary of Main Results ...... 188
7.2 Contributions and Limitations of the Study ...... 191
7.3 Policy Implications ...... 194
LIST OF REFERENCES...... 198
VIII LIST OF TABLES
TABLE PAGE
3.1 The Growth of Real Income ...... 34
3.2 Macroeconomic Indices of the Korean Economy 1980 - 1990 ...... 35
3.3 Monetary Deepening of Korean Economy ...... 43
3.4 Interest Rate Level per Year by Type of Financial Institution 1992 ...... 51
3.5 Changes in the Market Share of Total Deposits
by Type of Financial Institutions (Deposit Markets) ...... 53
3.6 Composition of an Average Farm Household's Financial Assets, 1980, 1985,
and 1990 ...... 55
3.7 Borrowing Sources for an Average Farm Household
by Type of Financial Institution ...... 57
3.8 The Growth of Assets, Equity, Members, and Employees
of an Average PAC, 1980 to 1991 ...... 68
3.9 Labor Allocation within a PAC (National Average) ...... 69
3.10 The Changes in Business Structure of PACs (National Average) ...... 71
3.11 The Growth of Business Gross Income of PACs (National Average) 72
3.12 The Break Even Point Business Size per Employee in 1991 ...... 74
4.1 Sample Size and Regional Distribution ...... 82
4.2 Selected Characteristics of PACs (Equity, Employees, and Market
Environment Variables), Sample and National Average, 1991 ...... 85
ix 4.3 Balance Sheet: Average Assets of Sample PAC ...... 88
4.4 Balance Sheet: Average Value of Liability and Equity per PA C ...... 89
4.5 The Structure of PAC Gross Business Income; Sample and National Averages. 93
4.6 Differences in Average PAC Gross Income between Rural and Urban ...... 94
4.7 An Example of Labor Allocation to Cooperative Activities by the Class of
Employee: One of the Sampled PACs for 1991 ...... 98
4.8 Banking Cost Structure of Average PAC ...... 99
5.1 The Case 1 of EPSCE with No RSCE of Large Firm ...... 131
5.2 The Case 2 of Diseconomies of EPSCE with No RSCE of Large Firm ...... 132
5.3 The Case 3 of EPSCE with Scale Economies of Large Firm ...... 132
6.1 The Mean Value of Cost Function Variables : Aggregate Sample of Total
Observations ...... 156
6.2 The Variable Means of Average Rural and Urban PACs ...... 157
6.3 The Distribution of Sample PACs by Deposit Size Class by Region ...... 158
6.4 Product Mix of Deposit Size Class ...... 159
6.5 Test for Heteroscedasticity ...... 161
6.6 Estimation Result of the Frontier Cost Function ...... 163
6.7 Joint Test Results of Production Technology ...... 167
6.8 Ray Scale Economies of Banking ...... 169
6.9 Expansion Path Cost Efficiency to Next High C lass ...... 174
6.10 Expansion Path Cost Efficiency to Scale Efficient C lass ...... 175
6.11 Expansion Path Subadditivity (EPSUB) by Deposit Size Class ...... 177
6.12 Distribution of Cost Inefficiency Groups^ ...... 178
6.13 The Sources of Input Inefficiency ...... 181
X LIST OF FIGURES
FIGURE PAGE
1.1 The Hierarchy of Administrative District in Korea ...... 9
3.1 Financial Institutions in Korea (As of the end of June 1990) ...... 47
3.2 Trends in Nominal Interest Rates in the Rural Financial M arkets ...... 59
3.3 Trends in Real Interest Rates in the Rural Financial M arkets ...... 59
3.4 Organizational Structure of the Agricultural Cooperatives ...... 64
4.1 Distribution of Sample Region ...... 81
4.2 Total Banking Cost Distribution ...... 101
4.3 Average Costs Measured by Banking Assets ...... 101
4.4 Average Interest Costs of Borrowings and Deposits ...... 102
4.5 Average Interest Costs as Sum of Borrowings and Deposits ...... 102
4.6 The Borrowing Cost Share of Total Bank Costs ...... 103
4.7 Average Bank Costs Excluding Borrowing Costs ...... 104
5.1 Cross Sections of Multiproduct Cost Surface ...... 122
5.2 Multiproduct Cost Surface and Output Efficiency Measures ...... 123
5.3 Expansion Path Cost Efficiency ...... 134
XI CHAPTER I
INTRODUCTION
This study proposes to analyze the efficiency o f banking production of the primary agricultural cooperatives (PACs) in Korea by using the concepts of scale economies, product mix economies, and input efficiencies related to the overuse of inputs. The PACs are the most important rural financial institutions in the country, and they also provide a diverse set of economic function such as marketing of farm products, purchasing farm inputs and consumer goods, operating cooperative insurance, etc. The banking business of
PACs is recognized as a key business for their viability since it is a major income source of
PACs. Korea, one of the countries with the strongest controlled banking system, has proceeded with a broad and gradual financial liberalization program since the early 1980s but competition in the market is restrictive and the management of banks is still controlled by the government (Chung). The completion of financial liberalization is expected to bring
about a much different market environment in terms of competitiveness in markets and the degree of managerial autonomy granted to the financial institutions.
A question about banking production efficiency is closely related to the context o f
regulation in financial markets. The regulation of the banking industry of most countries
has been justified either because of the existence of externalities and asymmetric
information problems in financial markets or for purposes of resource allocation to stimulate economic development. The externality problem is supported by the historical experiences of bank failures and their domino effects on financial markets such as occurred in the Great Depression in the 1930s. The asymmetric information problem is related to the unequal market power existing between consumers and financial institutions that are specialized in dealing with information. These two reasons have been used to justify banking regulations in developed countries, while in developing countries such as Korea the emphasis on resource allocation for economic development has been used to justify regulations. Since there exists market failures in financial markets, the best policy to maximize social welfare requires regulations of the banking industry as a solution to the constrained-Pareto efficiency problem (Greenwald and Stiglitz).
A question naturally arises, however, about whether or not the regulations actually result in the constrained-Pareto efficiency, since there is no guarantee that the regulatory authority will have more exact information about the financial markets than financial intermediaries operating in the markets (Timothy). If the regulator as social planner does not have better information, regulations will likely bring about inefficiency in resource allocation in spite of the best efforts of the regulation. If the industry is inefficient, existing regulations cannot be justified as the best remedy for the solution of the constrained-
Pareto efficiency problem.
It is generally recognized that banking is over-regulated even in developed countries such as the U.S. and the EEC. As McKinnon (1973) and Shaw argued, the financial markets have been repressed in most developing countries. Furthermore, the
Ohio State University research program has extensively explored the problems of market regulations in developing countries, and showed the comparative efficiency of nonregulated financial activities (Adams, Graham, and Von Pischke; Von Pischke, Adams, and Donald; Adams and Fitchett). The empirical evidence suggests that over-regulation leads to the protection of inefficient banks at the expense of the public interest. Moreover, it is argued that the costs of a somewhat higher rate of bank failure, resulting from fewer restrictions on bank activities, may be offset by the enhanced efficiency of banking production (Kidwell and Peterson). This recognition naturally induces financial liberalization.
But financial liberalization does not lead the complete abolishment of regulations for the industry but the removal of the unnecessary interventions. There is still a conflict between the goals of market stability, a high degree of bank safety, and a high degree of competitiveness. In other words, the regulator's task is to find the proper balance between bank safety and economic stability on the one hand and an efficient banking system on the other (Kidwell and Peterson). The problem at hand is how to do it.
One of the ways to examine the efficiency of banking industry is to test whether or not the industry provides, with a given technology, financial services with minimum costs; that is, whether or not the banks as multiproduct firms exhaust scale and scope economies without committing an overuse of inputs. Since the path-breaking works of Benston
(1965, 1972) and Bell and Murphy about the economies of scale in the U.S. commercial banking industry, extensive research has been devoted to answer this question. Adar,
Agmon, and Orgler noted the effects of output mix and jointness on banking costs, which led to exploring the existence of scope economies in the banking firms (e.g., Gilligan,
Smirlock and Marshall; Lawrence and Shay; LeCompte and Smith (1985), etc.). Suspicion about the regulator's ability has led to the recent literature concerning the possible existence of the systemic overuse of inputs, owing to a lack of competitiveness in the banking industry.
The change in emphasis about the production efficiency of banking was possible thanks to the development of multiple product (multiproduct) firm theory (Baumol, Panzar, and Willig), of duality theory about the production process and the cost function or the profit function (Diewert; McFadden), of flexible functional forms such as the translog function (Christensen, Jorgenson, and Lau), and of the frontier function concepts
(Pareil; Fare, Grosskopf, and Lovell; Aigner, Lovell, and Schmidt). To explore the banking production efficiency o f PACs as multiple products firms, this study will use the frontier cost function depending on the duality theory.
1.1 The Research Problem
The primary agricultural cooperatives in Korea are multipurpose cooperatives which engage in a number o f profit and non-profit activities. Among them, the banking business has contributed to the growth of rural financial markets and agricultural production through successful mobilization of savings in both rural and urban areas, supplying agricultural Amds, and allocating funds geographically through a well organized network of cooperatives. Moreover, profits from the banking services of cooperatives have contributed significantly to their viability.
However, if the regulation of financial markets were used to support the viability of small financial institutions such as the PACs, this performance might reflect the government support. It is true that the PACs have been protected from competition with banks by allowing the PACs to hold high interest rates for both deposits and loans and a large margin between the deposit and lending rates of interest. This support might have enabled even inefficient cooperatives to be viable. On the other hand, the market area for one PAC is restricted to the administrative district where the PAC is located so it cannot increase the scale of production even if it is under significant economies of scale. This possible existence o f the PAC banking production inefficiency provides the motivation to analyze the production process. Furthermore, this study also is motivated by two recent issues related to the PACs in Korea. The first issue is whether the agricultural cooperatives will be viable in a new financial market environment which is expected to occur in the near future. The changes in financial markets will result from; i) financial liberalization, ii) the opening of capital markets to foreign investors, iii) the need to change agricultural policies as alternatives for post Uruguay Round of GATT, and iv) a
decreasing rural population that implies a decrease in the base of the PACs (Park; Kim,
B.J.; Sul). The changes in market environment will accelerate market competition, and
eliminate government support for the agricultural cooperatives. If the PACs do not reach
minimum costs, these changes in market environment may be severe threats for their
viability. The PACs may face the problem of scale economies, as small banks in the U.S.
experienced in 1980s'. Rural PACs may suffer from small market size and a decrease in
agricultural policy loans supplied by the government which incur handling costs but
substantial revenues for the PACs. It may be very hard for an urban PAC to compete with
large financial institutions. Therefore, for the PAC as well as all financial institutions, it is
very important task to find the most efficient scale and product mix.
The second issue is whether rural financial markets which are geographically
fragmented are too small to be competitive. It is argued that since rural financial markets
are too small to obtain scale economies, the livestock cooperatives and the postal savings
should be consolidated with the agricultural cooperatives (Kim, B.J.; NACF, 1989). In
* The interest rate ceilings (regulation Q) for the U.S. banks were removed in 1980. This deregulation increased interest rates and a severe recession of economy in the early 1980s pressed the operation of small banks, so that bank failures increased sharply. The bank failures from 1983 to 1988 were 817, while the failures from 1943 to 1982 were 262 (Kidwell and Peterson: 254). order to examine the nature o f a natural monopoly situation in rural financial markets, it needs to test the cost subadditivity of rural financial institutions.
These issues are important not only for the viability of the banking sector, but also for the viability of the PACs. Since the gross revenue o f the PACs is highly dependent on their banking business, a failure of the baion may lead the PACs to become economically unvia the competitive viability of the banking secto
A financial institution is viable if i serves an ever-increasing number of customers, nking business due to market competitA providing new financial services and actively improd transaction costs (Meyer, 1988). Toble. Therefore, this study will focus on insure viability,vironment should be based on appropriate informa production technology r.of financial institutions. The ronometric theories for multiproduct firms provides king work bt is self-sustaining, covers its costs, andy Baumol, Panzar, and Willig introduced the concept of ray scale economies, economies of viable financial institution is dynamic in scope, and cost subadditivity, as major properties of multiproduct production. Thves its efficiency as reflected in reducee measurements of expansion path scale economies and expansion path subadditivit the strategies for a changing market eny developed by Berger,
Hanweck and Humphrey (1986, 1987) provide a more consistent tion about the market structure and theway to analyze economies of scale and economies of scope in multiproduct production procesecent development of economic and ecses. Production frontier models, which are based on the notion of optimizing behavior of firmsinsights for this purpose. The path brea, provide useful insight into the existence of input inefficiencies such as technical and allocative inefficiencies.
This study will adopt these methods to explore the characteristics of banking production of PACs. The study is expected to provide important policy implications for both the agricultural cooperatives and the regulatory authorities in establishing new strategies.
1.2 The Objectives of the Study
A banking firm is competitively viable when it produces outputs at minimum cost with given technology. Depending on this proposition, this study will evaluate the viability o f PAC banking by using the measured production efficiency. In measuring the efficiency, scale and product mix economies will be analyzed as representing output efficiency, and the degree of input overuse as input inefficiency will be measured by using a frontier cost function.
Since PAC banking produces multiple outputs, measuring the degree of production efficiency depends on specific cross-sections of a multiproduct cost surface. To measure economies of scale, two assumptions about the cross-sections of the multiproduct cost surface will be adopted: first, to increase production scale with a constant product mix; second, to increase the production scale with allowing changes in product mix. For the first case, ray scale economies by Baumol et al. will be used. For the second case, the concept of expansion path scale economies (EPSCE) developed by Berger, Hanweck and
Humphrey (1986) will be used. But the measure of EPSCE is not consistent in the general sense. Therefore, this study develops an alternative measure. Measuring product mix economies evaluates whether or not dividing a large PAC into two small PACs will result in an increase in costs. The specific objectives are as follows:
1) to measure the degree of scale economies of PAC banking production for the two
situations mentioned above,
2) to measure the degree of product mix economies of PAC banking production. 3) to measure the degree of input overuse in PAC banking production,
4) to test the systemic relationship between the measured input inefficiency from 3) and
possible sources of input inefficiency, if input inefficiency exists,
5) to develop an alternative measure of the expansion path scale economies for 1),
6) to evaluate the competitive viability of PAC banking with the measurements of
production efficiency as the results from 1) to 3), and
7) to develop appropriate policy implications from the analysis of PAC banking production
efficiency.
1.3 The Hierarchy of Administrative Districts in Korea
The agricultural cooperatives are established along administrative districts. Thus, to understand the system of agricultural cooperatives, it is necessary to understand the hierarchy of administrative districts. For the sake of convenience, the administrative district hierarchy is explained in this section.
As shown in Figure 1.1, the highest administrative districts are special
(municipality) cities and Do as the provincial level. There are six special cities whose population exceeds one million persons. The administrative districts o f special cities are divided into several Kus and a Ku is divided into many Dongs. On the other hand, a Do is composed of cities and Kuns as county level, and a Kun is composed of few Ups and about 10 Myouns. A community is officially recognized as a city when its population is over 50 thousand persons. If a community that belonged a Kun becomes a city, then it is separated from Kun; that is, a city is independent from the Kun. A community is officially recognized as an Up when its population exceeds 20 thousands, but it still remains under Nation
Do ^ Special City (population: over one million; 6)
City Kun ipulation: over 50,000: (136)
Myoun Doni (population: over 20,000; 178 (1.257)
Lie or Dong' (34,786) i
Village (67,396)
Source: Major Statistics o f Agriculture, Forestry, and Fisheries the 1992, Ministry of Agriculture,
Forestry, and Fisheries (MAFF), 1992.
Note: The number in parenthesis is the administration district number in 1991.
Figure 1.1 The Hierarchy of Administrative District in Korea
the administration district of the Kun. The Up functions as a center town of rural areas and its characteristics are close to urban in terms of dominant share of service industry or high population density. Myoun and Up are the lowest level of administrative districts of the local government. At the lower level of Myoun or Up, there are Lies or Dongs that consist of about two or three villages. 10
A primary agricultural cooperative (PAC) is authorized in a Myoun or Up, while the National Agricultural Cooperatives Federation (NACF) as the federation of agricultural cooperatives has its local offices corresponding to administrative districts such as provincial level offices or county level offices; that is, a Seoul city office. Do offices, city or Kun offices. The offices of the NACF have their branches in the administrative areas. In classifying a region into rural and urban, the Up is defined as either urban or rural by the purpose of classification. When functional approaches to rural areas are used, the Up as a center town o f rural area is defined as rural. According to this view, the Up is just a big rural community. This definition is consistent with establishing regional policies such as rural development projects. This definition of rural areas can be named as the broad concept of rural. But when the socioeconomic characteristics are emphasized, the
Up is defined as urban since the characteristics of Up are close to those of urban in terms o f industrial structure with a high share of non-agricultural sector, high population density, urban life style, etc. Thus, this concept defines only the Myoun as a rural area. Therefore, this definition can be named as the narrow concept of rural. CHAPTER II
A REVIEW OF BANKING PRODUCTION EFFICIENCY STUDIES: THE COST FUNCTION APPROACH
The purpose of this chapter is to provide a general base for building an analytical framework and selecting appropriate research methods for analyzing the banking production of Korean primary agricultural cooperatives (PACs). This study will use a cost function approach that has many advantages in exploring the production technology, compared to the production function or the profit function, as will be discussed in the following section. Therefore, the discussion focuses on the literature o f banking production efficiency using cost function analytical approaches.
Traditional macroeconomic analysis regarded banks as passive conduits of monetary policy (Benston and Smith). In contrast, the microeconomic approach, which views a bank as a firm producing specialized commodities under market competition, has focused on the behavior of the bank as a profît maximizer. The cost function approach to banking behavior is contrasted with the portfolio approaches that completely ignore the use of real resources to produce banking services and focus on analyzing how to allocate produced outputs according to the rates of returns or risks.
The portfolio or risk management approach cannot provide information about scale and scope economies of banking production that is important to regulatory authorities
11 12 as well as the bank decision makers (Benston and Smith; Sealey and Lindley). With duality theory and the methodology developed to analyze multiproduct firms, cost function analysis has been extensively used to explore the properties of banking production technology such as scale and scope economies.
The main concern of empirical studies about banking production has been the measurement of scale economies (1950s - 80s), scope economies (1980s), and input inefficiencies in banking production (since the late 1980s). These studies are closely related to concerns about the impact of market regulation on chartering (scale economies), creating commodities (scope economies), and deregulation of banking.
Scale and scope economies have been extensively explored to determine the impact of regulations by comparing unit with branch banking in the U.S.. The analysis of production inefficiency theory using the notion of the cost frontier opened up another field of banking studies. It is argued that production inefficiencies arising from the underutilization of inputs due to market regulation may result in even larger losses of resources than occur with suboptimal scale and scope economies. The results of input inefficiency studies presented striking evidence of inefficiency in U.S. banking, which contrasted with the evidence of a dearth of scale and scope economies in the industry.
For that reason, the studies of input inefficiency suggest that deregulation of financial markets will threaten the viability of many banks, not because o f scale and scope inefficiencies, but because of under utilization of inputs (Berger and Humphrey, 1991).
In the first section of this chapter, some advantages of the cost function approach are presented to give a justification for the use of the cost function approach in this study. In the second section, the discussion will focus on the analysis of output efficiency in banking production. The section includes a review of scale economy studies in the earlier stage, the methodological approaches taken, and the empirical 13 results and their implications. In the third section, a discussion about functional forms will be conducted to justify the choice of an appropriate form for this study. The fourth section will present the concept of the frontier function, and the alternative analytical methods including the stochastic frontier function, the thick frontier model, and the linear programming method. The fifth section focuses on the objective function of cooperatives and credit unions which is one of the controversial issues in the application of these neoclassical tools to cooperatives. Since cooperatives are organizations of interest groups whose interests are not necessarily homogeneous unlike the usual proprietary firms, it may be questionable to apply a simple profit maximization or cost minimization assumption. The sixth section summarizes the main points o f the discussion.
2.1 Some Advantages of the Cost Function Approach
McFadden pointed out that the practical advantage of the cost function lies in its computationally simple relation to the cost minimizing input demand functions. The partial derivatives of the cost function with respect to input prices yield the corresponding input demand functions, and the sum of the value of the input demand equals costs. In econometric applications, the cost function avoids the difficulty of deriving demand systems from production possibilities, while at the same time insuring consistency with the hypothesis of competitive cost minimization.
Binswanger summarized the advantages of a cost function in an empirical sense as compared with a production function: 14
(1) It is not necessary to impose the condition of homogeneity of degree one on the
production process to arrive at estimation equations. Cost functions are homogeneous
in input prices regardless of the homogeneity properties of the production function.
(2) Production functions use factor quantities as independent variables which, at the firm
or industry level, are not proper exogenous variables. It is more likely that
entrepreneurs will make decisions on factor use according to exogenous prices which
makes the factor levels endogenous decision variables.
(3) In deriving estimates of the elasticity of substitution or of factor demand in the many
factor case, the matrix of the estimates of the production coefficients has to be
inverted, which exaggerates the estimation error.
(4) There is less multicollinearity among input prices than among input variables.
Comparing the cost function approaches with profit function approaches, the advantages are as follows:
(1) A minimum cost plan will always exist as long as the input requirement set V(y) is
closed, but the maximal profit point may not (Varian)k
(2) The profit function approach requires the output prices as independent variables, but
accounting problems associated with income variables are generally more severe in
such cases than when using expense data (Srinivasan).
(3) The assumption of profit maximization appears to be less plausible for nationalized
banking systems than cost minimization (Srinivasan).
’ If tlie input requirement set V(y) is closed, costs are certainly bounded from below since tlie minimum cost the firm can produce y is zero. Hence, a minimum cost production plan will always exist. However, if the production technology exhibits constant or increasing returns to scale, the profits should be unbounded and thus the maximal profit point will not exist, i.e., the profit function may not exist (Varian, 1984:26 • 27). However, tliis problem may be not substantial in the short run because the firm with fixed factors likely to exhibit decreasing returns to scale with respect to the fixed variable and the short run restricted profit function for the ease almost always makes sense. 15
(4) The profit function approach assumes that firms are price takers in both the input and
output markets. In the case of the banking industry, the assumption of output price
taker may not be the case because the banks located in a specific area may be a
monopoly in supplying loans (Klein)
2.2 Studies of Banking Output EfRciencv
2.2.1 Scale Economy Studies in the Early Stages of Banking Production Efficiency
Early studies of banking production focused on measuring economies of scale^.
Since the theory and analytical methods for a multiproduct production process were not developed, the model for a single output production process was applied. A single output measurement was either total earning assets or total deposits (AlhadefF; Horvitz;
Schweiger and McGee; Gramley), or a weighted sum of earning assets (Kalish and
Gilbert). Some studies (e.g. Greenbaum) used total revenues as output under the assumption that total revenues of banks reflect the relative valuation of bank services of different categories of bank assets (Gilbert). A single output metric regards the outputs of banks as homogeneous commodities, so that it cannot capture the multiproduct nature of commercial bank outputs. In addition, it is likely to result in aggregation bias.
To avoid the nonhomogeneous nature of banking products. Bell and Murphy,
Benston (1965,1972), Longbrake and Haslem assumed that each of the bank's services is separately produced through technically independent production processes.
The Cobb-Douglas functional form was used for different categories of bank outputs.
2 See Gilbert's extensive literature survey and Clark's brief summary of this field. 16
However, their approaches not only cannot provide an overall picture of banking production but also exclude the problem of jointness that probably exists in banking production(Adar, Agmon, and Orgler). The study by Benston, Hanweck, and
Humphrey employed two indices of overall bank outputs to obtain an aggregated single output as the measurement of bank output. They used the number of accounts so that they could measure overall economies of scale without using the complicated notions of multiproduct firm theory. However, the nature of jointness was not considered. Employing indices as an overall output metric regards the output composite as constant results in significant bias in estimating economies of scale if significant cost complementarities exist (Mester). Another problem is that the Cobb-
Douglas function does not have desirable properties consistent with theory. It has restrictive properties, such as constant returns to scale, and a downward sloping cost curve asymptotically approaching a constant. U-shaped cost curves are also a priori excluded and cost complementarities are not allowed. Thus, these properties would likely lead to the conclusion that scale economies are limited to smaller banks, because the measured economies of scale are likely to be biased due to the nonmeasured impact of scope economies (Benston, Hanweck, and Humphrey). In addition, since the production technology is usually not known, use of an exact function like the Cobb-
Douglas requires strong assumptions, a priori, about the technology.
The problems that resulted from the separability assumption and the restrictive functional form were solved by the multiproduct firm theory developed by Baumol,
Panzar, and Willig and the application of the translog functional form. Multiproduct
firm theory permits the exploration o f jointness in production, while the translog cost function eliminates the need to make assumptions about the unknown technology since
it is a second order approximation to any kind of cost function. 17
2.2.2 Definition of Bank Outpnt and Analytical Approaches
The most controversial issue in the literature concerns banking output. A central issue is whether the production of deposits is a bank output or not. Most studies in the
1970s regard deposits as inputs, since the production of deposit services is considered as an intermediate payment and safekeeping services to depositors who provide funds as inputs for loans (Sealey and Lindley). This view produced a reduced-form model: the production o f deposit services is an intermediate input which will be netted out, but the production of loan services remains as output (Humphrey).
The production of deposit services, however, directly accounts for half of utilized real resources, labor and capital. Including deposit services as output reflects a value added criterion for defining outputs and represents the production structure of banking.
This view produces a structural model of a multiproduct banking firm, in which the production o f deposit services is not netted out and remains as a kind of final good
(Humphrey).
These views about banking output divide the analytical approaches into the production, the intermediation, and the user cost approach (Berger and Humphrey,
1992). The production approach views banks as producing specialized financial commodities, such as earning asset accounts of various sizes by processing deposits and making loans. In the process, banks incur capital and labor costs (Benston and
Smith). This view focuses on producing transaction contracts. Operating costs are specified as the dependent variable in the cost function and the number of accounts is used as the output metric. The number of deposit accounts is regarded as an input, while that of loan accounts is regarded as output. However, though this approach may more precisely measure operational efficiency in producing loans, it ignores the greater 18 part of banking costs, so it will likely fail to provide an overall picture of banking production. It cannot capture differences between produced deposits and purchased funds since both of them are netted out in the reduced form model. Furthermore, it may lead to contradictions when banks are merged (Berger and Humphrey, 1992)^.
In contrast, the intermediation approach stresses that the function o f financial institutions is to intermediate funds. The intermediation approach treats deposit services as outputs as well as loan services as outputs. It measures output in the dollar value of loans and deposits rather than the number of accounts. Total costs including both operating costs and interest expenses are treated as dependent variables; thus, the input characteristics of deposits are also specified through their associated interest expenses in total costs (Humphrey). This approach is more appropriate in analyzing the viability of a bank (Berger, Hanweck, and Humphrey, 1987)
The user cost approach defines banking output according to whether or not a financial product contributes to bank net revenues (Hancock; Fixler and Zieschang). If the financial returns on an asset exceed the opportunity costs, the instrument is output; otherwise, the instrument is an input. Though the criterion seems to be consistent with production theory, it is not easy to obtain financial revenues for each service and their marginal opportunity costs, which makes it difficult to use this method in practice
(Berger and Humphrey, 1992).
Despite the relative advantages of each approach, the choice is also dependent on data availability. It is not easy to obtain data or account numbers or the financial
^ Berger and Humphrey use the following example; Assume bank A is specialized to produce deposits and sells the mobilized funds to bank B that lends the funds to final borrowers. The production approach calculates the total outputs of the two banks by adding the produced deposits of bank A and the loans of bank B. But when the two banks are merged, the produced deposits are no longer considered output, even tliough there is no change in banking production. 19 returns on an asset relative to its opportunity costs. Therefore, the intermediation approach has been preferred by many researchers.
2.2.3 Research Findings of Scale and Scope Economies
Scale Economies : The findings of economies of scale reported in the literature are closely related to the definition of outputs (or inputs) used, the treatment of bank output as a single or multiple outputs, the functional forms used in the analyses, and market structure and regulatory conditions. Most studies in the 1980s concluded that scale economies are exhausted at relatively low levels of output
(Benston, Hanweck, and Humphrey; Murray and White, 1983; Clark; Kim, H.Y.), while Gilligan, Smirlock and Marshall showed no support for the existence of overall economies of scale. However, studies after the mid-1980s in the U.S. that used the data from large bank samples found evidence of scale economies (Hunter and Timme;
Lawrence and Shay; Noulas, Ray, and Miller; Cropper). These results suggest that the effect of recent regulatory and technological changes may have given larger banks a cost advantage over that which existed in previous years (Cropper).
Economies of Scope : Baumol, Panzar and Willig emphasize that the presence of economies of scope creates an incentive for specialized firms to merge and become multiproduct firms. Scope economies come fi'om the existence o f non-allocable inputs.
In the case of the banking industry, the main sources of cost complementarities are the spreading of fixed costs, information economies, risk reduction, and consumer cost economies (Berger, Hanweck, and Humphrey, 1987). However, even allocable inputs 20 may cause jointness in the short run as a special case (Shumway, Pope, and Nash,
1984, 1988; Leathers).
Gilligan, Smirlock and Marshall, and Lawrence and Shay reported findings of significant economies o f scope using U.S commercial bank data. Murray and White
(1983) and Kim, H.Y. reported scope economies in Canadian credit unions, while
Cuevas (1984, 1988) reported the same in the case of Honduras banks, and so did
Srinivasan for banks in Bangladesh. However, Benston et al., LeCompte and Smith
(1985), and Berger, Hanweck and Humphrey (1986, 1987) found no significant economies of scope in U.S. commercial banks or Savings and Loans. LeCompte and
Smith (1990) suggested that the cost complementarities may depend on the market structure; they found cost complementarities in the data for 1978, but did not find them in the data for 1983 for the same sample of Savings and Loan associations.
Many problems have been identified, however, in measuring economies of scope. First, many studies used the pairwise cost complementarities [i.e., 32C(Q)/9Qi8
Qj ^ 0], typically at only the mean value of the data, to examine the existence of scope economies, but they should be measured between all product pairs and all points smaller than the mean (Mester; Berger, Hanweck and Humphrey, 1986,1987).
Moreover, a translog function cannot meet these complementarity conditions (Berger,
Hanweck and Humphrey, 1986,1987)'’. Second, the costs at zero output are not defined with the translog function. Many studies adopt an arbitrary small value of output approximating zero (e.g., Kim, H.Y.; Mester), but the results highly depend upon the degree of approximation (Berger, Hanweck and Humphrey, 1986). Third, the estimates of scope economies cannot avoid extrapolation errors in most cases that do
Berger et al. (1986) proved that 92C(Q)/3Qi9Qj > 0 must hold for some output combinations in the case of tlie translog function, which gives inconclusive scope economy findings. 21 not have observations of zero output. That is, the zero output value is far outside the sample range over which the cost function is estimated (Berger, Hanweck, and
Humphrey, 1986, 1987).
Some researchers have tested for the existence of cost subadditivity, which is a sufficient condition for a natural monopoly with economies of scope. Evans and
Heckman developed a grid approach to examine cost subadditivity from the observed product mix among samples, and Hunter, Timme, and Yang (1989, 1990) extended it for analyzing large U.S. banks. Berger, Hanweck, and Humphrey (1986) developed the measurement of expansion path subadditivity of costs that is based on an idea similar to that of Evans and Heckman. Both did not find any evidence of subadditivity in U.S. banks.
The findings from these studies of scale or scope economies suggest that output inefficiency, i.e., scale or scope economies, is not large in U.S. banking. The studies of some developing countries showed that the measured scale or scope economies were significantly different depending upon the approach used, i.e., production versus intermediation (e.g., Srinivasan). Furthermore, the results of measured scope economies were questionable because of the drawback inherent in the translog function and the arbitrary approximation of zero output.
2.3 Functional Forms
Translog functional forms have been frequently used in the 1980s to explore the banking cost structure . The translog cost function is known to have many desirable properties. It is consistent with approximation theory. It contains a manageable number of unknown parameters that are easily estimable, and it is a 22 fiinction o f the class of flexible functional forms that do not restrict the second-order properties of the cost structure (Caves, Christensen, and Trethway; Guilkey, Lovell and Sickles; Mester). The translog fiinction allows testing for homotheticity, jointness in production, and the estimation of a U-shaped average cost curve. It is relatively easier to deal with in the econometric sense and is more parsimonious in number of parameters to estimate than other flexible functional forms (Christensen, Jorgenson, and Lau).
On the other hand, many researchers have identified shortcomings of the translog function in empirical studies. First, it does not permit zero output so that scope economies can not be appropriately explored without extrapolation error.
Second, the remainder term of the translog function may behave poorly away from the point of approximation. This problem introduces heteroscedasticity in the error term, which biases the standard error estimates and causes inefficient point estimation
(Wales; Caves and Christensen; Gallant; Barnett; Barnett and Lee). Third, the approximation region where the cost function is well defined, that is, where the cost function is nondecreasing and concave in input prices, may be small (Barnett and Lee;
Mester; LeCompte and Smith, 1990).
Several alternative functional forms have been suggested to compensate for the shortcomings of the translog function. The first problem of the translog can not be avoided as long as the function is logarithmic. To deal with the zero output problem, the generalized translog functional form using a Box-Cox transformation, recommended by Caves, Christensen, and Trethway, was used by Clark, Lawrence, and Kilbrides, McDonald, and Miller. It has not been used popularly, however, because of the intractability of its results (Baumol, Panzar, and Willig). The quadratic functional form recommended by Baumol et al. was used to measure scope economies 23 by Roller and Dermine and Roller. But the drawback of the quadratic functional form is the imposition of strong separability between outputs and inputs that should be tested. Pulley and Braunstein developed a composite cost function that combines the translog with the quadratic function and allows for zero outputs and jointness in production. But it requires the estimation of many parameters; moreover, it is highly nonlinear.
To solve the second and third problems, the minflex-Laurent Amctional forms have been used as an alternatives in banking studies (LeCompte and Smith, 1990;
Hunter, Timme and Yang, 1990). However, since this solution also requires an increase in the number o f parameters to estimate, it is not applicable in the small sample case.
Therefore, any other alternative functional form besides the translog also has shortcomings and there is no criterion for use in selecting the best functional form. For that reason, translog cost functions are still widely used, despite some drawbacks, due to this simplicity and ease of use.
2.4 Production Frontier Approaches to Banking Production
2.4.1 Rationale
The traditional cost function approach for determining characteristics of the production technology depends on the assumption that all banks or credit unions are approximately efficient, meaning that they are successful in reaching the efficient frontier where costs are minimized given output levels, input prices, and the existing technology (Stevenson; Berger and Humphrey, 1991; Reifschneider and Stevenson). 24
This assumption, however, is questionable in empirical studies. If banks or credit unions are not equally efficient, the average relationship estimated by ordinary least squares method might not reflect the frontier relationship (Stevenson). Second, the degree of economies or diseconomies observed in empirical studies has not been very significant in explaining the wide dispersion of costs among banks. Hence, the results from the traditional approach may not be valuable if even inefficient banks can be viable due to market distortions caused by regulations (Berger and Humphrey, 1991).
The measurement of production inefficiencies in banks is relatively recent.
Studies report that despite the dearth of scale and product mix economies, bank costs reveal a striking degree of dispersion (Aly, et al.; Bauer, Berger, and Humphrey;
Elyasiani and Mehdian; Evanoff and Israilevich, 1990; Ferrier and Lovell; Berger and
Humphrey, 1991). Berger and Humphrey (1991) argue that a substantial portion of the dispersion in costs appears to reflect inefficiencies that donunate measured scale and scope economies. This implies that cost reductions from a further deregulation of financial markets will result in the elimination of production inefficiencies. The approaches used to measure production inefficiency are classified into two categories: the econometric frontier approach and the linear programming (LP) approach (or data envelopment analysis(DEA)).
2.4.2 The Econometric Frontier Approach
Econometric stochastic frontier models have been applied recently to banking studies (Ferrier and Lovell; Bauer, Berger, and Humphrey). These models assume that managerial inefficiencies, such as over-utilization or under-utilization of resources are included in the error term of a standard cost function (Aigner, Lovell, and Schmidt; 25
Meeusen and Van den Broeck). The error term is assumed to be the sum of two independent components; that is, the error associated with integrated inefficiency differences and the error reflecting the stochastic characteristics of the frontier or the measurement error with respect to the dependent variable.
The problems with this model concern how to view the distribution of inefficiencies and how to construct the relationship between the error term of a cost function and that of its share functions (Bauer; Greene, 1980, 1990; Stevenson).
Aigner, Lovell, and Schmidt (ALS model) suggest half-normal and exponential distributions as the appropriate distributions of inefficiencies. This assumption implies that the likelihood of inefficient behavior monotonically decreases for increasing levels o f inefficiency, an assumption that is rather inflexible (Stevenson; Greene, 1990).
Thus, Stevenson suggests a truncated normal or normal gamma distribution to mitigate the problem. Greene (1980, 1990) suggests the gamma distribution for the problem. But the half-normal assumption by ALS is most widely used due to the simplicity of error specification.
To estimate the econometric frontier model, the maximum likelihood estimation method (MLE), following the ALS model, has been popularly used. Olson,
Schmidt, and Waldman suggested the method of corrected ordinary least squares
(COLS) as an alternative for estimation of stochastic frontier models. In the cost function approaches, the single equation and a system of equations are used, but the problem of the relationship between the cost function and its share equations is not yet solved when using the system of equations (Greene, 1980; Bauer). When using the single equation method, the COLS can be used as an appropriate method because of the ease in dealing with the model. Olson et al. showed that the COLS tended to outperform the MLE in sample sizes of less than 400. The main advantage of the 26
COLS is that the COLS estimators are easier to compute than the MLE ones, and the main disadvantage is that the former is less efficient than the latter (Schmidt and
Lovell).
2.4.3 The Thick Frontier Model
This approach, developed by Berger and Humphrey (1991), estimates two cost functions - one for the lowest average cost quartile of banks and the other for the highest average cost quartile of banks. The difference between the two functions is assumed to reflect market factors and inefficiency residuals. The inefficiency residual is then decomposed into several types of inefficiencies by using procedures similar to those of Kopp-Diewert-Zieschang. Another assumption is that the error terms within the lowest and the highest cost quartiles reflect only random measurement error and luck. Obviously, this approach lacks precision and imposes some rather ad hoc assumptions to develop the subgroups and produce the frontier (Berger and
Humphrey, 1991; Evanoff and Israilevich, 1991). Another problem is that the model requires either a priori information about efficient banks or large samples.
2.4.4 Linear Programming Approach
The linear programming (LP) approach is based on Farrell's work and its extension by Fare, Grosskopf, and Lovell. This approach defines the level of inefficiency of a bank as the proportional reduction in input quantities that can occur while maintaining a given level of output or, alternatively, the proportionate reduction in inputs that can be achieved when the firm moves from its actual production frontier. 27
This approach uses an LP technique to estimate piecewise linear production or cost frontiers that connect the inputs or the costs of the efficient firms. Firms on the vertices of the piecewise linear frontier are considered Ailly efficient, and the inefficiencies of other firms are measured relative to this frontier (Rangan, et al.; Aly, et al.; Elyasiani and Mehdian; Ferrier and Lovell; Berger and Humphrey, 1991).
This approach does not require any assumptions about the functional form; the efficiencies of a decision making unit are measured relative to all other decision making units with the simple restriction that all decision making units lie on or below the efHcient frontier (Seiford and Thrall). The efficiency of a specific firm is directly measurable without a complicated specification of error terms or indirect derivation as a parametric frontier model.
This approach, however, regards random errors or all variations fi'om the efficient frontier as reflecting inefficiencies, so even small changes in measurement error may have a large cumulative effect on the aggregate inefficiencies (Berger and
Humphrey, 1991). The results of the analysis are sensitive to the variables selected and the number o f constraints (Berger and Humphrey, 1991; Seiford and Thrall).
Moreover, this method does not provide information about statistical inferences, such as significance levels o f estimates, so we cannot statistically test hypotheses.
The frontier fimction approaches have revealed strikingly large input inefficiencies in U.S. banking. The range of overall input inefficiency was from about
20 percent (Berger and Humphrey, 1991, Ferrier and Lovell) to about 50 percent (Aly et al ), which contrasted remarkably with the degree of estimated scale inefficiency
(see Evanoff and Israilevich, 1991). These differences might be a result of the estimation methodology used. For example, the former was obtained by a parametric
(thick frontier function) approach, while the latter was obtained by a nonparametric 28
(LP) approach. But the LP approach is not intuitively appealing because it ignores random effects on the measured degree of inefficiency and does not provide any clues for statistical inference.
2.5 The Objectives of Cooperatives
So far the review presented the studies that depended on the assumption of cost minimization which is acceptable for an application to general firms. But a question arises about the usefulness of the cost minimization assumption when we analyze cooperatives that have a complex structure that combines a number of different groups whose interests may not always be consistent (LeVay). In other words, can cost minimization be considered as the objective of the cooperative ?
A wide spectrum of views about objectives is found in the literature. Smith et al. summarize two principle characteristics of credit unions that prevent the simple application of general firm theory. First, the members of a cooperative are both the owners of the organization and the consumers of its outputs or the suppliers of its inputs. Second, the membership of a cooperative provides both the demand for and supply of loanable funds.
Thus, many researchers argue that a multiple objective function model, considering two or more objectives simultaneously, may be appropriate for the analysis of credit unions, because of the possibility of inherent conflict, rather than the homogeneous objective of cost minimization (Taylor; Smith, Cargill, and Meyer; Smith, 1984). Smith, Cargil, and
Meyer argue that the objective function of credit unions is the maximization of total benefits, comprised of lender and borrower benefits. The optimal point depends on which members dominate the credit union. The cases are divided into borrower dominant, lender 29 dominant, and neutral cases. Homogeneous objective assumptions such as profit maximization or cost minimization are reasonable only if the objective fimction is neutral.
However, there are few empirical studies that have tested the hypothesis. Smith
(1986) tested the variant objective function hypothesis and did not find convincing evidence to support it. The results support neutrality between borrower and saver interests. Smith suggests three reasons for the neutrality: altruistic motivations o f the members, nonexclusiveness between lenders and borrowers, and the possibility of mass withdrawal. Altruistic motivations are the heart of the cooperative philosophy, so only the neutral operation of the cooperative is relevant. Borrowers and lenders are not always mutually exclusive in a dynamic sense, since the borrowers o f today can be the lenders of the future and vise versa. The balance sheet constraint of a credit union, the balance between savings and loans, is another source of tlie neutrality, because mass withdrawal or nonentry by the nonpreferred group could threaten the viability of credit unions. Thus,
Smith's results support the homogenous objective fimction models, under the cost minimization assumption, as specified by Koot, Wolken and Navratil, Murray and White
(1980, 1983), Fryetal..
Considering the agricultural cooperatives in Korea, the homogenous objective function model is acceptable. The above discussion depends on the assumption that the cooperatives can make decisions independently concerning the prices of inputs and outputs. But if the cooperatives are price takers, then the controversy about the objectives will be meaningless because the only thing that the cooperatives can do is minimize costs with a given output level for their members. 30
2.6 Summary
The development of multiproduct firm theory and econometric technology has enabled us to obtain more information about the cost structure of banking. However, many unanswered questions still remain, such as the appropriate definition of output, the models, functional forms, and the error distribution relationship between the production function and the cost function in the case of the stochastic frontier function to measure input inefficiency.
There are three types of banking production approaches using the cost function, i.e., the production, the intermediation, and the user cost approach. Of these, the intermediation approach that defines deposits as bank outputs is preferred to analyze the viability of banks which is the objective o f this study.
The findings obtained from the U.S. banking studies showed that output inefficiency, i.e., scale or scope economies, were not large, but the frontier function approaches presented strikingly large input inefficiencies. This suggests that even inefficient banks might be successful because of the protection from competitiveness provided by banking regulations. Another problem is that the measured scope economies were questionable because o f one drawback of the translog fimction, and normally its arbitrary approximation to zero output.
Considering a functional form as an instrument to explore the cost structure, there is no criterion to determine what is the best functional form. Any alternative functional form besides the translog also has shortcomings. For that reason, the translog cost function is still widely used, despite some drawbacks, due to simplicity and ease of use. 31
This chapter discussed three types of frontier function approaches, the stochastic frontier function, the thick frontier model, and the LP model. The LP approach is not intuitively appealing because it ignores random effects on the measured degree of inefficiency and does not provide any clues for statistical inference. The thick frontier model is intuitively rather than theoretically appealing.
But it requires a large sample. Therefore, the stochastic frontier function approach is appropriate for this study because it uses a small sample.
There is a controversy about whether or not the cost minimization assumption, which is a sufficient condition for the use of the cost function approach, is proper in the case of the study of cooperatives. The main argument against the homogenous objective assumption (cost minimization or profit maximization) is that the multiple objective function model, which considers two or more objectives simultaneously, may not be appropriate in the analysis of credit unions, because of the possibility of inherent conflicts.
But this argument may be reasonable if the cooperatives can independently make decisions concerning the prices of inputs and outputs. But if the cooperatives are price takers, such as in Korea, then the controversy about the objectives is meaningless because the only thing the c— eratives can do is minimize costs with a given level of output for its members. Therefore, considering the structure of agricultural cooperatives in Korea, the homogenous objective function model seems acceptable. CHAPTER III
KOREAN FINANCIAL MARKETS AND AGRICULTURAL COOPERATIVES
This chapter is designed to describe Korean formal financial markets, the performance of PAC banking in the market, and the contribution of banking to the viability of the PACs. This chapter consists of four sections.
In the first section, the characteristics of economic growth and financial development in Korea in the 1980s will be summarized in order to describe the current financial market environment. The discussion in this section will focus on macroeconomic performance, the relationship between economic growth and financial development, and the process of financial liberalization that resulted in the current financial market environment.
In the second section, the current formal financial market structure will be analyzed in terms o f financial institutions, interest rates by type of financial institution, and the market share of each financial institution. In addition, the structure o f the rural financial market, where most PACs operate, will be discussed independently. For the discussion of market share and the rural financial market structure, time series data will be used to capture the trend in the financial market structure. In this process, the performance of the PAC banking will be revealed. In the last part of this section, the
32 33 factors that affect the successful performance of PAC banking will be interpreted, and these factors will be interpreted in terms of the future of PAC banking.
In the third section, the agricultural cooperative system and the importance of the banking business for the PACs will be presented by describing the structure and relationship of the PACs with their federation, the NACF. The analysis of business structure will consider the economic contributions of the banking for the viability of the
PACs. Other functions, such political and sociological contributions to members or to the community, are beyond the scope of this study. The last section will summarize the major points of the discussion in this chapter.
3.1 Economic Growth and Financial Development of Korea
3.1.1 Economic Growth
The Korean economy has grown rapidly through successful export policies in the past 30 years. From 1970 to 1991, the annual growth rates o f real GNP and per capita income were 9.5 and 7.6 percent, respectively. In terms of the nominal value of the US dollar, per capita income increased from 289 dollars in 1971 to 6,498 dollars in 1991. In the process of economic growth, the government played the lead in economic planning, investments, exports, resource allocation, etc. until the late 1970s. But the failure of the heavy and chemical industrialization policy, high inflation, and the stagnated economic growth in the late 1970s forced the government to change the policy direction towards more of a market-oriented economy by reducing market intervention. In the early 1980s, the government made strong efforts to stabilize the price level and gradually reduced market intervention. The deregulation of the financial sector that had been strongly 34
Table 3.1 The Growth of Real Income
Y ear GNP (A) Agri. (B) Agri. Share Per Capita Farm Household (B/A) Income Income billion won billion won % thousand won thousand won 1975 37,143 7,373 23.4 1.053 3,198
1981 55,354 7,238 15.8 1,429 4,487
1985 78,088 8,799 13.3 1,914 5,736
1990 130,685 8,681 7.8 3,048 8,404
1991 141,602 8,633 7.2 3,273 9,007
Annual Growth Rate
1975-81 8.3 -0.4 6.3 7.0
1981-91 9.8 1.8 8.6 7.2
Source: The Ministry of Agriculture, Forestry & Fisheries (MAFF), Major SlaiisUcs o f Agriculture,
Forestry & Fisheries I992\ The Bank of Korea (BOK), National Accounts.
Note: Won is money unit in Korea.
regulated in order to allocate resources to the preferred sectors was stressed.
The policy changes produced very successful results such as maintaining a high rate o f economic growth while stabilizing the inflation rate at a very low level, a positive current balance in foreign trade and reduction of foreign debt, a reduction in the unemployment rate, etc. The annual growth rates of real GNP and per capita income from 1981 to 1991 were 9.8 and 8.6 percent respectively, which were higher than the respective values from 1975 to 1981 (Table 3.1). The inflation rate declined sharply from about 20 percent in the late 1970s, on average, to about 6 percent from 1982 to 1991, 35 when measured by the GNP deflator* (Table 3.2). In addition, the rapid increase in exports resulted in a considerable surplus in the current account from 1986 to 1989
(Table 3.2). The successful economic growth in the mid 1980s encouraged not only the government to conduct more deregulation of markets but also foreign countries to press
Table 3.2 Macroeconomic Indices of the Korean Economy 1980 - 1990
Year Inflation Current Unemploy- Budget Savings Investment Rate * Balance ^ ment Rate surplus c Rate Rate % million dollars % billion won % % 1980 24.0 -5321.0 5.2 - 23.1 32.0 1981 16.9 -4646.0 4.5 -1827.0 22.7 29.9 1982 7.1 -2650.0 4.4 -1727.0 24.2 28.9 1983 5.0 -1606.0 4.1 -584.0 27.6 29.2 1984 3.9 -1373.0 3.8 -757.0 29.4 30.3 1985 4.2 -887.0 4.0 -1115.0 29.1 29.9 1986 2.8 4617.0 3.8 23.0 32.8 28.9 1987 3.5 9854.0 3.1 1198.0 36.2 29.6 1988 5.9 14161.0 2.5 2770.0 38.1 30.7 1989 5.2 5055.0 2.6 1900.0 35.3 33.5 1990 10.6 -2179.0 2.4 -2598.0 36.0 37.1
Source: Economic Siatislics Yearbook 1991,the Bank of Korea, 1991.
3 Based on the GNP deflator.
^ The current balance in international trade.
® The budget surplus or deficit of the government, the central government and local governments.
* * Average inflation rates which were measured by the whole sale price index, consumer price index and GNP deflator from 1982 to 1990 were 1.5 % and 4.7 %, and 5.4 % respectively. 36 for the opening of the domestic market such as for agricultural products and the financial market.
On the other hand, the stabilized price levels resulted in positive real interest rates in banks despite the fact that nominal interest rates declined in 1981 and 1982, which increased the domestic savings rate measured by the ratio of total domestic savings to
GNP. From 1986 to 1989, the increased domestic saving rate exceeded the domestic investment rate measured as the ratio of total investment to GNP, which enabled the government to reduce the foreign debt due to a substantial surplus in the current account. Furthermore, the economic boom of this period increased tax revenues from the increased income and consumption levels, which resulted in a substantial budget surplus.
With these favorable macroeconomic conditions, the government was able to proceed with financial liberalization.
The agricultural sector suffered from several problems, however, in the process of structural change, which increased the instability of the rural economy that is politically and socioeconomically heavily burdened. Agricultural production stagnated in the 1980s. Agricultural production decreased by 22.7 percent because o f bad weather in
1980, showed negative growth rates again in 1987, and also from 1989 to 1991 because of a reduction in grain production. The main reason for declining agricultural production in the late 1980s was that farmers gave up cultivating the marginal land, due to low profitability and the lack of labor, which rapidly increased in recent years. Thus, the share o f the agricultural sector in GNP sharply declined from 23.4 percent in 1975 to 7.2 percent in 1991 (Table 3.1).
For the farm households, though average real income has steadily increased, the income gap between rural and urban areas has been growing, and the debts of farm households have increased remarkably. The debt problem of the farm households became 37 an especially important political issue in the mid 1980s and forced the government to return to the preferential interest rate policy for the agricultural sector in 1987^. The increased debts were closely related to policy failures in stabilizing the prices of agricultural products, an increase in capital investments to commercialized farming, and a worsening of the socioeconomic environment in rural areas including education and health care facilities.
Considering rural financial markets, factors such as commercialization of farm output and the increase in both income and debts increased the demand for financial intermediation. With stagnated agricultural production, non-farm income became a more important income source for farm households, which increased dependency on money as opposed to in-kind methods of payment. The commercialization of farming means monetization of the farm household economy. The increase in debts requires more financial intermediation. The factors that increased liquidity in rural areas that must have resulted in internal expansion of rural financial markets.
3.1.2 Socioeconomic Changes and the Demand for Financial Services
In the process of rapid economic growth, Korea has been rapidly industrialized, urbanized, and internationalized in the past 30 years. The industrialization of the Korean economy reallocated resources from the agricultural sector to the non-agricultural sector, increased market integrity by developing social infrastructure and communication media and through dismantling geographically segmented regional markets, and brought
^ The government removed the preferential interest rates of policy loans that were kept to support to specific sectors such as the international trade or the agricultural sector in 1982. But as the farm household economy became worse with rapidly increasing debts, the rural economy became unstable and political pressure to support the rural economy increased. Thus, the government returned the interest levels of policy loans to the preferential interest policy as before. 38 about westernized urbanization. This process, in other words, means that the homogenous agrarian community was dismantled and a heterogeneous urban or industrialized life style began to dominate the Korea community. The influence o f these socioeconomic changes on the financial markets can be summarized in the following way.
First, the economy became too large for the government to control directly, and socioeconomic development produced interest groups too diverse for the government to optimize utility through direct control of the economy. Thus, the government was obliged to change the policy direction towards a market-oriented economy in the late
1970s, which was important background for the financial liberalization o f the 1980s
(Chung).
Second, migration from rural to urban areas continued. The farm household share of total population that was 55.1 percent in 1965 and 28.4 percent in 1980 decreased to
14.0 percent in 1991. This concentration of population in urban areas implies an increase in total liquidity in the economy since the dependency on money by urban residents is generally higher than that of farmers. This change implies the expansion of the urban financial markets and the reduction of rural financial markets.
Third, market integration was accelerated in terms of both domestic and international markets. As social infrastructure and communication media develop, geographically fragmented markets have been integrated into broader market areas, which resulted in an improvement in financial intermediation efficiency. The integration of the market also implies an increase in competition among the financial institutions that already existed. In addition, the opening of the domestic market resulted in international market integration. 39
Fourth, the increased heterogeneity of the community resulted in diverse patterns of consumption that produced a structural change in production, which in turn enhanced the heterogeneity of the community. For the agricultural sector, the changes in the pattern of food consumption resulted in rapid commercialization of farming into vegetable, fruit, livestock, and flower farming. The commercialization of farming stimulated the monetization of the farm household economy, which resulted in the expansion o f rural financial markets through the increased liquidity of the rural economy, and stimulated market integration. On the other hand, the diverse demand for financial services from the increasingly heterogeneous community forced financial intermediaries to innovate in the production of financial services.
Fifth, democratization has proceeded steadily and social welfare programs have been improved. Chung pointed out that the government, which was captured by a military group in 1980, was not viewed favorably by the people, so that it tried to deregulate the financial sector to get political support from the people. The social welfare programs, such as the introduction and expansion of national health insurance and the national pension program, created an increase in savings for health care and old age.
Consequently, the socioeconomic changes together with the expansion of the economy towards more a diverse dimension forced the government to deregulate the financial sector, while the financial institutions conducted technical innovations in response to the changes in demand for financial services. Examples of the new financial services available in the 1980s, are the sales of government and public bonds under repurchase agreements (RPs), cash management accounts (CMA) as a Korean version of money market funds (MMF) in the U.S., credit card businesses, etc.. These services were widely introduced with the help of computerization and technological developments. 40
It can be said that the socioeconomic changes have had a favorable effect on the incumbent financial institutions so far. The financial market was substantially extended and financial deepening proceeded remarkably in the 1980s as will be presented in the following section. But the financial markets are not yet competitive enough to give efficient services to their clientele. Therefore, if financial liberalization continues and the financial institutions are currently inefficient in terms o f scale and product mix economies, some financial institutions that do not have or cannot achieve optimal scale and product mix will face a severe problem o f viability.
3.1.3 Financial Liberalization and Money Deepening
The financial sector was strongly regulated to support key industries and exports.
Generally, the existence of externalities and the problem of asymmetric information in financial transactions justifies financial regulation in developed countries; however, a more important reason for financial regulation is support for development o f the economy in Korea (Chung). Chung argued for financial regulation purposes economic development has always been given priority over other purposes such as the sound operation of financial institutions, protection of depositors, and maintenance of the financial system.
There are several types of regulatory legal acts, administration processes, and routine orders or instructions of the monetary authority. These include bank chartering as an entry barrier in the financial industry, the required level of equity and reserve requirements to control balance sheets, the regulation of kinds of assets and debts, the regulation of interest rates for both deposits and loans, and audits and supervision. 41
Among the regulations, low interest rates and credit allocations were the major financial policies followed until the late 1970s. In the 1960s, a Keynesian belief that low interest rates stimulate investment played a dominant role in planning economic policies.
In the 1970s, a fear o f the high financial cost burden on firms and cost-push inflation induced the government to keep low interest rates policies (Cole and Park). The depression of the financial sector undermined financial development, especially in the
1970s; for example, the M2 ratio to GDP did not rise from 1970 (32.4 %) to 1980 (33
%) as shown in Table 3.3.
Since the early 1980s, however, the government implemented a broad and gradual deregulation of financial markets to promote competition and enhance efficiency through the market mechanism. Financial liberalization was broadly recognized as an effective strategy for financial development as well as for economic development. Entry barriers to the financial industry were lowered and the ceilings on interest rates were deregulated. In addition, the internationalization of the Korean financial markets has been proceeding in response to the pronounced internationalization of the rest of the economy. The major series o f financial deregulation efforts in the 1980s can be summarized as follows (The Bank of Korea, 1990):
1) To lower the entry barriers into the financial industry, the government chartered many
nationwide commercial banks, securities investment companies, and life insurance
companies in the 1980s. Furthermore, restrictions on opening branches for local banks
and savings institutions were partly deregulated.
2) The imposition of credit ceilings on individual banks that were supposed to directly
control total banking credits was replaced with an indirect control system in 1982 in
order to increase the extent of managerial autonomy of banking institutions. 42
3) The preferential rates on policy loans made by commercial banks were abolished in
1984.
4) As the first stage in the deregulation of interest rates, the ceilings on inter-bank call
rates and the issuing rates of corporate bonds without guarantee were lifted in 1984.
5) As a measure toward creating a universal banking system, a realignment of the
business boundaries between different kinds o f financial institutions was made, so that
commercial banks' ancillary and peripheral businesses were diversified and non
banking financial institutions such as investment and finance companies, merchant
banking corporations were allowed to deal with new financial commodities such as
commercial paper, and cash management accounts in the early 1980s.
6) In 1988, most of the lending rates of banks and non-bank financial institutions were
liberalized, while the interest rates on deposits were partially liberalized.
7) As a process of internationalization of financial markets, many foreign bank branches
were allowed to open in the 1980s, and the securities market was open to foreigners
since 1992. In addition, the domestic life insurance market has been open to foreign
life insurance companies since 1987.
Further financial liberalization will be continued until the financial market is deemed as competitive as in the developed countries. It is expected that determination of interest rates will be completely deregulated after 1997, and the current program of financial liberalization will be complete. The scope of business of financial institutions will be extended; the government is planning to change the banking system toward the universal banking system that means little differences in the scope of business among banks, investment companies and institutions in the securities market.
So far, the financial liberalization of Korea has been cited as a successful case
(Cho; Fry, M.J.; McKinnon, R.I., 1988). Cho reported that the efficiency of credit 43 allocation has improved following financial liberalization. McKinnon pointed out that
Korea successfully performed financial deregulation while maintaining price stabilization and tight control of exchange rates, which resulted in positive real interest rates and increases in domestic savings. Certainly, the financial liberalization and rapid economic growth accelerated the financial deepening of the Korean economy as shown in Table
3.3.
Table 3.3 Monetary Deepening of Korean Economy
Year GDP M2 M3 M2/GDP M3/GDP
billion won billion won billion won % Vo 1970 2,768 898 N.A 32.4 N.A 1975 10,302 3,150 3,903 30.6 37.9 1980 38,041 12,535 17,811 33.0 46.8 1981 47,482 15,671 23,243 33.0 49.0 1982 54,443 19,904 30,965 36.6 56.9 1983 63,833 22,938 37,648 35.9 59.0 1984 72,644 24,706 45,204 34.0 62.2 1985 80,847 28,565 54,764 35.3 67.7 1986 93,426 33,833 70,710 36.2 75.7 1987 108,428 40,280 92,040 37.1 84.9 1988 127,963 48,939 118,135 38.2 92.3 1989 143,001 58,638 150,774 41.0 105.4 1990 172,724 68,708 193,410 39.8 112.0
Source: Economic Statistics Yearbook 1991,BOK, 1991.
Note: All monetary values are expressed in nominal terms. 44
The ratio of money demand measured by currency and time and savings deposits o f the banking institutions (M2) to GDP slowly increased from 33 percent in 1980 to
39.8 percent in 1990, but the ratio of total liquidity in the economy (M3) to GDP sharply increased from 46.8 percent to 112 percent for the same period. The sharp difference between these two financial ratios implies that financial assets at non-banking financial institutions have grown rapidly compared to the assets of the banking institutions. The government uses M2 as an index to control money supply, but the disparity between M2 and M3 is an incentive to use M3 instead, which would imply that some of the non-bank financial institutions such as PACs may be subject to the government's money supply policies.
However, Chung argued that the most important factors for financial liberalization are still strongly controlled by the government. The management of the banking institutions is not independent of the monetary authority; first, the government affects the management of commercial banks in decision making by appointing the presidents of commercial banks; second, the commercial banks can not determine their lending rates by the principle of profit maximization, although most lending rates were legally deregulated in 1988; third, the banks used to get permission from the monetary authority to make loans over a certain amount, which undermines their ability to evaluate, screen, and select borrowers. Though government does not intervene as strongly in the non-bank financial institutions, they are indirectly influenced by government control.
As Chung appropriately argued, the essential part of financial liberalization is not completed, and the banking institutions are the most strongly regulated compared to the other financial institutions. Therefore, the direct impact of financial liberalization will be greater for the banking institutions than for the savings institutions such as the PACs 45
(Lee, D.H et al.). However, direct impacts, through the expanded influence of the deregulated banking institutions on the financial markets, can be substantial for small institutions such as the savings institutions. It is expected that the banking institutions and some o f the non-bank flnancial institutions will become bigger through mergers and expansion o f their scope of business. The banking institutions or other big flnancial institutions whose major clientele are large firms may extend their services to households through market competition. Then small financial institutions will face the problem of securing good clientele, and obtaining optimum scale and scope economies.
In the case of PACs, they will lose their current advantage of price competition, and high interest rates in deposits and loans, which may cause a substantial amount of the non-member clientele in urban areas to change their financial institutions. Another problem may come from the change in the method of controlling the money supply from
M2 to M3, which means that deposits at non-bank financial institutions will become subject to monetary policy (NACF, 1992). In that case, the PACs will be required to deposit reserve requirements at the central bank, where they would get no return. Since the PACs have deposited their reserve requirements in Mutual Credit Special Accounts at the National Agricultural Cooperatives Federation (NACF) and earned a 10 percent interest rate, the change in money supply policy implies that the PACs will face a substantial cost burden. 46
3.2 Formal Financial Market Structure and PAC Banking
3.2.1 Financial Institutions
Banking Institutions: The financial institutions are categorized into four groups, i.e., the central bank, banking institutions, non-bank financial institutions, and institutions operating in securities market, as shown in Figure 3.1. The banking institutions (or deposit money banks) consist of commercial banks that were established according to the General Banking Act, and specialized banks that were established according to a series of separate acts. But there are few differences in the scope o f business and regulations affecting commercial banks and specialized banks. The non bank financial institutions include diverse financial institutions such as development institutions, savings institutions, investment companies, and life insurance companies.
Considering the scope o f this study, it is necessary to narrow the range of financial institutions. Since the main functions of the savings institutions, such as PACs, are mobilizing and lending funds acquired through deposits and borrowings, which is, in part, the same as the functions of banking institutions, discussion henceforth will focus on only the banking institutions and the savings institutions.
There are three types of commercial banks such as nationwide commercial banks, local banks, and foreign bank branches. At the end of 1990, eleven nationwide commercial banks operated with 1,695 banking offices nationwide. Five of the eleven nationwide commercial banks were chartered in the 1980s and another one changed its status from a specialized bank to a commercial bank in 1989. In addition, the other nationwide commercial banks, except for one bank that was privatized in 1972, were denationalized in the early 1980s. In other words, the ownership of the commercial 47
Central Bank The Bank of Korea
Nationwide Commercial Banks (11) Commercial Banks Local Banks (10)
Banking Institutions Foreign Bank Branches (67) (Deposit Money Banks)
Specialized Banks IBK, CNB, NACF, NFFC, NLCF
Development Institutions
Non-Bank Financial Institutions Trust Accounts of Banks (53), Savings Institutions MSFC (237), Credit Unions (1,318) MCF (1,678), NCFA (3,265), Postal Savings (Post Offices)
Investment Companies
Life Insurance Companies
Other Institutions
Securities Market
Source: Financial System in Korea, The Bank of Korea, 1990. p. 9.
Note: Figures in parentheses represent the number of institutions
Figure 3.1 Financial Institutions in Korea (As of the end of June 1990) 48
banks was completely denationalized in 1980s; however, the government still appointed
the presidents of the commercial banks.
At the end of 1990, ten local banks operated with 638 banking offices. All local
banks have been privately owned from the outset in contrast to the nationwide
commercial banks and the specialized banks. The local banks maintain a branch banking
system like the nationwide commercial banks, but the opening of branches is allowed
within the provinces where their head offices are located, except for Seoul city and
neighboring provinces. The local banks are relatively small banks whose total assets were
on average about one-fifth of the total assets of the nationwide commercial banks at the
end of June 1990 ( The Bank o f Korea, 1990).
The influence of foreign bank branches on the financial markets considerably
increased, both in numbers and also in the range of banking services. The number of
foreign bank branches increased from 18 in 1979 to 69 at the end o f 1990. In addition,
there were 24 foreign bank representative offices as liaison offices at the end of 1990.
Six specialized banks were designed to serve specific target groups; that is, the
Industrial Bank of Korea (IBK) to support small and medium enterprises, the Citizens
National Bank (CNB) to provide banking services to households and small enterprises,
the Korea Housing Bank (KHB) to finance housing fiands for low income households,
and the credit and banking sectors of three types of cooperative federations such as
National Agricultural Cooperatives Federation (NACF), the National Federation of
Fisheries Cooperatives (NFFC), and the National Livestock Cooperative Federation
(NLCF). The specialized banks had 1,557 banking offices at the end of 1990, which was
similar to the number of banking offices of nationwide commercial banks.
Most banking institutions, i.e., both the commercial and specialized banks,
operate in urban areas; that is, 90.2 percent of total banking offices were located in cities 49 in 1989. In other words, most of the financial institutions in rural areas are the savings institutions that are small and have high prices as will be presented in the next section.
Savings institutions: Savings institutions consist of trust accounts of banking institutions, Mutual Savings and Finance Companies (MSFCs), credit unions (CUs),
New Community (Saemaul) Finance Associations (NCFAs), Postal Savings (post offices), and Mutual Credit Facilities (MCF) such as PACs, fisheries cooperatives (FCs), and livestock cooperatives (LCs). The CU, NCFA, and Mutual Credit facilities are treated as the same kind of institution, i.e., credit unions, that are regulated by the Credit
Union Act of 1972 and may transact only with their members, so that they will be henceforth indicated as a credit union group to avoid ambiguity between the CU and the terms representing the set of CUs, NCFA, PACs, FCs, and LCs. Most savings institutions were legally recognized as financial institutions with the promulgation of the
Presidential Emergency Decree in 1972, designed to induce the movement of informal financial market funds into the formal financial market.
The MSFCs were introduced by the Mutual Savings and Finance Company Act in 1972 to absorb small savers' funds from the curb markets and to provide financial services to households and small business firms. At the end of June 1991, there were 237
MSFCs with 97 branches. The main business of the MSFC consist of mutual installment savings, which give entitlements to borrow funds, mutual time deposits, small unsecured loans and discounts of bills for mutual investment savers (BOK).
Since 1960, many credit unions have been organized as cooperative associations in offices, churches, and other private groups in order to facilitate financing for their members and to promote mutual economic benefits, but they were not subject to 50 administrative guidance and supervision until 1972. At the end of February 1992, there were 1,362 credit unions with 2,307 thousand members.
The NCFA are also a kind of credit union, which are organized as cooperative associations among people who reside in the same neighborhood in rural or urban areas in order to promote the Saemaul Movement (New Community Movement) under the auspices of local governments. At the end of June 1990, there were 3,265 NCFA with
5,730 thousand members.
The Mutual Credit facilities such as the PACs and the specialized agricultural cooperatives (SACs), FCs, and LCs are different from the CU or the NCFA in that they have strong and sound federations as banking institutions. Among them, the PACs that are the focus of this study have the largest and most systematic organizational network and are well equipped with on-line systems, which is important in order to gain a competition edge. There were 1,425 PACs and 41 SACs at the end of 1991,90 fisheries cooperatives and 160 livestock cooperatives at the end of June 1990.
The postal savings system that was consolidated with the PACs in 1977 was resumed in 1983. The postal savings are closer to being banking institutions than other savings institutions. But the interest rates paid on deposits are 0.5 percent higher than those of banks and they are free from reserve requirements because the system is to provide a public service rather than to make profits. For that reason, the government guarantees payments against deposit liabilities. The postal savings have well organized on-line systems. 51
3.2.2 Interest Rates
The current interest rates for deposits and loans are different by type of financial institution. Based on the interest rate for one year time deposit in 1992, the rate paid by the banking institutions is 10 percent per year, and those of the credit union group -
PAC, CU, NCFA, LC, and FC - are 11 percent, while that of MSFC is 14 percent per year (Table 3.4). For the credit union group, the interest income accruing from deposits is exempt from income taxes, education taxes, and residence taxes for amounts below
Table 3.4 Interest Rate Level per Year by Type of Financial Institution 1992
Financial Institutions Interest Rates of Deposits Interest Rates of Loans
a
Deposit Money Banks 10.0% 11.0- 13.0% b
Savings Institutions
M SFCc 14.0 17.2
PAC, CU, NCFA 11.0 14.0 d
Postal Savings 10.5 -
Source: Economic Statistics Yearbook 1991,BOK, 1991; Rhu et al., 1992.
^ Based on one year time deposit.
1) Based on commercial loans in 1988. Although the lending rates were deregulated since 1988, there is
little change in tlie rates.
® Rhu et al., 1992.
^ Based on Mutual Credit Loans of PACs. 52 the ceiling for deposits per clientele. Therefore, the PACs have a price advantage in mobilizing funds compared to the banking institutions, but a disadvantage compared to the MSFCs.
The lending rates among the various types of financial institutions are also different in accordance with these different deposit rates. For commercial loans, the lending rates of the banking institutions are from 11 to 13 percent per year, and that of
PAC is 14 percent, while that of MSFC is 17.2 percent per year.
3.2.3 Market Share
The market share in total deposits by type of financial institution and changes from 1985 to 1990 are presented in Table 3.5. These data include only banking and savings institutions. The banking institutions are still dominant in deposit markets but their market shares declined from 79.6 percent in 1985 to 71.3 percent in 1990, which was reflected by the slower annual growth rate of deposits (9 %) compared to the rates of savings institutions (13.4 %).
In contrast, the savings institutions grew rapidly during this period. Above all, the growth rates of the FC and LC (24 %), NCFA (17.1 %), CU (14.5 %) were remarkable.
The rapid growth of FCs and LCs was due to the increase in the number of offices from
281 in 1987 to 548 in 1990. On the other hand, the NCFA and the CU have successfully grown by providing more kinds of services and strengthening relationships with their members. They have better systems against the problem of delinquency so they overcame low creditworthiness caused by the financial accidents which occurred frequently in the 53 past (Rhu et al. 1992; NACF, 1992)^. The growth rate of the PACs was also remarkable
(13.8 %). This resulted in the largest market share among the savings institutions in
1990.
Table 3.5 Changes in the Market Share of Total Deposits by Type of Financial
Institutions (Deposit Markets)
1985 1990 Annual Growth
Financial Rates
Institutions Deposits Share Deposits Share 1985 ~ '90
billion won % billion won %%
Banking Institutions 31,023 79.6 8,4054 71.3 9.0
Savings Institutions 7,930 20.4 33,795 28.7 13.4
PACs 2,628 6.7 11,600 9.8 13.8
CU 557 1.4 2,658 2.3 14.5
NCFA 921 2.4 5,652 4.8 17.1
MSFC 2,765 7.1 8,940 7.6 10.7
Postal Savings 872 2.2 2,722 2.3 10.4
FC,LC 186 0.5 2,223 1.9 24.0
Total 38,953 100.0 117,849 100.0 10.1
Source; Economic Statistics Yearbook 199EBOK, 1991.
^ Some presidents or managers of these institutions misappropriated client's deposits for other purposes such as informal lending or investments in real assets. This diversion of money caused insolvency of the institutions due to failures of collecting repayments. 54
3.2.4 Rural Financial Markets
The analysis presented so far was done at the national level, but the situation of rural financial markets is much different than the national averages. While the market structure at the national level suggests considerable competition, the rural formal financial markets are dominated by the PACs. The survey of the farm household economy conducted by the Ministry of Agriculture, Forestry, and Fisheries (MAFF) gives detailed information about the rural financial market structure, although it excludes information about non-farm households.
Considering the savings markets, on average 80.2 percent o f the total financial assets o f a farm household were deposited or invested in formal financial institutions
(FFI) in 1990. The financial assets deposited in banking or savings institutions represented 67.6 percent, while the financial assets of life insurance companies and securities were 8.3 and 4.4 percent, respectively, in 1990 (Table 3.6).
The share of FFI in farm household savings steadily increased during the 1980s.
It increased from 52.3 percent in 1980 to 80.2 percent in 1990 and the agricultural cooperatives, the PACs and the NACF, have absorbed most of the incremental savings, as reflected in the changes in the market share of those institutions (see Table 3.6). The financial assets represented by life insurance and securities have increased considerably in absolute value but their shares were still not large in 1990.
In contrast to the rapid growth of the FFI in the rural savings market, the share of the informal financial market (IFM) declined substantially in the 1980s, especially, the share of Kye, which is the traditional form of rotating savings and credit associations
(ROSCAs). Private loans, loans made to friends, relatives or others, showed a relatively small decrease; that is, the share o f Kye decreased from 22.9 percent in 1980 to 13.7 55
Table 3.6 Composition of an Average Farm Household's Financial Assets, 1980,
1985, and 1990
Annual Financial Assets 1980 1985 1990 Growth Rate Amount Share Amount Share Amount Share 1980-'90 thousand % Uiousand % thousand %% won won won FFMs 263 52.3 924 59.5 4,984 80.2 34.2 Deposits 194 38.6 681 43.8 4,199 67.6 36.0 Agri. Coop. & Banks 168 33.4 572 36.8 3,773 60.7 36.5 Postal Savings 6 1.2 36 2.3 185 3.0 40.9 Others 20 4.0 73 4.7 241 3.9 28.3 Securities 37 7.4 88 5.7 272 4.4 22.1 Insurance 32 6.4 155 10.0 513 8.3 32.0
IFMs 240 47.7 630 40.5 1,227 19.8 17.7 Private Loans 125 24.9 415 26.7 813 13.1 20.6 Kye 115 22.9 215 13.8 414 6.7 13.7
Total 503 100.0 1,554 100.0 6,211 100.0 28.6
Source: The Survey Report o f Farm Household Economy, MAFF, each year.
percent in 1990, while that of private loans decreased from 24.9 percent to 20.6 percent.
The decreasing share of the Kye implies that its importance as a saving instrument has decreased as the financial services of the FFI have improved The positive real interest rates paid on savings in the FFI must have attracted some of the informal market funds.
^ According to a recent survey, the main purpose of the Kye was reported to be to maintain friendship and mutual help rather than to use it as savings instruments. Sul and Kee (1988) reported that 77.4 percent of surveyed farm households were using at least one Kye and 74.7 percent of them used the Kyes 56
On the other hand, the dominant role of the PACs in the credit markets is clear when the borrowing sources of farm household debts are considered. In the total farm household debts at the end of 1990, 86.1 percent were borrowed from FFMs and 76.3 percent were borrowed from the PACs (Table 3.7). The composition of sources of loans has substantially changed as the debts have rapidly increased, which is similar to the trend noted in the savings markets. The debts of a farm household, on average, increased from 338 thousand won in 1980 to 4,734 thousand won in 1990, which is equivalent to a
30.2 percent annual increase. Most of the increase in loans were borrowed from the agricultural cooperatives, especially from the PACs. The share of the PACs as a source o f debts increased from 40.8 percent in 1980 to 76.3 percent in 1990, while the share of the other FFIs slightly declined from 10.1 percent in 1980 to 9.8 percent in 1990. The market share of the FFIs increased from 50.9 percent in 1980 to 86.1 percent in 1990 due to the sharp increase in the share of the PACs. One interesting fact is that the declining share of the NACF was offset by the increasing share of the other institutions except for the PACs and the NACF, which means that the other institutions became important borrowing sources for the farm households. The increasing rate of indebtedness from some of non-agricultural cooperative institutions considerably exceeds that of the total debts and even that of the PACs, though the amounts were still very small (Table 3.7).
The share of the rural IFMs in both savings and borrowing markets substantially declined in the 1980s due to the rapid growth of the PACs. The reasons for this decline must be related to the positive real interest rates of FFMs and the improved financial services of the financial institutions through technological innovation that substitutes for to maintain friendship and mutual help. Rho reported that the share of farm households which were using the Kye as a savings instrument was 8.8 percent of the surveyed farm households in 1987, which contrasted with 22.9 percent in a 1977 survey by Chang. 57
Table 3.7 Borrowing Sources for an Average Farm Household by Type of Financial
Institution
Annual Financial Institutions 1980 1985 1990 Growth Rate Amount % Amount % A m ount % 1980 ~ '90 Uiousand Uiousand Uiousand % w on won w on FFMs 172 50.9 1,440 71.1 4,078 86.1 37.2 PAC 138 40.8 1,191 58.8 3,612 76.3 38.6 NACF 27 8.0 146 7.2 245 5.2 24.7 Other Banks 1 0.3 17 0.8 36 0.8 43.1 NCFA 4 1.2 15 0.7 48 1.0 28.2 Credit Unions 1 0.3 7 0.3 30 0.6 40.5 Others 1 0.3 64 3.2 107 2.3 59.6
IFMs 166 49.1 584 28.9 656 13.9 14.7
Total 338 100.0 2,024 100.0 4,734 100.0 30.2
Source: The Survey Report o f Farm Household Economy, MAFF, each year.
some of the advantages of the IFM. Furthermore, the successful absorption of rural IFM funds by rural FFIs reduced the pervasive interest rate gap between rural IFM and FFM in the 1980s as shown in Figures 3.2 and 3.3. The trend in interest rates in both the formal and informal markets revealed a relatively constant gap between the two markets until the real interest rates of the FFMs became positive and substantially increased in the early 1980s. This fact supports the argument that the successful performance of the rural
PACs enhanced the efficiency o f financial intermediation in the rural financial markets
(NACF, 1992). On the other hand, the contribution of the rural IFMs to financing farm 58 households was still substantial even though the growth rate o f the IFM was not as fast as that of the FFM. The absolute size of savings or borrowings from the IFM steadily increased as presented in Tables 3.6 and 3.7.
3.2.5 The Factors Explaining the Successful Growth of the PAC Banking and Prospects for the Future
As discussed in the previous section, the PACs became the dominant financial intermediaries, in both the savings and credit markets, by absorbing most incremental funds in the rural financial markets. Furthermore, banking by the PACs successfully grew nationally. What are the main factors for this success, and will these factors be effective in the future ?
The successful savings mobilization of the PACs can be explained by the following factors. First, the PACs have a diverse relationship with their members, who are most of the rural residents, through providing more services related to production, marketing, purchasing of consumer goods, etc., allowed the accumulation of clientele information while maintaining relationships. The PAC functions as an one-stop center that provides diverse services in one building, which enables the clientele to save transaction costs when they have other transactions in the non-banking sectors (Meyer,
1984). Furthermore, the PAC that has the most extensive communication channels with its members is one of the most influential institutions for rural residents, and provides the employees of the PACs with sufficient opportunities to have good relationships with the residents. Good human relationships are a very important influence on savings mobilization in rural artas (Suh and Park; Sul and Kee; NACF,). 59
50 -r
FM
HCliOAH 40 HCTD 35 . . BNKTD %30-.
.S 20
C l 5 f
10
1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
Nole: IFM = Rural informal financial market; MCLOAN = Mutual Credit loans;
MCTD = 1 year time deposits at PACs; BNKTD = 1 year time deposits at banking institutions
Figure 3.2 Trends in Nominal Interest Rates in the Rural Financial Markets
20 FH
MCLOAN
Î HCTD .1 BNKTD g
76 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990
- 1 0 -L
Figure 3.3 Trends in Real Interest Rates in the Rural Financial Markets 60
Second, the PACs have competitive advantages in the security of savings, convenience of savings transactions, and profitability of savings (Lee, N.S.). There have been no cases of insolvency in the PACs, but there were many cases of bank failure in
MSFCs, CUs, or NCFAs from the late 1970s to the early 1980s. In addition, the PACs provide clientele with more interest income through tax exemption as well as higher interest rates than banks do. These factors must have been effective in urban financial markets.
Third, the growth of income and commercialization of farming increased the liquidity of rural areas, and the PACs could effectively absorb the increased liquidity because of the factors mentioned above. The commercialization of farming, since the mid
1970s, enhanced liquidity in rural areas through the monetization of the rural economy as well as increased income of farm households.
Fourth, the development of social infrastructure stimulated market integration that expanded market area and reduced the clientele's transaction costs fi'om traveling, communication, monitoring, etc., as discussed in 3.1.2. In the past, the geographically segmented rural financial markets undermined the transactions between the PACs and remote area residents. This physical barrier has been lowered with the development of social infrastructure such as transportation facilities and communication media.
Fifth, positive real interest rate savings in FFIs attracted informal financial market funds. The outflow of funds from the IFMs reduced the gap in interest rates between the
FFM and the IFM.
Some factors that contributed to the successful mobilization of savings may be no longer effective in the current phase o f market competition or in the near future. Other savings institutions, major competitors of PACs, have healthier systems against insolvency problems than previously, through strong government supervision, audit, and 61 establishing special funds to meet bank insolvency (Suh, C.H.). The PACs have well trained employees, better equipment, and better technology compared to CUs, NCFAs,
PCs, and LCs. This gap will not be easily removed but reduced as other institutions expand investments in high technology and training. In the near future, except for the convenience o f savings and the advantages of a nationwide network between the PACs and the NACF, the other competitive factors may not be favorable to PACs. If complete financial liberalization is achieved, the competitive advantage o f high interest rates compared to banking institutions will disappear, and the banking institutions may extend their services to households that are major clientele o f savings institutions. Small savings institutions like PACs will face tough competition in order to secure a good clientele
(Rhu et al.). Moreover, though the PACs are still the largest savings institutions holding
34.3 percent of total deposits o f savings institutions in 1990, the advantage of a large market share is likely to decrease in the future because of the decreasing weight of rural financial markets. This means that the PACs should make more efforts to be internally efficient through optimization o f scale and product mix.
3.3 Agricultural Cooperatives and Banking Operation
3.3.1 Historical Background
The current agricultural cooperatives were established in 1961 by amalgamating the Agricultural Bank and the Agricultural Cooperatives ( the old agricultural cooperatives). Before amalgamation, the Agricultural Bank was assigned to agricultural credit businesses, while the old agricultural cooperatives were assigned to operate agricultural marketing business. However, the performance of the old agricultural 62 cooperatives was too poor to be viable because of a shortage of funds and insufficient management skills. The Agricultural Bank did not adequately coordinate with the old agricultural cooperatives, even though it was designed to support the cooperatives. In effect, the old agricultural cooperatives were mere fertilizer agents of the government, while the Agricultural Bank was an exclusive credit organization in the rural society. The conflict between the Agricultural Bank and the old agricultural cooperatives resulted in severe criticism, which forced the two institutions to be consolidated.
The newly established agricultural cooperative had a three-tier system in its structure: village cooperatives (Lie or Dong cooperatives) as PACs, county cooperatives
(Kun or city cooperatives) as the regional federations, and the National Agricultural
Cooperatives Federation (NACF). In the early stages, the village cooperatives had an average of less than 100 members which was too small to operate such diverse activities as farming and livelihood guidance purchasing, selling, banking, cooperative insurance, processing farm products, etc. (Yun). Although 2.2 million of the 2.5 million farm households were members of agricultural cooperatives, they were not active but members in name only. Thus, most cooperative business was handled at the county level cooperatives.
The merger of the primary cooperatives (Lie or Dong cooperatives) was undertaken from 1969 to 1973. This reduced the number of primary cooperatives from
16,089 in 1968 to about 1500 in 1973, and increased the average membership per PAC from 139 to 1400. According to this reorganization, the system o f one PAC per Myoun or Up was set up. Along with the enlarged scale and improved management skills, the
PACs could effectively handle the cooperative activities assigned to them. In this period, the Mutual Credit program was introduced to enhance the efficiency of rural financial
^ For example, education for rational consumption, clothing, food, health, housing, etc. are included in the livelihood guidance (Yun). 63 intermediation by absorbing rural savings and lending those funds at lower interest rates than informal financial market rates, to secure a certain volume of operations that would provide for managerial viability, to strengthen the membership through understanding the functions and roles of cooperatives, and to help rural residents to increase their propensity to save (Yun). This program was successful and substantially contributed to the growth o f PACs. By 1977, most PACs were rated as viable in terms of this basic management (Yun). As the PACs grew, the functions of county cooperatives gradually tended to weaken and overlapped with those of PACs. Accordingly, the county cooperatives were reorganized as branch offices of the NACF and the PACs took over warehouses, branch offices and other farmer-related service facilities from the former county cooperatives in 1981. The current system is a two-tier system, the PACs and the
NACF.
The agricultural cooperatives had been under the control of government, so that many members perceived the PACs as not theirs but government institutions. However, the change of the presidential election system for PACs and the NACF in 1989 improved the managerial autonomy of the agricultural cooperatives by permitting the members to choose their presidents.. In the past, the president of NACF who was appointed by the government appointed the presidents of PACs. These reforms meant that the PACs should be responsible for their viability instead of relying on government support.
3.3.2 Organization
The organizational structure of Korean agricultural cooperatives is shown in
Figure 3.4. Farmers organize PACs and special agricultural cooperatives (SACs), both of which in turn are authorized members of the NACF. The members of PACs consist of 64
Mutual Credit National Agricultural Special Account Cooperatives Federation (NACF) (MCSA)
Provincial I City Office
County OfTice a NACF: Banking Instltulton Branch Ofiice O Savings Institution
Primary Agricultural Special Agricultural Cooperatives (PACs) Cooperatives (SACs)
Branch Office
Farm Households
Figure 3.4 Organizational Structure of the Agricultural Cooperatives
farmers who live in a specified administrative area such as an Up (a center town of a county level) or a Myoun (township), where a single PAC is established. Thus, a financial market area for the primary cooperative is determined by the administrative area
(Suh and Park). The SACs consist of farmers who are engaged in special farming enterprises such as vegetables, fruits, fiowers, etc. In 1991, the numbers of PACs and
SACs were 1,425 and 41, respectively. On average, a PAC had 1,376 members and about 30 employees. 65
All PACs have 3 business departments, an extension and an administration department. The activities of each business department will be explained below. The extension departments consist of a guidance department and a women's department. The guidance department conducts farming guidance, education and publicity, income development projects, while the women's department conducts the activities related to women such as guidance on the improvement of living conditions.
The NACF had 15 provincial or city offices, 143 county offices, and 341 branch offices in 1990. All offices of the NACF except for 9 provincial city offices offer banking services. The NACF performs such non-economic functions as management guidance to the PACs, auditing of PACs, research and policy-related activities as well as many kinds of business activities.
The NACF has Mutual Credit Special Accounts (MCSA) for operating payment reserves or managing fund surpluses of PACs. At least 10 percent of savings mobilized by PACs must be deposited at the MCSA as reserves for payments. If PACs have fund surpluses available for more than three months after making loans, they can deposit those funds as time deposits at the MCSA. Temporary surplus funds are kept in the internal fund account between PACs and the NACF including local offices. All the required reserves deposited at the MCSA must be deposited at banking institutions, while time depoisits at the MCSA are either loaned to PACs with fund shortages or invested in securities with high interest rates. The interest rate of required reserves of PACs is 11 percent per year and that of time deposits is 12.5 percent (NACF, 1992; MAFF, 1990). 66
3.3.3 The Organizational Structure of a PAC
The business activities of a PAC can be divided into four sub-sectors: marketing, purchasing, banking, and cooperative insurance. The marketing business covers a very broad range such as selling farm products of members, operating warehouses, providing transportation services, processing farm products or manufacturing farm inputs, and operating farm machinery joint utilization facilities. Marketing activities are closely related to farm production so that the cooperative members regard them as the most valuable businesses. But the overall performance of the marketing business has not been successful in achieving its objective to promote the sales of agricultural products on more favorable terms and conditions in the interest of the cooperative members. The purchasing business deals with farm inputs and consumer goods. The farm inputs include all kinds of materials such as feed stuffs, farm machines, gas, pesticides, fertilizer, etc.
The purchasing of consumer goods was started to protect farmers from merchants who had monopolistic power in remote rural areas. With a chain-store system introduced in
1969, this business has steadily grown. Cooperative insurance which has grown rapidly since 1980 deals with life insurance and other insurance agmnst losses such as fire insurance, machinery insurance, livestock insurance, etc.
The banking business covers 11 types of deposit accounts, 7 types of Mutual
Credit Loan accounts that are cooperative loan commodities independent of government policy loans, and about 100 types of policy loan accounts. Two of the eleven deposit accounts are demand deposits, and the others are savings and time deposits. The savings and time deposits can be used as collateral for loans. Three of the seven loan accounts are connected with a particular type o f deposit as collateral. The deposit and saving 67 services are open to anyone. But loans to non-members are limited to one third of the total loans made by a PAC excluding the policy loans of government.
The Mutual Credit loans can be classified into three categories; short-term loans, mid-term loans, and Mutual Credit policy loans. Mid-term loans have a term o f five years with 2 year grace terms and have been supplied since 1986 but this share in total Mutual
Credit loans is trivial. Mutual Credit policy loans refer to the cooperative funds that are used as policy loans when there is a shortage of policy loan funds. Except for the mid term loans, however, the interest rates on all loans are the same.
The policy loans have many kinds of sources and targets. The NACF provides these funds to the PACs and the borrowers of these loans are targeted by the government so the PACs incur few operating costs in receiving the funds and selecting the borrowers. However, the volume of funds is substantial, and managing the accounts and collecting the payments is undertaken by the PACs, so this requires considerable labor. On the other hand, PACs deposit required reserves and residual deposits in the
MCSA in the NACF, or may borrow funds from the MCSA if their mobilized deposits are too low to meet loan demand.
Table 3.8 presents the average total assets, amount of equity, and number of members and employees o f the PACs. The PACs have on average of 16.5 billion won
(Korean money unit) in assets that was equivalent to 22 million U.S. dollars in 1991.
Equity represented only 3 percent, so the PACs have a very high debt-equity ratio which is related to the dominant share of banking activities in the total operation o f the PACs.
The share of members' investments to total equity was only 31.9 percent, which implies that the growth o f a PAC depends significantly on the capital accumulated from business surpluses. For the 1980s, the PACs grew rapidly; for example, average total assets grew at a 24.8 percent annual rate in the period from 1980 to 1991. 68
Table 3.8 The Growth of Assets, Equity, Members, and Employees of an Average PAC, 1980 to 1991
1980 1985 1990 1991 Annual Growth Rate (1980 ~'91)a Total Assets (million won) 1,434 3,757 13,446 16,462 24.8
Fixed Assets 110 269 692 962 21.8
Equity (million won) 112 284 465 536 15.3
Members' Investment 85 125 159 171 6.5
Accumulated Surplus 16 137 248 295 30.3
Current Surplus 10 23 58 71 19.0
Members (persons) 1,302 1,416 1,375 1,376 0.5
Employees (persons) 17.6 20.7 27.8 30.0 5.0
M embers per Employee 74.2 68.4 49.4 45.9 -4.3 (persons)
Source: Managerial Performance Report o fNACF, PACs, each year.
® The unit is percent.
The average membership in the PACs has grown slowly and there are a number of inactive members In contrast, the number of workers substantially increased during the period, so that the number of members per employee decreased from 74.2 persons in
^ In fact, many members do not regard membership in a PAG as valuable. Many farmers do not know tlieir investments or dividends from the investments, which reflects farmers recognition of PACs. Thus, when they moved out from mral to urban areas, they do not report either the migration or an intention to give up their membership. Thus, PACs often have a larger number of members than farm households that exist in their authorized areas. 69
Table 3.9 Labor Allocation within a PAC (National Average)
Types of Cooperative Average Number of Standardized Number of Activities Employees Employees ^ Number Percent persons persons %
Banking 11.17 9.36 35.0 Policy Loans 1.03 0.97 3.6 Others 10.14 8.38 31.3
Purchasing 6.45 5.22 19.5 Farm Inputs 3.50 2.96 11.1 Consumer Goods 2.95 2.26 8.4
M arketing 4.65 3.58 13.4 Farm Products 1.71 1.54 5.7 Others 2.94 2.05 7.6
Cooperative Insurance 1.03 0.92 3.5
Extension 1.79 1.68 6.3
Administration 5.24 5.99 22.4
Total 30.34 26.76 100.0
Source: The Managerial Performance Report o fNACF, PACs, 1992.
^ Refers to the number of employees as the normalized number of third class employees. A detailed
explanation is presented in Chapter IV.
1980 to 45.9 persons in 1991. This suggests that the cooperatives could provide more services to their members (Huh).
Table 3 .9 presents labor allocation to each of the PAC activities using national average data for 1991. On average 11 out of 30 total employees worked in banking and 70 one o f them worked on policy loans, while 6 and 5 persons worked in the purchasing and marketing sectors, respectively. In addition, 1 person was allocated to the cooperative insurance business. About 7 persons also worked for extension and administration activities. This information on labor allocation is used to evaluate the economic performance of each business sector. Since the characteristics of multipurpose cooperatives make it hard to analyze the input-output relationship of each business sector separately, each PAC is required to record labor allocations to each cooperative activity.
Considering the structure and economic performance of the individual business sector, the banking business has been very significant in contributing to the rapid growth of the PACs. Since there is no simple measure o f aggregate business size in multibusiness firms, the size of each business sector was measured to represent changes in the structure of a PAC using the criteria shown in Table 3.10. From 1980 to 1991, the end of year stock value of deposits increased at an annual rate of 30.4 percent, while the total business size increased at 26.2 percent. Hence, the share of banking to total business volume increased from 34.4 to 49.4 percent. This increased share suggests that PACs may emphasize banking more than other sectors or that banking was more profitable than other businesses. The growth rate of cooperative insurance was also very high, while the growth of the marketing and purchasing sectors was slow. The declining share of the purchasing sector was related to a policy change in the supply of fertilizer and farm machines that previously had been supplied through the agricultural cooperatives.
Beginning in the early 1980s, the government deregulated the supply of those farm inputs so the agricultural cooperatives lost monopolistic position.
The share of banking in total business gross income in 1991 was 73.1 percent when this measure includes indirect costs such as labor expenses of employees and 71
Table 3.10 The Changes in Business Structure of PACs (National Average)
Business Sector 1980 1985 1990 1991 Annual Growth Rate (1980 ~ '91)
Banking (Deposits) " 555 1,795 8,084 10,279 30.4 % (million won) % (34.4) (47.1) (48.9) (49.4)
M arketing ^ (million 308 655 1,931 2,417 20.6 won) % (19.1) (17.2) (11.7) (11.6)
Purchasing ^ (million 378 685 1,380 1,583 13.9 won) % (23.4) (18.0) (8.3) (7.6)
Cooperative Insnrance ^ 374 676 5,132 6,533 29.7 (million won) % (23.2) (17.7) (31.1) (31.4)
Total (million won) 1,615 3,811 16,527 20,812 26.2 % (100.0) (100.0) (100.0) (100.0)
Source: Managerial Performance Report o fNACF, PACs, each year.
Note: The number in parenthesis is the share relative to total volume. The totals represent a simple
summation of the total sectors without regard to the conceptual differences between activities
such as banking and insurance.
® The stock value of end of year deposits.
^ Total turn over.
^ Amount of insurance contracts outstanding. 72 capital expenditures. This source of gross income grew much faster than the other business sectors (Table 3.11). The annual growth rate of banking gross income was 25 percent from 1981 to 1991, while that of the marketing and purchasing sectors was 13.1 and 13.7 percent, respectively. However, criticism about the biased structure towards banking has increased as banking has become more important. Many cooperative members and scholars have argued that more effort should be placed on the marketing sector in order to promote the sales of agricultural products on more favorable terms and conditions. They argue that a more important agricultural problem is that farmers cannot sell their products at appropriate prices because of inefficient marketing channels and unstable market conditions.
The marketing performance of the agricultural cooperatives (both the NACF and
PACs) has not satisfied members' expectations. The cooperatives have not effectively
Table 3.11 The Growth of Business Gross Income of PACs (National Average)
Y ear Marketing Purchasing Banking Cooperative Total Insurance (thousand won) 1981 20,289 23,956 52,438 3,987 100,670
1985 26,804 28,260 107,750 7,130 169,944
1990 61,186 69,272 366,995 19,326 516,779
1991 69,209 86,521 487,705 23,655 667,090
Annual Growth 13.1 % 13.7 25.0 19.5 20.8 Rate (1981 ~ '91)
Source; Managerial Performance Report o fNACF, PACs, each year. 73 competed with merchants for either agricultural products or inputs but they have been successful in financial markets. Whether each business sector is profitable or not can be used as a criterion of viability. But it is not simple to evaluate the profitability of each individual business because of indirect costs that cannot be directly allocable.
The NACF uses some alternative measures to evaluate profitability by using the concept of the break-even point in business size. Table 3.12 presents two alternative break-even measures based on specific assumptions. The first alternative assumes that each business sector should at least cover its direct costs and labor costs. The second column of Table 3.12 represents the break-even business size per employee for each sector, and the fourth column represents the actual business size per employee.
Therefore, the difference between the second and the fourth column (the fifth column) indicates whether each business presents profits or deficits. According to this criterion, many business sectors presented positive profits, except for fertilizer, pesticides, utility business such as repairing farm machines, and transportation. But this criterion does not reflect capital costs used in the business or the labor costs allocated to the administrative department. Therefore, the values in the fifth column have limitations.
The second method assumes that the business revenues should cover both business costs and extension costs including administrative costs. Since the extension activities cannot earn income, the income received from businesses should cover the costs of extension. For these reasons, the second assumption is a more reasonable criterion of business viability for cooperative activities. The third column in the table presents the value of business size depending on the second assumption and the last column indicates profitability. In this case, all business sectors except for the banking and the warehouse services presented negative profits, but the positive profits of the 74 warehouse sector is negligible. In other words, the banking sector is the only profitable and sizable income source for PACs and these activities subsidize the others.
Table 3.12 The Break Even Point Business Size per Employee in 1991
Break-Even Point Business Business Size per Employee Actual Difference Labor Cost Extension Cost Business C-AC-B Base (A) Base(B) Size (C) Banking (million won) Deposits 273 740 978 705 238 Purchasing Fertilizer 344 935 221 -123 -714 Pesticides 167 454 102 -65 -352 Consumer Goods 167 453 285 118 -168 Others 381 1,035 450 69 -585 M arketing Farm Products 1,144 3,106 1,525 381 -1581 Processing 55 151 94 39 -57 Warehouse 18 51 55 37 4 Utilization 23 62 14 -9 -48 Transportation 33 90 18 -15 -72
Cooperative Insurance 214 582 473 259 -109
Source; Managerial Performance Report of NACF, PACs, 1992, p.l33.
Note: The value bases of banking, cooperative insurance, and marketing and purchasing business size
are average stock of deposits, insurance premium, and the values of total sales. 75
3.4 Summary
Successful economic growth and gradual financial liberalization has brought about remarkable financial deepening of the Korean economy and the development of the financial industry. In the process of financial deregulation, the entry barriers were lowered and many restrictions on financial institutions were abolished. But a good deal o f regulation still remains and the financial institutions are not completely autonomous in decision making.
The banking business of the PACs successfully grew in urban financial markets as well as rural financial markets so that efficiency of financial intermediation increased as represented by lower market interest rates. In addition, the growth of the banking sector significantly contributed to the viability of the PACs. The other business sectors of the
PACs did not achieve satisfactory performance in the sense that they had to be subsidized and their performance did not satisfy member expectations.
In these circumstances, it is clear why the PACs are sensitive to the changes being made in the financial market environment through financial liberalization. It seems that these changes so far have not been unfavorable to the PACs. They still have some competitive edge over other savings institutions as well as banking institutions which could actually cause inefficiency in the utilization of inputs if there is a lack of competition. But, many PACs may not reach an optimum scale or product mix due to the remaining regulations of the financial industry.
However, if financial liberalization continues, then the competitive edge o f the
PACs will disappear. Large financial institutions such as banks will have competitive a edge through economies of scale and scope. The current price advantage of savings institutions such as the PACs will be abolished; then the interest rates in markets will 76 represent not the regulated prices but the minimum production costs o f financial intermediaries. These changes will substantially affect the viability of the PACs through decreasing production costs of financial services. Any inefficiency o f either input utilization or production scale and product mix will be a heavy burden to the PACs.
Therefore, this investigation into the existence of production efficiency should be valuable for the PACs as well as for policy makers concerned about the financial markets in Korea. CHAPTER IV
DATA AND DESCRIPTIVE ANALYSIS OF BANKING COST STRUCTURE
This Chapter will focus on explaining the data set that will be used in the empirical analysis and on descriptively analyzing the PAC business structure, i.e., the relationship between the banking sector and the other business sectors of a PAC, and the banking cost structure of the sampled PACs. The explanation about the data set and of the sampling procedure will help in the evaluation of the sample information. The sampling procedure was designed to represent a typical rural province of Korea which includes small and medium cities. Since there are substantial differences in the rural and urban financial markets, we may not treat all observations with the same weight in order to obtain reasonable information about the banking production of the PACs. For this reason, a second province was added to supplement the small size of the urban sample of PACs.
A comparative analysis is needed to determine the structural differences between the rural and urban PACs. This analysis will provide important information about the characteristics of the observations before preceding to econometric analysis. Furthermore, it is desirable to suggest appropriate policy implications for not only the specific sample area but also for the entire nation. This possibility is explored through simple comparative analyses of the average values of the sampled PACs and the national average for some indices.
77 78
In the first section, the data set is explained and the regional distribution of observations will be presented. The second section will focus on differences in the business structure of PACs between the sample and the national average, and between rural and urban samples. The first comparison will concern the operating characteristics of a PAC such as equity, labor, members, population size o f a market area, etc. Following the discussion, the accounting system of balance sheet and gross business income will be presented. In the third section, the cost structure of banking costs will be discussed as an introduction to the analysis of banking cost frontier. This section is planned to give an intuitive understanding of the banking cost structure rather than on theoretical interpretation. Therefore, the analysis may help to specify the model in Chapter V and to explain the results from evaluating the cost frontier surface that will be presented in
Chapter VI. The fourth section will summarize the main points from the discussion in this chapter.
4.1 Data and Sampling
4.1.1 Data
The data used in the empirical analysis are divided into a main data set and a supplemental data set, which contain information about the performance of PACs in
1991. The main data set consists of income statements (profit and loss sheet), balance sheets at the end of 1991, employee allocation sheets for 1991, a subdocument o f the income statement showing a detailed classification of operating costs, and the
Managerial Performance Report of PACs (MPRP) of 1992 by NACF. From the main data set, the value of all variables was obtained to estimate the frontier cost function. 79
The variables of total banking costs and input prices were obtained from the expenditure accounts of the income statement, while the output variables were obtained from the accounts of bank assets and liabilities in the balance sheet. The employee allocation sheets present labor allocation to individual business, extension, and administrative sectors by type of employee class. This information was used to calculate the banking portion of operating costs and capital as well as wage rates. The subdocument o f the income statement enables us to precisely classify operating costs into expenses for labor and for capital. By using this information, fnnge benefits, capital expenditures and other management costs can be appropriately classified. The
MPRP presents very diverse information about the individual PAC; that is, the number of cooperative members, the number of farm households, quasi-members, regional characteristic indices, the volume of each business sector, the number of branches, average stock value of deposits. Mutual Credit Loans, and policy loans, etc.
The supplementary data set was used to obtain market environment variables for each PAC such as the farm household ratio and population density. The data are the Annual Report of Population Survey by the Economic Planning Board (EPB), and the Annual Report of Kun (or city) Statistics of the sample administration area. To analyze sources of input inefficiency, information from the supplementary data set was used.
4.1.2 Sampling
The 211 PACs selected for this study were non-randomly sampled from two provinces, Choongnam and Kyounggi, of Korea. Choongnam is a province of a typical rural area located in the west of South Korea, while Kyounggi is the suburban province of 80
Seoul city that has many urbanized areas (see the map in Figure 4.1). All PACs in
Choongnam were sampled, including 5 cities, 22 Ups, and 148 Myouns, as administrative districts. In contrast, the PACs of Kyounggi were intentionally sampled to supplement the small size o f the Choongnam urban sample. Thus, the administrative districts of Kyounggi
Do (province) sample were chosen to represent a relatively urbanized region. The
Kyounggi sample consists of observations from 3 cities and 4 Kuns that have 8 Ups and
29 Myouns. Among them, all cities and one Kun that has four Ups and three Myouns are adjacent to Seoul city.
To classify the observations into rural and urban PACs, the narrow concept of rural area that was explained in Chapter I was applied to the empirical analysis o f this study; thus, all cities and Ups were classified as urban, while Myouns except for one adjacent to Seoul were classified as rural. The reason that the Up is defined as urban is that the financial market environment of the Up is close to that of a city rather than that of a Myoun. The financial market of the Up is relatively competitive. There are usually deposit money banks (branches of nationwide banks and local banks) as well as non-bank financial institutions such as many kinds of savings institutions and branches of life insurance companies in the Up, while there are usually two or three savings institutions, i.e., one PAC and one post office or others, in a Myoun. In addition, the Up has a relatively large population and non-agricultural industries, while the small population of
Myoun is geographically distributed in a large area and most of the Myouns are highly dependent on the agricultural sector. The high dependency on agriculture implies that the liquidity of the rural financial markets is lower than that of the urban markets. Therefore, there exist differences in the context of financial transactions as well as market environment, between the Up and the Myoun. These differences should affect the banking production of the PACs. 81
South Korea Sample Area
Kyounggi P q
^ S o / O ' Seoul Incheon
•■• Choongnam Do . y , ^ I . igs-m, • ♦ « • • • * . I i/ * • • ••’ •' ■(• Daejoeon
) p f
Kwoangju (b Pusan
&
'S d ^
Figure 4.1 Distribution of Sample Region 82
Table 4.1 Sample Size and Regional Distribution of PACs Included in the Study
Administrative Area Regional Distribution of Sample Total
Choonginam Kyounggi Observations
Urban 27 17 44
City 5 8 13
Up 22 8 30
Myoun 1 1
Rural (Myoun) 139 28 167
Total 166 45 211
Along the regional classification criteria, the samples consist of 167 rural and 44 urban PACs. However, the data for 3 PACs were not complete; that is, some information was missing from the data sets. In addition, although 208 observations were used to estimate the frontier cost function, there were two outliers which were excluded so that
206 observations were used in the empirical analysis. Among the excluded observations, the rural and urban PACs are 4 and 1, respectively. 83
4.2 Operating Characteristics of the Sampled PACs
4.2.1 Equity, Labor, and Market Environment Variables
The differences between rural and urban financial market environments mentioned in the previous section may affect the PAC banking production in terms of the size of banking business, the production efficiency, and the production technology. The favorable conditions of the urban financial markets will allow the PACs to increase the scale of banking production. The high competitiveness of the urban market may lead the PACs to be more efficient. These differences may result in different production technologies to adapt appropriately to the market conditions.
The size of banking production can be easily captured by a descriptive analysis, but the differences in the production efficiency and technology may not be easily captured in the regime o f multiple output production. If the urban PACs have more efficient technology so they can produce outputs more cheaply than the rural PACs, then the cost function curve of the urban PACs should be lower than that of the rural PACs. But even if we do not observe any evidence of the differences in efficiency or technology by the descriptive analysis because of the complicated cost structure of multiproduct firms, we may find some evidence from the econometric model that is consistent with theory.
This chapter will explore descriptively this kind of evidence. If this evidence is found, we would have sure intuition about the PAC banking production and can choose proper technologies to deal with the estimation of the frontier cost function. For that reason, this section will conduct comparative analyses between rural and urban PACs, concerning how efficiently rural PACs produce financial services. Furthermore, another comparative analysis between the overall sample means and national averages is added to 84 evaluate information from the sample in terms of its applicability to nationwide policy formation. Since the observations were selected from specific areas, the sample information may be biased in the case of the specific areas. Therefore, it is desirable to evaluate whether the results of the analysis are representative nationwide. The results of this comparison are presented first as follows.
If the operating characteristics of sample PACs such as business volume, composition of businesses, own capital (equity) or labor input, etc., are not significantly different from the national average, the sample information would be applicable for nationwide policy recommendations. In this section, the comparison will focus on the general indices of cooperative operations such as equity of a PAC, the number of employees, the number of members, population size o f the market, etc. These are the basic variables used to measure cooperative activities. Table 4.2 presents differences in these factors between overall sample means and national averages. Generally speaking, there are no substantial differences in these variables; that is, the differences in equity, the number of employees, and the number of members are within 5 percent of the national average.
Although the difference in the number of branches of the sample is about 12 percent, it is meaningless since the difference comes from less than one branch. On the other hand, there are relatively big differences in the number of quasi-members, members with no right to vote*, current surplus in 1991, and members' investments, which members financed to supply funds for their cooperative's own capital. The difference in current surplus comes from the data base. The average of the sample does not reflect dividends to members, while the nationwide average excluded them. Thus, it may not be significantly different from the national average. In addition, the number of quasi-members does not
* Most clientele of urban PACs are non farmers so that they cannot be regular members of PACs. Some of them who invest as money as a regular member's investment are treated as members in terms of receiving economic benefits from the PAC such as dividends, but they cannot participate in decision making of the PAC as a member. 85
Table 4.2 Selected Characteristics of PACs (Equity, Employees, and Market
Environment Variables), Sample and National Average, 1991
Characteristic Rural PACs Urban PACs Total National Samples Variables Value % Value % Value % Averages B/A JA)______(Average Value per PAG) Equity (million won) 393 100.0 1,122 100.0 547 100.0 543 0.99 Member's Investments 128 32.6 232 20.6 150 27.4 177 1.18 Accumulated Surplus 189 48.1 669 59.6 291 53.1 295 1.02 Current Surplus 76 19.3 222 19.7 107 19.5 71 0.67
No. of Employees 24 100.0 51 100.0 29 100.0 30 1.03 (persons) Banking 7 29.2 22 43.1 10 34.5 11 1.05 Non-banking 17 70.8 29 56.9 19 65.5 19 1.00
No. of Members 1,326 1,897 1,451 1,376 0.95 (persons) Quasi-members 123 1,915 441 738 1.67
Banking Branches 0.4 1.6 0.7 0.7 1.12
Population of a market 7.4 38.1 13.9 ._ area (thousand persons )
Source: Survey Data; Management Performance Report o fNACF, PACs, 1992.
Note: The won is money unit of Korea. 8 6 have an important meaning in a practical sense, and the overall sample information is not likely to be biased from the national average level due to the differences in the values of members' investments. Therefore, if we consider differences in these few variables as not so important, the sample information can be used for nationwide use.
Table 4.2 provides information about the differences between urban and rural
PACs, and compares the average FAC with the national average. The average value for the urban sampled PACs are about 2 - 3 times as big as those for the sampled rural PACs.
The difference may reflect differences in market environment as well as production efficiency. The market environment is much more favorable for the urban than the rural
PACs. For example, the rural PACs are restricted by market size in one or two administration areas with small population; As presented in Table 4.2, the population size per rural PAC is 7,000, but that per urban district is 38,000. Furthermore, income level, total liquidity, and high population density are conditions favorable to urban PACs.
Therefore, the market size for the urban PACs may not be a restriction for them to increase their production scale. The difference in the number o f branches also represents the difference in market environment. The branch of the urban PACs is open to maximize business surplus, but that of the rural is open for members' convenience in remote areas^.
The urban PAC allocates a greater portion of labor to the banking sector (43.1 %) than rural PAC (29.2 %). This can be explained by differences in market environment and individual business cost structure. The rural PAC that must carry out more activities for producers who are widely distributed needs more labor for non-banking sectors, i.e., marketing, purchasing, extension, etc. Most non-banking activities can be done outside of the office so that the rural PAC may need more labor to carry out the same activities.
2 The managers of urban PACs, who were interviewed, said that new branches should be over the break even point of banking business within 2 years, but those of rural PACs said that new branches were open for members' convenience in remote areas rather than business surplus. 87
According to Huh who analyzed the PACs of Kyounggi Do, the non-banking business
sectors had significant scale economies, while the banking sector had constant returns to
scale. This result suggests that since the business volume of the urban PAC is much larger than that of the rural PAC, differences in labor productivity between rural and urban can be higher in non-banking business sectors than in banking business sector. That is, the
difference in labor allocation share between rural and urban might be related to these
characteristics of individual business cost structure.
The composition of equity is also different between rural and urban PACs. The
members' investment share to total equity of the rural PAC is 32.6 percent, but that of the
urban is 20.6 percent, which implies that the rural PACs are more dependent on their
members to provide the funds used in the business. In other words, it simply reflects that
since the business conditions of rural PACs are unfavorable, so the rural PAC accumulated
less capital from business surplus.
4.2.2 Balance Sheet Structure
Tables 4.3 and 4.4 present the structure of the balance sheet which compares
average values for rural and urban sampled PACs. There is little difference, between rural
and urban PACs in their basic structure such as the share of banking and general business
accounts that includes all accounts except for the bank accounts. At the overall sample
mean, the banking asset share to total assets is 86.7 percent (Table 4.3), while the banking
liability share is 91.6 percent (Table 4.4), which suggests that some of banking liabilities
used for non-banking assets were larger than the amounts of some cash reserves for the
non-banking activities. For example, the banking liability for the non banking sector
includes borrowings or subsides from the NACF to invest in fixed assets or income 88
Table 4.3 Balance Sheet: Average Value of Assets Held by the Sample PACs
Account R ural % Urban % Total %B/A (A) (B) Samples (million won) Banking 10,886 86.0 27,283 88.0 14,355 86.8 2.51 Deposits at MCSA 1,271 10.0 5,999 19.4 2,271 13.7 4.72 Mutual Credit Loans 5,369 42.4 15,629 50.4 7,539 45.6 2.91 Policy Loans 3,680 29.1 4,526 14.6 3,859 23.3 1.23 Others 565 4.5 1,128 3.6 684 4.1 2.00
General Business 1,773 14.0 3,715 12.0 2,184 13.2 2.10 Business Claims 597 4.7 643 2.1 607 3.7 1.08 Inventory 283 2.2 435 1.4 315 1.9 1.53 Coop. Insurance Loans 217 1.7 434 1.4 263 1.6 2.00 Fixed Assets 587 4.6 1,997 6.4 885 5.4 3.40 Others 88 0.7 206 0.7 114 0.7 2.34
Total 12,659 lOO.fl1 30,998 100.0 16,538 100.0 2.45
Source: Survey Data
generating projects for members.
On the other side of the balance sheet, the banking share of rural PAG (86 %) is slightly lower than that of the urban (88 %). On the other hand, there is no difference in the share of liability between rural and urban. This figure is contrasted with the labor allocation share, i.e., higher share of labor allocation to the banking sector in the urban
FAC than the rural FAC, as presented in Table 4.2. This contradiction can be explained by three factors: differences in market environment, different cost structures between banking 89
Table 4.4 Balance Sheet: Average Value of Liability and Equity per PAC
Account Rural % Urban % All % B/A w (B) (million won) Banking 11,594 91.6 28,368 91.5 15,142 91.6 2.45 Deposits & Savings 6,674 52.7 22,557 72.8 10,034 60.7 3.38 Borrowings 4,569 36.1 4,971 16.0 4,654 28.1 1.09 Others 351 2.8 840 2.7 455 2.7 2.39
General Business 672 5.3 1,507 4.9 848 5.1 2.24 Business Liability 214 1.7 439 1.4 262 1.6 2.05 Coop. Insurance 219 1.7 441 1.4 266 1.6 2.02 Others 44 0.4 122 0.4 61 0.4 2.74 Allowances 194 1.5 505 1.6 260 1.6 2.60
Equity 393 3.1 1,122 3.6 547 3.3 2.85 Contribution by Members 128 1.0 232 0.7 150 0.9 1.80 Accumulated Surplus 189 1.5 669 2.2 291 1.8 3.54 Current Surplus in 1991 76 0.6 222 0.7 107 0.6 2.92
Total 12,659 100.0 30,998 100.0 16,538 100.0 2.45
Source: Survey Data
and other businesses, and different market value of assets between rural and urban. The first reason is that the rural market environment needs more labor to conduct the non banking activities, as explained in the previous section. The second also was suggested in the previous section; that is, this relationship implies that an employee of an urban PAC works more efficiently in the non-banking business sector because of the existence of 90 greater scale economies there than in the banking business. Tables 4.3 and 4.4 may support this hypothesis; that is, the banking employee number o f the urban PAC is 3 .1 times that of the rural PAC, but the ratio of banking assets is 2.5 times: On the other hand, the non-banking employee number o f the urban PAC is 1.7 times that of rural PAC, but the ratio of non-banking assets (the assets of general business) is 2.1 times. The fourth factor is that a high market value of land in urban areas results in a high value o f fixed assets so that non-banking asset shares might be higher in urban than rural PACs^. But this effect seems to be trivial when the composition of liability and equity is considered.
Although the differences in the account shares between the rural and urban PACs are small, there are substantial differences in subaccount composition between them.
Among banking assets. Mutual Credit Loans have the largest share in both rural and urban, but the magnitude of the share for the urban PAC (50.4 %) is greater than that for the rural PAC (42.4 %). Policy loans are second in the banking assets of the rural PAC
(29.1 %), while they are third in the case of the urban PAC (14.6%). Deposits in Mutual
Credit Special Accounts (MCSA) are more important assets to the urban PAC (19.4 %) than policy loans are. Comparing the deposit size at MCSA between the rural and the urban PAC, the value of deposits at MCSA for the urban PAC is 4.7 times that for the rural PAC, which substantially exceeds the ratio of total banking assets (2.5 times). This differences reflect the characteristics of rural and urban PACs.
That is, since there is no restriction on mobilizing savings, but there are restrictions on loans to non-members, the restrictions on loans to non-members must have an effect on the urban PACs with many non-member clients, resulting in greater fund surpluses after lending. But the restriction is not be effective on the rural PACs with a relatively small non-member clientele so that it has relatively little fund surplus. Consequently, the urban
^ The average price of land for the urban PACs was 9.6 times that of the rural PACs in 1987. 91
PACs play a role of mobilizing funds for not only their members but also for the members of other PACs. The MCSA plays the role o f a central bank by keeping required reserves and fund surpluses of member cooperatives and operating as a funds pool for PACs with shortages to fulfill demand for money in their markets.
Among the general business asset accounts, which includes all accounts of the
PAC except for the banking sector, there are few differences in uncollected business claims income from non-banking businesses and in inventories of non-banking sectors, such as materials, unsold goods, stored goods, etc. In contrast, the value of fixed assets, including tangible and intangible fixed assets and other assets, is substantially larger in the case of urban than in rural PACs
In the subaccounts of liabilities, the rural PACs highly depend on borrowings, comparing with the urban PACs, which is a reflection of asset accounts, i.e.. Mutual
Credit loans and policy loans. In the sub accounts of general business, there are no differences in the shares, but the value of each account for the urban PACs is about 2 - 3 times that of the rural. The share of total equity to total assets also shows no difference between the rural and urban PACs.
In summary, the average balance sheet structure suggests that the banking sector is the dominant one in both assets and liabilities and there are no large differences in the large categories of banking, general business and equity. However, the subaccount shares and their relative value present substantial differences between rural and urban PACs because of different market environments. In addition, the banking sector may have relatively less scale economies than non-banking businesses, when the labor allocation share and the large account shares are considered simultaneously. However, it is not possible to precisely evaluate the differences in production eflRciency with this limited information. 92
4.2.3 Gross Business Income
Banking provides a major source of income for the sampled PACs as discussed in
Chapter III. At the overall sample mean, 74.0 percent of this total business income comes from banking, which is close to the national average level (Table 4.5). The second and third most important sources of gross income are the purchasing (14.9 %) and marketing sectors (8.3 %). There is no large difference in this ordering between rural and urban
PACs. But the gross income share from banking for urban PACs (79.5 %) is higher than for rural ones (70.2 %), while the shares of other income are lower (Table 4.6). In contrast with the composition o f assets or liabilities, this result is relatively consistent with the labor allocation ratio as presented in Table 4.2. But the degree of difference in share is not as great as the difference in labor allocation share, which may suggest the existence of scale economies for non-banking businesses as discussed in the previous section.
The ranking of non-banking business shares represents the characteristics of the
PACs; that is, the highest ranking order of shares, in the case of rural PACs, are purchasing farm inputs, purchasing consumer goods, warehouse service, and selling members' farm products. In other words, although most members consider it desirable to promote sales of farm products, the business returns to PACs are too small, which reflects that the business is not competitively viable. Recently, the NACF and PACs made efforts to promote the marketing business, especially focusing on processing or selling farm products. But the business of selling farm products is very competitive and PACs do not yet have a competitive edge. Considering the other non-banking businesses, markets are very competitive with very little market intervention. Because they do not have a competitive edge in these businesses, the banking businesses of PACs also may not be 93
Table 4.5 The Structure of PAC Gross Business Income : Sample and National
Averages
Business Sector Sample Nation Average Cost Revenue Gross Income % Gross % (A) (B) (B-A) Income (million wott) Banking 1,291 1,783 493 74.0 488 73.1
Purchasing 1,661 1,760 99 14.9 87 13.0 Farming Inputs 1,066 1,122 55 8.3 44 6.6 Consumer Goods 595 638 43 6.5 42 6.4
M arketing 637 692 55 8.3 69 10.4 Farm Products 595 617 23 3.5 34 5.1 Processing 5 6 2 0.3 2 0.3 Warehouse 17 38 20 3.0 26 3.9 Utilities 2 3 1 0.2 1 0.1 Transportation 18 28 9 1.4 7 1.0
Coop. Insurance 35 54 20 3.0 24 3.5
Total 3,679 4,297 666 100.0 667 100.0
Source: Survey Data; Managerial Performance Report o fNACF, PACs, 1992.
Note: The cost in this table does not include indirect cost not allocable to each sector so the gross income should not be understood as profits. 94
Table 4.6 Differences in Average PAC Gross Income between Rural and Urban
Business Sector Rural Urban Value (A) % Value (B) % B/A (million won) (million won) Banking 356 70.2 1,005 79.5 2.82
Purchasing 84 16.6 154 12.2 1.83 Farming Inputs 56 111 51 4.0 0.90 Consumer Goods 28 5.4 103 8.2 3.75
Marketing 50 9.9 74 5.9 1.48 Farm Products 18 3.6 41 3.2 2.25 Processing 2 0.4 0 0.0 0.14 Warehouse 21 4.1 19 1.5 0.93 Utilities 0 0.0 5 0.4 25.81 Transportation 9 1.8 9 0.7 1.00
Coop. Insurance 17 3.3 31 2.4 1.86
Total 507 100.0 1,264 100.0 2.49
Source: Survey Data
profitable when market regulation is removed as financial liberalization proceeds. Many managers of PACs argue that the success of their banking businesses will enable the PACs to expand their marketing activities despite deficits in the marketing sector. But the success of their banking businesses should come not from regulation benefits but from 95 efficient operations. Since dependency on banking is very large, a small change in market environment may have a substantial impact on the viability o f PACs. Therefore, all persons concerned about the agricultural cooperatives are very sensitive to the possible impact of expected changes in financial market environments from financial liberalization and socioeconomic changes discussed in Chapter III.
4.3 Descriptive Analysis of the Banking Cost Structure
This section presents a more detailed explanation of banking costs of the sample
PACs. The direct banking costs that can be explicitly classified are interest expenses for deposits and borrowings, and other banking costs such as loan losses and miscellaneous expenditures. Banking operating costs consisting of wages, fringe benefits, capital expenditures, retirement pay, other management costs, etc., are included in total cooperative operating costs as indirect costs since they are not explicitly allocable to individual cooperative activity. Thus, to obtain banking operating costs an index is needed to estimate the portion of banking relative to total cooperative operating costs. Although there is no method to precisely calculate the bank portion of operating costs, the NACF uses, as an alternative, the ratio of labor allocation to each business sector to evaluate the performance of each business sector. This study follows this alternative of the NACF. The labor allocation index, or the banking labor ratio (BLR), is the ratio of a normalized number of banking employees to that of total employees of a PAC as presented in equation (4.1). The normalized number of employees is the number obtained from the number of employees multiplied by the ratio of the each class employee's basic wage and retirement payments to that of the standard class employee's. 96
This index can be justified, at least, to obtain labor expenses of the banking sector since it is normalized by the wages o f the standard class employee. Moreover, the banking business is still a heavily labor intensive business so that labor expenses are the dominant portion of banking operating costs. That is, these estimates are not likely to result in a large bias relation to the true banking operating costs.
However, to specify capital expenditures, a strong assumption is needed that the capital-labor expenditure ratio is constant between each business sector. Clearly, this assumption may be too strong, but there is no better alternative except to deal with the entire businesses of a PAC. But even if we could construct a large model of the entire businesses, it would not avoid the problems of aggregation bias or degrees of freedom with the given sample size since a PAC operates many heterogeneous businesses. That is, it is not certain whether or not a large model would result in unbiased and reasonable implications. Therefore, for these reasons, this study will use a small model dealing with only the banking sector, by using the BLR. The formulation of the BLR is
(4,.) where
BNj = the number of the ith class banking employees,
BS| = the basic salary o f ith class employee,
BS4 = the fourth class employee's basic salary^,
TNj = the number of ith class total employees,
SNBE = standardized number of banking employees,
SNTE = standardized number of total employees.
The class II and III are regarded as the same group of the employee class in Korean so the class IV is the third class in Korean. This study ignores the ordering in Korean so one should be careful to use these data of PACs. The fourth class employee are an intermediary staff of the PAC and has responsibilities for decision making such as loans. 97
An example for calculating the BLR is presented in Table 4.7 which is a actual labor allocation case in the sample for 1991. The banking operating costs were obtained by multiplying the total cooperative operating costs by the BLR.
At the overall sample mean, interest expenses on deposits and on borrowings including policy loan funds are 82.2 percent of total banking costs, while operating costs are 13.3 percent (Table 4.8 ). Interest expenses on deposits and on borrowings are 66.1 and 16.1 percent of total costs, respectively. Loan losses are only 0.8 percent of total costs. In operating costs, wages and fringe benefits are the largest (8.7 percent) share and retirement pay is 1.5 percent of total banking costs. Thus, the costs related to labor are
10.2 percent of total banking costs, which is equivalent to 77.3 percent of banking operating costs. On the other hand, the management expenditures represent capital costs including the other costs such as depreciation, repairs, maintenance of fixed capital, materials, communication, advertisements, etc.. These capital costs are 3 percent of the total banking costs which is equivalent to 22.7 percent o f the banking operating costs.
This cost structure implies that the banking is a labor intensive business of the PACs. In recent years, an on-line computer system was introduced in the PACs, but its impact is not yet clear. These capital intensive technologies will bring about changes in the banking cost structure in the fiiture. This study did not include a variable representing the effect of computer systems, since it is too early to capture its impact and the data are not available.
The bank cost composition of a rural PAC is different from that of an urban PAC.
The rural PAC operating cost share to total costs (12.5 %) is slightly lower than for urban
PACs (14.4 %). This implies that rural PACs might make more effort to minimize costs than urban PACs, or that urban PACs have to pay more to compete with other financial institutions. Among direct costs, the borrowings cost share (interest expenses) of urban
PACs (7.9%) is much lower than that of rural PACs (22 %), 98
Table 4.7 An Example of Labor Allocation to Cooperative Activities by the Class of
Employee: One of the Sampled PACs for 1991
Cooperative Labor Allocation by the Class of Employee (Yearly Average) Activities I n III IV V VI vn Total (persons) Banking 0.2 12.5 11.4 24.1 Deposit 0.1 5.2 8.4 13.7 Loan 0.1 5.8 5.9 Mutual Credit 0.1 3.7 3.8 Policy Loan 2.1 2.1 Others 1.5 3 4.5
Purchasing 5.9 0.1 3.8 9.8 Marketing 2.4 2.4 2 6.8 Coop. Insurance 0.7 0.1 0.8 Extension 1.5 1.3 2.8 Administration 1 2 4.9 2.1 1.8 11.8 Total 1 2.2 27.9 15.3 7.9 1.8 56.1
Source: One of the Sampled Primary Agricultural Cooperatives in Kyounggi province, 1992.
Note: This table excludes the president of this PAC but he is reflected in calculating BLR. The weight of
each employee class, BSi/BS^, in equation (4.1) is obtained from the salaries of the ith class divided by
tliat of IV class. The weight in 1991 is :the president, 1.23; 1,2.31; II, 1.82; III, 1.49; IV, 1; V. 0.64; VI,
0.63; Vll, 0.53 {Managerial Performance Report o fNACF, PACs, 1992, p 131). The BLR for this PAC
is: SNBE = (0.2*1.49 + 12.5*1 + 11.4*0.64) = 20.09, and SNTE = (1.23 + 1*1.82 +2.2*1.49 + 27.9*1+
15.3*0.64 + 7.9*0.63 +1.8*0.53) = 49.95. Therefore, BLR = 20.09/49.95 = 0.402. 99
Table 4.8 Banking Cost Structure of Average PAC
Banking Costs R ural U rban Total Samples
Amount % Amount % Amount %
(million won) Direct Costs 964 87.5 2512 85.6 1291 86.7
Deposit Interest Expenses 670 60.8 2157 73.5 983 66.1
Borrowing Interest Expenses 242 22.0 232 7.9 240 16.1
Other Costs 52 4.7 123 4.2 67 4.5
(Loan Losses) 6 0.5 38 1.3 12 0.8
O perating Costs 137 12.5 424 14.4 198 13.3
Wage and Fringe Benefits 93 8.4 268 9.1 130 8.7
Retirement Pay 13 1.2 60 2.1 23 1.5
Management Expenditures 32 2.9 95 3.2 45 3.0
Total Banking Costs 1101 100.0 2936 100.0 1489 100.0
Source: Survey Data
which reflects that the policy loans are not a substantial portion due to the large amount of deposits mobilized, as well as the small borrowings from MCSA.
Since the PAC banking business produces multiple outputs, average costs are not measurable in a simple way. But if we use a proxy measure such as total costs divided by bank total assets or deposit size, some notion of cost structure can be obtained. Figure 4.2 presents total bank costs that shows a strikingly linear relationship 100 with total assets over all PACs in rural and urban areas. This figure suggests that there may be no difierence in the banking production technology between the rural and urban
PACs, because if the technology o f the rural PACs is significantly different from that of the urban ones, then the total cost curve should be discontinuous around the point connecting the rural and urban PACs. But this argument needs to be statistically tested rather than relying solely on this graphic presentation. Another point to be noted is that the observations are quite dense on the total cost curve in the range less than 15 billion won asset size. This figure also may imply that input inefficiencies may not be substantial.
When the proxy average costs are measured by total banking costs divided by total banking assets, the average cost curve is a horizontal line which implies constant returns to scale or no scale economies (Figure 4.3). This striking results suggests that there is no possibility of cost saving by increasing the scale of banking production. In other words, the banking costs increase proportionately along the expansion path of outputs as shown in Figure 4.2.
However, Figure 4.3 hides the relationship between the cooperative's own banking products and the externally determined products of policy loans. The interest rates on policy loans, which because of data availability are measured as average interest rates of borrowings for both policy loans and borrowings from MCSA, are much lower than that of deposits (Figure 4.4). Thus, as the policy loan share to total bank debts increases, the total average interest cost will decline. This policy loan impact of reducing total average costs must be much greater for the small rural PACs than for the large urban PACs as shown in Figure 4.6. This inverse weight of policy loan interest expenses with banking asset size resulted in the flat average cost curve of Figure 4.3. Therefore, to evaluate bank cost efficiency without the policy loan impact, the interest expense on borrowings must be 101
7
6
I; 5 g 4 I ■■
3
2
1
0 0 S 10 15 20 25 30 35 40 45 50 toUl assets (billion won)
Figure 4.2 Total Banking Cost Distribution
0.18 T
0.16 ■
0.14 --
0.12 ■■
0.1 g> 0.08 - •
0.06 -•
0.04 -
0.02 --
0 10 20 25 30515 35 40 45 50 total assets (billion won)
Figure 4.3 Average Costs Measured by Banking Assets 102
0.2 j
0.18 -■ " borrowing costs
0.16 o deposit InteresU
_ 0.14 ••
I 0.12 -
I 0.1 -■
I 0.08 ■ •
" 0.06 ■■
0.04 -
0.02 -■ ■ ■
0 - -H ------1------1------—I 10 20 30 40 50 60 70 total assets (billion won)
Figure 4.4 Average Interest Costs of Borrowings and Deposits
0.25 T
0.2
0.15 ■
0.1 ■■
0.05 -
4- 10 20 30 40 50 60 70 total banking assets (billion won)
Figure 4.5 Average Interest Costs as Sum of Costs of Borrowings and Deposits 103
0.6
0.5
0.4
I 0.3
0.2
0.1
0 10 20 30 40 SO 60 70 total banking assets (billion won)
Figure 4.6 The Borrowing Cost Share of Total Bank Costs
separated from total costs, and policy loans must be excluded from outputs. Actually, excluding borrowing costs from total costs excludes Mutual Credit loans using borrowings from MCSA because the data do not allow the interest expenses on policy loans to be separated from total borrowing expenses. Thus, for this case, deposit size can be an appropriate variable to measure average costs.
There is another reason for the effect of low interest rates on policy loans to create the flat average cost curve in Figure 4.3. That is, small rural PACs paid high prices for deposits and borrowings compared to large urban PACs as shown in Figure 4.5. In the case of deposit interest rates, this reflects the fact that the share of demand deposits with low interest rates is lower in the rural PACs than in the urban PACs. In rural areas, demand deposits are not preferred since there are relatively few clientele who need high 104 liquidity. Thus, the share of savings and time deposits is generally high in rural areas, which results in high costs of deposits. In the case of borrowing interest rates, since small rural PACs borrow a large amount of funds from MCSA with high interest rates compared to policy loans, the average rates of borrowing interest appear higher than large PACs that seldom use the MCSA funds. In other words, the average costs of small rural PACs would not be as low as those of large urban PACs if there were no cheap policy loans.
When the policy loan interest costs are excluded from the total costs and average costs are measured by deposit size, the average cost curve appears to be a convex curve to the origin but closer to L-shaped rather than U-shaped (Figure 4.7). This implies that there exist scale economies in small PACs but they can be exhausted fairly quickly as the size of
PAC increases.
0.25
•I 0.2
0.15
0.1
a a 0.05
10 20 30 40 50 GO deposit size (billion won)
Figure 4.7 Average Bank Costs Excluding Borrowing Costs 105
The impacts of cheap policy loan on bank cost structure should be considered first when analyzing banking scale efficiency. Even though the small PACs seem to be scale efficient when considering total costs, this may be an incorrect interpretation since the measure ignores the scale efficiency determined by the cooperative's own production. So far, the studies in this field have not clearly show this relationship. Most of the studies did not separate or measure the impacts of policy loans (external funds with low interest rates). Y.S. Kim and Huh failed to show the existence of banking scale economies of rural
PACs at the sample mean. Thus, their suggestion to merge the small PACs could not be justified by the empirical evidence. Even in the other studies about developing countries
(e.g., Cuevas, 1984; Srinivasan) that used the intermediation approach, the measured scale efficiency might be distorted by the impact o f external funds. For example, Srinivasan reported that the scale economies of rural bank branches in Bangladesh were not found by the intermediation approach but were found by the production approach. If the results from the intermediation approach are accepted, there are no policy implications about the scale of rural branches. However, if the scale economy measurement was related to the relationship found above, such policy implications would be meaningless. Overall scale economy analysis needs to consider both cases as discussed above. When the cost decreasing effect of policy loans on average interest costs is removed, it will become clear if there are possibilities to reduce costs by Increasing the scale of banking or not.
On the other hand, the figures presented in this section show relatively smooth cost curves, which suggests that structural differences in banking costs between rural and urban
PACs are not likely to exist. The cost curves of large, mostly urban, PACs were smoothly connected with those of small PACs in spite of substantial differences in sub accounts of the banking sector as shown in previous sections. That is, the cost curves reflect only the size differences between small and large PACs. If this result is statistically supported, the 106 rural and urban PACs can be pooled in the estimation of the frontier cost function as the same technology group.
Furthermore, Figure 4.1 also suggests that the samples are highly dense, especially for deposits less than 15 billion, along the total cost line. This may imply that total cost variation is very small among sample PACs with the same production scale, so that it likely shows a very small degree of input inefficiency from the estimated cost frontier.
4.4 Summary
So far, the discussion has focused on the business structure and the banking cost structure of the sample PACs. The business structure of the sampled PACs was very close to that of the national average data. This suggests that information obtained from this study may be applicable for policy recommendations for the whole country. Rural and urban PACs exhibited different scales due to market environment.
However, banking production technology was not significantly different from the rural to the urban PAC, although there exist substantial differences in production scale and product mix. Using graphical analysis, total costs or average cost curves presented very smooth and continuous lines in the range between rural and urban PACs. This suggests that the rural and the urban PAC are not drawn from different technology groups but from different groups in scale of production. Furthermore, there was no indicator to suggest that the urban or the rural PAC was more efficient in terms of scale or productivity, except for small rural PACs that seemed to experience scale economies. Considering the density o f total banking costs along the production scale, it is not likely that there exists substantial deviations in utilizing inputs among PACs producing the same scale and product mix. In other words, measured input inefficiency also may be very low. 107
Banking is a dominant business sector o f the PACs and has been a major income source affecting the viability of the PACs. The size differences in banking between rural and urban PACs is related to limited market area as well as market environment. To date, the merger o f small rural PACs has been recommended as an effective strategy to save banking production costs (Huh), a recommendation based on the possible existence of scale economies. However, previous studies (Kim, Y.S.; Huh) have failed to statistically show the existence of scale economies for rural PACs. As seen in Figures 4.2 and 4.3, scale economies may not be found if total costs include the interest cost of policy loans. If the impact of policy loan costs is excluded from the scale economy measure, then the existence of scale economies may be found as seen in Figure 4.4.
However, the average cost concept applied in this section is not appropriate in the regime of multiproduct firms. It is only a proxy measure to improve our intuitive understanding of the banking cost structure of the PACs. Thus, the results from the above descriptive analysis may not be upheld under more precise measurement. More precise measurement can be obtained from multiproduct functions such as the cost, production, or profit functions. The following chapter will provide a basis for estimating the multiproduct cost function and evaluating the existence of scale economies in the PACs. CHAPTER V
METHODS AND MODEL SPECIFICATION
This chapter will discuss the model building and estimation methods, which will justify the results of the frontier cost function presented in the next chapter. The model is constructed to follow the intermediation approach that is appropriate to evaluate the competitive viability of banks, as discussed in chapter II. The production approach is not applicable simply because of data constraints.
For a firm which produces multiple products to be competitively viable, the production costs o f the firm should be as low as those of any competitor with the same scale and product mix. In other words, the production costs of the firm should be on the cost frontier of the industry, and the production scale and product mix should be kept at a minimum cost point on the cost frontier surface. Therefore, in a practical sense, a firm will want to know: ( 1) what is the cost frontier of the industry with a given technology; ( 2) how much does the firm's input utilization deviate from the cost frontier and what factors undermine the full utilization of inputs; (3) how far does the firm deviate from its optimum scale on the cost frontier; and (4) what is the optimum product mix. For these questions, this chapter will present a practical methodology to explore the multiproduct cost structure.
This chapter consists of six sections. In the first section, the theoretical aspects of the dual relationship between the production process and the cost function are presented to provide justification for use of the cost function. Since the duality relationship is
108 109 satisfied when firms behave as cost minlmizers, that is, when they are on the cost frontier, the determination o f the cost frontier becomes an empirical problem. Thus, in the second section, the method for estimating the frontier cost function and firm specific inefficiency are presented. In the third section, cost concepts related to multiple products, and the measures of output efficiency such as scale economies and scope economies are presented, which leads to the empirical analysis in chapter VI. The fourth section consists of general assumptions to build the model, definition and measurements of variables, functional forms, and model specification. In the fifth section, estimation methods for the frontier cost function are presented; this also includes a justification for pooling the sample of rural and urban observations and for the data scaling by the geometric means of the sample. In the sixth section, variable selection is presented to find the relationship between the measured input inefficiency and the possible sources of input inefficiency such as employee's motivation, the market environment, managerial strategy, or uncertainty.
5.1 Production Process and Cost Function
5.1.1 The Nature of Banking Production
Banking production can be characterized as providing services for financial transactions, multioutput production, and jointness in production. Financial transactions that depend on specific contracts stating the conditions of lending and repayment are constrained by the divisibility of assets, transaction costs, prices of the assets, and risk of nonpayment. The divisibility constraint incurs search costs for a customer who wants just a certain amount of money, while risk of nonpayment incurs costs in evaluation and in the monitoring of borrowers. Besides these costs, financial transactions entail various kinds of 110 costs such as documentation, information gathering, transportation, securing of collateral, among others. Transaction costs which subsume all of these costs provide a reason for the existence of financial intermediary* (Benston and Smith).
Financial institutions produce specific financial commodities by transforming claims on generalized purchasing power, and furthermore provide financial services by acting as a broker or dealer (Benston and Smith). The transformation function enables the financial institutions to divide large-denomination financial assets into smaller units so that it can solve the divisibility constraint of both borrowers and lenders (Klein). The intrinsic risks in financial transactions are diversified through the transformation process. Financial institutions are specialized to evaluate credit risks for the uninitiated depositor; they filter the value signals o f financial assets in a financial market with limited information (Leland and Pyle; Santomero). Institutions with advantages coming from specialization and economies of scale provide financial intermediary services with lower costs^. A comparative advantage in transaction costs comes from processing documents, acquiring information about borrowers' ability to repay debts, and monitoring instruments that can be easily converted into generalized purchasing power (Benston and Smith). These tasks depend heavily on labor and capital equipment.
Financial institutions generally tend to produce more than one kind of financial commodity. They tend to have many sources and uses of funds. They can obtain funds through equity, borrowing, accepting deposits of various kinds, and so on. They can
* The basic form of financial intermediation is the market maker who provides a market place. By gathering potential buyers and sellers in the same place, information costs are reduced. A dealer who takes a position at his own risk in the assets transacted is a more sophisticated form of financial intermediary. A financial institution is the most complex form of intermediation that produces financial commodities (Benston and Smith). 2 If we assume funds mobility to be given, then how to allocate the mobile funds to each asset will be the problem to solve in order to obtain maximum revenues. In this case, optimization of portfolio management must be an appropriate approach, which has been a major field of endeavor in banking behavior studies (see Baltensperger; Santomero). I l l employ these funds by making loans, purchasing securities, building offices, buying equipment, etc. Labor and capital are utilized as inputs to add value to funds obtained and to create financial commodities as outputs. Thus, the production of financial institutions can be characterized by a multiproduct production technology;
Q = f(X) (5.1) where Q is an output vector (qj, q 2, ..., q„) and X is an input vector (x^, % 2, • .,Xm). The production function is equivalent to an input requirement set
V(Q)={XinRn+:(Q,X)} (5.2) where R”+ is non-negative n-dimensional plane and (Q, X) is a net put bundle. By convention, (Q, - X) is an output if it is positive and an input if it is negative. If there is no way to produce more output with the same inputs or to produce the same output with less inputs, a production plan is efficient. The assumption of input requirement set convexity from below guarantees the existence of an efficient input requirement set V*(Q)
(McFadden). This physical version of the production process can be converted to a cost version that will be discussed in the following section. Before analyzing the cost version of the production process and discussing production efficiency, we need to justify the use of the cost function in exploring production technology.
5.1.2 Cost Function as the Dual of the Production Process
Cost function approaches to study the banking industry have been extensively used in lieu of the production or profit function approaches. The reasons for the popularity of the cost function in empirical studies can be summarized as follows: (I) Duality theory 112 provides a justification for the cost function approach to banking production. (2) The function has some advantages in empirical analysis over the production or profit function.
McFadden demonstrated that if the input requirement set is a closed non empty, convex, and monotonie technology, then there exists a unique cost function,
C = C(Q, W) = min {W-X| X in V(Q)} (5.3) where W = (w%, W], ...,Wm) is an m-dimensional input price vector. The cost function states the least cost required to produce the output bundle Q with the input bundle X. It also satisfies the properties o f the cost function, i.e., non-negative, non-decreasing in input prices, homogeneous of degree one in input prices, concave in input prices, and continuous in input prices and fixed outputs. In addition, it is usually assumed that the cost function is twice differentiable with respect to input prices.
Varian briefly summarized the duality relation between production technology and the cost function as follows:
(1) Given a cost function we can define an input requirement set;
(2) If the original technology is convex and monotonie, the constructed technology will be
identical with the original technology
(3) If the original technology is non convex or nonmonotonic, the constructed input
requirement will be a convexified, monotonized version of the original set, and, most
importantly, the constructed technology will have the same cost function as the
original technology. That is, the cost function is well defined regardless o f the
functional form of the production function.
Thus, the cost function of a firm summarizes all of the economically relevant aspects of its technology. This property of duality in production opens a wide field for empirical studies in production technology by using both cost and profit functions. 113
The duality relationship between the production and the cost function is mathematically expressed as follows:
C(Q,W) = MinW'X (5.4) s.t.Q = f(X).
The solution to this problem is obtained by using the Lagrangian form:
MinI = W’X + MQ-f(X)] (5.5) where À is a Lagrangian multiplier. The first order conditions of the problem are:
Solving (5.5) and (5.6) gives the conditional input demands given the desired level of output Q* = (qi*,...,qn*); > e.,
Xj = Xj*(wi,...,Wm; Q*) for j = l...m (5.8)
Substituting X* in W'X gives the indirect cost function;
C* = C(qi*,...,qn*; wi,...,Wm). (5.9)
If C* is well behaved (continuous, homogeneous of degree one in W, and concave in W), then a nice relationship between C* and xj* can be obtained by Shephard's lemma; 114
Xj* = 3C*(Q, W)/awj (5 10)
Therefore, we can explore the production technology by using the indirect cost function,
C*.
5.2 Frontier Cost Function and Estimation Method
5.2.1 Cost Frontier and Possible Sources of Input Inefliciency
The duality relationship between the production process and the cost function is satisfied when a firm operates on the cost frontier as a cost minimizer or profit maximizer, but it is not easy to empirically determine the cost frontier since there exists a substantial variation of costs among the observed firms with the same scale of production and product mix. One may regard the costs of the best firm as the cost frontier (deterministic frontier approach) but that does not allow for any random effects on costs. This approach ignores the external effects on production costs that firms cannot control. Others may regard the average level of costs as the cost frontier (traditional econometric cost function). Since sample information is not complete in the sense that it carmot discriminate
among the sources of cost variation among observations such as random shocks or
measurement errors, the deterministic frontier approach does not have intuitive appeal.
But the traditional average cost function approach that assumes all deviations from
average costs are due to random errors may also be unreasonable when there exist some
conditions that allow even inefficient firms (non-cost minimizers) to be viable due to
market distortions such as market interventions or a non competitive market structure. If
market distortions result in the existence of a substantial number o f inefficient firms, the
estimated average cost function should deviate from the tme cost frontier. 115
It is arguable that even technically and / or allocatively inefficient banks are likely to be viable since the banking industry is heavily regulated. PACs have played important roles not only for their own members but also for the execution of government policies.
As discussed in chapter III, they have been protected by the government from unfavorable market conditions. Thus, regulation of the banking industry and government protection might result in the existence of inefficient PACs. For this reason, this study will examine whether or not the cost frontier of PACs differs from the traditional average cost function.
If the cost frontier deviates from the average cost function, the frontier cost function will be used to explore the cost structure of PACs. Intuitively, the deterministic frontier function is not appealing, so that this study will use the stochastic frontier function approach that allows for random effects on costs.
If there exists a frontier cost function different from the average cost function, the deviation between the cost frontier and the average cost function implies the degree of input inefficiency. It is clearly desirable to estimate not only the average level of input inefficiency but also the degree of firm specific inefficiency. Furthermore, if the sources of input inefficiency can be captured, information from the analysis will provide more practical implications for bank managers and policy makers.
The factors of measured input inefficiency may be endogenous, exogenous, or simple measurement errors. The endogenous factors o f inefficiency may result from the misallocation of inputs, over-investment for future operations, structural adjustment processes, unskilled technology in the case of newcomers, or simple ineptitude. The
exogenous factors affecting inefficiencies may be market environment constraints such as geographical segmentation, low level of social infrastructure, or regulation of the markets.
In the case of the banking industry, above all, regulation of the industry is considered to be a main source of production inefficiency as discussed in Chapter II. However, the factors 116 o f input inefficiency can be highly related to each other. The effects o f regulation may be so broad that it may not be possible to exactly distinguish the effects of regulation from others. For example, the requirement to charter a bank functions as an entry barrier that may reduce market competition and result in a monopolistic market structure. A monopolistic market structure likely results in a firm earning some rents over normal profits. Likewise, a regulated interest ceiling imposed to support some specific groups may also result in rents for some range of banks. The existence of rents may induce a lack of motivation which undermines the full utilization or optimal allocation of inputs. Of course, the lack of motivation may be due to other reasons.
Because this study employs cross-section data, the direct effects of regulation are not considered. Therefore, the sources o f input inefficiency can be classified into four categories: employees' motivation, market environment, managerial strategy, and uncertainty^. First, the lack of motivation is a principal-agent problem that results in a failure to fully utilize employees' abilities. Second, the measured degree o f inefficiency may result from different market environments. For example, rural financial institutions may need more inputs to produce the same services produced by urban financial institutions because of their geographically broad market area, small units of financial transactions, underdeveloped social infrastructure, etc. Third, measured inefficiency may also be related to a managerial strategy that focuses on one or a few specific businesses o f the PAC. For example, a manager may emphasize the banking business since it is profitable, while another manager may emphasize the marketing business since he considers it more effective in increasing members' income. A biased managerial strategy toward a specific business sector may induce employees to neglect full utilization of inputs in the other business sectors. Fourth, market uncertainty may result in a misallocation of inputs. These
^ Since the PACs in general are uniform institutions, there are no differences in regulations affecting them when we use cross section data. 117 impacts must be a random event to each individual PAC, but if the degree of market uncertainty is particularly substantial to specific observations, then the measured inefficiency of those observations must be reflected by this variation of random shock. For example, if a financial market is unstable because it is located either in a rapidly developing area or in an area under unexpected economic stress, the financial institutions in that area will be unable to make precise decisions resulting an appropriate allocation of inputs.
According to the sources o f input inefficiency categorized above, an empirical analysis will be conducted in Chapter VI. Variable selection and the expected relationship between measured inefficiency and these variables will be more concretely discussed in the section which deals with sources of input inefficiency.
5.2.2 Estimation of the Frontier Cost Function
This study uses a corrected ordinary least square (COLS) method to obtain an estimate of the frontier cost function. The COLS estimation follows the method of Olson,
Schmidt, and Waldman that used higher central moments of the ordinary least square
(OLS) residuals to get estimates of the inefficiency error term. The idea of the COLS is based on the assumption that the OLS estimator of parameters, when the error term of the function consists of composite terms such as inefficiency and random error, is unbiased and consistent except for the intercept. Thus, if we correct the bias of the OLS estimate of the intercept, we can obtain the frontier cost function via OLS.
Let the frontier cost function with the input price vector W and the output vector
Q be expressed as follows:
C(Q, W) = f(Q, W) + e = p'xt + e = P'xt + u + V (5.11) 118 where P, Xj, u, and v are parameters to be estimated, the independent variables of observation t, the inefficiency error term, and the random error, respectively. Since the central moments of the OLS residuals consistently represent the true distribution of error terms, we can estimate the expected value of the inefficiency error term u, E(u) = ft, by using the higher moments if we know the distribution of u and v. If we assume u ~ N(0, a
and V ~ #(0, then E(u) = p. = ( 2/jt)^/2 and the second and third moments of the OLS residual are
7 7t~ 2 7 jU2=t^+ of 7t (5.12)
(5.13)
These moment equations give
0 f = K t e > ' (5.14) where A2 and A 3 are the second and third moments of the OLS residuals'*.
We can then correct the constant term by subtracting E(u) = p = ( 2/jr)*/^ from the OLS estimated constant term. With the completion of this procedure, we obtain the frontier cost function consistent with the underlying assumption of the error distribution. If firms are allocatively efficient, then the inefficiency error o f the cost function p will represent the cost of technical efficiency. When the firms are not allocatively and technically efficient, the inefficiency error will represent overall inefficiency.
'* Olson et al. derived the moments for estimating the production frontier function. For the frontier cost function estimation, the sign of the inefficiency error u is changed so that the bracket term of the third moment is (1 • 4/rc) in the production frontier model. 119
Observation-specific estimates o f inefficiency, Uj, can be obtained by using the distribution of the inefficiency term conditional on the estimated entire error term ê ,, as suggested by Jondrow, Lovell, Materov, and Schmidt. To estimate u,, one can use either the expected value E(u|e) or the mode of the conditional distribution M(u|e). Let the subscript i be omitted here, then ;
M(w|e) = e(o^/(f) ife^O (5.16) = 0 if e < 0 where (J)(-) is the standard normal density function, d>( ) is the standard normal distribution function, ^ since and are not known, the estimated parameters and are used to obtain E(u|e) or M(u|e). This study uses equation (5.16) to estimate the degree of input inefficiency of a specific PAC by using ô^, ô^, and the residual ej from the COLS.
5.3. Multiproduct Cost Concepts and Output Efficiency Measures
Duality theory enables us to use cost functions for analysis of the production process. In a multiproduct setting, however, it is not easy to calculate the exact structure of the multiproduct cost surface, even if we have an exact cost function. It is practical to classify cases into some interesting areas and then to evaluate the multiproduct cost function according to established measures.
5.3.1 Multiproduct Cost Concepts and Measures of Scale and Scope Economies 120
Berger, Hanweck, and Humphrey (1986) introduced the concept of competitive viability to analyze whether or not a firm can be viable in the long run. A firm is competitively viable when its costs do not exceed the scale-adjusted cost of competitors with the same production mix. Formally it is defined as
C(Q) < l/tIiC(Qi). (5.17) where Q is an output vector, C(Q) is the cost function, Q' is any nonnegative output vector. For any positive number t, it is true such that SiQ* = t Q. Competitive viability is equivalent to cost subadditivity if and only if t = 1. That is, costs are subadditive if the division of an output bundle among two or more firms cannot produce the same bundle as cheaply as one firm; thus, the firm producing Q is competitive viable.
In the case of a single product, a firm is competitively viable at the minimum point of the average cost curve. Moreover, average cost (AC) and marginal cost (MC) derived from total costs (TC) can unambiguously describe the firm's cost structure. For example, scale economies (SCE) are typically measured as
Since AC will fall (increase) as long as MC lies below (above), if SCE is greater, equal to, or less than one, then the firm's technology reveals increasing, constant, or decreasing returns to scale. If SCE is greater than one in any range of the cost function, a natural monopoly exists in the industry. In other words, AC and MC give complete information about the competitive viability condition. 121
In a multiproduct setting, however, it is not easy to explore the structure of the multiproduct cost surface. There are two reasons for complications in determination of the cost structure (Baumol, Panzar, and Willig; Bailey and Friedlaender). First, there is no meaningful definition of average cost, such as AC in the single product case, since there is no meaningAil way to aggregate multiple products into a single product measure. Second, the composition of products can affect costs, so that costs depends on the existence of scope economies as well as scale economies. The existence of scope economies or cost complementarities among products characterizes the multiple production process.
To completely describe the cost structure of a multiproduct firm, we need to examine every point on the cost surface, but this is not feasible econometrically. A practical way to explore the shape of the cost surface is to introduce the concepts of cross sections of the cost surface from some base points of outputs. The cross-sections are divided into four cases as shown in Figure 5.1. These cross-sections are simultaneously presented on the output plane in Figure 5.2. For each cross-section, multiproduct cost concepts and economy measures that are related to the cross-sections are explained in the following section.
Ray Average Scale Economies (RSCE); Let a firm A produce output y^ and y i so the cross-section of the cost surface on the ray OA will be as shown in Figure (a) of
5.1, which represents scale expansion from the origin to the output bundle A with a constant product mix. In this case, Baumol et al. introduced the concept of ray average cost (RAC) that is a straightforward extension of a single product average cost. RAC is defined as 122
(a) cost cost (b)
total cost total costs
O A (yl,y2) O
cost
cost
total cost total costs
C(yl,0) C(0,y2'
O A (yl,y2)
Figure 5.1 Cross Sections of Multiproduct Cost Surface 123
cost
total cost surface
1C0PE(Q )
EPSUB(Q ) B RSCEi
O EPSUB(Q )
SCOPE(Q )
y 2
Figure 5.2 Multiproduct Cost Surface and Output Efficiency Measures 124
RAC{Q) = ^ ^ ^ (5.19)
where is the unit bundle for a particular product mix and t is the number o f units in the bundle Q = tQ®. That is, RAC regards multiple products as a composite commodity.
RSCE is a measure of overall ray scale economies that is defined as the elasticity o f output with respect to cost measured along a ray. Let an output vector Q = (q^, q 2,q„). Then
RSCE{Q) = d\nC{tQ)ld\x\t\,^, = Y,d\nC{Q )ld\nq, (5.20)
If RSCE(Q) = 1, there are constant returns to scale; but if RSCE(Q) is less than or greater than one, then there are economies or diseconomies when outputs are changed equiproportionately*. If RSCE(Q) = 1, then the firm is competitively viable, because it produces outputs with minimum costs.
Product Specific Scale Economies (FSCE) : While RSCE describes the behavior of costs as output expands or contracts along a given ray, product specific scale economies (FSCE) describe the way in which costs change as the output of one commodity changes while the output of the other commodity is fixed. Geometrically, the cost change is measured above the cross-section FB or EB in figure 5.2, where the other output is held fixed in each case as depicted in Figure (b) of 5.1. To measure FSCE,
Baumol et al. conceptualized average incremental cost (AIC); that is, the average added cost to total cost resulting from a given output of product i. More formally, it is defined for a given output vector and the product i e N as
* Equivalently, RSCE(Q) is measured by using the elasticity of RAC(tQ) with respect to t, at output Q. If the elasticity of RAC(tQ) is defined as e, then RSCE(Q) = l/(l+e) (Baumol, Panzar, and Willig, 1982:51). 125
(5 21)
where Q®n-1 is a vector with a zero component in place of qj® and components equal to those of Q® for the remaining products. Then the PSCE for product i at the output vector
Q® is
= (5.22)
where MC, = (5.23) dq^
If PSCE, (Q) is greater, equal to, or less than one, there are increasing, constant, or decreasing returns to scale with respect to product i. However, PSCE, (Q) reflects only the partial effect of product mix on costs since it measures the effect on costs o f a ceteris paribus change in product i. It is unlikely that an expansion of this kind will occur in a real-world firm.
Economies of Scope (SCOPE): Cost savings may result from the simultaneous production of several different outputs in a single firm; that is, economies may result from thescope of the firm's operation (Baumol, Panzar, and Willig). The presence of economies of scope creates an incentive for specialized firms to merge and become multiproduct firms. The condition of scope economies is satisfied if the sum o f the cost of the specialized firms exceeds the costs of a multiproduct firm. For example, in the case of
Figure (d) of 5 .1, if the costs of the multiproduct firm A are below point b on the 126 hyperplane OBD, then economies of scope exist, while if the costs are higher than b, then there are diseconomies of scope. If the costs of firm A are equal to b, there are weak economies of scope and either specialized or multiproduct firms are all competitively viable.
The scope economies come from the existence of non-allocable inputs. Berger,
Hanweck, and Humphrey (1987) summarized the source of scope economies in the case of the banking industry, such that:
(1) Spreadingfixed costs: If excess capacity exists, fixed or quasi-fixed brick-and-mortar
branch costs, data processing costs, or loan officer and teller expenses may be spread
over an expanded product mix.
(2) Information economies: Information generated from supplying services to clientele
may be reused without incurring substantial costs to produce additional products such
as evaluating default and delinquency probabilities.
(3) Risk reduction: Portfolio and interest risks can be reduced through asset diversification
and asset-liability matching.
(4) Consumer cost economies: By joint production of demand deposits, savings account,
and loan services, a bank can reduce clientele-incurred banking costs due to
transportation cost savings, ease of inter-account funds transfers, etc.
When we use a statistical cost function, only the cost savings from the source of (I) and
(2) can be captured, while those from (3) and (4) are ignored. Hence, the total economies from joint production may be understated in empirical studies (Berger, Hanweck, and
Humphrey, 1987)
For the output vector Q® in Figure 5.2, economies of scope (SCOPE(Q^)) are measured by the percent difference in total costs resulting from a single firm versus a pair of firms producing Q®. 127
SCOPE ( QB) = [ZjCCqjB) - C(QB)]/ C(QB) (5.24) where C(Qi®) is the cost of producing qj® on a stand alone basis. For the case of two outputs,
SCOPE (QB) = [C(0, qz) + C(qi, 0) - C(QB)]/ C(QB). (5.25)
If SCOPE (QB) is less than zero, then firm B producing output vector QB is not competitively viable. Firm B should be divided into a set of firms producing a reduced or a specialized output.
In the presence of economies of scope, costs may be spread over more products.
Thus, even if product specific scale economies present decreasing returns to scale so that a ceteris paribus increase in output q; will cause incremental costs to rise more than proportionately, the effect of cost savings from the scope economies may be greater than the effect of any other product specific scale economies. In this case, the cost function is said to be transray convex (Baumol, Panzar, and Willig). That is, as a firm changes its product mix while holding the level of some aggregate measure of output, cost will be lower for a diverse rather than a specialized product mix. In Figure (d) of 5.1, the cost surface C(yi,0)-C(0,yz) represents transray convexity. In this case, firm A' will increase the product mix ratio of output yl to y 2 to reach the minimum point on a given cost surface. The concept of transray convexity of the cost function leads to cost subadditivity that is useful to test for natural monopoly. However, it is not practical to verify econometrically whether the combined costs are higher over every division of Q. Berger,
Hanweck, and Humphrey (1986, 1987) consider two subcases. One compares the costs of a larger firm with that of combinations of small firms producing the same complement of 128 outputs as the larger firm (expansion path subadditivity). In the other subcase, each small firm produces one component of the output vector (scope economies).
Expansion Path Subadditivity (EPSUB): Let firm A and B compete in a market with different scales and product mixes as shown in Figure 5.2. For firm B to be competitively viable, it should have a competitive edge over not only firm A but also a potential competitor firm D that produces a residual output bundle, Q^, which is equivalent to - Q^. This is measured by comparing the cost sets of a pair of firms producing output vector Q® versus the costs of the single firm B.
EPSUB (QB) = [C(QA) + C(QD) - C(QB)]/ C(QB) (5.26)
where QA +q D = q B if EPSUB (Q B) is less than zero, then firm B cannot be viable. It will be driven out of a competitive market by a combination of firms A and D. If EPSUB
(Q B) is greater than zero, firm A may face inducements to expand its size and/ or acquire or merge with other firms such as firm D®.
Expansion path scale economies (EPSCE) measure cost change effects along the impacts of changing scale and product mix simultaneously; this corresponds to the cross section of Figure (c) of 5.1. In the output plane of Figure 5.2, the concept of EPSCE is represented by the line OA and OB. However, the measure of EPSCE by Berger et al. has some drawbacks as will be presented in the following section. Therefore, an alternative method is developed in this study, which is based on the idea of EPSCE. The alternative measure will be presented in the following section.
^ Evans and Heckman introduced a grid approach to examine cost subadditivity. Grosskopt, Hayes, and Yaisawarang (1992) developed economies of diversification (DIVERS) as an intermediate case between SCOPE and EPSUB. 129
5.3.2 The Drawbacks of EPSCE and Expansion Path Cost Efficiency
The Drawbacks of EPSCE: A firm may grow along an expansion path with a varying product mix; for instance firm A expands into firm B in Figure 5.2. To measure the cost change effects along the expansion path AB, Berger et al. used equation (5.27), the elasticity of incremental cost with respect to a changing product mix as scale increases, and named this measure the expansion path scale economies (EPSCE):
(5.27) _ Y . dlnC(Q^y ^[C(e")-C(e'‘)]/C(ô") JlnQ
Qi® and are the ith outputs contained in the output vectors o f firms A and B respectively. Berger et al. interpreted the evaluation criteria of EPSCE as follows. If
EPSCE ( Q \ Q®) is greater than one, then there are diseconomies on the ray AB. If firm
A is competitively viable, then firm B is viable if and only if EPSCE ( Q \ Q®) = 1. If
EPSCE (Q\ Q®) is less than one, firm B would be driven out by a larger firm with the same product mix. The ray scale economies are a special case of EPSCE.
However, there are problems in the application of this measurement. First, the interpretation of the inequality sign given by the authors is wrong. Second, the measurement is not general enough to represent a possible set of changes in both the scale and product mix. Third, it is not appropriate to refer to it as a measure of scale economies, since the measure represents changes not only in scale but also in product mix. The following interpretation shows the problems of EPSCE. 130
Berger et al. explain that if EPSCE is greater than one, then there are diseconomies on the ray AB. However, this interpretation is wrong because the term of the numerator
(QBj-QAj)/QjB implies a percentage increase in output and the term of the denominator
[C(QB)-C(QA)]/C(QB) implies a percentage increase in costs along the expansion path.
Let us assume that the sum of the cost elasticities of the individual outputs equals 1. If the percentage increase in outputs is greater than the percentage increase in costs, economies o f scale must exist, so E PSC E (Q \ QB) > l. This relationship means that Berger et al. interpreted the meaning of EPSCE in the opposite direction. Perhaps they applied the same intuition to this case as in the case of simple cost elasticity with respect to output to this case.
The cost elasticity o f output is the last term of the equation, 91nC(QB)/dlnQ;. If the sum of the elasticities with respect to each output, RSCE, is equal to 1, then the firm exhausts overall scale economies. It is frequently observed that the sum of output elasticity is less than I. In this case, the last term, the cost elasticity of output, reduces the magnitude of the ratio of the increasing rate of output to that of cost. Therefore, the measured EPSCE depends not only on the ratio of the increasing rate of output to that of cost, but also on the degree of RSCE of large firm B. Even if the increasing rate of output is greater than that of cost, EPSCE can be less than I if the RSCE of the larger firm exhibits scale economies (i.e., the sum of the cost elasticities with respect to each output is less than 1). This means that EPSCE is not general, so that even if we correct Berger et al.'s mistake in interpreting the inequality sign, we can apply this measure only to the specific case when a large firm exhausts ray scale economies.
The change in product mix is related to the cost complementarity effect rather than the scale effect. A firm may save costs solely from the effects of product mix economies 131 without realizing cost savings from an expansion of scale. Therefore, the term "expansion path scale economies" needs to be changed to represent a more appropriate meaning.
The following examples are presented to illustrate the problems associated with
EPSCE. Let firm A be small and firm B be large. EPSCE ( Q \ Q^) is measured for three cases. Case 1 shows that outputs increase faster than costs (scale economies). The ray scale economies of the larger firm, B, are assumed to be 1 (no scale economies). As presented in Table 5.1, EPSCE(Q^, Q®) is greater than one under the existence o f scale economies if the large firm exhibits no overall ray scale economies. Case 2 represents outputs that increase more slowly than costs (diseconomies of scale). The ray scale economies of firm B are assumed to be 1. Table 5.2 presents the results of EPSCE under diseconomies of scale; that is, E P S C E (Q \ Q®) is less than 1 if the large firm exhibits no
RSCE. Case 3 also shows that output increases faster than costs (scale economies) but the ray scale economies of the larger firm B are assumed to be less than 1 ; that is, the larger firm operates under scale economies. However, despite the scale economies on the
Table 5.1 The Case 1 of EPSCE with No RSCE of Large Firm
Firm (QB..QA.)/Q.Bor ainC(QB)/3inQi
Small (A) Large (B) 1C(QB)-C(QA)1/C(QB)
Output 1 10 40 (40 - 10)/40 = 3/4 0.4
Output 2 50 150 (150-50)/150 = 2/3 0.6
Costs 100 200 100/200 = 1/2 -
EPSCE(QA, Q®)= [0.4*(3/4) +0.6*(2/3)]/(1/2) = 16.8/12 > 1. 132
Table 5.2 The Case 2 of Diseconomies of EPSCE with No RSCE of Large Firms
(Q®i-Q^)/Qi® or 3lnC(QB)/ainQi Small (A) Large (B) fC(QP)-C(QA)l/CfO»)______
Output 1 30 40 (40 - 30)/40 = 1/4 0-4
Output 2 80 150 (150-80)/150 = 7/15 0-6
Costs 100 200 100/2 00 = 1/2
EPSCE(QA, QB)= [0.4*(1/4) + 0.6*(7/15)]/(l/2) = 45.6/60 < 1 ______
Table 5.3 The Case 3 of EPSCE with Scale Economies of Large Firms
______(Q»i-Q^;)/Q;B or ainC(QB)/3lnQi
Output 1 10 30 ( 3 0 . io )/30 = 2/3 0 4
Output 2 50 100 (100 - 50)/100 = 1/2 0.4
Costs 100 200 100/200 = 1/2
______EPSCE(QA QB)= [0.4*(2/3) +0.4*(l/2)]/(l/2) = 5.6/6 < 1. ______133 expansion path, EPSCE(Q^, Q^) is less than one because the large firm B is under overall ray scale economies. This result contradicts case 1. Therefore, these examples clearly show that EPSCE is not a relevant measure.
In their 1986 working paper, Berger et al. tried to explain the results obtained by their analysis of the U.S. banking industry, but their interpretation was not clear. The interpretation was omitted in this 1987 article. Hunter et al. got similar results in their study but did not interpret the implications. Perhaps the mistake in explaining the EPSCE measure undermines its wider application in empirical analysis, while the measure o f expansion path subadditivity that was simultaneously developed by Berger et al. has been popular.
Expansion Path Cost Efficiency (EPCE): To measure the degree of potential cost saving realized by changing the scale and product mix along an expansion path, this study extends the idea of examining the concavity condition. Since this measures overall cost saving from both scale and cost complementary effects, it is named the expansion path cost efficiency (EPCE).
Let a frontier cost surface be f(Q, W) = f(x) as shown in Figure 5.3. In addition, let firm A and firm B have output bundles x* and y, respectively, as shown in Figure 5.3.
Then total costs will be A and B on the frontier cost surface. If firm A grows to the level of firm B on the expansion path AB of Figure 5.2, then the impact on costs from changing scale and product mix can be measured in the following way.
We can obtain a tangent plane of point A on the frontier cost function by taking the derivative at A with respect to the independent variables 134 cost
total cost
tangent plane AC=MC
output 0 X* y
Figure 5.3 Expansion Path Cost EfTiciency
f{x)+Df{x){y-x) (5.28) where f(x*) = predicted value of costs at point A,
X * = a vector of independent variables at point A,
dx„
y = a vector of independent variables at the scale efficient point B.
Equation 5.28 is equivalent to
(5 .29) 135
If costs increase at a constant ratio as outputs increase from x* to y, the cost surface will expand along the tangent plane. If this is the case, total costs will be C on the tangent plane A where the output coordinate coincides with y. However, if there is a cost saving from expanding x* to y, total costs will be lower than point C. The distance between point C and the actual cost, point B, on the cost function, implies the degree of potential cost saving obtained by expanding firm scale and changing the product mix. The measurement of cost efficiency on the expansion path is the ratio of the distance BC to
Cy. Therefore, the degree of expansion path cost efficiency is formally stated as:
EPCE is a measure of the impact of total cost changes, while the other measures such as RSCE, PSCE, or EPSCE depend on an analogy of average cost concepts. Thus,
EPCE gives direct information about the degree o f cost saving from changing the scale and product mix, while the other measures give only indirect information. For example,
RSCE suggests only whether or not a firm with a specific output bundle is operating under scale economies, and does not present how much cost can be saved if a scale inefficient firm reaches the scale efficient point. Therefore, EPCE is not only more general than
RSCE but also presents more valuable information about the degree of potential cost
saving. Furthermore, this advantage can be used to compare the degree of cost saving experienced with scale efficiency as opposed to that achieved from removing input inefficiency, if a firm is under scale inefficiency and deviates from the cost frontier.
So far the discussion has focused on how to measure output efficiency on the cost frontier surface without regard to the appropriate econometric technology. However, all 136 measures introduced in this section cannot appropriately employ a consistent econometric method, since the choice of a specific functional form accompanies the cost of its selection, as will be discussed in the following section. For example, if we use the translog cost function, we cannot get meaningful results about product specific economies or economies of scope, because the measures for those economies require a zero output evaluation that is not relevant in logarithm form. Therefore, we need to make a decision about what information should be obtained in order to achieve the objectives of the study in question. This study will use only three measures of output efficiency, i.e., RSCE,
EPCE, and EPSUB, since the benefit of choosing other functional forms rather than the translog are not expected to be great. The following section will provide a Justification for, and present a model specification based on this decision.
5.4 Model Specification for the Frontier Cost Function
5.4.1 General Assumptions
The manager of a PAC is assumed to be a cost minimizer. PACs are price takers operating under government price controls for both deposits and loans. In this situation, cost minimizing behavior must be consistent with welfare maximizing behavior on the part of PAC members. Actually, cooperative banking is similar to commercial banking. The banking services of PACs are open to non-members as well as members, which is not consistent with the general rules of cooperatives. Non-members form the predominant share of total clientele in urban PACs. In rural areas, almost all farm households are members of a PAC. PACs directly compete with other financial institutions in urban as well as rural areas. This is different from the condition that is assumed in cooperative or 137 credit union theory based on closed membership. The objective of the banking business in
PACs is often viewed as profit maximization (Y.S. Kim). Therefore, the assumption of cost minimization must be consistent with the reality of the banking behavior of the PACs.
The cost minimizing assumption facilitates analysis of the PACs cost structure based on a duality theorem, which guarantees that the cost fiinction exactly reflects the properties of the production technology.
The second assumption is that the production processes of PACs are separable between banking and the other sectors. This separability means that resources used in a
PAC are allocable between banking and its other production activities. The PACs' accounting system separates these sectors so that most resource use can be captured by accounting documents. The data describing labor allocation to each business sector are available, but that o f fixed capital and managerial expenditures on materials are not. Thus, the joint goods nature of fixed capital and operating costs requires another assumption to calculate the fixed capital and operating costs used for banking.
The last general assumption is that banking's capital-labor ratio is the same with that of other sectors in the PACs. In this study, a standardized banking labor to total labor ratio of the PAC will be used, which is the usual index in analyzing an individual business income. This is clearly an arbitrary assumption because PACs may have different capital structures depending on the attributes of their businesses, such as processing factories, warehouses, buildings for utilization, or supermarkets etc. However, an analysis of the entire business due to the joint good nature of fixed capital may require too many parameters to handle with a small sample size. Therefore, it is practical to separate banking from the other business sectors. 138
5.4.2 Hypothesis
The major objective of this study is to obtain information about the production efficiency of PACs by exploring their cost structure, which depends on the dual relationship between the production process and the cost function. To fulfill these objectives, the study focuses on testing the following hypotheses:
Hypothesis 1: The banking sectors o f the PACs are operating under scale
economies.
Hypothesis 2: The banking businesses o f the PACs have economies o f cost
complementarity as multiproduct firms.
Hypothesis 3: Some PACs overuse inputs in their banking production, or there
exists systemic input inefficiency in the PACs’ banking production.
Hypothesis 4: The degree o f input inefficiency is related to employee's motivation,
market environment, managerial strategy, and uncertainty.
The above hypotheses will be tested by using cost economy measures introduced in the previous section and statistical tests. The test for hypothesis 1 will be conducted by using RSCE and EPCE, while the test for hypothesis 2 will be done by EPSUB.
Hypothesis 3 will be tested by the moment statistics of the OLS residuals, which will be explained in 6.2.2, while hypothesis 4 will be tested by simple OLS regression. The order of hypotheses does not reflect the order of the empirical tests, since they cannot be exclusively tested. That is, the test for hypothesis 3 is determined by whether or not the tests for the hypotheses1 and 2 should depend on a cost frontier function that is different from the traditional cost (average) function. Only if the null hypothesis of 3 is accepted, i.e., the cost frontier function exists, can hypothesis 4 be tested. 139
5.4.3 Definition and Measurement of Variables
Output ; Deposits, Mutual Credit loans, and policy loans are specified as the banking outputs of the PACs. The outputs are aggregate value; that is, deposits and
Mutual Credit loans include 11 and 7 commodities, respectively, while policy loans consist of about 100 different items. There are some differences in the same category of output such as differences between demand deposits and savings and time deposits in the deposit category, between general loans and mid-term loans in the Mutual Credit loan category, etc. However, it is not practical to try to capture all the characteristic differences in the same output category with a limited sample size. Flexible functional forms such as the translog function require many interaction terms between independent variables, so the number of parameters may exceed the number of observations or may cause problems with the degrees of freedom. The output classification of this study can be justified in a practical sense, because the differences within an output category can be ignored when they are compared with the differences between outputs.
The definition of output follows the value added criterion. Output should be measured in terms of what banks do that causes operating expenses to be incurred
(Benston et al. 1982). Thus, deposits are outputs, as discussed in chapter II. Policy loans must also be outputs in terms of incurring real costs for monitoring and collecting repayments. But the characteristics of policy loans are also different from the characteristics of other outputs since their amounts are not internally determined. This particular feature o f policy loans should be reflected in the model specification and in evaluating the frontier cost function. In contrast to treating deposits as outputs, time deposits in Mutual Credit Special Accounts (MCSA) of the NACF are used for the management of the FACs' surplus funds but are not defined as outputs because they do 140 not incur real costs. There is no risk of default or price variation in time deposits in the
MCSA. Time deposits in the MCSA are just the residual of clients' deposits that are subtracted from Mutual Credit loans. The characteristics of outputs are already reflected in clients' deposits. Therefore, the time deposits in the MCSA should be classified as outputs only if the deposits of its clientele are classified only as inputs. Since this study uses the structural model of banking production, the intermediation approach, time deposits in the MCSA are not classified as outputs. Borrowings, both policy loans and borrowings from MCSA, also are not specified as outputs according to value added criteria, but clients' deposits are specified as outputs. Borrowings are only input substitutes for the capital and labor used to produce deposits as a means of financing loans and other assets. Borrowings entail almost no labor and physical capital operating expenses or value added.
The value of outputs is measured by the average stock value in won in 1991 that is calculated by the NACF accounting method. This measurement has some advantages. It does not have a seasonal variation problem that may result in a bias of the estimated parameters. The seasonal variation of outputs is very substantial in the case of rural PACs as discussed in chapter III. Thus, if a study uses a specific date, such as the end o f year, the stock value as the metric of outputs is likely to provide biased estimation results. On the other hand, the use of flow values such as total transaction amounts per year may also result in biased estimation, since the total transaction amounts do not reflect transitory and stationary transactions. Therefore, the average stock value is a more reasonable measurement of outputs.
Inputs : Inputs are labor, capital, deposits, and borrowings. Labor is measured by the number of standardized employees, normalized by the salaries of each class of 141 employees. The basic salaries of employees in the same class are the same in all PACs, while fringe benefits may be different from one to another PAC. The standardized number of employees of the banking sector (SNEB) is obtained as followings:
where BN; is the number of ith class banking employees, BS j is the basic salary of the ith class employee, BS 4 is the fourth class employee's basic salary, and i = 1, ,V. The quantity of capital is measured as the non-labor expenditure portion of operating costs.
The proxy value of capital price will be explained below. Deposits are the average daily stock value as explained above. Borrowings, including policy loan funds and borrowings from MCSA, are the stock value at the end of year rather than the average stock value, since the data of average stock value o f borrowings from MCSA is not available.
Input Prices; The labor price is obtained by dividing the sum of wages and fringe benefits by the average standardized number of banking employees in 1991. The capital price is a controversial issue in service industry studies. One of the methods used to obtain capital price is to use the book value o f capital and capital expenditures. However, the drawback of the book value approach is clear since book value is not a market value but historical value. The book value for the machinery and equipment purchased in different years is reported in the then current capital prices. The value according to this method understates the actual value of capital purchased in previous years, so that banks with older machines appear to be more efficient. Another proxy value o f capital that has been popularly used in banking studies includes the rental cost of fixed assets and premises.
^ See Table 4.7 for an example of labor allocation to cooperative activities. 142
However, the rental costs are not obtained if banks own their fixed assets, such as in
Korea. The other alternative is to divide capital expenditures such as rents, depreciation, utilities, equipment, and furniture expenditure by the dollar value of deposits or assets
(Mester; Ferrier and Lovell). This study follows this last method, and capital expenditures are divided by the amount of deposits.
The deposit price is obtained by dividing interest paid to depositors by the average stock value of deposits. Even though the interest rate of every item of deposits is fixed, the interest rates at the aggregate level depend on the composition of the deposit commodities or time period of deposits. This implies that the price of aggregate deposits varies across the sample observations. Moreover, the price may depend on the characteristics of specific regional markets such that the ratio of demand deposits to total deposits in rural areas is likely to be lower than in urban areas. This characteristic results in a higher price of aggregate deposits. The deposit price was used as an effective yield
(Mullineaux; Mester; Cuevas). It may be a useful index to banking managers because actual interest costs o f deposits at some aggregate level might be more helpful in decision making. Banking managers are likely to use effective yields rather than nominal interest rate of every deposit commodity, when they determine banking scale.
For the same reason, the price of borrowings is obtained as interest expense on purchased funds divided by their stock value at the end of year. Purchased funds include policy loans and borrowings from the MCSA that are used for the cooperative's own purposes.
Total Costs: Total costs are the dependent vaiiable and consist of operating costs, interest expenses on both deposits and borrowings, and loan losses. 143
Incorporating a Branch Variable: The number of branches is incorporated as a control variable, with the variable measured by the number of banking offices, head office plus branches. This does not necessarily mean that the head office and branch offices are treated equally. Log terms do not allow for zero values and the value of log one is zero; therefore, adding the office to the number of branches does not affect estimation results.
5.4.4 Functional Form
This study uses the translog cost function as the primary functional form. The frontier cost function is obtained from the cost function by using the corrected ordinary least square (COLS) method. Among the reasons for choosing the translog cost fiinction are the following.
First, the translog cost function has many desirable properties as a flexible function. It is flexible enough to represent production technology without placing any prior restrictions on the full set of price and output elasticities at a base point (Caves and
Christensen). It is more parsimonious in parameters than other flexible functional forms. It allows testing for homotheticity, jointness, separability in production, and permits estimation of a U-shaped average cost curve. Furthermore, it is linear in parameters so that it is relatively easier to deal with and provides global information about an estimated cost surface.
Second, although the translog function has some drawbacks, such as inability to evaluate zero output and a narrow range of the approximation region where the cost function is well defined, the alternative functional forms that have been developed also have some drawbacks as discussed in Chapter II. For example, quadratic functional forms preclude the possibility of nonseparability between outputs and inputs. The generalized 144 translog form may solve this problem, but its analytical characterization is awkward and
difficult to work with, and does not represent the relevant cost properties as a tractable
expression of its parameters. The minflex-Laurent function and the composite cost
function can solve the problems of narrow approximation region or of incorporating zero
outputs, but it is not parsimonious in parameters so that it is likely to increase
multicollinearity problems. In addition, the generalized and the composite cost functions
are highly nonlinear so that they cannot guarantee provision of global information about
the cost surface.
Third, even if we do not evaluate zero output points on the cost surface, we may
obtain valuable information about the cost structure or production technology within the
sample range. Moreover, zero output evaluation may be often undertaken far outside the
sample range, which may result in unreliable conclusions. In fact, multiproduct firms
belong to an industry that produces the same kinds of outputs, which implies that we
cannot observe zero output cases. In this case, any inference from an estimated cost
function cannot avoid extrapolation error. Therefore, it will be more valuable to get
information within the range of sample information. The PACs produce almost the same
kinds of banking commodities. In fact, measurement of economies of scope and product
specific scale economies are directly related to the problem of zero output evaluation.
Since the translog function cannot appropriately treat the zero output case, most studies
that have tried to evaluate scope or product specific economies with the translog cost
function used an arbitrary approximation of a zero value. However, the evaluation results
varied along the degree of approximation, so that the implication of the results was
questionable (Berger, Hanweck, and Humphrey, 1986,1987)*. Thus, if we are concerned
* Thus, Berger et al. (1986, 1987) used a hybrid translog cost function to measure scope economies, while the standard translog cost function was used to measure overall scale economies, expansion path scale economies, and expansion path subadditivity. But the hybrid translog cost function properties are not yet well known. 145 about more reliable information from the observations rather than uncertain information about economies of scope and product specific economies depending on zero output points that are located far outside of observations, the translog cost function can be a reasonable instrument to explore a multiproduct cost structure. This study will use only sample information such as ray scale economies, expansion path cost efSciency, and cost subadditivity, which do not necessarily require a zero output evaluation.
In addition, since the approximation region of the translog cost function depends on the point of approximation, if we move the approximation point along the estimated cost surface, the problem of the narrow range of approximation region may be reduced
(Boisvert). This study will use the geometric mean scaling method for that purpose.
5.4.5 Model Specification
This study uses a single cost equation rather than a system of equations, employing the seemingly unrelated regression estimation (SURE) method. The SURE model depends on the idea that the use o f a likely existing contemporaneous correlation between a cost function and its input share equations derived from the cost function will improve the efficiency of the estimated parameters. In addition, since the use of share functions implies an increase in data, the SURE model may reduce the multicollinearity problem that can occur in flexible functional forms using many interaction terms. However, there is a theoretical problem when we estimate a frontier cost function. It is unclear how the error term of the cost fonction is related to that of the input share equation (Greene, 1980;
Bauer). The literature shows that either one makes strong assumptions about the relationship, or ignores it. Thus, the estimation results are different from each other 146 depending on these assumptions^. Thus, since the purpose of using SURE method is to improve the efficiency of parameters, there is no reason necessarily to use a SURE model in estimating cost frontier models. The single equation method can avoid such arbitrary assumptions about the error term relationship and does not necessarily obtain inefficient estimates of parameters if the contemporaneous correlation among the equations of the
SURE system are not significant (Judge et al.).
The model specification uses three pieces of prior information and one piece of ex post information. The first prior information is that there is no relationship between policy loans and input prices. Since policy loans are exogenousely determined by the government, there is no reason that they are related to the input prices. This information is imposed on the model by dropping the interaction terms between policy loans and the input prices. However, the interaction terms between policy loans and other outputs are included because the marginal cost of individual output is likely to be dependent on other outputs through the simultaneous use of information. Common use of information can reduce operating costs such as monitoring, evaluation, etc.
The second information is that the price of purchased funds, policy loans and borrowings from the Mutual Credit Special Account of NACF, do not affect scale economies through interaction terms with outputs. This relationship is imposed by dropping the interaction terms between outputs and the price of purchased funds, which follows Hunter et aVs (1989,1990) specification. This specification differentiates the characteristics of deposits from that of purchased funds. Deposits are specified not only as outputs, but also as inputs that affect scale economies through price effects, i.e., through the interaction terms associated with deposit price and outputs. However, purchased funds
^ For example, Ferrier and Lovell assumed allocation inefTiciency is fixed over the time period and decomposed input ineSlciency to technical and allocation inefficiency. Their results showed the existence of very high allocation inefficiency in U.S. banking industry, which is much diCTerent from the results of other studies. 147 affect total costs through only interest cost effects. It should be noted that this specification excludes only the price effect of purchased funds through cost functions.
There still exists a possible effect on the scale economy measure through interest paid on purchased funds already included in total costs, as discussed in the previous chapter. This effect will be captured by the partial output elasticity o f policy loans.
The third theoretical prior information is linear homogeneity in input prices and symmetry condition. Linear homogeneity is a precondition for the existence of a duality relationship between the cost and production transformation process (Caves, Christensen, and Tretheway). Linear homogeneity is imposed by 2^^.= 1, ZkPu = 0, = 0, and the symmetry restriction is aij = «j, and Py = Pik-
The ex-post information concerns the relationship between the number of branches and total costs. This variable is assumed to have a log-linear relationship with total costs, without allowing interaction terms with other independent variables such as output or input variables. Berger et al. (1986, 1987) argued that branches can be considered as substitutes for deposit interest, because the branch provides more convenient services for clientele, thus reducing clients' transaction costs such as transportation or waiting. This argument suggests that the use of interaction terms will be appropriate to represent the true cost relationship. However, the observations of this study did not show any significant interactive relationship with other output variables. That is, there is no practical advantage from such a specification to compensate for a loss of degrees of freedom.
According to the specification, the cost frontier model is specified as
3 j 3 3 4 \nTC = + + Inw* i= l ^ 1=1 J = \ *=1 j 4 4 2 3 + Inw* + ein^/(+M+v ^ *=i /=i 1=1 *=i (5 .3 2 ) 148 where
TC = total costs = operating costs + interest payments on deposits and purchased funds
+ loan losses, q, = won value of output i, 1) clientele's deposits, 2) Mutual Credit loans (cooperative
loans), and 3) policy loans, w i = average annual salary including fnnge benefits paid per employee,
W2 = total capital expenditure including management costs divided by average daily stock
value of deposits = (operating costs - salary and fringe benefits - retirement pay)/
average daily stock value of deposits,
W3 = deposit price = interest payments on deposits divided by average daily stock value of
deposits,
W4 = price of purchased funds = interest payments on borrowings from MCSA and policy
loans divided by the end of year stock of purchased funds,
BR = the number of branches plus head office = the number of banking offices, u = input inefficiency error term with half normal distribution, N(0, a„2), and
V = statistical noise with normal distribution, N(0, 0 ^2).
5.5 Estimation Method for the Frontier Cost Function
Pooling Sample Estimation : The urban subsample of PACs has only 43 observations, but the number of parameters to be estimated is 31. The sample size may therefore not be large enough to obtain reliable estimation results. An effective means of estimation which avoids the problem of degrees of freedom is either to pool both the rural and urban samples or use dummy variables with respect to all independent variables when 149 a pooling estimation cannot be statistically justified. A test is needed to determine whether the subsamples of both rural and urban should be pooled or not.
Separating the subsamples is intuitively appealing since the environment of the rural financial markets is different from that of urban financial markets. Different financial market environments may result in different production technologies, yielding different cost functions. The urban financial market is more competitive, dependent on the non-farm sector, has less seasonal variation in financial transactions, and has fewer limitations on market size for the individual PAC. On the other hand, there are also factors that suggest that the cost fonction of rural PACs may not be different from that of urban PACs. Both rural and urban PACs are organized within the same system. Though each PAC is legally independent, it operates as if it were a branch of NACF. The managers of PACs are employees of NACF and have been trained in NACF. They are allocated to each PAC and transferred after a certain period of employment. The principles of management for all
PACs is determined by NACF. Therefore, all PACs likely have the same knowledge about banking operations. All PACs can have the same banking production technology, so that the cost function may not differ among PACs despite different financial market environments. Therefore, whether or not the subsamples can be pooled should be statistically tested because of these conflicting factors.
The analysis used the Chow test, which has been used effectively in linear models to test for the existence of structural differences between two samples (McKinnon, 1992).
The Chow test did not reject the null hypothesis that the cost fonction of the rural PACs as structurally the same as the urban PACs. The F-value with degrees of freedom of 133 and 10 was 0.43993 that is far less than its critical value 2.58 at the 5 percent significant level. Since there was no significant difference in the two regressions of the rural and urban subsamples, they were pooled and regarded as a group with common technology. 150
Data Scaling Estimation : The data were scaled around the geometric means of all observations to conveniently evaluate the estimated cost frontier surface. In the translog function case, the estimates from the scaled data are equivalent to those from the non-scaled data (Boisvert). This does not mean that scaled estimates of the parameters remain the same as those of the non-scaled data. But they should change in an appropriate manner so that significance and other substantive statements should not be affected by scaling (Cramer).
Geometric mean scaling enables us to obtain an approximation to the true underlying frontier cost function in the neighborhood of the scaling point (Boisvert;
Akridge and Hertel). This advantage may reduce the criticism of a narrow range of approximation region where the cost âinction is well defined, if we move the scaling point along the sample class (Boisvert). This study uses this strategy to evaluate a specific point on the frontier cost function. The pooled samples were divided into seven classes and the scaling point was changed by the class mean when the class mean point needed to be evaluated. Another advantage is that scaling the data with the geometric mean facilitates calculation of scope and scale economy measures since the natural logarithm of one is zero. In the translog function, the examination of the estimated sign often depends on the first or second order derivatives, and scaling can provide a very simple calculation of the first and second order derivatives with respect to independent variables. In addition, for inference to be free from the arbitrary measurement of units, this can be effectively used
(Cramer).
Measuring the Degree of Overall Input Inefficiency : This study uses equation
(5.16) as the analytical method for obtaining a measure of firm specific overall input inefficiency. Overall input inefficiency can be decomposed into technical and allocative 151 inefficiency by using the Kopp and Diewert method, but this decomposition was not conducted, since the measured inefficiency is so low that decomposing the overall inefficiency into technical and allocative inefficiency is meaningless.
Estimation Excluding Outliers : The estimation of the frontier cost function was very sensitive to outliers. Thus, two outliers were excluded from the data; one was far outside the positive residuals, while the other was far outside the negative ones. The estimation with the outliers presented significant heteroscedasticity, but the estimation without the outliers did not. Since the presence of heteroscedasticity due to only two observations is not reasonable, the outliers were removed from the estimation. Thus, a total of 206 observations were used to estimate the frontier cost function.
5.6 Variable Selection for the Analysis of Input Inefficiency Sources
Possible sources o f input inefficiency can be categorized into four groups as discussed in 5.2.1 including the employees' incentive to work, market environment, managerial strategy, and uncertainty. To find a systematic relationship between measured overall input inefficiency and its possible sources, regression analysis will be conducted.
For this purpose, the explanatory variables were selected using the following information.
First, the total wage and fringe benefits paid to an employee were chosen to represent the employee's incentive to work. As Leibenstein proposed, lack of employee motivation can be an important source of input inefficiency. Thus, a large incentive should reduce the degree of inefficiency. It is assumed that total wages, including salaries and fringe benefits, represent the employee's motivation level, 152
Second, six variables were chosen to represent the market environment or other exogenous variables. The variables representing the market environment are population density, a farm-household ratio to total households, and geographical characteristic of the region (flat, hilly. Up, and city). These variables represent differences in market competitiveness, are completely externally determined, and thus may be correlated with each other. High population density, a low farm household ratio, and city or Up represent a competitive urban financial market environment. If market competitiveness encourages efficiency in financial institutions, these variables will have a positive relationship with the measured degree o f input efficiency. There are other variables that represent the market environment but are not completely externally determined such as the size of the PAC.
The selected size variables that are related to both market environment and internal decision making are deposit size and the number of cooperative members. Another variable is the ratio of policy loans to Mutual Credit loans, which represents dependency on externally determined funds in supplying money to the market. Since a high dependency on policy loans means that the PAC is either operating in an unfavorable market environment or is inefficient, the expected relationship to the degree o f input efficiency is negative.
Third, to represent the characteristics of managerial strategy in terms of business skewness, two variables were chosen, i.e., the share of banking business to total gross income and the ratio o f deposits to Mutual Credit loans. Since a large share of banking gross income means that the manager of the PAC emphasizes the banking sector as a strategic business, the production efficiency of banking is likely to be high. That is, the banking gross income share is expected to be positively related to the degree of input efficiency. The ratio of deposits to Mutual Credit loans represents a specific managerial strategy rather than the entire cooperative business strategy. A manager of a PAC may 153 prefer deposits in the MCSA to lending mobilized funds, since the deposits in the MCSA have no risk and low transaction costs. In other words, this ratio may represent an easy going strategy that is likely to be related to a lack of motivation to actively operate the cooperative businesses. If this variable represents only an easy going strategy, its expected sign will be negative. However, this variable also may reflect regulatory conditions for urban PACs that deposit residual funds at the MCSA because of restrictions on lending money to non members. In this case, the effects of efficiency can be positive due to the competitiveness of urban financial markets. Therefore, the expected sign is ambiguous.
Fourth, the ratio of loan loss to total outstanding loans, retirement pay, and annual growth rate of deposits are chosen to represent uncertainty that might affect the utilization of inputs. Loan losses reflect either a failure of credit management or uncontrollable externalities such as policy loans or borrowers' unexpected business failures. In many cases, an employee's premature retirement cannot be expected. Moreover, the retirement pay system in Korea is a lump sum payment, so that the value of these payments is very large. Therefore, if a relatively large number of employees retire, the operating costs will be very high in that year. These variables must be negatively related to the degree of efficiency.
The variables for the regression analysis of the input inefficiency sources are operationally defined as follows;
Dependent variable:
The degree of input efficiency = 1 minus the measured overall input inefficiency.
Independent Variables:
(1) Employees' Incentive (MOTIVE) = total wages and fringe benefits/ number of
employees, 154
(2) Population Density (PD) = population of the market area (thousand persons)/ market
area (km^),
(3) Farm Household Ratio (FHR) = number of farm households in the market area / total
households in the area,
(4) Geographical Characteristics : if flat, D1 = 1; others, D1 =0;
if hilly, D2 = 1; others, D2 = 0
if Up, D3 = 1; others, D3 = 0,
(5) Deposit Size (DS) = average stock value of deposits,
(6) Number of Cooperative Members (NCM),
(7) Policy Loan Ratio to Mutual Credit loans (PTOM) = average stock value of policy
loans / average stock value of Mutual Credit loans,
(8) Banking Business Income Share (BBIS) = gross banking business income / total
business gross income,
(9) Deposit Ratio to Mutual Credit loans (DTOM) = average stock value of deposits/
average stock value of Mutual Credit loans,
(10) Loan Loss Ratio to Total Outstanding Loans (LLOS) = value of loan losses /
(average stock value of Mutual Credit loans + average stock value of policy loans),
(11) Retirement Payment (RETP) = total retirement payments / number of employees, and
(12) Annual Growth Rate of Deposits from 1986 ~ 1991 (AGRD).
There is no a priori information about the functional relationship between measured input efficiency and the possible sources of that inefficiency. Since this study proposes simply to determine a relationship between measured input inefficiency and its possible sources, OLS can be used as an acceptable approach. OLS using the stepwise regression method was applied in this case. CHAPTER VI
EMPIRICAL RESULTS
In this chapter, the estimation results of the models and the evaluation results of the cost structure of the estimated cost frontier are presented. The first section presents the mean values of the variables and product mix by deposit size classes representing the size of the PACs banking business. The second section provides the results of the model specification tests for heteroscedasticity and the existence of a cost frontier. In the third section, the estimation results for the frontier cost function and its evaluation are presented. In the fourth section, the test results for production technology are presented.
Nonjointness in inputs, separability, and homotheticity were tested to capture the characteristics of banking production. The fifth section provides the evaluation results of output efficiency, such as economies of scale and product mixes, from the estimated frontier cost function. This section also contains estimates of ray scale economies (RSCE), expansion path cost efficiency (EPCE) measured by the formula specifically developed in this study, and expansion path subadditivity (EPSUB). The sixth section presents the degree of overall input inefficiency and the regression results about the sources o f this inefficiency. The last section summarizes the overall findings of the empirical analysis.
155 156
6.1 The Mean Value of the Variables and Product Mix bv Deposit Size Class
Table 6.1 gives the mean value of the variables in the sample data with their minimums and maximums. Table 6.2 presents a comparison of the mean values between rural and urban PACs. As discussed in Chapter IV, there are large size differences in total costs and output variables between rural and urban PACs. In contrast, the input price variables show an opposite relationship; that is, those for rural PACs are higher than those for urban PACs. In the case of wage rates, the rural PACs have relatively more high ranking employees than the urban PACs, which results in higher wage rates in the rural
PACs. The portion of third or higher class employees relative to total bank employees is
Table 6.1 The Mean Value of Cost Function Variables : Aggregate Sample of Total
Observations
Variables Mean Minimum M aximum Total Costs (million won) 1,433 481 6,439
Outputs (million won) Deposits 8,967 2,125 51,702 Mutual Credit Loans 6,901 1,194 29,578 Policy Loans 4,256 1,286 10,865
Input Prices Labor Price (thousand won) 1,4625 1,0405 1,8350 Capital Price (won/won) 0.0055 0.0027 0.0103 Deposit Price (won/won) 0.1150 0.0887 0.1873 Borrowing Price (won/won) 0.0499 0.0235 0.0761
N um ber of Branches per PAC 0.6585 0 6 157
Table 6.2 The Variable Means for the Average Rural and Urban PACs
Variables Rural (A) Urban (B) A/B Total Costs (million won) 1,045,600 2,878,000 0.36
Outputs (million won) Deposits 5,839,000 20,706,000 0.28 Mutual Credit Loans 5,013,200 13,999,000 0.36 Policy Loans 3,976,000 5,207,700 0.76
Input Prices Labor Price (thousand won) 14,755 14,144 1.04 Capital Price (won/won) 0.0056 0.0048 1.16 Deposit Price (won/won) 0.1170 0.1080 1.08 Borrowing Price (won/won) 0.0507 0.0471 1.08
Number of Branches 0.41104 1.5814 0.26
57.2 percent in the rural PAC, but only 54.0 percent in the urban PACs. The difference in capital price, which was defined in the previous chapter as total management expenditures for maintenance of capital or materials divided by deposits, means that rural PACs pay more costs for savings mobilization because of market environment. Deposit price and borrowing price are related to the composition of accounts and the time period to keep the accounts as discussed in Chapter IV. The number o f branches is very small; its maximum value of the rural PACs is only 2. Thus, this variable may not have enough variation to estimate a meaningful relationship with total costs.
The PACs were divided into seven classes by deposit size to evaluate the estimated cost frontier function. The distribution of deposit size classes is presented in Table 6.3. 158
Table 6.3 The Distribution of Sample PACs by Deposit Size Class by Region
Deposit Size Rural Urban Total CIasses(billion won) No. % No. % No. % Less than 3 7 4.29 0 0.00 7 3.40 3 - 5 77 47.24 0 0.00 77 37.38 5-7.5 46 28.22 2 4.65 48 23.30 7 .5 -1 0 19 11.66 6 13.95 25 12.14 10-15 13 7.98 8 18.60 21 10.19 1 5 -2 5 1 0.61 15 34.88 16 7.77 Greater than 25 12 27.91 12 5.83 Total 163 100.00 43 100.00 206 100.00
Most PACs fall in the range of 3 billion to 7.5 billion won (59.7 % of total observations).
The rural PACs tended to lie at the lower end of this range, while the urban PACs tended to have more than 15 billion won in total deposits.
Table 6.4 presents the product mix for each deposit size class of PACs measured at the class mean. As the total deposit size increases, the deposit share of the sum of total outputs steadily increases, while the policy loan share steadily decreases. On the other hand, the Mutual Credit loan share is fairly constant; that is, the largest difference between classes is only 4 percentage points (from 33 to 37 percent). Since the increase in deposit share or decrease in policy loan share is gradual, there are no big differences in product mix between neighboring classes. But there is a relatively large change in product mix between the third (5 - 7.5 billion) and the fourth class (7 .5 -1 0 billion won) in terms of the share gap between deposits and Mutual Credit loans. In the third class, the deposit share (38 %) is 4 percentage points higher than the Mutual Credit loan share, but in the 159 fourth class, the former (43 %) is 9 percent higher than the latter (34 %). Most rural PACs belong to the third or lower classes (80 % of rural PACs) that have deposits similar to
Mutual Credit loans; especially, the first and second class PACs must borrow substantially from the MCSA to supply funds to their markets*. In short, the product mixes of these classes are contrasted with those o f the highest three classes, i.e., urban PACs.
Table 6.4 Product Mix of Deposit Size Class
Deposit Size Deposits Mutual Credit Policy Loans Sum of Class Loans Outputs (billion won) Value^ Share Value® Shareh Value® Share** b Less than 3 2,710 0.33 2,751 0.34 2,630 0.33 8,091
3 - 5 4,006 0.36 3,743 0.33 3,515 0.31 11,265
5-7.5 6,181 0.38 5,502 0.34 4,607 0.28 16,291
7.5 - 10 9,008 0.43 7,186 0.34 4,873 0.23 21,067
10-15 11,710 0.45 8,805 0.34 5,229 0.20 25,744
15-25 19,749 0.51 14,367 0.37 4,698 0.12 38,815
Greater than 25 35,913 0.58 21,042 0.34 4,924 0.08 61,879
3 The unit of value is one million won. h The share is the ratio o f each output to the sum of outputs.
* Since the reserve requirement is 10 percent of deposits, the average borrowings for Mutual Credit loans in the three lower class PACs can be calculated from Table 6.4. Considering the required reserve rate, the first class fund deficits in the balance of deposits and Mutual Credit loans are 312 million won, and the second class fund deficits are 137 million won. The third class fund surplus is only 61 million won, while the fourth class surplus is 922. Since these estimates are averages, a temporal fund shortage might be substantial in the third class or lower classes. 160
6.2 Model Specification Tests
Before proceeding with the actual estimation of the model, model specification tests were conducted to obtain appropriate estimates of a frontier cost function consistent with econometric theory. This section presents two tests for model specification: the existence of heteroscedasticity in the disturbance term and the existence of a cost frontier different from the traditional cost (average) function. These tests will lead to an appropriate method to obtain estimates of the frontier cost function consistent with theory.
6.2.1 Heteroscedasticity Test
Classical econometric theory depends on the assumption of homoscedasticity of the error term. If this condition is violated, the ordinary least squares (OLS) parameter estimates do not attain the minimum variance among the possible estimation methods, though they are unbiased and consistent. This problem leads to unreliable tests of hypotheses. Since this study uses cross section data, the heteroscedasticity problem is likely to exist.
The Breusch-Pagan-Godfrey (B-P-G) test and the Glejser tests were applied as diagnostic tests for the presence of heteroscedasticity by using the Shazam computer package. The former test is a general test for the hypothesis that the variance of a function is some function of its explanatory variables. It is known that this test rejects the null hypothesis, when it is true, less frequently than indicated by the selected Type I error
(Judge, et al ). The Glejser test permits one to more specifically test for the nature of the heteroscedasticity problem. 161
Table 6.5 Test for Heteroscedasticity
Test Statistics Test Value Critical Value^ Degree of Freedom
e2 on X: B-P-G testa 35.851 43.7729 30
|e| on X: Glejser test ______37.803 ______43.7729 ______30______a e is the error term of the cost function and X signifies a vector of independent variables.
^ %2 statistics at 5 % significant level.
Both test values are less than their critical values at the 5 percent significance level, which means that the null hypotheses, that is, the error terms are homoskedastic, cannot be not rejected statistically as presented in Table 6.5. Thus, the application of OLS can be justified in the sense of independently identically distributed error terms.
6.2.2 Test for the Existence of the Frontier Function
The study assumed that the error term of the cost function is composed of both an inefficiency disturbance and statistical noise, and that the Ainction corrected for the inefficiency error term is a cost frontier. Intuitively, the cost frontier must be located below or equal to the conventional cost function because the existence of the inefficiency disturbance means additional costs from an overuse o f inputs. Thus, the inefficiency error term of the cost function should be positively skewed if the cost frontier is different from the conventional cost function. In the case of a half normal distribution of the inefficiency disturbance, the second and third moments of the OLS residuals can be used to test for 162 the skewness of the error terms. This method is known to be more robust than the maximum likelihood method (Schmidt and Lin; Waldman ). The test statistic is defined as
= m jml , (6 .1)
where m2 and m 3 are, respectively, the second and third moments of the OLS residuals.
One may use the table 34B in Biometrika Tables for Statisticians, Vol. 1, to determine the significance level of
The second and third moments of the OLS residuals from the pooled sample were 0.00111 and 0.0000208, yielding .^ = 0 .5 6 2 3 6 which exceeds the 1 % critical value of
0.403. Therefore, it is statistically justified to use a cost frontier function different from the conventional cost function.
6.3 Estimation Results of the Frontier Cost Function
The estimated results for the cost frontier function using the COLS method are presented in Table 6.6. Four criteria were used to examine whether the estimated cost frontier function is reasonable or not. The criteria are goodness of fit, statistical significance o f the coefficient estimates, the sign of marginal costs, and the concavity condition in input prices. The measures of goodness o f fit, and adjusted R^, are close to 1, which means that the estimated function nearly completely explains the cost structure of the sample PACs. However, this may result from multicollinearity problems that can occur in flexible functional forms. A multicollinearity problem will cause a very high R2 but insignificant estimates. Since the best remedy for this problem is to use more data, most studies which use flexible functional forms with a given amount of data can not avoid 163
Table 6.6 Estimation Result of the Frontier Cost Function
Coeffîcients (Variable) Estimate Standard Error t value oq (intercept) -0.0474** 0.0051 -9.3556 (Inqi, deposits) 0.5441** 0.0208 26.1070 «2 (lnq2. Mutual Credit Loans) 0.3032** 0.0228 13.3130 «3 (Inqg, policy loans) 0.1394** 0.0092 15.1130 «11 ((lnqilnqi)/ 2) 0.7282** 0.0814 8.9442 «22 ((lnq2N 2V2) 0.4781** 0.0565 8.4590 «33 ((lnq3lnq3)/2) 0.1090** 0.0174 6.2685 «12 ((lnqiInq 2)/2) -0.5645** 0.0662 -8.5296 «13 ((Nllnq3V2) -0.1116** 0.0327 -3.4142 «23 ((Inq2lnq3)/2) 0.0349 0.0448 0.7788 Pi (Inwi, price of labor) 0.1258** 0.0317 3.9633 p2 (inw2, price of capital) 0.1129** 0.0148 7.6424 P3 (inw3, price of deposits) 0.6052** 0.0244 24.7660 p4 (lnw4 , price of borrowings) 0.1561** 0.0187 8.3566 P ll ((lnwiInwi)/ 2) 0.9287** 0.3198 2.9038 P22 ((lnw2lnw2)/2) 0.0370 0.0967 0.3829 P33 ((lnw3lnw3)/2) 0.2756 0.2499 1.1028 P44 ((lnw4 lnw4 )/2) 0.2800** 0.1072 2.6124 P12 ((lnwilnw 2)/2) -0.1631 0.1227 -1.3297 Pl3 ((lnwilnw 3)/2) -0.3272 0.2289 -1.4293 Pl4 ((lnwilnw 4 )/2) -0.4384** 0.1555 -2.8184 P23 ((lnw2lnw3>/2) 0.0096 0.1151 0.8370 P24 ((Inw2lnw4 )/2) 0.1164 0.0630 1.8474 P34 ((Inw3lnw4 )/2) 0.0419 0.1343 0.3120 711 (Inqilnwi) -0.0367 0.2537 -0.1445 712 (lnqilnw 2) 0.0004 0.0829 0.4796 713 (lnqilnw 3) 0.6108** 0.1664 3.6703 721 (lnq 2lnwi) 0.0367 0.2537 0.1445 722 (Inq2lnw2) -0.0004 0.0829 -0.4796 723 (Inq2lnw3) -0.6108** 0.1664 -3.6703 0 (InBR, branches) 0.0046 0.0082 0.5616
|l = E(u) = 0.0365 (var(p)) = 0.0007618 (var(u)) = 0.0021 Oy^ (var(v)) = 0.000357 r 2 = 0.9960 ^ ~ «u^ + = 0.002457 Adjusted r 2 = 0.9955 A, = a„/av = 2.42615 164
* ; significant at 5 % level: ** ; significant at 1 % level the problem. Thus, the significance of estimates, is a more important criteria for evaluating the appropriateness of the estimated function, rather than R^. The portion of the estimates that are statistically significant is very high; that is, eighteen of thirty one parameters (58.1
%) are significant at the 1 percent level (t-value is greater than 2.576). This high ratio of significant parameters suggests that the results of the estimated function are reliable.
It is not easy to discriminate between the right and wrong sign o f the estimate in the translog function. In the cost function case, the signs of marginal costs which should be positive are a criteria of discrimination. Since the data were scaled by geometric means, the coefficient sign of first order terms of outputs, aj (i = 1, 2,3) can be conveniently used to examine the sign o f marginal costs. That is, if aj is positive, the marginal cost of an increase in that variable is positive^. The estimated results show that all coefficients o f the first order output terms are positive, > 0; thus, the estimates are reasonable.
Although the condition that costs are homogeneous degree one in input prices can be parametrically imposed on the translog cost function, the regularity condition of concavity of the cost function cannot be imposed. Thus, it is necessary to examine whether or not the condition of concavity in input prices is violated. This can be easily examined using the signs of coefficients when the data are scaled by geometric means. The cost function is concave if the matrix of second derivatives with respect to input prices is negative semidefinite. The matrix is negative definite if and only if the principal minor determinants of order k have the sign ( - 1)^ for k = 1,..., n. In the case of the estimated cost function scaled by the geometric means, the second derivatives with respect to input
dC C ainC d\nC ^ Marginal cost (MC) is —— = — —------, and — ------= (X, since the other log term is zero at the mean dq^ qi d\nqi dlnq^ value. Both cost (C) and output (qj) are positive so that the sign of MC depends on that of a . 165 prices are - pjC/wj^ where pj, C, and wj are the parameter of input price j in log terms, total costs, and price of input j, respectively^. This relationship means that we can evaluate the concavity condition with only the sign of Pj. If the Pj are all positive (the
-pjC/wj2 are all negative), then the matrix of second derivatives must be negative definite.
The estimates o f pj in Table 6.6 are all positive; thus, the concavity condition is satisfied around the geometric means of the observations. Therefore, the estimated frontier cost function satisfies the regularity conditions of the cost function, at least around the geometric means.
Information about the cost frontier that is different from the average conventional cost function is presented below the coefficient estimates in Table 6.6. The expected value o f the one-side disturbance, cost inefficiency (|i), is 0.0365, which implies that costs are, on average, almost 4 percent above the frontier. This is a strikingly low level even though it is somewhat expected considering Figure 4.2. In the following section 6.5, a more detailed discussion will be undertaken. On the other hand, the variance of cost inefficiency, 0^2 = 0^2 . 2)17t, is 0.0007618 and the variance of the random error is 0.000357, which implies that the variance o f cost above the frontier is, on average, about two times as large as the variance of the frontier itself. These two pieces of information indicate that although the variation of the one-side disturbance dominates the variation of the statistical
^ The first derivative of cost with respect to input price is , and the second derivative dWj Wj d\nWj
IS
^ d c d in c^^ ainc a c d ^ainc dWj dWj Wj dXnWj dXnWj dWj Wj Wj dWj dXnWj 166 noise, the variation of total costs is so small that it leads to a very low level o f cost inefficiency.
Since this overall evaluation of the estimated frontier cost fimction satisfies the regularity condition for the cost function and shows a good fit with a high portion of significant parameters, the derived cost structure information from the cost frontier function can be justified in the theoretical and statistical sense. This estimated cost frontier function will now be used as a base for exploring the multiproduct cost surface of PACs.
6.4 Tests for the Characteristics of Production Technology
Test for Nonjointness: It is highly likely that jointness in banking production exists as discussed in section 5.3.1. This possibility of jointness in inputs can be tested by using the following relationship. If inputs are nonjointly used, it must be true that the marginal cost of each output is independent of the level of any other output. That is,
(6.2, dq.dQj where C(Q, W) is the cost function, and q; is output i. The null hypothesis of nonjointness is; Hq: ay = 0, i j, where ay is an interaction term between output i and j. The joint test value for nonjointness is 156.6 which greatly exceeds the critical value 7.8. Therefore the test rejected the null hypothesis at the 1 percent significance level (Table 6.7).
Test for Separability: A necessary and sufficient condition for separability is that the cost function must be multiplicatively separable; that is, C(Q, W) = H(Q)*G(W). This means that if the technology is separable, then the ratio of any two marginal costs is 167
Table 6.7 Joint Test Results of Production Technology
Production Structure Test Statistics^ Critical Value (5 %) Degree of Freedom
Nonjointness 156.58819 a 7.81473 3
Separability 15.07245b 7.81473 3
Homotheticity 8.009323 7.81473 3
® Wald statistics
independent of factor prices. Thus, the changes in factor prices cannot influence decision
making about product mix. In mathematical form, this means
d\nC ld\nqi = 0 51nw, d\nCld\nqj (6.3)
This test can be performed as a parameter restriction such that "Yu = 0, where 7j| is the
interaction term of output i and input price 1. This joint test for separability rejected the
null hypothesis because the test statistics were 15.07 which is greater than the critical
value 7.8 at the 5 percent significance level (Table 6.7).
Test for Homotheticity of Production: The production technology represented
by the translog cost function is homothetic if and only if the cost elasticity with respect to
each output is constant. This requires the following parameter restrictions; Zoiÿ = 0 and
Z7ik = 0 The test rejected the null hypothesis that the cost fimction is homothetic as 168 presented in Table 6.7, though it seemed to be homothetic in the graph presented in Figure
4.2.
Therefore, the banking production of PACs is characterized by joint, non separable, and non-homothetic technology. The jointness of the production technology supports the multiproduct model used to deal with the banking production of PACs. In other words, if we use a single output model with an index such as Benston et al. used, the model would not capture the dependency of marginal costs between outputs'*. Non separability implies that decision making about the product mix is sensitive to input prices, so the use of interaction terms in the model between outputs and input prices is justified. This supports the use of the translog function rather than the quadratic functions that a priori exclude the possibility of non-separability. The non-homotheticity of production implies that the production costs do not necessarily increase at a constant rate as outputs increase. In other words, there is range where cost saving is possible by changing the scale of production.
6.5 Output Inefficiency
6.5.1 Ray Scale Economies
The measured overall ray scale economies show that the PACs exhaust scale economies in earlier stages on the cost frontier, which was expected as explained in chapter IV in the descriptive analysis of the banking cost structure. Table 6.8 presents the results for the estimated scale economies. RSC2 in the table refers to the overall ray scale economies when deposits. Mutual Credit Loans, and policy loans are regarded as outputs.
'* Benston et al. used a divisia index to aggregate outputs to a single output that describes the complete structure of the cost function when the aggregation method can be justified, as discussed in chapter II. 169
Table 6.8 Ray Scale Economies of Banking
Deposit Size Class Number of Partial Elasticity Scale Economies (billion won) or Observations Region Deposit MCL» PLb RSClc RSC2d
Less than 3 7 0.4263 0.3632 0.1714 0.7895** 0.9609*
3 - 5 77 0.4609 0.3481 0.1664 0.8090** 0.9754*
5 - 7.5 48 0.4996 0.3260 0.1620 0.8256** 0.9876
7.5 ~ 10 25 0.5664 0.2966 0.1306 0.8630** 0.9936
10 - 15 21 0.6327 0.2479 0.1183 0.8806** 0.9989
15 - 25 16 0.7291 0.2086 0.0575 0.9377** 0.9952
Greater than 25 12 0.8870 0.1026 0.0175 0.9896 1.0071
Rural PACs 163 0.4986 0.3292 0.1551 0.8278** 0.9829*
Urban PACs 43 0.7183 0.2018 0.0797 0.9201** 0.9998
All 206 0.5441 0.3032 0.1394 0.8473** 0.9867
^ Mutual Credit loans
^ Policy loans
^ RSCl is the ray scale economies when only deposits and Mutual Credit Loans are considered as outputs.
^ RSC2 measures the ray scale economies including policy loans as outputs
^ The asterisks mean that the measured ray scale economies are significantly different from 1 at the
respective significance levels which are based on the Wald test statistic; *, 10 %; **, 1 %
significance level. 170 while RSCl represents the overall ray scale economies when only deposits and Mutual
Credit Loans are regarded as outputs. Measuring RSCl was done to evaluate the ray scale economies independent of policy loans. Since the concept of outputs is meaningftil when they are internally controlled by the firm, we may regard the policy loans as a control variable given outside of the PACs. With this specification, we can obtain useful information about the cost structure of the banking business that PACs perform independently. As discussed in chapter IV, RSC2 may not give relevant information about scale economies due to the cheap interest cost effects and the substantially different shares of the policy loans across observations. The cheap policy loans may result in a flat average cost curve as shown in Figure 4.3. Therefore, RSCl may be a more relevant measure for policy implications based on scale economies.
RSC2 is not significantly different from 1 for the overall sample mean and for the average of the urban PACs, while RSC2 for the average rural PAC is statistically different from 1 at the 10 percent significance level. Therefore, only the rural PACs are under slight overall scale economies at their mean level when the policy loans are treated as outputs.
Considering the different deposit size classes, RSC2 consistently increases as the deposit class increases, which implies an increase from slight economies of scale to slight diseconomies of scale. But the existence of scale economies is statistically supported only in the smallest two classes with a low significance level because RSC2 for the lowest two classes show slight overall scale economies at the 10 percent significant level, but the
RSC2 for the other classes are not significantly different from l(no scale economies). In other words, if the deposit size exceeds 5 billion won, there are no cost gains from expanding scale with a given constant product mix when the effects of policy loans on the measured overall ray scale economies are included. This result implies that about a half of 171 the rural PACs cannot gain save costs by expanding the scale of banking production, and most urban PACs are also operating under constant returns to scale.
However, the measured ray scale economies excluding the policy loan effects,
RSCl, result in a substantially different implication. The estimates of RSCl for all classes of PACs, except for the largest class, suggest they are operating under significant economies of scale and that the degree of scale economies consistently decreases as deposit size increases. The regional sample means, RSCl, are 0.8473, 0.8278 and 0.9201, at the overall, rural and urban sample means, respectively. In all cases except for the largest class, the value o f RSCl is different from 1 at the 1 percent significance level. Only the largest class PAC exhausts scale economies when the effects of policy loans on the measured overall scale economies are excluded. This suggests that this class can be an effective standard to be used as the scale efficient class.
What are the implications of the differences between RSCl and RSC2 ? J.H. Kim found that the banks with relatively small policy loans exhibited scale economies, while the other banks with relatively large policy loans exhibited no scale economies in Korea.
Although he could not determine the degree of policy loan contribution to the measured scale economies and did not test for the statistical significance of differences in the measured scale economies, he suggested that policy loans might increase banking costs due to their contribution to unhealthy loans and loan losses^. This explanation means that policy loans raise marginal costs so that the distance between the average cost curve and the marginal cost curve is very small or the latter may exceed the former.
^ He could not classify policy loans as an independent output in his model because he defined revenue from asset accounts as the outputs. The Korean banking accounting system does not classify revenues of loans by type of individual account. Thus, he divided banks into two types: the banks with high policy loan ratio to total outstanding loans and the banks with lower policy loan ratio. Then he compared the overall ray economies of scale of the two groups. On the other hand, Y.S. Kim also found no scale economies at the overall sample mean of the PACs for the same reason. 172
Kim's results are not relevant for this study, because if the effect of raising marginal costs is substantial, the small class PACs should reveal higher unit costs of output than the larger class PACs and significant scale economies must be observed along the cost curve. The more important fact in this case must be that the cheap policy loans affect total average costs - including interest expenses on deposits and on borrowings - which are lowered by mixing funds, as discussed in chapter IV. Thus, since the shares of policy loans are relatively constant over the PACs, the smaller the bank size, the greater the effect of lowering total average costs. Consequently, the total average costs of small
PACs with a large share of policy loans becomes lower, which likely flattens the total average cost curve over the entire range of observations as shown Figure 4.3. For this situation, overall scale economies cannot be observed. Even in the case of deposit money banks that Kim studied, this effect might be true. Of course, the factors that Kim suggested cannot be ignored either, but the effect of policy loans on increasing marginal cost must be offset by reducing average costs through the effects of mixing funds. The fact that the share of loan losses, at the overall sample mean, was only 0.8 percent of total banking costs supports the argument that the effect of mixing funds is to lower the average costs of policy loans.
The conclusion appears to be that when the measured ray scale economies exclude the effect of policy loans, the PACs can save substantial costs by increasing the scale of production up to the largest class of PACs. For the smallest rural PACs with deposits of less than 7.5 billion won, the overall marginal cost is only about 80 percent o f the overall average cost if we exclude from banking activities the policy loans made by PACs. 173
6.5.2 Expansion Path Cost Efficiency
Expansion path cost efficiency (EPCE) measures the potential cost savings that could be obtained by changing the scale of production and/ or product mix as explained in
Chapter V. As presented in Table 6.4, each deposit size class had a different product mix.
There were substantial differences in product mixes between the smaller size rural PACs and the larger urban PACs, but no large differences in product mixes between the neighboring size classes. This suggests that it will not be a relevant strategy for a small
PAC to increase the scale of production while holding product mix constant. Therefore, in a more practical sense, the EPCE that allows for the effects on costs of changes in both scale and product mix can give valuable information about cost structure to the manager of a small PAC. But the EPCE has a drawback in treating the effects of policy loans. The
EPCE is not developed enough to measure the effect of a partial change like RSCl in the previous section, so the degree of cost changes along the expansion path independent of policy loans was not analyzed in this study. The following evaluation results, therefore, are based on the specification of all three outputs - deposits, Mutual Credit loans, and policy loans.
Tables 6.9 and 6.10 present the measured EPCE, the degree of total cost saving along a expansion path, for both cases of neighboring classes and between each class and the most scale efficient class, i.e., the largest class. Though a number of class comparisons are possible, only two sets of comparisons are reported. Table 6.9 shows the degree of cost saving when a lower size class PAC changes its scale o f production and product mix to equal those of the next higher class. For example, if the PAC in the lowest class evolves to the next class whose deposit size is in the range from 3 to 5 billion won, then total cost savings will be 0.5 percent. 174
Table 6.9 Expansion Path Cost Efliciency to Next High Class
Deposit Size Class (billion won) Expansion Path
Lower end Higher end Cost Efficiency
0 - 3 3 - 5 0.9819626
3 ~ 5 5 ~ 7.5 0.9954621
5 ~ 7.5 7.5 ~ 10 0.9783021
7.5 ' 10 10 - 15 0.9988095
10 ~ 15 15 ~ 25 1.0020450
15 ~ 25 25 ~ 0.9960276
The degree of cost saving between two neighboring classes is not substantial, i.e., less than 2 percent. This result is consistent with the measured overall ray scale economies
(RSC2) that presented no large differences in RSC2. In fact, since there were no substantial differences in product mixes between neighboring classes (Table 6.4), the
EPCE measured between neighboring classes must be close to RSC2. The result of the
EPCE between neighboring classes does not provide enough information to evaluate competitive viability since the information is too limited in the narrow ranges of these observations. Therefore, to evaluate competitive viability, a comparison was made between each individual class and the most efficient class.
Table 6.10 presents the degree o f cost saving when the small size classes change both their scale and product mix to that of the largest class. The results show that the smallest three size classes can save costs if the scale of production and product mix reach those of the largest class, but the large size classes with deposits greater than 7.5 billion 175
Table 6.10 Expansion Path Cost Efficiency to Scale Efficient Class
Deposit Size Class (billion won) Expansion Path
Lower end Higher end Cost Efficiency
Less than 3 Greater than 25 0.8519486
3 - 5 Greater than 25 0.9153974
5 - 7.5 Greater than 25 0.9324724
7.5 - 10 Greater than 25 0.9949096
10 - 15 Greater than 25 0.9977796
15 - 25 Greater than 25 0.9960276
won do not realize any significant cost saving. The measured EPCE suggests that if the smallest class reached the largest class, total costs would be reduced by IS percent, while for the second and third classes the total costs saved reach 8 and 7 percent, respectively.
However, the larger three classes would save less than 0.5 percent, which is meaningless.
Therefore, a PAG that belongs to the deposit size class over 7.5 billion won is competitively viable while the smaller ones are not. A substantial portion of both rural
(11.6 %) and urban (12.1 %) PACs fall in the class of deposit size between 7.5 and 10 billion won which is output efficient compared to the largest class. It appears, therefore, that the small rural PACs can save costs by simultaneously making a large increase in scale and in changing product mix. 176
6.5.3 Expansion Path Subadditivity
The expansion path subadditivity (EPSUB) evaluates whether or not the combined output of a representative PAC in size class i would be produced at more or less cost by a separate representative PAC in the next smallest size class h plus a separate potential PAC which produces the residual or complementary output bundle (Berger et al., 1986). In other words, it evaluates whether or not it is desirable to divide a representative PAC in a class into a smaller PAC (incumbent) and a potential PAC (new PAC) that wants to enter the financial market.
A problem for obtaining EPSUB concerns how to deal with the input price variables and the branch variable. There are two sets of input price variables (for smaller and larger class representative PACs) and the number of offices at the class mean may not be an integer. By following Berger et al.'s method, this study used the input prices of the next higher class PAC, and the number of offices was treated as one if the office number o f the higher class PAC minus that o f the smaller class PAC was less than one. Actually, the difference in the number of offices between two class PACs was less than one, so that every potential entrant (PAC) was assumed to have only one office (no branches).
The evaluation result of the EPSUB as reported in table 6.11 shows that it is not efficient, over all classes, to divide a representative PAC into two smaller PACs, i.e., a
PAC in the next smallest size class and a separate potential PAC which produces the residual or complementary output bundle. Therefore, all size classes of PACs are competitively viable against the smaller PACs. The degree of diseconomies obtained by dividing the representative PAC in a class into two smaller PACs shows a relatively large variation from 2 to 21 percent (Table 6.11). Dividing the largest two size classes of PACs into smaller PACs will lead to a substantial increase in costs. For the largest class PAC 177
Table 6.11 Expansion Path Subadditivity (EPSUB) by Deposit Size Class
Deposit Size Class (billion won) EPSUB Higher Class Lower Class
3 - 5 Less than 3 0.0384
5-7.5 3 - 5 0.0168
7.5 - 10 5-7.5 0.0570
10-15 7.5-10 0.0289
15-25 10-15 0.0790
Greater than 25 15-25 0.2092
(greater than 25 billion won), the cost advantage of maintaining one multiproduct PAC rather than two smaller PACs is about 21 percent of the current production costs, while dividing the second largest PAC will increase cost by 8 percent.
Since the largest two size classes of PACs are urban, the evaluation result of
EPSUB implies that urban financial markets are more favorable to the large multiproduct
PACs than are rural financial markets. This finding supports the existence of large banking institutions in urban areas. Furthermore, the impact of financial liberalization will be substantial for urban PACs. On the other hand, the class of 7.5 - 10 billion won is the largest class of rural PACs that shows the highest degree of diseconomies (6 %) as a result o f dividing it into smaller PACs. That is, rural PACs also will be affected by the entrance in the RFMs of large banking institutions. 178
6.6 Input Efficiency
6.6.1 Degree of Input Inefficiency
The degree of firm specific input inefficiency was at a strikingly low level as mentioned in section 6.3. The expected value o f cost inefficiency, E (u) = |X, means that average total costs at the overall sample mean exceed the cost frontier, but this measure does not give information about the cost inefficiency for each observation. According to
Jondrow et al., observation specific cost inefficiency can be obtained by either equation
(5.15) or (5.16). The measured inefficiency by using both equations was less than E (u); that is, these results suggested an inefficiency, on average, of 1.8 and 1.1 percent respectively. Table 6.12 presents the distribution o f observations divided into inefficiency groups, which was obtained by equation (5.16) that showed a wider distribution than equation (5.15) . Only 5.3 percent of the observations showed an inefficiency level greater than 5 percent inefficiency, and 18.9 percent of the observations showed from 2 to 5 percent inefficiency when equation (5.16) was used.
Table 6.12 Distribution of Cost Inefficiency Groups*
Observation Specific Inefficiency < 2 % 2 - 5 % 5 % < Total
Observations 156 39 11 206
Percent 75.7 18.9 5.3 100
* These inefficiency groups are classified by using the mode of p. 179
These results are in striking contrast to the results of U.S. commercial banking studies that reported about 20 ~ 40 percent input inefficiency. Since there are no studies about input inefficiency in the Korean banking industry, there is no basis o f comparison for these results. There are several reasons that may explain why the measured input inefficiency is so low for the PACs.
First, the PACs essentially employ the same banking technology and almost all the existing information about cooperative performance is well known to the PACs, while
U.S banks operate in a competitive environment in which there are important economic advantages to those banks that can develop cost efficient methods and can prevent or slow the transfer of information about them. Further, the performance of the PACs is evaluated by the NACF which in turn provides each PAC with the results®. Thus, the technology of every PAC might be standardized so that the utilization level of inputs will not deviate from a standard level.
Second, the measured cost inefficiency might be affected by the estimation method used to generate the frontier cost fimction. That is, maximum likelihood estimation
(MLE) might have given different results. Although the literature on frontier function studies does not provide consistent evidence of differences in the measured inefficiency among the different estimation techniques, Greene (1990) reported substantial differences in the average inefficiency E(u) obtained by using different estimation methods; that is,
E(u) of the COLS was 0.0065, while that of the Ægner, Lovell, and Schmidt (ALS) model and that of the Greene gamma model were 0.0988 and 0.1146, respectively. The expected values of inefficiency obtained from MLE were about sixteen times larger than those from
COLS, which implies that the degree o f measured inefficiency can be substantially affected
® A more detailed discussion was presented in the section of 5.5 to explain the reasons why there may be limited differences in technology between rural and urban PACs. 180 by selection of the estimation technique. If future estimates show a consistently lower result from the COLS method, it is possible that the estimates obtained in this case will also underestimate the true inefficiency.
6.6.2 Sources of Input Inefficiency
In an attempt to identify the possible sources for the measured inefficiency, a stepwise regression analysis was conducted using the variables specified in section 5.6.
But the result was very poor. When all specified variables in section 5.6 were used as explanatory variables, no parameters were statistically significant. When stepwise regression was used, the results revealed only three statistically significant variables (Table
6.13). The explanatory power of the regression was poor, but the F statistic rejected the null hypothesis of all coefficients being equal to zero. The and F-value are 0.085 and
6.258, respectively. This result is similar to those of Aly et al. that produced = 0.06 and an F-value = 4.91 in a U.S. commercial bank study. The significant variables were the ratio of policy loans to Mutual Credit loans (PTOM), ratio of loan losses to total outstanding loans (LLOS), and the annual growth rate of deposits from 1986 to 1991
(AGRD).
The coefficients have reasonable negative signs implying that dependency on policy loans, loan loss ratio to total outstanding loans, and annual growth rate of deposits increase and input efficiency decreases. Since a PAC cannot screen the borrowers of policy loans, there is a higher probability of delinquency or non-repayments of these loans compared to Mutual Credit loans. The bad loan accounts increase managerial costs for collection o f repayments through legal proceedings or loan losses which implies the overuse of inputs. On the other hand, if a PAC is inefficient in utilizing inputs, the policy 181
Table 6.13 The Sources of Input Inefficiency
Variable Coefficient Standard Error tv a lu e
PTOM -0.0037** 0.0014 -2.6937 LLOS -3.3452** 1.1260 -2.9708 AGRD -0.000025* 0.000011 -2.3121 GONSTANT 1.0048** 0.0050 202.6700
R-Square 0.0850 R-Square Adjusted = 0.0715
F = 6.258**
Note: *; significant at 5 %; **; significant at 1 %.
loan ratio to Mutual Credit loans could be higher. The high ratio of the loan losses ratio may come from either failures of careful management of loans or unexpected accidents, but if other things are equal, the high ratio of loan losses implies the failure of loan management. If a PAG has bad loans, it should pay more expenditures to collect payments, which implies the overuse o f inputs. The high annual growth rate of deposits reflects that the financial market expanded rapidly, so a PAG may have increased its capital investment for the future or failed to use inputs optimally because it could not precisely predict market changes.
The excluded variables - market environment variables representing market competitiveness, employee's motivation, and managerial strategy - did not have a statistically significant relationship with input efficiency. One possible explanation of this is the presence of consumer economies that reduce consumer's transaction costs. In the competitive markets, financial institutions may use more inputs to attract more customers 182 by providing convenient facilities, advertising, etc. Since these consumer economies are not included in the model, the costs o f the PAC in the more competitive markets, such as the urban market, might be offset by the cost savings from efficient utilization o f inputs.
More importantly, input inefficiency may be related to the human capital factor that was not adequately captured in the model. It may be reasonable to assume that a complete explanation of input efficiency sources is impossible even if the measured efficiency is correct.
6.7 Summary
In this chapter, the frontier cost function of the PACs was estimated by COLS in order to evaluate the production efficiency of the PACs through the cost structure. The evaluation of the cost frontier was conducted to test the following hypotheses; 1) the
PACs are operating under economies of scale in the production of banking services, 2) the
PACs have economies o f scope in producing banking services, 3) some PACs do not fully utilize inputs in producing banking services, and 4) input inefficiency is related to employee motivation, market environment, managerial strategy, and uncertainty.
Before proceeding to test the hypotheses 1), 2) and 4), a test for proposition 3) was conducted to determine if the frontier for the cost function is different from the traditional cost function (average function). The existence of a cost frontier different from the average function implies the existence o f systemic input inefficiency. The test supported the existence of systemic input inefficiency on average. Therefore, the use of the cost frontier function rather than the traditional average function was justified in evaluating the banking cost structure of the PACs. All evaluations conducted cover the 183 overall sample, the rural and urban subsamples, and the means of the seven deposit size classes in order to determine how the measures vary across regions and sizes of PACs.
The ray scale economies (RSCE) measure developed by Baumol et al. and the expansion path cost efficiency (EPCE) measure developed by the author were used to test the hypothesis 1) concerning cost changes from changing the scale of operation. When the policy loans were considered as normal outputs, the RSCE measure supported the hypothesis, but only at the mean level for the rural PACs and at the class means for the two smallest size classes of all PACs. An interesting result is that diseconomies of scale were not found in the range of PACs included in the sample although constant returns to scale were found over a wide range in sizes of observations. This suggests that there is no bound to profit maximization; that is, as long as a PAC can increase scale within this size range, its profits will increase. The exhaustion of scale economies in fairly small size PACs and the wide range of no scale economies appears to be explained by the existence of cheap policy loans and their share in total output of banking services. When the policy loans were excluded, there were significant scale economies over all size groups of observations, except for the largest deposit size class. Thus, the results for most groups of observations supported hypothesis 1).
On the other hand, the EPCE measure supported hypothesis 1) only when the lowest three size classes of PACs increase their production scale and change their product mix to the scale and product mix of the largest size class. This result was not supported, however, for the upper three size classes. Since EPCE could not evaluate the impact of policy loans, the test was based on the assumption that the policy loans are regarded as normal output. This result implied that small rural PACs can save unit costs in producing banking services by increasing scale and changing product mix, for example through mergers o f two or three PACs. 184
The test for hypothesis 2) was conducted by using the measure of expansion path subadditivity (EPSUB). The result supported the proposition that monitoring all financial services in a single PAC results in cost savings compared to dividing the function among two or more specialized small PACs.
The degree of input inefficiency as measured by the COLS method was very low.
This might be related to either the technology itself or to the method of estimating the cost frontier function. There are logical reasons to expect that the technology used by the
PACs is fairly similar, but the possibility of estimation bias could not be verified since the study employed only one estimation method. The analysis of possible sources of input inefficiency did not strongly support hypothesis 4). The estimated function did not produce a very good fit and only few variables were statistically significant in explaining the degree o f input inefficiency. But the signs of the significant variables were reasonable.
The results suggested that a high dependency on policy loans, large loan losses, and unstable market conditions represented by a large annual growth rate of deposits tended to increase the degree of input inefficiency of the sampled PACs. CHAPTER VII
CONCLUSION
The banking businesses of the primary agricultural cooperatives (PACs) in Korea have successfully grown during the past twenty years. This successful growth in PAC banking must be strong evidence of viability in the sense of dynamism. However, this does not necessarily mean that PAC banking is efficient enough to provide financial services at minimum costs. It must be a natural question whether or not the PACs efficiently perform financial intermediation with a given technology. Considering the regulations on financial markets that may undermine the creative operations of financial institutions, a suspicion of this kind becomes stronger. Moreover, as financial liberalization continues in the future, the viability of PAC banking needs to be examined in terms o f production efficiency in order to determine if it will be able to withstand competition.
The banking industry in Korea is strongly regulated, even though it has undergone gradual deregulation since 1980. The pervasive regulation creates suspicion about the production efficiency of the industry. It is generally recognized that regulation has undermined the creative operations of financial institutions and has fostered production inefficiency through failing to stimulate competitiveness, through frequent and severe intervention in decision making of the banks by the monetary authority, and through substantial subsidies to protect the industry from loan losses. Some of the regulations resulted from a belief that a bank as a public institution cannot go bankrupt under any
185 186 circumstances. The banking sector of the PACs is no exception, although the FACs are one form of non-banking financial institution whose viability is not guaranteed by the government. To date, not a single PAC has gone bankrupt since they were launched in
1973, which implies not only the success of their businesses but also the existence of strong support for the PACs. But changes underway in the financial market may change the situation. Financial liberalization, through removing some of the major interventions in the markets and the financial institutions, will enhance the efficiency of the financial industry. In this situation, only efficient institutions should be viable. Thus, concerns about production efficiency, especially economies o f scale and scope, have increased in recent years. But beyond these questions, if the regulations resulted in passive operations of the banking industry, some of the financial institutions must have underutilized inputs. If this is true, the information about how much they underutilized inputs is also important in order to know the optimal scale and product mix. With these questions in mind, this study started to evaluate the cost efficiency of the banking business of PACs.
To answer these questions, the study used the frontier cost function approach relying on the dual relationship between the production process and the cost function. The general hypotheses for the study were established as 1) the banking businesses of the
PACs operate under economies of scale, 2) the banking production of the PACs has economies of product mix, 3) some o f the PACs overuse inputs, i.e., systemic input inefficiency exists, and 4) this input inefficiency can be explained by the motivation of employees, the characteristics of market environment, managerial strategy, and uncertainty. To test these hypotheses, some measures concerning the multiproduct cost surface were used.
To select the appropriate measures for the analyses of the cost efficiency, the choice of functional form and a concept of the frontier consistent with theory must first be 187 determined. This study selected the translog cost function because of its many desirable properties and the corrected ordinary least square (COLS) method to obtain the frontier cost function. However, the selection of the translog functional form did not allow the evaluation of economies of scope and product specific scale economies since the function does not allow zero terms in variables which is necessary for obtaining meaningful measures of economies of scope and product specific scale economies. But there is second reason for foregoing the measurement of those economies; and that is that the results of the measures o f scope economies and of product specific scale economies may be questionable since they depend on information from well outside of the sample range.
Thus, this study regards only the information within the sample range as valuable. To obtain more meaningful information from the observations, the sample was divided into seven deposit size classes as a measure of banking business volume and the three regional groups, rural, urban, and overall. The mean of each grouping was regarded as representative for its group or class.
The ray scale economies (RSCE) and the expansion path cost efficiency (EPCE) were used to measure scale economies. The latter measure was developed in this study to replace the expansion path scale economies (EPSCE) measure developed by Berger et al.
(1986). The EPSCE was found to have some drawbacks that do not allow consistency in evaluation of results. The EPCE, which extended the idea of examining the concavity (or convexity) condition of a function, is general enough to measure potential overall cost savings by simultaneously changing production scale and product mix.
To test economies of cost complementarity, the expansion path subadditivity
(EPSUB) was used to find whether or not dividing a PAC into two smaller PACs with same multiple products can bring cost savings. Considering no PAC to be specialized in a specific output, measuring economies of scope may not be meaningful, but the EPSUB, 188 which depends only on sample information, must be a reasonable measure in terms of reality.
The test for the hypothesis 3), the existence of the frontier cost function, was conducted by using the second and third moments of the OLS residuals. Output efficiency such as scale economies and cost complementarities were evaluated by the frontier cost function. Hypothesis 4), the relationship between input efficiency and its possible sources, was tested by using a measured degree o f input inefficiency specific to the PAC and the specified variables to represent the characteristics of the PACs.
Accounting data for 206 rural and urban PACs in Choongnam Do, a typical province in the rural areas, and Kyounggi Do, the suburban province of Seoul city, were used. The sample, that includes all the PACs in Choonam Do and 45 PACs in Kyounggi
Do, consists of 163 rural and 43 urban PACs. From the accounting data, that includes nonallocable operating costs due to the nature of multibusinesses, the banking sector was separated by using the labor allocation index which has been used by the national agricultural cooperatives federation (NACF).
Unfortunately the data have only won (Korean money unit) values of output and no information about the number of transaction accounts which would allow using the production approach. Thus, this study used the intermediation approach that is preferred to the production approach in evaluating the viability o f banking institutions.
7.1 Summary of Main Results
The test to determine if the cost frontier is different from the traditional average function supported the use of the frontier cost function. The tests for production technology supported the use of the multiproduct cost function and of the interaction 189 terms between input prices and outputs. With these statistical justifications, the evaluation of cost efficiency was conducted using the estimated frontier cost function.
The measured RSCE showed that only the rural sample mean and the two smallest deposit size classes were operating with slight scale economies, while the other five classes, the urban sample mean, and the overall mean exhibited no scale economies - they experienced constant returns to scale, when policy loans are regarded as normal outputs.
This result was clearly questionable since the scale economies were exhausted early, and the range of constant returns to scale was large without the process undergoing diseconomies of scale. The fact that no scale economies existed in the large range of observations was related to the policy loans, whose interest rates were at a very low level, and to the different portions of the policy loans along the observations. The policy loans are different from normal outputs in the aspect of control. The amounts of policy loans are externally determined by the government, so that a PAC cannot increase or decrease the quantity o f policy loans by employing the marginal cost and marginal revenue principles.
Therefore, we may regard the policy loans as an externally determined control variable. By specifying the policy loans in this way, the ray scale economies depending on only the outputs that the PAC determined independently were measured. In contrast to the overall ray scale economies including the policy loans as output, the ray scale economies measure without policy loan effects showed significantly and consistently decreasing scale economies as the deposit size class increased. The only case of no scale economies was the largest class. In other words, the impact o f policy loans on the measured overall ray scale economies hid the possibility of cost savings through expanding the production scale of the banking business. Therefore, the measured ray scale economies without the policy loan impact should be considered as the true measure of scale economies, representing the location of a specific PAC on the cost surface. 190
Unfortunately, however, the specification of the cost frontier which excluding the policy loans from the definition of output could not be done in measuring EPCE and
EPSUB. Thus, the information from the EPCE and EPSUB should be considered in relation to the overall ray scale economies including the policy loan impact on scale economies.
The alternative measure of expanding scale effects accompanied by changing the product mix, EPCE, tested two cases; that is, the possibility of cost savings when the lower class PAC expands its scale to the scale of the next higher class, and when the lower class expands its scale to the scale of the largest class. The first case showed that there is no advantage in cost savings in all classes. The second case demonstrated that the lower three classes can save costs; and the lowest size class can save costs up to IS percent.
That is, the small rural PACs can save unit costs only by largely expanding and making big changes in product mix. Therefore the two scale efficiency measures, RSCE and EPCE, supported idea that efficiency could be increased by the expansion o f production and changes in product mix in the small PACs.
The last output efficiency measure, EPSUB, supported the conclusion that any class PAC has cost advantages over an existing smaller PAC and its potential supplementary PAC with the same amount of additional output. The degree of cost advantages from being an incumbent became larger as the class was larger, which supported the hypothesis that urban financial markets are favorable to large multiproduct banking businesses.
The level of input inefficiency, on average, measured fi-om the estimated frontier cost function was at a very low level (about 4 %), which contrasts with the estimates for
U.S. commercial banks (20 ~ 50 %). This was supported by measuring PAC specific input inefficiency. There may be two reasons for the low level of measured input inefficiency: 191 first, there was no large difference in banking technology among the PACs and, second the estimation technique may result in a substantial bias in the measured degree of input inefficiency.
Regression analysis was undertaken to determine if there existed any systemic relationship between input inefficiency and its possible sources, even though the input inefficiency was lower. Most explanatory variables were proxy variables because of data constraints. The result was very poor and any respecification of the proxy variables provided the same results. Among twelve variables tested, only three were found to be statistically significant using the stepwise regression method. These results suggest that the
PACs with a high dependency on policy loans, with high loan losses, and with a rapidly expanding deposit size tend to overuse inputs. But the other variables representing employee's motivation, market environment, and managerial strategy were not statistically significant in explaining the differences in input inefficiency.
7.2 Contributions and Limitations of the Study
The important contributions of this study in terms of methodology are three. The first contribution was in identifying the drawback of the expansion path scale economies measure developed by Berger, Hanweck and Humphrey(1986). This formula cannot give consistent implications because of the lack of generality. This drawback seems to undermine the popular application of this measure to empirical studies.
The second contribution was to develop an alternative measure of the EPSCE.
This study extended the idea of examining the concavity (or convexity) condition to measure the potential cost savings by expanding the scale of production and changing the product mix. This new measure was named as the expansion path cost efficiency (EPCE). 192
The EPCE is general enough to measure the effects of cost changes between two coordinates on the cost surface, and this gives direct information about the degree of cost change rather than using the indirect information such as RSCE that gives only information about whether or not scale economies exist.
The third one is that the policy loan effect on the measured ray scale economies is explained, and because of that this study can be used to obtain appropriate policy implications. This finding should be considered when using the intermediation approach in similar circumstances in developing countries where many development banks use subsidized funds or cheap policy loans.
On the other hand, some limitations of this study require careful explanation when interpreting the analysis results. First, this study used the strong assumption o f separability between the banking sector and the other business sectors of a PAC in order to apply the econometric model. This assumption excludes a priori the effects of other businesses on the banking business, and these effects may be substantial considering the competitive edge of multibusinesses. Furthermore, the labor allocation index that was used to separate the banking and non-banking portions of operating costs may not be an exact measure of resource allocation; therefore, the results cannot avoid a possible bias due to the use of this index. Therefore, in future research, a more complete model should be constructed to represent the characteristics of the multibusinesses without incurring the aggregation bias or the inconsistent measurement of outputs. So far, some studies were found in Japan and
Korea, which used gross business income of the individual business sector as the appropriate output metric to handle the entire business sector. But the appropriateness of using this output metric has not yet been theoretically supported in the literature.
A second limitation is that this study focused only on cost advantages to the PACs themselves without considering the cost benefits for PAC clients and these may be an 193 important factor in explaining the competitive edge of the PACs. In other words, a PAC can appear inefficient in utilizing inputs, in operating scale, or in combining product mix, but if we consider the total transaction costs of both the PAC and its clients, the PAC may be evaluated as efficient. This bias was appropriately pointed out by Berger et al. and
Hunter et al. When this possible bias is considered, we may infer why the urban PAC in more competitive financial markets did not show any evidence o f more efficiency in utilizing inputs. The competitiveness can result not only in efficiency in utilizing inputs but also more expenditures to benefit the clientele. That is, the operating costs of the urban
PACs may be higher than that those of rural PACs. Without any information about these circumstances, the measured inefficiency cannot represent the true behavior o f the PACs.
Third, this study used the corrected ordinary least squares (COLS) method to estimate the frontier cost function, which has some advantages such as robustness, representing the true distribution of OLS residuals, and computational ease. But this method is known to be inefficient compared to the maximum likelihood method (Schmidt et al., 1979). Moreover, Greene's study (1990) presented substantial differences in the estimated inefficiency between the two methods, even though it is not known whether or not this difference is general. In other words, the robustness of the results about input inefficiency was not clearly verified in this study.
Fourth, in measuring EPCE and EPSUB, a method to effectively separate the effects of policy loans on these measurements was not developed, so that the study was not consistent in explaining the measures between RSCE and the others.
Fifth, evaluation information depends on only one financial institution that had well unified technology. Thus, the input utilization efficiency might be over evaluated, since the other efficient institutions technology, if they exist, could not be compared. 194
7.3 Policy Implications
The findings reported in this study suggest some policy implications for both the individual PACs and the regulatory authority for financial markets. It should be noted that the limitations o f the study presented above may result in a bias compared with the real world, so that policy implications need to be carefully interpreted.
Above all, the banking businesses of most rural PACs are too small to achieve minimum cost points on the frontier cost function. Therefore, expanding the scale of banking will enhance competitive viability. This suggestion is supported by the ray scale economies which result from exclusion of the policy loan effects on the measured ray scale economies, and on the result of expansion path cost efficiency. If only the former result is accepted, all class PACs except for the largest class of urban PACs (deposit size greater than 25 billion won) should expand production scale to reach a point of cost minimization.
If the latter is accepted, most urban PACs do not need to increase their production scale in order to achieve cost savings. However, a more effective strategy is to expand the scale of banking production at least until it becomes that of the largest urban PACs whose deposit sizes are greater than 25 billion won. This suggestion is based on the fact that diseconomies of scale were not found over the range of observations in the sample. No diseconomies of scale were found even when the policy loans were considered as normal output, or when all cases of EPCE that measured possible cost savings through expanding scale of production and changing product mix are considered. This implies that if there exist any positive business surplus (profits), then even the classes with no scale economies can increase surplus without meeting an upper bound in the sample range. That is, the results support the argument in favor of the merging of small PACs. 1 9 5
The rural PACs with a deposit size of less than 7.5 billion won should be considered for mergers first since their market must be too small. In particular, the PACs with a deposit base of less than 3 billion won can save substantial total costs. The strategy of merging is stressed for the rural PACs for two reasons. One is that the possibility of scale expansion is limited in the existing market area. Another is that additional cost savings are possible through mergers by reducing the high class employees such as the presidents of PACs. In the case of the urban PACs, the market size constraint must not be so binding so that the strategy of merging is not a good alternative. They may expand the scale of banking through increasing branches. The estimated cost function did not show any significant relationship between the number of branches and total costs, which may suggest that the expansion of branches does not create diseconomies.
The existence of cost subadditivity (EPSUB) with scale economies implies the existence of natural monopoly conditions in the banking business of PACs. That is, one
PAC per financial market achieves minimum production costs. But it should be, ited that this information comes from only the banking cost structure of the PACs. With this in mind, the implication becomes different with different market situations in the sample region. In the urban financial markets, there are a number o f financial institutions which are much bigger than the urban PAC, and this study did not use information about those institutions. That is, the reported finding of EPSUB is limited to only the case of the
PACs. Therefore, a suggestion for the large urban PACs is to remain large rather than to divide into smaller PACs. Furthermore, the urban PACs should make more efforts to be viable in the competitive market circumstances of the future, because it may not have competitive viability in that situation, considering the findings of EPSUB. If another big financial institution such as a commercial bank has similar or more efficient technology - i.e., the reported EPSUB is also true for that big institution, then the urban PAC will not 196 be viable as long as it competes in the same market and with the same product mix. In that case, for the urban PACs to be viable, they should differentiate themselves from large financial institutions by creating unique commodities.
In the case of the rural financial markets, however, the PACs actually have a monopolistic market share. Thus, the reported EPSUB suggests that any entrant with the same product mix and with a smaller size will not be viable in competition with the PAC.
This suggestion is stronger than that for the case of urban financial markets since this information covers almost all sizes of large rural financial institutions. Considering the current rural financial market circumstances, it is not likely that other institutions larger than the PAC will enter the market, which means that the potential competitors o f the rural PACs will be small financial institutions. If these small institutions have the same products as the PACs and compete with them, they will not be viable as long as their technology is not more efficient than the technology of the PACs. From this point of view, it may not have been a wise policy for the government recently to allow some special agricultural cooperatives (SAC) to operate banking businesses in their rural financial markets. Furthermore, it may happen that the banking sectors of fishery and livestock cooperatives will need to be merged with that of PACs as long as they will not differentiate commodities or develop more efficient technology. Recently the forestry cooperatives tried to operate banking businesses. A carefiil decision as to whether or not the forestry cooperatives are allowed to operate the banking businesses should be made.
The findings about input inefficiency do not seem to provide strong evidence that the PACs overuse inputs because the estimated levels were so low. But it should be noted that these findings resulted from the comparison o f input utilization among only the PACs that have highly uniform systems, even if the measured input inefficiencies approximate the true magnitudes. 197
The regression analysis between measured input efficiency and its possible sources suggests that the PACs should improve the abilities in handling o f policy loans, in managing the outstanding loans, and in predicting changes in markets. So far, there has been excess demand for funds in the formal financial markets so financial institutions could easily secure good borrowers. It seems that the PACs have emphasized savings mobilization, but the technology for management of loans has been relatively less developed. The PACs should be able to accurately evaluate the projects of borrowers, so that they can obtain reliability in processing information about the borrowers. The government also needs to regard information provided from the PACs as valuable in order to select the borrowers for the policy loans. As financial markets develop, the management of loans will become more important, and the business of selling funds rather than mobilizing funds will be a key issue for the viability of financial institutions in the future in
Korea as well as in developed countries. LIST OF REFERENCES
Adams, Dale W, Douglas H. Graham, and J.D. Von Pischke (ed.) Undermining Rural
Development with Cheap Credit. Boulder, Colorado: Westview Press, 1984,
Adams, Dale W and Delbert A. Fitchett (ed,), Informal Finance in Low-Income
Countries. Boulder, Colorado: Westview Press, 1992,
Adar, Z, A, Agmon and Y, Orgler, "Output Mix and Jointness in Production in the
Banking Firm," Journal o f Money, Credit .and Banking 7(1975):235-44,
Afriat, Sydney, " Efficiency Estimation of Production Functions," International
Economic Review 13(1972): 568-98,
Akridge, J,T, and T,W, Hertel, "Multiproduct Cost Relationships for Retail Fertilizer
Plants," American Journal o f Agricultural Economics 68(1986):928-38,
Aigner, Dennis, C,A, Knox Lovell, and Peter Schmidt, "Formulation and Estimation of
Stochastic Frontier Production Function Models," Journal o f Econometrics
6(1977):21-37,
Alhadeff, D A, Monopoly and Competition in Banking. California: University of
California Press, 1954,
Aly, Hassan Y,, Richard Grabowski, Carl Pasurka, and Nanda Rangan, "Technical,
Scale, and Allocative Efficiencies in U,S, Banking: An Empirical Investigation,"
Review o f Economics and Statistics 72(1990):211-8,
Bailey, Elizabeth and Ann F, Friedlander, "Market Structure and Multi-products
Industries," Journal o f Economic Literature 20(1982): 1024-48,
Baltensperger, E, " Alternative Approaches to the Theory of the Banking Firm," Journal 198 1 9 9
o f Monetary Economics 6(1980): 1-37.
The Bank of Korea. Financial System in Korea. The Bank o f Korea, 1990.
Barnett, William A. "The Miniflex-Laurent Translog Flexible Functional Form." Journal
o f Econometrics 30(1985):33-44.
Barnett, William A. and Yuk W. Lee. "The Global Properties of the Miniflex Laurent,
Generalized Leontief, and the Translog Flexible Functional Forms."
Econometrica 53(1985): 1421-37.
Bauer, Paul W. "Recent Development in the Econometric Estimation of Frontiers."
Journal o f Econometrics 46(1990):39-56.
Bauer, Paul W., Allen N. Berger, and David B. Humphrey. "Efficiency and Productivity
Growth in U.S. Banking." Forthcoming in The Measurement o f Productive
Efficiency: Techniques and Applications edited by Harold 0 . Fried, C. A. Knox
Lovell, and Shelton S. Schmidt: Oxford University Press, 1992.
Baumol, William, John Panzar, and Robert Willig. Contestable Markets and the Theory
o f Industry Structure. San Diago: Harcourt Brace Jovanovich, 1982.
Bell, Frederick W., and Neil B. Murphy. "Costs in Commercial Banking: Quantitative
Analysis of Bank Behavior and Its Reaction to Bank Regulations." Research
Report No. 41, Federal Reserve Bank of Boston, 1968.
Benston, George J. "Branch Banking and Economies of Scale." Journal o f Finance
20(1965):312-31.
______. "Economies of Scale of Financial Institutions." Journal o f
Money, Credit ,and Banking 4(1972):312-41.
Benston, George J., Gerald A. Hanweck, and David B. Humphrey. "Scale Economies in
Banking: A Restructuring and Reassessment." Journal o f Money, Credit ,and
BankingU{mi)-.A35-56. 200
Benston, George J. and Clifford W. Smith Jr. "A Transactions Cost Approach to the
Theory of Financial Intermediation." Journal o f Finance 31(1976):215-31.
Berger, Allen G., and David B. Humphrey. "The Dominance of Inefficiencies over Scale
and Product Mix Economies in Banking." Journal o f Monetary Economics
28(1991): 117-48.
and ______. "Measurement and Efficiency Issues in Commercial Banking."
Forthcoming in Output Measurement in the Service Sectors edited by Zvi
Griliches. Chicago, IL: University of Chicago Press, 1992.
Berger, Allen G., Gerald A. Hanweck, and David B. Humphrey. "Competitive Viability
in Banking: Scope and Product Economies." Paper, Division of Statistics, Board
of Governors of the Federal Reserve System, Dec. 1986.
______, ______, and ______. "Competitive
Viability in Banking: Scale, Scope and Product Mix Economies." Journal o f
Monetary Economics 20(1987):501-20.
Besley, Timmothy, "How Do Market Failure Justify Interventions in Rural Credit
Markets ?" Discussion paper # 162 prepared for the AGRAP division of the
World Bank, Research Program in Development Studies Center of International
Studies Woodrow Wilson School, Princeton University, July 1992.
Binswanger, H.P. "A Cost Function Approach to the Measurement of Elasticities of
Factor Demand and Elasticities of Substitution." American Journal of
Agricultural Economics 56(1974): 377-86.
Boisvert, Richard N. " The Translog Production Function: Its Properties, Its Several
Interpretations, and Estimation Problems." Department of Agricultural
Economics, Cornell University Agricultural Experiment Station, A.E. Res. 82-28,
Cornell University, 1982. 201
Caves, Douglas W., Lauritis R. Christensen, and Michael W. Thretheway. "Flexible Cost
Functions of Multiproduct Firms." Review o f Economics and Statistics 62(1980):
447-81.
Caves, Douglas W. and Lauritis R. Christensen. " Global Properties of Flexible
Functional Forms." The American Economic Review 70(1980): 422-32.
Chang, K.S. " A Structural Analysis of Rural Savings Kyes." Monthly Report of
Agricultural Cooperatives, NACF, September, 1979.
Christensen, Laurits, D.W. Jorgenson, and L. J. Lau. "Conjugate Duality and the
Transcendental Logarithmic Production Function." Econometrica 39(1971):255-
6.
Cho, Yoon Je. "The Effect of Financial Liberalization on the Efficiency of Credit
Allocation: Some Evidence from Korea." Journal o f Development Economics
29(1988): 101-10.
Chung, W.C. Financial Reform. Seoul Korea: Peobmoonsa, 1991. (in Korean).
Clark, Jeffrey A. "Estimation of Economies of Scale in Banking Using a Generalized
Functional Form." Journal o f Money, Credit .and Banking 16(1984):53-68.
Cole, David C. and Yung Chul Park. Financial Development o f Korea 1945-1978.
Cambridge, Mass.: Harvard University Press, 1983.
Cramer, J.S. Econometrics Applications o f Maximum Likelihood Methods. Cambridge
University Press, 1986.
Cuevas, Carlos E. "Intermediation Costs and Scale Economies of Banking under
Regulations in Honduras." Unpublished Ph.D. dissertation. The Ohio State
University, 1984.
Cuevas, Carlos E. "Intermediation Costs in an Agricultural Development Bank: A Cost-
Function Approach to Measuring Scale Economies." American Journal o f 202
Agricultural Economics 70( 1988):273-80.
Dermine, J. and Lars-Hendrick Raller. " Economies of Scale and Scope in French Mutual
Funds." Journal o f Financial Intermediation 2(1992): 83-93.
Diewert, E. " Applications of Duality Theory." in Frontiers o f Quantitative Economics,
vol. 2, edited by M. Intriligator and D. Kendrick. Amsterdam: North Holland,
1974.
Elyasiani, Elyas and Seyed M. Mehdian. "A Non-Parametric Approach to Measurement
of Efficiency and Technological Change: The Case of Large U.S. Commercial
Banks." Journal o f Financial Services Research 4(1990): 157-68.
Evanoff, Douglas D. and Philip R. Israilevich. "Productive Efficiency in Banking."
Economic Perspectives (July/ August 1991), Federal Reserve Bank of Chicago:
11-32.
Evanoff, Douglas D. and Philip R. Israilevich. "Deregulation, Cost Economies and
Allocative Efficiency of Large Commercial Banks." Working Paper Series WP-
1990/19, Federal Reserve Bank of Chicago, 1990.
Evans, David and James Heckman. "A Test for Subadditivity of the Cost Function with
an Application to the Bell System." American Economic Review 74( 1984):615-
23.
Fare, Rilf, Shawana Grosskopf, and C.A.K. Lovell. The Measurement o f Efficiency of
Production. Boston: Kluwer-Nijhoff, 1985.
Farell, M.J. "The Measurement of Productive Efficiency." Journal o f the Royal
Statistical Society. Series A, General 120(1957):253-81.
Ferrier, Gary and C.A.K. Lovell. "Measuring Cost Efficiency in Banking." Journal o f
Econometrics 46(1990):229-45.
Fixler, D. and K. D. Zieschang. "Output and Price Measurement in Commercial 203
Banking." Unpublished Manuscript, Bureau of Labor Statistics, February 7, 1990,
Fry, C., C. Harper, and S. Stansell. " An Analysis of Credit Union Costs: A New
Approach to Analyzing Cost of Financial Institutions, Journal o f Bank Research
12(1982): 239-49.
Fry, M.J. Money, Interest, and Banking in Economic Development. Baltimore: The Johns
Hopkins University Press, 1988.
Gallant, A.R. "On the Bias in Flexible Functional Forms and an Essentially Unbiased
Form." Journal of Econometrics 15(1981):211-45.
Gilbert, R. Alton. "Bank Market Structure and Competition: A Survey." Journal o f
Money, Credit, and Banking 16(1984):617-45.
Gilligan, Thomas W., Michael L. Smirlock, and William Marshall. "Scale and Scope
Economies in the Multiproduct Banking Firm." Journal o f Monetary Economics
13(1984):393-405.
Gramley, L.E. A Study o f Scale Economics in Banking. Kansas City: Federal Reserve
Bank of Kansas City, 1962.
Greenbaum, S.I. "Competition and Efficiency in the Banking System - Empirical
Research and Its Policy Implications." Journal Political Economy 75(I967):461-
79.
Greene, William H. "On the Estimation of a Flexible Frontier Production Model."
Journal o f Econometrics 13(1980): 101-15.
______. "A Gamma Distributed Stochastic Frontier Model." Journal o f
Econometrics 46(1990): 141-63.
Greenwald, B. and J. Stiglitz. "Externalities in Economies with Incomplete Markets and
Imperfect Information." Quarterly Journal o f Economics 101(1986): 229-64.
Gropper, Daniel M. "An Empirical Investigation of Changes in Scale Economies for the 204
Commercial Banking Firm, 1979 - 1986." Journal o f Money. Credit, and Banking
23(1991):718-27.
Grosskopf, S., K. Hayes, and S. Yaisawamg, "Measuring Economies of Diversification:
A Frontier Approach." Forthcoming in Journal o f Business and Economic
Statistics.
Guilkey, David K., C.A. Knox Lovell, and Robin Sickles. "A Comparison of the
Performance of Three Flexible Functional Forms." Working Paper, University of
North Carolina, August 1980.
Hancock, Diana. "Bank Profitability, Interest Rates, and Monetary Policy." Journal o f
Money. Credit, and Banking 14(1985): 179-92.
Horvitz, P.M. "Economies of Scale in Banking." in Private Financial Institutions.
Englewood Cliffs, New Jersey: Prentice-Hall, 1962:1-54.
Huh, Gill Haeng. "A Study on the Economies of Scale in Operations of Agricultural
Cooperatives in Korea." Unpublished Ph.D. dissertation, Seoul National
University, 1992a.
Humphrey, David B. " Flow Versus Stock Indicators of Banking Output: Effects on
Productivity and Scale Economy Measurement." Journal o f Financial Services
Research 6{1992): 115-35.
Hunter, William C., Stephen G. Timme. "Technical Change, Organizational Form, and in
the Structure of Bank Production." Journal of Money. Credit, and Banking
18(1986): 152-66.
Hunter, William C., Stephen G. Timme, and Won Keun Yang. "An Examination of Cost
Subadditivity and Multiproduct Production in Large U.S. Banks." Working Paper
89-3, The Federal Reserve Bank of Atlanta, April 1989.
Hunter, William C., Stephen G. Timme, and Won Keun Yang. "An Examination of Cost 205
Subadditivity and Multiproduct Production in Large U.S. Banks." Journal o f
Money, Credit, and Banking 22(I990):504 - 525.
Jondrow, J., C.A.K. Lovell, I.S. Materov, and P. Schmidt. " On the Estimation of
Technical Inefficiency in the Stochastic Frontier Production Function Model."
Journal o f Econometrics 19(1982): 233-38.
Judge, G.G. et al. Introduction to the Theory and Practice of Econometrics. 2nd ed. John
Wiley & Sons, 1988.
Kalish, Lionel and R. Alton Gilbert. "An Analysis of Efficiency of Scale and
Organizational Form in Commercial Banking." Journal o f Industrial Economics
21(1973):293-307.
Kilbride, Bernard J., Bill McDonald, and Robert E. Miller. "A Reexamination of
Economies Scale in Banking Using a Generalized Functional Form." Journal o f
Money, Credit, and Banking 18(1986):519-26.
Kidwell, David S. and Richard L. Peterson. Financial Institutions and Money, fourth ed.
, Chicago, IL: The Dryden Press, 1990.
Kim, B. J. "Localization, Financial Liberalization and Development of Mutual Credit."
paper presented at the celebration symposium for the 30th anniversary of the
agricultural cooperatives. " The Development of Credit and Banking Business of
the Agricultural Cooperatives toward 2000s.' Seoul: NACF, 1991.
Kim, J.H. "A Study of Scale and Scope Economies of Multiproduct Firms: The Bank
Industry in Korea." Unpublished Ph.D. Dissertation, Joong Ang University, 1990.
Kim, H. Y. "Economies of Scale and Economies of Scope in Multiproduct Financial
Institutions: Further Evidence from Credit Unions: A Note." Journal o f Money,
Credit and Banking 18(1986):220-26.
Kim, Y.S. " An Empirical Analysis of Scale and Scope Economies of Financial 206
Institutions; The Primary Agricultural Cooperatives in Korea." Unpublished
Ph.D. Dissertation, Yonsei University, 1991.
Klein, M.A. "A Theory of the Banking Firm." Journal o f Money, Credit, and Banking
3(1971): 205-18.
Koot, R.S. " On Economies of Scale in Credit Unions." The Journal o f Finance
33(1978): 1087-94.
Kopp, Raymond J. and W.Erwin Diewert. "The Decomposition of Frontier Cost Function
Deviations into Measures of Technical and Allocative Efficiency." Journal o f
Econometrics 19( 1982):319-31.
Lawrence, Colin. "Banking Costs, generalized Functional Forms, and Estimation of
Economies of Scale and Scope." Journal o f Money, Credit, and Banking
21(1989):368 - 379.
Lawrence, Colin and Robert P. Shay. "Technology and Financial Intermediation in a
Multiproduct Banking Firm: An Econometric Study of U.S. Banks 1979-82." In
Technological Innovation, Regulation and Monetary Economy edited by
Lawrence and Shay, pp. 53-92. Cambridge, Mass.: Ballinger, 1986.
Leathers, Howard D. "Allocable Fixed Inputs as a Cause of Joint Production: A Cost
Function Approach." American Journal o f Agricultural Economics
73(1991): 1083-90.
LeCompte, Richard L.B. and Stephen D. Smith. "An Empirical Analysis of Scale and
Scope Economies in the Savings and Loan Industry." Working Paper, Texas
Christian Univ., Sept.,1985.
LeCompte, Richard L.B. and Stephen D. Smith. "Changes in the Cost of Intermediation:
The Case of Savings and Loans." The Journal o f Finance 45(1990): 1337-46.
Lee, D.H. et al. The Process ofFinancial Liberalization and Development Strategies for 207
Mutual Savings and Finance Companies. Korea Development Institute (KDI),
1991.
Lee, N. S. " Strategies and Mechanisms for Mobilizing Rural Savings: Korean
Experiences from the Mutual Credit Scheme of Agricultural Cooperatives." in
Mobilization of Rural Savings in Asia and the Pacific, Report of an APO
Symposium 27th Nov. - 4th Dec., 1990, Tokyo, Japan. Tokyo: Asian Productivity
Organization (APO), 1992.
LeVay, C. " Agricultural Co-operative Theory: A Review." Journal o f Agricultural
Economics 34(1983): 1-44. The Agricultural Economics Society, The University
College of Wales. 1983.
Liebenstein, Harvey. "Allocative Efficiency vs. 'X-Efficiency'." The American Economic
Rev/ew56(1966):392-415.
Longbrake, William A., and John A. Haslem. "Productive Efficiency in Commercial
Banking: The Effects of Size and Legal Form of Organization on the Cost of
Producing Demand Deposit Services." Journal of Money, Credit, and Banking
7(1975):317-30.
McFadden, D. " Cost, Revenue, and Profit Funcitons." in Production Economics: A Dual
Approach to Theory and Applications, edited by M. Fuss and D. McFadden,
Amsterdam: North Holland, 1978:3-109.
McKinnon, R.l. Money and Capital in Economic Development. Washington D.C.:
Brookings Institution, 1973.
. Financial Liberalization and Economic Development: A Reassesment o f Interest
Rate Policies in Asia and Latin America. San Francisco, California: International
Center for Economic Growth, 1988.
McKinnon, J.G. "Model Specification Tests and Artificial Regressions." Journal o f 208
Economic Literature 30(1992): 102-46.
Meeusen, Wim and Julien van den Broeck. "Efficiency Estimation from Cobb-Douglas
Production Functions with Composed Error." International Economic Review
8(1977):435-44.
Mester, Loretta J. "A Multiproduct Cost Study of Savings and Loans." Journal o f
Finance 42(1987):423-45.
Meyer, Richard L. "The Viability of Rural Financial Intermediation." paper prepared for
The Fourth Technical Consultation on the Scheme for Agricultural Credit
Development (SACRED), FAO, Rome, Jan. 21 -23, 1988.
______. "Farm-Household Heterogeneity and Rural Financial Markets:
Insights from Thailand." in Undermining Rural Development with Cheap Credit
edited by Dale W Adams, Douglas H. Graham, and J.D. Von Pischke, pp 22-35,
Boulder, Colorado: Westview Press, 1984.
Mullineaux, Donald J. "Economies of Scale and Organizational Efficiency in Banking: A
Profit Function Approach." The Journal o f Finance 33{197S):259-2S0.
Murray, John D., and Robert W. White. "Economies of Scale and Economies of Scope in
Multiproduct Financial Institutions: A Study of British Columbia Credit Unions."
Journal o f Finance 38(1983):887-902.
______and ______(1980). "Economies of Scale and Deposit-Taking Financial
Institutions in Canada: A Study of British Columbia Credit Unions," Journal o f
Money, Credit and Banking, 12(1980):58-70.
NACF, Theory and Practice o f Agricultural Finance. Seoul: NACF, 1989.
. Current Issues and Development Strategies of Mutual Credits. Seoul: NACF,
1992.
Noulas, Athanasios G., Subhash C. Ray, and Stephen M. Miller. "Returns to Scale and 209
Input Substitution for Large U.S. Banks." Journal of Money, Credit, and
Banking 22{\99Qi):9A-m.
Olson, J.A., P. Schmidt, and D.M. Waldman. " A Monte Carlo Study of Estimators of
Stochastic Frontier Production Funcitons." Journal o f Econometrics 13(1980);
67-82.
Park, J. Y. "The Changes in Financial Market Environment and the Development of
Credit and Banking Business of the Agricultural Cooperatives." paper presented
at the celebration symposium for the 30th anniversary of the agricultural
cooperatives 'The Development of Credit and Banking Business of the
Agricultural Cooperatives toward 2000s.' Seoul: NACF, 1991.
Pulley, L.B. and Y.M. Braunstein. " A Composite Cost Function for Multiproduct Firms
with an Application to Economies of Scope in Banking." The Review o f
Economics and Statistics, 74(1992): 221-230.
Rangan, Nanda, Hassan Y. Aly, Carl Pasurka, and Richard Grabowski. "The Technical
Efficiency of U.S. Banks." Economics Letters 28(1988): 169-75.
Reifshneider, David and Rodney Stevenson. "Systematic Departures from the Frontier: A
Framework for the Analysis of Firm Inefficiency." International Economic
Review 32(1991):715-23.
Rho, W.S. " The Analysis of Kyes in Farm Households." Monthly Report o f Agricultural
Cooperatives, NACF, August, 1988: 227-36.
Rhu, S.K. et al. The Changes in Financial Market Environment and Development
Strategies for Credit Unions. Korea Development Institute (KDI), 1992.
Roller, L.H. "Proper Quadratic Functions with an Application to the Bell System." The
Review o f Economics and Statistics, 72(1990): 202-10.
Santomero, A.M. " Modeling the Banking Firm." Journal o f Money, Credit, and Banking 210
16(1984): 576-602.
Schmidt, P. and C.A. Knox Lovell. " Estimating Technical and Allocative Inefficiency
Relative to Stochastic Production and Cost Frontiers." Journal o f Econometrics
9(1979): 343-66.
Schmidt, P. and T.F. Lin. " Simple Tests of Alternative Specifications in Stochastic
Frontier Models." Journal o f Econometrics 24(1984): 349-61.
Schweiger, I. and J.S. McGee. "Chicago Banking: The Structure and Performance of
Banks and Related Financial Institutions in Chicago and Other Areas." Journal o f
Business M{}\x\y 1966):34-41.
Sealey, C.W., and J.T. Lindley. "Inputs, Outputs, and the Theory of Production and
Costs at the Depository Financial Institutions." Journal o f Finance
32(1977):125I-65.
Seiford, Lawrence M. and Robert M. Thrall. "Recent Developments in DE A: The
Mathematical Programming Approach to Frontier Analysis." Journal o f
Econometrics, 46 (1990):7-38.
Shumway, R., R. Pope, and E. Nash. "Allocable Fixed Inputs and Jointness in
Agricultural Production: Implication for Economic Modeling." American Journal
of Agricultural Economics 66(1984):72-8.
______, ______, and ______. "Allocable Fixed Inputs and Jointness in
Agricultural Production: Implication for Economic Modeling: Reply." American
Journal o f Agricultural Economics 70(1988):950-2.
Smith, D.J. "A Theoretic Framework for the Analysis of Credit Union Decision-
Making." Journal o f Finance 39( 1984): 1155-68.
. " A Test for Variant Objective Functions in Credit Unions." Applied Economics
18(1986): 959-70. 211
Smith, D. J., T. Cargill and R. A. Meyer. "An Economic Theory of a Credit Union."
Journal o f Finance 36(1981): 519-28.
Srinivasan, Aruna. "A Multiproduct Cost Study of Rural Bank Branches in Bangladesh."
Unpublished Ph.D. dissertation, The Ohio State University, 1988.
Stevenson, Rodney E. "Likelihood Functions for Generalized Stochastic Frontier
Estimation." yowr/ia/ o f Econometrics 13(1980):58-66.
Stigler, Gorge J. "The Xistence of X-Efficiency." The American Economic Review
66(1976):213-6.
Suh, J.H. " Rural Savings Mobilization in Korea: Issues, Policies and Suggested
Measures.” Journal o f Rural Development 14(1991): 1-13.
Suh, J.H. and S. J. Park. A Study on the Structure o f Rural Formal Financial Markets.
Seoul: Korea Rural Economics Institute, 1986.
Sul, K. E. "The Structural Adjustment of Agriculture and Agricultural Financial
Development Post Uruguay Round." Paper presented at the celebration
symposium for the 30th anniversary of the agricultural cooperatives 'The
Development of Credit and Banking Business of the Agricultural Cooperatives
toward 2000s.' Seoul: NACF, 1991.
Sul, K.E. and C.N. Kee. A Study to Improve the Supply System o f Agricultural Credit.
Seoul: KREI, 1986.
Taylor, R. " The Credit Union as a Cooperative Institution." Review o f Social Economy
29(1971): 207-17.
Varian, Hal R. Microeconomic Analysis. New York: W.W. Norton & Company, 1978.
Von Pischke, J.D., D.W Adams, and Gordon Donald (ed.). Rural Financial Markets in
Developing Countries; Their Use and Abuse. Baltimore: The Johns Hopkins
University Press, 1983. 212
Waldman, D. " A Stationary Point for the Stochastic Frontier Likelihood." Journal o f
Econometrics 18(1982): 275-79.
Wolken, J. and F. Navratil. " Economies of Scale in Credit Unions: Further Evidence."
The Journal o f Finance 35(1980): 769-77.
Yun, K. H. Agricultural Cooperative in Korea. Seoul: NACF, 1987.
Zieschang, Kimberly D. "A Note on the Decomposition of Cost Efficiency into
Technical and Allocative Components." Journal o f Econometrics 23(1983): 401-
5.