Alternative tests of agency theories of callable corporate bonds.
Which of these agency problems has a significant impact on observed financing behavior? Empirical research on corporate bond calls has revealed limited information about the separate effects of each agency theory, in part, because the theories have two important overlapping predictions. First, each of the agency problems is exacerbated when default risk is greatest, implying that riskier debt offerings are more likely to include a call provision. Second, each of the theories implies that both bond call features and shorter-maturity bonds can equally well mitigate the agency problem.
Thus, empirical research that relies heavily on default risk and maturity to explain the prevalence of call provisions is not able to assess the importance of the separate agency theories of callability. Moreover, the strong explanatory power of these two variables, while consistent with agency theories, may be attributable to other, unrelated, forces driving the use of call options on corporate bonds.
Our objective is to determine which, if any, of the agency theories explains the use of call options. While previous research has found evidence favoring the agency theories as a whole, the studies do not attempt to weigh the merits of the individual theories separately. We construct tests of the three theories either separately or in pairs using four types of data. First, we investigate bond ratings changes to determine if call options resolve information problems. Second, we look at the distribution of call dates, which play a pivotal role in the asymmetric information and underinvestment models. Third, we test for a relationship between investment activity and the use of call options, as the underinvestment theory predicts. Fourth, we evaluate the risk incentive or asset substitution theory by examining the value of the call options.
Our tests are based on a sample of 958 bonds issued during 1987-1991, when issuance of noncallable bonds was quite common. Indeed, after 1981 the use of call options by investment-grade firms decreases dramatically (see Crabbe (1991)), effectively nullifying the attempts of theorists in earlier decades to explain why nearly all bonds are callable.(1) In contrast, previous empirical studies on call options have used samples dominated by bonds issued prior to 1982 that were almost always callable. The authors of those studies have been forced either to analyze special circumstances, such as the use of two-tiered versus one-tiered call provisions (Thatcher (1985)), or to overweight the observations of noncallable bonds (Kish and Livingston (1992)). In our sample, the numbers of callable bonds and noncallable bonds are about equal, greatly improving the reliability of the inferences.
The results of our alternative tests lend far less support to agency theories than have previous studies. First, counter to the predictions of the asymmetric information theory, we find that firms issuing callable debt do not subsequently reveal favorable information. Second, the distribution of first call dates exhibits a degree of uniformity that seems inconsistent with the notion that call dates are set close to the date when new information will be revealed, as the asymmetric information and underinvestment theories state. Third, we find no relationship between the level of investment activity and the use of calls. Fourth, our estimates of call values suggest that riskier firms are including less valuable call options than high-grade firms, a result completely at odds with the prediction of the risk incentive theory.
We conclude that the three agency theories of callability do not have sufficient empirical support to retain their position as the most prominent explanations of the use of call options on corporate bonds.(2)
I. The Use of Call Options to Resolve Agency Problems: Theory and Evidence
This section reviews the existing theory and empirical evidence concerning the use of call options on corporate bonds.
Bodie and Taggart (BT) (1978) show theoretically that call options on corporate bonds can resolve the underinvestment problem identified by Myers (1977). This agency problem arises because, after issuing bonds in period t, managers may receive information in period t + 1 that reveals that their investment opportunities only provide a payoff to bondholders. Managers, who maximize shareholder wealth, then would prefer not to make the full investment expenditures in period t + 1. Bondholders, knowing that suboptimal investment policy reduces the expected value of the firm, will demand a risk premium on the bonds when they are issued in period t.
The numerical example in BT's paper highlights the critical role default risk plays, showing that firms with very little bondholder protection are more likely to find themselves in such a situation. Indeed, BT state that "the greater is the probability of default, the more new investment tends to 'bail out' the bondholders and hence the more noncallable debt tends to weaken the investment incentive" (page 1193). Including a call option on the bond alleviates this problem by allowing the firm to replace the existing debt with new debt in period t + 1, in what is essentially a recontracting based on the new information. BT show that short-maturity bonds can equally well solve this agency problem.
Barnea, Haugen, and Senbet (BHS) (1980) extend the theoretical work of BT and Myers to include two more agency theories that are based on asymmetric information and the incentive to take on greater risk. In the BHS asymmetric information model, managers have private information in period t when the bonds are issued, and that information is revealed to all in period t + 1.(3) Managers with favorable information will not want to issue bonds in period t because of the wealth transfer from shareholders to bondholders that will occur in period t + 1.
By issuing callable bonds in period t, however, managers mitigate the effects of bond mispricing because they simultaneously issue another instrument (the call option) that is equally mispriced in the opposite direction. The call option offsets the loss suffered by shareholders because bondholders, with their imperfect information, will underestimate the likelihood of a call, just as they underestimate the value of the debt.
As pointed out by Myers and Majluf (1984), the riskier the bond, the greater the effect of asymmetric information on its pricing. BHS show that a short-maturity bond is also useful in solving this problem.(4)
In the second BHS model, management has an incentive to take on lower-valued, riskier projects after issuing bonds. Such a shift will result in a wealth transfer from debtholders to shareholders and a decrease in the value of the firm. If debtholders understand the incentives of shareholders to shift toward higher-variance investments, they will pay less for the debt when it is issued.
The solution to the risk incentive problem is to issue callable debt. Then, if managers shift projects, they would decrease the value of the shareholders' long position in the call option. To eliminate the incentive to increase risk, a firm will include a call option whose value equals the potential gain from switching investments. This problem may also be solved with shorter-maturity debt.
In summary, BT and BHS consider three models involving agency problems that justify the pervasive use of call options: the underinvestment problem, asymmetric information, and the risk incentive problem. Each theory implies that callable bonds and short-term bonds are substitutes. The three theories also imply that call options are more useful as default risk increases because agency problems increase with the risk of the firm. Thatcher (1985) asserts that "as firms use more debt and bonds become riskier, stockholder actions have more effect on bond payoffs, and the magnitude of agency costs from all sources is increased" (page 557). Masulis (1988) notes that "call provisions also involve negotiation and enforcement costs of their own, which limits their profitable application. In this regard, agency cost benefits are likely to be greatest in firms with high leverage and large proportions of intangible or growth assets" (page 38).
Three empirical studies examine agency theories of callable bonds: Kish and Livingston (1992), Mitchell (1991), and Thatcher (1985). Each study involves the estimation of a qualitative dependent variable equation. Of the three, only the Kish and Livingston study has a sufficient number of noncallable bonds to estimate a zero/one equation on the inclusion of a call option, although they are forced to overweight the sample of noncallable bonds. Mitchell examines the joint decision to include a call option, a sinking fund provision or a refunding provision, and the choice of maturity, while Thatcher investigates the decision to issue a two-tiered call option.
The explanatory variables in these studies are the term to maturity of the bond and proxies for agency costs (such as bond ratings, leverage ratios, profitability, and dividend payout ratios) that are highly correlated with default risk. The models are relatively good at predicting whether a bond includes a call option. The authors interpret these findings as supporting the agency theories.
While the empirical research to date is consistent with the agency theories as a group, the tests are not powerful enough to differentiate the three agency theories from each other. Each theory predicts that the use of call options will increase as default risk rises, and each shows that shortening the bond's maturity is a good substitute for a call option. Because they rely heavily on default risk and maturity as explanatory variables, these researchers are unable to distinguish the relative importance of the three theories. Moreover, the declining use of call options in recent years brings to the forefront the need to re-examine the empirical validity of these theories as a group.
II. Call Provisions: The "Stylized Facts" Have Changed
Much of the existing literature on call options refers to the stylized fact that nearly all bonds are callable. Since 1982, however, an increasing percentage of corporate bond offerings have been noncallable, especially among investment-grade issuers. Crabbe (1991) shows that call options, which had been included in nearly all (95%-99%) high-grade bonds between 1977 and 1981, were attached to fewer than one-quarter of such bonds by 1990 (Table 1, column 1). In our sample of bond offerings between 1987 [TABULAR DATA FOR TABLE 1 OMITTED] add 1991, the trend away from callable bonds among investment-grade firms is similar (Table 1, column 3).(5)
Among the speculative-grade firms in our sample, the fraction of bonds that are callable remained near 95% between 1987 and 1989 (Table 1, column 5). Few speculative-grade bonds were issued during 1990-1991, causing the fraction of callable bonds to be affected by small sample problems. Nevertheless, the number of noncallable junk bonds rose substantially in 1991. During 1992-1994, about 20% were not callable.(6)
The trend away from call options does not stem from an increase in the use of other solutions to agency problems. For example, if firms had chosen to issue more short-term debt, the decline in the use of call options would be consistent with the agency theories. Crabbe's analysis, however, shows that the declining use of call options is unrelated to the choice of maturity. At every maturity, the use of call options has declined over the 1982-1990 period.
Nor have sinking funds replaced call options on noncallable debt. Sinking fund provisions can serve as substitutes for call provisions (Kalotay and Williams (1992) and Ogden (1988)). If firms had increasingly substituted sinking funds for call options after 1981, most of the noncallable bonds in our sample would have had sinking funds. The first column of Table 2, however, shows that the vast majority of noncallable bonds had no sinking fund provisions. Rather, sinking funds are more frequently used in conjunction with call options.
A possible explanation for the decline in the issuance of callable debt over the last decade is that agency costs may have diminished over time. Such a decline could have occurred if firms became less risky. This is an unlikely explanation for the trend toward noncallable bonds, however, because default rates have increased substantially since the 1970s. Columns 7 and 8 of Table 1 present default rates on investment-grade and speculative-grade bonds, respectively, calculated by Fons, Carty, and Girault (1993). Default rates rose for both investment-grade firms and speculative-grade firms after 1981 and more than doubled for the latter group.
Table 2. The Use of Sinking Fund Provisions This table presents data on the fraction of noncallable and callable bonds that included sinking funds by rating category. Noncallable Bonds Callable Bonds Moody's Rating (%) (%) Aaa 0 25 Aa 0 54 A 0.4 35 Baa 0 50 Ba 7 54 B 17 63 C or lower 9 100 Investment-Grade 0.2 41 Speculative-Grade 9 62
Strong patterns in the use of call options have developed since 1982. Call options on corporate bonds are now mostly observed on speculative-grade bonds and on long-term investment-grade bonds (e.g., bonds with maturities of 30 years or longer). Table 3 presents the frequency of call options in our sample of 958 industrial straight bonds (see Data Appendix for details). Nearly all of the speculative-grade bonds are callable, while fewer than 30% of the investment-grade bonds are callable. Within the speculative-grade category, call options are used more often by firms with B ratings than by firms with Ba ratings, which have lower default risk.
Figure 1 shows the distribution of maturity and callability for investment-grade and speculative-grade bonds. In our sample, speculative-grade bonds are almost exclusively 7 to 12 years in maturity, most are callable, and none has a maturity over 20 years. In contrast, in the investment-grade sector, maturities often extend beyond 20 years, and the only bonds likely to be callable are those with very long maturities.
These strong correlations among default risk, maturity, and callability are consistent with agency theories of call provisions. As noted, Barnea, Haugen, and Senbet (1980) emphasize that the three agency problems can be solved either by a call provision or by issuing short-term noncallable debt. Thus, each theory implies a correlation between maturity and callability: Bonds will be either short-term and noncallable or long-term and callable, a prediction that is consistent with the stylized facts in Figure 1. Similarly, each of the theories suggests that riskier firms are more likely to use callable debt.
Table 3. Ratings and the Frequency of Call Options in New Offerings of Industrial Bonds, 1987-1991 This table reports the fraction of bonds in our sample that are callable. The sample is described in the Appendix. Number of Moody's Rating Percent Callable Offerings All Ratings 46.03 958 Investment-Grade 29.33 699 Speculative-Grade 91.12 259 Aaa 18.18 33 Aa 32.00 100 A 29.82 399 Baa 28.74 167 Ba 68.89 45 B 95.75 212 Caa 100.00 2
Table 4 shows that maturity and default risk alone have significant power in predicting whether a bond includes a call provision. Maturity is measured in years, and default risk is proxied by an indicator variable for investment-grade status. McFadden's R-squared, a measure of the explanatory power of the probit, is 0.27. The maturity and default risk variables correctly classify 79% of the total sample, of which 74 and 83% are correctly classified for callable and noncallable bonds, respectively. By comparison, in a model with eight additional variables, Kish and Livingston (1992) are also able to classify 79% of the total correctly, and very similar percentages - 78 and 81%, respectively - of the callable and noncallable bonds.
Unfortunately, as with previous research, the probit results are consistent with the hypothesis that call provisions mitigate agency problems, but they do not shed light on the relative importance of the specific agency problems of asymmetric information, underinvestment, and risk incentives.
III. New Tests of the Agency Theories
Previous studies have had difficulty distinguishing among the agency theories because the tests have focused on whether a bond includes a call provision. In [TABULAR DATA FOR TABLE 4 OMITTED] our tests, we examine several areas in which the theories are distinguishable. Specifically, we examine the subsequent rating changes of the bonds, the date that the bond can be called, the investment activity of the issuing company, and the theoretical value of the call option. Table 5 shows the tests and their implications for the three theories.
In Panel A of Table 5, we use bond rating changes to test the asymmetric information theory. According to the theory, when favorable information is released about the firm in period t + 1, the call feature allows the firm to recontract at a lower yield. Upon the release of favorable information, the rating agencies would be more likely to upgrade the issuer's bond rating. Therefore, we expect callable bonds to receive more upgrades than noncallable bonds.(7)
Similarly, if rating downgrades proxy for the revelation of unfavorable information, downgrades of callable bonds should occur less frequently than downgrades of noncallable bonds. Basically, a firm should not issue callable debt if it is likely to reveal information that will negatively affect the value of its bonds, such as lower-than-expected profitability or leverage-increasing capital structure changes.(8)
Our second test focuses on the distribution of first call dates among the callable bonds in the sample. According to the asymmetric information and underinvestment theories, the first call date should be set so as to minimize the costs of the agency problems. In the underinvestment model, firms pass up positive-net-present-value opportunities that are revealed in period t + 1 unless they can extinguish existing bonds. Thus, managers should select a call date that corresponds to the date when prospects for the outcome of the project are revealed. Similarly, in the asymmetric information model, managers will select a call date as close as possible to the date of revelation (period t + 1), since failure to do so will result in a wealth transfer from shareholders to bondholders.
In both models, if the call date precedes the revelation date, the firm will not benefit from calling the bond, because no new information will be available to facilitate the recontracting. If the call date is too late in the life of the bond, the call option solves neither agency problem - either the investment opportunity will be passed up or the bondholders will have been paid excessively high rates for too long.
Unfortunately, the theories' predictions about call protection cannot be tested directly because the date of information revelation typically is unobservable. The theories can be tested indirectly, however, by examining the [TABULAR DATA FOR TABLE 5 OMITTED] variation in call dates. According to the theories, firms will select a call date close to the date when information is revealed. Because financial contracting is costly, however, managers must weigh the advantages of choosing an exact call date that best resolves the agency problem against the cost of deviating from a "standard" call date.
Thus, the optimal call date depends on the potential reduction in agency costs and the costs of financial contracting.(9) High-risk firms, which face higher agency costs, receive more benefit from choosing a call date that matches the revelation date, while safer firms have little incentive to use discretion in selecting a call date. Our test, then (shown in Table 5, Panel B), is to examine whether low-rated bonds display a greater dispersion in call dates.
Next, we test for a relationship between investment activity and the use of call options. If recontracting is desirable because of new information about investment opportunities, then we would expect the use of call options to increase with investment activity (Table 5, Panel C). Evidence that investment activity is highly correlated with the use of call options would support both the underinvestment theory, which is strictly about investment, and the asymmetric information model, in which the private information of managers may be about investment opportunities.
We investigate the investment hypothesis in two ways. First, using time series data, we examine the correlation between the prevalence of callable bonds and business fixed investment. The underinvestment and asymmetric information theories suggest that call options will be used less frequently in times when investment activity is low. Second, using cross-sectional data, we test whether firms with high investment activity are more likely to issue callable bonds. The theories would be supported by evidence that firms with high capital expenditures are more likely to issue callable bonds.
If either of these tests is rejected, it would be evidence against the underinvestment theory, but it would not necessarily be inconsistent with the asymmetric information theory because, in that model, the information may be unrelated to investments.
Finally, we investigate the value of the call options. According to the risk incentive theory, when shareholders have much to gain from switching to riskier investments, a firm should issue debt with a high option value. Conversely, when the potential wealth transfer to shareholders is small, a firm will issue debt with a low option value. As discussed earlier, firms with high default risk have the greatest incentive to switch to riskier investments. Thus, the risk incentive theory may be tested by examining whether firms with low credit ratings issue debt with high option values.
Theoretical option models (Buser, Hendershott, and Sanders (1990) and Ho and Lee (1986)), have shown that the value of a call option is determined by several factors: the volatility of interest rates, the term structure of interest rates, the schedule of call prices, the period of call protection, and the maturity of the security. An individual firm, of course, must take interest rate volatility and the term structure as given. The firm, then, sets the option value on its bond by choosing a call price, the period of call protection (the first call date), and the bond's maturity.
To increase the call option value, an issuer would select a low call price, choose an early call date, lengthen the maturity, or choose some combination of these three determinants. To estimate the value of call options, we use the option pricing model developed by Kalotay, Williams, and Fabozzi (1993), which incorporates all the features that determine the value of the call. The risk incentive theory would be supported by evidence of a positive correlation between option values and default risk, while a negative correlation would weigh against the theory. This test, however, must be viewed as a joint test of the risk incentive theory and the accuracy of the Kalotay, Williams, and Fabozzi model.
IV. Bond Rating Changes and Asymmetric Information
We constructed a Moody's rating history for each bond from the time it was issued to July 1, 1992. To eliminate the bias arising from the correlation between call provisions and maturity, we eliminate bonds that matured before July 1992. Panels A and B of Table 6 show the percentage of bond upgrades and downgrades, respectively, for the seven major categories of Moody's ratings. The statistics are calculated for all rating changes, regardless of whether the rating change pushes the bond into a different major category (e.g., a downgrade from Baa1 to Baa2 would count as a Baa downgrade).
The data in Panel A indicate that bonds with high credit ratings are upgraded infrequently, while bonds with low original ratings have a higher incidence of upgrades. Notably, 25% of the B-rated bonds were upgraded by mid-year 1992, compared to 12% of the Ba-rated bonds and fewer than 10% of the bonds issued with investment-grade ratings. Moreover, the 46 upgrades of B-rated bonds nearly equal the combined total of 47 upgrades for the other rating categories.
The higher incidence of upgrades among speculative-grade firms is consistent with the asymmetric information theory, but it may only be indicative of the riskiness of low-rated bonds. If the upgrades truly reflect asymmetric information, the theory predicts that upgrades would be more frequent among callable bonds.
Panel A of Table 6 shows that callable bonds and noncallable bonds experience about the same fraction of upgrades. In the speculative-grade sector as a whole, the difference between the percentages of upgrades between callable and noncallable bonds is insignificant. Among the B-rated bonds, Moody's upgraded 60% of the noncallable bonds and 24% of the callable bonds, a difference that is statistically significant at the 0.05 level. Although the sample [TABULAR DATA FOR TABLE 6 OMITTED] includes only a small number of B-rated noncallable bonds, this finding is exactly opposite to the prediction of the asymmetric information hypothesis.
For the investment-grade sector as a whole, the results also do not indicate that callable bonds are upgraded more frequently than noncallable bonds. However, for bonds issued with Aa ratings the difference is significant at the 0.05 level. For the investment-grade bonds with Aaa, A, and Baa ratings, the differences between the percentage of callable upgrades and noncallable upgrades are not statistically significant at conventional levels of confidence.
Shown in Panel B, the evidence on downgrades strongly contradicts the asymmetric information theory. For each rating category, the percentage of callable downgrades exceeds the percentage of noncallable downgrades.(10) For bonds with original ratings of Aaa, A, Baa, and B, this difference is statistically significant at the 0.05 level, a finding that is exactly opposite to that predicted by the asymmetric information hypothesis.
Because the use of call options is so strongly related to a firm's rating, some ratings categories had small samples of either callable bonds or noncallable bonds. While tests of significance incorporate these sample size effects, large samples are understandably considered more reliable. The samples of bonds rated Aa, A, and Baa are much larger, and each sample provides at least 30 observations of both callable and noncallable bonds. Of these three major rating categories, only one shows a significantly greater propensity for upgrades of callable bonds compared to noncallable bonds, and two show a significantly greater frequency of downgrades of callable bonds. The evidence based on the large sample sizes does not support the asymmetric information theory.
The pool of callable bonds is somewhat more heavily weighted toward bonds issued earlier in the sample period. Few noncallable junk bonds were issued prior to 1990, and the sample of investment-grade callable bonds includes a higher fraction of bonds issued before 1990. Thus, our results may be driven by the fact that the chances of being downgraded are greater if the bond were issued prior to the 1990 recession.
There are two ways to check this: 1) test for upgrades and downgrades one or two years after issuance, thereby putting more emphasis on rating changes in 1988 and 1989 or 2) test upgrades and downgrades for the sample of bonds issued prior to 1990. Both these tests are less reliable because they reduce the sample sizes of rating changes. Nevertheless, the results of these two tests also indicate that callable bonds are no more likely to be upgraded than noncallable bonds and are more likely to be downgraded, although the difference is not usually statistically significant.
Overall, these findings do not support the hypothesis that firms include call provisions because of asymmetric information. If anything, the data curiously suggest that the choice of including a call provision may have resulted in a wealth loss to shareholders because a higher proportion of the bonds with call provisions were subsequently downgraded.
We caution that the results are specific to the 1987-1992 sample period, a period that includes a recession and a record volume of downgrades of corporate debt, and that similar results may not obtain in different time periods. At the same time, we have no reason to presume, a priori, that the phase of the business cycle should make downgrades of callable bonds more likely than downgrades of noncallable bonds. Indeed, the asymmetric information theory suggests that callable bonds will suffer fewer downgrades in a period of frequent downgrades, a prediction completely at odds with the empirical evidence.
V. The Choice of Call Dates
The distribution of first call dates in our sample is highly concentrated around five dates: years zero, three, five, seven, and ten (Table 7). These call dates tend to be associated with specific maturities (Table 7, columns 1-10). For example, most five-year bonds are callable after three years; most seven-year bonds are callable starting in year five; most ten-year bonds can be called in either five or seven years; and nearly all 30-year bonds are callable in ten years or refundable in ten years and callable immediately.
Among callable speculative-grade bonds, 93% have a first call date three or five years after issuance. First call dates on investment-grade bonds are more evenly split across years zero, five, seven, and ten, reflecting the greater dispersion in maturities in the high-grade sector.
Among bonds with maturities of seven years or more (which represent over 90% of the callable bonds in the sample), investment-grade bonds show greater variation in call dates (Table 7, columns 11 and 12). The coefficient of variation for the investment-grade bonds in this group is 81, compared to only 28 for the lower-rated bonds. This result is evidence against the asymmetric information and underinvestment theories.
Under a strict interpretation of the theories, the concentrated distribution of call dates implies that almost all firms expect information to be revealed on one of only a few possible dates after issuance. It seems likely that information will be revealed more uniformly over time. Taking the theories at face value, the data also suggest that a sizable minority of investment-grade firms, especially those issuing 30-year bonds, expect an information release so soon after issuance that they make the bonds callable immediately. Moreover, the data also imply that many investment-grade firms receive the information later than riskier firms. These findings do not suggest that speculative-grade firms use more discretion in selecting call dates, as suggested by agency theories.
VI. Investment Activity
To investigate the time series patterns of investment and [TABULAR DATA FOR TABLE 7 OMITTED] call options, we use annual data on the growth rate of nonresidential business fixed investment (Bureau of Economic Analysis) and data from Crabbe (1991) on the percentage of bonds that were noncallable each year between 1977 and 1990. Panel A of Table 8 presents the data on investment and the use of call options in each of the 13 years. Panel B shows the results of regressing the fraction of bonds that are noncallable on the yearly growth rate of business fixed investment.(11) No significant relationship exists between the two variables, reflecting the fact that the use of call options declined steadily between 1977 and 1990 while investment activity largely followed the business cycle.
We estimate a probit equation to test whether individual firms' capital expenditures can predict the use of call options. We were able to obtain COMPUSTAT data on capital expenditures corresponding to the year of issuance for 886 of the bonds in the sample. Capital expenditures are scaled by total assets in order to measure the expenditures in relation to the income and debt-servicing ability of the company.
Shown in the first two numerical columns of Table 9, the results of the probit show that capital expenditures are a highly significant predictor of noncallability. This result is exactly the opposite of the theory's prediction. The strong negative relationship between investment and callability may be the consequence of the high correlation between ratings and capital expenditures. Holding ratings and maturity constant, as in the last two columns of Table 9, the coefficient on capital expenditures is substantially smaller in absolute value, although still significantly different from zero.
The investment data provide no evidence that call features are related to the underinvestment problem. And the data suggest that if the asymmetric information model is correct, the private information does not concern the firm's investment plans.
Table 8. Aggregate Investment and the Use of Call Provisions over Time Panel A reports time series data on aggregate investment as measured by business fixed investment (source: Bureau of Economic Analysis) and the fraction of newly issued noncallable investment-grade bonds each year. The data on the fraction of bonds with call options are from Crabbe (1991). Panel B reports the result of a regression of the percent of bonds that were not callable each year on a constant and the growth rate of business fixed investment (BFI). T-statistics are reported in parentheses below the coefficient estimates. Panel A. Time Series Data Business Fixed Percent of Bonds That Are Year Investment Not Callable (Crabbe) 1977 364.3 1 1978 412.9 1 1979 448.8 2 1980 437.8 5 1981 455.0 3 1982 433.9 21 1983 420.8 21 1984 490.2 34 1985 521.8 25 1986 500.3 32 1987 497.8 49 1988 530.8 45 1989 540.1 63 1990 546.5 78 Panel B. Regression of Percent Noncallable on Growth Rate of Business Fixed Investment Percent Noncallable = 30.67 -90.59 x Growth of BFI (3.96)(-0.89) [R.sup.2] = 0.06
VII. The Value of Call Options and Risk Incentives
A firm can increase the option value on its bond by extending the maturity, setting an earlier call date, or reducing the call price. We have already seen from Figure 1 that riskier firms typically issue callable bonds with short maturities (low-grade callable bonds have an average maturity in our sample of 10.7 years, compared with 17.6 years for the investment-grade callable bonds). The much shorter maturity of bonds issued by speculative-grade firms is evidence against the risk incentive theory but may be offset by other factors. For example, call dates among riskier firms are slightly earlier (4.35 years versus 5.02 years).
Given the typical maturities and call dates of riskier bonds, call option values on speculative-grade bonds are unlikely to surpass those of high-grade bonds unless the lower-tier issuers set sharply lower call prices. In our sample, about a third of the callable bonds have par call prices, while the remainder include call prices set at a premium over par. The risk incentive theory suggests that bonds with low call prices (par call prices) will be issued by firms with low ratings. Table 10 shows that only 4% of the speculative-grade offerings have a par call price. By contrast, 52% of the investment-grade bonds have a par call price.
The differences between investment-grade and speculative-grade bonds are even more striking when the data are disaggregated by maturity. Virtually all the long-term, investment-grade bonds have a first call price set at a premium. Thus, most of the investment-grade bonds with par call prices are similar in maturity to the speculative-grade bonds with premium call prices. In the subsample of issues with 1-20 year maturities, 82% of the investment-grade bonds have a call price of par versus only 4% of the speculative-grade bonds. Moreover, at each maturity, the percentage of investment-grade bonds with a par call equals or exceeds the speculative-grade percentage.
This evidence on call prices does not support the risk incentive theory. Instead of including options with low exercise prices, speculative-grade issuers tend to include call options with high exercise prices. Combined with their short maturities and only slightly earlier call dates, the evidence suggests that riskier firms issue bonds with low option values.
The value of a call option, however, is jointly determined by the schedule of call prices, the period of call protection, and maturity, in a complex relationship. It is possible that the slightly earlier call dates of speculative-grade issuers (two-thirds of a year earlier on average) more than offset the effects of higher call prices and sharply shorter maturities.
We estimate option values for each of the 441 callable bonds in the sample using the bond option pricing model of Kalotay, Williams, and Fabozzi (1993). Their one-factor model is arbitrage-free with a short rate that follows a lognormal random walk. In the estimation, we assign 10% interest rate volatility for the short-term interest rate. Other [TABULAR DATA FOR TABLE 9 OMITTED] [TABULAR DATA FOR TABLE 10 OMITTED] inputs in the model include the term structure of Treasury yields at the time of issue and the term structure of credit spreads.
The estimated option values for investment-grade bonds average 2.4% of the gross proceeds, compared with 1.46% for the speculative-grade bonds. This difference is driven in large pan by the longer maturities of investment-grade bonds. Holding maturity constant, the option values of investment-grade and speculative-grade bonds are closer. For example, among bonds with ten-year maturities, the option values for investment-grade bonds average 1.199% of proceeds, compared with 1.167% for speculative-grade bonds. Even holding maturity constant, however, the riskier firms show no propensity to include more valuable call options, as the risk incentive theory predicts.
Table 11 presents the results of a regression of estimated option values on indicator variables for ratings. The omitted rating category is B. The first set of coefficients shows that nearly all the investment-grade callable bonds have significantly higher option values than the B category. The second set of coefficients is from a regression that includes maturity and maturity squared. This regression shows that the earlier call dates nearly offset the negative effects of the higher call premiums in the speculative-grade sector: Holding constant the effects of maturity, only the Aa-rated bonds have a significantly higher option value than that of the B-rated bonds. These estimates suggest that most investment-grade callable bonds have sharply higher call options in large part because of their longer maturities.
To summarize, speculative-grade firms tend to issue bonds with high call premiums and short maturities, seriously decreasing the values of their call options. The only feature that they consistently set to boost option values is the call date, which the riskier firms tend to set at three or five years, compared to seven or ten years more commonly found for investment-grade callable bonds. Assuming the Kalotay, Williams, and Fabozzi model is accurate, the call dates of the speculative-grade bonds are not sufficiently early to offset the effects of their shorter maturities and higher call prices. We find no evidence that firms with high agency costs issue bonds with high option values, as predicted by the risk incentive theory.
VIII. Estimating a Multivariate Logit
Our four new tests of agency theories provide evidence against the agency theories of call options. Each test, however, is designed to test one or two theories separately, and none can support or reject all three agency theories. Consequently, we may be increasing the odds of rejecting each hypothesis because we fall to hold all else constant. We can further test the three theories as a whole by estimating a multivariate logit of the probability of including a call option.
Shown in Table 12, the explanatory variables in the logit equation include default risk (investment-grade status), maturity, an indicator variable for bonds that were upgraded within one year of issuance, a similar variable for bonds that were downgraded, and capital expenditures in the year the bond was issued. The estimates from this logit support our previous results. Maturity and default risk are Strong predictors of the use of call options; subsequent upgrades and downgrades have no explanatory power, and capital expenditures are negatively related to the use of call options.
An examination of bond offerings during 1987-1991 reveals that callable bonds fall primarily into two groups: bonds with speculative-grade ratings and long-term investment-grade bonds. Noncallable bonds are almost exclusively of investment grade. Default risk and maturity, therefore, are highly correlated with call protection.
These stylized facts about call protection, default risk, and maturity are consistent with the predictions of the three major agency theories of call provisions. Each of the theories predicts that default risk increases the probability of including a call feature, and each suggests that short-term bonds serve the same purpose as callable long-term bonds. These predictions imply that default risk and maturity cannot be used to distinguish among the theories. We have examined several areas in which the various theories are distinguishable: the subsequent rating changes of the bonds, the period of call protection, the investment activity of bond issuers, and the value of the call option.
Overall, the evidence from our alternative tests shows little support for any of the theories. We find that, on average, favorable asymmetric information is not revealed after callable bonds are issued. Call dates are tightly concentrated around standardized dates, a finding that does not suggest that firms select the period of call protection to correspond to the time when information is expected to be revealed. High investment expenditures are generally associated with firms that issue noncallable debt, rather than callable debt. Moreover, the secular decline in the use of call options is unrelated to changes in aggregate investment activity. Finally, we find that speculative-grade firms intentionally reduce the call option value by choosing high exercise prices and short maturities.
While none of these tests can conclusively prove that the agency theories are incorrect, as a whole, they provide more [TABULAR DATA FOR TABLE 11 OMITTED] [TABULAR DATA FOR TABLE 12 OMITTED] than ample evidence to question the acceptance of the agency theories as the most probable explanation for callable corporate bonds. Because increasing numbers of firms have switched to issuing noncallable bonds, we must reconsider models based on the premise that all firms prefer callable debt. Future theories should also be consistent with the new stylized fact that call features are found mainly on risky bonds and very long-term, investment-grade bonds.
1 For example, Bodie and Taggart (1978) begin their article by asking the question, "Why do corporations almost never issue noncallable long-term bonds?"
2 Our results on upgrades and downgrades suggest that call options do not serve as signals of credit quality (such as in Robbins and Schatzberg (1986)). Other possible explanations for the inclusion of call provisions are related to taxes. Boyce and Kalotay (1979) and Marshall and Yawitz (1980) offer tax-based rationales for call provisions that emphasize the tax deductibility of the call premiums. Mauer, Barnea, and Kim (1991), however, show that the tax advantages to issuers are more than offset by the tax disadvantages to bondholders, given the correlation between interest rates and tax rates.
3 Although formulations of the asymmetric information theory usually consider the firm's investment opportunities, the theory applies to other situations in which investors have asymmetric information about the prospects of the issuing firm. For example, some firms reportedly issue callable bonds because they expect successful asset sales or equity offerings will allow them to deleverage their capital structure before the bond matures.
4 BHS do not explicitly consider the use of a signal, be it a call option or some other instrument, to resolve the information asymmetry. Rather, they assume that no costless signal is available to solve the problem and that a call option is likely to be at least as cost-effective as a signal in solving the problem. Robbins and Schatzberg (1986) present a signaling model in which risk-averse managers issue callable bonds to persuade the market that their firms have good prospects. In their separating equilibrium, firms with unfavorable information issue equity. Thus, their model appears to explain the pervasive use of call options in corporate bends observed prior to the 1980s. Their model also implies that riskier firms, for whom asymmetric information problems are more troublesome, should issue callable debt more frequently than safe firms.
5 Our sample has slightly fewer callable bonds than Crabbe because we exclude the telecommunications industry, which, like utilities, still frequently includes call options on bonds.
6 Based on Securities Data Corporation data on industrial straight bonds with speculative-grade ratings, through May 1994.
7 The news that prompts a racing change may not reflect asymmetric information. In this case, such revelations would not be expected to affect callable bonds more frequently than noncallable bonds.
8 In the Robbins and Schatzberg (1986) model, the call option signals improving credit quality. If the signal works, one might expect the credit rating agencies to incorporate the knowledge that callable bonds are offered by higher-quality firms into their ratings at the time of issuance. If so, subsequent rating changes would not be expected to differ between callable and noncallable bonds because all the information from the signal would have been incorporated into the initial rating. Representatives of the rating agencies indicate to us that they do not have a policy of assigning a higher rating to bonds with call options, ceteris paribus, because they do not view the call option as a signal, regardless of whether the bond market does. This implies that ratings will be upgraded upon the release of favorable information even if it had previously been signaled to the bond market. Moreover, the rating agencies do not necessarily assign a rating at issuance for the life of the bond. Rather, the time horizon is only the foreseeable future, which they say is about three years for most industrial bonds. If favorable information is not likely to be revealed within that horizon, they have no reason to raise the initial rating. As time passes and the horizon includes the call date, rating changes would reflect the information signaled by a call. Therefore, if the time horizons of the ratings are generally shorter than the call protection period, we would expect to see more upgrades of callable bonds even if the call option works as a signal. Thus, our test for bond rating changes also provides a test of the signaling models.
9 See Smith and Warner (1979) on the costs of financial contracting.
10 In contrast to the correlation between original ratings and the incidence of upgrades, the data indicate no correlation between ratings and the frequency of downgrades.
11 The growth rate of business fixed investment is used because the level trends up with the growth of the economy.
Barnea, A., R. Haugen, and L. Senbet, 1980, "A Rationale for Debt Maturity Structure and Call Provisions in the Agency Theoretic Framework," Journal of Finance (December), 1223-1234.
Bodie, Z. and R. Taggart, 1978, "Future Investment Opportunities and the Value of the Call Provision on a Bond," Journal of Finance (September), 1187-1200.
Boyce, W.M. and A.J. Kalotay, 1979, "Tax Differentials and Callable Bonds," Journal of Finance (September), 825-838.
Buser, S.A., P.H. Hendershott, and A.B. Sanders, 1990, "Determinants of the Value of Call Options on Default-free Bonds," Journal of Business (January), s33-s50.
Crabbe, L., 1991, "Callable Corporate Bonds: A Vanishing Breed," Federal Reserve Board Financial and Economic Discussion Series (FEDS) Working Paper 155 (March).
Fons, J., L. Carty, and D. Girault, 1993, "Corporate Bond Defaults and Default Rates 1970-1992," Moody's Special Report (January).
Ho, T.S.Y. and S. Lee, 1986, "Term Structure Movements and Pricing Interest Rate Contingent Claims," Journal of Finance (December), 1011-1029.
Kalotay, A. and G. Williams, 1992, "The Valuation and Management of Bonds with Sinking Fund Provisions." Financial Analysts Journal (March-April), 59-67.
Kalotay, A., G. Williams, and F. Fabozzi, 1993, "A Model for the Valuation of Bonds and Embedded Options," Financial Analysts Journal (May/June), 35-46.
Kish, R. and M. Livingston, 1992, "Determinants of the Call Option on Corporate Bonds," Journal of Banking and Finance (August), 687-703.
Kraus, A., 1973, "The Bond Refunding Decision in an Efficient Market," Journal of Financial and Quantitative Analysis (December), 793-806.
Marshall, W. and J. Yawitz, 1980, "Optimal Terms of the Call Provision on a Corporate Bond," Journal of Financial Research (Summer), 202-211.
Masulis, R., 1988, The Debt/Equity Choice, Cambridge, MA, Ballinger Publishing.
Mauer, D., A. Barnea, and C. Kim, 1991, "Valuation of Callable Bonds Under Progressive Personal Taxes and Interest Rate Uncertainty," Financial Management (Summer), 50-59.
Mitchell, K., 1991, "The Call, Sinking Fund, and Tenn-to-Maturity Features of Corporate Bonds: An Empirical Investigation," Journal of Financial and Quantitative Analysis (June), 201-222.
Myers, S., 1977, "Determinants of Corporate Borrowing," Journal of Financial Economics (November), 147-175.
Myers, S. and N. Majluf, 1984, "Corporate Financing and Investment Decisions When Firms Have Information that Investors Do Not Have," Journal of Financial Economics (June), 187-221.
Ogden, J., 1988, "A Rationale for the Sinking-fund Provision in a Quasicompetitive Corporate-Bond Market," Journal of Business Research (May), 197-208.
Robbins, E. and J. Schatzberg, 1986, "Callable Bonds: A Risk-Reducing Signalling Mechanism," Journal of Finance (September), 935-949.
Smith, C. and J. Warner, 1979, "On Financial Contracting - An Analysis of Bond Contracting," Journal of Financial Economics (June), 117-161.
Thatcher, J., 1985, "The Choice of Call Provision Terms: Evidence of the Existence of Agency Costs of Debt," Journal of Finance (June), 549-561.
The data on new offerings of corporate bonds and call schedules were gathered from The Wall Street Journal and cross-checked for accuracy against security descriptions given on Bloomberg Business News, Corporate Financing Week, and Investor's Dealers Digest. The objective was to select bonds that would provide a strong test of the agency theories of callable bonds.
To this end, we eliminated bonds with special features that a firm might include to mitigate other agency costs. In addition, we eliminated bonds issued by firms in regulated industries because regulated firms often face constraints on the types of securities that may be issued. For example, first mortgage bonds issued by utility holding companies subject to the SEC's jurisdiction under the Public Utility Holding Company Act of 1935 are required to be callable immediately after issue and have restrictions in their use of refunding provisions.
From an initial population of 3,310 bond offerings, the resulting sample consists of 958 straight bonds sold in the U.S. market by industrial firms between 1987 and 1991. Here are the sample elimination criteria (and the resulting sample size):
1) Bonds issued by utility, telecommunications, and financial companies. (Reduces the sample from 3,310 to 1,631 bonds.)
2) Convertible bonds and bonds with equity features. (Reduces the sample from 1,631 to 1,348 bonds.)
3) Bonds with variable coupon rates, including floating-rate notes, reset notes, step-up coupon bonds, and step-down coupon bonds. Bonds issued at an original issue discount, such as zero-coupon bonds and deferred-coupon bonds. (Reduces the sample from 1,348 to 990 bonds.)
4) Bonds not rated by Moody's and S&P. (Reduces the sample from 990 to 958 bonds.)
By matching the sample of 958 industrial straight bonds with data on ratings changes provided by Moody's Investors Service, we were able to construct a rating history for 889 of the industrial bonds. The rating history runs from the time of the offering to July 1, 1992. Because bonds with short maturities are less likely to have a rating change than are bonds with long maturities, the 60 bonds that matured before July 1992 were eliminated from the sample. Thus, the sample used for tests of upgrades and downgrades includes 829 bonds.
The estimations involving capital expenditures and assets are based on a sample of 886 bonds for which COMPUSTAT data were available.
Leland Crabbe is at Merrill Lynch & Co., New York, NY. Jean Helwege is at the Federal Reserve Bank of New York, NY.
|Printer friendly Cite/link Email Feedback|
|Author:||Crabbe, Leland E.; Helwege, Jean|
|Date:||Dec 22, 1994|
|Previous Article:||A simple approximation of Tobin's q.|
|Next Article:||Calling nonconvertible debt and the problem of related wealth transfer effects.|