بازده معکوس در بازار اوراق قرضه: شواهد و دلایل

The finance literature has shown that equity returns are predictable using past returns. This study extends that literature by examining bond return predictability. Using returns constructed from dealer bid prices, we find short- to intermediate-term reversals in investment grade corporate bond returns. These reversals are larger in the first half of the sample period and consistent with the predictions of dealer inventory cost models. This supports Jegadeesh and Titman’s [J. Financ. Intermed. 4 (1995) 116] assertion that daily, weekly, and monthly reversals in equity returns come from dealer inventory considerations, not behavioral biases. Finally, unlike equity returns, we find no evidence of momentum in bond returns.

The predictability of equity returns based on past returns is a popular subject in both the academic literature and the investment community (see Dremen, 1977 and Dremen, 1979). Empirical research shows evidence of reversals over short horizons (1 week to 3 months); momentum over intermediate horizons (3–12 months); and reversals again over long horizons (3–5 years). Though the existence of these patterns in equity returns is well established, the explanation for why these patterns exist remains controversial. Many argue that investor irrationality is the primary cause and offer behavioral models consistent with the empirical findings. Daniel, Hirshleifer, and Subrahmanyam (DHS, 1998) and Odean (1998) are examples. Others argue for rational explanations that either take issue with empirical methodologies as in Conrad and Kaul (1998) or develop rational models consistent with the phenomena as in Berk et al. (1999). Still others differentiate the short horizon from the longer horizon patterns and argue these are separate phenomenon with distinct causes. For example, Lehman (1990) attributes the short horizon reversals to investor overreaction; Roll (1984) points to measurement issues such as bid–ask bounce; and Lo and MacKinlay (1990) suggest the cause is cross-autocorrelation across securities. Finally, Stoll (1989) and Jegadeesh and Titman (1995) show that dealer responses to inventory imbalances can cause negative serial correlation in returns and may be responsible for reversals in daily, weekly, and monthly equity returns. The purpose of this paper is to expand the scope of this research by examining the predictability of bond returns based on past returns by looking at the intertemporal bid pricing of Treasury and corporate bonds in the dealer market. There are two main reasons why we might find patterns in bond returns. First, the bond market is largely a dealer market. Hence, the dealer inventory cost models of Stoll (1978), Amihud and Mendelson (1980), and Ho and Stoll (1981) should apply. These models predict negative serial correlation in prices as dealers adjust bid–ask spreads relative to the “true” price to encourage transactions that will even out their inventory positions. Although these models are usually applied to equity data of a daily or weekly frequency, Jegadeesh and Titman (1995) show these models have implications for longer horizon equity returns as well. They provide evidence equity market dealers on the NYSE take several days to balance out their inventory positions and suggest that the monthly reversals in equity returns found by Jegadeesh (1990) are due to this phenomenon. If true, reversals in bond returns over longer horizons may be affected as well, since the corporate debt market is less liquid than the equity market, lengthening the time it takes for dealers to rebalance their inventories. A second reason we might find patterns in bond returns is found in behavioral models that predict patterns in risky asset returns due to investor irrationality. These models were crafted to explain patterns in equity returns, but imply we should find similar patterns in the returns of other risky assets. If the patterns in equity returns are due to overreaction, it is not unreasonable to expect similar patterns in bond returns. After all, if investors overreact to information in analyzing equity values, perhaps they do so in analyzing risky bonds as well, causing overreaction in the price adjustment behavior of dealers in the corporate bond market. This may makes sense given psychological evidence showing experts are more prone to overreact than others due to greater overconfidence (see Griffin and Tversky, 1992). In this study, we use dealer bid prices to calculate returns. The use of bid prices has advantages for our purpose because we eliminate the influence of bid–ask bounce and adverse information costs, which can cause spurious negative serial correlation in transaction prices. Also, because the bid quotes are taken at month-end, we avoid bias due to nonsynchronous trading, which can also cause spurious negative serial correlation. Based on monthly bid prices from 1978 to 1998, we find the existence of return reversals in investment grade corporate bonds in the short to intermediate horizons (1–6 months). These patterns differ from those found in equities, do not exist in Treasury bond returns, and are not likely due to data issues. To determine the cause(s) of these patterns, we perform two types of analyses. First, we determine the source of the reversals in terms of return components. If dealer inventory rebalancing is the cause, we would expect the reversals to be concentrated in the security-specific component of bond returns. Overreaction, on the other hand, could be consistent with either reversals in the factor component or the security-specific component of returns. We find the reversals are not caused by serial correlation in common bond risk factors; namely, changes in the term spread and default spread. Further, the lack of any patterns in Treasuries allows us to rule out yield curve effects as a contributing factor since changes in the yield curve would affect both Treasuries and corporate bonds in a similar fashion. Nor can the reversals be attributed to overreaction to changes in common risk factors or to lead–lag effects resulting from slower price reaction to a risk factor for some bonds versus others. Thus, the reversals appear to come from the security-specific component of bond returns, which is consistent with both overreaction and dealer inventory rebalancing. Second, we split the sample period into two halves as a means of distinguishing between overreaction and dealer inventory effects. We provide evidence of a significant increase in the liquidity of the corporate bond market between the first and second halves of the sample period, and show how this allows us to infer whether the reversals are caused by inventory rebalancing or overreaction. This is true because the magnitude of any reversals caused by inventory rebalancing is related to the size of bid–ask spreads. Thus, if dealer price adjustments due to inventory considerations are the source of the reversals, we should see a decline in the magnitude of reversals between the first and second halves of the sample period, as liquidity rises and bid–ask spreads shrink. If overreaction is the primary explanation, we should expect to find little difference between the two sub-periods. We find the reversals become largely insignificant in the second half of the sample period, which is consistent with the dealer inventory cost explanation. This finding that the bond reversals are due to dealer inventory costs supports Jegadeesh and Titman’s (1995) assertion that daily, weekly, and monthly reversals in equity returns are due to dealer inventory considerations, not overreaction. The difference between the reversals in the two markets is time horizon, which is driven by differences in liquidity. The rest of the paper is organized as follows. Section 2 summarizes how overreaction and dealer inventory cost can generate patterns in returns. Section 3 describes our sample. Section 4 presents the returns to winner–loser portfolios for corporate bonds. Section 5 discusses data issues. Section 6 presents a simple model of returns that allows us to decompose the sources of reversals. Section 7 splits the sample to differentiate between overreaction and dealer inventory adjustments as causes of the reversals. Section 8 concludes.

This study examines the predictability of bond returns based on past returns using dealer bid prices over 1978–1998. We find the existence of short- to intermediate-term reversals (1–6 months) in the corporate bond returns. These patterns are different from those found in equities and do not exist in Treasury bond returns. Further, the evidence indicates the return reversals for corporate bonds diminish at longer horizons (7–12 months) and are mainly confined to the first half of the sample period (1978–1987). We next examine the possible sources of the reversals and find they are not due to serial correlation in the common risk factors. They also cannot be attributed to overreaction to the default spread factor; lead–lag effects resulting from a delayed bond price reaction to the default spread factor for some bonds and not others; nor term structure effects. This leaves the firm-specific component of corporate bond returns as the source of the reversals, and overreaction or dealer inventory rebalancing as potential explanations. Our analysis of the first and second halves of the sample period combined with our analysis of the differences in liquidity between the two sub-periods provides evidence that the reversals come from dealer inventory effects, not investor overreaction. Thus, the results are unrelated to the intermediate and long-term patterns found in equity returns. However, the finding that the reversals are due to dealer inventory costs supports Jegadeesh and Titman’s (1995) assertion that daily, weekly, and monthly reversals in equity returns are due to dealer inventory considerations, not overreaction. The difference in the case of corporate bonds is that these microstructure effects exist in quarterly and semi-annual returns as well. Thus, our results serve as a warning that microstructure effects may be more severe for corporate bond returns than equity returns at the monthly horizons normally examined by researchers. This is especially true prior to the 1990s. Finally, it is interesting that there is no evidence of any momentum in the bond market, which contrasts with the equity market. This indicates the pricing processes in the two markets differ with respect to whatever causes intermediate momentum in the equity markets. If momentum in equity markets is caused by investor underreaction or overreaction, then clearly there is little evidence of such behavior in the bond market. This may indicate that marginal investors in the bond market are less prone to such behavioral biases and/or that there is simply less uncertainty and, hence, less room for error in valuing corporate bonds, which would reduce the potential for such biases to influence pricing. If momentum is due to cross-sectional differences in expected returns, then our duration quintiles have effectively controlled for such differences. Controlling for cross-sectional differences in expected returns in the equity markets is a considerably more difficult proposition.