Evidence from Firm-level and Industry

Has financial liberalization improvedeconomic efficiency in Korea?
: Evidence from firm-level and industry-level data
Jungsoo Park*
Yung Chul Park†
Sogang University
Korea University
Abstract
This study uses panel regression on firm-level and industry-level data in Korea from 1991
– 2007 to see the impact of financial liberalization on bank and non-bank financial
institution(NBFI)’s lending, before and after the 1997 crisis. Financial liberalization index is
developed incorporating multi-faceted nature of the liberalization process. The main findings
are as follow. First, firm-level analyses reveal that the financial liberalization had influenced
the allocation of the bank and NBFI’s loans for relatively smaller firms but not for the large
listed firms. As for the smaller firms, the loans are allocated to firms with better performance
history. Secondly, the financial liberalization had influenced the allocation of loans
eventually leading to the improvement of the TFP growth of the smaller firms but not for the
large firms. Thirdly, the heavier reliance on the direct financing during the financial
liberalization process does not seem to have influenced the firm’s efficiency. Finally, the
efficiency-enhancing effect of financial liberalization through bank lending channel cannot be
observed at the industry-level.
Keywords: Financial liberalization, economic efficiency, banking, external finance
JEL code: G20, O40
*
School of Economics, Sogang University, Seoul, Korea. Tel: (02) 705-8697, E-mail: [email protected] .
Division of International Studies,Korea University, Seoul, Korea. E-mail: [email protected] .
†
1. Introduction
Korea’s financial sector had been characterized as a repressive regime where financial
markets and instructions were regulatedthroughoutmuch of the post war period beforeits
government initiated a wide range of reform for financial market liberalization and openingin
the early 1980s. Most of the reform initiatives, however,ended up being superficial or
cosmetic as they were compromised to placate domestic opposition. It was not until the early
1990s when substantial reform for financial restructuring got underway.
Market interest rates went through several phases of deregulation to be fully liberalized
as part of the reform dictated by the IMF for the resolutionof the 1997 financial crisis.Since
then, most of the regulatory restrictions on asset and liability management at banks and other
non-bank financial intermediaries (NBFIs) and on capital account transactions have been
phased out to create a market oriented and open financial regime.
Theories predict that transition from a repressive to a deregulated regime would bring
about improvement in efficiency of resource allocation of the economy as it expands the
capacity of intermediation and screening of creditworthy and productive borrowers of
financial markets and institutions. Indeed,it is almost a matter of faith that in emerging
economies such a market oriented financial system would be more efficient in allocating
capital than a state controlled regime.
In his survey ofthe literature, Levine (2005)argues that finance has a positive causal
effect on growth as it improves the allocation of capital by easing external financing
constraints on firms and that countries with more efficient financial systems grow faster. A
large number of empirical studies on the finance-growth nexusconfirm Levine’s view,
although it should be noted that there are also pieces of evidence from the experiences of
emerging economiessuggesting that the positive effect is not universal.
2
This study attempts to throw light on whether the Korea’s experience with financial
liberalization bears out the positive effect of financial growth. To this end, this paper
conducts empirical analyses of causal linkages and their significance between market
oriented financial reform on the one hand and efficiency and economic growth on the other in
terms of the firm and industry level panel data before and after the 1997 crisis. More
specifically, this paper investigates whether financial liberalization has led to an increase in
total factor productivity of firms and industries through improvement of efficiency of the
allocation of loanable funds at banks and non-bank financial intermediaries(NBFI).
For this purpose, section 2 briefly surveys the literature on the effects of financial
liberalization on financial development, efficiency improvements, and economic growth.
Empirical studies attempting to analyze the finance-growth nexus needs to begin with the
construction and measurement of the degree of market orientation of the financial system. In
section 3 this study develops and estimatesan indexof financial liberalization using the data
reflecting changes in the behavior of financial markets and institutions attributable to the
market oriented reform. Section 4 develops three models for empirical estimation at the firm
and industry level. The estimation results are presented in Section 5. Section 6 concludes.
2. A Brief Survey of the Literature
Earlier studies on financial development and economic growth such as the ones by
Gurley andShaw (1955)3, Shaw (1973) and McKinnon (1973)identify potential channels
through which financial liberalization would bring about diversification and improvements in
efficiency of the financial sector with the attendant positive effect on economicgrowth.
3
Financial Aspects of Economic Development
John G. Gurley and E. S. Shaw
The American Economic Review
Vol. 45, No. 4 (Sep., 1955), pp. 515-538
3
According to their thesis, government controls on market interest rates and asset and liability
management at banks and NBFIs, which constitute financial repression,would cause
stagnation infinancial development and ultimately in economic growth.Greenwood and
Jovanovic(1990) and (Bencivenga and Smith, 1991) concur by showing that financial
intermediation in a liberal financial regimeenhances efficiency of the economy and growth
through a better information processing and investment screening.
King and Levine (1993), conducting a cross-country regression analysis, and
Levine,Loayza, and Beck (2000), performing country-level panel regression estimation, show
that financial developmentspawns positive effects on economic growth. Beck, Levine, and
Loayza (2000) find that expansion and efficiency improvements of financial intermediation
producepositive effects on total factor productivity growth,although they have little long-term
effects on investment and the savings rate. Favara (2003) disputes the findings of Beck,
Levine, and Loayza (2000) in a study using cross-country panel data that financial
development does not necessarily promote economic growth.
Rajan and Zingales (1998) present evidencethat industries differ on the degree of financial
dependency on external financing, suggesting that the lack of an adequate provision of capital
through financial markets would imply that financially dependent industries would be less
able to take advantage of the potential growth opportunities presented by financial growth
and development than those industries requiring less external financing. They argue that,
given the differences in external finance dependency, capital market development would
result in differential rates of growth in different industries.
However, Fisman and Love (2003) show that growth opportunities rather than the
technological external finance dependency explain better growth of different industries using
the dataset of Rajan and Zingales (1998). Using the firm level data from 30 countries,
Demirguc-Kunt and Maksimovic (1998) find that a wider access to external finance tends to
4
encourage long-run growth of firms. Using panel data on a large number of firms from 13
developing countries Laeven (2003) examines whether financial liberalization relaxes
financial constraints on firms. He finds that small firms are more financially constrained than
large firms before the start of financial liberalization, but afterwards financial constraints on
small firms are low whereas they are higher for large firms. This is because large firms have
access to preferential directed credit during financial repression.4.
As for the studies on the Korean experience, Kim (2003) considers five facets of financial
liberalization –changes in the regulations and government policy on interest rates, the foreign
exchange rate, the legal reserve requirement, capital account liberalization, and bank
privatization – to analyze the effect of financial liberalization on economic growth in Korea.
He uses a principal component analysis to create a single index representing the five aspects
of financial liberalization. Kim concludes that financial liberalization in Korea has had a
significantly positive impact on economic growth. Shin and Oh (2005) use 30 to 34 Korean
industry-level panel data from 1981 to 2001 to investigate the extent to which external
finance dependency (Rajan and Zingales, 1998) and growth opportunities (Fisman and Love,
2003) have contributed to the growthof Korean industries respectively. Their results imply
that industrial development in Korea was influenced by both the external finance dependency
and growth opportunities in the 1980s, but the growth opportunities were the main factor
contributing to industrial growth in the 1990s.
Ahnet. al (2008) use Korean firm-level data from 1991 to 2003 to see if there was a
differential effect of external financing on capital accumulation, R&D, and TFP growth
before and after the financial crisis of 1997. Separate regressions on the subsamples of 1991 –
4
,Does Financial Liberalization Reduce Financing Constraints?
Luc Laeven
Financial Management
Vol. 32, No. 1 (Spring, 2003), pp. 5-34
5
1996 and of 1999 – 2003 show that the availability of external finance is associated with a
faster capital accumulation but less so after the crisis. In contrast, the external finance effect
on total factor productivity growth was found to be relatively weak both before and after the
crisis.
3. Financial Liberalization Index
Empirical examinations of the effects of financial liberalization on the economy need to
begin with identification and estimation of variablesgauging quantitative changes in the
degree of deregulation and opening of financial markets. Since there is no generally accepted
measure of financial liberalization,most studies develop graded indices over time in terms of
a multi-faceted measure of financial liberalization that covers a number of aspects of
financial reform (Abiadet al.2008). Following a similar approach, this study usesprincipal
component analysis to constructs an index based on several measures reflecting changes in
the financial system brought about by the reform for financialliberalization for the 1990-2007
period.
This study usesthe following sixmeasures in constructing a financial liberalization
index:
i) The first four are the measuresestimating changes in entry barriers, banking supervision,
liberalization of capital flows, and deregulation of security markets in the process of financial
liberalization. They are formulated in index and drawn from theIMF financial reform
database.5(What are the numbers on the vertical axis-index?)
5
See Abiadet al (2008). Originally, IMF financial reform index was created by summing the following 7
indices: creditcontrols(credit controls),intratecontrols(interest rate controls),entrybarriers(entry barriers),
bankingsuperv(banking supervision),privatization (privatization),intlcapital(international capital flows), and
securitymarkets(security markets). However, intratecontrolsindex exhibits an unrealistic downward movement
after 2000. The index privatization seems to capture the flow of privatization as the index has the value of 2 for
pre-1997 period, while 0 for the post-1997. Since the latter two indices do not seem to be suitable for the degree
6
ii) The fifth one is a measure of the volatility of the nominal exchange rate. Since capital
account liberalizationand the adoption of a more flexible exchange rate system would lead to
larger swings in the exchange rate than before, it captures the progress in financial market
opening. Thismeasure is constructed first by estimatinga series ofannual standard deviations
of the rate of changein the monthly nominal effective exchange rate obtained from the BIS
for the period from 1990 to 2007 (excluding the currency crisis period of 1997 and
1998).This annual series of standard deviation is further normalized by subtracting its sample
mean, and then dividing by its sample standard deviation.
iii) The sixth one is a measure of domestic financial deregulation covering the three phases of
the interest rate decontrol undertaken by the Korean government. Throughout the 1980s,
deposit and lending rates in the banking sector were strictly under the control of the
government. The relaxation of the control over these interest rates went through three
different phases of liberalization in the 1990s. As a result of the deregulation, there was a
substantial increase in the movements of interest rates on bank deposit and lending
throughout the 1990s, which may indicate that banks became more sensitive to changes in
thedemand for their loans (see Figure 2-1). Thismeasurestarts at 1 for the 1991-92 period,
then increases incrementally in steps of 0.5 for 1993-94, 1995-96, 1997-04, and post-2005
periods. It is reasonable to assume that full liberalization of interest rates was completed
during the post-2005 period where the raw measure rose to the highest level 3.
Using principal component analysis, these six measures are then converted intoa single
index representing the six aspects of financial liberalization. Details of the construction are
provided in appendix.Figure 3-1 illustrates the six aspects of financial liberalization and the
index from the principal component analysis.
of financial liberalization concept, they were excluded from consideration. Furthermore, creditcontrolsis not
considered either since it does not vary after 1990.
7
.6
.4
2
0
.2
2.2
2.4
entrybarriers
2.6
bankingsuperv
.8
2.8
3
1
Figure 3-1. Six measures and financial liberalization index
1990
1995
2000
year
2005
1990
2010
2005
2010
3
2.5
1
1.5
2
securitymarkets
2.5
1
1.5
2
intlcapital
2000
year
Banking supervision
3
Entry barrier
1995
1990
1995
2000
year
2005
2010
1990
2000
year
2005
2010
0
-1
1
0
finlib2e
intrate_reform
2
1
2
Security markets
3
International capital flow
1995
1990
1995
2000
year
2005
2010
Interest reform index
1990
1995
2000
year
2005
Exchange rate volatility index
8
2010
2
0
-2
-4
Scores for component 1
1990
1995
2000
year
2005
2010
Financial liberalization index (derived from principal component analysis)
Note) Entry barrier, banking supervision, international capital flow, security markets,
interest reform index are all measured on the scale of 0 to 3, 3 being the highest
degree of liberalization. Exchange rate volatility index is normalized using the
sample mean and sample standard deviation as described in the Section 3.
Financial liberalization index is derived based on principal component analysis.
Throughout the sample period, firms-in particular large ones-continued to rely less on
traditional bank financing in favor of raising funds on the capital market. The ratio of direct
financing (bond and equity) to total financing is used to examine the effect of the relative
increase in capital market financing on firms’ efficiency. Changes in the ratio are shown in
Figure A1 of Appendix.
4. Empirical Implementation and Data Description
4.1 Main questions and empirical models
This study constructs three linear regression models for firm-level and industry-level
panel data from 1991 – 2007 to assess the impact of financial liberalization on the lending
behavior of banks and NBFI’s before and after the 1997 crisis.
More specifically, this study poses the following three questions:
1) Have banks and NBFIs rationalized theirlending operations to allocate their loanable
funds more efficiently, taking into accountfirm’s and industry’s fundamental
characteristicssuch as the ROA or value-added history more than before? Put it
9
differently, in the process of financial liberalization have the individual firms and
industries with a better ROA or value added record become favored borrowers at
financial intermediaries
2) Havethe firms and industries with a greater access to indirect bank financing (or with
easing of external financing constraints) done better in improving their TFPs?
3) Has the growing reliance on capital market financing by firms had any real impact on
their productivity growth? Here, the hypothesis is that capital markets, as opposed to
banks and NBFIs, are more efficient in screening out more viable and
potentiallysuccessful firms.
In order to answer the three main questionsboth at the firm and industry level, this
study uses the firm-specific (industry- specific) fixed-effects panel regression estimation.
•
Model 1
The first model, which is designed to address firm-specific (industry-specific) question
on the rationalization of bank lending is specified as follows:
𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠)𝑖𝑡 = 𝛼0 + 𝛼1 ∗ 𝑋𝑖,𝑡−1 + 𝛼2 ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡 + 𝛼3 ∗ 𝑋𝑖,𝑡−1 ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡
+𝛾′𝑍𝑖𝑡 + 𝑢𝑖 + 𝑣t + 𝜀𝑖𝑡
(1)
The dependent variable–𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠𝑖,𝑡 ) −is the rate of growth of total loans extended
by both banks and NBFIs.The variable𝑓𝑖𝑛𝑙𝑖𝑏𝑡 is the financial liberalization index at time
t.𝑋𝑖𝑡−1 is the rate of return on asset (ROA) of firm iin year t-1(or the growth of value-added of
industry i ) and represents firm (industry)i’s main performance characteristic of the past. In
this specification, the regressor, 𝑋𝑖𝑡−1 , represents an independent effect of the past
10
performance on loan acquisition.The interaction term (𝑋𝑖𝑡−1 ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡 ) capturesthe extent to
which financial liberalization has changedthe effect of the past performanceon acquiring
loans.If this latter term is significant and positive, then it would suggest that financial
liberalization prevailed on banks and NBFIs to consider firm’s (industry’s) past performance
more
than
before
in
allocatingtheirloans.
Z
represents
a
vector
of
control
variables.Variables𝑢𝑖 and 𝑣𝑡 stand for firm-fixed (industry-fixed) effects and year-dummies.
•
Model 2
The model for analyzing the second question-the contribution of loans by banks and
NBFIs to firm’s (industry’s) TFP growth- takes the following form:
𝑑𝑙𝑛(𝑇𝐹𝑃)𝑖𝑡 = 𝛽0 + 𝛽1 ∗ 𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠𝑖,𝑡−1 ) + 𝛽2 ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡 + 𝛽3 ∗ 𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠𝑖,𝑡−1 ) ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡
+𝛿′𝑍𝑖𝑡 + 𝑢𝑖 + 𝑣t + 𝜀𝑖𝑡
(2)
The dependent variable is firm’s (industry’s) TFP growth (𝑑𝑙𝑛(𝑇𝐹𝑃)𝑖𝑡 ).The regressor,
𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠𝑖𝑡−1 ), captures an independent effect of the past loan growth on TFP growth. The
interaction term (𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠𝑖,𝑡−1 ) ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡 )measuresthe role of financial liberalization in
influencingcontribution of loans to firm’s (industry’s) productivity. If this term is significant
and positive, then it would suggest that financial liberalization has been instrumental to
effecting a larger contribution of loans of banks and NBFIs to TFP growth of firms
(industries).In other words,the interaction term could be understood as measuring the
significance of the effect of the past loan growth on TFP growth.
•
Model 3
11
The third modelfor analyzing the effect of change in the structure of financing on
firm’s efficiency is designed as follows:
𝑑𝑙𝑛(𝑇𝐹𝑃)𝑖𝑡 = ρ0 + 𝜌1 ∗ 𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠𝑖,𝑡−1 ) + 𝜌2 ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡 + 𝜌3 ∗ 𝑑𝑙𝑛(𝑙𝑜𝑎𝑛𝑠𝑖,𝑡−1 ) ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡
+ρ4 ∗ 𝑠ℎ𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑡−1 + ρ5 ∗ 𝑠ℎ𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑡−1 ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡 + 𝜑′𝑍𝑖𝑡 + 𝑢𝑖 + 𝑣t + 𝜀𝑖𝑡 (3)
whereshdirect is the share of direct financing in total financing ofan individual firm.In this
specification,𝑠ℎ𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑡−1 captures the effect of the structural change in firm’sfinancing
method on its productivity (TFP) growth, while 𝑠ℎ𝑑𝑖𝑟𝑒𝑐𝑡𝑖𝑡−1 ∗ 𝑓𝑖𝑛𝑙𝑖𝑏𝑡 measureshow much
financial liberalization has changed the effectiveness of the direct financing on improving
firm’s productivity. Industry-level analysis for this model is not conducted due to the lack of
relevant direct financing share data.
Inaddition to the full sample regressions of the three models presented above,the
sample of firm-level data is divided into two sub-groups:Korea Exchange (KRX)-listedfirms
and the remainingother small-and medium-sized(SME) ones.6The sub-sample regressions are
conducted to examine whether small and medium sized firms have gained a greater access to
banks and BBFIs for indirect financing than larger ones in the process of financial integration.
4.2Data description for firm-level data
This study uses 33,162 annual observations of 6,233 firms in the KIS-VALUE database
for the 1990 to 2005 period.KIS-VALUE database compiled by NICE Information Service
provides financial and non-financial data for the listed and external-audited firms. Financial
6
The KRX-listed firms are relatively more established and large due to demanding requirements for the
listing.The remaining firm group includes KOSDAQ-listed firms, external audited firms, and firms under
government supervision.
12
companies and the financial crisis period from 1997 to 1999 are excluded from the sample.
The bank financing variable- loans- is the sum of short-term and long-term loans from
banks and non-bank financial intermediaries (NBFI).7ROA (roa) is the return on asset for
each firm. Firm-level TFP growth rates –dlnTFP– are obtained from Baeket. al. (2009) who
calculated TFP growth rates using the approach suggested by Bailey, Hulten and Campbell
(1992). The index of change in the funding structure –shdirect- is the share of bond and
equity financing in total financing of each firm.
As the model 1 is a model to identify determinants of bank and NBFI financing, it
includes control variables such as the ratio of the R&D expenditure to gross sales (rnd_sales)
to account for innovative activity of the firm, the log of total assets (lnassets) to proxy the
size of the firm, the ratio of debt to equity (liab_eq) to measure the degree of capital
adequacy, and a post-1997 crisis dummy(dum_after97),which takes a value of one after
1997.8All these variables ( except for the dummy ) enter lagged in regression. As for
themodels 2 and 3, which are TFP growth determinant model, the set of control variables
include rnd_sales, lnassets, and dum_after97. The monthly exchange rates and financial
reform indicesare collected from BIS andAbiad et al (2008), respectively
Summary statistics and definition of variables are provided in the Table A1 of
theAppendix.
4.3Data description for industry-level data
7
Short-term loans are the sum of bank overdraft, other short-term borrowings, short-term borrowings in foreign
currency, current portion of long-term borrowings, and current portion of long-term borrowings in foreign
currency. Long-term loans are the sum of long-term borrowings and long-term borrowings in foreign currency.
8
Extreme value observations with TFP growth rate over 0.5 or less than -0.5, percentage change in loans in
excess of 100% or less than -100%, external financing to sales ratio greater than 1, or R&D ratio greater than 1
are all excluded from the sample. As for the regressions regarding the structural change in financing, the
observations with percentage changes in bond and equity in excess of 200% or less than -200% are additionally
excluded.
13
The data are obtained from the annual industry-level panel data set at the two-digit
Korean industries from 1990 to 2007. The financial crisis period of 1997 – 1999 is excluded
from the sample as before.
The bank financing variable–loansi–is the sum of total short-term and long-term loans
to each industry from banks and non-bank financial intermediaries (NBFI). The data are from
Statistical Yearbook, BOK.Industry-level value-added (vai) and TFP growth rates were
collected from Korea Productivity Center which provides Korean input and output data for
the EUKLEMS database. 9 TFP growths were calculated based on gross output. R&D
intensity –rnd_vai–defined as the ratio of current R&D expenditure to current value-added is
the only control variable in all models.
R&D expenditure data are from the KISTEP
database. The sample comprises 218 annual observations of 22 different manufacturing and
non-manufacturing industries for the period of 1993 to 2007.10
Summary statistics and definition of variables are provided in the Table A2 of the
Appendix.
5. Empirical Analysis
5.1 Results fromfirm-level analysis
•
Rationalization of Bank Lending Operations
Table 5-1presents the results of the estimation of equation (1) for individual firms.
Columns (1) ~ (3) include a post-1997 crisis dummy to weigh up the main effect of financial
liberalization. Column (4) includes year dummies to control unobserved year-specific
9
In principle, EUKLEMS database lists data for 72 different industries for each country. However, not all data
are available for each country.
10
There are 19 manufacturing industries, electricity, gas & water industry, construction industry, and transport,
storage & communication industry. Extreme value observations with TFP growth rate over 0.2 or less than -0.2,
percentage change in loans in excess of 50% or less than -50%, or R&D ratio greater than 0.2 are all excluded
from the sample.
14
shocks.The coefficient estimatesforroa(lagged ROA) are significant and positive in all
columns (1) ~ (4), suggesting that the ROA history was an important factor taken into
account in screening borrowers at banks and NBFIs during the sample period. The coefficient
estimates for finlibare significant and positive in columns (2) ~ (3). This finding implies that
financial liberalization was also accompanied by an expansion of the intermediation capacity
of banks and NBFIs.
The coefficient signsof the interaction terms between firm’s ROA and the
liberalization indexarepositive and significant as shown in columns (3) and (4). The inclusion
of the interaction terms does not change either the coefficient estimates or significance of
other variables. This finding provides a piece of evidence for a positive impact of financial
liberalization on asset management at banks and NBFIsas it induces them totakes into
consideration more than before firm’s ROA history in allocating their loans.
Tables5-2 and 5-3 present the sub-sample regression results forKRX-listed subsampleand for the remaining other small- and medium-sized (SME)firm sub-sample,
respectively.Columns (3) and (4) in Table 5-2 and Table 5-3 respectively show that the
interaction term is significant for the SME group, whereas it is insignificant for the KRXlisted firm sample. This finding suggests that financial liberalizationallowed a larger room
and generated incentives for accommodating the loan demand by SMEs at banks and NBFIs,
while it did not affect to any considerable degree the access of large and more established
firms to indirect financing. The overall positive effect of the interaction term in the full
sample results presented in Table 5-1 is due to the fact that the SME sample dominates KRXlisted firm sample in terms of number of firms.11 Although these results indicate that the
financial liberalization have had positive impact on the rationalization of the bank lending to
11
The number of KRX-listed firms account for 22% in 1995, 12% in 2000, and 8% in 2005 of the total number
of firms in the database.
15
a typical firm in Korea, the overall impact on the whole economy cannot be assessed
conclusively since each firm’s relative size and importance to the economy differs in great
magnitude.12The ensuing industry-level analysesmay provide clearer picture since these firms
are aggregated into relevant industries.
•
Bank Loans and Productivity
Table 5-4displays regression resultsfor equation (2).The dependent variable is TFP
growth (dlnTFP).As in equation (1), columns (1) ~ (3) include a post-1997 crisis dummy to
isolatethe effect of financial liberalization and columns (4) ~ (6) include year dummies.
Estimation of equation 2 is expected to reveal whether loans extended by banks and NBFIs (a
measure of access to indirect financing)made any contribution to improving firms’ TFP
growth with financial liberalization going forward during the sample period.
The coefficient estimatesof the interaction terms between lagged loan growth and
financial liberalization index (dlnloans*finlib)are all positive and significant in columns (3) ~
(6). Again, the inclusion of this interactive term has little effect on either the
coefficientestimates or signs of other regressors. This empirical findingrendersvalidity tothe
chain of events observed during the sample period in which financial liberalization first leads
to improvement of efficiency of lending operations of financial intermediaries, which in
turnboosts firms’subsequent TFP growth.
Tables5-5 and 5-6 present the sub-sample regression results for the KRX listed and
SME groups. As in model 1, the results suggest that the firms belonging to the SME group
benefitted more from financial deregulation. Columns (3) ~ (6) in Table 5-5 and Table 5-6
12
The KRX-listed firms’ sales account for 71% in 1995, 72% in 2000, and 60% in 2005 of the total sales of the
full sample. The median sizes of the SME firms are 21% in 1995, 11% in 2000, and 12% in 2005 of the median
size of the KRX-listed firm sample. The distribution of firms in terms of sales are provided in Figure A2 of
Appendix.
16
display the interaction terms(dlnloans*finlib) for the SME sample are significant, but not for
the KRX listed group. This finding, which is consistent with that of Table 5-4, suggests
thatSMEs were able to enhance theirTFP growthby gaining a greater access to indirect
financing in the process of financial liberalization. Again, these results indicate that the
financial liberalization played positive role in enhancing TFP growth of a typical borrowing
firm in Korea. However, the overall impact on the whole economy cannot be confirmed since
each firm’s relative size and importance to the economy differs as discussed earlier.
•
Capital Market Financing and TFP Growth
Estimation of equation (3) of which results are presented in columns (5) and (6) in
Table 5-4show that change in the financing structure (shdirect, a lagged share of direct
financing) and the interaction terms between the lagged loan growth and the financing
structure (shdirect*finlib) fail to show any significant effect on TFP growth (dlnTFP ). The
sub-sample results in columns (5) and (6) of Tables 5-5 and 5-6 present qualitatively the
same results. Thus, the relative increase in capital market financing had little impact on firm’s
TFP growth andthat this insignificance was not related to financial liberalization during the
sample period, implying that capital market financing is a good substitute for bank financing
for larger firms in Korea.
5.1 Results from industry-level analysis
In this section, the firm-level analysis is extended to examine the effects of financial
liberalization at the industry level using the same models except for a couple of modifications.
First, in Model 1 for industry level analyses, value-added (VA) is chosen in place of the ROA
as the potential factor influencing the lending behavior of banksand NBFIs. This choice is
17
made because of the unavailability of industry data for the ROA.Second,the industry-level
analyses use a different set of control variables. Since it is very difficult to find variables
identifying industry-specific characteristics, estimation of three models include only a limited
set of control variables. All estimates are obtained from industry-fixed effects panel
regressions.
•
Rationalization of Bank Lending Operations
The results of estimation of equation (1) at the industry level in Table 5-7 show that
dlnvai(lagged
value-added
growth)
andthe
liberalization
index
are
statistically
insignificant.The coefficient signsof the interaction term between industry’s value-added
growth and financial liberalization index-dlnvai*finlib-are alsoinsignificant as presentedin
columns (3) and (4).13
Overall, the results suggest that unlike in the firm level analysis, financial
liberalization had little bearing on the lending behavior of banks and NBFIs at the industry
level.These results contradict those of the firm level analysis. One possible explanation for
this disparity is that the magnitude of the positive effect of financial liberalization on the
SMEgroup may not have beenlarge enoughat the industry level to produce the results similar
to those of the firm-level analysis.
•
Bank Loans and Productivity
Table 5-8displays regression results of equation (2).The dependent variable is TFP growth
(dlnTFP).As before, columns (1) ~ (3) insertsa post-1997 crisis dummy to isolate the effect of
financial liberalization. Columns (4) ~ (6) include year dummies. The coefficient
13
The results are qualitatively the same for the manufacturing only sub-sample.
18
estimatesofthe financial liberalization index (dlnloans*finlib) andthe interaction terms
between lagged loan growth and arenot significantas shown in columns (3) ~ (6).
Although the firm-level analyses in the earlier subsection show that financial
liberalization had a significant impact only on the group of firms not listed on the KRX, the
relative importance of this impact at the industry level may not have been large enough to
produce the results similar to those of the firm level analysis. These results may be
interpreted as implying that financial liberalization has not improved efficiency of lending by
banks and NBFIs to foster industry’s subsequent TFP growth.
5. Concluding remarks
19
[References]
Abiad, Abdul, EnricaDetragiache, and Thierry Tressel, "A New Database of Financial
Reforms," IMF Working Paper WP/08/266, December 2008
Ahn, Sanghoon, Joon-Ho Hahm, and Joon-Kyung Kim, “External Finance and Productivity
Growth in Korea:Firm Level Evidence Before and After the Financial Crisis”KDI
Journal of Economic Policy, Vol. 30, No.2, 2008, pp.27~59.
Baek,C, Y. Kim, and H. U. Kwon, “Market Competition and Productivity after the Asian
Financial Crisis: Evidence from Korean Firm Level Data,” CEI Working Paper Series
No.2009-12, 2009.
Baily, Martin Neil ,Charles Hulten and David Campbell,“Productivity Dynamics in
Manufacturing Plants,”Brookings Papers on Economic Activity. MicroeconomicsVol.
1992, 1992, pp. 187-267
Beck, T., R. Levine, and N. Loayza, “Finance and the Sources of Growth,” Journal of
Financial Economics, Vol. 58, No. 1, 2000, pp.261~300.
Bencivenga, V. R. and B. D. Smith, “Financial Intermediation and Endogenous Growth,”
Review of Economic Studies, Vol. 58, No. 2, 1991, pp.195~209.
Demirguc-Kunt, A. and V. Maksimovic, “Law, Finance and Firm Growth,” Journal of
Finance, Vol. 53, no. 6, 1998, pp.2107~2137.
Favara, G., “An Empirical Reassessment of the Relationship between Finance and Growth,”
IMF Working Paper, No. 03/123, 2003.
Fisman, R. and I. Love, “ Financial Dependence and Growth Revisited,” NBER Working
Paper No. 9582, 2003.
Greenwood, J. and B. Jovanovic, “Financial Development, Growth and the Distribution of
Income,” Journal of Political Economy, Vol. 98, No. 5, 1990, pp.1076~1107.
Kim, Han-Ah, “The Relationship between Economic Growth and Financial Liberalization
and Development,” KDIC Financial Review, Vol. 4, No. 2, 2003, pp.5~30. (in Korean)
King, R. G. and R. Levine, “Finance and Growth: Schumpeter might be right,” Quarterly
Journal of Economics, Vol. 108, No. 3, 1993, pp.717~737.
Levine, R., N. Loayza, and T. Beck, “ Financial Intermediation and Growth Causality and
Causes,” Journal of Monetary Economics, Vol. 46, 2000, pp. 31-77
McKinnon, R., Money and Capital in Economic Development, Washington DC: Brookings
Institution, 1973.
20
Rajan, R. G. and L. Zingales, “Financial Development and Growth,” American Economic
Review, Vol. 88, No. 3, 1998, pp.559~586.
Shaw, E. S., Financial Deepening in Economic Development, New York: Oxford University
Press, 1973
Shyn, Yong-Sang and Wankeun Oh, “Financial Dependence, Growth Opportunity, and
External Finance and Productivity Growth in Korea Industrial Growth in
Korea,”KyungjehakYeonku (Journal of Korea Economic Association), Vol. 53, No. 3,
2005, pp.49~76. (in Korean)
21
Tables
Firm-level data regressions
Table 5-1. Determinants of bank and NBFI financing: firm-level regressions for full sample
(dependent variable = dln(loans))
(1)
(2)
(3)
(4)
0.371***
(10.297)
0.378***
(10.489)
0.168***
(6.236)
0.372***
(10.264)
0.167***
(6.210)
0.225*
(1.798)
-0.066
(-0.991)
-0.093***
(-16.395)
0.000
(0.743)
-0.145***
(-10.031)
0.365***
(10.092)
VARIABLES
roa
finlib
roa*finlib
rnd_sales
-0.076
(-1.138)
-0.080***
(-15.066)
0.000
(0.788)
-0.066***
(-10.168)
lnassets
liab_eq
dum_after97
-0.071
(-1.054)
-0.092***
(-16.296)
0.000
(0.718)
-0.147***
(-10.143)
0.272**
(2.171)
-0.070
(-1.048)
-0.089***
(-14.665)
0.000
(0.774)
Firm fixed effects
Yes
Yes
Yes
Yes
Year fixed effects
No
No
No
Yes
Observations
33162
33162
33162
33162
R-squared
0.039
0.040
0.040
0.045
Number of firms
6233
6233
6233
6233
Note) All estimates are firm-fixed effects panel regression estimates. Variables roa, rnd_sales, lnassets,
andliab_eqare all one-year lagged. The financial liberalization variable finlibis an index derived from PCA.The
variableroa*finlibis an interaction term between roaand finlib. t-statisticsare in parentheses. *** p<0.01, **
p<0.05, * p<0.1.
Table 5-2. Determinants of bank and NBFI financing: firm-level regressions for KRX-listed
firm sample (dependent variable = dln(loans))
VARIABLES
roa
(1)
a1
(2)
a2
(3)
a3
(4)
a4
0.211**
(2.032)
0.220**
(2.114)
0.097*
(1.746)
0.209**
(1.961)
0.097*
(1.738)
0.149
(0.473)
0.036
0.185*
(1.744)
finlib
roa*finlib
rnd_sales
0.000
0.039
22
0.169
(0.540)
-0.023
lnassets
liab_eq
dum_after97
(0.000)
-0.048***
(-3.778)
0.000
(0.589)
-0.120***
(-8.108)
(0.084)
-0.058***
(-4.163)
0.000
(0.610)
-0.165***
(-5.552)
(0.079)
-0.059***
(-4.182)
0.000
(0.606)
-0.164***
(-5.488)
(-0.050)
-0.037**
(-2.536)
0.000
(0.743)
Firm fixed effects
Yes
Yes
Yes
Yes
Year fixed effects
No
No
No
Yes
Observations
5187
5187
5187
5187
R-squared
0.059
0.060
0.060
0.077
Number of firms
555
555
555
555
Note) All estimates are firm-fixed effects panel regression estimates. Variables roa, rnd_sales, lnassets,
andliab_eqare all one-year lagged. The financial liberalization variable finlibis an index derived from PCA.The
variableroa*finlibis an interaction term between roaand finlib. t-statisticsare in parentheses. *** p<0.01, **
p<0.05, * p<0.1.
Table 5-3. Determinants of bank and NBFI financing: firm-level regressions forfirms not
listed in KRX(dependent variable = dln(loans))
VARIABLES
roa
(1)
a1
(2)
a2
(3)
a3
(4)
a4
0.391***
(10.166)
0.398***
(10.342)
0.185***
(6.004)
0.393***
(10.173)
0.185***
(5.983)
0.241*
(1.764)
-0.067
(-0.991)
-0.099***
(-15.985)
0.000
(0.625)
-0.141***
(-8.531)
0.390***
(10.101)
finlib
roa*finlib
rnd_sales
lnassets
liab_eq
dum_after97
-0.077
(-1.134)
-0.086***
(-14.734)
0.000
(0.680)
-0.053***
(-7.338)
-0.072
(-1.062)
-0.098***
(-15.891)
0.000
(0.597)
-0.143***
(-8.621)
0.296**
(2.154)
-0.072
(-1.066)
-0.101***
(-14.940)
0.000
(0.629)
Firm fixed effects
Yes
Yes
Yes
Yes
Year fixed effects
No
No
No
Yes
Observations
27975
27975
27975
27975
R-squared
0.035
0.037
0.037
0.041
Number of firms
5678
5678
5678
5678
Note) All estimates are firm-fixed effects panel regression estimates. Variables roa, rnd_sales, lnassets,
andliab_eqare all one-year lagged. The financial liberalization variable finlibis an index derived from PCA.The
variableroa*finlibis an interaction term between roaand finlib. t-statisticsare in parentheses. *** p<0.01, **
p<0.05, * p<0.1.
23
Table 5-4. The Effect of Bank and NBFI Financing on TFP Growth: firm-level regressions
for full sample (dependent variable = dln(TFP))
(1)
(2)
(3)
(4)
(5)
(6)
0.008***
(4.431)
0.008***
(4.610)
0.034***
(4.112)
0.008***
(4.461)
0.008***
(4.479)
0.008***
(4.489)
0.012*
(1.816)
0.157***
(7.748)
-0.014***
(-7.760)
0.012*
(1.822)
0.157***
(7.737)
-0.015***
(-7.708)
0.002
(0.492)
0.013*
(1.881)
0.157***
(7.725)
-0.015***
(-7.718)
0.002
(0.514)
0.007
(0.655)
VARIABLES
dlnloans
0.152***
(7.474)
-0.015***
(-9.204)
0.153***
(7.521)
-0.017***
(-10.059)
0.008***
(4.594)
0.033***
(4.017)
0.013*
(1.847)
0.152***
(7.507)
-0.017***
(-10.120)
0.013***
(6.287)
-0.003
(-0.786)
-0.003
(-0.684)
finlib
dlnloans*finlib
rnd_sales
lnassets
shdirect
shdirect*finlib
dum_after97
Firm fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Year fixed effects
No
No
No
Yes
Yes
Yes
Observations
33163
33163
33163
33163
33163
33163
R-squared
0.006
0.006
0.006
0.010
0.010
0.010
Number of firms
6233
6233
6233
6233
6233
6233
Note) All estimates are firm-fixed effects panel regression estimates. Variables dlnloans, rnd_sales, lnassets,
andshdirectare all one-year lagged. The index finlibis an index derived from PCA.The variabledlnloans*finlibis
an interaction term between dlnloansand finlib. t-statisticsare in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Table 5-5. The Effect of Bank and NBFI Financing on TFP Growth: firm-level regressions
for KRX-listed firm sample (dependent variable = dln(TFP))
(1)
(2)
(3)
(4)
(5)
(6)
0.005
(1.617)
0.005*
(1.692)
0.016
(1.320)
0.006*
(1.749)
0.017
(1.369)
-0.008
(-0.711)
0.065
(0.645)
0.008**
(2.459)
0.008**
(2.299)
0.008**
(2.301)
-0.011
(-0.977)
0.108
(1.083)
-0.011
(-0.982)
0.109
(1.094)
-0.009
(-0.762)
0.104
(1.036)
VARIABLES
dlnloans
finlib
dlnloans*finlib
rnd_sales
0.060
(0.590)
0.065
(0.646)
24
lnassets
-0.004
(-1.414)
-0.006*
(-1.832)
-0.006*
(-1.802)
0.005
(1.566)
-0.002
(-0.345)
-0.003
(-0.414)
-0.007**
(-2.239)
shdirect
-0.007**
(-2.099)
-0.002
(-0.298)
shdirect*finlib
dum_after97
-0.007**
(-2.079)
-0.002
(-0.318)
0.021
(1.065)
Firm fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Year fixed effects
No
No
No
Yes
Yes
Yes
Observations
5187
5187
5187
5187
5187
5187
R-squared
0.001
0.001
0.002
0.028
0.028
0.029
Number of firms
555
555
555
555
555
555
Note) All estimates are firm-fixed effects panel regression estimates. Variables dlnloans, rnd_sales, lnassets,
andshdirectare all one-year lagged. The index finlibis an index derived from PCA.The variabledlnloans*finlibis
an interaction term between dlnloansand finlib. t-statisticsare in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Table 5-6. The Effect of Bank and NBFI Financing on TFP Growth: firm-level regressions
for firms not listed in KRX(dependent variable = dln(TFP))
(1)
(2)
(3)
(4)
(5)
(6)
0.009***
(4.192)
0.009***
(4.335)
0.037***
(3.718)
0.009***
(4.147)
0.009***
(4.199)
0.009***
(4.200)
0.018**
(2.257)
0.157***
(7.318)
-0.017***
(-7.818)
0.018**
(2.268)
0.157***
(7.305)
-0.017***
(-7.804)
0.003
(0.655)
0.019**
(2.277)
0.157***
(7.302)
-0.017***
(-7.805)
0.003
(0.661)
0.003
(0.198)
VARIABLES
dlnloans
0.153***
(7.136)
-0.017***
(-9.222)
0.154***
(7.174)
-0.020***
(-9.932)
0.009***
(4.339)
0.036***
(3.624)
0.017**
(2.151)
0.154***
(7.157)
-0.020***
(-10.00)
0.014***
(5.925)
-0.004
(-0.683)
-0.003
(-0.587)
finlib
dlnloans*finlib
rnd_sales
lnassets
shdirect
shdirect*finlib
dum_after97
Firm fixed effects
Yes
Yes
Yes
Yes
Yes
Yes
Year fixed effects
No
No
No
Yes
Yes
Yes
Observations
27976
27976
27976
27976
27976
27976
R-squared
0.006
0.007
0.007
0.010
0.010
0.010
Number of firms
5678
5678
5678
5678
5678
5678
Note) All estimates are firm-fixed effects panel regression estimates. Variables dlnloans, rnd_sales, lnassets,
andshdirectare all one-year lagged. The index finlibis an index derived from PCA.The variabledlnloans*finlibis
an interaction term between dlnloansand finlib. t-statisticsare in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
25
Industry-level data regressions
Table 5-7. Determinants of Bank and NBFI Financing: industry-level regressions (dependent
variable = dln(loansi))
dlnvai
(1)
(2)
(3)
(4)
-0.018
(-0.212)
-0.020
(-0.234)
0.139
(0.794)
-0.015
(-0.166)
0.138
(0.785)
0.203
(0.469)
0.133
(0.236)
-0.180**
(-2.290)
0.117
(1.593)
finlib
dlnvai*finlib
rnd_vai
0.115
(0.205)
-0.118***
(-5.214)
dum_after97
0.122
(0.217)
-0.177**
(-2.265)
0.459
(1.296)
-0.158
(-0.342)
Industry fixed effects
Yes
Yes
Yes
Yes
Year fixed effects
No
No
No
No
Observations
218
218
218
218
R-squared
0.137
0.140
0.141
0.467
Number of industries
22
22
22
22
Note) All estimates are industry-fixed effects panel regression estimates. Variables dlnvaiandrnd_vaiare oneyear lagged. The index finlib1 is an index derived from PCA.The variabledlnvai*finlibis an interaction
termbetween dlnvaiand finlib. t-statisticsare in parentheses. *** p<0.01, ** p<0.05, * p<0.1.
Table 5-8. The effect of bank and NBFI financing on TFP growth: industry-level regressions
(dependent variable = dln(tfpi))
dlnloansi
(1)
(2)
(3)
(4)
0.008
(0.394)
0.007
(0.329)
0.028
(0.643)
-0.004
(-0.176)
Yes
Yes
218
0.111
22
0.068
(0.482)
-0.007
(-1.234)
0.070
(0.496)
-0.020
(-0.985)
0.005
(0.252)
0.024
(0.537)
0.042
(0.338)
0.067
(0.475)
-0.020
(-0.984)
Yes
No
218
0.016
22
Yes
No
218
0.018
22
Yes
No
218
0.019
22
finlib
dlnloansi*finlib
rnd_vai
dum_after97
Industry fixed effects
Year fixed effects
Observations
R-squared
Number of industries
26
0.155
(1.211)
0.039
(0.279)
Note) All estimates are industry-fixed effects panel regression estimates. Variables dlnloansiandrnd_vaiare oneyear lagged. The index finlib1 is an index derived from PCA. The variabledlnloansi*finlibisan interaction term
between laggeddlnloansiandfinlib. t-statisticsare in parentheses. *** p<0.01, ** p<0.05, * p<0.1
27
Appendix
Details on using principal component analysis to derive the financial liberalization index.
Figure A1. Ratio of Direct Financing to Total Financing in Korea (1990 – 2007)
.65
.6
.55
.5
finstruct1
.7
.75
Unit: percent
1990
1995
2000
year
2005
2010
Note: The index is the ratio of direct financing (bond and equity) to total financing
Source: Flow of Funds Tables, Bank of Korea
28
1.5e-08
Density
1.0e-08
0
0
5.0e-09
2.0e-09
4.0e-09
6.0e-09
8.0e-09
Density
2.0e-08
2.5e-08
Figure A2. Distribution of firms in terms of sales (1991 – 2005)
0
2.000e+08
SALES(NET)
4.000e+08
6.000e+08
KRX-listed firm sub-sample
0
2.000e+08
SALES(NET)
4.000e+08
6.000e+08
Non-KRX firm sub-sample (SME)
Table A1. Summary statistics and definition of variables : firm-level data
Variable
dln(tfp)
dlnloans
ROA
finlib
rnd_sales
lnassets
liab_eq
shdirect
Description
Growth rate of TFP
Growth rate of external financing
Return over asset
Financial liberalization index
Ratio of R&D to sales
log of total assets
ratio of liability to equity
share of direct financing
Obs
32881
32881
32873
32881
32881
32881
32771
32881
Mean
0.004
0.053
0.065
0.792
0.01
24.10
6.53
0.29
Std.
0.104
0.358
0.077
0.268
0.06
1.36
262.02
0.31
Min
Max
-0.497
0.497
-1.000
1.000
-0.653
0.971
0.323
0.991
-0.04
4.24
19.54
31.80
-1547.40 39200.84
0.00
1.00
Note) All variables except for finlibare firm-level annual frequency data. finlibis year-specific
variable.
Table A2. Summary statistics and definition of variables: industry-level data
Variable
dln(tfpi)
dlnloansi
dlnvai
finlib
rnd_vai
Description
Growth rate of TFP
Growth rate of external financing
Growth rate of value-added
Financial liberalization index
Ratio of R&D to value-added
Obs
218
218
218
218
218
Mean
0.007
0.055
0.067
0.889
0.035
Std.
Min
0.034
0.138
0.101
0.184
0.039
Max
-0.126
-0.393
-0.409
0.349
0.000
0.198
0.456
0.421
1.000
0.182
Note) All variables except for finlibare industry-level annual frequency data. finlibis yearspecific variable.
29