capital read

8/8/2019 Capital Read

1/33

Tests of the Pecking Order Theory and the Static

Tradeoff Theory of Optimal Capital Structure

by

Soku ByounSchool of Business

University of Southern IndianaTel: (812) 4641719Fax: (812) 465-1044

Email: [email protected]

Jong C RhimSchool of Business

University of Southern IndianaTel: (812) 4651637Fax: (812) 4651044

Email: [email protected]

March, 2003


2/33

Abstract

We investigate implications of the static tradeoff theory and the pecking order theory.

The results suggest that firms adjust their debt levels according to target debt ratios as

well as the pecking order. By combining the two theories we find that firms are more

likely to increase debt when they face financing needs as well as below-target debt level

while they are more likely to decrease debt level when they have surplus cash flows as well

as above-the-target debt level. The pecking order is found to be a much more binding

force for small firms, supporting the hypothesis that small firms are more likely to follow

the pecking order because of the difficulty in accessing external financing sources due to

asymmetric information. We also find that small firms are significantly slower than largefirms in adjusting the debt level when the adjustment requires an increase in debt level

according to the target adjustment model.


3/33

1 Introduction

The existence of debt financing generates agency costs of debt under informational asym-

metry: the stockholders incentive to take sub-optimal risky projects which transfer wealth

from bondholders to stockholders (Jensen and Meckling, 1976) and to abandon profitable

projects in some future states (Myers, 1977). If debt is used as a valid signal of a more

productive firm (Ross, 1977), an increase in the amount of debt may reduce the agency

costs associated with informational asymmetry. Also, debt reduces the agency costs of free

cash flow (Jensen, 1986). The tradeoff theory views a manager as trading off the benefits

from debt financing against the various costs of debt. The marginal agency cost of debt

is regarded as an increasing function of debt in a capital structure. Therefore, a man-

ager, acting as a shareholder value maximizer, should borrow up to the point where the

marginal value of the benefits from debt financing including interest tax shields is equal to

the marginal cost of debt including agency and financial distress costs. Barnea, Haugen and

Senbet (1981) argue that a firm reaches an optimal capital structure when the costs associ-

ated with agency problems are balanced by the benefits associated with different financial

contracts in terms of their inherent ability to resolve agency problems and tax exposure.

Another idea is that informational asymmetries between insiders and outsiders introduce

incentive problems in financial relationship. The pecking order theory states that firms

prefer internal financing and if external financing is required, they issue the safest security

first. Managers will choose to issue debt when investors undervalue the firm and issue equity

when they overvalue the firm. Recognizing this policy of managers, investors will perceive

an equity issue as bad news, making the cost of issuing equity higher. If the firm can use

internal financing sources or issue low-risk debt, then the cost of asymmetric information

can be minimized. If the manager has better information than investors, it is better to issue

debt than equity (Myers and Majluf, 1984). That is, firms issue debt first, then possibly

hybrid securities such as convertible bonds, then equity as a last resort.

Previous studies provide mixed empirical evidence for the two theories. Evidence in

favor of the tradeoff theory includes industry effects of optimal ratios, the negative relation

of leverage ratios to intangible assets proxied by research and development expenditures,

2


4/33

and mean reversion in debt ratios. Bradley, Jarrell and Kim (1984) find that firms optimal

leverage is inversely related to the expected costs of financial distress and to the amount

of non-debt tax shields. They also find the highly significant inverse relation between firm

leverage and earnings volatility. MacKie-Mason (1990) provides evidence that firms issue

less debt when they have tax loss carry-forwards.

According to Myers (1993), the most telling evidence against the tradeoff theory is the

inverse correlation between profitability and financial leverage. Titman and Wessels (1988)

find a significant negative relationship between profitability and debt ratios.1 However,

the tradeoff theory predicts the opposite relationship unless profitable firms incur more

agency costs than less profitable firms as the debt ratio increases. Titman and Wessels(1988) find no relationship between debt ratios and a firms expected growth, non-debt tax

shields, volatility, or the collateral value of its assets. The pecking order theory suggests

that there is no well-defined optimal capital structure; instead the debt ratio is the result

of hierarchical financing over time (Myers, 1984). Kester (1986), in his study of debt policy

in U.S. and Japanese manufacturing corporations, finds that the return on assets is the

most significant explanatory variable for actual debt ratios. MacKie-Mason (1990) asks

the question, Do firms care who provides their financing? His result suggests that the

importance of asymmetric information gives a reason for firms to care about who provides

the funds (e.g., between public and private debt) because different fund providers have

different access to information about the firm and different ability to monitor firm behavior.

This is consistent with the pecking order theory implied by Myers and Majluf (1984) since

private debt will require better information about the firm than public debt.

However, distinguishing these theories in empirical tests has proven difficult as variables

describing one theory can also be classified for the other theory. For example, as pointed

out by Booth et al. (2001), profitability is highly correlated with growth opportunities,

suggesting that the negative relation between profitability and leverage can be a reflection

1 Rajan and Zingales (1995) and Booth et al. (2001) also report some evidence of anegative correlation between profitability and leverage among G7 countries and developingcountries, respectively.

3


5/33

of higher financial distress costs for high-growth firms. Shyam-Sunder and Myers (1999)

argue that many of the current empirical tests lack sufficient statistical power to distinguish

between the theories.

We adopt a direct approach in testing the theories by examining firms capital structure

decisions facing financing needs and/or deviation from target debt ratios. With the new

approach, our study focuses on several issues. First, we test the pecking order theory

against the trade-off theory and vice versa for overall firms capital structure decisions. We

investigate potential asymmetric decisions when firms face external financing needs and

when they face financial surplus. Similarly, we allow differential adjustment speeds for

increasing and decreasing the debt level to an optimal debt level according to the tradeofftheory. If the adjustment cost of increasing a debt level is higher than that of decreasing,

it may not be surprising to see low R-square and a lack of power in the linear models.

Second, rather than taking the theories as exclusive, we explore the potential complementary

workings of the two theories. It is possible that firms have a target debt level and still prefer

internal funds to finance new investments rather than outside funds, ceteris paribus. Firms

may also adjust their debt levels to the targets when facing financing needs or having surplus

cash flows. Accordingly, we investigate firms capital structure decisions considering both

the financing needs and the target debt ratio. Third, we examine the possibility of different

financing decisions by small vs. large firms, and by dividend-paying vs. non-paying firms.

Since firm size and dividend payment provide important information about the firms agency

costs and the degree of asymmetric information, small or non-dividend-paying firms may

have quite different capital structure policies from large or dividend-paying firms.

We find evidence that firms adjust their debt levels according to target debt ratios.

The pecking order theory also explains a considerable portion of the variation in debt

levels. Both theories are not distinguishable in explaining firms capital structure decisions.

The two theories appear to be supplementary rather than contradictory. Firms are more

likely to increase debt level when they face external financing needs as well as below-target

debt level and use surplus cash flows to lower debt toward target levels. Overall, firms

financing decisions are in accordance with both the pecking order and the tradeoff theories.

4


6/33

The pecking order is found to be a much more binding force for small firms, supporting

the hypothesis that small firms are more likely to follow the pecking order because of

the difficulty in accessing external financing sources. We also find that small firms are

significantly slower when the adjustment requires an increase in debt level according to the

target adjustment model.

The rest of the paper proceeds as follows. In Section 2, we discuss the models. In Section

3, the data and estimation procedures are discussed and estimation results are reported

with their implications. Also, non-nested model tests are applied to test the tradeoff theory

against the pecking order theory and vice versa. Concluding remarks are in Section 4.

2 Static Capital Structure Choice Models

As mentioned in the introduction, the existence of debt financing generates various

agency costs. Unprofitable firms with high long-term debt levels may be reluctant to is-

sue equity because the wealth transfer from shareholders to bondholders may exceed the

increased value associated with improving the capital structure. These agency costs are

mitigated if the firm issues short-term debt rather than long-term debt (Barnea, Haugen

and Senbet, 1980). The firm may also rely on short-term debt to finance the retirement

of its long-term debt because of the low agency costs of short-term debt. 2 The firm may

deviate from the average long-term debt level but not from the target ratio; that is, it

is possible that the long-term debt alone does not reflect the firms real decision on its

capital structure. Thus using only long-term debt may bias the results against the target

adjustment model. Rajan and Zingales (1995) also suggest that the effects of past financing

decisions are best represented by the ratio of total debt to capital. Accordingly, we develop

the target adjustment model as follows:

2 For example, the commercial paper (CP) market is a large source of corporate short-term funds. Nayar and Rozeff (1994) report that there were $525 billion CP outstandingin 1991. They also provide evidence that the announcements of CP ratings influence stockreturns.

5


7/33

TDit = 1 + 2TDEit + 3TDEitMit + it (1)

where

TDit = Changes in total debt for firm i from time t 1 to t,

TDEit = TD

it TDit1,

TDit

= Target total debt (including short-term as well as long-term) for firm i

at time t,

TDit1 = Total debt for firm i at time t 1, and

Mit = Dummy variable equal to one ifTDEit > 0, zero otherwise,

where TDE represents the deviation of the total debt level from its target.3 Note that the

specification in equation (1) allows different adjustment costs for increasing and decreasing

the debt level. If firms are strictly following the target adjustment model, we expect 2 = 1

and 3 = 0. However, if the adjustment costs are higher when the firm increases its debt

level than when the firm decreases it, we may expect a negative 3.

For the pecking order model, consider the cash flow identity:

OCFit Xit Wit = INTit TDit + DIVit Eit (2)

where

OCFit = Operating cash flows after interest and taxes for firm i at time t,

Xit = Net capital expenditures of firm i at time t,

Wit = Change in net working capital excluding short-term debt for firm i from

time t 1 to t,

3 One may consider using total liabilities in defining the debt ratio which can be viewedas a proxy for what is left for shareholders in case of liquidation. However, as pointed outby Rajan and Zingales (1995), total liabilities include items such as accounts payable whichare used for transactions purposes rather than for financing and overstate the amount ofleverage.

6


8/33

INTit = Interest payment of firm i at time t,

TDit = Change in total debt outstanding for firm i from time t 1 to t,

DIVit = Dividend payments of firm i at time t,

Eit = Change in equity outstanding for firm i from time t 1 to t.

Now we define financial deficit as4

DEFit = Intit + DIVit + Xit + Wit OCFit (3)

And we specify the pecking order model as

TDit = 1 + 2DEFit + 3DEFitQit + it (4)

where Qit is a dummy variable which equals one for positive DEFit and zero otherwise. Note

that from the cash flow identity of equation (2), the financial deficit should be supplemented

either by debt or equity.5 The pecking order theory predicts that firms with a positive

financial deficit are likely to issue debt, relying on equity as a last resort. Therefore the

hypothesis to be tested is 1 = 0 and 2 > 0. The motivation for the interaction term

is not as clear for this model as it is for the target adjustment model. Shyam-Sunder and

Myers (1999) maintain that the pecking order theory predicts that the firm will only issue

or retire equity as a last resort. In that case, we would expect 3 = 0. However, no other

study documents firms decisions to retire capital and it would be interesting to see how

the pecking order actually works in reducing the capital with financial surplus.

For estimation of the target adjustment model and the pecking order model, we scale all

the variables either by book value or market value of total assets. In measuring the target

debt ratio, we use the historical mean of debt ratio (total debt / total assets or long-term

4 When we consider only long-term debt, we define DEF with short-term debt includedin net working capital.

5 Here our implicit assumption is that the dividend policy and investment decision areexogenous and independent of the financing decision.

7


9/33

debt / total assets) times total assets as the instrument for the target debt ratio.6 While

Shyam-Sunder and Myers (1999) and Myers (1984) argue that there are rational reasons

for managers to specify debt targets in terms of book values, Titman and Wessels (1988)

incline to the use of debt level measured at market value. Accordingly, we also estimate

the same models using total debt over the market value of total assets. For the market

value of total assets, we replace the book value equity with its market value. We also

consider both total debt and long-term debt in measuring debt ratios. Without a theory

to guide whether the capital structure decision should be based on book value or market

value and whether total debt or long-term debt, any capital structure study implicitly takes

a position by choosing one method against another. By considering alternative estimationapproaches and examining the robustness of the results, we can make confident conclusions

in the investigation of the tradeoff theory and the pecking order theory.

3 Empirical Analysis

3.1 Data

The time period analyzed in this study is 1981 through 2000. Our primary data source

consists of the Annual Industrial COMPUSTAT files. Financial firms and regulated utilities

are excluded from the sample because these firms have very different capital structures and

the financing decisions of these firms may not convey the same information as for non-

financial and non-regulated firms.7 For example, a relatively high leverage ratio is normal

for financial firms, but the same high leverage ratio for non-financial firms may indicate

a possibility of financial distress. We also require firms to have at least $3 million of

assets to be included in our sample. This requirement may bias our sample toward large

6 Shyam-Sunder and Myers (1999) use this definition but they only include firms withfull 20-year data. Accordingly, their sample size is only 157 firms. We also use the 10-yearmoving average; that is, for a given year the target debt level is the average of past 10-yeardebt ratios times total assets. However, the results are identical and not reported.

7 Financial firms are represented by SIC codes 6000-6799 and utilities are in SIC codes4800-4999. Accordingly we exclude these industries from our sample.

8


10/33

firms. However, large firms should have relatively easy access to the debt market and our

sample can mitigate the concern about the liquidity constraints on real investment due

to asymmetric information problems, as we assume real investment is exogenously given.

Calomiris and Hubbard (1990) show that the allocation of new funds across classes of

borrowers can ration funds away from some classes of borrowers who would receive credit in

the absence of asymmetric information. Hence, the terms under which intermediary credit

is available are key determinants of investment especially for firms lacking easy access to

direct credit (Bernanke, 1983). We will further discuss the effect of firm size on the capital

structure decision in the next section. Finally, we include only firms that have a complete

record over at least 11 consecutive years of the variables considered in our analysis. In thisway, we identify 1,236 firms.

Table I presents summary statistics on the book value of assets, the market value of

equity, long-term debt ratio and total debt ratio for the sample for the years 1981, 1990

and 2000 as well as the full sample of 18,236 firm-year observations. Overall the average

total debt to asset ratio is .2587 and the average long-term debt to asset ratio is .1977.

Table I

3.2 Estimation and Results

Common problems found in panel data are detected in this sample; the ordinary least

square (OLS) assumption of independent errors is unlikely to be satisfied. The most serious

problem of OLS estimation comes from the dependence of the residuals. Residuals show

the presence of autocorrelation and homoskedasticity.8 The coefficients of skewness are

significantly negative and the coefficients of excess kurtosis are unduly large, indicating thedistribution of residuals is negatively skewed and has much higher peak and fat tails than

8 The Breusch-Pagan tests for homoskedasticity are rejected for all regressions at 5 percentsignificance level. The White test also indicates the presence of significant heteroskedastic-ity.

9


11/33

a normal distribution.9

To see if there are outliers in the data, we calculate the diagonal values of the hat

matrix, DFFITS and DFBETAS as described in Belsley, Kuh and Welsch (1980). All the

diagonal values of the hat matrix are less than .05 and all the DFBETAS are less than

.35 in their absolute values.10 The OLS regressions with outlier exclusion produce almost

identical results. The problem of dependent errors seems to be more important with this

data set than that of outliers.

To address the error-dependence problem associated with the panel data, it is necessary

to use a special estimation procedure. The mixed effects model described in Goldstein

(1995) and Venables and Ripley (1999) employs a set of assumptions on the disturbancecovariance matrix that gives a cross-sectionally heteroskedastic and time-wise autoregressive

model. The estimation procedure first estimates the variance structure by maximizing the

marginal likelihood of the residuals from a least-squares fit of the linear model, and then

the fixed effects are estimated by maximum likelihood assuming that the variance structure

is known, which amounts to fitting by the generalized least square (GLS).

Table II reports the OLS and mixed effects GLS estimation results for the target ad-

justment and pecking order models. As a precaution against heteroskedasticity, we scaled

all the variables by the book value of total assets.

Panel A shows the estimates of OLS and GLS regressions of changes in total debt level

on various combinations of explanatory variables. Compared with the results from the OLS

estimation, the magnitude and significance of the GLS-estimated target adjustment coeffi-

cients are considerably increased (from .4392 to .5835 in row A1). The coefficient estimate of

the target adjustment interaction dummy variable remains significantly different from zero

9

Chi-square goodness of fit tests and Jarque-Bera (1980) Lagrange Multiplier tests fornormality of the OLS residuals are rejected for all regressions.

10 However, there are some significant DFFITS values. We estimate the OLS regressionsafter deleting the observations of which DFFITS are greater than .34 in their absolutevalues. This procedure is rather arbitrary. However, Maddala (1992) suggests an estimationprocedure which uses smaller weights for observations with absolute values of DFFITSgreater than .34 (See pp. 487490 in Maddala, 1992.).

10


12/33

(.1869 for the OLS and .1764 for the GLS in row A1), suggesting that the adjustment

costs are indeed different between increasing and decreasing the total debt level. On the

other hand, the estimates of pecking order coefficients remain the same (.8054 for the OLS

and .8071 for the GLS in row A2). The pecking order model indicates even a greater dif-

ference between the coefficient estimates of the positive financial deficit interaction dummy

variable (.6282) and the financial deficit variable (.8071). This result suggests that firms

use up their financial surplus to pay back outstanding debt but they are much less sensitive

in increasing the debt level facing the financing needs.

Table II

We also report estimation results including the target adjustment coefficients and the peck-

ing order coefficients in the same equation (row A3). The magnitude and significance of

the pecking order coefficient estimates and the counterparts of the target adjustment co-

efficients show little change. This result is contrary to the finding of Shyam-Sunder and

Myers (1999). In their results, the magnitude and significance of the coefficients of the tar-

get adjustment model are significantly reduced while those of the pecking order coefficients

change little when the two models are nested. Allowing different adjustment costs between

increasing and decreasing debt level appears to contribute to the different results.

We also estimate the models using long-term debt ratios. The results are provided

in Panel B. The explanatory power of the target adjustment model is indistinguishable

from that of the pecking order model in terms of R-squares. One thing to note is that

the estimated coefficient of the target adjustment interaction dummy variable becomes

insignificant in both magnitude and significance. On the other hand, the coefficient estimate

of the financial deficit interaction dummy variable remains significant and negative even

when the two models are nested. The two models are very similar in all other respects.

We further estimate the models using the market value of total assets as a denominator

in calculating the debt ratios in Table III. We also scale all the explanatory variables by the

market value of total assets. We report only the GLS estimation results. The market-value-

11


13/33

based debt ratios produce larger adjustment coefficient estimates for the target adjustment

model in row A1. In row A2, the estimated coefficient of the financial deficit interaction

dummy variable is 1.4072, greater, in its absolute value, than the coefficient estimate of

financial deficit, 1.0245. The magnitudes of these coefficient estimates remain unchanged in

the presence of the target adjustment variables in row A3. The pecking order seems to be

working only in reducing debt when there is financial surplus. We also estimate the same

model with long-term debt only and do not find any significantly different results compared

to the results in Table II. The estimation results seem to be influenced by the fluctuations in

market-value-based debt ratios due to changes in market value of equity. Even when there is

no change in debt level, the debt ratio changes as the market value of equity changes. Thismay induce a spurious relationship between the changes in debt and the financial deficit or

the deviation from the target.

Table III

Overall, both the target adjustment model and the pecking order model explain

considerable variation in firms debt financing. Consistent with our conjecture that adjust-

ment costs are higher for debt increasing than for debt decreasing, firms are much slower

in increasing debt ratio. Firms also use up financial surplus more easily to pay back out-

standing debt than they use debt to finance financial deficit.

3.3 Non-Nested Model Tests

Observing predicted signs with reasonable significance of estimated coefficients from the

GLS estimation, we still cannot make a definite conclusion for one theory against the other.In this section we apply the non-nested model test called the J test (See Davidson and

MacKinnon (1993) pp. 381-388.). The test is based on artificial nesting. Consider the

artificially nested model in our framework:

HN : TDit = (1)(1 +2DEFit +3DEFitQit) +(1+2TDEit +3TDEitMit) +it

(5)

12


14/33

where is a parameter that has been introduced to nest the pecking order model (say P1)

and the target adjustment model (say T1) within HN. When = 0, HN collapses to P1

and when = 1, HN collapses to T1. The problem is that the artificial model, HN, is not

estimable because the parameters are not separable. The solution to this problem is to

replace one of the two models with estimates. Suppose it is T1 we wish to test. Then we

can replace the s in T1 with their estimates (possibly GLS estimates) and estimate the

following artificial regression model:

TDit = 1 + 2DEFit + 3DEFitQit + (1 + 2TDEit + 3TDEitMit) + it. (6)

The J test uses the t statistic for the null hypothesis = 0.11

If the null hypothesis isrejected, then T1 is not rejected by P1. If the null hypothesis is not rejected, then T1 is

rejected by P1. Davidson and MacKinnon (1993) show that under the null hypothesis, ,

the estimate of, is asymptotically normal. Therefore the J test is a one degree of freedom

test irrespective of the number of observations. It is also possible to test P1 against T1, and

the artificial regression for doing so is essentially the same as the above, but with the roles

ofP1 and T1 reversed.

Table IV

The results of J tests are provided in Table IV. In calculating the t-test statistics, we

estimate the parameters using the GLS estimation procedure. The test statistics can be

compared with t-distribution with one degree of freedom. Even though the numbers are

slightly higher for the target adjustment model (or lower p-values in parentheses), the two

models are indistinguishable. We cannot reject one model against the other in all cases.

This result contrasts the results of Shyam-Sunder and Myers (1999) which leads to the

conclusion that the underlying financing behavior of firms is in accordance with the pecking

order model but not with the target adjustment model.

11 It is called the J test because and the s are estimated jointly in s.

13


15/33

3.4 Nested Models

The test results so far show that two theories are supplementary rather than contradic-

tory. Now it is interesting to see how the financing decisions are affected when both the

target adjustment and the pecking order are taken into consideration. The target adjust-

ment model can be rewritten as:

TDit = b1+b2TDEit+b3TDEitMit+b4TDEitMitQit+b5TDEit(1Mit)(1Qit)+it. (7)

All the variables are as defined in the previous section. Note that the interaction term

MitQit indicates the presence of financial deficit as well as the downward deviation of the

debt level from the target. In previous results we have seen that firms are much slower to

increase the debt level as shown by the significant and negative coefficient estimate of 3

in equation (1). We can investigate how the adjustment speed changes as firms face not

only the deviation from the target debt level but also external financing needs. Positive

estimate ofb4 would suggest that firms tend to make adjustments to their target debt when

they face external financing needs. Also, the interaction term (1 Mit)(1 Qit) indicates

the presence of financial surplus with the above-the-target debt level. If firms are making

downward adjustments of their debt levels when they have financial surplus, the estimate

of b5 would be positive.

Similarly, we specify the pecking order model as

TDit = a1+a2DEFit+a3DEFitQit+a4DEFitQitMit+a4DEFit(1Qit)(1Mit)+it. (8)

If the firms are making adjustments to their target debt with surplus cash flows or facing

financing needs, the estimates ofa4 and a5 would be positive.

Table V

As before, we scale all the variables by either book or market value of total assets.

The estimation results are reported in Table V. The coefficient estimates of the additional

interaction variables in equation (7) are all positive and significant. These results suggest

14


16/33

that firms are more likely to increase debt level when they face financing needs as well as

below-target debt level. This sharply contrasts to the results found in Tables II and III that

show much slower upward adjustment of debt level toward the target. Also, firms are more

likely to decrease debt level when they have surplus cash flows as well as above-the-target

debt level.

The estimated coefficients of the additional interaction variables in equation (8) are also

positive and highly significant. In fact, we cannot reject the null hypothesis ofa2 + a4 = 1

in all cases at 5 percent significance level. This suggests that the entire financial deficit

tends to be financed by debt when firms have below-the-target debt level. Firms are also

using surplus cash flows to lower debt toward their targets. The results in Table V provideevidence that firms financing decisions are in accordance with both the pecking order and

the tradeoff theories and that we can better explain firms capital structure decisions by

taking the two theories as complimentary rather than exclusive ones.

3.5 Financing Decisions and the Roles of Dividend and Firm Size

In this section we investigate whether the financing decisions are different between small

and large firms, and between dividend-paying firms and non-paying firms. The cost of is-

suing debt or equity is much higher for small firms than large firms (Titman and Wessels,

1988). Small firms are subject to severe asymmetric information problems but less agency

problems. Also, small firms have less access to external funds than do large firms (Bernanke,

1983). This suggests that the adjustment speed can be different between small and large

firms. Specifically, we hypothesize that the pecking order model is a more profound de-

scription of the financing decisions of small firms. Also, the adjustment speed differential

between increasing and decreasing the debt level is likely to be greater for small firms thanfor large firms.

In the above analysis we take dividend as predetermined. In our sample, 682 out of 1,236

firms (55 percent) did not pay any dividend during the sample period. Fazzari and Peterson

(1993) argue that low-dividend firms are the most likely to face financial constraints and

show that non-dividend-paying firms rarely make use of new equity financing. Also, Barclay

15


17/33

and Smith (1995) provide evidence that firms debt levels are related to dividend yields.

To address these issues, we estimate the pecking order model and the tradeoff target

adjustment model by dividing firms into small and large firms based on the median of aver-

age total assets, and by splitting the sample into dividend-paying and non-dividend-paying

groups.12 We expect that the pecking order model will be a better description of financing

decisions of non-dividend paying firms.

Table VI

In Table VI we compare the target adjustment and pecking order models between small andlarge firms. All the estimation results are based on book value debt ratios. Clearly the esti-

mates of the adjustment coefficients (in the third column of Panel A) are significantly greater

for small firms (.6279) than for large firms (.4927). Whether we use total debt (in Panel A)

or long-term debt (in Panel B) in defining the debt ratio, the results are almost identical.

The results also show that small firms are much more likely to follow the pecking order,

possibly because of the difficulty in accessing external financing sources. The coefficient

estimates ofDEF is .9322 (.7286) for small firms and .4263 (.3015) for large firms based

on the total (long-term) debt. Also, highly significant and negative coefficient estimates

for the interaction variables provide evidence that small firms are much slower when the

adjustment requires an increase in debt level. We further estimate the same groups of firms

with market value debt ratios. The results are qualitatively the same and not reported here.

Table VII

Table VII reports the comparison of the target adjustment and pecking order models be-

tween dividend-paying and non-dividend-paying firms. The estimates for the target adjust-

12 We first calculate the average of total assets for each firm over the sample period andthen the median of the averages across the cross sectional firms. Also, dividend-paying firmsare defined as those that paid at least one cash dividend during the sample period.

16


18/33

ment coefficient (the third column) tend to be greater for dividend-paying firms than for

non-dividend-paying firms (.6878 vs. .6025 for total debt and .6019 vs. .5432 for long-term

debt). On the other hand, the estimates for the financial deficit coefficients are greater for

non-dividend-paying firms than for dividend-paying firms (.5451 vs. .4613 for total debt

and .4891 vs. 4012 for long-term debt). The significant differences between the coefficient

estimates for interaction variables in the pecking order model suggest that non-dividend-

paying firms use less debt to finance their financial deficits. Overall, the pecking order seems

to be a more binding force for non-dividend-paying firms as it does for small firms.

4 Conclusion

The paper investigates implications of the tradeoff theory and the pecking order theory.

We find that the difference between the target debt ratio and actual debt ratio is an im-

portant determinant of the change in debt level. The results suggest that firms adjust their

debt levels according to target debt ratios but the upward adjustment of the debt ratio is

much less sensitive to the deviation from the target ratio, reflecting higher adjustment costs

for debt increasing than decreasing.

The pecking order model also captures a significant portion of variations in debt ratio.

When we allow different slopes for positive and negative financial deficit, the results suggest

that firms use financial surplus to pay back their outstanding debt but the increase in debt

level is much less responsive to their financing needs. Firms appear to consider both long-

term and total debt levels in marking their optimal capital structure decisions.

We formally test the tradeoff theory against the pecking order theory and vice versa

using non-nested model tests. The test results are not clear between the tradeoff theory

and the pecking order theory. We further explore the possibility of supplementary workings

of the theories. The results suggest that firms are more likely to use debt to finance their

financial deficits when they have below-the-target debt level while they tend to use surplus

cash flows to lower debt toward the targets. We conclude that firms capital structure

decisions can be better explained by taking the two theories as complementary rather than

17


19/33

exclusive ones.

We also compare the pecking order model and the tradeoff target adjustment model

between small and large firms and also between non-dividend and positive-dividend paying

firms. The pecking order is found to be a much more binding force for small firms, supporting

the hypothesis that small firms are more likely to follow the pecking order because of the

difficulty in accessing external financing sources. The results also provide weak evidence

that non-dividend-paying firms are more concerned about the pecking order than dividend-

paying firms. We also find evidence that small firms are significantly slower when the

adjustment requires an increase in debt according to the target adjustment model.

18


20/33

References

Barclay, M.J. and C.W. Smith, JR., 1995, The Maturity Structure of Corporate Debt,

Journal of Finance 50, 609631.

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.

Barnea, A., R.A. Haugen and L.W. Senbet, 1981, Market Imperfections, Agency Prob-

lems, and Capital Structure: A Review, Financial

Management, Summer, 722.

Belsley, D., E. Kuh and R.E. Welsch, 1980, Regression Diagnostics, Wiley.

Bernanke, B., 1983, Non-Monetary Effects of the financial Crisis in the Propagation of

the Great Depression, American Economic Review 73, 257276.

Booth, L., V. Aivazian, A. Demirguc-Kunt, and V. Maksimovic, 2001, Capital Structures

in Developing Countries, Journal of Finance 56, 87130.

Bradley, M., G.A. Jarrell and E.H. Kim, 1984, On the Existence of an Optimal Capital

Structure: Theory and Evidence, Journal of Finance 39, 857878.

Calomiris, C.W. Hubbard, R.G., 1990, Firm Heterogeneity, Internal Finance and Credit

Rationing, Economic Journal 100, 90104.

Davidson, R. and J.G. MacKinnon, 1993, Estimation and Inference in Econometrics, Ox-

ford University Press.

Fazzari, S.M and B.C. Peterson, 1993, Working Capital and Fixed Investment: New

Evidence on Financing Constraints, RAND Journal of Economics 24, 328-342.

Goldstein, H., 1995, Multilevel Statistical Models, Halsted Press, New York.

19


21/33

Jarque, C.M. and A.K. Bera, 1980, Efficient Tests for Normality, Homoskedasticity and

Serial Independence of Regression Residuals, Economic Letters 6, 255259.

Jensen, M.C., 1986, Agency Costs of Free Cash Flow, Corporate Finance Takeovers,

American Economic Review 76, 323339.

Jensen, M.C. and W.H. Meckling, 1976, Theory of the Firm: Managerial Behavior,

Agency Costs and Ownership Structure, Journal of Financial Economics 3, 305

360.

Kester, W.C., 1986, Capital and Ownership Structure: A Comparison of United States

and Japanese Manufacturing Corporations, Financial Management 15, 516.

MacKie-Mason, J.K., 1990, Do Firms Care Who Provides Their Financing? Asymmetric

Information, Corporate Finance, and Investment, NBER Project Report edited by R.

Glenn Hubbard, The University of Chicago Press, 63 103.

MacKie-Mason, J.K., 1990, Do Taxes Affect Corporate Financing Decisions? Journal

of Finance 45, 14711494.

Maddala, G.S., 1992, Introduction to Econometrics; Second Edition, Macmillan.

Myers, S.C., 1977, Determinants of Corporate Borrowing, Journal of Financial Eco-

nomics 5, 147175.

Myers, S.C., 1984, The Capital Structure Puzzle, Journal of Finance 39, 575592.

Myers, S.C., 1993, Still Searching for Optimal Capital Structure, Journal of Applied

Corporate Finance 39, 414.

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 Eco-

nomics 13, 187221.

Nayar, N. and M.S. Rozeff, 1994, Ratings, Commercial Paper, and Equity Returns,

Journal of Finance 49, 14311449.

20


22/33

Rajan, R.G. and L. Zingales, 1995, What Do We Know about Capital Structure? Some

Evidence from International Data, Journal of Finance 50, 1421 1460.

Ross, S.A., 1977, The Determination of Financial Structure: the Incentive-Signaling Ap-

proach, The Bell Journal of Economics, 2340.

Shyam-Sunder, L. and S.C. Myers, 1999, Testing Static Trade-Off Against Pecking Order

Models of Capital Structure, Journal of Financial Economics 51, 219-244.

Titman, S. and R. Wessels, 1988, The Determinants of Capital Structure Choice, Journal

of Finance 43, 119.

Venables, W.N. and B.D. Ripley, 1999, Modern Applied Statistics with S-PLUS, Third

Edition, Springer.

21


23/33

Table ISummary Statistics for Selected Variables

The sample consists of 18,236 firm-year observations and the time period is 1981through 2000. Our primary data source consists of the Annual Industrial

COMPUSTAT files. The sample includes only firms that have at least $3 million ofassets as of 2000 and a complete record over at least 11 years of the variablesconsidered.

StandardMean Minimum Maximum Deviation

Book value of total assets1981 410 4.828 6961 11011990 1100 3.013 77734 37292000 2599 3.239 87495 6695

1981-2000 1352 3.001 92473 4238

Market value of equity1981 279 2.387 6267 7631990 986 2.097 54093 33282000 4947 2.094 274428 20143

1981-2000 1622 2.005 274428 7956

Total debt / total assets1981 0.2076 0.009 0.7230 0.13481990 0.2790 0 0.9153 0.1828

2000 0.2832 0 0.9547 0.18381981-2000 0.2587 0 0.9992 0.1754

Long-term debt / total assets1981 0.1508 0 0.7045 0.11961990 0.2087 0 0.9140 0.16572000 0.2215 0 0.9084 0.1723

1981-2000 0.1977 0 0.9873 0.1596


24/33

Table IIEstimation Results of the Target Adjustment and Pecking Order Models Based on

Book Value

The sample consists of 1236 firms with relevant Compustat data for 1981 to 2000, 18,236total observations. The target debt level is based on the historical mean of debt to total asset

ratio and all the explanatory variables are scaled by the total assets ( A ). In panel B, thetarget debt level is based on the historical mean of debt to total debt plus market value ofequity ratio and all the explanatory variables are scaled by the total debt plus market value ofequity (S). TDE(LDE) represents the deviation of the total (long-term) debt level from itstarget and DEFthe financial deficit. Mis a dummy variable equal to one if TDE(LDE) ispositive, and zero otherwise. Qis equal to one if DEFis positive and zero otherwise.Numbers in parentheses are t-values. *Not different from zero at the significance level of 1%.

Panel A. The Dependent Variable is Changes in Total Debt Scaled by Book Value ofTotal Assets (A).

Constant TDE/A (TDE/A).M DEF/A (DEF/A).Q R2

(A1) OLS .0187 .4392 -.1869 .1266(12.95) (35.61) (-7.307)

GLS .0184 .5835 -.1764 .1493(12.71) (39.11) ( -6.86)

(A2) OLS .0248 .8054 -.6310 .1883(20.94) (57.09) (-33.35)

GLS .0243 .8071 -.6282 .1907(20.12) (57.13) (-33.09)

(A3) OLS .0302 .4639 -.1649 .7440 -.5841 .2381(21.51) (34.29) (-7.02) (55.78) (-32.49)

GLS .0285 .5115 -.1501 .7452 -.5780 .3070(20.06) (37.83) (-6.35) (56.45) (-32.36)


25/33

(Table II Continued)

Panel B. The Dependent Variable is Changes in Long-Term Debt Scaled by Book Valueof Total Assets (A).

Constant LDE/A (LDE/A).M DEF/A (DEF/A).Q R2

(B1) OLS .0010 .4410 -.0119* .1381(8.28) (32.95) (-.5070)

GLS .0095 .5038 -.0046* .1567(7.80) (37.23) (-.1963)

(B2) OLS .0192 .6087 -.4506 .1622(19.14) (50.32) (-29.33)

GLS .0187 .6113 -.4492 .1663(18.18) (50.44) (-29.14)

(B3) OLS .0188 .3707 .0136* .5552 -.4318 .2632(15.87) (29.55) (.6124) (48.36) (-29.44)

GLS .0173 .4266 .0202* .5569 -.4302 .2874(14.37) (33.67) (.9068) (48.98) (-29.51)


26/33

Table IIIEstimation Results of the Target Adjustment and Pecking Order Models Based on

Market Value

The sample consists of 1236 firms with relevant Compustat data for 1981 to 2000, 18,236total observations. The target debt level is based on the historical mean of debt to total debtplus market value of equity ratio, and all the explanatory variables are scaled by the total debt

plus market value of equity (S). TDE(LDE) represents the deviation of the total (long-term)debt level from its target and DEFthe financial deficit. Mis a dummy variable equal to one ifTDE(LDE) is positive, and zero otherwise. Qis equal to one if DEFis positive and zerootherwise. Numbers in parentheses are t-values.

Panel A. The Dependent Variable is Changes in Total Debt Scaled by Market Value ofTotal Assets (S).

Constant TDE/S (TDE/S).M DEF/S (DEF/S).Q R2

(A1) GLS .0238 .7026 -.4749 .1586(8.617) (30.09) (-11.71)

(A2) GLS .0521 1.0245 -1.4072 .1939(27.03) (62.78) (-62.92)

(A3) GLS .0554 .4528 -.1222 0.9448 -1.3440 .2373(21.53) (21.03) (-3.282) (58.49) (-60.82)

Panel B. The Dependent Variable is Changes in Long-Term Debt Scaled by MarketValue of Total Assets (S).

Constant LDE/S (LDE/S).M DEF/S (DEF/S).Q R2

(B1) GLS .0081 .5041 -.1249 .0866(4.223) (27.97) (-3.916)

(B2) GLS .0317 .6110 -.7954 .1198(22.59) (43.94) (-49.75)

(B3) GLS .0243 .3579 .1318 .5365 -.7448 .1909(13.27) (20.44) (4.255) (39.45) (-47.69)


27/33

Table IVResults of J-Tests for Artificially Nested Target Adjustment and Pecking Order Models

The first row represents the models that are tested as null hypothesis in the nested model. P1and T1 are the pecking order and target adjustment model without the interaction dummyvariable, and P2 and T2 are the pecking order and target adjustment model with theinteraction dummy variable included. The reported test statistics have t- distributions with one

degree of freedom. Numbers in parentheses are one-tailed p-values.

Dependent Variable T1(vs. P1) T2 (vs. P2) P1 (vs. T1) P2(vs. T2)

TD/A 55.89 65.89 22.26 65.15

(.0057) (.0143) (.0048) (.0049)

LD/A 35.97 66.95 23.86 58.43

(.0088) (.0133) (.0088) (.0054)

TD/S 39.66 62.82 27.26 31.45(.0080) (.0117) (.0080) (.0101)

LD/S 34.89 48.16 23.86 39.51

(.0091) (.0066) (.0133) (.0081)


28/33

Table VGLS Estimation Results of the Nested Target Adjustment and Pecking Order Models

The sample consists of 1236 firms with relevant Compustat data for 1981 to 2000, 18,236total observations. ?TDis a change in total debt and ?LDis a change in long-term debt. A isthe book value of total assets and Sis the market value of total assets in which the book

value of equity is replaced with the market value of equity. TDE(LDE) represents thedeviation of the total (long-term) debt level from its target and DEFthe financial deficit. Mis adummy variable equal to one if TDE(LDE) is positive, and zero otherwise. Qis equal to one ifDEFis positive and zero otherwise. Numbers in parentheses are t-values.

Panel A. The Target Adjustment Model

Dependentvariable

Constant TDE/A (TDE/A)xM

(TDE/A)xMQ

(TDE/A)x(1-Q)(1-M)

R2

?TD/A .0219(15.92)

.3232(21.93)

-.3356(-11.73)

.4830(20.80)

.5221(24.81)

.1526

?LD/A .0112(10.37)

.2522(18.24)

-.1539(- 5.55)

.4427(19.37)

.4585(23.78)

.1579

?TD/S .0246( 9.16)

.5114(20.13)

-.5891(-11.77)

.4324(18.77)

.5384(24.55)

.1442

?LD/S .0097( 4.98)

.3334(15.88)

-.2563(- 6.02)

.4142(19.80)

.5370(23.91)

.1362


29/33

(Table V Continued)

Panel B. The Pecking Order Model

Dependent

variable

Constant DEF/A (DEF/A)x

M

(DEF/A)x

MQ

(DEF/A)x

(1-Q)(1-M)

R2

?TD/A .0349(35.15)

.2718(14.41)

-.4835(-22.29)

.9089(38.15)

.2930(30.97)

.2000

?LD/A .0261(29.38)

.1662(10.18)

-.3208(-16.71)

.7917(37.30)

.2298(25.91)

.1728

?TD/S .0137( 7.28)

.1981( 5.51)

-.2218(- 6.16)

1.1189(27.00)

.2291(16.25)

.1539

?LD/S .0386(35.79)

.1980( 8.92)

-.7978(-34.97)

.8506(31.76)

.5600(131.4)

.4665


30/33

Table VI

Comparison of the Target Adjustment and Pecking Order Models between Small and

Large Firms

The sample consists of 1236 firms with relevant Compustat data for 1981 to 2000, 18,236total observations. The target debt level is based on the historical mean of debt to total assetratio and all the explanatory variables are scaled by the total assets ( A ). TDE(LDE)represents the deviation of the total (long-term) debt level from its target and DEF thefinancial deficit. Mis a dummy variable equal to one if TDE(LDE) is positive, and zerootherwise. Qis equal to one if DEFis positive and zero otherwise. Numbers in parenthesesare t-values.

Panel A. The Dependent Variable is Changes in Total Debt Scaled by Book Value of

Total Assets (A)

Constant TDE/A (TDE/A)M DEF/A (DEF/A)Q

Small .0140 .6279 -.1990(5.69) (27.50) (-5.05)

Large .0208 .4927 -.1035(13.08) (26.65) (-3.24)

Small .0274 .9322 -.8164(14.28) (48.97) (-32.05)

Large .0113 .4263 -.0300(7.69) (19.30) (-.9876)


31/33

(Table VI Continued)


Constant LDE/A (LDE/A)M DEF/A (DEF/A)Q

Small .0134 .5381 -.1134(2.60) (27.05) (- 6.38)

Large .0123 .4250 .0309(8.71) (23.48) (1.83)

Small .0205 .7286 -.6126(12.93) (44.84) (-30.18)

Large .0089 .3015 -.0392(6.963) (16.11) (1.54)


32/33

Table VII

Comparison of the Target Adjustment and Pecking Order Models betweenDividend-Paying and Non-Dividend-Paying Firms

The sample consists of 1236 firms with relevant Compustat data for 1981 to 2000, 18,236total observations. The target debt level is based on the historical mean of debt to total asset

ratio and all the explanatory variables are scaled by the total assets ( A ). TDE(LDE)represents the deviation of the total (long-term) debt level from its target and DEF thefinancial deficit. Mis a dummy variable equal to one if TDE(LDE) is positive, and zerootherwise. Qis equal to one if DEFis positive and zero otherwise. Numbers in parenthesesare t-values.

Panel A. The Dependent Variable is Changes in Total Debt Scaled by Book Value ofTotal Assets (A).

Constant TDE/A (TDE/A)M DEF/A (DEF/A)Q

Dividend .0162 .6878 -.1683(10.44) (34.86) (-5.10)

No Dividend .0122 .6025 -.0626(4.19) (20.46) (-1.25)

Dividend .0185 .4913 -.2172(12.86) (30.52) (- 9.81)

No Dividend .0206 .5451 -.3945( 7.80) (39.44) (-13.49)


33/33

(Table VII Continued)


Constant LDE/A (LDE/A)M DEF/A (DEF/A)Q

Dividend .0071 .6019 -.0582(5.18) (32.39) (- 1.88)

No Dividend .0051 .5432 .0263(2.07) (21.59) ( .577)

Dividend .0160 .4014 -.1735(12.33) (41.35) (- 8.71)

No Dividend .0183 .4891 -.3738(7.77) (25.14) (-14.09)

capital read

Documents