The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
http://clevelandfed.org/research/workpaper Best available copy Working Paper 8604 THE DEFAULT PREMIU:V A N D CORPORATE BOND EXPERIENCE By Jerome S. Fons Jerome S. Fons is a n economist at t h e Federal Reserve Bank of Cleveland. The author would like t o thank Jim Balazsy for excellent research assistance, Bill Gavin and Ross S t a r t for helpful comments, and t h e participants of t h e Federal Reserve System's Financial Analysis Committee Meeting at t h e Board of Governors in November 1985. Working Papers of t h e Federal Reserve Bank of Cleveland a r e preliminary materials, circulated t o stimulate discussion and critical comment. The views s t a t e d herein a r e t h e author's and not necessarily those of t h e Federal Reserve Bank of Cleveland or of t h e Board of Governors of t h e Federal Reserve System. June 1986 Federal Reserve Bank of Cleveland http://clevelandfed.org/research/workpaper Best available copy The Default Premium and Corporate Bond Experience I. Introduction The emergence of organized markets for low-rated corporate (or junk) bonds has provided financial researchers with an opportunity t o address a fundamental question: a r e holders of default- prone debt compensated (actuarially) f o r t h e risk of default? Past research on this market has focused on t h e default experience of corporate debt. A quite different a r e a of research involves modeling t h e spreads between t h e returns of bonds of different credit quality. Few (if any) research e f f o r t s have combined these approaches by using past default experience t o explain differential r a t e s of return on lowrated bonds. In this study we develop a risk-neutral model of t h e expected probability of default for low-grade bonds as a function of t h e additional required r a t e of return on these instruments over default- free bonds. Within this framework, securities a r e priced as functions of t h e first moment of t h e return distribution. The techniques a r e used t o express this pricing relationship in terms of t h e yields t o maturity of risky bonds as well as their holding period returns. W e then compare t h e default r a t e s implied in corporate bond yields t o a series based on r e c e n t corporate bond default experience. We also discuss why implied default r a t e s cannot b e obtained from measured holding period returns. Finally, attention is paid t o macroeconomic indicators of expected default rates. In a n early paper on t h e subject of default risk premia, Fisher (1959) suggested t h a t t h e risk premium required on a corporate bond (holding maturity constant) depends on t h e likelihood t h a t t h e issuing firm will default (defined here as a failure t o pay any coupon or principal payments when due) and on . http://clevelandfed.org/research/workpaper Best available copy t h e "marketability u of t h e bond. In addition, modern approaches acknowledge t h e influence on required returns t h a t result from call provisions, t h e t a x effect f o r deep discount bonds (due t o t h e d i f f e r e n t t a x treatment of o r d ~ n a r y Income vs. capital gains), and sinking fund payments (which reduce t h e avera g e maturlty of a firm's debt). Isolating t h e influence of default likelihood on interest- rate differentials involves controlling f o r these o t h e r effects. As Jones, Mason and Rosenfeld (1984) pointed out, studies t h a t a t t e m p t t o explain corporate bond prices c a n b e identified 3s being e i t h e r "macro N in nature, in t h a t relative bond prices a r e modeled as being "functions of t h e supply and demand of various assets, and/or t h e position of t h e economy in t h e business cycle," o r as being "microw in t h a t relative prices a r e modeled .as a function of firm specific characteristics. The approach taken in this paper i s macro; t h a t is, t h a t a g g r e g a t e returns on a sample of bonds a r e used t o infer a v e r a g e default probabilities for t h e population of bonds with similar characteristics. In order to test hypotheses concerning t h e models derived in this paper, w e assume t h a t bond market participants a r e in complete agreement as to t h e probability of d e f a u l t f o r a particular issue. W e f u r t h e r assume t h a t all bonds a r e perceived, and t h e r e f o r e priced, as having t h e same likelihood of default as o t h e r s in t h e same rating category. Ideally, t h e assigned rating gives, in a single measure, t h e rating agency's estimate of t h e issue's probability of default. Studies t h a t h a v e a t t e m p t e d to measure t h e importance of rating changes on bond price movements tend to differ in t h e i r conclusions. Hettenhouse and Satoris (1976) as well as Weinstein (1977) conclude t h a t market participants incorporate new information before a rerating. On t h e o t h e r 2 http://clevelandfed.org/research/workpaper Best available copy hand, Eder~ngton, Yawitz, and Roberts (1984) flnd t h a t t h e market responds t o rating changes in addition to publicly available information. They also conclude t h a t ratings by Moody's and Standard & Poor's a r e equally reliable indic a t o r s of an issue's creditworthiness. In order to isolate expected default probabilities, we will restrict our attention to two broad classes of ratings: investment grade and speculative grade. Investment grade corporate bonds carry a rating of Baa3 or higher (from Moody's) and/or BBB- or higher (from Standard & Poor's), while speculative grade bonds consist of issues with ratings below these, a s well a s corporate bonds t h a t a r e nonrated. Section I1 contains a discussion of t h e construction of t h e default r a t e series used in this paper. In section 111 w e present a model of t h e pricing ofdefault- prone bonds in terms of their required y ~ e l d st o maturity and compare derived implied default r a t e s with a c t u a l default experience. Section IV rep e a t s this exercise for holding period returns and discusses t h e complications of using holding period returns. Section V investigates t h e relationship between changes in expected corporate default r a t e s and certain macroeconomic measures. In section VI w e present a summary of t h e paper and some closing remarks. 11. Measuring Corporate Defaults Economists have tracked t h e performance of corporate d e b t beginning with a study by W.B. Hickman (1958). His and almost every subsequent study defines t h e r a t e of default as t h e value of issues defaulting during t h e period examined divided by t h e value of bonds outstanding during some part of t h e p e r ~ o d(usually t h e beginning). Altman and Nammacher (1985) (A&N hereafter) 3 -- http://clevelandfed.org/research/workpaper Best available copy a r g u e that, since almost all defaults over t h e last few years have occurred in t h e low-rated sector, t h e appropriate measure of t h e corporate default r a t e includes only t h e value of low-rated bonds in t h e denominator. In A&N, and in this study, only publicly held, straight (non-convertible) corporate debt with a speculative (or no) rating is included in t h e denominator. Convertible bond defaults w e r e included in t h e numerator for our measurements, however, because of t h e likelihood t h a t market participants do not differentiate between losses in this s e c t o r and t h e straight bond sector. In t h e six-year period from t h e beginning of 1980 through 1985, a total of $3.586 billion of corporate d e b t defaulted, roughly $1.021 billion of which consisted of convertible issues. Eliminating t h e s e defaults would substantially reduce our measured default rates. The a c t u a l default r a t e series presented below i s therefore biased upwards. r A complication arises in t h e construction of a measure of t h e default r a t e f o r bonds of a given rating: by t h e time a n issue defaults, i t h a s usually descended in rating until i t h a s reached t h e rating D (for Default). We, therefore, limit our analysis t o t h e performance of all low-rated corporate debt. Figure 1 presents a monthly time series plot of annualized default r a t e s for January 1980 through December 1985. This was constructed by dividing t h e par value of bonds defaulting at e a c h month by t h e par value of outstanding low- rated bonds at e a c h date. T h e par value of defaulting issues was obtained from A&N (up to December 1984) a n d Standard & Poor's Bond Guide (through December 1985). Observations o n t h e par value of speculative grade bonds outstanding were taken at t h e end of e a c h year from Standard & Poor's Bond Guide. Estimates of outstanding bonds, by month, w e r e obtained by interpolating annual measures. The mean of t h i s default series is 1.883 percent, with a 4 http://clevelandfed.org/research/workpaper Best available copy standard deviation of 3.297 percent. The series reaches a minimum of zero a t several points, including part of 1981, and attains a maximum of 19.504 perc e n t in April 1982. For some of t h e analysis, a smoothed default r a t e series is employed, partly because of t h e volatile nature of actual default rates. A smoothed version was constructed by summing at e a c h d a t e defaults occurring over t h e past 12 months and dividing by outstanding low-rated bonds six months earlier. This is essentially a 12-month moving average of t h e monthly default r a t e series. From January 1980 through December 1985, t h e average value of our smoothed default r a t e series is 1.796 percent, roughly corresponding t o A&N1s estimate of 1.507 percent (obtained from year-end observations for January 1978 through December 1984). Our smoothed default r a t e series h a s a standard deviation of 1.273 percent. The maximum value of 4.756 percent occurs in November 1982, while t h e minimum of 0.045 percent occurs in October 1981. A summary of these series, and all subsequent d a t a series, c a n be found in table I. W e note t h a t t h e constructed series a r e based on the assumption t h a t defaults result in a t o t a l loss t o bondholders. In fact, AhN find t h a t defaulting bonds continue to t r a d e at 41 p e r c e n t of par within one month following t h e default. Therefore, t h e a c t u a l ttloss rate t 1 i s somewhat lower than our default r a t e estimates. 111. Default R a t e s and Yields t o Maturity Our theoretical model is based on t h e pioneering work of Bierman and Hass (1975), with subsequent extensions by Yawitz (1977). The proposed model is in t h e same spirit as t h a t used by Yawitz, Maloney,and Ederington (1983) 5 http://clevelandfed.org/research/workpaper Best available copy to model yield spreads in the municipal bond market on the basis of differing default probabilities and tax effects. 1 Most asset-pricing models are based upon the first and second moments of the return distribution. With risk-neutral preferences, on the other hand, agents consider only the first moments of the distributions of return: the security's expected return completely determines i t s market price. This framework facilitates the construction of -a certainty-equivalence pricing relationship. Assume that a promised coupon (or principal) payment will be rendered at the end of a given period with a perceived probability P. A payment pro- .A t mised t periods from now i s expected to be received with probability P default occurs (and applies only to payment streams for which there havebeen no previous defaults) with probability (1-PI. In the event of a default, a fraction of the promised coupon and principal payments is received, denoted here by p. I f capital markets are frictionless, and information is costless, arbitrage will force the market price of a certainty-equivalent (default-risk-adjusted) payment stream, discounted at the riskless rate of interest to be equal t o a risky stream, discounted at the appropriate risky rate of interest. Algebraically: where i is the riskless rate of interest, r is the risky rate of interest, C is 6 http://clevelandfed.org/research/workpaper Best available copy t h e promised risky coupon rate, and N is t h e number of years to maturity. Using a geometric sum formula to express (1) without summations, we have: The yields to maturity, i and r, a r e for bonds t h a t a r e identical in all respects except for t h e likelihood of default. Further simplifications of t h e above expression a r e possible, if one approximates t h e finite-maturity coupon bond with a perpetuity, and if one assumes t h a t default results in a total loss 2 t o debt holders. A more general specification of equation ( I ) would involve time-sub- scripts for t h e variable P, so t h a t payment r a t e s would b e allowed t o vary over calendar time (hence t h e term structure). The product of t h e PIS from t h e initial d a t e t o t h e relevant payment d a t e s would replace P t in t h e first term in t h e numerators of t h e l e f t hand side of (1). The product of t h e P's from t h e initial d a t e t o t h e d a t e preceeding t h e payment d a t e would b e multiplied by 1 minus t h e expected payment r a t e in t h e relevant period for t h e second term. Of course, t h e r e is no way to identify t h e values of t h e separ- ate expected probabilities of payments. In addition, a closed-form solution like t h a t of equation (2) could not b e found. The use of a single, constant measure of P c a n b e interpreted as a n "average" likelihood of payment, summarizing expectations of f u t u r e payment rates. 7 http://clevelandfed.org/research/workpaper Best available copy A. Yield t o Maturity Data. An index of yields to maturity for low-rated bonds was obtained from Salomon Brothers 1 Corporate Bond Research department. The index used begins at t h e end of 1979 and is constructed from a sample of 176 bonds (as of September 19851, weighted by t h e outstanding principal amount of each issue (to control for e a c h issue's relative influence on market rates) t h a t meet t h e following criteria: 1) more than $25 million in principal outstanding (assuring adequate marketability), 2) ratings below Baa31BBB-, o r not rated but of lower than Baa31BBB- quality, 3) a coupon of 10 percent o r more, and 4) longer than 10 years in maturity.) In addition, we w e r e able t o obtain t h e weighted coupon r a t e s and weighted maturity d a t e for t h e sample at e a c h point in time. Defaulting bonds a r e removed from t h e sample, as a r e issues t h a t are upgraded t o investment- grade status. Complications arise in t h e analysis because of several uncontrolled factors. First, nearly all corporate bonds contain call provisions. In a sample of 702 currently outstanding, publicly held, low- rated (or nonrated) issues, all but 32 had call provisions, and 97 w e r e being called as of January 1986. In practice, many (high-coupon) low- rated bonds t r a d e on a yield-to-call basis. Of course, high- grade corporate bonds also c a r r y call provisions. The f a c t t h a t t h e low-rated sample consists of high-coupon issues, increases t h e likelihood t h a t they would b e called if interest r a t e s fall significantly (or if t h e f irmls financial condition warrants a n up-grading). This and o t h e r f a c t o r s imply t h a t t h e r e is no comparable high-grade index t h a t will exactly match e a c h of t h e characteristics (apart from default risk) of t h e low- rated sample. As a compromise, w e chose to use Salomon Brothers' New Medium Term 8 http://clevelandfed.org/research/workpaper Best available copy Industrials index f o r Aaa rated bonds, found in their Analytical Record of Yields and Yield Spreads. This series is based on estimates (by Salomon's Syndicate Department) of t h e required yields on issues coming t o market t h a t a r e rated Aaa and will mature in 10 years. These estimates were made a t t h e beginning of month t + l and were aligned with t h e low-rated index observations t h a t were taken on t h e last day of month t. It has been observed t h a t new issues a r e priced at yields slightly higher than t'seasoned" issues, d u e t o their relative lack of liquidity. The Aaa/AAA rated yields were chosen t o represent t h e default- risk- free r a t e s largely because of t h e lack of defaults by bonds originally issued with this rating in t h e past 15 years. Using t h e yields on long-term U.S. Treasury issues a s t h e default- free yield would complicate t h e analysis, because these securities lack-. call provisions, and because their returns a r e subject t o different t a x t r e a t ment. In addition, t h e sheer volume of transactions involving Treasury bonds introduces t h e possibility t h a t yield differentials r e f l e c t a marketability factor. 4 Finally, cross- sectional variations in t h e measured returns of a sample of representative bonds can b e attributed t o firm-specific idiosyncrasies. It i s assumed t h a t t h e a v e r a g e measured returns will vary systematically a s t h e result of a f a c t o r t h a t is tied t o t h e default experience of corporate bonds. The use of a weighted average of bond returns causes t h e influence of t h e idiosyncratic variations t o cancel one another. Hickman (1958, p.66) discusses t h e difficulties of using (weighted) average returns as measures of t h e return on a pooled investment portfolio. He concludes t h a t under most conditions, t h e e r r o r will b e negligible. http://clevelandfed.org/research/workpaper Best available copy 8. Comparison of Actual and Implied Default R a t e s A numeric solution program was employed t o solve (2) for t h e expected "payment rate", P, given supplied values for r, I, p, C, and N, at t h e end of each month t. This is a measure of t h e cross-sectional average of ~ m p l i e d expected payment rates, based on t h e yields of a cross-section of low-rated bonds. A problem emerges, however, because of t h e aggregation procedures used. L e t us assume t h a t P i s a n implicit function of r (with i, C, p, and N held fixed). Since (2) cannot b e solved explicitly f o r P, a computer simulation was employed to graph t h e implicit function with restrictions on t h e values of t h e o t h e r variables and a n assumption about t h e relationship between C and r. Figure 2 i s a graph of t h e simulation. Note t h a t when t h e payment r a t e P i s ' equal t o 1, t h e risky r a t e takes on t h e supplied value of t h e riskless r a t e (10 percent here). The relationship between P and r is shown t o b e convex in t h e relevant range. Jensenls inequality, therefore, suggests t h a t t h e cross-sectional average of P will b e g r e a t e r than, o r equal to, t h e measured payment rate. This implies t h a t our e s t i m a t e of (I-PI, t h e implied expected default rate, is biased downwards. A plot of (I-PI, t h e expected default r a t e implied by our model of yield differentials, is presented in figure 3 along with a plot of t h e moving average default series. The f a c t t h a t t h e implied expected default r a t e series appears t o track, and even lead, ltactualll default r a t e s so well i s surprising, given t h a t t h e implied r a t e represents a n average of expected f u t u r e default rates. This behavior indicates a d e g r e e of myopia on t h e p a r t of market participants. The spread between implied and a c t u a l (smoothed) default r a t e s is also surprisingly large and persistent over this period. 10 http://clevelandfed.org/research/workpaper Best available copy Acknowledging t h e statistical complications introduced by t h e construction of these variables, one may gain additional insight by using regression techniques. Cochrane- Orcutt adjusted regressions of (1-P) on constants and the "raw tt default r a t e series, ADR, a s well a s t h e smoothed default r a t e series, SADR, a r e presented below in table 2. These regressions indicate t h a t t h e r e is some connection between measured implied default r a t e s (based on risk-neutral preferences) and t h e two actual default r a t e series. The adjusted R-squares of 10.2 p e r c e n t and 11.1 percent, respectively, indicate t h e perc e n t a g e variation in t h e implied default r a t e series t h a t is "explained" by t h e two measures of a c t u a l default rates. The large t- statistics for t h e constant terms cause us to r e j e c t t h e null hypothesis t h a t t h e market's (risk-neutral) estimate of default r a t e s equals actual default r a t e experience. In fact,'. evidence suggests t h a t market prices imply default r a t e s t h a t exceed a c t u a l default r a t e s by roughly 5 percentage points. IV. Default Experience, Holding Period Yields, and Ex-post Performance In this section, w e apply t h e default-risk- neutral framework t o t h e pricing of risky d e b t in terms of t h e expected holding period yields on default-prone and default- free bonds. A bond's holding period return embodies changes in t h e market price as well as coupon earnings (pro-rated for t h e holding period). Define Bt t o b e t h e default- free bond's market price at t h e end of period t, and C t t o b e t h e promised coupon payment earned in period t. Now l e t t h e holding period return f o r a default-risk-free bond b e defined by Ht, such that: http://clevelandfed.org/research/workpaper Best available copy The corresponding gross return to t h e holder of a default- prone bond with price B l i and coupon C l i in period i is represented by: Note t h a t H and h a r e period-specific returns, in turn, convertible t o annual t t rates. Now l e t mt b e t h e percieved probability t h a t a n issuer will not default over period t, conditional upon a default not having previously occurred. If t h e period under consideration is a single month, then (mt)12 i s t h e expected likelihood t h a t t h e firm will not default over a given year. L e t us further assume t h a t in t h e e v e n t of a default, t h e holder of t h e risky bond will r e c e i v e with certainty a fraction p of t h e beginning period price B*t-l. The investor's expected (net of default) r e t u r n o n t h e risky bond, E(ht), i s therefore given by: http://clevelandfed.org/research/workpaper Best available copy For t h e certain case in which m equals 1, E(ht) will equal h t t' whereas in t h e c a s e of certain default, E(h ) will equal (p-l), resulting in a loss to t h e t bondholder. In t h e absence of market imperfections, equilibrium in t h e riskneutral setting requires t h a t t h e expected net-of-default return on t h e default-prone and the default- free bond will be equal. Setting t h e right-hand side of ( 5 ) equal to Ht and using equation (4), we have: Subtracting both sides of (6) from ht and rearranging, gives: where (I-mt) is t h e period- specific expected default r a t e embodied in t h e holding period yields of t h e default-prone and default- free securities, given a n assumed recovery rate, pO5 Note t h a t (7) represents a risk-neutral, e x a n t e relationship between expected holding period returns and expected default rates. A bond's realized holding period return, however, i s a n e x post measure of performance. Conversely, measured yields-to-maturity a r e based on expect e d performance and embody e x a n t e expected default rates. Bond holding period returns may deviate from expected returns, limiting our ability to measure implied default r a t e s from t h e difference between holding period yields of risky and risk-f r e e bonds. http://clevelandfed.org/research/workpaper Best available copy Consider a short-run increase in t h e expectation of corporate defaults. C e t e r i s paribus, this would have t h e effect of reducing t h e prices of outstanding low-rated bonds, thereby reducing t h e measured holding period return h . IJnder t most conditions, this would lower t h e "implied" default r a t e (1-mt). 6 Indeed, below we show t h a t relatively short-run price movements (resulting from new default information) c a n cause t h e r~ght-hand side of (7) t o take on negative values, thereby violating t h e definition of a probability. Therefore, (7) cannot b e used t o obtain implied default rates. What o n e obtains from applying this formula t o e x post returns is a differential "performance rate" for low-rated bonds. A. Holding Period Data. A proxy for ht was constructed monthly by Blume and Keim (1984) based on t h e price movements and coupon payments of t h e bonds used in Salomon Brother's Low-Rated (or High Yield) Bond Index (discussed above). The "merged1' series s t a r t s at t h e end of January 1980 and covers through J u n e 1984. It has a mean of 1.14 p e r c e n t (for a n equivalent annual average return 7 of 14.57 percent) and a standard deviation of 4.09 percent. As a measure of t h e holding period returns on default- free bonds, Ht, w e used Salomon Brother's High G r a d e Index f o r t o t a l rate-of-return found in their Analytical Record of Yields and Yield Spreads (up to December 1985). The index was formed by calculating t h e t o t a l returns of roughly 900 issues with weights based on issue size. The weights a r e revised monthly, and bond issues a r e included and deleted as ratings a r e updated. The a v e r a g e weighted maturity of t h e issues at t h e end of 1985 was 22.1 years. This series i s also used as a benchmark return in Blume and Keim (1984). 14 http://clevelandfed.org/research/workpaper Best available copy 8. Comparison of Actual and Implied Default Rates. In order to minimize extraneous influences on holding period yields, a holding period of one year was selected, in addition to the one-month holding period. The Blume-Keim series was converted t o an annual return series by accumulating monthly returns over t h e past year a t each month. That is, the measured annual holding period return at each d a t e is based on t h e returns to bond holders who sold a security purchased one year earlier (and collected coupon payments for t h e period). With these measures of return, (7) implies t h a t annual performance r a t e s a r e estimated. In figure 4, we present a plot of t h e performance r a t e implied by equation (7), obtained from annual holding period measures, along with t h e historic moving average default rate, SADRt. Confirming our intuition, negative performance r a t e s exist when actual d e f a u l t experience is highest. The performance rate, (I-mt), reaches a minimum value of -0.1009 in November 1982, t h e month following t h e maximum value reached by t h e smoothed actual default r a t e series. It i s clear t h a t periods corresponding t o negative performance r a t e s a r e those in which one- year holders of low-rated bonds realized significant losses. In general, t h e performance r a t e series descends as actual default r a t e s rise, and vice-versa. In t a b l e 2, w e present t h e regressions of t h e measured performance r a t e s (expressed in annual terms and based on one- and 12-month holding period yields) on t h e two actual default r a t e series. The low R-squares indicate t h a t relatively l i t t l e of t h e variation in performance r a t e s is explained by a c t u a l default rates. The negative coefficients on actual default r a t e s and t h e significant t- statistic on t h e smoothed default r a t e series supports t h e observa15 http://clevelandfed.org/research/workpaper Best available copy tion of a negative correlation between a c t u a l default r a t e s and t h e performance of low-rated bonds. The significant (and positive) t- statistics on t h e constant terms of t h e regressions, using a performance r a t e series formed from 12-month holding period yields suggest t h a t , on average, holders of lowr a t e d bonds realized significant holding period gains relative t o their highgrade counterparts. V. Default Expectations and Macroeconomic Measures In this section, a n a t t e m p t is made t o allow for t h e influence of o t h e r macroeconomic variables, in addition t o a c t u a l corporate default rates, on implied default and performance rates. P a s t studies of differential quality spreads have used a n assortment of macroeconomic indicators. J a f f e e (1975) examines f a c t o r s t h a t influence t h e risk spread of corporate yields in a cyclic a l fashion. He finds t h a t t h e most significant variable in explaining t h e risk spread i s a measure constructed by Fair (19711, based on d a t a collected by t h e University of Michigan Survey Research Center, which acts as a proxy f o r consumer sentiment. This f a c t o r was also used by Cook and Hendershott (19781, in addition t o others, to explain t h e spread between high-grade corpor a t e and Treasury securities. Rather than t a k e this approach, implied default and performance r a t e s a r e t e s t e d f o r correlation with new default information and surprises in macroeconomic measures. I t i s well known t h a t in periods of (unanticipated) rising prices, firms with fixed nominal contractual obligations tend to benefit. Conversely, (unanticipated) reductions in prices may c a u s e hardship to some firms. Since expect e d inflation will already b e incorporated into t h e contracts, it is t h e unan16 http://clevelandfed.org/research/workpaper Best available copy ticipated part of inflation t h a t will a f f e c t t h e probability of default. Therefore, a natural macroeconomic proxy is t h e deviation of t h e percentage change in t h e consumer price level from expectations. Other indicators of macroeconomic activity a r e the Board of Governors of t h e Federal Reserve's industrial production index and t h e Labor Department's unemployment r a t e estimate. Two characteristics of our sample period tend t o limit t h e effectiveness of this exercise, however. The first is t h e relatively short sample period available to us. The size of t h e market f o r low-rated bonds approached significance only towards t h e end of t h e 1970s. The identification of long-run relationships is, thus, seriously hampered. Secondly, in t h e sample period of this study, t h e overall inflation r a t e was, on average, falling, a f t e r a long period of accelerating inflation. The e f f e c t s of this regime switch on t h e reported results is indeterminate, introducing t h e possibility t h a t t h e behavior of market participants over a longer period may well differ from t h e behavior exhibited here. To test f o r a relationship between unanticipated inflation r a t e s and our estimates of implied default and performance rates, w e constructed a n unanticipated inflation series by subtracting one-month-ahead forecasts of t h e perc e n t a g e change in t h e CPI (obtained from Money Market Services) from a c t u a l monthly percentage changes. Similar series were constructed f o r measures of t h e unemployment r a t e and t h e percentage change in industrial production (a monthly proxy for GNP). O n e would expect that, if a g e n t s incorporate new information about t h e economy (in addition t o firm-specific factors) into their expectations of default rates, these proxies will b e related to changes in expected default rates. 17 http://clevelandfed.org/research/workpaper Best available copy For vat-ious sample periods, we regressed t h e first differences of implied default rates, obtained from differential yields t o maturity (I-P), on a constant, a c t u a l d e f a u l t r a t e s (in levels a s well as first differences), unanticipated inflation, unanticipated industrial production, and unanticipated unemployment. The macroeconomic surprises were lagged o n e month, a s the tirning of t h e actual series normally lags t h e reported period by a f e w weeks. The results, found in t a b l e 3, indicate t h a t of t h e t h r e e macroeconomic indicators, surprises in reported measures of industrial production have t h e highest correlation with implied expected default rates, although t h e level of actual d e f a u l t r a t e s contributes slightly more. When t h e f i r s t differences of a c t u a l default r a t e s a r e used, t h e surprise in inflation appears t o have t h e highest (negative) correlation with expected default rates. However, no variable e n t e r s significantly in e i t h e r regression at t h e 95 percent confidence level. The low adjusted R-squares also leads us t o conclude t h a t current macroeconomic surprises a r e poor indicators of expected default rates. The same regressions, adjusted for serial correlation of t h e error terms, were run using t h e implied performance r a t e (based on one-month holding period yields, converted t o annual rates) in place of expected default rates. Also found in t a b l e 3, t h e s e results suggest that, though insignificant at t h e 95 percent confidence level, surprises in inflation a r e most closely related (positively) t o performance rates. I t must b e t h e case t h a t firm-specific f a c t o r s dominate t h e formation of default expectations t o t h e point t h a t surprises in macroeconomic measures a r e poor predictors of overall quality spreads. VI. Summary & Conclusions This paper represents t h e f i r s t e f f o r t t o tie together the. differential 18 http://clevelandfed.org/research/workpaper Best available copy returns required by holders of low-rated corporate bonds and t h e actual default experience of these issues. A model of t h e behavior of low-rated bond pricing was developed in a risk-neutral setting. W e applied t h e model t o the observed returns of a sample of bonds and compared t h e default r a t e s implied in these returns to t h e default experience of low-rated debt. W e conclude t h a t t h e default r a t e s implied in corporate bond returns exceed those experienced in recent years. In this sense, holders of well-diversified portfolios of low-rated corporate bonds a r e rewarded for bearing default risks. It was also shown t h a t measured holding period returns cannot be used t o e x t r a c t implied default rates. Finally, w e examined t h e relationship between a s e t of macroeconomic variables and expected measures of default and performance rates. W e conclude t h a t expected corporate default r a t e s a r e not related t o any of t h e macroeconomic variables at t h e 5 percent critical level, although expected default r a t e s were most strongly related t o surprises in inflation measures and actual default rates. Surprises in output proxies appear t o have less of a relationship t o expected default rates. Further study in this a r e a will require t h e accumulation of b e t t e r (and more detailed) measures of corporate bond returns. The construction of a standardized d a t a base, modeled a f t e r t h e C e n t e r for Research on Security Prices (or CRISP) tapes, would most benefit f u t u r e endeavors in this field. In addition, a longer sample period would increase our understanding of both t h e pricing of default risk and t h e relationship between expected default r a t e s and macroeconomic activity. http://clevelandfed.org/research/workpaper Best available copy Footnotes 1. Yawitz, Maloney, and Ederington (1983) do not compare their estimates of default r a t e s in t h e municipal market with a c t u a l rates. 2. If o n e approximates t h e finite- maturity coupon bond with a perpetuity-- multiply t h e left-hand side of (2) by ( ~ + i ) - ~ / ( l + i )and ' ~ t h e right hand side by ( ~ + r ) - ~ / ( l + r and ) - ~ l e t N approach infinity--then (2) becomes: If we assume t h a t default results in a t o t a l loss t o holders (pol, then this becomes: 3. For t h e 176 issues in Salomon Brother's Low-Rated Index, as of September 1985, 23 w e r e r a t e d BB (by Standard & Poor's), 26 w e r e r a t e d B+, 34 w e r e r a t e d B, 45 w e r e r a t e d B-, and 48 w e r e r a t e d CCC. AdtN find t h a t t h e highest default- risk group (in terms of rating at issuance) w e r e bonds r a t e d single 0. This index, therefore, represents t h e average returns of t h e riskiest corporate bonds. 4. Coupon payments received from Treasury securities a r e currently exempt from state and local income taxes. Ibbotson and Sinquefieldls (1982) measured default premium, constructed by subtracting t h e ex-post holding returns o n 20 http://clevelandfed.org/research/workpaper Best available copy Treasury bonds from AAA/Aaa rated corporates, may mostly reflect this tax differential. 5. As in footnote 2, the assumption that default results in a total loss to bondholders (that is, p=0) gives: 6. The partial derivative of (1-mt) with respect to ht is: {Ht + (l-p)IAht + (1-p# 2 , and will be positive when Ht>(p-1). The smallest value reached by the onemonth holding period return on the high-grade series from January 1980 through June 1984 is -0.0799. The 12-month holding period minimum for this rate is -0.1296. Based on Altmants estimates, (p-1) equals -0.59, implying that this condition will be met under most circumstances. 7. See Blume and Keim (1984) for a description of this series. http://clevelandfed.org/research/workpaper Best available copy Table 1 Summary of Measured and Constructed Series Mean Standard deviation Minimum (date) 12/79 through 12/85 ADRt 0.01857 (I-month) ADR . 0.0 1776 (12-month) 1/80 through 6/84 12/80 through 6/84 (1-m) 0.08858 (fr. 12-month HPY) t Expressed as an annual rate. 0.00000 (23 points) Maximum (date) http://clevelandfed.org/research/workpaper Best available copy Table 2 Implied Default and Performance Rates on Actual and Smoothed Default Rates Dependent variable Const. ADR SADR R~ DW (I-m) (fr. I-month HPY) (1-m) (fr. 12-month HPY) (I-m) (fr. 12 month HPY) Note: All regressions were run using the Cochrane-Orcutt procedure for first-order serial correlation. The reported Durbin-Watson statistics are less powerful when the serial adjustment technique is used. The t-statistics are reported in parentheses. http://clevelandfed.org/research/workpaper Best available copy Table 3 Implied Default and Performance R a t e s on Actual Default R a t e s and Macroeconomic Surprises (Dependent Variable First-Differenced) Dependent variable Const. ADR Infl. 1nd.Prod. Unemp. (Dependent Variable and ADR First-Differenced) *-. 'K DW . t Run using t h e Cochrane-Orcutt procedure for first-order serial correlation. The t- statistics a r e reported in parentheses. http://clevelandfed.org/research/workpaper Best available copy Figure 1: Actual Default R a t e Series (in Annual Rates). Source: Altman and Nammacher (1985), and Standard & Poor's Bond Guide. 25 http://clevelandfed.org/research/workpaper Best available copy Figure 2: Plot of r against P (with p=.41, i=.l, C=.913rt,and N=14). t: The partial adjustment for coupon rates is based on the fact that corporate bonds were trading, on average, at 91.3 percent of par value. Source: See text. 26 http://clevelandfed.org/research/workpaper Best available copy Figure 3: Time S e r i e s of (1-P) and Smoothed Actual Default Rates. - SADR a - 0 - 0 - - - I I 1 - /- - - I I I I --J ~ ~ Source: See text. -1 I I I ~ ~ ~ ~ /\ \ I I ' \" /'' \,*I 4 ' t b 1I I I I ', / \ ,\ I J / I' I\ ~ \ I I I I I I I I - - I I ~ ~ ~ ~ I ~ ~ ~ ~ ~ ~ ~ ~ ~ I I I I ~ I I I http://clevelandfed.org/research/workpaper Best available copy F ~ g u r e4: Time Series Plot of (I-m), Constructed from Annual Holding Period Yields and Smoothed Actual Default Rates. Legend - -------- - I --- I - - - - - 1 1 1 1 1 1 1 1 '@------e- 1 1 Source: See text. 1 1 1 1 1 1 1 1 1 t l l 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 I I I I I . http://clevelandfed.org/research/workpaper Best available copy References Altman, Edward I., and Scott A. Nammacher. "The Default Rate Experience on High Yield Corporate Debt," Morgan Stanley & Co., Incorporated, March 1985. Bierman, Harold Jr., and Jerome E. Hass. "An Analytic Model of Bond Risk Differentials," Journal of Financial and Quantitative Analysis, vol. 10, no.5 (December 1975), pp. 757-73. Blume, Marshall E., and Donald B. Keim. "Risk and Return Characteristics of Lower-Grade Bonds,I1 Rodney L. White Center for Financial Research, The Wharton School, University of Pennsylvania, December 1984. Cook, Timothy Q., and Patric H. Hendershott. "The Impact of Taxes, Risk and Relative Security Supplies on Interest Rate Differentials," Journal of Finance, vol. 33, no. 4 (September 19781, pp. 1173-86. Ederington, Louis H., Jess B. Yawitz, and Brian E. Roberts. "The Informational Content of Bond Ratings," NBER Working Paper No. 1323? April 198 4. Fair, Ray C. A Short-Run Forecasting Model of the United States Economy. Lexington, MA, D.C. Heath and Company, 1971. Fisher, Lawrence. "Determinants of the Risk Premiums on Corporate Bonds," Journal of Political Economy, vol. 67, no. 3 (June 1959), pp. 217-37. Hettenhouse,. C.,. and W. Satoris. "An Analysis of the Information Value of Bond-Rating Changes,!! Quarterly ~ e v ~ eofwEconomics and Business, vol. 16, no. 2 (Summer 1976), pp. 65-78. Hickman, W. B. Corporate Bond Quality and Investor Experience. Princeton: Princeton University Press, 1958. Ibbotson, Roger G., and Rex Sinquefield. Stocks, Bonds, Bills, and Inflation: The Past and The Future. Charlottesville, VA: Financial Analysts Research Foundation, 1982. Jaffee, Dwight M. "Cyclical Variations in the Risk Structure of Interest Rates," vol. 1, no. 3 (July 19751, pp. 309-25. Jones, E. Philip, Scott P. Mason, and Eric Rosenfeld. "Contingent Claims Analysis of Corporate Capital Structures: an Empirical Investigation," Jouinal vol. 39, no. 3 (July 19841, pp. 611-25. http://clevelandfed.org/research/workpaper Best available copy Weinstein, Mark. "The Effect of a Rating Change Announcement on Bond Price," Journal of Financial Economics, vol. 5, no. 3 (December 19771, pp. 329-50. Yawitz, Jess 8. "An Analytical Model of Interest Rate Differentials and ~ l f f e r e nDefault t ~ecoveries,"Journal of Financial and Quantitative Analysis, vol. 12, no. 3 (September 19771, pp. 481-90. Yawitz, Jess B., Kevin J. Maloney, and Louis H. Ederington. "Taxes, Default Risk, and Yield Spreads," NBER Working Paper, No. 1215, October 1983.