STATISTICS on the movements of interest rates over time suggest that there can be only limited adjustments of rates for the purpose of carrying out anti-inflationary and anti-deflationary monetary policies. Although short-term rates, which may be significant for inventory changes and for balance of payments problems, show considerable variability, those rates which are most likely to influence investment outlays do not. The statistics indicate that changes as large as two percentage points (e.g., from 3 per cent to 5 per cent) in long-term interest rates have rarely been brought about within the space of 12 months. Within the span of a few months—perhaps the longest adjustment period that successful monetary policy can permit—the variability of long-term rates is considerably smaller; changes of even one percentage point have been rare.
One observer states that the limit to variations in the yield of British consols for countercyclical purposes is “a fraction of 1 per cent.” And up to the 1950’s the cyclical variations in the U.S. corporate bond yield series had been practically imperceptible.1
For Great Britain, two characteristics of the statistical series used have misrepresented the variability of long-term interest rates. First, the conventional procedure has been to represent long-term interest rates by the yield on consols (perpetual government bonds); and second, the series relied on to represent yields on bonds of business companies (i.e., the Actuaries’ Investment Index) has reflected only the “current” (“running”) yield, i.e., the ratio of the nominal (“coupon”) interest rates on the securities to their prices2 which, as explained below, is only a portion of the variation in their total yield. Data available for the United States suggest a third factor which may be more important than either of these, viz., that the cyclical variations in the long-term rates paid by borrowers are substantially in excess of those in the series for average yields, which are based on a sample of outstanding, established (“seasoned”) bonds. Analysis of this third factor will be the chief concern of this paper.
Concentration of attention on yields on consols has been misleading because of the large decline in the price of such a security caused by, say, a rise of one percentage point in its yield. When the yield rises from 4 per cent to 5 per cent, the price of the security drops by one fifth. Knowledge of this relationship induces investors to refuse to buy consols in recession periods if an attempt is made to push the yield below, say, 4 per cent.3 The 4 per cent level may therefore be a minimum to which consol yields can be pushed in time of recession. But perpetual bonds are not representative of the debt issues of industrial countries. The prices of the 14-year to 20-year range of maturities, which includes most of the issues of industrial borrowers, fall only by half or five eighths as much as the prices of consols for a 1 per cent increase in yield. This means that investors would be willing to buy these bonds even after their yield had fallen below the 4 per cent level. If investors were willing to accept the risk of just as great a price decline for industrial bonds as for consols, they would be willing to buy the industrials as long as yields were no more than 2 per cent (rather than 1 per cent) below the levels foreseen. This suggests that the cyclical variability of interest rates on bonds actually issued by business is (almost) twice as great as the variability of the yield series usually considered—yields on consols. The possibilities for stabilizing investment are therefore much greater than observation of the limited variability of consol yields suggests.
Use of the current yield to represent yields on industrial debentures understates the cyclical amplitude of the movements in the yield from these bonds, because it makes no allowance for the additional return to the lender when market rates are high and the average outstanding bond is selling below par: viz., the profit to be received on redemption of the bond at a price of 100, which is reflected in a gradual rise in the bond’s price as it moves closer to its redemption date. Conversely, when market rates fall to low levels and the price of the average outstanding bond rises above par, the observed decline in the issue’s current yield is reinforced by the pending loss that will be realized when the bond is redeemed at par.
The variability of yields (and hence prices) on existing bonds is of chief concern to the financial wealth holder, whose actions may prevent bond yields from dropping far during recession periods. But the stabilizing effects of monetary policy can be exerted most effectively through its influence on borrowers; and the would-be borrower is influenced by the amplitude of variation in the interest rate that he pays, which is the rate on new bond issues. The underrepresentation of the cyclical variability in long-term bond yields that is due to the citation of the average yield on existing (“seasoned”) rather than newly issued bonds may be deduced from data on bond issues in the United States. It will be shown in this study that the amplitude of variation of the yield on new issues is systematically ½-¾ percentage point wider than the amplitude of the conventional series of yields on seasoned long-term bonds, on which economists’ opinions have been based. Thus, in June and July 1957, when interest rates were at a peak, the yields on all new issues of long-term public utility bonds exceeded the current yields on seasoned bonds by between 0.48 and 0.98 percentage point. The weighted average excess was 0.94 percentage point in June and 0.69 percentage point in July.
Some recognition of the wider amplitude of yields from new issues has developed since the middle of the 1950’s, but even that recognition has been limited by a view that the wider oscillations observed were, to a great extent, reflections of the rapidity of change in long-term rates; this implies that if economic conditions made it necessary to stabilize rates at high levels, the excess yield paid by new borrowers would shrink to a small amount, and the cycle-stabilizing benefits from the wider amplitude of yields on new issues would almost disappear. Although some of the observed widening of amplitude is probably a transitory reflection of expected changes in yields, it will be shown that a major part of the excess yield observed is permanent, a reflection of the difference in characteristics between the new bond issues and those older issues which happen to be chosen to represent the long-term interest rate.4
Charts 1 and 2 illustrate the relationship over time between interest rates paid by long-term borrowers and the reported rates usually taken to represent long-term interest rates in the United States. They show, for Grade A and Grade Aa public utility bonds, the yields on new 30-year issues and the comparable Moody’s average yields on outstanding bond issues.5 Three periods of high interest rates are covered—1953, 1957, and 1959. In each period the yield on new issues was much in excess of the reported yields on seasoned issues, while during the adjacent low interest periods the new issue yield was approximately the same as the seasoned yield.6
Mr. White, economist in the Finance Division, received his undergraduate and graduate training at Harvard University. He has contributed articles to a number of economic journals.
R. F. Harrod, Towards a Dynamic Economics (London, 1948), p. 120, and Richard Goode and Eugene A. Birnbaum, “The Relation Between Long-Term and Short-Term Interest Rates in the United States,” Staff Papers, Vol. VII (1959–60), pp. 224-43.
See description of the series in U.K. Annual Abstract of Statistics, 1959.
Harrod, loc. cit., and James Tobin, “Monetary Policy and the Public Debt,” Review of Economics and Statistics, Vol. XXX (1953), p. 126.
Those doubting the possibility of wider cyclical variations in interest rates often doubt their desirability as well, on the grounds that (a) wider variations would entail higher average rates to counteract the additional risks to bondholders of wider price movements; (b) in prosperous times very high rates tend to be inflationary rather than deflationary owing to their effect on government expenditures for debt service and the rise in living costs associated with higher mortgage rates and business costs of production.
It may be briefly noted that the second of these two points disregards the usually cited demand-deflating effect of high interest rates. The first point is not relevant to the widening of the range of variation discussed in this paper. It refers to the amplitude of variations in the yields on seasoned bonds, and to the consequential risk of capital loss to the holders of such bonds, bought when interest rates were depressed, should these rates subsequently rise. The amplitude of variations considered in this paper is the larger one which is found in a different series, viz., the interest rates paid by borrowers on successive new issues.
It may be noted that the cyclical variability of the short-term bank loan rate is also understated by the customary U.S. statistics. The unique U.S. banking tradition, that of all loans received a fraction must be left on deposit with the lending bank (roughly 10 per cent in times of easy money and 20 per cent when credit is tight), caused the 1½ percentage point rise in bank charges reported between autumn 1958 and autumn 1959 to be an effective increase of rates of 2.36 percentage points. (In addition to understating the effective variations in banks’ interest rates, disregard of this consideration has led to overstatement of the cyclical variation in the effective money supply.)
Through 1955 the data are Moody’s series of weekly averages of the published daily interest rates, i.e., the average yield on ten Grade A and Grade Aa quality seasoned public utility bonds having 25-35 years to maturity. After 1955 the data are the tabulations in Moody’s Bond Survey of monthly average yields on seasoned issues and of the weighted average of yields on new issues during the months. Bonds having any of the characteristics of shares are excluded. However, bonds with restrictions on the time of call for redemption (which first appeared in late 1956) are included in the averages of yields on new issues.
This paper is based on public utility securities rather than on industrial bonds because the former provide a larger number of fairly homogeneous new issues. Similarly, while Baa bonds might be more representative of industrial issues than A or Aa bonds, the yields on Baa bonds are too diverse for efficient comparisons between different issues and over time. Hence Aa bonds are given chief attention. Industrial bonds do, however, appear to show the same yield patterns as Aa bonds.
A major part of corporations’ bond financing is through “privately placed” issues for which no yield statistics are available. However, since the chief market for such security issues comprises the same investors who dominate the new public issues market (life insurance companies, etc.), it is to be expected that the interest costs paid by borrowers would be about the same for new private issues as for new public issues (see E. R. Carey, Direct Placement of Corporate Securities, Boston, 1951, pp. 109-10, and George J. Leness and others, New Money for Business: Techniques of Long-Term Corporate Financing, New York, 1956, pp. 91-92). Hence the exclusion of data on private placement yields does not limit the generality of the present investigation.
Data recently made available for Germany show, even more clearly than the U.S. data, conformity in the timing of changes between yields on seasoned issues and new issues but a much greater amplitude of variation in yields on new issues. During 1959–60, yields on new issues rose 1½ percentage points—over twice the rise in yields on outstanding issues; and in a subsequent reversal, a portion of these movements was retraced. The yield differences associated with a given level of yield on seasoned issues appeared to be the same whether rates were rising, falling or stable. (See Monthly Report of the Deutsche Bundesbank, January 1962, p. 14.)
Sources of information on the U.S. bond market are given in footnotes 13, 22, 25, and 26 (pp. 121, 128, 135, and 136).
A fifth explanation often given is that the comparative inactivity of the seasoned bond market causes seasoned bond prices to adjust to prices in the new issues market only after some delay. This explanation is highly questionable, and in any case it would, like the fourth explanation in the text, constitute no barrier to the maintenance of high yields on new issues. Therefore, it is dealt with only as a part of the test of the validity of the fourth explanation.
This is not a logically necessary relationship. All that is logically necessary is a smaller amplitude of variation for yields on seasoned bonds. In fact, in the 1930’s, when interest rates on bonds were very low, yields on new issues tended to be below the average yields reported for seasoned bonds; see David Durand and W. J. Winn, Basic Yields of Bonds, 1926–1947 (National Bureau of Economic Research, New York, 1947), pp. 38-39. As will be made clear below, in part this was the consequence of the fact that bond yields had been declining continuously during the 1930’s.
The erratic and biased character of the yields on seasoned bonds has led one financial writer to condemn the popular series of average bond yields and to propose that some average of yields on new issues be used in place of high-grade (seasoned) bond yield averages to represent “the” interest rate; see R. F. Murray, “What Yield Do You Use?” Security Analysts Journal, August 1955, p. 15. But despite the fact that some series for yields on new issues have become available in the last few years, economic and financial analysis has continued in general to be based on averages of yields of seasoned issues.
The close proximity of the D and S0 curves along most of their length is correct only on the assumption that seasoned bonds can be bought and sold without commission. In fact, the buyer of the seasoned bond may pay a ¼ per cent commission (above the “asked” price), which is equivalent to a decline of 0.01 per cent in the yield and can be ignored. The seller pays the same commission and, in addition, has usually to accept a “bid” rather than an “asked” price, so that he typically receives a price 1½-2 per cent below the asked price. This means, since the yield differential is computed on the basis of the seasoned issues’ asked price, that the seller of seasoned bonds requires an extra yield of 0.09-0.12 percentage point on new issues before he will switch to them. To allow for this qualification, the horizontal portion of S0 can be drawn above the zero level by
The strength of the taxation inducements to take losses should not be overestimated; the banks’ holdings of long-term corporate securities are relatively small, and individuals and their trust funds can deduct the full amount of capital losses from income only to the extent of $1,000 in a given year (plus a carryforward of $5,000 more into the next five years) and must include any capital losses on other kinds of assets. On the other hand, as shown in Diagram 1, actual transactions in seasoned bonds will probably involve only a very small part of the outstanding holdings of those who are prepared to participate in the seasoned market. Bondholders who have no aversion to losses should therefore account for a major part of the potential market that would be activated by a small change in the yield differential.
Instances of search for maximum income regardless of capital losses realized are cited in L. S. Wehrle, “Life Insurance Investment—The Experience of Four Companies,” Yale Economic Essays, Spring 1961, p. 129.
These conclusions might be challenged on the ground that a prolonged period of high rates, say, three years, might be expected to use up the infinitely elastic range of the supply of low-coupon bonds (the flat segment of curve S1 in Diagram 1 just above and to the right of a’ becoming shortened and finally disappearing). This would cause a’ to rise appreciably. That is, however, unlikely to occur for two reasons. In the first place, the available evidence suggests an annual volume of transactions in public utility and telephone bonds, excluding the large proportion of transactions that are between brokers, of only about 10 per cent of the outstanding total. (A tabulation of non-new-issue bond sales in the over-the-counter market during three months of 1949 showed a total of $514 million sold; outstanding bonds were about $13½ billion. See Irwin Friend and others, The Over-The-Counter Securities Markets, New York, 1958, p. 122.) Since a major part of the final sales included in this figure will have been made for special reasons by those who were averse to the realization of capital losses (cf. the leftward steeply-rising segment of curve S1 in Diagram 1), it follows that relatively little will have come from the infinitely elastic segment of the supply curve representing those bondholders who are not averse to realizing capital losses. Even if only, say, 15 per cent of the outstanding bonds are in the hands of those not averse to realizing capital losses, several years of high rates would have to pass before the infinitely elastic segment of the supply curve could disappear. In the second place, the infinitely elastic range tends to reconstitute itself, for insofar as net sales to others at a loss occur, the purchasers will be holding bonds which they acquired at a discount and hence will face no capital-loss obstacle, or only a small one, to the resale of the securities later in the high interest rate period.
This conclusion requires two qualifications: First, this sort of rise in yields on new issues should have some psychological effect on yields in the seasoned market, especially if the higher yield is long continued. Second, some extra funds may be diverted from the seasoned market by the extra yield; this can occur insofar as the large savings institutions—which buy most of the new issues but are excluded from the seasoned market because the minimum size of a transaction which it is efficient for them to undertake is too large for that market—are the ones to lose out by the rationing of new issues, while buyers normally operating in the seasoned market are able to obtain a part of the new issue under the rationing arrangements and thereby divert some funds from the seasoned market.
There is one further reason for concern over the importance of the underwriter’s interest-rate risk factor in the observed yield spread. That factor is probably much weaker for the bond issues of manufacturing companies than for the public utility issues used here to represent the entire bond market. This is so because public utility issues must normally be awarded to underwriters by competitive bidding on the day before the date of issue, whereas industrial companies usually have continuing arrangements with a single underwriter which give the latter time to test the market, adjust the yield offered, and secure enough customers in advance of the issue.
If, in fact, yields on new issues did decline more quickly than yields on seasoned issues, this would be highly advantageous, for it would mean that antideflationary policies take effect more rapidly than commonly thought by those who worry about the difficulty of lowering long-term interest rates.
Leness, op. cit., p. 148, reports that a successful public utility issue “normally” sells out in a few hours.
If the yield on a new issue should prove inadequate, the price of the bond may have to decline until it is at a point where a more than “adequate” yield is offered. Although this consideration justifies enlargement of the extra yield on new issues, it ought to leave the total allowance for interest risk within the ¼ percentage point range. Raising the yield on a 25-30 year issue by ¼ per cent would increase the borrower’s interest payments by an amount that has a present value equal to 3½-4 per cent of the value of the issue.
See Leness, op. cit., p. 113. This variability in underwriters’ charges is significant in another connection also: it increases the cyclical flexibility of long-term borrowing costs somewhat beyond what is shown by the data for yields on new issues.
Refunding is profitable for the debtor company if the interest rate at which it can reborrow is lower than the interest rate it ceases to pay when it redeems an existing debt. But if the price of the high-coupon bond issue were free to rise as market rates fell, the situation would be the same as that of an investment in the bonds of other firms that were issued at a time of high interest rates. As market interest rates decline, the prices of these other bonds will rise so that their yields remain equal to market rates. And it is pointless to borrow at the market rate to purchase bonds which yield no more than the market rate. In the case under consideration, however, the price at which the company can buy back its own bonds cannot rise above the “call price” (usually the original issuing price plus one year’s interest), and large savings in interest payments do become possible.
Moody’s Bond Survey, January 21, 1957, p. 747.
The differential income tax treatment of the interest yield to maturity on bonds selling close to, or well below, par could logically explain an important part of the observed spread between yields on seasoned and new bond issues, and require the existence of such a spread even if interest rates were at their cyclical minimum. Individuals paying high rates of income tax should strongly prefer bonds selling at a discount because that part of their yield to maturity which represents the rise in their prices toward par (or the redemption price) will be taxed at the low capital gains tax rate. In practice, the uncertainty of receiving the favored tax treatment on the low-priced bonds in all years (the capital gains tax applying only to the excess of the year’s gains over any losses realized), and the unimportance in the seasoned bond market of those paying high tax rates, make this an unimportant explanation for the yield differential. The seasoned bond market is dominated by pension funds and other tax-exempt or partially tax-exempt institutions. Even when yields are at cyclical peaks, there is still “only a scattering of individual investor interest” in the high-grade bond market. (See D. M. Kelly, “Basic Structure of the Corporate Bond Market,” Commercial and Financial Chronicle, July 20, 1961, p. 24.) Wealthy individuals who pay very high taxes on current income are not interested in the capital gains tax advantages of bonds standing at a discount because fully tax-exempt state and local bonds offer much higher net-of-tax yields.
Individuals might be attracted by the higher levels of peak interest rates offered on the medium-grade bonds that industrial firms commonly issue. The present discussion, based on high-grade bonds, is nevertheless conclusive, for it establishes that an important discrepancy in yield exists even in the absence of strong tax considerations.
See, for example, A. F. Brimmer, “Credit Conditions and Price Determination in the Corporate Bond Market,” Journal of Finance, September 1960, pp. 365-66.
For the two issues of January 1957 shown at the bottom of the table, a relative reduction in yield of
Thus David Durand and W. J. Winn (op. cit.) made the point in 1947 that the seasoned bonds then available for bond yield averages sold above par and therefore had to have unrepresentatively high yields. But for conditions like those of today they merely said: “If interest rates ever rise substantially so that high-grade bonds are selling at less than par, the coupon-yield relationship may present new problems for analysis” (ibid., p. 40, note 6).
The present paper, which was originally prepared in 1957 for use within the International Monetary Fund, was brought to date for publication. Apart from the peripheral indications contained in the Durand-Winn footnote, no support by the financial community, written or oral, for the call-risk/capital-gains factor was discovered in the course of preparation. Subsequently, however, a description of the bond market by an officer of a leading bond house was published in which the seasoned bond market (apparently a very active one) was described as dominated by sophisticated institutional investors; private individuals were considered a negligible factor; and the call-risk factor was mentioned as an explanation of the difference between the yields on new issues and on low-coupon seasoned issues. The differential caused by this factor was said to be very small, however, because bond buyers could not well judge the prospects for long-term interest rates and because of the prevalence of prolonged call restrictions in the large proportion of new issues that were privately placed with institutional investors, especially life insurance companies. Aversion to capital loss was not mentioned as a factor in the yield differential. See D. M. Kelly, op. cit., pp. 24-25.
R. F. Murray, loc. cit., criticized the customary yield averages partly on the ground that the new issues market is insulated from the market for seasoned issues, being dominated by the large life insurance companies, pension funds, etc., which must deal in such large blocks that they would disorganize prices on the market for seasoned issues (which must deal chiefly in “odd lot” amounts of any given bond). Although the large institutions which dominate the new issues market are excluded from the seasoned market, the dominant groups in the latter (small insurance companies, small pension funds, and banks and trust funds, as well as individuals) can move with complete freedom between the two markets. (See Investment Bankers Association of America, Fundamentals of Investment Banking, New York, 1949, p. 762, for verification of the fact that the smaller life insurance companies operate easily on the “seasoned” market; see M. Hamberg and others, Characteristics of Transactions in Over-The-Counter Markets, Philadelphia, 1953, pp. 51-55 and 76, and Irwin Friend, Activity on the Over the Counter Market, Philadelphia, 1951, p. 27, on the domination of non-new-issue bond transactions by the various financial institutions and the comparatively large size of their purchases and sales.)
An officer of one large life insurance company, who agreed that large units could not buy and sell efficiently on the market for seasoned bonds, nevertheless considered that insurance companies as a whole annually “purchase and sell tremendous amounts” of seasoned bonds. That these do not reflect specially arranged, irregular acts is indicated by the fact that the company normally does not know the identity of the seller or buyer of the bonds transferred. (See R. B. Patrick, “Management of the Life Insurance Investment Portfolio,” in Investment of Life Insurance Funds, David McCahan, ed., Philadelphia, 1953, pp. 64-65.) Some more recent statistics are available to support this view. Voluntary sales of corporate bonds from portfolios, plus (probably very minor) sales of municipal bonds and corporate stocks, which “typically arise as investment officers face particular opportunities to switch their holdings to improve yield, strengthen quality, or broaden diversification of the portfolio,” are estimated to have totaled the large amounts of $700 million in 1957 and $800 million in 1958 (K. M. Wright, “Gross Flows of Funds Through Life Insurance Companies,” Journal of Finance, May 1960, p. 156). Since this estimate is an extrapolation of data on the larger companies (which have difficulty in making open market sales of seasoned securities) to represent all companies, it presumably understates total sales of seasoned securities but may correctly represent total sales of seasoned corporate bonds.
The frequency and reliability of bond price quotations tend to decline as the bond ages and, apparently, disappears from circulation. This suggests a tendency toward a diminishing frequency of over-the-counter transactions. Even though the stock exchange is only a secondary market place for (nonspeculative) corporate bonds, the published data on transactions show that six months after an issue has appeared transactions are still quite large; and that lapse of time is sufficient to establish the coupon rate-yield relationship.
These yields are Moody’s figures adjusted to an “asked” price basis from the reported average of “bid” and “asked” prices. The spread between the two prices is typically 1½-2 per cent; Moody’s bond prices were normally, therefore, ¾-l per cent below those used in this study. The lower prices mean that Moody’s yields as reported are about 0.05 percentage point higher than those used here.
This is derived from the arithmetical relationship between market price and yield to maturity. A rise in the bond’s price to 1 per cent above par reduces the yield to maturity in two ways: it reduces the ratio of the interest payment to the amount invested by 1 per cent, and at the time of redemption (at the par price of 100) it subjects the investor to a 1 per cent capital loss. The 1 per cent reduction in a 4 per cent interest payment costs slightly less than 0.04 percentage point, and the 1 per cent capital loss incurred after, say, 30 years is equivalent to less than a one-thirtieth or 0.03 percentage point reduction in annual interest receipts. The two elements together cost a little over 0.06 per cent.
The path traced by bcd represents the arithmetical relationship between a bond’s price and its yield to maturity. The slope of bcd is given by the fact that for 30-year bonds a 1 per cent change in price causes an opposite change of 0.06 percentage point in yield to maturity. In Charts 3, 4, and 5 a long arrow is drawn to indicate this slope.
Represented by a steep slope (not shown in Diagram 2) in the yield curve to the left of the origin.
When the yield on seasoned issues rises from ob to og, the yield on new issues rises (in a first approximation) by the additional amount gh. The ratio of this extra rise gh to the rise bg in the basic yield is easily seen to equal the ratio (sign disregarded) of the slope of the rising segment of the yield curve, ch, to the slope of the line bcd. The first slope is approximated by the slope of straight lines fitted to the rising segments of the yield curves in Charts 3, 4, and 5. The slope of bcd is determined arithmetically (an 0.06 percentage point rise in yield for each 1 per cent decline in price).
With the ratio of the two slopes called a, the seasoned yield is, the cyclical minimum yield on seasoned issues is.min., and the estimated yield on new issues in, the above relationship may be represented algebraically as follows:
The above formula ignores the temporary additions to new issue yields caused by difficulties of marketing and fears of intervening rises in interest rates. A slight exaggeration of the permanent rise in yields on new issues is created by the simplifying assumption that the new issue will be found at h on curve ch rather than at the intersection of that curve with line af (see next paragraph of text).
W. J. Winn and A. W. Hess, Jr., “The Value of the Call Privilege,” Journal of Finance, May 1959, p. 189. A representative of the Securities and Exchange Commission (which discourages the electric power companies it supervises from accepting call restrictions) considered that five-year call restrictions had no significant effect on yields; see J. A. Pines, “Discussion,” Journal of Finance, May 1959, p. 219.