Conceptual Issues Directory UMM :Data Elmu:jurnal:J-a:Journal of Economics and Business:Vol50.Issue4.July1998:

crises, these forces have recently played a diminished role in the business cycle [see Friedman and Kuttner 1994 and Emery 1996]. Also, it has been argued that changes in the supply of commercial paper and Treasury bills, unrelated to business cycle conditions, have played a more dominant role in driving the spread. An alternative explanation for the decline in the spread’s predictive power has its origins in the dramatic rise in liquidity of the paper market over the past two decades. As Stigum 1990 has observed: A commercial paper trader’s primary responsibility is to assist issuers, but he also has a second responsibility, to provide liquidity to investors. More than any other aspect of the commercial paper market, it is the secondary market that has, in recent years, been developed. [Stigum 1990, p. 1051] As the liquidity of commercial paper rises, it becomes a closer substitute to Treasury bills in investors’ portfolios. As a result, shocks to either the paper or bill market cause rates in both markets to change proportionately. Thus, the spread is less sensitive to changes in market conditions linked to the business cycle. Although this explanation has been discussed in the literature, it has not been formally analyzed. 2 The objective of this paper is to fill this void in the literature. We do so by constructing a theoretical model which shows precisely how the relationship between the spread and its determinants is conditional on the substitutability between paper and bills in investors’ portfolios. Our second contribution is empirical. Using two different variants of the trend ratio of bill to paper volumes outstanding as proxies for the relative liquidity of commercial paper, we show that the coefficients linking the spread to its determinants fall as the liquidity of commercial paper rises. Moreover, we show that the paper-bill spread contains a liquidity premium which became less sensitive to interest rate uncer- tainty during the 1980s. The paper is outlined as follows. The theoretical model and a discussion of previous work are presented in Section II. Data and measurement issues are discussed in Section III. The results are presented in Section IV, and Section V concludes the paper.

II. Conceptual Issues

This section constructs a reduced-form model for the paper-bill spread derived from general specifications for supply and demand in the paper and bill markets. Our objective is to show that the spread’s sensitivity to exogenous changes in supply and demand declines as substitutability between paper and bills rises. The remainder of the section links the general model to previous work and discusses competing theories about the breakdown in the spread’s predictive ability. The Model Supply and demand for Treasury bills are expressed, respectively, as: B t S 5 a z XBS t 2 a o z r B,t 1 « t BS , 1 B t D 5 b z XBD t 1 b o z r B,t 2 r P,t 2 r B,t 1 « t BD 2 2 For example, Kashyap et al. 1993 speculated that, “as the commercial paper market deepens, the price pressure generated by a given Fed tightening should decline.” 362 J. P. Ferderer et al. where r B,t and r P,t are the Treasury bill and commercial paper rates; XBS t and XBD t are vectors of exogenous variables which have a systematic impact on the supply and demand for bills; a and b are vectors of structural parameters; a , b and h are scalar parameters; « t BS and « t BD are zero-mean shocks to the supply and demand for bills. Supply and demand for commercial paper are given by: P t S 5 c z XPS t 2 c o z r P,t 1 « t PS , 3 P t D 5 d z XPD t 1 d o z r P,t 1 h~r P,t 2 r B,t 1 « t PD 4 where XPS t and XPD t are vectors of exogenous variables which impact the supply and demand for commercial paper; c and d are structural parameter vectors; c o and d o are scalar parameters; « t PS and « t PD are zero-mean shocks to the supply and demand for paper. 3 Although the demand curves are not explicitly linked to an optimization problem, it is easy to see how they could result from the wealth maximization of investors. For example, in a world with three assets—Treasury bills, commercial paper and money—the second terms on the right sides of equations 2 and 4 are the relative yields between bills and money equation 2 and paper and money equation 4 where money earns zero interest. Similarly, the third term on the right sides of equations 2 and 4 are relative yields between bills and paper. The key parameter of interest is h—the cross-elasticity between bills and paper. It reflects the willingness of investors to substitute between paper and bills. Low substitut- ability is reflected by small values of h. In this case, equilibrium is restored following a disturbance primarily by an interest rate adjustment in the market where the disturbance occurred. In contrast, large h implies high substitutability. In this case, equilibrium is restored following a shock to one market by yield adjustments in both markets. For example, an increase in the supply of paper causes the paper rate to rise so that investors are induced to hold less money and more paper. However, the initial rise in the paper rate is moderated by arbitrage across the paper and bill markets by investors who view these two assets as close substitutes. In practice, four factors make paper and bills imperfect substitutes: 1 income earned from paper is taxable while bills are tax-free; 2 paper is subject to default; 3 bills provide services e.g., banks can use bills to post margin, collateralize overnight repurchase agreements, and satisfy capital adequacy requirements not provided by paper, and 4 bills have been more liquid than paper. If any of these factors changes, h should take on different values. Solving for the reduced form equation for the paper-bill spread, we get 4 r P,t 2 r B,t 5 2 a z XBS t 1 b z XBD t 1 x z XPS t 2 d z XPD t 1 u t 5 where: a 5 a z ~c o 1 d o c . 0; b 5 b z ~c o 1 d o c . 0; x 5 c z ~a o 1 b o c . 0; 3 The shocks are assumed to be uncorrelated. This is a simplifying assumption and does not change the main results of the model. As we see in the empirical section below, variables used to measure the impact of a particular influence on the supply and demand in one market certainly embody information about other influences as well. 4 See Appendix 1 for the derivation. Increasing Liquidity of the Paper-Bill Spread 363 d 5 d z ~a o 1 b o c . 0; c 5 h~a o 1 b o 1 c o 1 d o 1 ~a o 1 b o ~c o 1 d o ; and the variance of the error term in the reduced form model for the spread is given by: var ~u t 5 S c o 1 d o c D 2 z var ~« t BD 1 var ~« t BS 1 S a o 1 b o c D 2 z var ~« t PD 1 var ~« t PS . 6 Equations 5 and 6 are used in the rest of this section to discuss previous work in this area and competing explanations for the breakdown in the spread’s predictive power. Previous Work What factors should be included in the vectors of exogenous variables discussed above? Previous work has focused on three factors which influence the demand-side of the paper and bill markets in a way that can account for the paper-bill spread’s ability to predict business cycle movements. 5 First, an expectation of reduced economic activity leads to the belief that paper issuers are less likely to service their debt and, as a result, investors reduce the ratio of paper to bills in their portfolios. Second, investors may decrease the ratio of paper to bills in their portfolios as the level of nominal interest rates rise to take advantage of the tax-free nature of interest income earned on bills. If tight monetary policy causes nominal interest rates to rise prior to and during recessions, this could explain the strong predictive performance of the paper-bill spread. Finally, the paper-bill spread may contain a liquidity premium which is related to the business cycle. In contrast to other factors, the liquidity premium has received little attention in the literature. 6 This is surprising given the large body of work in finance which shows that an asset’s liquidity has a significant effect on its expected return and, by implication, the spread between expected returns on assets of differing liquidity [see Amihud and Men- delson 1986, 1991; Shen and Starr 1992; Kamara 1994]. For example, Kamara 1994 has shown that the spread between yields on relatively illiquid Treasury notes and liquid bills of equal maturity increases when interest rate uncertainty rises. 7 This logic suggests that the paper-bill spread might also contain an uncertainty-driven liquidity premium that rises prior to and during recessions. 8 Two supply-side shocks linked to the business cycle may influence the paper-bill spread. First, tight monetary policy raises the paper-bill spread by forcing bank-borrowers into the paper market. The credit-crunch version of this hypothesis posits that tight policy produced disintermediation out of the banking industry when deposit rate ceilings were binding prior to the removal of Regulation Q [see Cook 1981 and Rowe 1986]. 5 Friedman and Kuttner 1993 provide an excellent taxonomy of these effects. 6 Friedman and Kuttner 1993 discuss the liquidity premium, but do not attempt to directly measure it. 7 He [Kamara 1994] argues that an asset’s liquidity premium is the product of the: a expected length of time it takes to transact in the asset, and b volatility of its price. The difference between note and bill liquidity premia is a function of the difference in the expected length of time to transact in each asset multiplied by asset-price uncertainty. As long as notes are less liquid, the note-bill yield spread is sensitive to interest-rate uncertainty. 8 Evans 1984 and Ferderer 1993 provided empirical evidence that interest-rate uncertainty has a depress- ing effect on economic activity. 364 J. P. Ferderer et al. Disintermediation raised paper rates relative to bill rates because funds from the banking sector flowed disproportionately into the bill market because of the lower minimum denominations of bills and the nonpecuniary services provided by bills i.e., their use in posting margin, collateralizing overnight repurchase agreements, and satisfying capital adequacy requirements limited arbitrage across the markets, thus preventing equalization of the rates. Moreover, disintermediation reduced the ability of banks to make loans and thus forced firms to issue more commercial paper. The simple imperfect-substitutability version of the monetary hypothesis suggests that all episodes of monetary tightening lead to higher costs of funds for banks, causing them to restrict the supply of andor raise the interest rate on bank loans [see Bernanke 1990; Kashyap et al. 1993]. Some borrowers respond by issuing more commercial paper and, as long as paper and bills are imperfect substitutes in investors’ portfolios, the paper-bill spread rises. A second supply-side shock which may affect the spread in anticipation of business cycles is a change in business inventories [see Friedman and Kuttner 1993; Calomiris et al. 1994]. The primary motive for commercial paper issuance by nonfinancial firms is the financing of inventory accumulation. As firms issue more commercial paper in response to rising inventories at business cycle peaks, upward pressure is put on the paper-bill spread. Explaining the Decline in Predictive Power Several papers have documented the recent deterioration in the spread’s predictive power [for example, see Hess and Porter 1993; Friedman and Kuttner 1994]. One explanation for the declining predictive power is that recent business cycle movements have been driven by shocks which do not have a strong influence on the paper and bill markets. For example, Friedman and Kuttner 1994 argued that the 1990 –1991 recession resulted from adverse supply and fiscal policy shocks which do not impact the paper-bill spread. Also, Emery 1996 observed that there has been a downward trend, dating back to the 1970s, in “point estimates of the impact of spread movements on economic activity” p. 1. Emery did not explore the cause of this parameter instability other than to point out that the Lucas Critique was at work. 9 A second possible explanation is that factors unrelated to the business cycle have played a larger role in driving the spread in recent years. For example, Friedman and Kuttner 1994 contended that idiosyncratic changes in paper and bill supplies— changes unrelated to business cycle conditions— caused the paper-bill spread to rise during the economic expansion of the late 1980s. 10 A third possible explanation for the declining predictive power of the spread is that the forces driving the business cycle have not changed, but that the ability of the spread to respond to these forces has diminished due to an increase in paper market liquidity. Conceptually, this can be seen by examining the reduced form parameters in equation 5. 9 Along similar lines, Hafer and Kutan 1992, and Thoma and Gray 1993 argued that the predictive power of the spread prior to the mid-1980s arose mainly from two outliers in the data: the period around the collapse of the Franklin National Bank in 1974 and 1980 when the Carter Credit controls were imposed. When these two periods are excluded, the predictive power of the spread falls considerably. 10 Aggregate net issuance of commercial paper rose dramatically during 1988 and the first half of 1989 in response to the corporate leverage movement, while net issuance of Treasury bills was approximately zero on average from 1987 to mid-1989. Increasing Liquidity of the Paper-Bill Spread 365 Note that shifts in market supply and demand affect the paper-bill spread as long as paper and bills are imperfect substitutes i.e., h , `. When the substitutability rises— due to increased liquidity of commercial paper—the spread becomes less sensitive to exogenous changes in supply and demand. 11 If this explanation is correct, we should observe that the reduced form parameters in equation 5 are systematically linked to paper market liquidity. 12 Equation 6 reveals an additional way to test the liquidity explanation for the breakdown in the spread’s predictive performance. Note that the error term variance falls when the substitutability between paper and bills rises. If the error term in equation 5 displays heteroskedasticity induced by changes in the liquidity of paper, there is additional evidence that the decline in the spread’s predictive power resulted from a reduction in its ability to reflect supply and demand shocks.

III. Data Preliminaries