1、 外文翻译 A Note on Credit Risk of Vertical Keiretsu Firms: Preliminary Evidence from the Japanese Automobile Industry Material Source: Asia-Pacific Financial Markets Author: NAOYA TAKEZAWA, NOBUYA TAKEZAWA Abstract. This paper empirically examines the relationship between the credit risk of Toyota, Nis
2、san and Honda keiretsu-affiliated firms and the credit risk of the respective parent company. As credit spread data for keiretsu-affiliated firms were not available we create a keiretsu default index, as a proxy, using expected default probabilities obtained from the KMV and Leland and Toft(J.Financ
3、e 51,9871019,1996)option pricing models. We find parent credit spreads do not Granger cause our keiretsu default index and vice versa in a divaricated vector autoregressive (VAR) framework. 1. Introduction Much of the theoretical and empirical literature on Japanese business groups has Focused on is
4、sues related to the horizontal or main bank centered keiretsu (Hoshi and Kashyap (2001) among others).Yet one could argue much that the strength of the Japanese industrial economy is in the technological innovation of firms tied to the vertical keiretsu system (i.e.manufacturing-supplier,supply chai
5、n oriented keiretsu)(Lincoln et al.,1998;Namiki,1999).The recession in the 1990s placed stress on the vertical keiretsu system to the extent that many keiretsu-linked companies face financial distress problems and have been party to restructuring programs. The Nissan revitalization plan is one such
6、example of a corporate restructuring plan affecting the keiretsu-affiliated firms directly (Economist, 2000).In a recent study, Packer (1999) finds that yield spreads on corporate bonds for keiretsu-affiliated Firms were higher than non-keiretsu firms in 1999.In their 1997 sample, however, The mean
7、yield spread is higher for non-keiretsu firms indicating a reversal in credit ratings in the late 1990s.This suggests that keiretsu-affiliated firms have higher credit risk on average in recent years. In this paper we focus our attention on the Japanese automobile industry and ask to what extent is
8、the credit risk inherent in the keiretsu-affiliated firm related to the credit risk of the parent company. This can be addressed empirically by examining the relationship between the credit spread of the parent firm and the credit spread of the keiretsu firm. The credit spread, a benchmark corporate
9、 bond yield less than the benchmark government bond yield, is available for our sample of parent auto manufacturing firms.Unfortunately,as credit spread data for keiretsu-affiliated firms were not available we need to establish a proxy. To this end, we create a default risk index for publicly listed
10、 keiretsu-affiliated companies. Our default index provides us with a benchmark for the credit risk of the keiretsu as a whole. The index itself could provide useful information for banks making loans and investors in assessing the quality of securities issued by affiliated firms within each vertical
11、 keiretsu group. The default risk index for keiretsu firms first involves using an option pricing-based approach to obtain the expected default probabilities for the vertical keiretsu firms. A weighted average of these expected default probabilities is our keiretsu default index. In this paper we ad
12、opt the option pricing methodology as developed by KMV.1 and for comparative purposes; we also compute the expected default probability using the Leland and Toft (1996) Model. This research note hopes to shed some light on the relationship between the credit risk of parent companies and their keiret
13、su-affiliated firms in the context of the Japanese automobile industry. The need to diversify credit risk for keiretsu firms is expected to increase with the development of the credit market and as keiretsu firms issue securities to finance technological innovations.Furthermore,the market has introd
14、uced indices for convertibles(such as the Goldman Sachs,Bloomberg convertible index etc.)Which contain Toyota and Nissan securities and on the equity side, the Nagoya Stock Exchange recently listed the Toyota Group Fund which weights 50%in Toyota Motor stock and the reminder in 20 firms strongly aff
15、iliated with Toyota motors. As a consequence we foresee a greater need to gauge corporate credit risk for such basket type products and possibly hedge such credit risk. 2. KMV Credit Risk Model The KMV Credit Risk Model estimates the firms credit risk based on the market value of assets.2 Cossin and
16、 Priotte (2001), Duffie and Singleton (2003), and Ong (1999) among others provide an overview of the related methodology. This option-based approach has been widely adopted in academic research to measure the credit risk of financial institutions, for example.Moridaira (2000) among others provides a
17、n overview of empirical work covering Japanese financial institutions. In another interesting piece, Hung and Takehara (2002) empirically investigate the relationship between default probabilities and macroeconomic indicators over time for the firms listed on the first section of the Tokyo Stock Exc
18、hange. The intuitio n behind this approach to estimating default probabilities is quite simple. When the firm defaults, debt holders assume seniority and are paid from the assets of the defaulting firm prior to any form of distribution to the shareholders. As a consequence, the amount the shareholde
19、r can claim is the difference between the market value of assets and debt. If the market value of assets, MVA, is below the value of debt, D, the shareholder receives nothing.Therefore, the payoff to the shareholder or market value of equity, MVE, is similar to that of a call option: M VE=maxM VA D,
20、0. We will assume that the dynamics of(change in)the market value of assets is governed by the following geometric Brownian motion with drift d M VA=rWACC M VA dt+A M VA d z (1) Where rWACC is the expected return on the assets, dt is the time increment, and dz is the change in the Wiener process.The
21、n, the market value of equity MVE, can be defined in the context of the BlackScholes Model for a call option as MVE=MVAN(d1) erWACCD N(d2) (2) d1= t )r(lnA 2A2WA C C tDMV A (3) d2=d1 A t (4) Where A is the asset volatility, rWACC is the cost of capital, t is the time to maturity, And N (.)is the cum
22、ulative normal distribution. Note the valuation of the option is based on the trading of an asset that grows at the firms cost of capital (aggregate growth rate, i.e.WACC).Under the risk neutral measure, the option pricing procedure involves trading a cash asset instead of corporate assets that grow
23、 at the firms cost of capital or WACC.The existence of the option price is assured by a change of numeraire.In other words, we treat part of the corporate growth as non-systematic risk associated with the firm. Although this equity premium risk may be diversified through the capital market, we assum
24、e this risk to be corporate specific and assume that the option is priced based on trading a“typical”project of the firm as opposed to cash assets.This“typical”project will yield a growth rate consistent with the cost of capital of the firm. This becomes critical in our study because we are assuming
25、 individual firms are measuring their growth activity based on the relative performance of the typical firm project rather than monetary units based on cash assets, for example. The firms asset volatility is estimated from the observed equity volatility of the firm as the following, A=EAEEDA DEDADEA
26、E MVMVMV MVMVMVMVMVMV 2)()( 22Where A is the volatility of assets, E is the volatility of equity; D is the volatility of debt, and ED the covariance between equity and debt. We assume that the volatility of debt, D, is zero, and that debt and equity are uncorrelated, i.e.ED=0. The major objective in
27、 applying the KMV Credit Risk Model is to estimate the expected default probability of a firm. The expected default probability, EDP, or The conditional default probability empirically obtained in this paper is defined as Pr(M VA,TDT|M VA,0)=Pr(lnM VA,TlnDT|lnM VA,0) 1 N t trDMVAAW ACCTA )2/()/0( 2,
28、ln The subscripts 0 and T refer to time today and a future point in time, respectively. T is assumed to be 1 year3 to match the accounting period applied for most of the Financial data. Thus, EDP=Expected default probability=1-N (d2) (6) Where d2 is defined in Equation (4).This provides us with the
29、probability that the market value of assets will fall below the current debt level over a specified time interval (1 year in this paper).The implied market value is obtained by using the BlackScholes Model and then substituted back into Equation (4) to obtain N (d2). 6. Keiretsu Default Probability
30、Index For purposes of this paper, keiretsu affiliation is simply determined by the notes In Nihon no Kigyo group published annually by Toyo Keizai.The notes indicate whether the firm is a Toyota keiretsu firm, a Nissan keiretsu firm, a Honda keiretsu Firm, or an independent. These firms are then cro
31、ss-checked to see whether the Parent company is one of the main corporate shareholders (equity ownership). Furthermore, we check to see whether any member(s) of the board is directly Associated with the parent company. We find the keiretsu classification scheme to be consistent with both the equity
32、ownership and board of director classification criteria. Finally, a limited number of firms in the sample provide subsidiary to parent sales data (sales to Toyota, Nissan and Honda). We started with a total sample of 69 keirestu firms for Nissan, 75 keiretsu firms for Toyota and 10 keiretsu firms fo
33、r Honda listed on at least one stock exchange in Japan. Several firms were eliminated from the sample due to their line of business. Toyota Tsusho (trading company), for example, is deleted as the main line of business of the firm is not directly related to producing automobile parts. Other firms ar
34、e also deleted due to a lack of stock price data as discussed in Section 4. Over our 7-year time frame, several firms were acquired, other firms exited from the keiretsu, or new firms joined the keiretsu and thus the number of firms in the Keiretsu index differed year by year. For Nissan, the number
35、s ranged from a low of 5 in 2003 to a high of 22 firms in 19971999.For the Toyota group we had low of 11 firms in 2000 and high of 22 firms in 20012002.For Honda, the number ranged from 3 in 1997 to 7 in 2003.The data is collected for each firm over a 7-year period from 1997 to 2003. We created an i
36、ndex, which averages the expected default probability, EDP, across all firms associated with a particular keiretsu.The EDP are weighted by the strength of affiliation to the keiretsu in question. As a proxy for the strength of affiliation, we rely on the equity ownership by the parent company. An al
37、ternative approach would be the use of sales transaction data within the keiretsu, however, this proved to be difficult due to data limitations. To explore the predictive ability further we run Granger causality tests and find the default indices do not Granger cause credit spreads (Table VI). Table
38、1: Granger causality tests results for the Nissan, Toyota and Honda keiretsu Null hypothesis: F-statistic Probability Credit spread and Nissan Keiretsu index Granger causality test KMVNissan does not Granger Cause CSNissan 0.470 0.637 CSNissan does not Granger Cause KMVNissan 0.781 0.482 Credit spre
39、ad and Toyota Keiretsu index Granger causality test KMVToyota does not Granger Cause CSToyota 1.849 0.163 CSToyota does not Granger Cause KMVToyota 0.846 0.433 Credit spread and Honda Keiretsu index Granger causality test KMVHonda does not Granger Cause CSHonda 5.303 0.012 CSHonda does not Granger C
40、ause KMVHonda 0.136 0.874 These results hold for the Nissan and Honda keiretsu as well. The reverse hypothesis that credit spreads do not Granger causes the default indices cannot be rejected as well as at conventional levels of statistical significance. Consistent with our finding using correlation
41、s, we find that there is a weak relationship between the default indices of keiretsu firms and credit spreads of the parent firm at best. As a consequence, the credit spreads of the parent firm may not reflect the default risk of affiliated keiretsu firms. Or it could be that investors in parent com
42、pany corporate debt do not perceive the risk of keiretsu firms as affecting the risk profile of the parent firm(as reflected in the credit spread). 8. Concluding Remarks This note had the modest objective of empirically documenting the relationship between the credit risk of a parent company and the
43、 credit risk of keiretsu-affiliated firms. As credit spreads for the keiretsu-affiliated firms were not available we created a default index for publicly listed Toyota, Honda, and Nissan keiretsu firms. The default probabilities for listed firms affiliated with the Toyota, Honda and Nissan keiretsu
44、were obtained via a modified BlackScholes Model developed by KMV and the LelandToft Model. We find that our default index is not correlated with the parent company credit spreads.Furthermore, in a bivariate VAR framework, the default indices do not Granger cause the credit spreads and vice versa at
45、conventional statistical significance levels. In other words, we find little empirical evidence, with our limited data set, suggesting a relationship between parent company credits spreads and the default risk of keiretsu-affiliated firms. Investors in debt instruments issued by keiretsu-affiliated
46、firms or equity related products such as the Toyota Motor fund traded on the Nagoya Exchange could hedge the credit risk of the keiretsu-affiliated firm by taking positions in parent company related securities. For example, one possible hedge could involve using credit default swaps on the parent co
47、mpany. Another possibility is a put option written on the parent company stock or a synthetic equivalent. Such hedges assume a positive empirical relationship between the credit risk of the parent firm and the credit risk of the keiretsu-affiliated firm. We were unable to uncover such an empirical r
48、elationship.Thus; our preliminary empirical findings imply that it may be difficult to hedge the credit risk associated with keiretsu-affiliated firms using parent company credit instruments or related securities. Future research will be directed to extending the time domain of the empirical analysis, and improving the econometric analysis. We also intend to expand the data set to include non-listed keiretsu firms and eventually link t
