Corporate debt issuance and the historical level of interest rates.
In the neoclassical theory of investment, a drop in the cost of capital results in increased investment spending. As firms increase their investments, all else equal, they also tend to issue new debt to finance a portion of those investments. Thus, debt issuance can be a function of the level of capital investment by firms. Alternatively, firms may try to time debt issuance if they believe that interest rates are advantageous relative to where rates were previously. We define "timing of interest rates" as the practice of issuing debt, relative to financing needs and capital expenditures, when interest rates are low compared to historical levels.
We provide evidence that links debt issuance to changes in interest rates relative to their historical levels. To distinguish between a timing theory and a more standard neoclassical story, we ask whether firms issue more debt relative to their investment plans (capital expenditures) or relative to equity issuance, following changes in interest rates. We find that past interest rates have prominent effects on the issuance of debt: when current interest rates are low relative to historical levels, firms tend to issue more debt. These effects occur even after we account for other factors that can lead to debt issuance, such as capital expenditures, refinancing, firm size and profitability, the level of corporate cash flow, interest rate spreads, and equity valuation.
Consistent with Graham and Harvey's (2001) survey evidence that "managers attempt to time interest rates by issuing debt when they feel that market interest rates are particularly low," we find that the amount of debt issued is substantially higher when interest rates are low relative to historical levels. We also find that issuance depends not just on the absolute level of interest rates, but also on their recent history. Thus, not only the levels, but also the changes in the level of interest rates, seem to affect managers' issuance decisions.
When interest rates are low, there can also be greater incentives to issue equity or to increase the level of capital expenditures. Therefore, we examine the ratio of debt issuance to the sum of debt and equity issuance, and the ratio of debt issuance to capital expenditures. Both ratios increase when interest rates are low relative to recent history.
Since refinancing is natural when interest rates fall and could account for the increased issuance when rates are low, we also examine the results after we account for refinancings. We find that our conclusions continue to hold even when we exclude refinancings.
When we disaggregate the data, we find some evidence that firms with fewer financial constraints exhibit more evidence of timing debt issues than do more constrained firms. Larger firms appear to be more able or willing than smaller firms to issue when rates are relatively low. The same is true for firms with high, compared to low, free cash flow and for more profitable firms compared to less profitable firms. Conversely, firms with higher capital expenditures are less sensitive to the effect of historical rates than are firms with lower capital expenditures.
Managers appear to consider the history of recent interest rates when deciding to issue debt. Regardless of whether or not such considerations actually lower the cost of debt capital, which is a subject of an ongoing debate, they appear to affect corporate financial policy. This observation suggests that they should be considered more prominently in capital structure investigations.
The paper proceeds as follows. Section I summarizes our data. Section II presents our empirical results, and Section III concludes the paper.
I. Data and Sample Characteristics
Our study period is January 1970 to April 2001. For this period we obtain data for new, nonconvertible, public debt issues from Thomson Financial's Securities Depository Center (SDC) new issues database. There are 85,724 debt issues during this period. We obtain data on the issue date, the identity, and characteristics of the borrower (such as industry and nationality); and various characteristics of the bond issue, such as proceeds in nominal dollars, maturity, yield to maturity (YTM) at issuance, credit rating, whether the issue is floating or fixed rate, and whether it is callable or puttable. We adjust the nominal proceeds with the monthly consumer price index (CPI) to obtain proceeds in constant dollars (measured as of January 1, 2001) so that we can make valid comparisons of the quantities issued across time. We exclude issues by non-US firms, issues outside of the United States, issues by firms in Standard Industrial Classification (SIC) codes 6000-6999 (financial companies) or 9000-9999 (government-related), issues by nonprofit organizations (such as churches or universities), and issues that are components of derivative instruments. Those screens reduce the number of issues with available issuance data to 14,623.
We also obtain study-period data for equity issues in the US public markets from the SDC. We exclude initial public offers and then apply the same exclusion criteria as described above for debt issues. After applying these screens, we have 10,209 equity issues.
In addition to data on new issues, we obtain interest rate, inflation, and interest rate spread data. We obtain time series of the monthly yields on Treasury and corporate bonds with various maturities and on inflation rates, measured from the CPI, from the Federal Reserve Bank of Saint Louis's FRED database. These include the 90-day Treasury-bill yield (denoted as T-bill 90), the 10-year constant maturity Treasury yield, and the Baa corporate yield. We measure the term spread as the difference between the 10-year constant maturity Treasury yield and the T-bill 90 yield, and we measure the default spread as the difference between the Baa yield and the 10-year constant maturity Treasury yield.
We also obtain Compustat data for the issuers. We obtain firm Compustat data for 9,614 of the 14,623 issues, or approximately two-thirds of our issuance sample. For some of our tests, we obtain firm-level data on capital expenditures (Compustat data item 128). For each debt-issuing firm, we aggregate the annual amount of debt issued (from the SDC) and scale it by annual capital expenditures. The number of observations in our firm-level tests varies due to the availability of different data items. We also collect data that characterize (in the aggregate) the stock market, such as the price-earnings ratio and market-to-book value ratio of the S&P 500 index (from CRSP and Compustat).
In Figure 1, Panel A shows the patterns of the SDC's monthly debt issuance. Panel B shows the Compustat annual net debt issuance compared to mean monthly Baa yields. Compustat debt data include bank debt, which firms could use as a substitute for public debt. We calculate Compustat net debt issues, which we obtain from the cash flow statement, as data item 111 (issuance of long-term debt) minus data item 114 (reduction in long-term debt). We measure Compustat gross debt issuance as data item 111. We use the CPI to convert the dollar amounts into January 2001 dollars. For ease of comparison, we divide annual Compustat issuance by 12 to show the average monthly issuance. We exclude financial firms and foreign issuers on Compustat.
Both panels in Figure 1 show that debt issuance tends to increase when interest rates drop. When we compare the panels, Panel A shows considerable within-year variations in yields and debt issuance, thus confirming the benefit of using monthly data to examine timing attempts. The correlation between the annualized SDC debt issuance and annual Compustat net (gross) debt issued is 35% (78%), and the correlation between the annualized SDC debt issuance and the annualized Baa yield series is -54%. In comparison, the correlation between the Baa yield and the Compustat net (gross) issuance is -18% (-47%).
Table I provides stylized facts about debt issuance and shows yields on issued debt, Treasury bills, and 10-year Treasury securities over the sample period. Measured in January 2001 dollars, the total amount of debt issued in our sample is $2.60 trillion. The largest dollar amount of issues in a single year is $182 billion in 1993, which coincides with the second largest number of issues (1,049). The year 1993 shows the lowest median level of short-term rates (measured by T-bill 90s) and the fourth lowest median level of long-term rates (measured by 10-year Treasuries) across the 32 years of our sample. When we exclude 2001, for which we use only five months of issuance data, the smallest annual amount raised ($28 billion) and fifth smallest number of issues (221) occur in 1983, when median short-term rates and long-term rates are at their sixth and fifth highest levels, respectively, across the sample.
Table I shows the ratio of debt issued to the sum of debt and seasoned equity offerings (SEOs). The lowest mean and median values of the ratio occur in 1983, the year that also has the minimum amount of total debt issuance. The highest ratios occur in 1988 and 1989.
Table I also provides an overview of the fraction of issues that were callable or puttable across time. As noted above, we do not report data on callable or puttable debt prior to 1976, since the SDC did not consistently report those features prior to 1976. Overall, about 42% of the issues were callable. Callability was very common prior to the 1990s, and was used in an average of more than 75% of all debt issues. Since 1990, the callability has averaged less than 25% of the debt issues. The year 1981 has the maximum callability level in our sample (88.5%), the same year for which our data show the maximum yields on debt issues. This finding shows that when the cost of debt is highest, the greatest percentage of callable debt is issued. Callability provides companies with the ability to reduce the effective maturity of their debt issues at the highest interest rate levels.
[FIGURE 1 OMITTED]
We also examine the relation between call features and interest rate levels. The correlations between callability and the yields on T-bills, 10-year Treasury bonds, and Baa debt are 61%, 69%, and 68%, respectively. We also measure the correlation between the fraction of proceeds that are callable and interest rate volatility, defined as the standard deviation of the yields on Baa-rated debt over 12, 36, and 60 months prior to issuance. The correlations are 40%, 57%, and 60%, respectively. In unreported multivariate analyses, we regress the fraction of proceeds that are callable on interest rates, interest rate volatility, time dummies, and control variables. We then confirm the positive relation between callables and interest rates and rate volatility and the secular trend toward a decline in callables.
About 8% of the overall debt issues include put provisions, and a similar percentage of issues have floating rates (not reported). The period prior to 1989 shows very small percentages of puttable debt issues. During the peak interest rate year (1981), just over 1% of debt issues include put options. The year 1994 was a year in which the interest rates of debt issues were at about one-half the level for 1981. Thus, the 1994 issues have the maximum percentage of put options (21.5%). Despite the relatively low interest rates during the latter few years of our sample, the years following 1994 show sharply lessened levels of put options.
II. Interest Rate Levels and Debt Issuance
Many factors affect the decisions by companies to issue new debt and the features they select for the debt. (1) Graham and Harvey (2001) find that managers are interested in the current levels of interest rates relative to historical rates. They describe the finding as follows: "We inquire whether executives attempt to time interest rates by issuing debt when they feel that market interest rates are particularly low." They provide "... moderately strong evidence that firms try to time the market in this sense." (2) We examine the extent to which debt issues are associated with the level of interest rates relative to historical levels, and we account for financing needs and capital expenditures. Graham and Harvey also find that "... market timing is especially important for large firms," an item that we examine in addition to other characteristics of issuing firms. Bancel and Mittoo (2004) find similar results from their survey of European financial managers. Other studies that suggest timing considerations might affect debt decisions include Taggart (1977), Marsh (1982), and Faulkender (2005).
A. Interest Rate Levels and Historical Rates
We construct two measures of historical interest rates. The first measure identifies the decile ranking of current interest rates against rates over the previous 10 years. The first decile refers to rates that were below the 10th percentile of average monthly rates in the prior 10 years. The second decile pertains to those rates from the 10th percentile to below the 20th percentile, and so on. Like Graham and Harvey (2001), we also refer to the rates as "low" if they are in one of the bottom three deciles, as "high" if they are in one of the top three deciles, and "medium" for rates in the middle deciles. A given interest rate could be high or low relative to historical rates. For example, each of the 10-year constant maturity Treasury rates from 5% to 9% are relatively low rates in some periods and relatively high rates in other periods. A manager who faces a 10% cost of new debt might be more willing to issue debt at that rate if it is among the lowest rates in recent memory.
Our second method of accounting for historical rates is to use the lagged values of interest rates, in addition to their current value.
Figure 2 illustrates the point using quintiles of current interest rates relative to historical levels. (3) Panel A shows the average monthly SDC debt issue proceeds across quintiles of monthly average Baa yields. We define 1 as the lowest quintile and 5 as the highest. Clearly, more debt is issued when yields are lower.
Panel A also shows the average monthly proceeds across yield differential quintiles. We identify yield differential quintiles based on the difference between the current month's Baa yield and the average yield over the prior 10 years. Thus, quintile 1 represents the lowest differential. More debt is issued when yields are low as well as when they are low compared to the average yield over the prior 10 years.
Panel B of Figure 2 illustrates both the level effect (the raw yield level) and the change effect (the deviation from the past 10-year average) by using a two-way sort to compare the average monthly proceeds against the quintiles of monthly yields and yield differentials. For example, Panel B shows that the highest amount of debt issues occurs when yields are low and the 10-year yield differentials are also low. Panel B shows that both the level of the yield and the 10-year yield differential appear to affect issuance. Proceeds are generally highest in months in which rates are lower than the prior 10-year average. Thus, Panel B demonstrates a tendency toward debt issuance when interest rates are low relative to prior levels.
We note that several of the two-way sort categories have no observations. For example, there are no months in which the yield is at the highest-level quintile and also at the lowest quintile of the 10-year yield differential. Similarly, there are no months with lowest-quintile yields but highest-quintile 10-year yield differentials.
B. Debt Issuance, Interest Rates, Spreads, and Growth
Table II provides ordinary least squares (OLS) tests that incorporate a number of variables that might affect debt issuance. These variables include interest rates (either 10-year constant maturity Treasury rates or Baa rates), rate deciles (relative to historical rates), credit and term spreads, and price-earnings and market-to-book ratios, factors that indicate market growth. In half of the regressions, instead of interest rate deciles, we use five-year and 10-year lagged rates.
We are concerned that debt issuance might occur not because of low rates relative to historical levels, but because of attractive credit spreads and term spreads. Thus, the regressions in Table II incorporate such spreads. The term spreads, which we measure as the difference between 90-day T-bills and 10-year Treasuries, are not significantly associated with debt issuance. Credit spreads, which we measure as the difference between the 10-year constant maturity Treasury rates and Baa rates, are significantly and positively associated with debt issuance. This finding indicates higher credit spreads during periods of high issuance. The positive association is not consistent with attempts to time the credit spread, but it is consistent with timing the level of rates. Unless the demand to purchase debt issues increases along with the increased supply of debt issues at low rates, then during periods of high supply relative to demand, the demand and supply interaction should lead to lower relative bond prices and thus to higher credit spreads.
[FIGURE 2 OMITTED]
A second concern on the apparent relation between issuance and relative interest rates is that low relative interest rates might also reflect high contemporaneous levels of corporate growth expectations. Such a result would be consistent with the neoclassical theory of investment, since increased investment could be a natural part of the growth expectations. To account for the possibility of such effects, in Table II we include two measures that characterize market-wide growth expectations, the price-earnings (P/E) ratio, and the market price-to-book value ratio (M/B) for the S&P 500 index.
Even after we incorporate the spread and growth variables, the historical interest rate deciles remain highly significant. The highest levels of debt issuance occur when growth proxies are high and interest rates are historically low. The lowest levels of debt issuance occur when rates are high and expected growth is low. Therefore, the effects of relative interest rates on debt issuance remain even after we account for the effect of expected overall economic growth.
Variation in rates affects issuance in an economically meaningful way. For example, Panel A of Table I indicates that if the decile rank of the relative rate increases by 1, on average, issuance decreases by $361 million in a month. For comparison, if the rate itself increases by 1%, the amount issued in a month decreases by $257 million.
In Table II, the panels extend the regressions to utilize historical rate lags instead of the deciles of current rates relative to historical rates. The five- and 10-year lagged rates are positive and significant in each regression. The current level of 10-year constant maturity Treasury rates and Baa rates at the time of issuance remain negative and significant. Therefore, whether we measure interest rates as deciles relative to historical rates or whether we use lagged rates, we see that they significantly affect debt issuance.
In an untabulated analysis (available on request), we also investigate the horizon of relevant rate changes. We do not have a strict theoretical justification for choosing the 10-year horizon over which we calculate the decile ranking of interest rates relative to past rates (our first measure). The relevant horizon should reflect the "memory" of the decision maker for the history of interest rates. For example, one proxy for the length of the decision maker's "memory" could be the tenure of the CEO or the CFO. Therefore, we examine a model with the current Baa rate and its lagged value plus control variables, where the lags range from one year to 20 years. The lagged rate is positive and significant as early as three years back and as far back as 19 years, with the adjusted [R.sup.2] and the t-value of the lagged variable maximized at 11 years.
C. Debt Issuance as a Proportion of Total Issuance
An important question is whether the effect is a case of issuing debt to time the interest rates, or if it is a case of raising funds in general. For example, Burch, Christie, and Nanda (2004) specifically examine whether firms time equity offerings. In Panel B of Table II, we examine the issuance of debt and equity and calculate the ratio of debt issued to the sum of debt and equity issued. The equity issues are in the form of SEOs that, like public debt issues, are reported in the SDC. As noted earlier, during the period of our study we find more than 14,000 debt issues and 10,000 equity issues. The question is whether the ratio of the amount of debt issued to the total amount of debt and equity issued increases as interest rates become lower relative to historical levels. If so, interest rates relative to historical rates do specifically affect debt issuance, not just total firm issuance of new funding.
The results in Panel B of Table II demonstrate significantly higher debt-to-total issuance as a function of lower deciles of current rates relative to historical rates or higher lagged interest rates. The rate deciles remain highly significant, as do the five-year lagged rates, although the 10-year lagged rates do not. The results hold whether we use 10-year Treasuries or Baa rates. The levels of interest rates also continue to hold strongly.
Thus, even if more attractive economic conditions do attract equity issuance, the amount of debt issuance relative to the total of debt and equity issuance is still greater when interest rates are lower relative to historical rates. These results remain consistent with the survey results in Graham and Harvey (2001).
D. Debt Timing and Capital Expenditures
When interest rates decline to relatively low levels, the net present value of some capital investment projects may become positive because of the lower discount rate. Thus, what we observe may not be debt timing per se, but the result of firms issuing debt because of increased opportunities for capital spending.
We measure debt issuance relative to capital expenditures and examine the impact of changes in rates on this ratio. We obtain data on capital expenditures from Compustat (item 128). For each firm-year within our SDC sample for which the capital expenditure data are available from Compustat, we calculate the annual amount of SDC debt issued and divide it by the annual amount of capital expenditures. We exclude the extreme top 1% of the observations, limiting the ratio to a range of 0 to 38.4. The resulting sample size is 4,313 firm-years. We then regress the ratio of the quantity of debt issued divided by capital expenditures of each of the issuing firms on the interest rate level, interest rate deciles, or lagged interest rates, and control measures.
Panel C in Table II presents the results of regressions with a dependent variable equal to the quantity of debt issued divided by capital expenditures of the issuing firms. The historical decile ranks of current interest rates are strongly associated with the ratio of debt issued to capital expenditures. The lagged historical rates with a five-year lag remain significant, and, as in Panel B, the 10-year lags are not significant. Unlike the previous tests, the current level of interest rates (as opposed to its historical decile) does not remain significant when we include the historical rate deciles, but they do remain significant when we use lagged historical rates instead.
The results of Panel C in Table II further illustrate that debt timing, based on the current rates relative to historical rates, remains significant and is not caused merely by the need for capital expenditures. This evidence of timing is a noteworthy result, since the neoclassical theory of investment suggests increased capital expenditures when the cost of capital is low. Although increased expenditures do occur, they do not fully account for the quantity of debt issued. To account for the potential impact of interest rate persistence, we re-estimate the regressions in Table II by using a first-order autoregressive error model. The autoregressive results produce lower regression [R.sup.2] and lower t-values for the independent variables. However, the overall inferences on the relation between issuance and historical rates are consistent with those we find when we use OLS, in that the coefficients and the significance of the relative rate variables remain qualitatively similar. For example, in the first model specification in Panel A of Table II, the coefficient on the Baa decile variable changes from -360.91, with a t-value of-5.30 and a p-value less than 0.0001, to -284.62 with a t-value of -2.89 and a p-value of 0.004. The [R.sup.2] of the autoregressive model is 0.56 (including the autoregressive term), and 0.29 for the original regressors, compared to 0.48 in the original model. When we use second- or third-order autoregressive models, the variable remains significant and has a coefficient of similar magnitude.
E. Debt Timing, Refinancing, and Net Stock Issuance
The timing of debt issuance relative to historical rates can be influenced by the desire to refinance expensive existing debt. Therefore, we examine the robustness of the interest rate timing results with respect to the use of the proceeds. We also study the effects of net stock repurchases compared to net stock issuance, since repurchases could represent another type of refinancing. If firms issue new debt to retire expensive existing debt, then that decision might be driven entirely by the potential for reducing the cost of existing debt rather than for the issuance of further debt. Therefore, we divide our sample into subsets of firms that are net issuers of debt compared to firms that apparently use the proceeds to refinance existing debt. Since share repurchases can also be a form of refinancing, we break the samples into those with debt refinancing that do or do not make net share repurchases.
We calculate the net issuance of debt for each firm in each fiscal year as the difference between the amount of debt issued (Compustat item 111) and the amount of debt retired (Compustat item 114). We then divide firms into a net issuers subsample if the difference is positive, and a refinancing subsample if the difference is 0 or negative. We present our results in Panel A of Table III.
Panel A of Table III documents the significant effects of relative interest rates in both the refinancing subsample and the net issuers subsample. Refinancing is strongly associated with the relative levels of interest rates compared to historical levels (measured, again, by low, medium, or high deciles). Nevertheless, even when there is no refinancing, there continues to be a significant association between debt issuance and the historical decile levels of current interest rates.
We further split the net issuers into firms that use some of the proceeds for refinancing and firms whose net debt increases by more than the amount of the public debt issued, indicating that in addition to issuing public debt, those firms borrow from other sources. Once again, although it appears that there is more timing by the firms that use at least part of the proceeds for refinancing, there continues to be significant timing by the firms that use all of their proceeds as new financing. Therefore, consistent with the evidence discussed earlier, the results continue to show that whether or not refinancing is a part of their debt issuance decision, managers tend to issue more debt when interest rates are at low levels relative to historical rates.
We also break each of the three debt samples into subsamples with positive or negative net stock issuance. We calculate the net issuance of common and preferred stock for each firm in each fiscal year as the difference between the amount of stock issued (Compustat item 108) and the amount of stock repurchased (Compustat item 115). The results from this further subsampling are presented in Panel B of Table III. Our results for historically based debt timing remain robust across the stock issuance categories. In all cases, the evidence supports the hypothesis that the amount of debt issuance depends on where interest rates are relative to historical levels.
We also perform multivariate regressions, using the same structure and the same explanatory variables as those reported in Table II. In these regressions we redefine the dependent variable to be the aggregate monthly ratio of net debt to the sum of net debt and equity issued. This specification requires the assumption that annual net equity issuance and net debt issuance are evenly distributed across the months of the fiscal year. We define net long-term debt issuance as the minimum of either the total annual SDC debt issues or the increase in (annually reported) Compustat debt, thus eliminating all potential refinancing from the sample. We allocate each firm's net annual debt increase to months by using each firm's monthly proportion of its annual issuance. We then allocate net equity issuance, which is available only annually, to months by assuming that the issuance of net equity during the year is on the same monthly schedule as debt issuance.
The untabulated results (available on request) strongly support the results in Table II. When current rates are low relative to historical rates, companies issue a significantly higher proportion of net debt compared to net equity. For example, for the specification that uses the level of the Baa yield, the Baa decile, the credit spread, the term spread, and the S&P 500 P/E ratio as dependent variables, the adjusted [R.sup.2] is 0.074, and the coefficient on the Baa decile is -0.006 with a t-value of 2.97, and a p-value of 0.003. The results are similar for the rest of the specifications.
F. Corporate Characteristics, Debt Issuance, and Interest Rates
Graham and Harvey (2001) report that managers of larger firms tend to attempt interest rate timing more than do the managers of smaller firms. There can be other characteristics of firms that also influence the tendency to time debt issuance based on relative interest rates. For example, firms with greater financial flexibility may be better able to time their issuance than can firms that are financially constrained, either due to lower cash flows or large investment intensity. In Table IV we consider the effects of corporate profitability, the amount of free cash flow after funding investments (scaled by total assets), and the intensity of capital expenditures upon the tendency to issue debt. Table IV presents regression results based on the comparisons of large firms compared to small firms, highly profitable firms compared to less profitable firms, firms with high free cash flow compared to firms with low free cash flow, and firms with high capital expenditures compared to firms with low capital expenditures. In each of the paired subsamples, we break the samples into the upper and lower half of firms relative to the median of the variable of interest.
The results in Table IV show that for all the cases we examine, the level of interest rates measured as deciles based on historical rates is consistently significant in determining debt issuance. Small firms and large firms both issue more debt when the rate decile is lower. The rate coefficient is more negative for larger firms, indicating that larger firms are more able or willing to issue when rates are relatively low. Highly profitable firms and less profitable firms all issue more debt when rates are historically lower, although, like the results for larger firms, the more profitable firms show stronger negative coefficients than do the less profitable firms. The same pattern holds for firms with high compared to low free cash flow. Conversely, firms with higher capital expenditures are less responsive to where interest rates are relative to historical rates than are firms with lower capital expenditures.
We test for the differences in coefficients across firm types. We pool the subsamples to combine firm types and add interactive variables consisting of each dependent variable multiplied by a dummy variable for the firm type. We then examine the significance of the coefficients on the interactive terms.
The coefficient for the large firm dummy (set equal to 1 for large firms) times the Baa decile variable is -109.3 with a t-value of-2.64 and p-value of 0.009. Thus, issuance of debt by large firms is more sensitive to the relative level of interest than is small firm issuance. The coefficient for the more profitable firm dummy (set equal to 1 for more profitable firms) times the Baa decile variable is -94.5 with a t-value of -2.47 and p-value of 0.014. This result shows that more profitable firms exhibit more timing than do less profitable firms. The coefficient for the higher capital expenditure firm dummy (set equal to 1 for high capital expenditure firms) times the Baa decile variable is 124.3 with a t-value of 3.01 and a p-value of 0.003, showing that firms with higher capital expenditures exhibit less timing than do firms with lower capital expenditures. Finally, the coefficient for the greater free cash flow firm dummy (set equal to 1 for high free cash flow firms) times the Baa decile variable is -75.3 with a t-value of 1.82 and a p-value of 0.069, showing that firms with higher free cash flow exhibit more timing than do firms with lower free cash flow. Thus, it appears that larger firms and firms with less financial constraints are more likely to attempt to time debt markets.
We examine whether the timing of debt issuance is affected by the current level of interest rates relative to historical rates, after we have accounted for total financing and capital expenditures. While controlling for other factors that affect the issuance decision, we examine empirically whether corporate managers time the issuance of external debt with respect to interest rates. We find that debt issuance activity is very much affected by the level of interest rates relative to historical rates.
Our results on debt issuance are consistent with the survey results on debt timing in Graham and Harvey (2001) and Bancel and Mittoo (2004). Like these two studies, our results indicate that financial managers try to issue debt when rates are "particularly low" if the interpretation of "particularly low" is low relative to previous levels. We show that debt issuance, measured by total amount issued and number of issues of new debt, relates to the relative level of interest rates in comparison to their historical values over the prior 10 years and/or by the lagged historical levels. Our evidence indicates that debt issuance as a proportion of total capital issuance (including equity) reflects timing effects, and that debt issuance relative to capital expenditures also reflects timing.
We find strong evidence that the amount of debt issued and the number of debt issues are related, both to the absolute level of interest rates and to their levels relative to historical rates. After controlling for other market conditions or corporate characteristics that can affect issuance, we find that the firms in our sample issue significantly higher amounts of long-term debt when long-term interest rates are low in either absolute or relative terms.
When interest rates decline, companies tend to refinance past debt that is eligible for refinancing. Therefore, we also consider the effects of refinancing transactions on debt issuance related to levels of interest rates. We find that refinancing is more common in our sample when interest rates are at low levels relative to their history. Even when we focus on the non-refinancing transactions, we still find that debt issuance is significantly greater when relative interest rate levels are low. Thus, debt timing is not merely driven by refinancing.
The authors are grateful to Stan Block, Mike Cliff, Michel Dubois, Espen Eckbo, Wayne Ferson, John Graham, Robin Greenwood, Robert Kieschnick, Erik Lie, Steve Lim, Pete Locke, Ehud Ronn, Andy Waisburd, the seminar participants at the 2004 Winter Finance Workshop, the 2004 European Meetings of the Financial Management Association, the 2004 Meetings of the Financial Management Association, the University of Alabama, University of Connecticut, Ohio University, and the editor and referees for their helpful comments. We appreciate helpful discussions about corporate borrowing strategies and practices with Tom Hund, CFO, and Linda Hurt, Treasurer, of Burlington Northern Santa Fe. Professors Mann, Mihov, and Rodriguez acknowledge research funding from the Charles Tandy American Enterprise Center and the Luther King Capital Management Center for Financial Studies at Texas Christian University.
Bancel, F. and U. Mittoo, 2004, "Cross-Country Determinants of Capital Structure Choice: A Survey of European Firms," Financial Management 33, 103-132.
Barclay, M. and C. Smith, 1995, "The Maturity Structure of Corporate Debt," Journal of Finance 50, 609-632.
Brounen, D., A. de Jong, and K. Koedijk, 2004, "Corporate Finance in Europe: Confronting Theory with Practice," Financial Management 33, 71-101.
Burch, T., W. Christie, and V. Nanda, 2004, "Do Firms Time Equity Offerings? Evidence from the 1930s and 1940s," Financial Management 33, 5-23.
Diamond, D. W., 1991, "Debt Maturity Structure and Liquidity Risk," Quarterly Journal of Economics 33, 341-368.
Faulkender, M., 2005, "Hedging or Market Timing? Selecting the Interest Rate Exposure of Corporate Debt," Journal of Finance 60, 931-962.
Fisher, I., 1930, The Theory of Interest, as Determined by Impatience to Spend Income and Opportunity to Invest It, New York, NY, Macmillan.
Graham, J. and C. Harvey, 2001, "The Theory and Practice of Corporate Finance: Evidence from the Field," Journal of Financial Economics 60, 187-243.
Guedes, J. and T. Opler, 1996, "The Determinants of Maturity of Corporate Debt Issues," Journal of Finance 51, 1809-1833.
Houston, J. and C. James, 1996, "Bank Information Monopolies and the Mix of Private and Public Debt Claims," Journal of Finance 51, 1863-1889.
Hovakimian, A., T. Opler, and S. Titman, 2001, "The Debt-Equity Choice," Journal of Financial and Quantitative Analysis 36, 1-24.
Hoven, S. M. and D. C. Mauer, 1996, "The Determinants of Corporate Debt Maturity Structure," Journal of Business 69, 279-312.
Jung, K., Y. Kim, and R. M. Stulz, 1996, "Timing, Investment Opportunities, Managerial Discretion, and the Securities Issue Decision," Journal of Financial Economics 42, 159-185.
Marsh, P., 1982, "The Choice between Equity and Debt: An Empirical Study," Journal of Finance 37, 121-144.
Myers, S.C., 1977, "Determinants of Corporate Borrowing," Journal of Financial Economics 5, 147-175.
Myers, S.C. and N.S. Majluf, 1984, "Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have," Journal of Financial Economics 13, 187-221.
Taggart, R.A., Jr., 1977, "A Model of Corporate Financing Decisions," Journal of Finance 32, 1467-1484.
(1) Studies that examine factors related to decisions to issue debt include Fisher (1930), Myers (1977), Myers and Majluf (1984), Diamond (1991), Barclay and Smith (1995), Guedes and Opler (1996), Hoven and Mauer (1996), Jung, Kim, and Stulz (1996), Houston and James (1996), Hovakimian, Opler, and Titman (2001), and Brounen, de Jong, and Koedijk (2004), among others.
(2) John Graham explained in private communication with us on this topic that although CFOs in the Graham and Harvey (2001) survey were not explicit about what "particularly low" meant to them, it was his interpretation, based in part on his conversations with the executives who responded to their survey, that they were referring to current interest rates compared to their historical levels.
(3) We thank an anonymous referee for suggesting the figure.
Christopher B. Barry, Steven C. Mann, Vassil T. Mihov, and Mauricio Rodriguez *
* Christopher B. Barry is a Professor of Finance and the Robert and Maria Lowdon Chair in Business Administration, M.J. Neeley School of Business at Texas Christian University, Fort Worth, TX. Steven C. Mann and Vassil T. Mihov are Associate Professors of Finance, M.J. Neeley School of Business at Texas Christian University, Fort Worth, TX. Mauricio Rodriguez is a Professor of Finance and Real Estate, M.J. Neeley School of Business at Texas Christian University, Fort Worth, TX.
Table I. Characteristics of Debt Issues This table reports the annual characteristics of a sample of 14,623 public straight debt issues and 10,209 seasoned equity offers (SEOs) by US nonfinancial firms during January 1970-April 2001 reported by Thomson Financial. Amounts are in millions of constant (January 2001) dollars. We calculate debt as a proportion of total issues by using monthly debt issue amounts divided by the sum of debt issues and SEOs. "T bills" are US Treasury bills. Since the SDC does not consistently report data on callability, puttability, and call and put dates until 1976, the sample period for the fraction of callable issues, fraction of puttable issues, and effective maturity is January 1976-April 2001. Year Number Average Average Fraction of Debt Monthly (Median) Ratio of Issues Issues Amount of Debt to That Were of SDC Total Monthly Callable Debt Issue Amount Issued (Debt + SEOs) 1970 408 86,027 88% (88%) N/A 1971 373 75,002 68% (64%) N/A 1972 268 45,931 58% (57%) N/A 1973 185 37,997 70% (69%) N/A 1974 291 71,372 88% (88%) N/A 1975 398 85,772 80% (82%) N/A 1976 276 54,803 72% (72%) 0.738 1977 259 41,940 73% (78%) 0.765 1978 211 31,799 72% (72%) 0.831 1979 203 38,415 75% (75%) 0.877 1980 318 56,341 65% (69%) 0.840 1981 261 42,409 65% (65%) 0.885 1982 287 38,715 58% (61%) 0.850 1983 221 28,550 39% (36%) 0.828 1984 216 39,433 81% (82%) 0.704 1985 386 65,455 77% (77%) 0.772 1986 707 149,317 87% (87%) 0.748 1987 441 89,786 87% (90%) 0.700 1988 356 77,638 92% (93%) 0.801 1989 340 76,242 90% (91%) 0.644 1990 293 52,163 83% (85%) 0.314 1991 605 100,033 76% (75%) 0.188 1992 852 151,290 83% (83%) 0.344 1993 1,049 181,645 82% (82%) 0.410 1994 497 72,288 74% (77%) 0.306 1995 612 91,909 68% (70%) 0.196 1996 739 105,412 69% (67%) 0.165 1997 1,003 110,239 72% (73%) 0.062 1998 1,152 169,736 80% (79%) 0.131 1999 680 146,687 66% (69%) 0.128 2000 518 126,397 62% (66%) 0.257 2001 218 64,134 84% (85%) 0.349 All years 14,623 2,604,877 74% (77%) 0.418 Year Fraction Average Mean of Issues (Median) Yield (Median) That Were to Maturity Yield of 90- Puttable Day T-bills 1970 N/A 9.2 (9.10) 6.39 (6.48) 1971 N/A 7.9 (7.80) 4.33 (4.33) 1972 N/A 7.5 (7.50) 4.07 (3.95) 1973 N/A 8.0 (7.90) 7.03 (7.21) 1974 N/A 9.5 (9.40) 7.83 (7.84) 1975 N/A 9.7 (9.50) 5.78 (5.56) 1976 0.000 8.8 (8.70) 4.97 (4.96) 1977 0.004 8.5 (8.40) 5.27 (5.11) 1978 0.005 9.7 (9.30) 7.19 (6.87) 1979 0.025 10.8 (10.50) 10.07 (9.50) 1980 0.006 12.8 (12.50) 11.43 (11.81) 1981 0.011 15.8 (15.70) 14.03 (14.72) 1982 0.056 14.5 (14.70) 10.61 (11.72) 1983 0.081 12.5 (12.50) 8.61 (8.70) 1984 0.093 13.9 (13.60) 9.52 (9.72) 1985 0.044 12.2 (11.80) 7.48 (7.20) 1986 0.038 9.9 (9.40) 5.98 (5.95) 1987 0.052 10.4 (9.90) 5.78 (5.68) 1988 0.067 10.8 (10.00) 6.67 (6.60) 1989 0.159 10.7 (9.90) 8.11 (8.03) 1990 0.041 9.8 (9.80) 7.49 (7.63) 1991 0.046 9.0 (9.00) 5.38 (5.52) 1992 0.141 8.3 (8.10) 3.43 (3.43) 1993 0.189 7.5 (7.10) 3.00 (3.01) 1994 0.215 8.4 (8.00) 4.25 (4.24) 1995 0.113 7.7 (7.30) 5.49 (5.45) 1996 0.146 7.7 (7.20) 5.01 (5.01) 1997 0.077 7.1 (6.90) 5.06 (5.05) 1998 0.082 6.6 (6.50) 4.78 (4.96) 1999 0.063 7.0 (6.90) 4.64 (4.56) 2000 0.050 7.9 (7.80) 5.82 (5.78) 2001 0.041 6.9 (6.80) 4.58 (4.65) All years 0.087 9.0 (8.40) 6.61 (5.81) Year Mean (Median) 10-Year Treasury Constant Maturity Rate 1970 7.35 -7.39 1971 6.16 -6.13 1972 6.21 -6.16 1973 6.84 -6.77 1974 7.56 -7.56 1975 7.99 -8.06 1976 7.61 -7.74 1977 7.42 -7.40 1978 8.41 -8.42 1979 9.44 -9.15 1980 11.46 -11.49 1981 13.91 -13.70 1982 13.00 -13.74 1983 11.11 -11.12 1984 12.44 -12.42 1985 10.62 -10.35 1986 7.68 -7.44 1987 8.38 -8.53 1988 8.85 -8.94 1989 8.50 -8.24 1990 8.55 -8.54 1991 7.86 -7.97 1992 7.01 -6.95 1993 5.87 -5.89 1994 7.08 -7.21 1995 6.58 -6.39 1996 6.44 -6.52 1997 6.35 -6.36 1998 5.26 -5.48 1999 5.64 -5.85 2000 6.03 -6.02 2001 5.07 -5.12 All years 8.15 -7.57 Table II. Debt Issued across Relative Rates and Growth Measures This table presents OLS tests of the monthly amount of debt issued (Panel A), the ratio of debt to secondary equity offers (SEOs) (Panel B), and the firm-level ratio of the annual amount of debt issued to the annual capital expenditures (Panel C) as a function of the historical distribution of interest rates, interest rate levels, and control variables. The sample consists of 14,623 public straight debt issues and 10,209 SEOs by US nonfinancial firms during 1970-2001. Amounts are in millions of constant (January 2001) dollars. We measure all independent variables as of the month immediately preceding the month of issuance. Baa is the Moody's Seasoned Baa Corporate Bond Yield. The Baa decile is the decile rating of the rate at the time of issuance among rates over the prior 10 years. We define term spread as the difference between the 10-year Treasury constant maturity rate and the yield on 90-day T-bills. We define credit spread as the difference between Baa and 10-year Treasury constant maturity rate. Parameter estimates show p-values in parentheses. In Panel C, for each firm-year, we calculate the annual amount of debt issued reported by Thomson Financial. We then form the ratio of annual amount of debt issued to annual capital expenditures (Compustat item 128). We measure the independent variables as of the end of the year of issuance. Panel A. Amount Issued per Month in 2001 Constant Dollars Intercept 3,018.4 ** (0.033) 5,174.7 *** (0.000) Baa decile -360.9 *** (0.000) -411.0 *** (0.000) Level of Baa at -257.5 ** (0.014) -397.9 *** (0.000) issuance, % Credit spread 2,336.8 *** (0.000) 2,754.5 *** (0.000) Term spread -263.8 * (0.057) -100.8 (0.497) S&P 500 P/E ratio 249.4 *** (0.000) S&P 500 market/ 1,179.1 *** (0.000) book ratio Baa lagged 60 months Baa lagged 120 months Adjusted [R.sup.2] 0.48 0.464 Number of 375 375 observations Intercept -250.3 (0.854) Baa decile Level of Baa at -550.4 *** (0.000) issuance, Credit spread 2,471.7 *** (0.000) Term spread -278.1 ** (0.040) S&P 500 P/E ratio 241.6 *** (0.000) S&P 500 market/ book ratio Baa lagged 60 months 180.0 ** (0.018) Baa lagged 120 286.8 *** (0.000) months Adjusted [R.sup.2] 0.497 Number of 375 observations Panel B. Proportion of Debt Issued to Total Issues (Debt Plus SEOs) Intercept 1.115 *** (0.000) 1.111 *** (0.000) Baa decile -0.017 *** (0.000) -0.018 *** (0.000) Level of Baa at -0.018 *** (0.000) -0.017 *** (0.000) issuance, % Credit spread 0.031 * (0.071) 0.024 (0.126) Term spread -0.004 (0.568) -0.014 ** (0.027) S&P 500 P/E ratio -0.010 *** (0.000) S&P 500 market/ -0.070 *** (0.000) book ratio Baa lagged 60 months Baa lagged 120 months Adjusted [R.sup.2] 0.114 0.137 Number of 375 375 observations Intercept 0.901 *** (0.000) Baa decile Level of Baa at -0.030 *** (0.000) issuance, % Credit spread 0.018 (0.280) Term spread -0.003 (0.619) S&P 500 P/E ratio -0.007 *** (0.000) S&P 500 market/ book ratio Baa lagged 0.021 *** (0.000) 60 months Baa lagged 0.003 (0.330) 120 months Adjusted [R.sup.2] 0.179 Number of 375 observations Panel C. Firm-Level Ratio of Debt Issued Divided by Capital Expenditures Intercept 3.610 *** (0.000) 3.234 *** (0.000) Baa decile -0.022 *** (0.000) -0.019 *** (0.000) Level of Baa at -0.037 (0.391) -0.012 (0.783) issuance, % Credit spread 0.173 (0.216) 0.004 (0.977) Term spread 0.068 (0.133) 0.036 (0.445) S&P 500 P/E ratio -0.057 *** (0.000) S&P 500 -0.245 *** (0.002) market/book ratio Baa lagged 60 months Baa lagged 120 months Adjusted [R.sup.2] 0.021 0.020 Number of 4,314 4,314 observations Intercept 1.149 ** Baa decile Level of Baa at -0.156 *** (0.031) issuance, % Credit spread -0.251 * (0.000) Term spread 0.127 *** S&P 500 P/E ratio 0.018 (0.085) S&P 500 (0.004) market/book ratio (0.239) Baa lagged 0.258 *** 60 months Baa lagged -0.033 (0.000) 120 months Adjusted [R.sup.2] 0.029 (0.277) Number of 4,314 observations *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table III. Timing of Net Issues versus Refinancings The table presents comparisons of the timing of net debt issues and debt refinancings. Panels A and B show the monthly amount of debt issued across historical interest rate deciles and use of proceeds. The amount issued is in millions of constant (January 2001) dollars. "Low" indicates that the Baa rate falls into the lowest three deciles of relative rates in the preceding 10 years. "Medium" indicates that the rate falls into deciles four to seven. "High" indicates that the rate falls into the highest three deciles. We use Compustat data to classify issues as debt refinancings if the net amount of debt by the firm for the fiscal year is negative, and otherwise classify issues as net debt issues. Among the net debt issues, we classify those cases in which the net debt was lower than the annual amount issued as "some refinancing." We use Compustat data to further classify debt issues by net stock repurchasers if the net amount of stock issued by the firm for the fiscal year is negative, and otherwise classify issues as net stock issuers. We present in parentheses t-statistics for the test that the mean in category "Low" is equal to the mean in category "High." Panel A. Net Debt Issues Compared to Refinancings Net Debt Relative Category Debt Issuers, Some Net Debt of the Baa Rate Refinancing Refinancing Issuers Low 1,396.7 2,876.0 1,424.7 Medium 504.2 1,654.5 1,245.2 High 92.1 1,082.5 564.9 t-test: Low compared (11.57) *** (10.02) *** (7.93) *** to high Panel B. Net Debt and Stock Issues Compared to Refinancings Relative Category Debt Net Debt Issuers, of the Baa Rate Refinancing Some Refinancing Negative Positive Negative Stock Stock Stock Issuance Issuance Issuance Low 848.7 418.1 1,542.6 Medium 264.8 183.1 861.0 High 38.2 48.4 320.3 t-test: Low compared (10.83) *** (7.87) *** (12.76) *** to high Relative Category Net Debt Issuers, Net Debt of the Baa Rate Some Refinancing Issuers Positive Negative Positive Stock Stock Stock Issuance Issuance Issuance Low 1,098.43 734.6 519.1 Medium 698.0 719.9 464.9 High 733.7 175.1 380.3 t-test: Low compared (3.41) *** (7.80) *** (2.51) ** to high *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. Table IV. Issuance Activity across Firm Characteristics We calculate the median values for the sample of debt-issuing firms for total assets, measured in constant January 2001 dollars; profitability, defined as EBITDA divided by total assets; a proxy for free cash flows, defined as after-tax income plus depreciation minus capital expenditures; and capital expenditures scaled to total assets. We classify the firms in relation to the sample median as large or small, with high or low profitability, high or low free cash flows, and high or low capital expenditures. Then we calculate the aggregate monthly amount of debt issued by each type of firm and regress it against the level of the interest rate, the rate's decile among historical rates, the credit spread, the term spread, and the S&P 500 P/E ratio. We measure all independent variables as of the month immediately preceding the month of issuance. Baa is the Moody's Seasoned Baa Corporate Bond Yield. We define term spread as the difference between the 10-year Treasury constant maturity rate and the yield on 90-day T -bills. We define credit spread as the difference between the Baa and 10-year constant maturity Treasury rates. The parameter estimate p-values appear in parentheses. Amount Issued per Month in Constant 2001 Dollars Large Firms Small Firms Intercept -1,200.5 (0.144) 1,811.8 *** (0.000) Baa decile -179.2 *** (0.000) -117.6 *** (0.000) Level of Baa at 28.5 (0.637) -99.1 *** (0.001) issuance, % Credit spread 907.2 *** (0.000) 512.2 *** (0.000) Term spread -376.6 *** (0.000) 90.3 ** (0.025) S&P 500 P/E ratio 197.7 *** (0.000) 3.1 (0.796) Adjusted [R.sup.2] 0.479 0.300 Number of observations 363 374 High Profitability Low Profitability Intercept 284.2 (0.648) 324.8 (0.564) Baa decile -191.8 *** (0.000) -98.0 *** (0.000) Level of Baa at -3.1 (0.946) -85.0 ** (0.041) issuance, % Credit spread 684.6 *** (0.000) 794.0 *** (0.000) Term spread -127.0 ** (0.038) -147.0 *** (0.008) S&P 500 P/E ratio 95.8 *** (0.000) 102.6 *** (0.000) Adjusted [R.sub.2] 0.409 0.420 Number of observations 369 375 High Free Cash Flow Low Free Cash Flow Intercept -1217.0 * (0.067) 1,793.6 *** (0.002) Baa decile -171.0 *** (0.000) -139.9 *** (0.000) Level of Baa 3.9 (0.479) -92.3 *** (0.036) at issuance, % Credit spread 517.1 *** (0.004) 884.9 *** (0.000) Term spread -97.0 (0.130) -200.0 *** (0.001) S&P 500 P/E ratio 173.9 *** (0.000) 19.6 (0.261) Adjusted [R.sup.2] 0.523 0.237 Number of observations 350 367 High Low Capital Expenditure Capital Expenditures Intercept 1,418.0 ** (0.022) -765.5 (0.228) Baa decile -114.8 *** (0.000) -207.3 *** (0.000) Level of Baa -80.7 * (0.079) 26.0 (0.590) at issuance, % Credit spread 925.7 *** (0.000) 480.5 *** (0.005) Term spread -193.9 *** (0.002) -108.2 (0.085) S&P 500 P/E ratio 26.5 (0.148) 164.8 *** (0.000) Adjusted [R.sub.2] 0.205 0.571 Number of observations 369 345 *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level.
|Printer friendly Cite/link Email Feedback|
|Author:||Barry, Christopher B.; Mann, Steven C.; Mihov, Vassil T.; Rodriguez, Mauricio|
|Date:||Sep 22, 2008|
|Previous Article:||How much does expertise reduce behavioral biases? The case of anchoring effects in stock return estimates.|
|Next Article:||Earnings and equity valuation in the biotech industry: theory and evidence.|