Professional Documents
Culture Documents
DECEMBER I977
I
NWA net working assets
(=inventories, net Tc
trade credit and other SDBT short-term temporary
assets) debt capital
NK net capital stock
(=gross capital
stock minus accumulatfd
depreciation)
LDBT
E
long-term
debt
CSuitY
(3
I
cumulative gross
stock issues minus
PCB
permanent
capital
@oak
value)
cumulative stock
retirements plus
cumulative retained
earnings)
A net assets A
Assistant Professor of Finance, Northwestern University. While retaining full responsibility for the
final product, the author is indebted to Frank Brechling, Stuart Greenbaum, Jerry Hausman, Franc0
Modigliani,Stewart Myers and Eugene Savin for many helpful suggestions.
1. This balance sheet is meant to illustrate the basic nature of the financing decision. A more
complete balance sheet is included with the discussion of the estimates.
1467
1468 The Journal of Finance
All the financial items have been placed on the right side of the balance sheet
while real assets (net of certain liabilities such as trade credit, which will be taken
as exogeneous) appear on the left. This scheme separates firms’ real uses of funds
from their possible financing sources. Liquid assets are placed on the right side,
preceded by a negative sign, because the drawing down of liquid assets is consid-
ered a source of funds. Total net assets, A, are then understood to be net of liquid
assets. Changes in the balance sheet items in any period are constrained by the
sources and uses identity:
AA=ANWA+ANK=ASDBT+ALDBT+AE-ALIQ.
Each period’s change in equity can be decomposed into gross stock issues, stock
retirements or retention of earnings:
A A - RE=ASDBT+ALDBT+AGSTK-SRET-ALIQ. (3)
That is, to the extent that firms’ expenditures on plant and equipment and working
assets exceed their cash flow, they incur an external financing deficit which must be
made up by changes in the right-hand side items.
Several empirical studies to date have worked from a similar sources and uses
identity to explain changes in various balance sheet items? All but Spies’ study
take the size of the external financing deficit as exogenous and seek to explain the
composition of financing. Both Jaffee and White aggregate the financing sources
into long-term and short-term categories, while Bosworth’s studies have separate
equations for bond issues, stock issues, changes in short-term debt and changes in
liquid assets. Spies takes only cash flow as given and formulates additional
equations for dividends and changes in physical assets. Each of the studies assumes
that there are balance sheet targets toward which firms slowly adjust, and each
author estimates his stock-adjustment model by ordinary least squares.’
Although each of these studies recognizes that the sources and uses identity
imposes constraints on the coefficients of the equations, none allows for the
possibility that balance sheet interrelationships may enter through the error terms
as well. But it seems likely that shocks which affect one of the equations could
influence the other equations, resulting in correlations among the error terms.
Zellner4 has suggested a technique which exploits these cross correlations to
achieve more efficient estimates, and this technique will be used in the present
study.
A second difficulty with previous studies is that they do not make systematic use
of the theory of optimal capital structure. While it is true that a unified theory of
the financial side of the balance sheet has not yet been developed, the work of
Modigliani and Miller' has spawned a large literature on the optimal debt-equity
ratio: Points of controversy persist, but there seems to be some consensus that a
finite optimal debt-equity ratio exists, determined by such considerations as
corporate taxes, the costs of financial distress and rationing by lenders.' Since all of
the Modigliani-Miller analysis deals with the market value of securities, this
optimal ratio refers to the market value of debt relative to the market value of
equity, It is not a book value debt-equity ratio. If a corporation were initially
satisfied with its capital structure, for example, and then the market value of its
equity rose relative to the market value of its debt (say, because investors were
predicting enhanced prospects for the size and safety of its earnings stream), this
would be taken as a signal from the market that it could afford to issue more debt.
Development of the Model
Equation (3) is the basic identity upon which the model is built. AA - RE will be
taken as exogenous*, so the problem for firms in each period is to decide what
changes they will make in each of the five right-hand side items, given the size of
their external financing deficit. Although market values will be a determinant of
firms' actions, it is book values that they control. If more financing is required,
firms can increase the book value of their permanent capital, PCB (long-term debt
plus equity), or they can increase their temporary capital, TC (short-term debt
minus liquid assets). Changes in permanent capital will be represented by the
stock-adjustment equation:
5. See [a].
6. See [13] for a detailed survey of this literature.
7. See [l 11 and [9] for example.
8. The equations in this model are treated as one block in a larger block recursive system that would
explain the full financing and investment decision. The cost of capital will affect the firm's investment
(hence AA - RE) and also its desired level of financing. But AA - RE is determined independently of
the composition of current financing flows. This assumption would be unjustified if there were
widespread credit rationing or if corporate balance sheets were so far out of equilibrium as to cause a
curtailment of investment. Dhrymes and Kurtz [4], however, found no empirical evidence of such
feedback effects. This issue is discussed in more detail in Tagart, op. ciL, Chapter IV.
9. This study follows previous studies in using the stock-adjustment model. The stock-adjustment
mechanism can be rationalized as an attempt to balance costs of adjustment against the costs of being
out of equilibrium. See Taggart, JOC. cit., for further discussion.
10. See [14] for a discussion of this hedging strategy.
1470 The Journal of Finance
Movements in the target will affect permanent capital flows'in (4), but interest
rate timing considerations, RT, may exert an opposing influence. If firms expect
long-term rates to decline, for example, it may be better to postpone raising
long-term funds and borrow short in the interim."
The current period's retained earnings, RE, are also included in (4) since the
book value of equity rises as earnings are retained, and the level of permanent
capital is augmented.I2
Once adjustments in permanent capital are known, changes in temporary capital
are determined as a residual. Because PCB* and TC* must sum to A , while APCB
and ATC must sum to AA, it follows that:
ATC=6,(TC*- TC-,)+(l -S,)(AA -RE)-62RT. (6)
According to (6), temporary capital adjusts to its target at the same speed as
permanent capital, and also serves a shock-absorbing function. Any part of the
external financing deficit which is not made up by changes in permanent capital
must necessarily be made up by changes in temporary capital. Likewise, timing
considerations must be the opposite of those for permanent capital.
The composition of PCB will be governed by the desired debt-equity ratio, which
in turn depends on the market value of firms' debt, LDM, and the market value of
their equity, STOCK. The desired ratio will be defined by
where b is determined by the tradeoff between tax savings and expected bank-
ruptcy costs. This market value target must then be translated into a target for the
book value of long-term debt, LDBT*. If f is the average contractual interest rate
on long-term debt outstanding, and i is the current new issue rate on long-term
debt, the market value of the debt is approximately
L D M = LDBT~' ~
i '
Equation (8) is strictly true only for consols, but it will be a reasonable approxima-
tion if the bonds are of fairly long term to maturity. From (7) and (8), the target
level of LDBT can then be formulated as
11. The notion that corporate treasurers profit at the expense of lenders because they make superior
forecasts of future interest rates runs counter to the idea of efficient capital markets, but it can be
viewed as a hypothesis to be tested.
12. It is assumed here that dividend policy is determined independently of financing considerations.
This is consistent with the Modigliani-Miller[8] view that the investment and dividend decisions can be
separated from the financing decision. If the firm can always issue new stock to increase its dividend,
then desired dividend policy may affect the amount of financing needed, not uice wrsa. Empirical
support for this separability assumption can be found in E. Fama IS].
A Model of Corporate Financing Decisions 1471
rate timing considerations will affect long-term debt issues, and in addition
short-term movements in stock prices may affect the timing of bond issues. If the
stock market is unusually depressed, bond issues may be temporarily substituted
for stock issues13,so a stock market timing variable, STOCKT, will be included.
The adjustment equation for LDBT will now read
ALDBT=aI(LDBT* - LDBT- I)+ aZ(PCB*- PCB- I - R E )
+ P4RT
SRET=qI(LDBT*-. LDBT-,)+q,(PCB*- P C B - , - RE)+qqRT. (12)
Together, (lo), (11) and (12) allow for the substitution of debt for equity when the
market value of firms' stock goes up, but the extent of this substitution is restrained
by the inflow of retained earnings and the accumulation of permanent assets.
Changes in temporary capital can be decomposed into changes in liquid assets
and changes in short-term debt. The equations for ALZQ and ASDBT will be
similar in appearance to (6), but an additional target will be specified for the stock
of liquid assets. Liquid assets are necessary to conduct transactions, and their
target level should be positively related to the level of sales. Interest rates should
also enter, since these would determine the cost of maintaining any given level of
liquid assets, so the liquid assets target will take the form
LZQ* = c,SALES + c,RA TE (13)
where RATE represents an interest rate variable which will be specified more
precisely below. Since the stock of liquid assets can be changed quickly and
relatively cheaply, even if short-term borrowing must be undertaken to finance
them, it will be hypothesized that there is no lag in adjustment to the liquid assets
13. See, for example, [a]. Again the notion that corporate treasurers possess superior timing ability is
contrary to the efficient markets hypothesis, but we can still look for evidence that treasurers attempt to
time their stock issues.
14. Since STOCK can potentially increase faster than E, the book value of equity, it would appear
that LDBTL could ex& PCB* so that E* <O. But if a firm actually triad to approach such a solution,
the increasing debt would feed back on the value of equity, thus lowering future debt targets. As
expected financial distress costs began to outweigh tax savings, the market would signal to the firm to
stop issuing debt.
15. STOCKT has been omitted from the SRET equation, because the literature suggests that this
conaiderationprimarily influences security issues.
1472 The Journal of Finance
The Data
The equations estimated are (lo), (ll), (12), (14) and (15). The basic data were
taken from the Federal Reserve Flow of Funds data for nonfinancial corpora-
tions16,but several adjustments were made.” Quarterly flow data were available for
all balance sheet items, but stock data were not available for real assets (inventories
and fixed capital) or equity. Real asset and equity stocks were obtained by
cumulating the flows, using the IRS Statistics of Income as a base. Flow of Funds
figures begin in 19521 so the 1951 IRS data for nonfinancial corporations were
used. There were small discrepancies between the IRS data and comparable Flow
of Funds stocks (available for financial assets and liabilities) so the FOF stock
figures were accepted and the IRS real asset and equity figures were adjusted
upward to keep balance sheet proportions the same. Stock retirement figures were
also needed, and since these were available in the SEC Statistical Bulletin only
from 195711 onward, five years of data were available to wash out the effect of the
initial values from the IRS data.
The balance sheet stocks were then grouped as follows:
accts. receivable
-accts. payable
I -cash
- govt. securities
-tax accruals - LIQ -security r.p.’s
consumer credit -open market paper
N WA misc. assets
statistical disc. bank loans
-US.govt. loans SDBT finance co. loans
inventories open market paper
equity=cumulative stock
E issues (common and preFerred)
+ retained earnings
16. Although previous studies cited all use aggregate data, this practice might be questioned on the
ground that the model is constructed on the basis of individual firm decisions. Data on single firms do
not lend themselves to regression analysis, however, since stock and bond issue series would be discrete.
17. Readers wishing further details on these adjustments should consult [13] or write to the author for
a data appendix.
A Model of Corporate Financing Decisions 1473
Several liability items appear with a negative sign under NWA. Firms were thought
to exert little control over these items, so they were netted against assets and
treated as exogenous.
The remaining series to be constructed are market values of long-term debt and
equity, the stock market timing variable, STOCKT, and the interest rate timing
variable, RT. The market value of long-term debt was constructed using the
approximation given in equation (8). The new issue rate is for AAA utility bonds.
The average interest rate is developed by going back to 1947, a time of relative
interest rate stability when i and i might be considered equal; is a weighted
average of the previous quarter’s i and the new issue rate, where the weights are
determined by the proportion of new bond issues and retirements during the
quarter. New issue and retirement data are available from the SEC Statistical
Bulletin.
The market value of equity is approximated by taking dividend payments and
capitalizing them with a dividend yield series for Standard and Poor’s 500 stocks.
This approximation will only be as good as the assumption that the average
dividend yield for all nonfinancial corporations is the same as for the SBCP 500, and
it may be that the S8zP 500 is too heavily weighted toward low-dividend growth
stocks, in which case the market value of equity is overstated. The problem may
not be serious, though, as long as the true value and the estimated value are highly
correlated. The Flow of Funds figures contain annual estimates of the market value
of all corporate stock (nonfinancial and financial), and this series moves quite
closely with the estimated series for nonfinancial corporations constructed from
dividend yields. The series used in the long-term debt target is an average of the
past twelve quarters’ market value of equity, reflecting the idea that the target is an
element of firms’ long-run financing strategy which will not change with every
fluctuation in stock prices. The market value of equity was also used in construc-
ting the stock market timing variable. This consists of the average market value of
equity over the last two quarters divided by the average market value over the last
twelve quarters. Low values of this variable would be an indication that stock
prices are unusually depressed, which in turn would stimulate a substitution of
bond issues for stock issues. Hence the coefficient of STOCKT is expected to be
negative in the bond issue equation and positive in the stock issue equation.
The interest rate variable used in the target for liquid assets is a series on
commercial loan rates from the Federal Reserve Bulletin. A short-term Treasury
Bill rate might also be thought to be relevant, but recall from the balance sheet
above that liquid assets consist of both government securities and cash. A higher
Treasury Bill rate would positively influence the desired holding of government
securities but negatively influence desired cash balances. If these two effects
roughly cancel, then the target for total liquid assets might depend only on sales
and the commercial loan rate, the rate at which firms can borrow to replenish
liquid assets of all types. An eight-quarter moving average of sales, SSALES, was
used for the other liquid assets target variable.
The interest rate timing variable, RT, presented some conceptual difficulties.
Although it is reasonable to suppose that individual corporate treasurers would try
to time their issues to take advantage of expected movements in interest rates, it is
1474 The Journal of Finance
difficult to specify a simple variable that will capture aggregate expectations. The
use of aggregate data also makes it difficult to establish causal connections. White,
for example, argued that when the spread between long and short interest rates is
unusually large, interest rates are expected to move up, so firms will borrow long.
But perhaps it is long borrowing that has caused long-term interest rates to rise
relative to short, rather than vice versa. The hypothesis of corporate speculation on
the term structure of interest rates is probably best tested using disaggregated data,
but in view of widespread insistence that the phenomenon occurs,'* an attempt was
made here to capture its effects. The variable chosen was WDRCP, a weighted
average (with weights .67and .33) of the two most recent quarters' changes in the
commercial paper rate. If interest rates follow a cyclical pattern, but if the extent
and duration of cyclical movements are never known in advance, a few firms might
raise more permanent capital at the first sign of a downturn in rates, in expectation
of a reversal in the near future. The more rates dropped, the larger would be the
number of firms guessing that the trough was at hand. The actual trough and the
beginning of the upturn would bring forth the last flurry of long-term issues. Low
values of WDRCP are expected to stimulate long-term issues, then, while the
higher values occurring when interest rates turn up should lead to postponements
of permanent financing.
Finally, values were imposed for the long-term debt and permanent capital target
coefficients in equations (lo), (1 1) and ( l2).I9 The long-term debt target coefficient
is the optimal ratio of LDBT to STOCK (i/:). The average actual value of this
ratio between 1957 and 1972 was .2865 with a standard deviation of .037.On the
assumption that actual values move toward the target on average, .2865 was taken
to be the target coefficient. For the permanent capital target, NK in equation ( 5 ) is
obtainable from the data, but NWAP must be estimated. On the assumption that
temporary fluctuations in inventories and other working assets will be smoothed
out over a two-year period, a simple average of net working assets over the last
eight quarters was taken to reflect the permanent level of NWA.
The Estimates
Estimated coefficients for equations (lo), (ll), (12), (14) and (15) will now be
presented. Because the model represents the piecing together of several bits of
theory, rather than a single hypothesis, and because its purpose is to find empirical
regularities in corporate financing flows, it was felt that a certain amount of
experimentation was justified. Consequently, a basic set of estimates is presented
first, the results of a specification test are reported next, and finally, estimates of a
modified version of the model are presented. The essential picture that emerges
from all these results is the same. As an additional check, the basic model was
estimated using data from the FTC-SEC Quarter4 Financial Reports for Manufac-
turing Corporations, 1957-1972. These estimates further confirm the results re-
ported here?’
All estimates were camed out using Zellner’s technique” with the balance sheet
constraints imposed. Since ALDBT+AGSTK- SRET+ RE sum to APCB, ( a ,+
& - q J and (a3+/3,) must be constrained to zero. And since ALDBT-ALIQ
=A TC while A PCB + A TC =AA , it is necessary that (a, + /I2 - q,) =(A, - y,) = I -
+
(A, - y3) and (a4+ p4- q4 A, - y4) = 0. All balance sheet items were deflated by
net assets, A, as a heteroscedasticity correction, and seasonal dummy variables
were added to the liquid assets and short-term debt equations.u There were 62
observations reflecting quarterly data from 1957111 to 1972IV.
The first set of estimates appears in Table 1, and they seem sensible in several
respects. The adjustment to the long-term debt target appears to be an important
determinant of long-term debt issues, stock issues and stock retirements, while
adjustment to the permanent capital target exerts a significant influence on firms’
bond and stock issue decisions. As expected, when the debt-equity ratio is below
target, firms issue more debt and less stock, and when permanent capital is below
target, firms issue more of both bonds and stock. Looked at as partial derivatives,
though, these adjustment coefficients are small. A change in the long-term debt
target which is unaccompanied by a change in the permanent capital target elicits
slow responses on the part of the individual components of permanent capital. Any
swifter adjustment must await changes in the overall need for permanent capital, as
well as favorable timing considerations. Adjustment to the temporary capital target
proceeds at similar speeds (comparison of equations (4) and (6) suggests that this
would be so on apriori grounds). When temporary capital is below target, firms
draw down liquid assets and issue more short-term debt. The adjustment
coefficient in the ASDBT equation is not significantly different from zero, however.
A more important function for ALZQ and ASDBT shows up in their shock-
absorbing role, as evidenced by the large coefficients of (AA - RE). Because of the
relatively slow adjustment speeds to the long-term debt and permanent capital
targets, increases in firms’ external financing deficits must be largely met at the
outset by drawing down liquid assets and incurring additional short-term debt. The
determinants of the liquid assets target also influence ALIQ and ASDBT. The level
of sales has the anticipated positive effect on liquid asset holdings, but the
commercial loan rate has an inexplicably positive (and significant) effect. Perhaps
an average of past values, rather than simply the current value of ARCL, should be
used to reflect expectations. Finally, timing considerations appear to exert a
significant influence on corporate financing decisions. The coefficients of STOCKT
suggest that bonds are substituted for equity issues when the stock market is
depressed, while the coefficients of WDRCP indicate that permanent financing is
postponed when long-term rates are expected to fall, and liquid assets are drawn
down in the meantime. If short-term financing is substituted for long-term at such
times, WDRCP should have a positive coefficient in the ASDBT equation. Its
estimated negative coefficient is not significant, however.
In addition to looking at the coefficients,it is useful to examine the correlation
matrix of residuals for the four equations estimated directly. If there were no
disturbance relations between the equations, this matrix would be the identity
matrix. To the extent that it is not, an idea is given of the gain in efficiency from
using Zellner’s technique. The matrix is
ALDBTIA AGSTKIA ALIQIA ASDBTIA
AL D B T / A 1.oo
AGSTK/A .07 I .oo
ALZQ/A .56 .30 1.oo
ASDBT/A - .52 - .17 .24 1.oo
The largest residual covariances are between the ALDBT equation and the equa-
tions for ALZQ and ASDBT. This suggests that the extent to which the proceeds
from bond issues are used to replace temporary capital (build up liquid assets,
retire short-term debt) has not been fully captured by the model. The same is true
of the covariance between ALIQ and ASDBT. As sources of temporary capital,
these items should move in opposite directions. Their positive residual covariance
suggest that firms’ short-term borrowing in order to build up liquid assets is not
completely reflected in the liquid asset target. It was thought that these covariances
might be partially attributable to incomplete explanation of the seasonal patterns.
Examination of the residuals, in fact, uncovered a particularly large first-quarter
seasonal in the bond issue equation.
Since the market value debt-equity ratio in the long-term debt target is a
distinguishing feature of the model, a test of this specification was thought to be
worthwhile. Recall that L D B P - b S T O C K (i/i),where a value of .2865 was
imposed for 6. If the coefficients of tho difference between target and actual
long-term debt are now freed, the bond issue equation will read
hLDBT bSTOCK ( i / i ) LDBT-,
-=
A a1 +as A +... (16)
A
where a,= - as if the hypothesized specification holds. To assess the explanatory
power of the target variable in this form, it is not necessary to know the value of b.
ALDBTIA can simply be regressed on S T O C K ( i / $ / A , L D B T - , / A and the
other variables entering the equation. Similarly, alternative target variables could
be tested without knowledge of the target coefficients. A reading of the textbooks,
for example, might suggest that a book value debt-equity would be more relevant.
If LDBT* is specified as a fraction, c, of the book value of equity, and if an
TABLE 1
ESTIMATES OF &MIIONS (lo),(1 I), (12), (14)AND (15)
ALDBT
A
AGSTK
A A
ASDBT
A
-
SRET
A
-
( L D B P LDBT- 1)
.013 - .0066
A
(4.21) (2.62)
A
- .w7 .0m5
(6.85) (a351
(M - R E )
A
- .738 212
(1 5.95) (4.72)
Q4 .ooso .0050
(10.29) (10.29)
Q3 .ooo9 .0009
(1.88) (138)
Q2 .0042 .0042
(8.28) (8.28)
1477
1478 The Journal of Finance
average book equity figure over the last twelve quarters, SEBV, is used to make it
comparable to STOCK, the bond issue equation would read
--
ALDBT - cSEBV + ‘xg LDBT- I + . . .
A a1 7 A
Equations (16) and (17), and analogous equations for gross stock issues, were
first estimated by ordinary least squares. The results appear in Table 2.23Simply on
the basis of RZ,the standard error of the regression or the t-statistic of the target
variable, the market value formulation seems superior. The coefficients of both
target variables have the wrong sign in the bond issue equation, however, and while
they have correct signs in the gross stock issue equations, the implied target ratio of
the market value of debt to the market value of equity appears unreasonably
hi&.24 These results provide further evidence of the advantage of Zellner’s tech-
nique over ordinary least squares.
The same test was then conducted using Zellner’s technique, and the results are
shown in Table 3.25 When Zellner’s technique is used, it can be argued that the
equations should be judged as a group, and when the sums of squared residuals are
added across all four equations, they come to .001030 for the market value
formulation and .001038 for the book value formulation. If the equations are
examined singly, though, the results are ambiguous. The book value formulation
provides a better fit in the bond issue equation, but a worse fit in the stock issue
equation. This is puzzling, since the coefficient of SEBVIA is insignificant and has
the wrong sign in the bond equation, while the coefficient of S T O C K ( i / i ) / A has a
t-statistic of 3.07. In an ordinary least squares regression, it would be impossible for
a variable which is insignificant to have more explanatory power than one which is
highly significant. It might be supposed, then, that the two target variables have
different effects on the cross correlations between equations.
For the market value formulation, the correlation matrix of residuals is
23. Seasonal dummies have now been added to the bond and stock issue equations.
24. This ratio is obtained by dividing the coefficient of SSTOCK(i/l)/A by the coefficient of
GSTK-,/A-cf. equation (24)-which yields a value of 1.0.
25. The A L f Q / A and ASDBTIA equations were included in each Zelher estimation run, but their
coefficients were not sensitive to the alternative long-term debt targets, so the results are not reported.
A Model of Corporate Financing Decisions 1479
TABLE 2
TEST
OLS BOND AND STOCK ISSUE EQUATIONS-SPECIFICATION
Bond Issue Equations Stock Issue Equations
Market Value Book Value Market Value Book Value
Formulation Formulation Formulation Formulation
STOCK( i / T )
A
- .w55 - .Oo40
(3.56) (4.58)
A
- .0150 - .0044
(1.21) (.59)
LDBT- I
A
- .0625 - ,0432 .oO40 .021
(3.90) (2.59) (.a) (2.10)
Comparison of the matrices shows the market value formulation to have slightly
larger correlations between the residuals of .the long-term debt equation and those
of the liquid assets and short-term debt equations. Since it is these last two
equations which have the largest residual variances, it may be that the influence of
them renders the bond issue equation less precise in the Zellner estimates. On the
other hand, in the book va1u.e formulation, the residuals from the gross stock issue
1480 The Journal of Finance
TABLE 3
STOCK (i / T )
A
.0032 - .0010
(3.07) (1.21)
SEBV -. m 3 - .0015
A
c57) (.26)
LDBT- I
-.0156 - .033 .018 .023
A (3.12) (2.31)
(1.57) (2.91)
(PCB. - PCB- I - R E )
.021 .031 .022 .018
A
(2.32) (2.73) (3.47) (2.13)
STOCKT - 1.04
A
- .889 .889 1.04
(2.47) (3.06) (2.47) (3.06)
equation have a larger correlation with those in the liquid assets equation than they
do in the market value formulation. Accordingly, it is the market value formulation
which fits better for the gross stock issue equation.
This suggests that in the market value formulation, some influence has been
omitted from the bond issue equation which would also affect liquid assets
positively and short-term debt negatively. Possibly the smoothed book value of
A Model of Corporate Financing Decisions 1481
equity acts as a proxy for this influence (proxy, because SEBV does not seem to
have a significantinfluence of its own on bond issues) and thus serves to reduce the
residual covariance between the three equations.
At the same time, it was noticed that the seasonal dummies inserted in the bond
and stock issue equations, though significant, were still not catching the full effect
of the first quarter seasonal in bond issues. The seasonal is probably variable,
rather than constant as the use of dummy variables assumes, and it was thought
that some additional explanatory variable might be found to account for it.
Examination of the data on asset flows revealed that net investment expenditures
on plant and equipment have a similarly negative seasonal in the first quarter. This
raised the possibility that asset flows might have some explanatory power as timing
variables in addition to the stocks. On the basis of target stocks of long-term debt
and permanent capital in a particular quarter, firms may feel that they need more
long-term debt. But if they know at the same time that investment outlays are small
in this quarter and will be much larger in the next quarter, they may wish to
postpone their bond issues by one period in order to synchronize payments and
receipts better. Timing the issues to coincide with actual outflows can help to
reduce the transaction costs of switching funds in and out of liquid assets until they
are needed. Firms’ targets will tell them what the ideal balance sheet composition
is, but changes in the targets will not precisely mirror expenditures. The pattern of
expenditures could then influence the timing of security issues, reflecting firms’
attempts to coordinate their inflows and outflows.
When net investment divided by A ( N I / A ) was tried in the long-term debt,
liquid asset and short-term debt equations, its inclusion added substantially to the
fit of the long-term debt equation, and the first dummy became insignificant,
confirming the notion that investment outlays are responsible for the first quarter
seasonal in bond issues. In addition, the residual correlation between the bond
issue equation and the liquid assets and short-term debt equations was reduced.
The correlation matrix of residuals suggested that other variables may also have
been omitted from the original specification. There was still a substantial correla-
tion between the residuals of the liquid assets and short-term debt equations, for
example, which might be related to the strong second and fourth quarter seasonals
in these two equations, Tax accruals appeared to be a possible determinant, since
firms might borrow short and build up liquid assets prior to a tax payment.
Insertion of tax accruals divided by A (TAXA C / A ) added some explanatory power
to the liquid assets and short-term debt equations, but it still did not reduce the
importance of the seasonals.
A final set of estimates, incorporating these changes, is shown in Table 4. N I / A
and TAXACIA are included as timing variables in the ALDBT, ALIQ and
ASDBT equations. In addition, an eight-quarter moving average of the commercial
loan rate is used in the liquid assets target, in place of the current quarter’s rate.
And finally, an autocorrelation parameter is inserted in each of the equations.=
All of the new timing coefficients are significant, and the fit of the bond issue
equation is now up to the level of the book value formulation in Table 3. Addition
26. Because of the balance sheet constraint, the autocorrelationparameter must be constrained to be
the m e in all q ~ a t i oSee
~ . Berndt and Savin [I].
1482 The Journal of Finance
TABLE 4
WITH AUTOCORRBUTION
ESTIMATES CORRECTION TIMINO
AND ADDITIONAL VARIABLES
A LDBT
A
AGSTK eL/e
A
ASDBT
A
SRET
A
-
A
( L D B P - LDBT-1)
A
.010 - .0072 .003 1
(6.20) (5.62) (2.63)
(PCB* - P C B - , - R E )
.015 .027 .00032
A
(3.16) (7.18) (0.24)
(TC*-TC-,)
- .056 -.014
A
(4.47) (1.11)
(AA - R E )
A
- .563 .395
(14.70) (10.29)
STOCKT
A
- .481 .48 1
(- 2.44) (2.44
ASSA LES
,034 .035
A
(3.25) (3.25)
of the timing detail does not seem to blur the other effects. The adjustment
coefficients are still significant, with the exception of that for the temporary capital
target in the short-term debt equation. The interest rate timing variable still exerts a
significant influence, but the stock market timing variable has now become less
significant. Possibly its true effect is masked by the appearance of the other
variables, but it is also possible to argue that this term was previously picking up
some missing influence which has now been accounted for. The effect of the
commercial loan rate now appears with the correct sign in the smoothed form. The
autocorrelation parameter is significant, but it is comforting that its presence does
not drastically alter the remaining coefficients.
The correlation matrix of residuals for this set of estimates is:
ALDBTIA AGSTKIA A L I Q I A A S D B T / A S R E T I A
ALDBTIA 1.oo
AGSTKIA -.O9 1.oo
ALIQ/A .34 .3 1 I .oo
ASDBTIA -.49 .oo5 .52 1-00
A Model of Corporate Financing Decisions 1483
TABLE 4
ESTIMATES WlllI AUTOCORRELATION
CORRECTION AND ADDITIONAL TIMING v U l & S
--
ALDBT
A
AGSTK
A
WQ
A
ASDBT
A
SRET
A
-
NI
A
.425 .285 -.la
(14.73) (5.00) (2.42)
-
TAXAC
A
- M1 .237 298
(2.4) (4.45) (5.51)
Q4 .0053 .OO53
(21.20) (21.20)
Implied R 2
(Undeflatcd
Variables) .87 .84 .87 .70 .22
The decrease in correlation between the bond issue equation's residuals and those
of the liquid assets and short-term debt equations is readily apparent. The positive
residual correlation between liquid assets and short-term debt, on the other hand,
has increased. If a variable could be found to better account for the seasonal in
short-term debt and liquid assets, this residual correlation might be reduced.
It should be noted that the influence of the market value target for long-term
debt remains strong in the face of these modifications. While the specification tests
do not favor the market value formuation as resoundingly as one might like to see,
one would certainly not conclude that it is inferior to the alternative specification.
And since the book value specification may have been related to the added timing
factors (the correlation of SE B V /A with N I / A is .33) it is hard to believe that its
contribution to the bond issue equation was related to the influence of the book
value debt-equity ratio as a target.
Conclusion
The present study contributes to the understanding of corporate financing
patterns by including a market value debt-equity ratio as a determinant of
1484 The Journal of Finance