ریسک های سرمایه گذاران اوراق قرضه ژاپن: آزمون فرضیه "رسیدن به عملکرد" در بازار اوراق قرضه ژاپن

This paper attempts to test the “reach for yields” hypothesis in the Japanese bond markets to explore the cause of extremely low credit spreads on Japanese bonds, especially BBB-rated bonds, using a three-factor CAPM (γ-CAPM) with (co)skewness as an additional market risk factor. Under the γ-CAPM, risk premium can be expressed as a weighted average of β-risk and γ-risk. Empirical results support the γ-CAPM against the β-CAPM. The estimated weight of γ-risk is 2.6 percent in Japan, compared with 12.5 percent in the United States. This difference mainly reflects a lower degree of relative risk aversion in Japan.

This paper attempts to test the “reach for yields” hypothesis in the Japanese bond markets to explore the Japanese version of the “credit spread puzzle.” To that end, we explicitly consider the risk stemming from negative (co)skewness under a three-factor CAPM framework. Very few empirical studies have been conducted so far about the pricing of Japanese bonds, corporate bonds in particular. This is due mainly to the facts that (i) the Japanese corporate bond market is a relatively new market, which started to develop after the implementation of major liberalization measures such as a removal of eligibility criteria in 1996, and (ii) the basic trend of price developments has been quite different between Japan and other major advanced countries like the United States.1 Regarding the latter point above, Fig. 1 compares the 5-year credit spreads between Japan and the United States as of the end of April 2005. Note here that the credit spreads on Japanese corporate bonds are significantly less than the U.S. credit spreads. In particular, the Japanese BBB-credit spread is much lower than the U.S. BBB-credit spread. The large spreads on corporate bonds in the United States relative to their historical default rates has been termed the “credit spread puzzle” by Amato and Remolona (2003). In Japan, contrary to the case in the United States, bond investors have witnessed extremely narrowed spreads particularly on BBB-rated corporate bonds in recent years. This might be called the Japanese version of the “credit spread puzzle.”2 Transactions in the secondary bond markets have been much less active in Japan than in the United States. This implies that liquidity in the Japanese bond markets is much lower and this is especially true in the Japanese corporate bond market. The lower liquidity further deepens the Japanese version of the “credit spread puzzle.” In this paper, we try to minimize the effects of the low liquidity in the Japanese corporate bond market by using the NIKKO-BOND Performance Index that excludes the issues with outstanding amount less than one billion yen. Full-size image (28 K) Fig. 1. Comparison of 5-year credit spreads between Japan and the United States. Notes. (1) Credit spread is defined as the difference in yields between corporate and government bonds with the same maturity. (2) The default rate is an annualized average of the historical default rate in the past 5 years. Sources: Japan Securities Dealers Association, Moody's, R&I, Merrill Lynch. Figure options Why are the “credit spread puzzles” in the opposite directions between Japan and the United States? In this regard, let us raise the following two hypotheses. First, the Japanese corporate bond market has so far witnessed only a few cases of publicly issued (straight) bond default including that of MYCAL Corporation in September 2001.3 Also, even in the case of a default, the so-called “main banks” covered major portions of potential loss. Although the Japanese main bank system has been falling apart since the middle of 1990s, Japanese investors seem to have rarely cared about the default risk, as suggested by a number of Japanese investors and credit analysts.4 Second, as stated by Baba et al. (2005), under the zero interest rate policy adopted by the Bank of Japan in February 1999, some of the Japanese government bond (JGB) investors have stepped into the investments in credit instruments to look for higher returns.5 Whether or not the degree of risk aversion of Japanese bond investors is so low as to cause them to engage in “reach for yields” activities has attracted a lot of attention from both market participants and central bankers. Theoretically, preceding studies that investigated corporate bond pricing can be classified into the following two categories. One is the structural approach initiated by Black and Scholes (1973) and Merton (1974). This approach views corporate liabilities as contingent claims on the value of the firms. The other approach is the reduced-form approach proposed by Jarrow and Turnbull (1995), Jarrow, Lando, and Turnbull (1997), and Duffie and Singleton (1999), among others. This approach uses a hazard rate that describes the process of default probabilities, which is not necessarily related to firms’ credit standing represented by accounting information. Both approaches assume a risk-neutral investor. Recent empirical studies on the U.S. credit spreads show that the structural factors are important, but can explain only a small portion of the credit spreads. For instance, Collin-Dufresne, Goldstein, and Martin (2001) show that structural factors can explain only a quarter of the changes in the U.S. credit spreads, indicating that systematic factors, common to the aggregate corporate bond market, have much more contribution to the changes in credit spreads.6 Also, Driessen (2004) reports that the longer the maturity, the larger the contribution of common factors to credit spreads, while the contribution of the tax effect and default event risk is almost similar across maturities.7 In analyzing Japanese corporate bond returns to explain the Japanese version of the “credit spread puzzle,” we use a CAPM framework rather than other more conventional methodologies to explore the “reach for yield” activities by investors. This is due mainly to the following two reasons. First, the conventional methodologies do not take into account the “reach for yield” aspect in which risk averseness of investors is likely to be the key to understanding very low risk premiums. Second, Japanese bond investors do not seem to have cared much about default risk since they experienced only a few cases of publicly issued bond defaults. More specifically, we use a CAPM framework with (co)skewness as an additional risk factor, which we call the γ-CAPM in this paper. By using the γ-CAPM instead of the conventional β-CAPM, we can extract the “market-wide” degree of relative risk aversion from bond return data. The γ-CAPM has been applied to stock returns particularly in times of crash in the U.S. stock market and emerging markets by Harvey and Siddique (2000), Hwang and Satchell (1999), and Lim (1989), among others. To the best of our knowledge, however, this paper is the first one to apply it to bond returns. 8 Generally, when bond returns (prices) decline (rise) to the extent that the room for further declines (rises) is limited, negative skewness tends to expand for bond returns. This observation can be made for most of the fixed income returns in a low interest rate environment, particularly in the recent Japanese very low interest rate environment.9,10 Also, negative skewness is likely to capture part of the downside (negative) event risk. We often observe that once the negative events (or news) happen to a specific bond issuer or to the overall market, returns on specific or overall corporate bonds tend to fall suddenly and substantially. Put differently, although the probability of such an event is small, once it happens, it tends to cause prompt and large corrections in market valuation. On the other hand, there is no chance for a comparably large gain due to the property of fixed income securities.11 Thus, explicitly considering (co)skewness risk under the CAPM framework has an additional advantage in that we can also capture negative event risk as part of market risk. Note here that to fully assess the effects of such price changes on investment behavior as sources of risk, it is better for us to use the holding-period returns rather than the yields or credit spreads for the following three reasons. First, holding returns respond more sensitively than yields to the price changes over the holding period.12 Second, although it is quite rare for investors to buy-and-hold their bonds until maturities, yields are calculated under the assumption that they do so. Instead, investors are usually subject to the practice of mark-to-market in each pre-specified period. Third, holding-period returns fit better with the CAPM framework than yields. To make it easier to evaluate the degree of relative risk aversion implied by the Japanese bond data, we also apply the same model to the comparable U.S. bond data. Since Japanese domestic investors occupy a substantial portion of the Japanese bond markets, comparison between Japanese and U.S. bond markets enables us to observe the difference in the degree of risk aversion between Japanese and global investors.13 This comparison might shed light on the difference in the directions of the “credit spread puzzles” between Japan and the United States. The rest of this paper is organized as follows. Section 2 derives the formula of the γ-CAPM with (co)skewness as an additional risk factor. Section 3 applies the γ-CAPM to both Japanese and U.S. bond data and discusses the empirical findings. Section 4 concludes the paper.

This paper has tested the “reach for yields” hypothesis in the Japanese bond markets to explore the Japanese version of the “credit spread puzzle” under the γ-CAPM framework considering the risk stemming from negative (co)skewness (γ-risk). Under the γ-CAPM, risk premium can be expressed as a weighted average of β-risk and γ-risk. We have also analyzed the U.S. bond returns under the same framework for comparison. Main findings can be summarized as follows. (i) GMM specification tests support the γ-CAPM against the β-CAPM both for Japanese and U.S. bond returns. (ii) The estimated values of the degree of relative risk aversion are significantly positive both for Japanese and U.S. bond returns and its estimated value is much lower for Japanese bond returns than U.S. bond returns. (iii) The weight of γ-risk is 2.6 percent in Japan, compared with 12.5 percent in the United States. Thus, γ-risk is priced in the U.S. bond market to a higher degree than in the Japanese bond market. These results suggest that Japanese investors do not seem to have cared much about γ-risk in a recent low interest rate environment. The above finding suggests that Japanese investors’ activeness for “reaching for yields” that are higher than the suppressed yields on government bonds, as measured by the degree of relative risk aversion, should be part of the explanation of extremely narrowed credit spreads of Japanese corporate bonds.