Printer Friendly

Callable bonds revisited.

In light of the dramatic changes in the callable bond market, we reexamine the determinants of callable bonds. Using data from 1980-2003, we find that callable bonds are often issued by firms with both information asymmetry and underinvestment problems. However, risk-shifting does not appear to be a major factor. Furthermore, we find that interest rate hedging is an important factor for investment-grade bonds and when interest rates are high but not so for below-investment-grade bonds or when rates are low.

**********

The US callable bond market has experienced dramatic changes in the last two decades. First, callable bonds no longer dominate the public debt market (Crabbe, 1991). Prior to 1985, about 80% of all public corporate bonds included a call provision. This percentage has fallen to about 30% in recent years. Second, the callable bond market is now dominated by below-investment grade bonds, whereas investment-grade bonds used to be the norm. Along with these changes, the yield on the 10-year Treasury has fallen from more than 10% in the 1980s to about 5% in recent years. In light of these changes, this paper reexamines the determinants of callable bond issues.

Interest rate hedging is often thought to be the reason for callable bond issues. A call provision allows the issuing firm to refinance its debt at a lower rate if interest rates fall. In addition to this interest rate hypothesis, three theoretical explanations for callable bonds are often cited based on agency conflicts between shareholders and bondholders. These are: 1) the information asymmetry hypothesis, 2) the risk-shifting hypothesis, and 3) the underinvestment hypothesis. The information asymmetry hypothesis, developed by Barnea, Haugena, and Senbet (1980) and Robbins and Schatzberg (1986), argues that firms with information asymmetry problems issue callable bonds to either signal their positive future prospects or retain the option to refinance when their positive private information is revealed. The risk-shifting hypothesis, also developed by Barnea, Haugena, and Senbet (1980), argues that callable bonds discourage risk-shifting activities by limiting gains to shareholders from such activities. Barnea, Haugena, and Senbet (1980) illustrate that the value of the call provision, owned by shareholders, falls as the firm shifts into riskier assets, thus reducing shareholders' incentive for such activities. Finally, the underinvestment hypothesis, developed by Bodie and Taggart (1978) and Barnea, Haugena, and Senbet (1980), demonstrates that the Myers' (1977) underinvestment problem can be resolved with a call provision. Collectively, we call these three hypotheses the agency cost hypotheses of callable bonds.

Empirical tests of the agency cost hypotheses have mixed results. While several studies (Thatcher, 1985; Mitchell, 1991; Kish and Livingston, 1992) find some support for the hypotheses, other studies (Crabbe and Helwege, 1994) do not find any evidence consistent with the agency cost hypotheses. These mixed results are not surprising. First, most of the empirical studies use data from the 1970s and 1980s, decades where most bond issues were callable. Indeed, some of the studies include no straight bonds (Thatcher, 1985) or a limited number of straight bonds (Kish and Livingston, 1992). Second, the empirical studies examine each of the agency cost hypotheses in isolation. In a theoretical paper, Banko (2003) examines call provisions under multiple agency conflicts (e.g., a firm with both information asymmetry and underinvestment problems). The paper models conditions where a call provision is effective in resolving agency conflicts but also where other bond covenants are as effective (or better). (1)

These new ideas allow us to refine our empirical tests in two important ways. First, we reexamine the determinants of callable bonds using data on public bond offerings from 1980 to 2003. This sample period covers both high and low interest rate environments offering a better sample to test both the interest rate hypothesis and the agency cost hypotheses. Second, we examine the impact of combined agency conflicts on the issuance of callable bonds. While previous studies have tested multiple agency hypotheses, they examine each agency hypothesis separately. Alternatively, this study examines the role of call provisions in resolving a combination of two agency conflicts simultaneously.

Our empirical results find support for the interest rate hypothesis during high interest rate periods but not in low interest rate environments. In high interest rate settings, the use of call provisons is highly correlated with interest rate volatility, steepness of the yield curve, and changes in rates. On the other hand, in low interest rate environments, no significant relationship exists. In addition, we fail to find evidence supporting the agency cost hypotheses during high interest rate periods. Taken together, these results suggest that call provisions are primarily used to hedge interest rate risk when rates are high.

In terms of the agency cost hypotheses, we find the following. First, proxies for information asymmetry and underinvestment problems, when examined in isolation, are significant determinants of callable bonds. Alternatively, the potential for risk-shifting does not appear to be a primary determinant. Second, the combination of information asymmetry and underinvestment problems also significantly impacts the use of call provisions, while other combinations do not. This situation is particularly evident when management and outside investors differ in their respective expectations of future growth. A callable bond issuer tends to exhibit high historical sales growth, but outside investors have low expectations of the firm's growth opportunities as proxied by the market-to-book ratio.

In addition, we find that the motivation for issuing callable bonds is different between investment-grade and below-investment-grade firms. Investment-grade callable bonds are primarily used to hedge interest rate risk, while below-investment-grade callable bonds are mainly issued to alleviate agency conflicts.

Our findings are consistent with the changes in the callable bond market. As interest rates declined from historically high levels in the 1970s and 1980s, the need to hedge interest rate risk is no longer a concern for firms. Therefore, fewer firms issue callable debt. Furthermore, since below-investment-grade bond issuers tend to have severe agency problems (Alexander, Edwards, and Ferri, 2000), they are more likely to use call provisions to alleviate such problems. Consequently, the callable bond market is dominated by below-investment-grade bonds in recent years. Indeed, as Figure 1 illustrates, the rebound in the callable bond market in late 1990s and early 2000s coincides with the significant revival of the below-investment-grade bond market during that period. Our paper can also help to explain the mixed empirical results of previous studies. Since callable bonds are typically issued to hedge interest rate risk in high interest rate environments, regardless of the presence or absence of agency conflicts, it is not surprising that previous studies, using data in the 1970s and 1980s, cannot unambiguously test the agency cost hypotheses.

[FIGURE 1 OMITTED]

A major innovation of this paper is the examination of the combinations of agency problems. Previous studies only examine each agency conflict in isolation. Misleading conclusions may be drawn without consideration of the potential interaction between different agency problems. For example, a recent paper by Chen, Mao, and Wang (2007) also examines the determinants of callable bonds using the same sample period. This paper has many similar empirical findings. (2) Chen, Mao, and Wang (2007) conclude that firms with poor future investment opportunities (as proxied by market-to-book ratio, P/E ratio, and analyst earnings forecast) are more likely to issue callable bonds, thus rejecting the underinvestment hypothesis. However, these proxies for future investment opportunities are based on market prices or analysts' forecasts, reflecting the expectation of outside investors. In the absence of an information asymmetry problem, these market-based variables proxy future investment opportunities well. However, in the presence of a severe information asymmetry problem, a firm with future investment opportunities may not be able to convince outside investors of its good prospects, resulting in lower market-based expectations. Indeed, we demonstrate that firms with severe information asymmetry problems tend to have lower market-to-book ratios but higher historical sales growth. Because of the information asymmetry problem, outside investors cannot accurately assess and price the issuing firm's investment opportunities leading to the underinvestment dilemma. Call provisions are effective in solving this problem by allowing the firm to call and refinance its debt when its investment opportunities are realized. Thus, evidence of lower market expectations of future investment opportunities alone cannot unambiguously reject the underinvestment hypothesis. Indeed, we find that firms that have higher future investment opportunities but lower market expectations due to information asymmetry problems tend to issue callable bonds.

Another important distinction of this paper is the careful selection of proxies for each agency conflict. We use two proxies for each agency conflict. In addition, we distinguish these proxies from variables that can proxy for multiple agency conflicts. For example, bond ratings and debt ratios are often used to proxy agency conflicts (Kish and Livingston, 1992). The problem, though, is that bond ratings and debt ratios cannot separate the three agency hypotheses that all predict high default risk for callable bonds. As a result, the findings that callable bond issuers have lower ratings and higher debt ratios cannot unambiguously test and separate the three agency hypotheses. Furthermore, it is problematic to use bond ratings and debt ratios to proxy for the risk-shifting problem as they only measure the incentive of the issuing firm to engage in risk-shifting activities, but not its ability to do so. We use two variables that proxy the ability of issuing firms to engage in risk-shifting activities. We find no correlation between callable bonds and the potential for (or ability to engage in) risk-shifting activities, even for below-investment-grade bonds whose issuers have stronger incentives to engage in such activities. Thus, callable bonds are not a primary tool to alleviate the risk-shifting problem. This conclusion contrasts with that of Chen, Mao, and Wang (2007), who use bond ratings and debt ratios to proxy for the potential risk-shifting problem. The conclusion that callable bonds are used to solve this agency conflict based on lower bond ratings and higher debt ratios of callable bonds issuers is misleading.

The remainder of the paper is organized as follows. In the next section, we examine the previous literature regarding the determinants of call provisions. We review theoretical models as well as empirical tests of these models. Section II describes our sample and descriptive statistics. Section III presents our empirical findings and Section IV provides our conclusions.

I. Previous Research

Early research attempts to explain call provisions by modeling divergent interest rate expectations (Bowlin, 1966; Jen and Wert, 1967) and tax considerations (Boyce and Kalotay, 1979). However, following Jensen and Meckling (1976), a growing literature evolved into a general theory that we term the agency cost hypotheses for callable bonds. When firms seek external debt financing to fund investment projects, debt financing may create distortions in firm behavior, leading to suboptimal investment decisions. The agency cost hypotheses argue that embedded call provisions in callable bonds are useful to issuing firms and bond investors as they resolve agency conflicts caused by debt financing.

Bodie and Taggart (1978) and Barnea, Haugena, and Senbet (1980) are the first to exploit the agency story by exploring the ability of callable bonds to resolve the underinvestment problem. The underinvestment problem occurs in firms with positive NPV projects but high default risk. When the firm invests in its positive NPV project, gains from the project reduce the default risk of its outstanding debt issues, benefiting bondholders. As a result, equity holders may bypass positive NPV projects (Myers, 1977). In Bodie and Taggart (1978), a call provision allows the firm to call the bond and refinance it at a lower interest rate reflecting the lower default risk. Therefore, the shareholders will reap the full gains from the positive NPV project. By doing so, the issuing firm can resolve the potential underinvestment problem. The empirical implications of the underinvestment hypothesis are that: 1) callable bond issuers have more growth/investment opportunities or positive NPV projects, and 2) callable bond issuers have higher default risk. (3)

Another agency conflict is the risk-shifting problem. Shareholders of a leveraged firm can be viewed as owners of a call option on the firm's assets with the value of debt as the strike price (Merton, 1974). By substituting riskier projects for safer ones, managers can increase the volatility of the firm's asset value, thus increasing the value of this call option or stock price. The gain to shareholders comes at the expense of the bondholders. Higher volatility of the firm's asset value increases the default risk, lowering the bond value. Barnea, Haugena, and Senbet (1980) demonstrate that this problem can be mitigated by using an embedded call provision in the firm's debt. The value of the embedded call provision decreases if the price of the bond decreases due to higher default risk. Because this call provision is owned by shareholders, they will share the pain with bondholders if the default risk of the firm increases due to risk-shifting activities. Indeed, if designed properly, there will be no net gains to shareholders from risk-shifting activities. Thus, the embedded call provision can mitigate or eliminate shareholders' incentive to engage in risk-shifting activities.

The risk-shifting hypothesis has two testable empirical implications. First, callable bond issuers have greater default risk (or stronger incentives to engage in risk-shifting activities). Within the Black and Scholes (1973) options pricing framework, the value of a deep-in-the-money call option is less affected by changes in volatility than an at-the-money call option. Lower (higher) default risk implies that the shareholders' call option on the firm's assets is deep-in-the-money (at-the-money). Thus, an increase in the volatility of the firm's asset value will result in smaller (larger) gains to shareholders if the firm has lower (higher) default risk. Therefore, firms with higher default risk have a stronger incentive to engage in risk-shifting activities. Eisdorfer (2008) finds strong evidence of risk-shifting behavior for financially distressed firms. Second, callable bond issuers have more flexible asset structures and higher free cash flows (or the ability to engage in risk-shifting activities). Firms with fixed, illiquid assets cannot easily redeploy their resources to pursue other riskier projects (Prowse, 1990). Alternatively, firms with more free cash flows can easily divert their cash flow from investing in traditional/safer projects to riskier opportunities.

Barnea, Haugena, and Senbet (1980) and Robbins and Schatzberg (1986) investigate the use of call provisions in resolving information asymmetry problems. In their models, management possesses positive private information that cannot be credibly revealed to outside investors. Barnea, Haugena, and Senbet (1980) demonstrate that call provisions allow firms to refinance the bond at lower interest rates after the revelation of good news. Robbins and Schatzberg (1986) suggest that firms can use a callable bond to signal their positive private information to outside investors, thereby lowering their borrowing costs. While the two models differ in their use of callable bonds to resolve information asymmetry, both of them assume the presence of information asymmetry. Thus, a testable empirical implication of both models is that callable bond issuers are more opaque. In addition, the information asymmetry hypothesis also implies that callable bonds issuers have greater default risk. Positive private information tends to have a larger impact on riskier bonds. Refinancing a BB-rated bond with a BBB-rated bond after the revelation of good news probably saves a firm more money than refinancing an AA-rated bond with an AAA-rated bond.

Table I summarizes the testable empirical implications of the three agency cost hypotheses. It is important to note that the three agency cost hypotheses all imply that callable bond issuers have higher default risk. Accordingly, finding greater default risk among callable bond issuers is consistent with all three hypotheses, but it cannot unambiguously separate and confirm the presence of any specific agency problem. Conversely, finding lower default risk on callable bonds does reject all three hypotheses.

In terms of empirical studies, Thatcher (1985) presents the first test of the agency cost hypotheses. The paper examines two different types of protections against a bond call: 1) refunding protection and 2) call protection. The paper finds that firms with high agency costs retain their rights to call the bond during the refunding protection period, but firms with lower agency costs tend to offer bondholders more comprehensive call protection. Thatcher (1985) argues that the flexibility of the weaker protection, the refunding protection, used by firms with high agency costs, helps alleviate the underinvestment problem. Mitchell (1991) finds that firms with information asymmetry problems are more likely to use call provisions. Kish and Livingston (1992) conduct a comprehensive empirical study regarding the determinants of callable bonds from 1977 to 1986 to test several competing theories. They find that interest rate levels are a major determinant of call provisions. Firms with higher default risk and higher growth rates are also more likely to issue callable bonds. Crabbe and Helwege (1994) examine the impact of different agency costs on the decision to issue callable bonds. They find that callable bonds tend to have lower bond ratings and longer maturities. However, they find that callable bonds are no more likely to be upgraded than straight bonds, evidence against the information asymmetry hypothesis. Further, they do not find any correlation between the aggregate business investment and the number of callable bonds issued during their sample period. Finally, they do not find any support for the risk-shifting hypothesis.

In addition to the mixed empirical results, the callable bond market experienced significant changes during the 1990s. The percentage of corporate bonds that are callable dropped dramatically from the 1980s (Crabbe, 1991). Further, the callable bond market is now dominated by below-investment-grade bond issues. Along with these changes, interest rates decreased significantly. Furthermore, significant growth in the interest rate derivatives market offered firms hedging opportunities not available in the 1970s and 1980s. Guntay, Prabhala, and Unal (2002) argue that the rapid development of the interest rate derivatives market and easy access to derivative products allow issuing firms to manage interest rate risk without embedding a call provision in their debt issues. Finally, Banko (2003) offers new theoretical results that shed light on the mixed empirical results to date. The paper demonstrates that the usefulness of callable debt in resolving agency problems is more complicated than previously modeled. In particular, it demonstrates that: 1) a call provision is not the only mechanism to solve a stand-alone agency conflict, and 2) a call provision is most effective in alleviating the combined situation of information asymmetry and underinvestment problems.

II. Sample, Variables, and Descriptive Statistics

A. Data Collection

We collect data on fixed-rated, nonfinancial, US domestic callable and straight bond issues from the Thomson Financial SDC database. SDC does not have a specific data item to denote callable bonds. We use the SDC data item "Call Protection" to classify callable bonds. Bond issues with "Non-Call Life" in its "Call Protection" data item are classified as straight bonds and all others are classified as callable bonds. (4) The sample period covers 1980-2003. We exclude perpetual bonds, nonrated bonds, bonds with credit enhancements, and bonds issued by subsidiaries of parent companies. We also exclude make-whole callable bonds, bonds with poison call options, etc. Some firms have multiple bond issues within the same month that have the same call feature. Since these multiple bonds do not convey additional information, we select the earliest issue among the multiple issues. The issuer is matched with Compustat annual data one year prior to the bond issuance. We have 2,231 issues that have complete accounting information. Matching with I/B/E/S data reduces the sample to 2,109 issues, of which 1,114 are straight bonds and 995 are callable bonds.

Figure 1 illustrates the percentage of callable bonds in our sample from 1980 to 2003. Also plotted in Figure 1 are yields on the 10-year Treasury and the percentage of below-investment-grade bonds in the sample. The left axis is the percentage of bonds in the sample that are callable or below-investment grade. The right axis is the yields on the 10-year Treasury. Like Crabbe (1991), we find that callable bonds dominate the corporate bond market in the early 1980s, decline in the early 1990s, but rebound in the late 1990s and early 2000s. The decline in the issuance of callable bonds seems to be strongly correlated with the significant decline in the yield on the 10-year Treasury from 1980 to mid 1990s. The rebound of the callable bonds in the late 1990s and early 2000s coincides with the significant growth of the below-investment-grade bond market. Indeed, the majority of the callable bonds in the late 1990s and early 2000s are below-investment-grade bonds.

B. Variable Definitions

The Appendix provides a complete list of all the variables used in this study. Most are commonly used variables in similar studies. However, an important aspect of this paper is the careful classification of the various agency conflicts and the variables used as proxies. For these, we provide a detailed explanation.

1. Proxies for Information Asymmetry

Firm size is often used as a proxy for information asymmetry. Larger firms are followed by more analysts and are under more scrutiny from the financial media than their smaller counterparts. Further, Merton (1987) argues that large firms face fewer information asymmetry problems due to the greater awareness of the investing public. Also, large firms generally access capital markets more frequently and reveal more information to investors to lower their cost of capital. Another proxy for information asymmetry is analyst forecast errors. Large information asymmetry problems make it difficult for analysts to assess the value of the firm. As a result, analysts have a harder time forecasting earnings. Thomas (2002) uses this proxy to examine information asymmetry problems of diversified firms. Krishnaswami and Subramaniam (1998) also use this proxy to study changes in information asymmetry problems before and after subsidiary spin-offs. (5)

Based on these two proxies, we create a single variable for information asymmetry. Information Asymmetry equals zero if the firm size is greater than the median firm size and the earnings forecast error is less than the median forecast error. It equals two if the firm size is below the median and the earnings forecast error is above the median. Otherwise, the variable equals one. Thus, a high value for Information Asymmetry is associated with severe information asymmetry problems. (6)

2. Proxies for Risk-Shifting

Shome and Singh (1995) and Prowse (1990) use a firm's discretionary assets to proxy for risk-shifting potential, where discretionary assets are equal to 1 - (fixed assets/total assets). The idea behind this proxy is simple. Firms with a low level of discretionary assets cannot easily engage in risk-shifting activities. In addition to discretionary assets, we also use free cash flow as a proxy. Firms with more free cash flow can easily divert their cash flow from investing in traditional/safer projects to more risky projects. Following Freund, Prezas, and Vasudevan (2003) and Lehn and Poulsen (1989), we define free cash flow (FCF) as the ratio of the firm's operating free cash flow divided by the book value of its total assets in the year prior to the bond issuance.

Based on these two proxies, we again create a single variable, Risk-Shifting Potential. It is equal to zero if the discretionary assets and free cash flow for the firm are less than their respective medians. If both are greater than their medians, the variable equals two. Otherwise, the variable equals one. Thus, high Risk-Shifting Potential is associated with firms more likely to engage in risk-shifting activities.

3. Proxies for Growth Opportunities

Our first proxy for future growth/investment opportunities is the historical growth rate of sales. Firms with high sales growth are often young firms in new industries or untapped market segments. Chan, Karceski, and Lakonishok (2003) demonstrate that persistence in sales growth exists over the past 50 years; that is, firms with high historical growth in sales are more likely to exhibit above-average future growth in sales as well. Pagano, Panetta, and Zingales (1998) use historical growth in sales to proxy for a firm's growth opportunities.

We also use the firm's market-to-book ratio (MB Ratio) to proxy for future growth/investment opportunities. Myers (1977) argues that the market value of a firm can be divided into two parts: 1) the present value of existing assets and 2) the present value of future growth opportunities. Thus, the MB Ratio is often used to proxy for future growth opportunity because the book value of assets is a proxy for existing assets and the market value of assets is a proxy for both existing assets and growth opportunities. A high MB Ratio indicates that a firm has many investment opportunities relative to the assets currently in place.

Although both historical sales growth and MB Ratio proxy for future growth/investment opportunities, they may reflect different expectations. Historical sales growth is internally focused and achieved by the firm's management. Thus, it is a better proxy for management's expectation of future growth. (7) Alternatively, the MB Ratio is externally focused. The current stock price is a key determinant of the MB Ratio, which incorporates investors' expectations of the firm's future growth. As a result, the MB Ratio is a better proxy for investors' expectation of future growth.

We use three dummy variables to capture different growth scenarios. The first variable is HS (high sales growth), which is equal to one for firms with above-median historical sales growth of the sample, and zero otherwise. A firm with high sales growth probably has more growth/investment opportunities. We expect that these firms are more likely to issue callable debt as the potential for an underinvestment problem is also high. The second dummy variable, HSLR (high sales growth and low MB Ratio), is equal to one for issuing firms that have an above-median historical sales growth, but a below-median MB Ratio, and zero otherwise. Issuers in this group achieved high historical growth in sales, but outside investors have low expectations of future growth/investment opportunities. Since the potential future growth (as perceived by management) is not priced in the firm's securities, the underinvestment problem is likely to be severe. Thus, we conjecture that these firms are likely to issue callable bonds.

The last dummy variable, HSHR (high sales growth and high MB Ratio), is equal to one for issuing firms that have historical sales growth and an MB Ratio above their respective medians, and zero otherwise. A firm with HSHR equal to one is expected to have high growth opportunities. However, higher growth opportunities do not necessarily imply an underinvestment problem, particularly if investors fully anticipate and price the future growth. Therefore, the impact of this variable on the use of callable debt is left as an empirical question.

4. Proxies for General Agency Conflicts

Bond ratings and debt ratios are often used in the literature to proxy agency conflicts (Kish and Livingston, 1992). However, as discussed earlier, the three agency cost hypotheses all imply that callable bond issuers have higher default risk. As such, we use the bond ratings and debt ratios as proxies for all three agency conflicts. Empirical findings of lower bond ratings and higher debt ratios for callable bond issuers are consistent with all three hypotheses but cannot unambiguously separate them. Alternatively, opposite findings will reject all three hypotheses.

Following Kish and Livingston (1992), we create three dummy variables for bond ratings: 1) High Rating, 2) Moderate Rating, and 3) Low Rating. High Rating (Low Rating) is equal to one for bonds rated AAA or AA (below BBB) by Moody's and zero otherwise. Note that the Low Rating dummy variable corresponds to below-investment-grade bonds. Moderate Rating is equal to one for bonds rated A or BBB by Moody's and zero otherwise. (8) Debt ratio is defined as the firm's total debt divided by total assets.

C. Descriptive Statistics

Table II presents the descriptive statistics of the sample. Panel A reports the characteristics of the issuing firms. The mean (median) firm size is $8 ($3) billion. Firm size is adjusted for inflation expressed in 1983 dollars. The mean market-to-book ratio (MB Ratio) is 1.57 and the mean debt ratio is 33%. Panel B presents the bond issue characteristics. The average bond rating of the sample is 4.5, or between BBB and A. Furthermore, 22% of the bonds are of High Rating (AAA or AA rated by Moody's) and 18% are of Low Rating (below investment grade). Panel C reports the interest rate environment variables.

Some of the variables, noticeably Analyst Forecast Errors and Free Cash Flows, have very large kurtosis suggesting that these variables have large outliers. Such outliers may bias our results if we use the raw data to proxy for agency conflicts. Therefore, we rely more on the categorical variables derived from the raw data to proxy for agency conflicts.

Table III compares callable bonds and straight bonds. First, we compare the two types of bonds for the whole sample. Firms issuing callable bonds are smaller in firm size ($5.52 billion vs. $10.38 billion) and the difference is statistically significant. In addition, the Analyst Forecast Errors of callable bond issuers are significantly higher (2.62%) than straight bond issuers (1.36%). Consistent with this, the mean Information Asymmetry in Panel D is significantly higher for callable bonds than straight bonds, suggesting that callable bond issuers have more information asymmetry problems. Callable bond issuers also have significantly lower discretionary assets (43% vs. 50%) and a significantly lower free cash flow (6.36% vs. 7.41%) than straight bond issuers. The mean Risk-Shifting Potential in Panel D is significantly lower for callable bonds, indicating that callable bond issuers have a relatively low risk-shifting potential, inconsistent with the risk-shifting hypothesis.

The historical sales growth rate for callable bond issuers (13.23%) is significantly higher than that of the straight bond issuers (10.13%). However, the MB Ratio for callable bond issuers (1.36) is significantly lower than that of the straight bond issuers (1.76). Along the same line, the mean HSLR (high sales growth, low MB ratio) is higher for callable bonds, indicating that outside investors expect lower growth/investment opportunities from callable bond issuers in spite of the higher historical sales growth. Callable bond issuers may have (or at least management believes that they have) greater future sales and investment opportunities as indicated by their historical growth rates, but the firms cannot credibly convince outside investors. As a result, security prices reflect this difference in opinion. Such underestimation by outside investors (or overestimation by management) may prompt management to issue callable bonds. (9) This is consistent with the prediction of Banko (2003) that a combination of underinvestment and information asymmetry problems can best be resolved with a call provision.

Callable bond issuers tend to have higher debt ratios and lower bond ratings, indicating greater default risk. In addition, over 27% of callable bonds are below investment grade, while only 10% of straight bonds are below investment grade. Finally, the interest rate environment variables indicate that callable bonds are issued when the yield on the 10-year Treasury (T-10 Yield) is high, the yield curve is flatter, and interest rates are volatile. Furthermore, callable bonds are more prevalent following larger decreases in the Treasury yield, which continues to offer support for the interest rate hypothesis.

Next, we divide the sample into two groups based on the median 10-year Treasury yield at the time of the bond issuance. Issues offered when the T-10 Yield is above (below) the median are classified as high (low) interest rate environment (HIRE/LIRE) issues. (10) Thus, by construction, HIRE issues have an average T-10 Yield of 9.33% at the time of issuance, while the average T-10 Yield for LIRE issues is 5.79%. The number of callable bonds decreases from 648 issues in the HIRE periods to 347 issues in the LIRE periods. The simultaneous decreases in interest rates and the number of callable bond issues support the interest rate hypothesis. This is consistent with the findings of Guntay, Prabhala, and Unal (2002).

The impact of interest rates on the use of callable bonds is very different between the high and low interest rate environments. Similar to the whole sample, callable bonds in the HIRE subsample are generally issued when: 1) interest rates are higher and more volatile, 2) the yield curve is flatter, and 3) interest rates have declined significantly. In contrast, the interest rate environment does not seem to have an impact on the use of callable bonds in the LIRE period. These findings suggest that hedging interest rate risk is a determinant only when interest rates are at historically high levels, such as the 1980s. When interest rates are relatively low and stable, interest rate risk is no longer a concern for issuing firms.

Another interesting difference between the two groups is the ratings of callable bonds. Below-investment-grade bonds account for 11.42% of the callable bonds in the HIRE group but more than 57% of the callable bonds in the LIRE group. Callable bond issuers in the LIRE period exhibit higher default risk. This is confirmed by the significantly higher debt ratio. As the agency cost hypotheses predict that firms with higher default risk are more likely to issue callable bonds, the evidence of higher default risk for callable bond issuers in the LIRE group suggests that these callable bonds are driven more by concerns for agency conflicts than by the need to hedge interest rate risk.

III. Empirical Results

A. Univariate Analysis

1. Information Asymmetry and Risk-Shifting

First, we break our sample into three subsamples according to whether Information Asymmetry is equal to zero, one, or two, corresponding to low, medium, and high information asymmetry problems, respectively. Panel A of Table IV reports the percentage of callable bonds in these subsamples. The use of callable bonds increases monotonically with the severity of the information asymmetry problem. The difference in the percentages of callable bonds between the high and low Information Asymmetry subsamples is 27.77%, statistically significant at the 1% level. Next, we break the sample, following the same process, into three additional subsamples according to low, medium, and high Risk-Shifting Potential. Panel B of Table IV reports the percentage of callable bonds in these subsamples. The use of callable bonds significantly decreases with Risk-Shifting Potential. This pattern suggests that risk-shifting does not drive the use of call provisions.

Next, we examine the interaction of the two agency conflicts. We create nine subsamples along the two proxies for agency conflicts. Panel C of Table IV reports the percentage of callable bonds in each of the nine subsamples. At all levels of Risk-Shifting Potential, issuing firms in the high Information Asymmetry category are more likely to use callable bonds than issuing firms with low or medium Information Asymmetry. Alternatively, the use of callable bonds seems to be negatively correlated with Risk-Shifting Potential at all levels of Information Asymmetry. This reinforces the conclusions from Panels A and B. Information asymmetry problems drive the use of call provisions, while risk-shifting potential does not.

2. Information Asymmetry and Underinvestment Problem

In this section, we examine the impact of potential future growth opportunities on the use of call provisions. Since the two proxies for future growth reflect different expectations (management vs. outside investors), we first break the sample into two subsamples: 1) low sales growth (LS) and 2) high sales growth (HS) based on the median sales growth of the sample. In Panel A of Table V, we report the percentage of callable bonds in each group.

Close to 52% of the bonds in the HS subsample are callable, while around 43% of the bonds in the LS subsample are callable. The difference is significant at the 1% level. Thus, firms with more growth opportunities (at least as perceived by management) are more likely to issue callable bonds. To further examine the interaction of underinvestment and information asymmetry, we classify the high and low sales growth groups according to low, medium, or high Information Asymmetry. We have the following findings. First, there is no significant difference in the use of callable bonds between the HS and LS groups at the low level of Information Asymmetry (Panel A, Column 2). However, the HS group has a significantly higher percentage of callable bonds than the LS group at the medium and high levels of Information Asymmetry (Columns 3 and 4), and the difference between the HS and LS groups increases with Information Asymmetry. This suggests that in the absence of information asymmetry problems, higher future growth opportunities do not drive the use of callable bonds. However, the impact of future growth opportunities on the use of callable debt increases with the severity of the information asymmetry problem. Second, the use of call provisions increases monotonically with Information Asymmetry in both groups. However, the impact of information asymmetry problem on the HS group seems to be much stronger. For the HS group, the difference in the percentages of callable bonds between the high and low Information Asymmetry subsamples is more than 39%. For the LS group, the difference is only about 14%. Therefore, it seems that growth opportunities coupled with information asymmetry problems will lead firms to include call provisions in their debt offerings.

Panel B of Table V separates the HS group into two subgroups: 1) the above-median sales growth but below-median MB Ratio (HSLR) subsample, and 2) above-median sales growth and above-median MB Ratio (HSHR) subsample. We find that 36.14% of the HSHR issues are callable, while 68.95% of the HSLR issues are callable. HSLR issuers are more likely to have significant underinvestment problems (at least as perceived by management) and information asymmetry problems and, as such, are more likely to issue callable bonds. (11) Alternatively, HSHR issuers have greater growth/investment opportunities and outside investors have priced these investment opportunities in the firm's securities. Thus, these firms do not necessarily have underinvestment concerns. Therefore, they may or may not issue callable bonds.

Next, we examine the impact of information asymmetry problems on the use of callable bonds for firms with high sales growth. We further break the two HS subsamples into six subsamples based on the issuing firm's Information Asymmetry. For the HSHR subsamples, the use of callable bonds increases monotonically with Information Asymmetry. Only 24% of the issues with low Information Asymmetry include a call provision, but over 66% include a call provision when Information Asymmetry is high. This is consistent with Banko's (2003) theoretical finding that callable bonds are most effective in resolving the combination of information asymmetry and underinvestment problems. In the HSLR case (Panel B, second row), the degree of Information Asymmetry does not appear to affect the use of callable bonds. This suggests that the information asymmetry problem is a secondary determinant of call provisions after the potential underinvestment problem.

B. Multivariate Regression Analysis

In this section, we analyze the determinants of callable bonds in multivariate probit regressions. The dependent variable is a dummy variable equal to one for a callable bond and zero for a straight bond. The explanatory variables include proxies for agency conflicts, interest rate environment variables, and other control variables.

In Model 1 of Table VI, we include only the proxies for the agency conflicts and 31 industry dummies. (12) The base case is issues with low historical sales growth (LS). The coefficients on both Information Asymmetry and HS are positive and statistically significant, suggesting that information asymmetry problems and high sales growth increase the probability of callable bonds. (13) The coefficient on Risk-Shifting Potential is negative and significant, inconsistent with the risk-shifting hypothesis.

In terms of bond ratings, we use the High Rating as the base case. The coefficient on Lower Rating is positive and significant, indicating that below-investment-grade bonds are more likely to have call provisions than AAA- or AA-rated bonds, consistent with all three agency cost hypotheses. Interestingly, the coefficient on Moderate Rating is negative and significant, indicating that the correlation between default risk and callable bonds is not linear. Indeed, AAA- or AA-rated bonds are more likely to have a call provision than A- or BBB-rated bonds. (14) Kish and Livingston (1992) have similar findings. These findings suggest that default risk is a major determinant of below-investment-grade callable bonds but not for investment-grade callable bonds. The coefficient on Debt Ratio is negative and significant, which is inconsistent with the agency cost hypotheses. One explanation for this finding is that the Low Rating Dummy variable is a better proxy for default risk. When we leave out the rating dummy variables, the coefficient on Debt Ratio is no longer significant. Another explanation for the negative coefficient on Debt Ratio is the simultaneous impact of lower interest rates on the firm's use of debt financing and the call provision. When interest rates are lower, firms tend to use more debt financing (Barry et al., 2008). Table III demonstrates that the average Debt Ratio is higher in the Low Interest Rate Environment (LIRE) subsample. At the same time, firms issue fewer callable bonds when interest rates are lower, creating a spurious negative correlation between Debt Ratio and the use of the call provision. (15)

Model 2 further divides the HS group into HSHR and HSLR subgroups. The coefficient on HSLR is positive and significant, while the coefficient on HSHR is not significant. Thus, it seems that the positive and significant coefficient on HS in Model 1 is driven by the HSLR issues. In other words, not all firms with high historical sales growth tend to issue callable bonds. Only firms with high historical sales growth but lower MB Ratios are likely to issue callable bonds.

In Model 3, we examine the impact of combined agency conflicts on the use of the call provision. Specifically, we create two interaction terms: 1) Information Asymmetry * Risk-Shying Potential and 2) Information Asymmetry * HSHR. The coefficient on the first interaction term is not significant, and the coefficient on the second interaction term is positive and significant. Further, the coefficient on Information Asymmetry is no longer significant, suggesting that information asymmetry problems alone are not a typical characteristic of firms issuing callable bonds. Only firms that have potential investment opportunities coupled with information asymmetry problems are more likely to issue callable bonds. Note that after the inclusion of the interaction terms, the coefficient on HSHR becomes negative and significant, suggesting that firms with higher growth/investment opportunities but no information asymmetry problems are actually less likely to issue callable bonds than firms with lower historical sales growth. This further indicates that potential future growth opportunity, in the absence of an information asymmetry problem, does not necessarily lead firms to issue callable bonds.

In Model 4, we include the interest rate environment variables and other control variables. The coefficients on both the T-10 Yield and the standard deviation of the T-10 Yield are positive and significant, indicating that callable bonds are more likely to be issued when the general level of interest rates is high and volatile. Among the other control variables, the coefficient on Maturity is positive and significant. (16) This finding is also consistent with the interest rate hypothesis. Longer term bonds usually have higher interest rate risk. Thus, issuing firms are more likely to embed a call provision in longer term bonds. The coefficient on ROA is also positive and significant, signifying that profitable firms are more likely to issue callable bonds.

C. Interest Rate Environment and Callable Bonds

The univariate analysis indicates that the motivation for issuing callable bonds may be different in a low and stable interest rate environment than in a high and volatile interest rate environment. In this section, we further examine this issue using the multivariate regression model. We rerun Model 4 on the two interest rate subsamples, HIRE and LIRE. The results are reported in the first two columns of Table VII. There are important differences between the two subsamples. First, the coefficients on the interest rate environment variables are all significant in the HIRE subsample. In the LIRE group, the coefficients are either insignificant or have the "wrong" sign. Second, none of the coefficients on the proxies for agency conflicts is significant in the HIRE subsample. For the LIRE subsample, these coefficients are similar to the whole sample. These findings again support the idea that callable bonds are used to hedge interest rates when rates are high. In such times (e.g., the 1980s), most bonds contain call provisions regardless of the presence or absence of any agency conflict. This further explains the mixed results of empirical studies conducted in the 1980s and early 1990s based on data from a predominantly HIRE period. Furthermore, with the development of interest rate derivatives, firms now have convenient tools for managing their exposure to interest rate risk (Guntay, Prabhala, and Unal, 2002). The need to hedge interest rate risk via a callable bond is no longer necessary. Instead, callable bonds are primarily used to alleviate potential agency conflicts.

D. Credit Ratings and Callable Bonds

Theories suggest that agency conflicts are more severe and of greater concern to below-investment-grade bond issuers than firms with lower default risk. As such, it is plausible that firms of lower credit quality issue callable bonds to alleviate agency conflicts, while high credit quality firms issue callable bonds to hedge interest rate risk. To test this hypothesis, we rerun a modified Model 4 regression on the three credit rating subsamples: 1) High Rating, 2) Moderate Rating, and 3) Low Rating. As defined earlier, High Rating (Moderate Rating) issues are AAA- or AA- (A- or BBB-) rated bonds. Low Rating issues are below-investment-grade bonds (rated BB or below). The results are reported in the last three columns of Table VII.

For the High Rating subsample, none of the coefficients on the agency conflict proxies are significant, but three of the interest rate environment variables are significant at the 1% or 5% level. For the Moderate Rating subsample, only one agency conflict proxy (HSHR) is significant, but two interest rate environment variables are significant at the 1% level. In addition, the coefficient on Bond Rating, a proxy for general agency conflicts, is not significant in either subsample. These results strongly suggest that hedging interest rate risk is the main motivation for issuing investment-grade callable bonds.

For the Low Rating subsample, we leave out the two interaction terms from the regression to avoid a potential multicollinearity problem. (17) The results on the below-investment-grade bonds are quite different from the investment-grade bonds. The coefficients on Information Asymmetry and HSLR are significantly positive. In addition, the coefficient on HSHR is positive and marginally significant, in contrast to the significantly negative coefficients for the whole sample and the Moderate Rating subsample. The reason for the difference is that about 67% of below-investment-grade HSHR issues also have high Information Asymmetry, while less than 20% of investment-grade HSHR issues have high Information Asymmetry. This result further confirms earlier findings that callable bonds are used to resolve a combination of information asymmetry and underinvestment problems.

In terms of interest rate environment variables, only one is marginally significant in the Low Rating subsample, suggesting that below-investment-grade bond issuers do not use call provisions to hedge interest rate risk. Interestingly, the coefficient on Maturity is positive and significant for High and Moderate Rating subsamples but insignificant for the Low Rating subsample, further supporting the idea that hedging interest rate risk is a major motivation for investment-grade callable bonds, but not for below-investment-grade bonds.

Overall, the findings suggest that investment-grade callable bonds are primarily issued to hedge interest rate risk, while below-investment-grade callable bonds are mainly issued to alleviate agency conflicts. This is consistent with the fact that a majority of callable bonds issued in recent years are below investment grade.

Note that the coefficient on Risk-Shifting Potential is not significant in the Low Rating subsample. This finding can address a potential concern about our test of the risk-shifting hypothesis. The proxies we use to construct Risk-Shifting Potential (Discretionary Asset and Free Cash Flow) measure the ability (or potential) of issuing firms to engage in risk-shifting activities but not the incentive for such activities. A concern is that firms with higher free cash flow and discretionary assets are financially strong with low default risk; as such, they do not gain from risk-shifting activities. Thus, firms with high Risk-Shifting Potential do not have strong incentive to engage in such activities. However, this is not the case for below-investment-grade bond issuers, who have greater incentive to engage in risk-shifting activities due to higher default risk. Lack of significant correlation between the Risk-Shifting Potential and the use of call provisions for below-investment-grade bonds indicates that even firms with both incentive and ability to engage in those activities do not issue callable bonds. This is inconsistent with the risk-shifting hypothesis.

E. Firm Age and Callable Bonds

The economic literature has documented an inverse relationship between firm age and firm growth rate (Evans, 1987), suggesting that younger firms have greater investment and growth opportunities. Our sample has a similar pattern. Firms with above-median historical sales growth have an average age of 33 years (from the listing year) while firms with below-median historical sales growth have an average age of 44 years. Average firm age of the whole sample is about 39 years. In addition, younger firms may also have more severe information asymmetry problems as they do not have a track record or have few analysts or media coverage (Hyytinen and Pajarinen, 2008). In our sample, firms with the most severe information asymmetry problem (or those issues that have an Information Asymmetry variable equal to two) have an average firm age of 30 years while other firms have an average age of about 42 years. Hence, we do not use firm age as a proxy for either the underinvestment problem or the information asymmetry issue because it cannot distinguish between the two hypotheses.

However, firm age is an ideal proxy for the combined information asymmetry and underinvestment problems. Given that younger firms are more opaque but have more growth/investment opportunities, it is not surprising that callable bond issuers are younger (average firm age of 35 years) than straight bond issuers (average firm age of 41 years). The difference is statistically significant. In addition, we run the main probit regression (Model 4) with firm age as an additional variable. The coefficient on firm age is negative and significant, consistent with previous findings that firms with high growth opportunities, but severe information asymmetry problems are more likely to issue callable bonds.

F. Rule 144A Issues and Callable Bonds

An interesting development in the US bond market since the early 1990s is the significant growth of Rule 144A issues. Rule 144A issues are not registered with the SEC but can be traded among large institutional investors or Qualified Institutional Buyers (QIB) (see Fenn, 2000; Livingston and Zhou, 2002). Due to the lack of registration and lower standards for disclosure, firms with severe information asymmetry problems and other agency conflicts may prefer to issue in the Rule 144A market. Indeed, Livingston and Zhou (2002) report that over half of the Rule 144A bonds in their sample are issued by firms accessing the bond market for the first time and almost a quarter of the issuing firms do not file periodic disclosures with the SEC. Hence, a Rule 144A issue may signal severe information asymmetry problems and other agency conflicts. Therefore, we examine the relationship between Rule 144A issues and call provisions in this section.

The 275 Rule 144A issues in the LIRE subsample have a significantly higher level of Information Asymmetry than public issues. (18) In addition, a significantly higher percentage of Rule 144A issues are HSLR issues. (19) Given the higher level of agency conflicts, it is not surprising that over 63% of Rule 144A issues are callable as compared to only 22% of public issues. (20) Furthermore, we rerun the main regression (Model 4) on the LIRE subsample with a Rule 144A dummy variable. The coefficient on the dummy variable is significantly positive, indicating that Rule 144A issues are more likely to include a call provision. This further supports the hypothesis that call provisions are used to alleviate agency conflicts during low interest rate environments.

G. Robustness Checks

Our sample excludes issues that do not have analyst earnings forecasts. It is likely that these bond issues have severe information asymmetry problems due to the lack of analyst coverage. Thus, the sample may exclude bond issues with the most severe information asymmetry problems. An examination of the 122 observations with no analyst earnings forecasts indicates that about 67% of them also have below-median Firm Size, suggesting that these issues may have larger information asymmetry problems. (21) In addition, about 75% of them are callable bonds. To check whether exclusion of these issues has a significant impact on our results, we expand the sample by including these issues. We classify these issues as an above-median Analyst Forecast Errors group and reconstruct the Information Asymmetry variable accordingly. Column 1 of Table VIII reports the regression results on this expanded sample. The results are similar to those reported in Table VI, suggesting that exclusion of these issues does not have a major impact on our results.

We use the 10-year Treasury yields as a proxy for the general level of interest rates. However, yields on corporate bonds may not increase or decrease by the same magnitude as Treasury yields. There are periods when corporate bond yields are relatively high while Treasury yields remain low, and vice versa. To address this concern, we use Moody's AAA Corporate Bond Index Yields (obtained from the Federal Reserve website) as a proxy for the general level of interest rates. Similarly, we use Changes in AAA Yields and the Standard Deviation in AAA Yields to proxy for the changes and volatility of interest rates. Column 2 of Table VIII reports the regression results with corporate bond yields as interest rate environment variables. (22) The coefficients on the three interest rate environment variables remain significant. Thus, the results are robust to different proxies for the interest rate environment.

In creating the Information Asymmetry and Risk-Shifting Potential variables, we use the median values of the underlying proxies (Firm Size and Analyst Forecast Errors for Information Asymmetry and Discretionary Asset and Free Cash Flow for Risk-Shifting Potential) to classify issuing firms as having high or low agency conflicts. Median value is a somewhat arbitrary cutoff point. To check if the results are robust to different cutoff points, we use quintiles of each underlying proxy as cutoff points and convert them to an integer variable ranging from 1 to 5, where 1 indicates the least agency conflict and 5 indicates the most severe agency conflict. Then, we add the two integer proxies for information asymmetry (risk-shifting potential) to construct the Information Asymmetry (Risk-Shifting Potential) variable. Thus, firms with the most (least) severe information asymmetry problem, that is, firms with the smallest (largest) firm size quintile and the largest (smallest) analyst forecast error quintile, will have a value of 10 (2) for Information Asymmetry, etc. Similarly, firms with the most (least) severe risk-shifting potential will have a value of 10 (2) for Risk-Shifting Potential, etc. Column 3 of Table VIII reports the results of the regression model with the modified proxies of agency conflicts. The results are qualitatively the same as those reported in Table VI.

Our sample contains 632 unique issuing firms with 295 firms having only one bond issue and 337 having multiple issues. Among the 337 issuing firms, 76 issued straight bonds only (for a total of 261 issues), 78 issued callable bonds only (for a total of 236 issues), and 183 issued both straight and callable bonds (for a total of 1,317 issues). While the p-value in the probit regression has been corrected for the potential clustering problem due to multiple bond issues by the same issuing firm, we further check the robustness of the results by excluding firms that issued both callable and straight bonds. In addition, for the 154 firms that issued only callable or straight bonds, we choose the most recent issue by each firm. We reestimate the main probit regression on this subsample of 499 bond issues and report the results in Column 4 of Table VIII. The results are similar to those reported in Table VI.

In the previous section, we find that callable issues tend to fall in the high sales growth, low MB Ratio (HSLR) group. We argue that the discrepancy between the high historical sales growth rate and low MB Ratio leads such firms to issue callable bonds to avoid the underinvestment problem. However, an alternative explanation is that callable bond issuers are riskier, so investors simply require higher returns. Thus, even if investors factor in the higher future growth rate, they are not willing to pay a premium for the firms' stocks because of higher risks. To check this alternative explanation, we use the CRSP daily stock return files to estimate the market model beta for each bond issuer 46 days prior to the issuing date. Specifically, we use the value-weighted CRSP market index and the Scholes-Williams approach to estimate the market model beta. (23) Then, we sort the sample into 10 equally sized subsamples from the lowest to the highest beta. Table IX presents the historical sales growth rates and MB Ratio for straight bonds and callable bonds in each beta decile. By doing so, we hold the riskiness of the issuing firms constant. Callable bond issuers have higher historical sales growth in all 10 beta deciles and the difference is statistically significant in four of them. On the other hand, callable bond issuers have significantly lower MB Ratio in all 10 beta deciles. Thus, the lower MB Ratio for the callable bond issuers cannot be explained solely by the higher risk of these issuers. (24)

Absent information asymmetry problems, historical sales growth and the MB Ratio should be strongly correlated. For firms with severe information asymmetry problems, however, outside investors cannot accurately estimate future growth opportunities resulting in a weaker correlation between the two proxies. Firms are more likely to issue callable bonds when outside investors underestimate future growth. Consequently, we expect the correlation between the two proxies to be weaker for callable bond issuers than for straight bond issuers. The Pearson correlation coefficient for the two proxies is 0.068 for callable bond issuers and significant at the 5% level. The correlation coefficient is 0.126 for straight bond issuers and significant at the 1% level.

IV. Conclusion and Discussion

We examine the determinants of callable bonds in light of the significant changes in the callable bond market over the last two decades. The major findings can be summarized as follows. First, the level and volatility of interest rates have a major impact on the use of callable bonds in high interest rate environments but not in low interest rate environments. This supports the interest rate hypothesis. Second, the motivation for issuing callable bonds is different between investment-grade and below-investment-grade bond issuers. Investment-grade callable bonds are primarily used to hedge interest rate risk while below-investment-grade callable bonds are mainly issued to alleviate agency conflicts. Third, the potential for risk-shifting does not affect the use of callable bonds, either alone or in combination with the information asymmetry problem. Fourth, information asymmetry, when examined alone, seems to increase the use of call provisions. However, the impact of information asymmetry on the use of the call provision is significant only when it is combined with potential underinvestment problems. Thus, it is not simply information asymmetry that leads to the use of call provisions, but information asymmetry combined with the potential underinvestment problem. Finally, the potential underinvestment problem has a significant impact on the use of callable bonds, especially when combined with information asymmetry problems.

The evidence of high agency conflicts of callable bond issuers is consistent with findings by Harris and Piwowar (2006) and Edwards, Harris, and Piwowar (2007) that callable bonds have higher trading costs. While the two papers argue that the complexity of bond instruments increases trading costs, the results may also be explained by the more severe agency conflict problems of callable bonds. As the two papers demonstrate that informational opaque bond issues tend to have higher trading costs, their findings of higher trading costs of callable bonds may indicate that issuers of these bonds have information asymmetry problems consistent with the agency cost hypotheses of callable bonds.

The findings in this paper can help to explain why previous empirical tests of the agency cost hypothesis of callable bonds have mixed results. First, our empirical results suggest that the need to hedge interest rates is the primary motivation for firms issuing callable bonds in the 1980s, a period when interest rates were high and volatile. Most firms, with or without agency conflicts, issued callable bonds to hedge interest rate risk in the 1980s, making it difficult for previous studies to examine the impact of agency conflicts on the use of callable bonds. Second, our empirical results demonstrate that call provisions can best be used to resolve the combined problem of information asymmetry and potential underinvestment. Most previous studies examine each agency conflict in isolation, not in combination. This leads to weak support for the agency cost hypothesis.

Our findings also shed some light on the dramatic changes in the callable bond market, that is, the significant declines in the number of callable bonds and the higher percentage of the below-investment-grade callable bonds in recent years. As interest rates declined significantly from the 1980s, firms no longer issued callable bonds to hedge interest rate risk. Instead, callable bonds are primarily used to alleviate agency conflicts. Since firms with higher default risk tend to have more severe agency conflicts, it is not surprising that most of the callable bonds in recent years have below-investment-grade ratings.
Appendix: Variable Definitions

All the accounting variables used are year-end values one year prior
to the bond issue date (Year -1).

Variable                  Definition

Panel A. Issuing Firm Characteristics

Firm size                 Market value of the firm's equity.

Analyst forecast errors   The absolute errors of the consensus
                          earnings forecast scaled by stock price.

Discretionary assets      1--(fixed assets/total assets).

Free cash flow            (Operating income--tax--interest--total
                          dividend)/total assets.

Historical sales growth   Annualized sales growth from six years
                          prior to the bond issue to the year prior
                          to the bond issue.

MB ratio                  (Market value of equity--book value of
                          equity + total assets)/total assets.

Debt ratio                The firm's total debt divided by total
                          assets.

ROA                       Net income/total assets.

Industry dummies          Thirty-one industry dummies based on
                          two-digit SIC codes.

Panel B. Bond Issue Characteristics

Bond rating               An ordinal number ranging from one (below
                          B) to seven (AAA by Moody's).

High rating               A dummy variable equal to one for AAA or AA
                          rated bonds and zero otherwise.

Moderate rating           A dummy variable equal to one for A or BBB
                          rated bonds and zero otherwise.

Low rating                A dummy variable equal to one for BB or
                          lower rated bonds and zero otherwise.

Issue size                Total proceeds of the bond issue.

Maturity                  Years to final maturity of the bond.

Panel C. Interest Rate Environment

T-10 yield                Yield on the 10-year Treasury note at the
                          date of the bond issuance.

Changes in T 10 yield     The change in the T 10 yield in the 100
                          business days prior to the date of bond
                          issuance.

Yield curve slope         The yield difference between the 10-year
                          Treasury and one-year Treasury at the date
                          of bond issuance.

SD in T10 yield           The standard deviation of the T10 yields
                          over the 100 business days prior to the
                          date of bond issuance.

Panel D. Proxies for Agency Conflict Costs

Information asymmetry     A dummy variable equal to zero (two) if the
                          issuer's firm size is higher (lower) than
                          the median and the Analyst Forecast Error
                          is lower (higher) than the median.
                          Otherwise, the variable equals one.

Risk-shifting potential   A dummy variable equal to zero (two) if the
                          firm's Discretionary Assets and Free Cash
                          Flow are lower (higher) than the medians.
                          Otherwise, the variable equals one.

High sales growth (HS)    A dummy variable equal to one if the
                          issuing firm has above-median historical
                          sales growth, and zero otherwise.

High sales growth, high   A dummy variable equal to one if the
MB ratio (HSHR)           issuing firm has above-median historical
                          sales growth and above-median MB Ratio, and
                          zero otherwise.

High sales growth, low    A dummy variable equal to one if the
MB ratio (HSLR)           issuing firm has above-median historical
                          sales growth but below-median MB Ratio, and
                          zero otherwise.


References

Alexander, G.J., A.K. Edwards, and M.G. Ferri, 2000, "What Does Nasdaq's High-Yield Bond Market Reveal about Bondholder-Stockholder Conflicts?" Financial Management 29, 23-39.

Banko, J.C., 2003, "Three Essays in Corporate Bond Contract Design and Valuation," University of Florida Unpublished Dissertation.

Barnea, A., R.A. Haugen, and L.W. Senbet, 1980, "A Rationale for Debt Maturity Structure and Call Provisions in the Agency Theoretic Framework," Journal of Finance 35, 1223-1234.

Barry, C.B., S.C. Mann, V.T. Mihov, and M. Rodriguez, 2008, "Corporate Debt Issuance and the Historical Level of Interest Rates," Financial Management 37, 413-430.

Black, F. and M. Scholes, 1973, "The Pricing of Options and Corporate Liabilities," Journal of Political Economy 81, 637-654.

Blume, M.E., F. Lim, and A.C. MacKinlay, 1998, "The Declining Credit Quality of US Corporate Debt: Myth or Reality?" Journal of Finance 53, 1389-1413.

Bodie, Z. and R.A. Taggart Jr., 1978, "Future Investment Opportunities and the Value of the Call Provision on a Bond," Journal of Finance 33, 1187-1200.

Bowlin, O.J, 1966, "The Refunding Decision: Another Special Case in Capital Budgeting," Journal of Finance 21, 55-68.

Boyce, W.M. and A.J. Kalotay, 1979, "Tax Differentials and Callable Bonds," Journal of Finance 34, 825-838.

Brealey, R.A., S.C. Myers, and F. Allen, 2008, Principles of Corporate Finance, 9th Ed., New York, NY, McGraw-Hill/Irwin.

Chan, L.K.C., J. Karceski, and J. Lakonishok, 2003, "The Level and Persistence of Growth Rates," Journal of Finance 58, 643-684.

Chen, Z., C.X. Mao, and Y. Wang, 2007, "Why Firms Issue Callable Bonds: Hedging Investment Uncertainty," University of Texas Working Paper.

Crabbe, L.E., 1991, "Callable Corporate Bonds: A Vanishing Breed," Financial and Economics Discussion Series 115, Federal Reserve Board.

Crabbe, L.E. and J. Helwege, 1994, "Alternative Tests of Agency Theories of Callable Corporate Bonds," Financial Management 23, 3-20.

Edwards, A.K, L.E. Harris, and M.S. Piwowar, 2007, "Corporate Bond Market Transaction Costs and Transparency," Journal of Finance 62, 1421-1451.

Eisdorfer, A., 2008, "Empirical Evidence of Risk Shifting in Financially Distressed Firms," Journal of Finance 63, 609-637.

Evans, D.S., 1987, "Tests of Alternative Theories of Firm Growth," Journal of Political Economy 95, 657-674.

Fenn, G.W., 2000, "Speed of Issuance and the Adequacy of Disclosure in the 144A High-Yield Debt Markets," Journal of Financial Economics 56, 383-405.

Freund, S., A.P. Prezas, and G.K. Vasudevan, 2003, "Operating Performance and Free Cash Flow of Asset Buyers," Financial Management 32, 87-106.

Green, R.C., 1984, "Investment Incentives, Debt, and Warrants," Journal of Financial Economics 13, 115-136.

Guntay, L., N.R. Prabhala, and H. Unal, 2002, "Callable Bonds and Hedging," University of Maryland Working Paper.

Harris, L.E. and M.S. Piwowar, 2006, "Secondary Trading Costs in the Municipal Bond Market," Journal of Finance 61, 1361-1397.

Hyytinen, A. and M. Pajarinen, 2008, "Opacity of Young Business: Evidence from Rating Disagreements," Journal of Banking and Finance 32, 1234-1241.

Jen, F.C. and J.E. Wert, 1967, "The Effect of Call Risk on Corporate Bond Yields," Journal of Finance 22, 637-651.

Jensen, M.C. and W.H. Meckling, 1976, "Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure," Journal of Financial Economies 3, 305-360.

Kish, R.J. and M. Livingston, 1992, "Determinants of the Call Option on Corporate Bonds," Journal of Banking and Finance 16, 687-703.

Krishnaswami, S. and V. Subramaniam, 1998, "Information Asymmetry, Valuation, and the Corporate Spin-Off Decision," Journal of Financial Economics 53, 73-112.

Lehn, K. and A. Poulsen, 1989, "Free Cash Flow and Stockholder Gains in Going Private Transactions," Journal of Finance 44, 771-787.

Livingston, M. and L. Zhou, 2002, "The Impact of Rule 144A Debt Offerings upon Bond Yields and Underwriter Fees," Financial Management 31, 5-27.

Merton, R., 1974, "On the Pricing of Corporate Debt: The Risk Structure of Interest Rates," Journal of Finance 29, 449-470.

Merton, R., 1987, "A Simple Model of Capital Market Equilibrium with Incomplete Information," Journal of Finance 42, 483-510.

Mitchell, K., 1991, "The Call, Sinking Fund, and Term-to-Maturity Features of Corporate Bonds: An Empirical Investigation," Journal of Financial and Quantitative Analysis 26, 201-222.

Myers, S.C., 1977, "Determinants of Corporate Borrowing," Journal of Financial Economics 5, 147-175.

Pagano, M., F. Panetta, and L. Zingales, 1998, "Why Do Companies Go Public? An Empirical Analysis," Journal of Finance 53, 27-64.

Prowse, S.D., 1990, "Institutional Investment Patterns and Corporate Financial Behavior in the United States and Japan," Journal of Financial Economics 27, 43-66.

Robbins, E.H. and J.D. Schatzberg, 1986, "Callable Bonds: A Risk-Reducing Signaling Mechanism," Journal of Finance 41, 935-949.

Rogers, W., 1993, "Regression Standard Errors in Clustered Samples," STATA Technical Bulletin 13, 19-23.

Shome, D.K. and S. Singh, 1995, "Firm Value and External Blockholdings," Financial Management 24, 3-30.

Thatcher, J.S., 1985, "The Choice of Call Provision Terms: Evidence of the Existence of Agency Costs of Debt," Journal of Finance 40, 549-561.

Thomas, S., 2002, "Firm Diversification and Information Asymmetry: Evidence from Analysts' Forecasts and Earnings Announcements," Journal of Financial Economics 64, 373-396.

Weber, M. and D. Dudney, 2003, "A Reduced Form Coefficients Analysis of Executive Ownership, Corporate Value, and Executive Compensation," Financial Review 38, 399-441.

Wooldridge, J.M., 2002, Econometric Analysis of Cross Section and Panel Data, Cambridge, MA, MIT Press.

Wooldridge, J.M., 2003, "Cluster-Sample Methods in Applied Econometrics," American Economic Review 93, 133-138.

We would like to thank an anonymous referee and Bill Christie (the editor) for helpful suggestions and comments. We are also grateful to Mark Flannery, Gerald Jensen, Miles Livingston, Robert Miller, Ted Moorman, the seminar participants at the 2006 Meetings of the Financial Management Association and Northern Illinois University for helpful comments and discussions.

(1) For example, convertible bonds have also been shown to be an effective tool to resolve risk-shifting activities (Green, 1984).

(2) For example, Chen, Mao, and Wang (2007) also find that callable bond issuers have lower market-to-book ratios, lower bond ratings, smaller firm sizes, and callable bonds are more common in high interest rate environments.

(3) In discussing the underinvestment issues, Brealey, Myers, and Allen (2008) emphasize that these problems are "most serious when firms land in financial distress."

(4) We verify the accuracy of this classification with the Fixed Income Security Database (FISD). We can match about 90% of our sample with the FISD data by an issue's six-digit CUSIP number, issuing date, and maturity date. In 97% of the matched issues, our classification is identical to the FISD classification. We do not use the FISD database in this study as it does not contain bond issues that mature before 1990.

(5) We also use the standard deviation of analyst earnings forecast to proxy for information asymmetry. The results are qualitatively the same. However, using this measure eliminates a significant number of issues that only have one stock analyst.

(6) We identify two proxies for each of the three agency conflicts, primarily because these proxies are noisy, imperfect measures of their respective underlying problem. However, multiple proxies for each agency conflict make the examination of the interaction among the agency conflicts difficult. To mitigate this problem, we construct a single categorical variable for each agency conflict to capture the implications of the various proxies. Another advantage of using the categorical variables is that the results are less sensitive to outliers.

(7) An ideal proxy for management's expectation of future growth should incorporate management's private information about the availability of positive NPV projects. However, it is very hard, by definition, to find a proxy for such private information. As an alternative to historical sales growth, we have also used five-year growth in property, plant, and equipment (PPE) as a proxy for management expectation of future growth. As growth in PPE represents cumulative net investment in fixed assets by management, high (low) growth in PPE may indicate that management is more (less) confident about future growth opportunities. The results are similar when growth in PPE is used.

(8) We use this rating specification because the correlation between ratings and callable bonds may not be linear. We have also used a numerical rating variable and a series of zero/one rating dummy variables in the regression models. The results are robust to the different specifications of the rating variable.

(9) Note that management may issue callable bonds when they perceive a potential underinvestment problem. The underinvestment problem may be real (when there are real future investment opportunities unanticipated by outside investors) or not real (when management overestimates future investment opportunities).

(10) The two subsamples correspond very closely to two different time periods: pre-1991 and post-1991. Indeed, almost 90% of the bonds in the HIRE group are issued during or before 1991, and 99% of the bonds in the LIRE group are issued after 1991. If we break the sample into pre- and post-1991 subsamples, we have almost the same results. We choose not to divide the sample according to issue year because it is less intuitive and more arbitrary.

(11) Findings in Panel B support our conjecture that HSLR issuers have more severe information asymmetry problems. When we break the HSLR group into three subsamples along the Information Asymmetry dimension, only 74 issues (15%) are classified as Low Information Asymmetry and 229 (46%) are classified as High Information Asymmetry. On the other hand, for the HSHR group, 268 (48%) issues are classified as Low Information Asymmetry and only 122 (22%) are classified as High Information Asymmetry.

(12) To control for industry effect on the use of callable bonds, we create zero/one industry dummies based on the two-digit SIC code. Several industries have fewer than 15 observations. We combine these industries into one group and use this group as the base case in all the regressions.

(13) Multiple bond issues by the same issuing firm may create a clustering problem (Wooldridge, 2002, 2003). We use the Cluster option in STATA to adjust for the potential clustering problem (Rogers, 1993) and report the cluster-robust p-values.

(14) One possible explanation for this finding is the significant decrease in the number of AAA- or AA-rated corporate bonds in the United States since the 1980s (Blume, Lim, and MacKinlay, 1998). This coincides with the decline in the number of callable bonds in the 1990s, creating a spurious positive correlation between highly rated bonds and callable bonds. Indeed, when we control for the interest rate environment variables, the coefficient on Moderate Rating is no longer significant at the 5% level. Also, the coefficient is not significant in the two interest rate subsample (HIRE and LIRE) regressions as reported in Table VII.

(15) Indeed, when we estimate Model 4 for the HIRE and LIRE subsamples separately without the rating dummy variables, the coefficients on Debt Ratio are positive in both subsamples and significant in the LIRE subsample.

(16) Since maturity and call provision are jointly determined at the time of bond issuance, inclusion of a maturity variable in the regression may cause simultaneous equation bias. To check the robustness of our results, we also run a reduced form of the simultaneous equations by leaving out the maturity variable (Weber and Dudney, 2003). The coefficients on the remaining variables are not materially changed. We report the results with the maturity variable to be consistent with previous studies, such as Kish and Livingston (1992).

(17) Over 70% of the below-investment-grade bonds have high Information Asymmetry. As a result, the interaction term, Inf. Asy * Risk-Shifting, is highly correlated with Risk-Shifting Potential ([rho] = 0.82) and the interaction term, Inf. Asy * HSHR, is highly correlated with HSHR ([rho] = 0.91).

(18) As Rule 144A was introduced in 1990, only one Rule 144A issue in our sample is classified as a HIRE issue. Thus, we use the LIRE subsample to examine the relationship between callable bonds and Rule 144A issues.

(19) The majority (224) of the Rule 144A issues are below-investment-grade bonds. A comparison of below-investment Rule 144A issues and their corresponding public issues also indicates that Rule 144A issues have higher level of Information Asymmetry and a higher percentage of HSLR issues.

(20) Among below-investment-grade bonds, 75% of Rule 144A issues are callable while 39% of public bonds are callable.

(21) As a comparison, about 60% of issues with above-median analyst forecast errors have below-median Firm Size.

(22) Note that the number of observations is reduced to 1,865 because Moody's AAA Corporate Bond Index Yields is not available prior to 1983. We have also used Moody's BBB Corporate Bond Index Yields to construct the interest rate environment variables and the regression results are similar. However, data on BBB Corporate Bond Index Yields are not available prior to 1986, further reducing the sample size, especially bond issues during the high interest rate period.

(23) Following the standard event study methodology, we use an estimation period of 255 trading days. The estimation period is 46 trading days prior to the bond issuing date (or event date). Twenty bond issues do not have sufficient historical stock returns to estimate the beta.

(24) We have also created seven subsamples, one for each bond rating category. In each of the subsample, callable bonds consistently have higher historical sales growth but lower MB Ratios, and the differences are mostly significant.

John C. Banko and Lei Zhou *

* John C. Banko is an Instructor of Finance, Insurance, and Real Estate at the University of Florida in Gainesville, Florida, USA. Lei Zhou is an Assistant Professor of Finance at the Northern Illinois University in DeKalb, Illinois, USA.
Table I. Empirical Implications of Agency Cost Hypotheses

This table summarizes the testable empirical implications of the three
agency cost hypotheses.

                                                       Information
                     Underinvestment   Risk-Shifting    Asymmetry
                       Hypothesis       Hypothesis     Hypothesis

Growth opportunity        High             None           None
Free cash flows           None             High           None
Fixed assets              None              Low           None
Asset opacity             None             None           High
Default risk              High             High           High

Table II. Sample Descriptive Statistics

This table reports the descriptive statistics of the sample. Panel A
describes the issuing firm characteristics variables. Panel B reports
the bond issue characteristic variables. Panel C presents the interest
rate environment variables. All variables are defined in the Appendix.

                           Mean     Median    Std. Dev.   Skewness

Panel A. Issuing Firm Characteristics

Firm size ($billions)       8.09      3.28     15.35        5.37
Analyst forecast errors     1.95%     0.44%     7.26%      18.25
Discretionary assets       47.15%    47.23%    24.26%       0.86
Free cash flow              6.92%     6.83%     6.21%      -8.66
Historical sales growth    11.59%     9.27%    14.51%       4.92
MB ratio                    1.57      1.30      0.88        3.53
Debt ratio                 32.62%    31.90%    15.18%       1.17
ROA                         4.80%     4.86%     5.32%      -1.00

Panel B. Bond Issue Characteristics

Bond rating                 4.53      5.00      1.34       -0.59
of high rating             22.29%
of moderate rating         59.27%
of low rating              18.44%
Issue size ($ millions)   125.56    101.90     88.11        2.01
Maturity (in years)        14.57     10.02     10.00        0.85

Panel C. Interest Rate Environment

T-10 yield                  7.56%    7.06%      2.33%       1.00
Changes in T-10 yield      -0.34%   -0.33%      0.92%      -0.32
Yield curve slope           1.35%    1.39%      1.05%      -0.44
SD of T-10 yield            0.34%    0.26%      0.23%       2.05

                          Kurtosis     1%       99%

Panel A. Issuing Firm Characteristics

Firm size ($billions)       37.67      0.11    78.83
Analyst forecast errors    493.08      0.00%   24.47%
Discretionary assets        -1.19      7.85%   93.76%
Free cash flow             199.15     -3.19%   20.05%
Historical sales growth     57.16    -10.44%   64.86%
MB ratio                    20.87      0.81     5.04
Debt ratio                   4.00      2.16%   82.11
ROA                         14.26    -12.86%   17.79%

Panel B. Bond Issue Characteristics

Bond rating                  0.13      2.00     7.00
of high rating
of moderate rating
of low rating
Issue size ($ millions)      6.40      3.48   460.41
Maturity (in years)         -0.51      1.00    40.00

Panel C. Interest Rate Environment

T-10 yield                   0.54      3.88%   13.94%
Changes in T-10 yield        1.73     -3.10%    1.72%
Yield curve slope            0.99     -1.80%    3.28%
SD of T-10 yield             4.47      0.10%    1.25%

Table III. Univariate Comparison of Straight and Callable Bonds

This table reports the univariate comparison of straight and callable
bonds. The first two columns detail the comparison for the whole
sample. Then, we break the whole sample into two subsamples: 1) high
interest rate environment (HIRE) and 2) low interest rate environment
(LIRE). If the T 10 Yield is above (below) the sample median at the
time of issuance, the bond issue is classified as a high (low)
interest rate environment bond. All variables are defined in the
Appendix. Statistical significance is based on the difference in means
between callable and straight bonds.

                              Whole Sample

                          Straight     Callable

Panel A. Issuing Firm Characteristics

Firm size ($ billions)     10.38       5.52 ***
Analyst forecast errors     1.36%      2.62% ***
Discretionary assets       50.48%     43.42% ***
Free cash flow              7.41%      6.36% ***
Historical sales growth    10.13%     13.23% ***
MB ratio                    1.76       1.36 ***
Debt ratio                 31.31%     34.09% ***
ROA                         4.80%      4.79%

Panel B. Bond Issue Characteristics

Bond rating                 4.68       4.38 ***
of high rating             21.18%     23.52%
of Moderate rating         68.40%     49.05% ***
of Low rating              10.42%     27.43% ***
Issue size ($ millions)   131.46     118.96 ***
Maturity (in years)        11.68      17.81 ***

Panel C. Interest Rate Environment

T-10 yield                  6.81%      8.41% ***
Changes in T-10 yield      -0.28%     -0.40% ***
Yield curve slope           1.43%      1.27% ***
SD of T-10 yield            0.27%      0.41% ***

Panel D. Proxies for Agency Conflict Costs

Information asymmetry       0.84       1.18 ***
Risk-shifting potential     1.11       0.88 ***
High sales growth (HS)      0.46       0.55 ***
High sales growth, high     0.32       0.20 ***
  MB ratio (HSHR)
High sales growth, low      0.14       0.34 ***
  MB ratio (HSLR)
No. of obs.                1,114        995

                            High Interest Rate
                               Environment

                          Straight     Callable

Panel A. Issuing Firm Characteristics

Firm size ($ billions)      8.74       4.93 ***
Analyst forecast errors     1.20%      2.38% ***
Discretionary assets       46.39%     40.55% ***
Free cash flow              6.67%      5.89% ***
Historical sales growth     8.94%     11.74% ***
MB ratio                    1.54       1.24 ***
Debt ratio                 29.75%     30.76%
ROA                         5.06%      5.52% *

Panel B. Bond Issue Characteristics

Bond rating                 5.00       4.82 ***
of high rating             28.26%     27.01%
of Moderate rating         68.06%     61.57% **
of Low rating               3.68%     11.42% ***
Issue size ($ millions)   128.73     119.78
Maturity (in years)        11.26      19.11 ***

Panel C. Interest Rate Environment

T-10 yield                  8.40%      9.92% ***
Changes in T-10 yield      -0.22%     -0.49% ***
Yield curve slope           1.60%      1.15% ***
SD of T-10 yield            0.32%      0.50% ***

Panel D. Proxies for Agency Conflict Costs

Information asymmetry       0.84       1.13 ***
Risk-shifting potential     0.98       0.86 ***
High sales growth (HS)      0.47       0.58 ***
High sales growth, high     0.28       0.19 ***
  MB ratio (HSHR)
High sales growth, low      0.18       0.39 ***
  MB ratio (HSLR)
No. of obs.                  407        648

                            Low Interest Rate
                               Environment

                          Straight     Callable

Panel A. Issuing Firm Characteristics

Firm size ($ billions)     11.33       6.62 ***
Analyst forecast errors     1.45%      3.07% ***
Discretionary assets       52.84%     48.77% **
Free cash flow              7.83%      7.25%
Historical sales growth    10.82%     16.00% ***
MB ratio                    1.89       1.57 ***
Debt ratio                 32.21%     40.32% ***
ROA                         4.66%      3.42 ***

Panel B. Bond Issue Characteristics

Bond rating                 4.50       3.55 ***
of high rating             17.11%     17.00%
of Moderate rating         68.60%     25.65% ***
of Low rating              14.29%     57.35% ***
Issue size ($ millions)   133.01     117.43 ***
Maturity (in years)        11.92      15.39 ***

Panel C. Interest Rate Environment

T-10 yield                  5.89%      5.59% ***
Changes in T-10 yield      -0.31%     -0.24%
Yield curve slope           1.33%      1.51% ***
SD of T-10 yield            0.25%      0.25%

Panel D. Proxies for Agency Conflict Costs

Information asymmetry       0.84       1.27 ***
Risk-shifting potential     1.19       0.91 ***
High sales growth (HS)      0.45       0.48
High sales growth, high     0.34       0.23 ***
  MB ratio (HSHR)
High sales growth, low      0.11       0.26 ***
  MB ratio (HSLR)
No. of obs.                  707        347

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.

Table IV. Classification of Callable Bonds by Information Asymmetry
and Risk-Shifting Potential

This table reports the percentages of callable bonds in several
subsamples along two dimensions: 1) in formation asymmetry and 2)
risk-shifting potential. Each bond issue is ranked low, medium, or
high for both the Information Asymmetry and the Risk-Shifting
Potential variables. In Panel A (B), we report the percentages of
callable bonds in the low, medium, and high Information Asymmetry
(Risk-Shifting Potential) subsamples. Next, we create nine subsamples
along both agency conflicts and report the results in Panel C. The
number in parentheses is the number of observations in each subsample.

                           Low     Medium    High    High--Low

Panel A. Classification by Information Asymmetry

Information asymmetry     36.19%   42.68%   63.96%    27.77% ***
                          (641)    (827)    (641)

Panel B. Classification by Risk-Shifting Potential

Risk-shifting potential   56.12%   48.75%   35.99%   -20.13% ***
                          (613)    (882)    (614)

Panel C. Cross Classifications

                                      Information Asymmetry

                               Low        Medium   High      High--Low

Risk-shifting   Low          53.97%       46.48%    70.94%   16.97% ***
  potential                  (126)        (284)    (203)
                Medium       41.54%       43.03%    61.20%   19.66% ***
                             (260)        (323)    (299)
                High         21.96%       37.27%    59.71%   37.75% ***
                             (255)        (220)    (139)
                High--Low   -32.01% ***   -9.21%   -11.23%

*** Significant at the 0.01 level.

Table V. Classification of Callable Bonds by Information Asymmetry
and Underinvestment

This table reports the percentages of callable bonds in several
subsamples along two dimensions: 1) information asymmetry and 2)
underinvestment problems. Panel A reports the percentages of callable
bonds in six subsamples along the two proxies for agency conflicts: 1)
historical sales growth and 2) information asymmetry. Each bond issue
is ranked as low or high in sales growth, and low, medium, or high in
information asymmetry. Panel B presents the percentages of callable
bonds for the high historical sales growth subgroups, further
classifying them along their market-to-book (MB) ratio. The number in
parentheses is the number of observations in each subsample.

                         Total        Information Asymmetry

                                      Low          Medium

Panel A. Historical Sales Growth and Information Asymmetry

Low sales growth (LS)    42.79%       38.46%       39.57%
                         (1,054)      (299)        (465)
High sales growth (HS)   51.56%       34.21%       46.69%
                         (1,055)      (342)        (362)
HS--LS                   8.77% ***    -4.25%       7.12% **

Panel B. High Sales Growth, MB Ratio, and Information Asymmetry

High sales growth,       36.14%       24.25%       33.13%
  high MB ratio          (559)        (268)        (169)
  (HSHR)
High sales growth, low   68.95%       70.27%       58.55%
  MB ratio (HSLR)        (496)        (74)         (193)
HSHR--HSLR               32.81% ***   46.02% ***   25.42% ***

                         Information Asymmetry

                         High         High-Low

Panel A. Historical Sales Growth and Information Asymmetry

Low sales growth (LS)    52.41%       13.95% ***
                         (290)
High sales growth (HS)   73.50%       39.29% ***
                         (351)
HS--LS                   21.09% ***

Panel B. High Sales Growth, MB Ratio, and Information Asymmetry

High sales growth,       66.39%       42.14% ***
  high MB ratio          (122)
  (HSHR)
High sales growth, low   77.29%       7.02%
  MB ratio (HSLR)        (229)
HSHR--HSLR               10.90% **

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

Table VI. Probit Regression Results

This table reports the results of probit regressions. The dependent
variable is a dummy variable equal to one for callable bonds and zero
for straight bonds. All independent variables are defined in the
Appendix. Bond issues that have below-median historical sales growth
are the base case. The p-values (in parentheses) have been adjusted
for potential clustering problems that may arise from multiple bond
issues by the same issuing firm.

                             Model 1   Model 2   Model 3   Model 4

Intercept                     -0.333    -0.309    -0.148    -4.095
                              (0.12)    (0.15)    (0.52)    (0.00)
Information asymmetry          0.311     0.256     0.106     0.009
                              (0.00)    (0.00)    (0.32)    (0.93)
Risk-shifting potential       -0.205    -0.113    -0.172    -0.055
                              (0.01)    (0.17)    (0.14)    (0.67)
High sales growth (HS)         0.290
                              (0.00)
High sales, high MB (HSHR)              -0.053    -0.372    -0.501
                                        (0.57)    (0.01)    (0.00)
High sales, low MB (HSLR)                0.623     0.656     0.254
                                        (0.00)    (0.00)    (0.05)
Inf. asy * risk-shifting                           0.055     0.062
                                                  (0.44)    (0.42)
Inf. asy * HSHR                                    0.384     0.312
                                                  (0.00)    (0.01)
Moderate rating (A or BBB)    -0.318    -0.331    -0.352    -0.212
                              (0.01)    (0.01)    (0.00)    (0.07)
Lower rating (BB or below)     0.598     0.596     0.559     1.583
                              (0.00)    (0.00)    (0.00)    (0.00)
Debt ratio                    -1.292    -1.252    -1.281    -0.397
                              (0.00)    (0.00)    (0.00)    (0.33)
T-10 yield                                                   0.233
                                                            (0.00)
Change in T-10 yield                                        -0.124
                                                            (0.01)
Yield curve slope                                            0.040
                                                            (0.30)
SD in T-10 yield                                             1.012
                                                            (0.00)
ROA                                                          2.087
                                                            (0.01)
Maturity                                                     0.057
                                                            (0.00)
Log of issue size                                            0.110
                                                            (0.04)
Industry dummies                Yes       Yes       Yes       Yes
No. of obs.                    2,109     2,109     2,109     2,109
Pseudo [R.sup.2]                0.14      0.16      0.17      0.36

Table VII. Probit Regressions on Interest Rates and Credit Rating
Subsamples

This table reports probit regressions (Model 4) on five subsamples.
HIRE (LIRE) versions of Model 4 represent the regression results on
the High (Low) Interest Rate Environment subsamples, where the bonds
are issued when the T-10 Yield is above (below) the whole sample
median T-10 Yield. The High Rating (Moderate Rating) subsample
contains AAA- and AA- (A- and BBB-) rated bonds. The Low Rating
subsample contains below-investment-grade bonds. The p-values (in
parentheses) have been adjusted for potential clustering problems that
may arise from multiple bond issues by the same issuing firm.

                              HIRE     LIRE    High Rating

Intercept                    -4.404   -1.282     -6.934
                             (0.00)   (0.05)     (0.01)
Information asymmetry         0.054   -0.064     -0.098
                             (0.68)   (0.67)     (0.59)
Risk-shifting potential       0.205   -0.323      0.143
                             (0.28)   (0.05)     (0.56)
High sales, High MB (HSHR)   -0.387   -0.581      0.144
                             (0.19)   (0.02)     (0.69)
High sales, Low MB (HSLR)     0.013    0.199      0.048
                             (0.95)   (0.20)     (0.88)
Inf. asy * risk-shifting      0.039    0.115      0.157
                             (0.74)   (0.28)     (0.41)
Inf. asy * HSHR               0.166    0.403     -0.005
                             (0.48)   (0.02)     (0.99)
Moderate rating               0.039   -0.376
(A or BBB)                   (0.81)   (0.06)
Lower rating                  1.304    1.673
(BB or below)                (0.00)   (0.00)
Bond rating                                       0.034
                                                 (0.90)
Debt ratio                   -1.047   -0.346      0.282
                             (0.11)   (0.42)     (0.87)
T-10 yield                    0.267   -0.206      0.299
                             (0.00)   (0.01)     (0.00)
Change in T-10 yield         -0.197    0.035     -0.188
                             (0.00)   (0.69)     (0.05)
Yield curve slope            -0.137    0.108     -0.056
                             (0.00)   (0.04)     (0.40)
SD in T-10 yield              0.693    1.033      0.872
                             (0.01)   (0.11)     (0.04)
ROA                           1.673    2.219     -2.668
                             (0.33)   (0.02)     (0.37)
Maturity                      0.062    0.062      0.073
                             (0.00)   ('0.00)    (0.00)
Log of issue size             0.136    0.074      0.302
                             (0.18)   (0.20)     (0.07)
Industry dummies               Yes      Yes        Yes
No. of obs.                   1,055    1,054       470
Pseudo [R.sup.2]               0.37     0.38      0.49

                             Mod. Rating   Low Rating

Intercept                      -5.550         0.472
                               (0.00)        (0.62)
Information asymmetry          -0.092         0.486
                               (0.48)        (0.00)
Risk-shifting potential        -0.039        -0.200
                               (0.80)        (0.17)
High sales, High MB (HSHR)     -0.739         0.386
                               (0.00)        (0.06)
High sales, Low MB (HSLR)      -0.240         0.590
                               (0.21)        (0.00)
Inf. asy * risk-shifting        0.139
                               (0.20)
Inf. asy * HSHR                 0.246
                               (0.24)
Moderate rating
(A or BBB)
Lower rating
(BB or below)
Bond rating                     0.141        -0.565
                               (0.24)        (0.00)
Debt ratio                     -0.747        -0.375
                               (0.29)        (0.40)
T-10 yield                      0.359         0.015
                               (0.00)        (0.73)
Change in T-10 yield           -0.113        -0.173
                               (0.15)        (0.08)
Yield curve slope              -0.005         0.083
                               (0.92)        (0.31)
SD in T-10 yield                1.245         0.117
                               (0.00)        (0.87)
ROA                             4.851         0.836
                               (0.00)        (0.40)
Maturity                        0.064         0.025
                               (0.00)        (0.30)
Log of issue size               0.027         0.107
                               (0.80)        (0.30)
Industry dummies                 Yes          Yes
No. of obs.                     1,250         389
Pseudo [R.sup.2]                 0.45        0.18

Table VIII. Robustness Checks

This table reports robustness checks on the main probit regression
(Model 4 in Table VI). In Column 1, we expand the sample by including
issues that do not have analyst forecasts. We classify these issues as
a high Analyst Forecast Errors group and construct the agency
conflicts proxies accordingly. In Column 2, we use the Moody's AAA
Corporate Bond Index Yields to construct the interest rate environment
variables (except Yield Curve Slope). In Column 3, we modify the
construction of the Information Asymmetry and Risk-Shifting Potential
variables. Column 4 reports the regression results on a subsample of
499 bond issues by firms that issue either callable bonds only or
straight bonds only. The p-values are shown in parentheses.

                             Column 1   Column 2   Column 3   Column 4

Intercept                     -4.037     -4.967     -4.332     -7.271
                              (0.00)     (0.00)     (0.00)     (0.00)
Information asymmetry         -0.018     -0.003     -0.015      0.314
                              (0.85)     (0.98)     (0.81)     (0.24)
Risk-shifting potential       -0.035     -0.113      0.008      0.248
                              (0.78)     (0.40)     (0.91)     (0.49)
High sales, High MB (HSHR)    -0.543     -0.536     -0.900     -0.549
                              (0.00)     (0.00)     (0.00)     (0.25)
High sales, low MB (HSLR)      0.272      0.188      0.247      0.505
                              (0.03)     (0.16)     (0.05)     (0.02)
Inf. asy * risk-shifting       0.077      0.067      0.007     -0.080
                              (0.30)     (0.42)     (0.49)     (0.68)
Inf. asy * HSHR                0.368      0.305      0.118      0.605
                              (0.00)     (0.02)     (0.01)     (0.04)
Moderate ratings (A or BBB)   -0.220     -0.251     -0.234      0.125
                              (0.06)     (0.04)     (0.04)     (0.74)
Lower ratings (BB or below)    1.635      1.693      1.525      2.117
                              (0.00)     (0.00)     (0.00)     (0.00)
Debt ratio                    -0.223     -0.171     -0.307      0.308
                              (0.54)     (0.69)     (0.46)     (0.59)
T-10 yield                     0.241                 0.240      0.240
                              (0.00)                (0.00)     (0.00)
Change in T-10 yield          -0.130                -0.125     -0.027
                              (0.01)                (0.01)     (0.83)
Std. in T-10 yield             0.968                 1.007      2.921
                              (0.00)                (0.00)     (0.00)
AAA yield                                 0.359
                                         (0.00)
Change in AAA yield                      -0.225
                                         (0.00)
Std. in AAA yield                         1.170
                                         (0.00)
Yield curve slope              0.035      0.008      0.043      0.273
                              (0.34)     (0.83)     (0.26)     (0.00)
ROA                            1.999      2.090      1.826     -0.319
                              (0.01)     (0.02)     (0.03)     (0.84)
Maturity                       0.058      0.058      0.057      0.059
                              (0.00)     (0.00)     (0.00)     (0.00)
Log of issue size              0.093      0.073      0.113      0.330
                              (0.05)     (0.19)     (0.03)     (0.02)
Industry dummies                Yes        Yes        Yes        Yes
No. of obs.                    2,231      1,865      2,109        449
Pseudo [R.sup.2]                0.37       0.35       0.36       0.47

Table IX. Proxies for Growth across Beta Deciles

This table reports the means of the two proxies for growth across beta
deciles. The two proxies for growth are Historical Sales Growth and MB
Ratio. Column A presents the means of the Historical Sales Growth for
straight bond issuers and callable bond issuers. Column B provides the
means of the MB Ratio for straight bond issuers and callable bond
issuers. Column C reports the means of the market model beta for
straight bond issuers and callable bond issuers. Column D indicates
the number of observations. Statistical significance is based on the
difference in the means between callable and straight bonds.

Beta         A. Historical            B. MB Ratio
Deciles       Sales Growth

          Straight    Callable    Straight   Callable
            Bond        Bond        Bond       Bond
          Issuers     Issuers     Issuers    Issuers

1           7.76%     9.80%         1.77     1.40 ***
2           7.09%    11.29% ***     1.77     1.21 ***
3           7.88%    14.50% ***     1.77     1.31 ***
4           9.89%    12.00%         1.93     1.34 ***
5           9.33%     9.47%         1.70     1.31 ***
6          10.50%    13.31% *       1.77     1.27 ***
7          12.28%    13.83%         1.71     1.33 ***
8           9.92%    13.71%         1.75     1.35 ***
9          12.81%    13.46%         1.71     1.41 ***
10         14.40%    20.60% ***     1.78     1.56 ***

Beta            C. Beta         D. Observations
Deciles

          Straight   Callable   No. of Straight
            Bond       Bond       (Callable)
          Issuers    Issuers         Bonds

1           0.09     0.09          118 (90)
2           0.40     0.40          97 (112)
3           0.55     0.55          125 (84)
4           0.70     0.70          95 (114)
5           0.84     0.85          114 (95)
6           0.97     0.98 **       117 (92)
7           1.13     1.12          102 (107)
8           1.29     1.30          109 (100)
9           1.49     1.49          108 (101)
10          1.96     1.98          113 (96)

*** Significant at the 0.01 level.

** Significant at the 0.05 level.

* Significant at the 0.10 level.
COPYRIGHT 2010 Financial Management Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2010 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Banko, John C.; Zhou, Lei
Publication:Financial Management
Article Type:Report
Geographic Code:1USA
Date:Jun 22, 2010
Words:16364
Previous Article:The quality and price of investment banks' service: evidence from the PIPE market.
Next Article:Speed of issuance, lender specialization, and the rise of the 144A debt market.
Topics:

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters