PDF Ebook Bond Liquidity Estimation and the Liquidity Effect in Yield Spreads
We provide a model to estimate bond liquidity costs using only daily prices. Our liquidity estimates are 30% correlated with the bid-ask spread and are within five basis points of the bid-ask spread for investment grade bonds. Our median liquidity estimates are approximately $0.15 for investment grade bonds which compares well with those established by Schultz (2001) or Hong and Warga (2000). Regression tests indicate that our estimate of liquidity is associated with the bid-ask spread even after controlling for the commonly used liquidity determinants. Assessing the economic significance of our liquidity estimate in relation to the yield spread, we find that our estimate of liquidity is positively associated with the yield spread regardless of controlling for the commonly used yield spread determinants. These results imply that liquidity costs exact an demonstrable influence on bond returns and are consequently a priced element of the yield spread.
Corporate bond trading on exchange listed and over-the counter markets now exceeds $15 billion per day.1 Affecting these markets is the underlying liquidity that is of increasing concern to regulators, bond traders2 , and academicians. However, arguments concerning liquidity are often muted because of the complex problems of implementing liquidity aspects into empirical pricing analysis. Recent studies, Schultz (2001) and Hong and Warga (2000), highlight the obstacles in estimating the costs of trading corporate bonds, while Longstaff (2000) outlines the difficulties in incorporating liquidity into bond yield studies.3 While a comprehensive estimate of liquidity costs for corporate bonds is crucial for studying investment strategies and understanding bond yields, such an estimate is lacking in current empirical studies. This paper attempts to fill this void by presenting an empirical model to estimate bond liquidity and by assessing the economic significance of the relationship between liquidity and bond yields.
The crux of the problem in estimating liquidity and its affect on bond yields is the sheer lack of credible information on spread prices or bond quotes (Goodhart and O’Hara, 1997). Our goal is to provide a liquidity estimate that is estimable for any bond with daily prices, supplanting the need to obtain quotes or intraday trade-specific information. We extend the model of Lesmond, Ogden, Trzcinka (1999), which is applied to liquidity estimation of common equity, to the corporate bond market. We quantify our estimate with the underlying bid-ask spread and find that our liquidity estimate is 30% correlated with the bid-ask spread and is within five basis points of the bid-ask spread for investment grade bonds. Our median estimate for the investment grade trading cost, a measure most comparable to the more liquid bonds examined in the literature, is $0.15 per $100 value. This estimate compares well with the trading cost estimates of Hong and Warga (2000), Schultz (2001), and Chakravarty and Sarkar (1999), which range from $0.13 to $0.27 per $100 par value. We extend the analysis to demonstrate the economic significance of our estimates by showing that liquidity is priced in the yield spread.
At least two different approaches have been employed to estimate the transaction costs for corporate bonds. The first is to estimate the effective spread as the difference between the weighted average of the buy and sell prices during the day (Chakravarty and Sarkar, 1999, Hong and Warga, 2000), while the second is to invoke a methodology used by Schultz (2001) that regresses the difference between the trade price and the estimated contemporaneous bid quote on a dummy variable that is assigned a zero or one for sells or buys.
The effective spread method have two serious limitations. To be included in the sample, selected bonds must have at least one trade at the buy price and another at the sell price within the same day. This implies that only the most liquid bonds are used because this requirement removes non-traded bonds; and this potentially causes a non synchronous trading bias because trades at the bid or ask do not occur at the same time within the day. These biases are reflected in the relatively low explanatory power in their regressions of the estimated spreads over independent variables (e.g., the R-square is less than 2 percent in both Hong and Warga (2000) and Chakravarty and Sarkar (1999)). Schultz’s (2001) method, while more comprehensive and complete, must estimate trading costs based on monthly data to infer daily quotes by assuming price changes in corporate bonds are proportional to price changes in treasury bonds. This assumption is likely to hold only for investment grade bonds. In addition, this approach requires both traded and quoted bid prices.
To overcome these data limitations, this paper proposes an alternative method, termed the limited dependent variable (hereafter, LDV) model, that does not rely on the use of bid-ask quotes or spread prices. Rather, this method uses only closing daily prices to estimate liquidity costs by assuming that the trade prices reveal liquidity effects through the incidence of zero returns. Lesmond et al. (1999) and Lesmond (2002) find that this method works well for equity markets as evidenced by an 80% correlation between the LDV liquidity estimate and the bid-ask spread plus commissions.
Download
PDF Ebook Bond Liquidity Estimation and the Liquidity Effect in Yield Spreads
Posted in :
