金融硏究 Journal of Money & Finance Vol.26 No ) Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea* Won-Gi Kim **

金融硏究 Journal of Money & Finance Vol.26 No. 4 2012. 121) Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea* Won-Gi Kim ** Noh-Sun Kwark *** Abstract Interest rate term spreads and credit spreads are well known to have a predictive power for future fluctuations of output in many developed countries. This study examines the leading behaviors of interest rate term spreads and credit risk spreads in Korea in two ways. First, we apply various empirical methods to delineate the leading behavior of interest rate term spreads and credit risk spreads for business fluctuations between May 1995 and January 2012. Second, using structural VAR models, we decompose the sources of fluctuations of output and interest rate spreads into permanent real shocks and temporary financial shocks, and examine the impulse response of each variable to these shocks, focusing on the leading behavior of the spreads over the business cycle. We establish successfully that in Korea the leading behavior for the term spread and the credit risk spread comprises a tendency for the term spread to increase and for the credit risk spread to shrink about four to six months before an expansion. We also find that much of output fluctuation is attributed to real shocks while fluctuations in the interest rate spreads come from temporary financial shocks. Keywords : Term Spread, Credit Risk Spreads, Leading Behavior, Structural Var, Forecast-Error Variance Decomposition JEL Classification Number : E32, F3 Received 31 March 2012; Revised 1 June 2012; Accepted 22 October 2012. * The authors are grateful to seminar participants at Sogang University and the 2012 Economics Joint Conference, Hyungjoon Lim, and two anonymous referees for valuable comments. The usual disclaimer applies. This work was supported by the 2012 English paper program of the Korea Money and Finance Association. ** Department of Economics, Texas A&M University, College Station, TX 77845, U.S.A. (E-mail : wkim@econ.tamu.edu) *** Corresponding author, Department of Economics, Sogang University, Baekbum-ro 35, Mapo-gu, Seoul 121-742, Korea(Tel : +82-2-705-8770, Fax : +82-2-704-8599, E-mail : kwark@sogang.ac.kr) Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 1

Ⅰ. Introduction Information plays a very important role in financial markets because informational advantage such as inside information may easily create an excess return. Thus, financial market participants try to use all available information and are likely to want to gain as much knowledge of future movements as possible. Some financial variables have been known to move in advance of the real economy showing any direction. The index of leading indicator is composed of these variables, among which interest rate term spreads and credit risk spreads have shown a predictive power for fluctuations in future output. In light of this observation. much attention has been put on many related financial variables and a substantial amount of research has been carried out since Stock and Watson (1989). The term spread (or yield spread) is defined as a difference between a long-term risk-free government bond yield rate and a short-term risk-free rate. This extends when a boom is expected and shrinks when a recession is expected. That is, a steep yield curve slope has been used as a signal for economic recovery, whereas a flat or sometimes downward sloping yield curve indicates an impending economic recession. Lee (2007) shows that a situation in which short-term interest rates are higher than long-term rates for more than a month has been observed in the U.S. ten times over the period between 1960 and 2001 and in most of these cases a recession followed within a year. He concludes that the term spread has a predictive power for future fluctuations of output. A private U.S. economic research institute, the Conference Board, has included the term spread as a component of the index of the leading indicators since 1996. In Korea, Statistics Korea also added the term spread between three year government yield rates and call rates to the leading indicator variables. Extensive research on the predictability of term spreads over the business cycle has been documented (See: Ang, et al., 2006), Evans and Marshall (2007), Rosenberg and Maurer (2008), and Estrella and Hardouvelis (1991), for the U.S., and Lee (2012), Song and Choi (2008), Yie (2008), Park and Kim (2008), Lee (2007), and Ji and Park (2002), for Korea). Credit risk spreads have also often been considered as a leading variable over the business cycle, although the empirical evidence supporting their leading behavior is not as strong as that for the term spread. Bernanke (1990) and Friedman and Kuttner (1993) showed that the spread between three month commercial paper rates and three month T-bill rates are closely related to the future business cycle. Recently Gertler and Lown (1999) and Mody 2 金融硏究제 26 권제 4 호 2012

and Taylor (2003) have focused on more credit risk related spreads by using the spread between high yield bond (junk bond) rates and risk-free interest rates based on the financial accelerator theory. The purpose of this study is to examine the leading behavior of term spreads and credit risk spreads in Korea in two ways. First, we apply various empirical methods so as to show the leading behavior of interest rate term spreads and credit risk spreads for business fluctuations over the period from May 1995 to January 2012. Given that the monetary authority in Korea changed its operating target from monetary aggregates to short-term interest rates, call rates, after the financial crisis in 1998, it is worth examining the behavior of interest rates as a benchmark variable linking the financial sector to the real sector. Second, using structural VAR models, we decompose the sources of fluctuations of output and interest rate spreads into permanent real shocks and temporary financial shocks and examine the impulse response of each variable to these shocks, focusing on the leading behavior of the spreads over the business cycle. We establish successfully the leading behavior of the term spread and the credit risk spread in Korea, indicating that the term spread tends to increase and the credit risk spread tends to shrink about five to six months before an expansion. We also find that many of the output fluctuations are attributed to real shocks while fluctuations in the interest rate spreads come from temporary financial shocks. The rest of the paper is organized as follows. In Section II, the theoretical background for the leading behavior of the term spread and the credit risk spread is discussed. In Section III, the data used in the empirical analysis are explained and descriptive statistics and a long horizon analysis are provided. In Section IV a structural VAR model is constructed and, with some identifying restrictions, real shocks and financial shocks are identified to examine the leading behavior of the term spread and the credit risk spread. In Section V, we summarize our main result and conclude. Ⅱ. Theoretical Background 1. Term Spreads and Future Business Activities Following Plosser and Rouwenhorst (1994) or Mody and Taylor (2003), the slope of Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 3

the yield curve depends on the expected real interest rate, the present and future inflation, risks, and the term premium. Therefore, theoretical investigation addressing the relationship between the term spread and future economic activities should start by considering the link between real interest rates and macroeconomic fluctuations. Consider a simple Euler equation derived from the first order conditions of a representative agent s optimization problem. 1), (1), (2) where is the momentary utility function, is consumption, and are oneperiod and two-period annual interest rates at time, is the discount factor, and is the conditional expectation operator based on all information available at time t. The above equations are typical intertemporal first order conditions. From the above equations it is straightforward to get. (3) Thus, the interest rate term spread is related to the difference between the ratio of the expected marginal utility of consumption at time to the expected marginal utility of consumption at time and the ratio of the marginal utility of consumption at time to the expected marginal utility of consumption at time. For example, when consumption is expected to increase from time to time faster than is the case from time to time, the term spread between the two-period annual interest rate and the one-period interest rate widens. If the expected inflation rate is constant, this relationship can be applied to the spread between nominal interest rates with different maturity. Extending the above argument to the spread between the long-term interest rate and the short term interest rate, when the consumption level is expected to grow faster in the future, the term spread tends to increase. That is, the term spread increases before output rises, which is the leading behavior of the term spread for business fluctuations observed in the data. 1) This part modifies slightly Mody and Taylor (2003) s discussion. 4 金融硏究제 26 권제 4 호 2012

Harvey (1993) shows that the term spread is related to the movement of consumption using the intertemporal CAPM instead of a representative agent model. His argument starts with a consideration of consumption smoothing behavior, that is that when people expect a recession in the future, they increase saving by purchasing a long-term bond to prevent a decrease in future consumption for the purpose of consumption smoothing. This behavior increases the demand for long-term bonds and thus, the price of long-term bonds increases and the long-term interest rates fall, which makes the yield curve flat or even downward sloping when a depression is expected in the future. However, as Deaton (1992) and Taylor (1999) point out, the above argument holds at a marginal relationship and many empirical studies documents that the relationship between the current term spread and future movement in consumption is very weak. Some studies such as Chirinko (1993) and Taylor (1999) have tried to find a channel through investment but this has indicated that the current term spread is also weakly related to future movements in investment. Bernard and Gerlach (1998) emphasize the effect of monetary policy and expected inflation in the relationship between the term spread and future economic activities. The monetary policy argument views monetary policy change as being a leading behavior of the term spread. When monetary contractionary policy raises short-term interest rates, people expect this tight monetary policy to be temporary, which results in a relatively small increase in the future short-term interest rate and a decline in the term spread. Contractionary monetary policy affects real economic activities with a lag, which means the decline in the term spread is associated with future economic recession. The expected inflation argument focuses on people s expectations for the future economy. When an expectation of a future recession prevails, the expected inflation rate falls and thus long-term interest rates fall. If popular expectation is on average correct, the fall in the term spread or the flat yield curve is related to future economic recession. 2) 2. Credit Risk Spreads and Future Business Activities Credit risk spreads are defined as a difference between interest rates on risky bonds 2) Haubrich and Dombrosky (1996) find that the term spread between the 10-year treasury bond yields and the three month T-bill rates predicts the GDP growth rate in four quarters later but its predictive power changes over time and becomes significantly lower since 1980. Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 5

and risk-free interest rates with the same maturity. Most studies that address the leading behavior of the credit risk spread use the spread between three-month commercial paper (CP) rates and the three-month Treasury bill rates. Since only recognized firms can issue commercial papers, the default risk in commercial papers does not appear to be existent. However, commercial papers are ex ante not risk-free. Friedman and Kuttner (1993) find that a wider credit risk spread predicts an economic recession, whereas a smaller credit risk spread predicts an expansion. They explain this leading behavior of the credit risk spread through three channels. First, expectations of recession in the future raise the default risk of corporate bonds. Second, when monetary policy becomes contractionary and it affects economic activity with a lag, banks reduces loans to firms, which makes firms issue commercial papers and raises the interest rate of commercial papers. Thus, a wider credit spread between the commercial paper rate and risk-free Treasury bill rates is associated with a recession in the future. Third, when an economic recession is near, firms are likely to have trouble with cash flows and try to have liquidity, which makes firms issue new commercial papers. Due to the above three channels, an increase in the spread between the commercial paper rate and Treasury bill rates predicts a future recession. The leading behavior of the credit risk spread, defined as a difference between the three-month commercial paper rates and the three-month Treasury bill rates, has appeared to be weaker since 1980 due to an increase in the substitutability across financial assets coming from globalization and liberalization in financial markets (Gertler and Lown, 1999). As an alternative explanation many studies focus on high yield bond spreads based on the financial accelerator theory suggested by Bernanke et al. (1999). Gertler and Lown (1999) and Mody and Taylor (2003) define high yield bond spreads as a difference between junk bond yields and AAA-rated corporate bonds yields or T-bill rates and show that the junk bond yield spreads are closely related with future business fluctuations. Ⅲ. Data and Long Horizon Analysis 1. Data and Descriptive Analysis Even though the size of the Korean bond market is substantial and its history is long, 6 金融硏究제 26 권제 4 호 2012

this market was not much developed until recently. The trading volume in the Korean bond market was not heavy and was regulated by the government. After the financial crisis in Korea in 1998 the Bank of Korea changed its monetary policy target from monetary aggregates such as M2 to a short-term interest rate in the call market, which is an overnight loan between commercial banks with the purpose of managing bank reserves. 3) With liberalization and deregulation of the financial markets after the financial crisis, the Korean bond markets have developed significantly in both their qualitative and quantitative aspect. This study covers the period starting from about three years before the financial crisis in Korea, spanning from May 1995 to January 2012. 4) The frequency of the data is monthly. We use several financial data and real output data for a VAR analysis. To calculate the term spread we use three year government bond yield rates ( ) and five year government bond yield rates for long-term interest rates and call rates (overnight loans between commercial banks for reserves; ), three month CD (certificate of deposit, ) rates, and three month commercial paper rates ( ) for the short-term interest rates. 5) Since the empirical results for the terms spread with five year government bond yield rates are very similar to those with three year government bond yield rates, only those empirical results with three year government bond yield rates are reported. The certificate of deposit market has been growing significantly since the financial crisis in Korea and the CD rates have been used as a benchmark interest rate for collateralized mortgage loans. The term spread used in the index of economic leading indicator in the U.S. is defined as a difference between the ten year Treasury bond yields and the federal funds rate. However, the ten year bond market in Korea started to be traded only in 2007. Thus, we use three year government bond yield rates as a long-term interest rate. The term spread defined as a difference between three year government bond yield rates and call rates is currently one of ten components that comprise the leading economic indicator in Korea. For the credit risk spread, we use three year government bond yields 3) The call market corresponds to the federal funds market in the U.S. 4) The data coverage used in this paper is limited due to the availability of the three year government bond yield rates. 5) As a short-term interest rate the one year monetary stabilization bond yield rates are used but they do not show a close relationship with business fluctuations. The monetary stabilization bonds are issued by the Bank of Korea for the purpose of monetary policy. The total outstanding of the monetary stabilization bonds at the end of year 2011 is about 165 trillion wons. Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 7

for risk-free rates and three year AA- rated corporate bond yields ( ) for risky rates. 6) <Figure 1> shows the fluctuations of the term spreads calculated with three year government bond yields for the long-term rate over the sample period. The shaded areas indicate recessions. The term spreads were sharply negative in 1998 due to the credit crunch and financial crisis in Korea. It confirms the conventional observation that the yield curve becomes flat or downward sloping during economic recessions. The spreads became quickly reversed to the positive territory when the economy started to recover in the second half of 1998. The figure shows that the term spreads tend to reach a peak several months in advance of a peak of a business cycle. Since the starting point of a shaded area indicates a business cycle peak, the term spread tends to increase several months before the peak of the business cycle. The term spread also tends to decrease several months before a trough of a business cycle since the end point of a shaded area implies a trough of a business cycle. As such, the figure shows a leading behavior of the term spread over the business cycle. <Figure 1> Term Spreads over Business Cycle, May 1995-Jan. 2012 % 4 2 0-2 1995/05 1996/05 1997/05 1998/05 1999/05 2000/05 2001/05 2002/05 2003/05 2004/05 2005/05 2006/05 2007/05 2008/05 2009/05 2010/05 2011/05-4 -6-8 -10-12 -14 gb3-call gb3-cd gb3-cp To investigate the statistical properties of the term spreads over business cycles, <Figure 2> plots correlograms for each of the three term spreads with the industrial production 6) For a risky interest rate the BBB- corporate bond yield rates may be a better indicator measuring a default risk but the data are available only after October 2000. 8 金融硏究제 26 권제 4 호 2012

index. Each of the graphs indicates the correlation coefficients between each of the term spreads at time and the industrial production index at time where is the number at the horizontal axis. When the correlation coefficients are high in absolute terms in the left side of the origin, the term spread moves in advance of the fluctuations in the industrial production index. The highest correlation coefficients for all the three cases of the spreads,,, and occur at the correlation between the term spread at time and the industrial production index at time, which means there is a tendency for the term spread to move five months ahead of the movement in the industrial production index. The correlation coefficients between the industrial production index at time and the spreads of and at time, are 0.43 and 0.40, respectively but the term spread of shows a low correlation coefficient of 0.27 with the industrial production index. <Figure 2> Correlogram between Term Spreads and Industrial Production Index 0.4 0.3 0.2 0.1 0-12 -10-8 -6-4 -2 0 2 4 6 8 10 12 corr(ip, gb3-call) corr(ip, gb3-cd) corr(ip, gb3-cp) This analysis is also carried out for the credit risk spreads. <Figure 3> shows the fluctuations of the credit risk spreads calculated with a three year government bond yield rate ( ) for a risk-free rate and a three year AA- rated corporate bond yield rate ( ) and a three year BBB- rated corporate bond yield rate ( ) for a risky rate. The credit risk spreads extended significantly at the beginning of the financial crisis in 1998, several Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 9

<Figure 3> Credit Risk Spread over Business Cycle, May 1995-Jan. 2012 % 9 7 5 3 1-1 1995/05 1996/05 1997/05 1998/05 1999/05 2000/05 2001/05 2002/05 2003/05 2004/05 2005/05 2006/05 2007/05 2008/05 2009/05 2010/05 2011/05 aa-gb3 bbb-gb3 <Figure 4> Correlogram between Credit Risk Spread and Industrial Production Index 0.1 0-12 -10-8 -6-4 -2 0 2 4 6 8 10 12-0.1-0.2-0.3 corr(ip, aa-gb3) months before the trough of the business cycle in August 1998. The credit risk spread also tended to peak several months ahead of (or immediately before) a trough of the business 10 金融硏究제 26 권제 4 호 2012

cycle, identified as the end point of a shaded area. An increase in the credit rate spread seems to be associated with a future recession in the business cycle. The most recent sharp increase in the credit risk spread reflects the global financial crisis starting in September 2008. 7) The correlogram for the credit risk spread with the industrial production index is drawn in <Figure 4>. The figure shows that the correlation coefficient between the credit risk spread ( ) and the industrial production index becomes the largest in absolute term (-0.27) for the correlation between the credit risk spread at time and the industrial production index at time, which implies that a decline in the credit risk spread is associated with an increase in the industrial production index two months later. Thus, the credit risk spread leads the business cycle. 2. Long Horizon Analysis A more extensive method of showing a leading behavior of a variable over the business cycle, named the long horizon analysis, is suggested by Estrella and Hardouvelis (1991). The estimation equation used in the long horizon analysis is as follows:, (4) where and are the industrial production index and term spread (or credit risk spread) at time, respectively, and is the forecasting horizon. In essence, the estimation regression examines the predictability of the spreads for real activity. The first term, ( ) is for converting the coefficient into an annual percentage effect. <Table 1> reports the estimated coefficient on the term spread defined as the difference between three year government bond yield rates and call rates, CD rates, or CP rates. The numbers in parentheses indicate Newey and West (1987) heteroskedasticity and autocorrelation consistent 7) However, the leading behavior of the credit risk spread over the business cycle seems to be weaker than that of the term spread. Thus, currently the term spread is included as a component in the index of the leading indicator while the credit risk spread is not. Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 11

<Table 1> Long Horizon Analysis for Leading Behavior of Term Spread Forecast horizon 1 3.83 *** (3.56) 0.06 5.42 *** (3.66) 0.06 3.72 *** (3.74) 0.07 2 3.53 *** (4.65) 0.10 4.63 *** (4.41) 0.09 3.09 *** (4.36) 0.09 3 3.11 *** (4.84) 0.11 3.86 *** (4.74) 0.09 2.59 *** (4.29) 0.09 4 2.60 *** (4.55) 0.10 3.24 *** (4.09) 0.08 2.09 *** (3.89) 0.07 5 2.17 *** (4.17) 0.08 2.63 *** (3.65) 0.06 1.65 *** (3.38) 0.06 6 1.74 *** (3.59) 0.06 1.98 *** (2.94) 0.04 1.29 *** (2.85) 0.04 7 1.31 *** (2.87) 0.04 1.35 ** (2.13) 0.02 0.94 ** (2.20) 0.02 8 0.92 ** (2.14) 0.02 0.81 (1.37) 0.01 0.51 (1.27) 0.01 9 0.64 (1.58) 0.01 0.49 (0.88) 0.00 0.20 (0.53) 0.00 10 0.44 (1.15) 0.01 0.29 (0.56) 0.00 0.03 (0.08) 0.00 12 0.13 (0.38) 0.00-0.03 (-0.07) 0.00-0.17 (-0.55) 0.00 18-0.29 (-1.13) 0.01-0.56 (-1.59) 0.01-0.53 ** (-2.26) 0.03 24-0.49 ** (-2.50) 0.03-0.83 *** (-3.09) 0.05-0.74 *** (-4.16) 0.09 Note : 1) In parentheses are Newey and West (1987) heteroskedasticity and autocorrelation consistent standard errors corrected with twelve lags. All estimation equations have a constant term that is abstracted from the table. 2) ***, **, and * denote statistically significant at the 1%, 5%, and 10% level in a two-tailed test, respectively. 12 金融硏究제 26 권제 4 호 2012

standard errors corrected with twelve lags. 8) This shows that the term spread helps to predict the growth of the industrial production index up to five months ahead. A one percent point increase in the term spread with call rates as a short-term rate is associated with an 0.92 3.83 percent cumulative increase in annual term in the industrial production index over the following eight months period. For the term spread using CD rates as a short-term rate, a one percent point increase in the term spread is related with a 1.35 5.42 percent cumulative increase in annual term in the industrial production index over the next seven months period. The term spread with CP rates as a short-term rate shows a similar behavior to the other two term spreads: a one percent increase in the term spread is associated with an 0.94 3.72 percent cumulative increase in annual term in the industrial production index. The sign of the coefficient on the term spread turns to negative for the 24 month forecast horizon. This may reflect the mean-reverting behavior of business cycles 9) or monetary and fiscal policy responses facing business fluctuations. 10) Boudoukh et al. (2008) argue that the long horizon analysis may include other effects as the forecast horizon is longer. They accordingly suggest that caution is required in interpreting the relationship between the term spread and output fluctuation since the empirical results from using the long horizon method will be a mixture of many factors, the effects of which we need to decompose. The long horizon analysis for the credit risk spread is reported in <Table 2>. The 8) When Newey and West (1987) heteroskedasticity and autocorrelation consistent errors are computed, we chose two to twelve lags but the results were not qualitatively different. We fix the lag to 12 following Hamilton and Kim (2002). 9) This idea is supported by the fact that following the official turning point date announced by Statistics Korea, the business cycle expansion from a trough to the following peak lasts around one to two years. There may be a possibility of a missing relationship with other variables excluded in our long horizon model. However, the reversal of the signs of the coefficients with long lags is not the main focus of this paper. 10) We examined the term spread between the three year government bond yield rates and the one year monetary stabilization bond yield but the spread does not show any predictive power for future fluctuations in the industrial production index. We suspect that the one year monetary stabilization bond is not really considered as a short-term bond in the financial market or the maturity difference between the three year government bond yield rate and the one year monetary stabilization bond yield rate is too short and thus the term spread does not accurately reflect a change in inflation expectations. This empirical result differs from those found in Park and Kim (2008) and Ji and Park (2002), which found a predictive power of the term spread with one year monetary stabilization bond yield rates for future output fluctuations. Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 13

<Table 2> Long Horizon Analysis for Leading Behavior of Credit Risk Spread Forecast horizon 1 2 3 4 5 6 7 8 9 10 12 18 24-5.04 *** (-2.71) -3.20 ** (-2.39) -2.21 ** (-1.93) -1.23 (-1.25) -0.66 (-0.72) -0.24 (-0.29) 0.26 (0.32) 1.02 (1.38) 1.52 ** (2.23) 1.73 *** (2.71) 1.91 *** (3.39) 2.31 *** (5.72) 2.21 *** (7.47) Note : 1) In parentheses are Newey and West (1987) heteroskedasticity and autocorrelation consistent standard errors corrected with twelve lags. All estimation equations have a constant term that is abstracted from the table. 2) ***, **, and * denote statistically significant at the 1%, 5%, and 10% level in a two-tailed test, respectively. 0.04 0.03 0.02 0.01 0.00 0.00 0.00 0.01 0.03 0.04 0.06 0.15 0.24 credit risk spread is calculated as a difference between three year AA- rated corporate bond yield rates and three year government bond yield rates. An increase in the credit risk spread is associated with a decline in output two to five months later. A one percent point increase in the credit risk spread implies that the industrial production index tends to decrease in one to three months by -2.21-5.04 percent in the annual percentage term. Therefore, we 14 金融硏究제 26 권제 4 호 2012

can observe a leading behavior of the credit risk spread over fluctuations in the industrial production index by using the long horizon analysis. 11) Ⅳ. Structural VAR Analysis From the descriptive analysis and the long horizon analysis we found a leading behavior of the term spread and the credit risk spread over the business cycle. In this section, we attempt to analyze these spreads in greater detail to see how the industrial production index and the term spread (or the credit spread) respond to shocks identified by a set of identifying restrictions. Within this exercise we can also look for the leading behavior of the spreads over the business cycle, the size of the responses of the variables to each of the shocks, and the relative importance of the shocks in explaining fluctuations of the variables. 1. A Structural VAR Model and Identification Consider a simple bivariate unrestricted vector autoregressive model as follows:. (5) A structural VAR model starts with an unrestricted VAR model estimation and converts the estimated model into a structural model linking structural shocks with a set of identifying restrictions. To convert the two estimated residuals into structural shocks requires an identifying restriction. 12) We use the industrial production index and term spreads (or credit risk spreads) in the bivariate VAR model. First, we test an existence of unit root for each of the variables as in <Table 3>. <Table 3> reports unit root test statistics for the industrial production index and various spreads. Both the ADF test statistics and the Phillips-Perron 11) The reversal of the sign of the coefficient at the 9 to 24 months horizon is similar to the results for the term spread. 12) For a detailed discussion on the methodology of structural VAR models, refer to Kwark (2004). Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 15

test statistics suggest that the industrial production index has a unit root and the spreads considered are stationary. With this test, we can identify a permanent shock and a temporary shock from the unrestricted VAR estimation. We consider a permanent shock as a real shock and a temporary shock as a financial shock. An identifying restriction we adopt is that a financial shock does not have a long-run (permanent) effect on real output. 13) This identifying restriction can be expressed as in the following model:, (6) where is a real shock which has a permanent effect on the industrial production index and is a financial shock which has only a temporary effect on the industrial production index. We examined a bivariate structural model for term spreads with various measures and found very similar results. Thus, the following discussion of the empirical results will focus on the term spread between three year government bond yield rates and call rates ( ). <Table 3> Unit Root Test ADF Statistics -0.52-13.58 *** -4.34 *** -4.30 *** -4.44 *** -5.68 *** Phillips-Perron Statistics -0.54-13.60 *** -3.56 *** -3.69 *** -3.87 *** -4.41 *** Note : 1) The lag length is chosen by the SIC and a constant term is included. 2) *** denotes a rejection of the null hypothesis of unit root at the 1% significance level, respectively. 13) These identifying restrictions are similar to the ones used by Blanchard and Quah (1989) for identifying supply shocks and demand shocks in the bivariate VAR model composed of GDP and unemployment. They assume that a demand shock does not have a permanent effect on real output. 16 金融硏究제 26 권제 4 호 2012

2. Impulse Response and Variance Decomposition The purpose of this section is to quantitatively examine a leading behavior of the spread over the business cycle and the relative importance of structural shocks in explaining fluctuations in the variables concerned by a structural VAR model analysis. <Figure 5> and <Figure 6> show impulse responses of the industrial production index and the term spread ( ) to a one percent standard deviation innovation increase in a real shock. The dotted lines denote the 95 percent confidence interval calculated by the bootstrapping method with 1000 simulations. The impulse response of the industrial production index to a real shock is hump-shaped reaching a highest point of 2.66 percent increase six months after a one percent standard deviation increase in real shock is realized while the response of the term spread reaches a peak of 0.55 percent point increase two months after the real shock. The real shock has a permanent effect on the industrial production index by a roughly 2.11 percent increase while it has only a temporary effect on the term spread since the spread is stationary. The confidence intervals of the impulse responses of the spreads and the impulse responses of the industrial production and the spread to a temporary financial shock show a narrower band as the horizon extends. From this impulse response analysis we may conclude that the term spread tends to move four months earlier than the industrial production index, which is approximately consistent with results from previous studies (e.g., Lee, 2001 and Ji and Park, 2002), which found a leading behavior of the term spread over the business cycle of three to nine months. <Figure 7> and <Figure 8> show impulse responses of the industrial production index and the term spread to a one percent standard deviation increase in the financial shock. Since we assume financial shocks to be temporary, the long-run effect of a financial shock converges to zero. A one percent standard deviation increase in the financial shock increases the industrial production index, reaching a maximum point of 0.61 percent seven months after the shock is realized. The term spread shows a peak response of an 0.67 percent point increase in one month. The impulse responses clearly indicate a leading behavior of the term spread over the industrial production index by about six months. The forecast-error variance decomposition is provided in <Table 4>. The numbers in the table imply the relative importance of the contribution of each of the shocks in explaining Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 17

<Figure 5> Impulse Response of IP to a Real Shock 3.5 3 2.5 2 1.5 1 0.5 0 0 5 10 15 20 25 30 35 40 45 50 55 60 IP 95% Confidence Interval <Figure 6> Impulse Response of the Term Spread to a Real Shock 0.8 0.6 0.4 0.2 0-0.2 0 5 10 15 20 25 30 35 40 45 50 55 60 gb3-call 95% Confidence Interval unexpected fluctuations -period later. As expected from the impulse response analysis, much of the fluctuation in the industrial production index is explained by real shocks. Over all the horizons real shocks are responsible for 70 to over 90 percent of the fluctuation in the industrial production index. Financial shocks explain up to about 27 percent of the fluctuation 18 金融硏究제 26 권제 4 호 2012

in the industrial production index in one month. On the contrary, the term spread is largely explained by financial shocks over all of the horizons. Financial shocks explain about 60 percent of the fluctuation of the term spread. <Figure 7> Impulse Response of IP to a Financial Shock 2 1.5 1 0.5 0-0.5 0 5 10 15 20 25 30 35 40 45 50 55 60-1 -1.5-2 IP 95% Confidence Interval <Figure 8> Impulse Response of Term Spread to a Financial Shock 1 0.8 0.6 0.4 0.2 0-0.2 0 5 10 15 20 25 30 35 40 45 50 55 60 gb3-call 95% Confidence Interval Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 19

<Table 4> -Period Ahead Forecast-Error Variance Decomposition 1 2 4 8 12 16 24 40 60 real shock financial shock real shock financial shock 73.3 78.6 87.7 93.4 94.7 95.7 96.9 98.0 98.6 26.7 21.4 12.3 6.6 5.3 4.3 3.1 2.0 1.4 38.3 37.5 41.8 40.2 39.5 39.0 38.8 38.8 38.8 61.7 62.5 58.2 59.8 60.5 61.0 61.2 61.2 61.2 The same exercise with a structural VAR model is conducted for the credit spread. The results are given in <Figure 9>~<Figure 12>, and <Table 5>. The impulse responses in <Figure 9> and <Figure 10> show that the industrial production index responds to a real shock reaching a maximum of about 2.65 percent six months later while the credit risk spread responds negatively to the real shock. The maximum response of the credit risk spread is an 0.22 percent fall two months later. For a positive innovation of the credit risk spread the industrial production index responds with a fall of 0.59 percent six months after the shock and the credit risk spread shows a quick positive maximum response of 0.61 percent in a month. From the impulse response analysis, the credit risk spread shows a leading behavior <Figure 9> Impulse Response of IP to a Real Shock 3.5 3 2.5 2 1.5 1 0.5 0 0 5 10 15 20 25 30 35 40 45 50 55 60 IP 95% Confidence Interval 20 金融硏究제 26 권제 4 호 2012

of industrial production by about four to five months. The forecast-error variance decomposition for the credit risk spread shown in <Table 5> is similar to that for the term spread. Over all the horizons most of the fluctuation in the industrial production is explained by real shocks and the contribution of financial shocks is small. The fluctuation of the credit spread in the short-term is largely explained by financial shocks but the contribution of real shocks increases as the horizon expands. <Figure 10> Impulse Response of Credit Spread to a Real Shock 0.2 0.1 0-0.1 0 5 10 15 20 25 30 35 40 45 50 55 60-0.2-0.3-0.4-0.5 aa-gb3 95% Confidence Interval <Figure 11> Impulse Response of IP to a Financial Shock 1.5 1 0.5 0-0.5 0 5 10 15 20 25 30 35 40 45 50 55 60-1 -1.5-2 IP 95% Confidence Interval Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 21

<Figure 12> Impulse Response of Credit Spread to a Financial Shock 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0-0.1 0 5 10 15 20 25 30 35 40 45 50 55 60-0.2 aa-gb3 95% Confidence Interval <Table 5> -Period Ahead Forecast-Error Variance Decomposition real shock financial shock real shock financial shock 1 2 4 8 12 16 24 40 60 98.5 98.6 97.1 96.3 97.2 97.8 98.4 99.0 99.3 1.5 1.4 2.9 3.7 2.8 2.2 1.6 1.0 0.7 4.3 7.2 11.7 11.9 12.0 12.0 12.0 12.0 12.0 95.7 92.8 88.3 88.1 88.0 88.0 88.0 88.0 88.0 V. Concluding Remarks The term spread may reflect a change in monetary policy and a change in expected inflation. The credit risk spread also incorporates expectations about default risk in the future. These links make the term spread or the credit risk spread useful means of predicting future business cycles. The term spread has been used as a component of a leading index in many 22 金融硏究제 26 권제 4 호 2012

developed countries, including the U.S., and was recently added to a leading indicator in Korea. This study examined the leading behavior of the term spread and the credit risk spread in Korea in two ways. We first applied various empirical methods such as correlogram analysis and long horizon analysis to the term spread and the credit risk spread to find a leading behavior of the spreads in business fluctuations over the period from May 1995 to January 2012. Noting that after the 1998 financial crisis in Korea monetary policy changed its operating target from monetary aggregates to short-term interest rates, call rates, we think the interest rate variables may reflect substantial information regarding business fluctuations more quickly than before. The term spread as well as the credit risk spread showed a predictive power for future fluctuations in industrial production. Second, using structural VAR models, we decompose the sources of fluctuations of output and interest rate spreads into two sorts, permanent real shocks and temporary financial shocks, and examine the impulse response of each variable to the shocks focusing on the leading behavior of the spreads over the business cycle. From the impulse response analysis in the structural VAR model, we establish the leading behavior of the term spread and the credit risk spread successfully for Korea. The term spread tends to increase and the credit risk spread tends to shrink about four to six months before an expansion. We also find that much of output fluctuation is attributed to real shocks while fluctuations in the interest rate spreads come from temporary financial shocks. Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 23

<References> 1. Ang, A. Piazzesi, M., and M. Wei, What does the yield curve tell us about GDP growth?, Journal of Econometrics 131, 2006, 359-403. 2. Bernanke, B. S., On the Predictive Power of Interest Rates and Interest Rates Spread, New England Economic Review, Federal Reserve Bank of Boston, Issue November, 1990, 51-68. 3. Bernanke, B. S., M. Gertler, and S. Gilchrist, The Financial Accelerator in a Quantitative Business Cycle Framework, Handbook of Macroeconomics 1, Part C, Elsevier Science, 1999, 1341-1393. 4. Bernard, H. and S. Gerlach, Does the Term Structure Predict Recessions? The International Evidence, International Journal of Finance and Economics 3(3), 1998, 195-215. 5. Blanchard, O. J. and D. Quah, The Dynamic Effects of Aggregate Demand and Supply Disturbances, American Economic Review 79(4), 1989, 654-673. 6. Boudoukh, J., M. Richardson, and R. F. Whitelaw, The Myth of Long-Horizon Predictability, Review of Financial Studies 21(4), 2008, 1577-1605. 7. Chirinko, R. S., Business Fixed Investment Spending : Modeling Strategies, Empirical Results, and Policy Implications, Journal of Economic Literature 31, 1993, 1875-1911. 8. Deaton, A., Understanding Consumption, Oxford : Clarendon Press, 1992. 9. Estrella, A. and M. Hardouvelis, The Term Structure as a Predictor of Real Economic Activity, Journal of Finance 46(2), 1991, 555-576. 10. Evans, C. L. and D. A. Marshall, Economic Determinants of the Nominal Treasury Yield Curve, Journal of Monetary Economics 54(7), 2007, 1986-2003. 11. Friedman, B. M. and K. N. Kuttner, Why Does the Paper-Bill Spread Predict Real Economic Activity?, Why Does the Paper-Bill Spread Predict Real Economic Activity?, in Stock and Watson (ed.), New Research on Business Cycles, Indicators, and Forecasting, Chicago : University of Chicago Press for the NBER, 1993, 213-254. 12. Gertler, M. and C. S. Lown, The Information Content of the High Yield Bond Spread for the Business Cycle, Oxford Review of Economic Policy 15(3), 1999, 132-150. 13. Harvey, C. R., Term Structure Forecasts Economic Growth, Financial Analysts Journal, 1993, 6-8. 24 金融硏究제 26 권제 4 호 2012

14. Haubrich, J. G. and A. M. Dombrosky, Predicting Real Growth Using the Yield Curve, Economic Review, Federal Reserve Bank of Cleveland, Issue Q I, 1996, 26-35. 15. Ji, H. and S. Park, Evaluation of the Predictive Power of Interest Rate Spreads, Korean Journal of Financial Management 19(2), 2002, 233-252 (in Korean). 16. Kwark, N. S., Structural Vector Autoregressive Models : Methodology and Application, Sogang Economic Review 33(2), 2004, 1-20 (in Korean). 17. Lee, C., Study on the Predictive Power of Interest Rate Spreads for Business Cycles, Manuscript, LG Economic Research Institute, 2001 (in Korean). 18. Lee, D., Effects of Long-Term Bond Risk Premium on Term Spreads and its Implications, Monthly Bulletin, Bank of Korea, 2007, 23-48 (in Korean). 19. Lee, K. Y., Business Cycles and Financial Variables, Unpublished Manuscript, 2012 (in Korean). 20. Newey, W. K. and K. D. West, A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix, Econometrica 55(3), 1987, 703-708. 21. Mody, A. and M. P. Taylor, The High Yield Spread as a Predictor of Real Economic Activity : Evidence of a Financial Accelerator for the United States, IMF staff paper 50(3), 2003, 22. Park, K. and S. Kim, Study on the Predictive Power for Business Fluctuations, Proceedings of the Korea Money and Finance Association Annual Conference, 2008 (in Korean). 23. Plosser, C. I. and K. G. Rouwenhorst, International Term Structures and Real Economic Growth, Journal of Monetary Economics 33(1), 1994, 133-155. 24. Rosenberg, J. V. and S. Maurer, Signal or Noise? Implications of the Term Premium for Recession Forecasting, FRB of NY Economic Policy Review, 2008, 1-11. 25. Song, J. and Y. Choi, Bond Risk Premia and Business Cycle, Economic Analysis 14(4), 2008, 1-46 (in Korean). 26. Stock, J. H. and M. W. Watson, New Indexes of Coincident and Leading Economic Indicators, NBER Macroeconomics Annual 4, 1989, 351-409. 27. Taylor, M. P., Real Interest Rates and Macroeconomic Activity, Oxford Review of Economic Policy 15, 1999, 95-113. 28. Yie, M., An Analysis on the Predictability of Term Spreads over Output Gap, Bank of Korea Working Paper No. 343, 2008. Leading Behavior of Interest Rate Term Spreads and Credit Risk Spreads in Korea 25

金融硏究 Journal of Money & Finance Vol. 26 No. 4 2012. 12 1) 출자총액제한제도가대기업자산구성에미치는효과에관한연구 김민수 * 성인모 ** 소종일 *** 국문초록 본연구는출자총액제한제도의폐지가대기업자산구성의변화에미치는영향을분석하였다. 과거출자총액제한제도의적용대상이었던 14개그룹소속기업을대상으로, 2001 2011기간동안분석한결과, 출자총액제한제도의실질적폐지이후대기업의총자산중유형자산의보유비중이축소된반면투자자산과현금성자산의보유비중이확대된것으로나타났다. 특히이러한현상이상장사에서더강하게나타났으며, 출자총액제도폐지이후상장사가비상장사의지분을더욱확대하는반면상장사의실물투자는더욱감소하는행태가관찰되었다. 이러한결과는주요대기업들이 2007년출자총액제한제도가사실상폐지된이후계열사출자확대, M&A 등으로외형을확장한반면실물투자에는상대적으로인색했음을시사한다. 이는지속적인정책완화기조에도불구하고출자총액제한제도의존치가기업의계열사간출자에따른외형확장을억제시켰다고할수있다. 따라서출자총액제한제도의폐지를보완하고공정경쟁, 실물투자를통한성장동력확보등을위한대책을마련해야할것으로판단된다. 핵심단어 : 출자총액제한제도, 대기업집단, 경제력집중, 자산구성, 투자자산 JEL 분류기호 : G32, G34, G38 투고일 2012년 07월 02일 ; 수정일 2012년 07월 31일 ; 게재확정일 2012년 08월 28일 * 고려대학교경영대학 (Tel : 02-3290-1363, E-mail : equili@korea.ac.kr) ** 한국은행과장 (Tel : 02-759-5618, E-mail : imsung@bok.or.kr) *** 교신저자. 고려대학교경영대학 (Tel : 02-3290-2619, E-mail : soj405@korea.ac.kr) 출자총액제한제도가대기업자산구성에미치는효과에관한연구 27

Ⅰ. 서론 대기업집단으로의경제력집중을완화하고기업지배구조개선을위해시행되었던출자총액제한제도가 2009년 3월에폐지되었다. 전경련및대기업은동제도가기업의정상적인투자활동을제약하여설비투자를위축시키고신규유망산업진출, 주력업종중심의수직적통합등사업구조개편을저해하였다고주장하였으며, 사외이사제등내부감시장치도입으로대기업의지배구조가현저히개선되었다는의견을피력하였다. 공정거래위원회는대표적사전규제의상징인출자총액제한제도를폐지하면서, 기업규제완화에대한정부의확고한의지를시장에인식시키는한편기업의투자의욕고취로인한일자리창출및설비투자확대등경제활성화에기여할것으로기대하였다. 우리나라대기업들은미국서브프라임모기지사태로촉발된금융위기에도불구하고수익성, 성장성측면에서양호한실적을시현하였다. 이렇게창출된이익의일부는배당, 세금등으로사외유출되지만대부분자산증가의형태로사내에유보된다. 일부언론, 국회, 시민단체에서는높은유보율을근거로동사내유보자금을설비투자, 고용에활용해야한다는주장을제기하고있으나유보율은단지자본잉여금과이익잉여금의합을납입자본금으로나눈수치에불과하므로안정적인수익창출이가능한기업이라면유보율상승은불가피하다. 사내에자본잉여금, 이익잉여금의형태로유보된자금은총자산중현금성자산, 투자자산, 유 무형자산등으로배분되기때문에모두현금으로보유하고있다는일각의주장은근거가미약하다. 다만총자산증가속도에비해유형자산의증가속도가더딘것은대기업들이창출된수익의자산분배시설비투자비중이여타자산에비해작다는것으로보여진다. 이는금융위기에따른국내외경기에대한불확실성확대, 경쟁심화로미래수종사업에대한과감한투자보다는단기실적위주의경영에주력한결과라할수있다. 동제도의폐지이후대기업집단의계열사수는증가하고있으며대기업의중소기업고유업종으로의사업확장도늘어나는가운데내부지분율은상승하고계열사간내부거래비중도높은것으로나타났다. 정부가출자총액제한제도를폐지하며기대했던설비투자확대를통한경제활성화는미진한가운데대기업집단으로의경제력집중이심화되고있다. 정부는출자총액제한제도폐지를보완하기위해대기업집단의공시의무를강화하고동반성장위원회를설치하여대기업과중소기업간의격차해소를위해노력하고있으나가시적인효과는미미한실정이다. 28 金融硏究제 26 권제 4 호 2012

본연구는대표적인기업규제였던출자총액제한제도의의미와실효성을확인하고동제도가대기업집단의총자산의구성변화에미친영향을살펴보았다. 대기업집단의총자산구성비중추이를살펴보는한편, 출자총액제한제도폐지의효과를계량적방법으로분석하였다. 즉, 자산구성의주요요소인현금성자산, 계열사출자로대표되는투자자산, 그리고유형자산투자에어떠한요인들이영향을주고있는지를추정하는한편출자총액제한제도폐지변수를추가분석함으로써재벌정책의핵심으로유지해온출자총액제한제도의의미를되새겨향후의정책적시사점을도출해보고자한다. 제 Ⅱ장에서는출자총액제한제도의의의와변천과정, 동제도폐지이후의대기업집단으로의경제력집중현황을정리하고, 제 Ⅲ장에서는출자총액제한제도등과관련된그간의선행연구및이론적배경을살펴본다. 제 Ⅳ장에서는실증분석을위한모형의설계및변수등연구방법을설명하고, 제 Ⅴ장에서는실증분석을시도하며, 제 Ⅵ장에서는연구결과를요약하고정책적시사점을제시한다. Ⅱ. 출자총액제한제도의의의와변천 1. 출자총액제한제도의의의 출자총액제한제도는자산총액이일정규모이상인기업집단에속하는계열회사가당해회사순자산액 1) 의일정규모를초과하여다른국내회사주식을취득또는소유하지못하도록규제하며, 출자에대한사전적상한 (ceiling) 을두어출자한도액을초과하여다른국내회사의주식을취득 소유할경우공정거래위원회는그주식에대하여의결권행사를금지할수있다. 동제도는상호출자금지 2) 만으로는규제가어려운순환출자등간접적인상호출자에의한계열확장을억제하기위해도입되었다. 동제도는국내계열사의자산합계액이기준금액 3) 이상인기업집단의계열사에 1) 자본총계또는자본금중큰금액에서계열회사에대한출자액을제외한금액. 2) 특정기업집단에속하는국내회사들의자산총액합계액이 5조원이상인경우동집단소속회사가자기주식을취득또는소유하고있는계열회사의주식을취득또는소유하는것을금지하여실질적인출자없이자본금을가공으로늘리거나계열기업을확장하는것을막기위한제도 ( 공정거래법제9조 ). 3) 1986년 12월 4천억원 2000년 4월 30대기업집단기준 2002년 4월 5조원 2004년 12월 6조원 2007년 4월 10조원 2007년 7월자산규모 2조원미만인소속회사는적용제외. 출자총액제한제도가대기업자산구성에미치는효과에관한연구 29

적용되며, 최소한의합리적인규제를위하여국민경제적비중이큰기업집단소속회사를대상으로하였다. 또한기준금액이상인기업집단소속회사중에서금융 보험회사, 지주회사및워크아웃등관리절차가진행중인회사등은대상에서제외되었다. 기준금액이상의자산규모를가진독립기업이나, 계열사자산합계액이기준금액미만인기업집단도적용대상이아니며, 금융업만을영위하는기업집단은계열사의자산합계가기준금액이상이더라도적용되지않았다. 즉공정거래위원회는계열사간출자를통해형성된가공자본에의한무분별한지배력확장및소유 지배구조왜곡심화억제, 출자관계로형성된집단의힘을남용하는데따른독립중소 중견기업과의불공정경쟁차단, 계열사간복잡한출자고리에의한기업집단의동반부실화및금융시장마비를초래하는시스템리스크방지를위해출자총액제한제도를도입하였다. 4) 2. 출자총액제한제도의변천 5) 독점규제법의제정당시 (1980년) 에는대기업집단의경제력집중억제를위한규제가도입되지않았으나, 1986년 12월에출자총액제한제도가도입되었고 1987년 4월부터시행되었다. 출자총액제한제도는경제력집중억제시책으로상호출자금지, 금융 보험회사의결권제한, 지주회사설립금지와함께도입되었으며, 동제도의도입초기목적은직접적상호출자규제만으로는순환출자에의한경제력집중심화를억제할수없어이러한순환출자를간접적으로제한하는데있었다. 동제도도입당시의규제대상은자산총액 4,000억원이상기업집단소속계열사였으며, 출자한도는순자산의 40% 로설정되었으나, 이후대기업집단문제의핵심이소유 지배권집중에있다는인식이반영되면서 1994년 12월 4차독점규제법개정을통하여 1995년 4월부터출자한도를순자산의 25% 로낮추었다. 그러나 1997년말외환위기이후외국인의적대적 M&A허용등기업경영여건의변화와국내기업의구조조정등정책필요에따라비상경제대책위원회의결정으로 1998년 2월출자총액제한제도가폐지되었다. 6) 그러나제도폐지이후우려했던 4) 1986년 12월공정거래법개정취지문은다음과같이적시하고있다. 대규모기업집단에속하는회사에대하여다른회사에대한출자총액을일정수준이하로제한하고동일한기업집단에속하는회사간에는상호출자를금지하는등의제도를도입하여기업으로하여금무리한기업확장보다는내실있는기업성장에주력하도록유도함으로써대규모기업집단의과도한경제력집중을억제하고국민경제의활력및균형발전을도모함에있다. 5) 동제도변천내용은 < 부록 1> 에요약되어있다. 30 金融硏究제 26 권제 4 호 2012

외국인에의한적대적 M&A는발생하지않았고오히려대규모기업집단계열사들간의출자가큰폭으로늘어났다. 출자증가의주된요인은기업구조조정과정에서기업의부채비율을 200% 이하로낮추라는정부정책의요구에따라, 부채비율을낮추기위해부채를감축하거나또는자기자본을확충해야하는상황에서자기자본을늘리기위해이루어진유상증자에계열사가참여하게되면서, 계열사에의한내부지분율이크게높아졌으며대기업집단최대주주의소유권과경영권간의괴리가확대되었다. 공정거래위원회는출자총액제한제도폐지이후동일인 ( 총수 ) 이적은지분으로다수계열사를지배하는구조가심화되고, 일부계열사의부실이전체기업집단의동반부실을초래하는등선단식경영의폐해가발생하며, 실질적자기자본의증가가없는부채비율감축이이루어지고, 유상증자참여를통해부실계열사를지원하는등의문제가발생함에따라 1999년공정거래법을개정하여동제도를 2000년 1월부활시켰다. 이때출자총액은순자산의 25% 로제한하였다. 10대기업집단에대해서는대규모내부거래에대해이사회의결및공시제도를도입하였고, 2001년내부지분공시대상을 30대기업집단으로확대하였다. 2002년 1월공정거래법제10차개정에서는제도의적용대상을자산총액 5조원이상의기업집단소속계열사로변경하고, 기업의경쟁력강화와구조조정촉진을위해적용제외및예외인정범위를확대하였다. 기존가족중심으로선정된 30대기업집단제도를폐지하고자산규모 5조원이상인기업집단에대해출자총액제한을적용하였다. 이에따라가족이최대주주가아닌민영화된과거공기업이라도자산총액이 5조원이상인기업집단의경우출자총액제한제도를적용받게되었다. 2004년 12월, 출자총액제한제도대상기업집단을자산규모 5조에서 6조원이상으로올려대상기업을완화하였다. 또한, 네가지출자총액제한규정졸업제도를도입하여과도한지배력문제가없거나완화또는해소되었다고판단되는경우에졸업제도 7) 를적용하기로하였다. 사회간접자본 6) 1999년공정거래백서 는동제도의폐지배경에대하여 출자제한으로인해인수및합병, 자산거래, 사업교환등기업구조조정이제약되는한편, 기업지배구조개선, 금융감독강화, 외국인투자자유화, 적대적 M&A 허용등개혁추진으로차입금융에의한무분별한사업확장이어려워지고국내기업의경영권방어, 사업전략등이차별적으로제약된다는판단에따라출자총액제한을폐지하였다 고기술하였다. 7) (i) 내부견제시스템을잘갖춘지배구조모범기업, (ii) 계열회사간 3단계이상출자가없고, 계열회사수가 5개이하인기업집단, (iii) 소유지분율과의결지분율차이 ( 이하 괴리도 ) 가 25%p 이하이고, 의결권승수가 3.0 이하인기업집단등이출자총액제한제도로부터졸업할수있게됨. 공정거래법상 지배구조모범기업 이라함은다음 4가지중 3가지이상에해당되는기업을지칭한다. (i) 서면투표가능, (ii) 집중투표가능, (iii) 위원의수가 4인이상이고, 사외이사로만구성된내부거래위원회설치 운용, (iv) 출자총액제한제도가대기업자산구성에미치는효과에관한연구 31