تعامل سهام و اوراق قرضه در بازار: آیا مومنتوم اثر سرریز داشته است؟

This paper examines the interaction between momentum in the returns of equities and corporate bonds. We find that investment grade corporate bonds do not exhibit momentum at the three- to 12-month horizons. Instead, the evidence suggests that they exhibit reversals. However, significant evidence exists of a momentum spillover from equities to investment grade corporate bonds of the same firm. Firms earning high (low) equity returns over the previous year earn high (low) bond returns the following year. The spillover results are stronger among firms with lower-grade debt and higher equity trading volume, seem robust to various risk and liquidity controls, and hold even after controlling for past earnings surprises. In examining the source of the spillover, we find that the bond ratings of firms with positive (negative) equity momentum continue to improve (deteriorate) in the future, suggesting underreaction to the information in past equity prices about changing default risk is a likely source of the spillover effect. Overall, our results suggest that both equity and debt underreact to firm fundamentals, but past equity returns is a better proxy of firm fundamentals than past bond returns.

Jegadeesh and Titman (1993) show a pervasive momentum effect in stock returns at the three- to 12-month horizons. The momentum effect is robust across different time periods and stock markets, and it has proven difficult to reconcile with existing rational asset pricing models.1 Recent behavioral theories (see Barberis et al., 1998; Daniel et al., 1998; Hong and Stein, 1999) argue that momentum is the result of initial underreaction to private or public news on the part of naïve investors with biased expectations.2 These theories are a source of great controversy in the finance literature because, taken at face value, they present a challenge to market efficiency. This debate, however, has focused on equity markets with little attention paid to the possibility of momentum in other asset classes.3 Investigating other asset classes is of interest not only because they can provide important out-of-sample evidence unavailable in tests based on equities, but also because they can provide potential clues as to the sources of the observed momentum in equity markets due to the differences in the investor clientele and the information environment between the equity market and the other asset markets. In this paper, we evaluate the underreaction hypothesis and examine the robustness of the momentum phenomenon using corporate bonds. Corporate bonds provide a logical setting to examine these issues given that bonds and stocks are different claims to the same underlying operating cash flows and are affected by the same firm fundamentals. It is natural to ask, therefore, whether the momentum effect in stocks extends to bonds. In other words, do corporate bond prices also underreact to fundamental news? Furthermore, is information about firm fundamentals better reflected in equity prices or bond prices? Thus, which is a better proxy of fundamental news, recent change in stock prices or recent change in bond prices? The answer to these questions is not obvious because significant differences exist between the stock and the bond markets in terms of their investor clientele and information environment. The stock market contains more individual investors than the bond market. The bond market tends to be dominated by large institutional investors who are presumably more sophisticated and better informed than the individual investors. If the psychological biases causing underreaction are more applicable to individual investors than institutions, then bond prices should reflect firm fundamentals better, adjust to common firm information faster, and exhibit less underreaction. Alternatively, the corporate bond market might be less informationally efficient than the equity market because much more attention is paid to stocks compared with the bonds of the same firm. For instance, there are many more equity analysts than bond analysts. Sell side equity research is widely disseminated among buy-side firms and in the popular media. In contrast, research on bonds is mostly limited to the firms covered by rating agencies such as Moody's and Standard & Poor's (S&P). These ratings agencies do not revise a firm's bond ratings as often as an equity analyst might change her forecasts and recommendations. Stocks are also more actively traded than corporate bonds, which are often held till maturity, and privately informed investors might therefore prefer to trade in the stock market where they can profit at the expense of unsophisticated individual investors. Under this scenario, the stock prices should adjust to common firm information faster and exhibit less underreaction. In examining how corporate bond prices adjust to fundamental news, we first examine whether corporate bond returns exhibit momentum. Standard momentum tests cannot, however, provide unambiguous inferences on differences in speed of adjustment to common information across two assets.4 In other words, while the presence or lack of momentum can reveal whether bond prices underreact to the information in past bond returns, it cannot tell us whether bonds adjust to common firm information more slowly or quickly than do stocks.5 This is because past bond returns might not be a good proxy for firm fundamentals. Therefore, we consider an alternate set of tests that focus on the lead-lag predictability patterns between stocks and bonds at intermediate horizons, and we use these tests to draw our primary conclusions on differences in speed of adjustment between stocks and bonds. Our empirical tests use portfolio strategies in which we form portfolios based on past stock and bond returns, and then compute future stock and bond returns of these portfolios, as well as Fama-MacBeth cross-sectional regression tests. Our key results are as follows. First, we find no evidence of momentum among investment grade corporate bonds.6 Instead, we find evidence of significant reversals. For instance, based on a six-month ranking period, winners (R10) underperform losers (R1) by 45 basis points in the first month after the portfolio formation date and by additional 12 basis points per month over the next six months. These results are robust in various subsamples designed to alleviate concerns about possible data errors or illiquidity. Furthermore, the reversals appear to be strongest among the most risky bonds in our sample (BBB-rated bonds), suggesting that the lack of momentum is not driven simply by the returns of less risky bonds. Still, we cannot entirely rule out data errors as the source of the observed reversal patterns. However, there clearly is no momentum in corporate bond returns. Thus, momentum does not appear to be a general property of asset returns and might be security-specific. Does the absence of momentum in corporate bonds imply that bonds adjust to common information faster than stocks? In other words, do bond returns predict stock returns? The answer is in the negative, as shown by our more direct tests examining the lead-lag relationship between stocks and bonds. The results from these tests show that stock returns predict bond returns, but not vice versa. Specifically, we find bonds of past six-month equity momentum winners outperform bonds of past six-month equity momentum losers by 21 basis points in the first month after the portfolio formation date and by an additional 66 basis points over the subsequent six months. We refer to this as the equity momentum spillover. Our results show that the spillover effect is as economically significant as the momentum effect in equities. For instance, at the six-month holding period (skipping the first month after the portfolio formation date), we find that the annualized Sharpe ratio of a zero-investment strategy that is long the bonds of firms with equity winners and short the bonds of firms with equity losers is 0.72, while the Sharpe ratio for the corresponding zero-investment equity momentum portfolio is only 0.54. The spillover effect is also robust to various liquidity and risk adjustments, is concentrated among firms with lower grade debt, and is stronger among firms with high equity trading volume. One concern about our finding is that, because bond returns and equity returns are contemporaneously correlated, the spillover effect is merely a restatement of the equity momentum effect, which could cause bond prices to rise as stock prices increase in response to past stock price increases. However, using cross-sectional regression tests that control for the firm-specific contemporaneous correlation between bonds and stocks, we show that the spillover effect is not merely a restatement of stock momentum. Overall our results show that, while both stock and bond prices underreact to the common firm information in past stock returns, stock prices adjust to this information more quickly than do bond prices. In conjunction with the finding that bond prices do not underreact to the information in past bond returns, these results suggest that past stock returns is a better proxy of firm fundamentals than past bond returns. Additional tests controlling for past quarterly earnings surprises, which are a more direct proxy of firm fundamentals, show that past equity returns continue to predict future bond returns suggesting that past equity returns are still a good proxy of firm fundamentals in this context. Earnings surprises, by contrast, are able to predict future bond returns only in the absence of past equity returns. In investigating the sources of the spillover effect, we find that the spillover effect is related to future changes in bond ratings. In particular, during the first year after the portfolio formation date, the winner portfolio contains more upgrades and fewer downgrades than do the loser portfolio. In other words, high stock returns in the past predict better bond ratings in the future. This suggests that the ratings agencies could be reacting sluggishly to information in past equity prices about changing default risk of bonds. This does not necessarily imply that bond investors also underreact to the information in past equity returns. If rating agencies underreact, but investors do not, then past equity returns would predict ratings revisions, but not bond returns. The fact that past equity returns predict both ratings revisions and bond returns suggests that there might be underreaction on the part of both groups. Consistent with this view, we find that the spillover effect is diminished when we include future revisions in bond ratings as a control variable in our cross-sectional regressions. In summary, we reach two major findings. First, there is no evidence of momentum in corporate bonds, suggesting that momentum perhaps is not a general property of asset returns and might be security-specific. Secondly, corporate bonds underreact to information in past equity prices about changing default risk, suggesting that the stock market tends to be the primary place of information discovery. These findings provide mixed support for existing behavioral theories. On the one hand, support exists for the view that security prices underreact to information, although the kind of underreaction we show is more complex, involving multiple asset classes, than that modeled by existing behavioral theories. On the other hand, the lack of momentum in corporate bonds appears inconsistent with existing behavioral theories, which do not explicitly distinguish between various risky assets. The rest of the paper is organized as follows. Section 2 describes the data used in the study and the methodology employed when forming portfolios, Section 3 examines the evidence on bond momentum strategies, and Section 4 examines the lead-lag interactions between stock and bond returns of the same firm. Section 5 examines the relation between the equity momentum spillover effect and revisions in bond ratings. Section 6 concludes.

This paper finds that while no momentum exists among investment grade corporate bonds, there is a momentum spillover effect from past equity returns to future bond returns. This spillover effect is statistically and economically significant, and is robust to various risk and liquidity controls. Our results suggest that while both stock and bond prices underreact to information about firm fundamentals, past stock returns are a better proxy of such firm fundamentals than past bond returns. As a result, bond prices underreact to the information in past stock returns, but not to the information in past bond returns. The spillover effect is related to predictable changes in bond ratings conditional on past equity momentum. Winners experience more upgrades and fewer downgrades than do losers in the first year after portfolio formation, and future ratings revisions partly explain the spillover effect in cross-sectional regression tests. These results suggest that bond prices underreact to past changes in equity prices because bond ratings react sluggishly to past changes in equity prices. What are the implications of our findings for recent behavioral theories? While our spillover results are generally consistent with security price underreaction, the kind of underreaction we document is more complex, involving multiple asset classes than modeled by existing behavioral theories. In addition, the lack of univariate momentum in corporate bond returns is troublesome for existing behavioral theories because, in general, they do not distinguish among different risky assets. (An exception is the model of Barberis et al. (1998), which applies specifically to equities.) A potential explanation for these differences might lie in the differences in the investor clienteles between the stock and the bond market. What our results do point to is the need for more elaborate behavioral asset pricing models involving multiple assets. Our results also pose a challenge to rational asset pricing models because our results suggest that the bond market does not pay sufficient attention to the information in stock prices. Bond market traders might be prevented from exploiting this effect by constraints to arbitrage in the sense of Shleifer and Vishny (1997) involving, among other things, risk aversion, short investment horizons, transaction costs, short-sale constraints. There could also be the issue of imperfect information of not being aware of the predictability pattern. Therefore, some learning could be involved as well.