Full text of Economic Review (Federal Reserve Bank of San Francisco) : Fall 1987, Number 4

View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Opinions expressed in the Economic Review do not necessarily reflect the views of the
management of the Federal Reserve Bank of San Francisco, or of the Board of Governors of
the Federal Reserve System.

The Federal Reserve Bank of San Francisco's Economic Review is published quarterly by the Bank's
Research and Public Information Department under the supervision ofJack H. Beebe, Senior Vice President
and Director of Research. The publication is edited by Gregory J. Tong, with the assistance of Karen Rusk
(editorial) and William Rosenthal (graphics).
For free copies of this and other Federal Reserve publications, write or phone the Public Information
Department, Federal Reserve Bank of San Francisco, P.O. Box 7702, San Francisco, California 94120.
Phone (415) 974-2163.

Three Questions Concerning Nominal and Real Interest R a te s ................. 5
Carl E. Walsh

II.

Off-Balance Sheet Banking .............................................................................21
Christopher James

III.

Tax Policy and Corporate Capital S tructure..................................................37
Randall Johnston Pozdena

Editorial Committee:
John P. Judd, Brian Motley, Michael Keeley, Frederick Furlong, Bharat Trehan, and Reuven Glick

Carl E. Walsh*

Do increases in real interest rates tend to be followed by declines? Are
market interest rate movements a reflection of variations in expected
inflation or expected real rates o.freturn? Are monetary andfiscal policy
responsible for the behavior of real rates in the 1980s? These three
questions, and their implications for monetary policy, are addressed in
this paper.

The 1980s have witnessed unusually high levels
of market interest rates relative to measured rates of
inflation. The importance of this phenomenon
depends critically on the extent to which these high
ex post real rates have reflected high expected, or ex
ante, real rates of interest, since it is the expected
real return that should affect the savings, investment, and portfolio choices of the public. Because
these aspects of economic behavior are related to
expected real rates of interest, it is important to gain
a fuller understanding of the relationship between
market interest rates, expected real rates, and macroeconomic policies.
This paper examines three empirical questions
related to the behavior of nominal and expected real
interest rates:
Does the real rate ofinterest have a random walk
component? If the real rate does not have such a
component, it would tend to revert to a constant
average value after any changes, that is, rate
changes would be temporary in nature. Deviations

of the real rate from its average therefore may
provide information about the business cycle that
would be useful for the conduct of monetary policy.
One could interpret the average value of the real rate
as its "equilibrium" value, and indeed several analysts have suggested that monetary policy act to
stabilize the real rate around its equilibrium value. 1
If, however, the real rate does not tend to revert to
any constant level, then there is no sense in which
the real rate has a constant, long-run equilibrium
value around which it might be stabilized.
To what extent are unpredicted movements in
market interest rates due to movements in real rates
as opposed to movements in expected inflation?
Using U.S. data from the 1950s and 1960s, Fama
(1975) concluded that nominal interest rate movements were consistent with a constant expected real
rate of interest and that all market interest rate
changes were attributable to changes in expected
inflation. More recent U. S. experience suggests that
the expected real rate has moved quite sharply, so
that nominal rate changes may reflect a more equal
balance of movements in the real rate and expected
inflation.
Federal Reserve monetary policy has often been
characterized as designed to smooth market interest

* Associate Professor, University of California at
Santa Cruz, and Visiting Scholar, Federal Reserve
Bank of San Francisco. This article was written
while the author was a Senior Economist at the
Federal Reserve Bank of San Francisco.
5

The paper is organized to discuss each question in
order. The next section examines the stochastic
processes followed by nominal interest rates and
inflation to test for whether changes in these variables tend to persist or to be temporary. The results
have implications for the existence of a constant
average ex ante real rate around which the real rate
fluctuates. They also are useful in determining the
specification of the variables to use in the later
empirical analysis.
Section II uses the results from a Vector Autoregression (VAR) to decompose innovations in the
nominal rate into ex ante real rate innovations and
expected inflation revisions. By comparing such
decompositions over different sample periods, one
can obtain a sense for the changing informational
content of nominal rate innovations. Section 1lI
presents the decomposition of the real rate into
components attributable to monetary and fiscal disturbances, respectively. Conclusions are summarized in Section IV.

rates, and the appropriateness of such a policy
depends, in part, on whether movements in market
rates tend to be generated by changes in real rates or
by inflationary expectations. An expansion in the
money supply designed to offset a rise in market
rates that originates from expectations of higher
inflation may simply fuel an actual increase in
inflation.
To what extent are monetary and fiscal policy
disturbances re~ponsible for the movements of the
real rate, particularly over the last ten years? The
initial apparent rise in real rates in the early 1980s
has generally been attributed to restrictive monetary
policy actions designed to reduce inflation; the
continued high level of real rates is often blamed on
large federal budget deficits. However, manyeconomists argue that deficits have little impact on interest
rates. Decomposing the real rate into components
due to monetary policy shocks and fiscal policy
shocks may shed light on this debate.

Does the Real Rate Have a Random Walk Component?
1987). Factors, such as changes in tax policy, that
produce persistent shifts in the real rate may call for
a different policy response than factors, such as
fluctuations in the demand for money, that produce
temporary changes in the real rate.
To be more specific, a rise in the demand for
money, in the absence of a policy response, would
temporarily raise the real rate and contract aggregate demand. Policy might respond by expanding
the money supply to keep the real rate from rising.
But if the initial rise in the real rate were due to a
permanent shift in consumer preferences towards
current consumption and away from saving, then no
such monetary policy action would be called for. In
other cases, it may be less important to respond to
temporary movements in the real rate, since the
costs of failing to act would presumably be smaller
than a failure to respond to more persistent disturbances. At the time the change in the real rate is
observed, however, it may be difficult to determine
whether permanent or temporary factors are at work.

Until relatively recently, economists generally
assumed that most macroeconomic variables tended
to fluctuate randomly around either a constant average value or around a trend line. When a variable
rose above its trend, it was expected subsequently to
fall back towards the trend line.
In the last few years, this standard view has been
questioned. For example, Nelson and Plosser
(1982) argue that most macroeconomic variables
are better characterized as having a random walk
component to their behavior. 2 A random walk has
the property that changes are permanent, that is, if
the variable goes up, there is no tendency for it to
return to any average or trend value. Thus, shocks to
a variable containing a random walk component
will have permanent effects on the level of the
variable.
A finding that the ex ante real rate has a random
walk component would have important implications
for suggestions that real rates be used to guide the
conduct of monetary policy (Jenkins and Walsh

A common test for a random walk is based on the
least squares regression of the first difference of a
variable on its lagged level and lagged first differences. A constant and a time trend also may be
included. The test statistic is simply the standard
t-statistic for the coefficient on the lagged level.
Under the null hypothesis that the variable has a
random walk component, the coefficient on the
lagged level should equal zero. 3 A large, negative
t-statistic would indicate rejection of the null in
favor of the hypothesis that the variable is stationary
(perhaps with a trend).
Table 1 presents the results of the test described
above. Quarterly data were used, and the nominal
interest rate is the daily average of secondary market
yield on 3-month Treasury bills for the first month of
the quarter. Two price indices were used to calculate
'IT: the GNP Price Deflator and the Consumer Price
Index. Results for various sample periods are
reported.
The test statistics consistently fail to reject the
presence of a random walk term in the real rate. (In
no case can the null hypothesis be rejected at the 5
percent level.) Rose (1987) reports similar findings
for annual, quarterly, and monthly data for the U.S.
He also finds evidence of a random walk component
in the real rate for 17 other countries.

Testing for a Random Walk
Testing for random walk behavior in the ex ante
real rate of interest is complex because the expected
real rate cannot be observed. Nevertheless, it is
possible to draw some conclusions about the real
rate process by examining the behavior of the ex
post real rate and its two components, the nominal
interest rate and the rate of inflation.
To define some notation, let it denote the nominal
interest rate from t to t + 1. Let TIt + I be the rate of
inflation from t to t + 1, and let Etx t +j be the
expectation, formed at time t, of a variable xt+j'
Then, ignoring taxes, the ex ante real rate, rt, and the
ex post realized real rate, exrt, are given by equations 1 and 2:
(1)

(2)
Equations 1 and 2 imply that
(3)

so that the ex ante real rate and the ex post real rate
differ by the error made in forecasting future inflation.
Chart 1 plots the nominal interest rate on 3-month
Treasury bills and their ex post real return. The
apparent upward drift in the nominal rate from 1960
to 1981 was primarily a reflection of rising expectations of inflation; the ex post real return, far from
mirroring this upward trend, remained negative for
most of the 1970s. The sharp rise in the ex post real
rate from 1978 to 1982 was interrupted only during
early 1980 by the Federal Reserve's imposition of
credit controls.
The realized real rate differs from the expected
real rate by the error made in forecasting the rate of
inflation. Under any reasonable model of expectations formation, this inflation forecast error should
be transitory, or stationary, in nature (that is, have no
random walk component). Since the sum of a variable with a random walk component and a variable
without one will contain a random walk component,
equation 3 shows that r is stationary if and only if exr
is stationary. If exr contains a random walk component, then so must r.

Chart 1
Nominal and Realized
Real Rate on 3-Month
Treasury Bills
Percent

12
8
4

o
-4
-8 -j,..,.................................................,....,..............,....,....................................,
60

84 86

If the after-tax nominal interest rate and the rate of
inflation contain the same random walk component,
then when one is subtracted from the other to obtain
the after-tax real rate, the random walk components
will cancel, leaving an after-tax real rate with no
random walk element. If both i and 1T have random
walk components, then exr can also have a random
walk component, as indicated by Table 1, even ifthe
after-tax real rate does not.
When two variables contain random walk components but some combination of the two does not, the
variables are said to be cointegrated (see Engle and
Granger, 1987 and 1986, and Hendry, 1986). The
nominal interest rate and the rate of inflation will be
cointegrated if they contain the same random walk

The Effect of Taxes
One possible explanation for these results is that
exr is the wrong way to combine the nominal rate
and realized inflation to obtain a measure of the real
rate. Many economists would argue that the relevant
real interest rate should be an after-tax real rate.
Letting 1" denote the marginal tax rate, the ex post
after-tax real rate is (1 - 1") it - 1T t + I' When both i
and 11" contain independent random walk elements,
the two variables will tend to drift apart over time
since there are no forces acting to keep them close
together. But if the after-tax real rate tends to fluctuate around a constant value, then i and 1T cannot drift
too far apart. This implies that the random walk
elements in i and 1T must be related.

and temporary shocks to the real rate. A finding that
the random walk component accounts for almost all
the movement in the ex ante real rate would suggest
monetary disturbances have not been important.
Such evidence would support proponents of real
business cycle theories, which de-emphasize the
importance of money. 4
Cochrane (1986) has recently proposed a method
of measuring the relative importance of the random
walk component of an economic time series. 5
Applied to the ex post real rate for the period 1961 QI
to 1985QIV, Cochrane's measure of persistence
approaches approximately .12, implying that
roughly 12 percent of the total unpredicted change,
or innovation, to the ex post real rate represents a
permanent innovation associated with the random
walk component. Cochrane's measure suggests that
innovations to the ex post real rate are predominantly temporary in nature. Since monetary disturbances have only temporary effects on the real rate,
this finding is consistent with the view that monetary disturbances are an important source of real rate
movements.
The evidence provided by Cochrane's measure of
persistence must be qualified, however, by noting
that it has a downward bias when used to measure
the importance of the random walk component in
the ex ante real rate. 6 Thus, the appropriate interpretation is that at least 12 percent of real rate
shocks have permanent effects.

component. In this case, there will exist a constant
a, called the cointegrating parameter, such that
ai - 1T is stationary. If the after-tax expected real
rate has no random walk component, then a will just
be equal to one minus the marginal tax rate. Tests for
cointegration and estimates of a are reported in Part
Aof the Appendix.
The results from the cointegration tests are
mixed. Evidence of cointegration is found for the
1961QI - 1979QIII period, but cointegration is
rejected when the sample is extended through
1985QIII. In addition, if the after-tax rate has no
random walk component, the cointegrating parameter should equal one minus the marginal tax rate,
that is, the estimated value of a should be around
0.6 to 0.7. Unfortunately, the actual estimates generally fail to fall in this range. Hence, the evidence
seems to suggest that both the ex ante real rate and
the after-tax rate contain random walk components.
If this finding were to hold for other real rates,
particularly for longer term real interest rates, it
would have important implications. For example,
most modern macroeconomic theories imply that
monetary forces have only temporary effects on real
rates of interest. The presence of apparently permanent shifts in the real rate must then be due to
nonmonetary phenomena.
However, the evidence of a random walk component in the real rate still leaves unanswered the
question of the relative importance of permanent

II.

Movements in Real Rates or Expected Inflation?
some economists have blamed the Federal Reserve
for high real interest rates as nominal rates have, it is
argued, fallen less than has expected inflation. 7

Central banks have quite frequently relied on
nominal interest rates as both instruments of monetary policy and as informational variables to be used
as guides in the formulation of monetary policy.
However, the use of nominal rates has inherent
limitations because of the difficulty of determining
whether nominal rate movements reflect movements in expected real rates or in expected inflation.
In the 1970s, for example, the Federal Reserve was
criticized for failing to allow nominal interest rates
to rise sufficiently in the face of inflationary pressures. As a result, it was argued, monetary policy
was insufficiently anti-inflationary. More recently,

Decomposition
It is possible to use historical data to decompose
nominal interest rate movements into expected real
rate and expected inflation changes. This allows an
assessment to be made of the relative importance of
these two components during different sample periods. Of particular interest is the decomposition of
the unpredicted changes - or innovations - in the
nominal rate. Such innovations are important as
9

of real GNP, the GNP price deflator, M 1, the relative
price of fuels, and the real value of federal defense
purchases. All variables were entered into the VAR
in first difference form with a lag length of four. 10
The estimation period was 1961 QI to 1984QIY.
Data from 1985 to 1986 were dropped because of
the apparent shift in the relationship between M 1
and other macroeconomic variables that occurred in
1985. Details of the construction of it and Et7r t + I
can be found in Part B of the Appendix.
The estimation results show expected inflation
innovations to have been much more volatile than
nominal rate innovations. For the 1961 QI 1984QIII period, the variance of Et7r t + I was four
times that of it (2.03 versus O.5l). Since the October
1979 change in Fed operating procedures, the variance of it has risen (to 1.05), while that of Et7r t + I
has fallen (to 1.52), putting the expected inflation
innovation variance at less than twice that of it.
The series on it and Et 7r t + I can be used to
construct a series on f t, the innovation to the ex ante
real rate. II The results of this decomposition for
various subperiods are given in Table 2.
For the entire estimation period (1961 QI 1984QIII), a one percent innovation in the nominal
rate reflected, on average, a .56 percent real rate
innovation and a .44 percent expected inflation
innovation. 12 This division, however, is far from

they represent "new information" that may be useful for the conduct of monetary policy.
If E t Iit is the best linear forecast of the nominal
rate it based on information available at t - 1, then
the nominal rate innovation, denoted it, is just the
forecast error:

(4)
Since it = rt + Et'iT t + I and Et - lit = Et Irt +
Et-I'iT t + 1,8 the nominal rate innovation can be
written as the sum of the innovation to the expected
real rate and the revision, or innovation, to expected
inflation:

Given any two of the three innovations - il' f t and
Et7r t + I - the third can be calculated from equation
5. Armed with estimates of the three innovations,
the relative importance of the expected real rate and
expected inflation for nominal interest rate innovations can be gauged. 9
Estimates of both it and Et7r t + I were obtained by
estimating a six-variable VAR system. The variables included in the VAR were quarterly observations on the three-month Treasury bill rate, the logs

constant. During the period prior to the Fed's October 1979 change in operating procedures, nominal
interest rate innovations appear to have predominately reflected expected inflation innovations.
In contrast, nominal rate innovations since the
fourth quarter of 1979 have primarily reflected innovations in the re'al rate. A one percent nominal rate
innovation during the period 1979QIV - 1984QIII
was equal, on average to a .8 percent real rate
innovation and a .2 percent expected inflation innovation.
The decompositions of the nominal rate innovations that are reported in Table 2 are based on a
single VAR estimated over the entire 1961Ql 1984QIV period. This has the effect of implying
individuals knew the behavior of inflation and nominal interest rates during the 1980s when forming
expectations in, say, 1970. Such an implication is
not implausible if the underlying structure generating inflation, interest rates, and the other macroeconomic variables had remained unchanged over
the entire sample period. However, the increased
importance of aggregate supply shocks, such as the
oil price increase and oil embargos in the 1970s, the
shift in monetary policy procedures in 1979, the
rapid decline in inflation in the 1980s, and the
historically unprecedented deficits of the Reagan
Administration suggest that such an assumption of
structural constancy may yield a poor approximation when used to characterize the recent macroeconomic experience of the U.S. Huizinga and
Mishkin (1986), for example, present evidence to
suggest a shift in the structure in the real rate process
in October 1979.
To obtain a rough check on the robustness of the
innovation decompositions, the VAR system was reestimated over two subsamples: 1961 QI - 1979QIII
and 1970QI - 1984QIV While the results differed
somewhat from those obtained using the entire
sample, the basic message was the same. For example, estimates from 1961 Ql - 1979QIII imply that
almost all nominal rate innovations (98 percent in
fact) were the result of expected inflation innovations. This is consistent with Fama's assumption
that for the post-war period prior to 1972, all nominal interest movements were due to changes in
expected inflation (Fama, 1975). When the VAR is

estimated over the 1970QI - 1984QIV period,
expected inflation innovations are estimated to
account for 75 percent of the nominal rate innovations during 1970QI - 1979QIII and only 33 percent
during the 1979QIV - 1984QIII period.
Findings
The changing composition of the innovations to
the nominal rate reflects the changing relative
importance of expected inflation and real rate movements over the last twenty-fi ve years. The late 1960s
and most of the 1970s were periods of high and
variable rates of inflation. Real rates were far from
constant then, and ex post real rates were negative
during the 1970s (see Wilcox, 1983), but the dramatic increases in inflation appear to have dominated nominal rate innovations. The 1980s have
witnessed large movements in both inflation and
real interest rates. In a reversal of the 1970s, a
falling rate of inflation has been associated with very
high ex post real rates. Nominal rates have been
much more volatile, and, according to the VAR
estimates, nominal rate innovations have predominately reflected innovations to the ex ante real
rate of interest.
This evidence indicates that monetary policy cannot reliably respond in a simple way to movements
in market interest rates. For example, increases in
the nominal rate due to upward revisions of
expected inflation would, in general, call for a more
contractionary monetary policy. If nominal rate
changes were always dominated by such expected
inflation changes, a simple automatic policy
response might be possible. But nominal rate
changes are sometimes, as in the 1980s, dominated
by real rate changes.
Real rate changes pose more difficult problems
for monetary policy. If they were due to money
demand shifts, then they should be offset. In contrast, real rate effects due to aggregate spending
fluctuations should generally not be offset. The
changing informational content of movements in
market interest rates means that simple policy rules
based on market rates are unlikely to produce a
satisfactory monetary policy. Additional information is required to interpret the changing nature of
nominal interest rate movements.

The innovation decompositions provide interesting evidence on the information contained in unanticipated movements in nominal rates. Such movements primarily revealed information on expected
inflation in the 1970s and expected real rates in the
1980s, although they provide no explanation of the

underlying causes of either inflation or expected real
rate movements. In the next section, an attempt is
made to assess the role of macroeconomic policy
shocks in explaining the high real interest rate
during the first half of the 1980s.

What Raised Real Rates in the 1980s?

A number of alternative explanations have been
offered to account for the high real interest rates that
the U.S. has experienced during the past eight
years. Two of the most prominent attribute high real
rates to macroeconomic policies. The first views the
rise in the real rate beginning in 1979 (see Chart I)
as a result of a restrictive monetary policy aimed at
reducing the rate of inflation. The second attributes
the continued high level of real rates, particularly
since the 1981-82 recession, to current and expected
future federal budget deficits.
A measure of the contribution of monetary and
fiscal policy actions to the behavior of the ex post
real rate can be obtained from the same VAR system
used in the previous section to decompose nominal
interest rate innovations. The manner in which
movements in the ex post real rate are attributed to
the various disturbances is detailed in Part C of the
Appendix. The observed value of the ex post real
rate is expressed, for each period during the sample,
as the sum of six independent terms, one for each of
the six disturbances in the VAR system. Since the
purpose of this section is to focus on the behavior of
the measured real rate during the 1970s and 1980s,
the sample period over which the VAR was estimated was shortened by dropping the decade of the
1960s and estimating the system over 1970QI 1984QIY.
As described in the previous section, the VAR
system used to decompose nominal rate movements
used real federal defense expenditures as a measure
of fiscal policy. It is more common to use either the
federal deficit or total government purchases of
goods and services as proxies for the impact of fiscal
policy. Results will be reported for each of these
proxies, but each is an imperfect measure. The
deficit, or the deficit corrected for the business cycle

(the· high employment deficit), implicitly imposes
the assumption that expenditures· have the same
impact on real interest rates as do tax revenues. Yet
this is an assumption that macroeconomic theories
imply is wrong.
Simple Keynesian models predict that tax
changes are partially financed out of both consumption and savings so that the impact of taxes on
aggregate demand is less than an equal dollar
change in government purchases of goods and services. Other models predict that only government
expenditures will affect real rates. According to
these models, the impact of current taxes on private
spending would be offset by the effect of the accompanying change in future expected taxes when government expenditures are held constant. 13
Total government purchases of goods and services, however, will not provide a perfect measure
of the impact of fiscal policy even when taxes do not
matter. To the extent that some government programs (health, public transportation, etc.) substitute for private purchases, a rise in government
purchases may produce an offsetting decline in
private spending, leading to little net impact on
aggregate demand. This possibility suggests that a
category of government expenditures for which no
close private substitute exists should be used in
calculating the impact of government expenditures
on real rates. Federal defense expenditures constitute one such category.
In light of these considerations, three fiscal
proxies were used: the real federal budget deficit
(National Income and Product Account basis), real
government purchases of good and services, and
real federal defense purchases. The VAR system
was estimated using each of the three fiscal proxies
in tum. Then, the predicted path of the ex post real

rate was generated under the assumption that either
fiscal or monetary shocks were equal to zero. This
assumption yields an estimate of the contribution of
each type of shock over the sample period. Unfortunately, the estimated contributions of fiscal and
monetary shocks to real rate movements are sensitive to the fiscal proxy used. Table 3 summarizes the
results for the period since 1979QIY.

Chart 2
Impact of Money Supply
Shocks on the Real Rate
Percent

Impacts of Fiscal and Monetary Shocks
The deficit measure (rows 1 and 4 of Table 3)
attributes relatively little of the rise in the real rate
since 1979 to either monetary or fiscal shocks.
There is some indication that monetary shocks have
contributed less to the level of real rates since the
end of the last recession in I982QIV, whereas fiscal
policy has contributed more. This result supports the
view advanced by, among others, Cecchetti (1986).
Somewhat similar results were obtained by using
defense expenditures, although the absolute contribution of both monetary and fiscal shocks in this
case was much larger. Although the impact of monetary shocks falls slightly after the end of 1982, it is
estimated to have added more than fiscal shocks to
the real rate even in the 1983-1984 period. In
marked contrast, the contribution of fiscal shocks is

raised significantly when proxied by total real purchases of goods and services. With that proxy, fiscal
shocks are estimated to have added roughly 450
basis points on average to the ex post real rate
between 1979QIV and 1984QIII.
The time pattern of the impact of fiscal and
monetary shocks implied by the estimates using
either total purchases or defense purchases are fairly
similar. Using the results obtained when defense
purchases proxy for fiscal policy, Chart 2 illustrates
the role played by Ml shocks on the path of the ex

post real rate. This chart plots the ex post real rate 14
and an alternative path, eXrt , in which the estimated
effects ofMl shocks on exrt are removed. Whenever
exrt exceeds exrt , money disturbances are estimated
to have raised the ex post rcal rate. When exrt is less
than eXrt , the net impact of Ml shocks was to lower
the ex post real rate.
The evidence in Chart 2 appears to agree with
otheranalyses of U.S . real interest rates in the early
1980s (for example, Blanchard and Summers,
1984,and Ceeehetti, 1986). If money supply shocks
are interpreted as reflecting the impact of monetary
policy, the estimated decomposition of the ex post
real rate suggests that monetary policy began to
push up the real rate during the fourth quarter of
1979 and continued to contribute to the high level of
exr through 1982. Apparently not until the fourth
quarter of 1982 did the net contribution of monetary
policy fall to zero. During the three-year period
(l979QIV - 1982QIII), monetary policy actions
added an estimated 2.6 percentage points to the real
rate. To place this in perspective, exr, net of the
estimated effects of credit controls, averaged 5.4
percent during this three-year period.
More surprising is the apparent effect of monetary policy in pushing exr above exr during 1983
and 1984. For example, Cecchetti (1986) attributes
high real rates in 1984 to fiscal policy (high
expected future budget deficits) on the basis of
evidence from the term structure of interest rates.
From 1982QIV to 1984QIII, exr averaged 2.88
while exr averaged only 0.97. Monetary policy is

Chart 3
Impact of Fiscal Shocks
on the Real Rate
Percent

Ex Post
Real Rate

6
4
2

°tt~cM~k;;J-----~-2
Real Rate with Fiscal
Shocks set to zero

-4

therefore estimated to have contributed almost 2
percentage points to exr during this period.
The estimated impact of fiscal policy is shown in
Chart 3, where the path of the ex post real rate when
the effects of defense spending shocks have been
removed is shown. Fiscal policy is estimated to have
raised exr throughout the 1979-1984 period. This
rise reflects the increase in real defense expenditures
that began in 1979 under President Carter and that
continued under President Reagan. The average
effect of fiscal policy during 1979QIV - 1982QIII
was to raise exr 1.2 percentage points, roughly half
the impact of monetary policy. Since the end of the
1981-82 recession, fiscal policy, as measured by
defense spending shocks, has added 1.4 percentage
points on average to exr.

Conclusions
The second question concerned the respective
importance of innovations to the expected rate of
inflation and the expected real rate in accounting for
innovations in the 3-month Treasury bill rate. Test
results showed clearly that the division of nominal
rate movements between real rate and expected
inflation rate movements has changed quite
dramatically during the last twenty-five years. In the
1970s, almost all unpredicted nominal rate changes
were associated with variation in the expected rate
of inflation. In contrast, unpredicted changes in the
nominal rate during the first half of the 1980s

This paper has attempted to address three empirical questions related to the behavior of nominal and
real interest rates. The first asked whether a ranCR)m
walk component plays a role in the ex ante real rate
of interest. Test results were consistent with the
hypothesis that the real rate does contain a random
walk component. However, they also indicated that
the permanent effect of an unpredicted change in the
real rate is probably relatively small, although the
results showed that at least 12 percent of an unpredieted change would have a permanent effect on the
real rate.

predominately reflected changes in the expected
real rate of interest.
The third question addressed the role of fiscal and
monetary shocks in explaining the high real rates of
the 1980s. Decomposing the history of the ex post
real rate into the independent contributions of
various shocks provided a means of assessing the
impact of monetary and fiscal effects. Using defense
purchases by the federal government as a proxy for
fiscal policy, the evidence suggests that monetary
policy added just over 2 percentage points to the ex
post real rate between 1979QIV and 1984QIII .

Fiscal policy raised the ex post real rate on average
just over 1percentage point during this same period.
These results, however, were sensitive to the
measure of fiscal policy· employed. When total
federal purchases of goods and services was used,
fiscal policy was estimated to have added 450 basis
points on average to the real rate between 1979QIV
and 1984QIII. When the federal deficit was used,
neither fiscal nor monetary policy was estimated to
have contributed much to the behavior of the real
rate over the same period.

ApPENDIX

Part A
Table A.I presents the outcomes of stationarity
tests for the 3-month nominal Treasury Bill rate and
the two measures of inflation. The test statistics
indicate that the random walk hypothesis is not
rejected for the nominal interest rate, with one
exception: the nominal rate behavior during 1961QI
- 1979QIII is consistent with that of a variable
stationary around a constant trend. Since both i and
'IT appear nontrend stationary for the sample period
as a whole, the hypothesis of cointegration is tested;
that is, does there exist some combination of i and 11'
that is stationary? Can we find a constant a such that
ai t
11't+! is stationary?
Engle and Granger (1987) propose several tests of
cointegration based on the "co-integrating regression" of either it on 'ITt + ! or 'ITt +! on it.! If the real
after-tax rate of interest is stationary, then the coefficient on it in a regression of 'ITt + Ion it should equal
one minus the marginal tax rate. This coefficient
should therefore be of the order of magnitude of 0.6
to 0.7. The reverse regression of it on 'ITt + !' should
yield a consistent estimate of one over one minus the
marginal tax rate, which should be in the approximate range 1.4 to 1. 7.

Under the null hypothesis of no cointegration, the
residuals from the cointegrating regression should
be nonstationary. This implies that the DurbinWatson statistic will approach zero. Thus, a "large"
D-W indicates cointegration. In addition, the
residuals can be subjected to standard tests for a
random walk. Critical values from a Monte Carlo
experiment are reported by Engle and Granger.
Results from the cointegrating regressions are
reported in Table A.2. The column labeled CRDW
gives the Durbin-Watson statistic, the D-F (for
Dickey-Fuller) column gives the t-statistic from a
regression of the first difference of the residuals on
their lagged level, while the ADF column adds four
lagged first differences to the residual regression.
The evidence for cointegration is mixed. For the
1961QI - 1979QIII period, none of the lagged first
differences of the residuals is significant, so D-F
provides the appropriate test, and both it and
CRDW indicate rejection of no cointegration.
When the post 1979QIII period is added to the
sample, some of the lagged first differences are
significant, suggesting the ADF statistic should be
used. In all cases, this fails to reject no cointegration. The CRDW statistics rejects no cointegration
when 'ITt + 1 is regressed on it but not when it is
regressed on 11't+ l ' The estimates of the cointegration parameters (reported in the column labeled a)
also yield mixed results. The estimated coefficient

1. Barsky [1987] discusses the effects of regressing
it on lagged 'ITt's as proxies for Et'IT t +! when
inflation is stationary. However, the results in his
Table 2 suggest that 11' is non-stationary for the
1960-1979 period.

operator L (that is, A(L) = A o + AIL + A 2 U
+ . . . and Li xt = Xt _), and Ut is the vector of
one-step ahead forecast errors. Let S7T be a selection
vector such that S7TZt = TIt (that is, S7T just picks out
TI from the list of variables in z). The equation for TIt
is given by

in aU odd numbered regressions should be around
0.7. Only equation 4 comes close.

Part B
The nominal interest rate innovation, it, is simply
the one-step ahead forecast error for the nominal bill
rate as implied by the estimated VAR. The inflation
forecast innovation, E t1rt + I' can also be obtained
from the VAR in the following manner. Suppose Zt is
the 6x 1 vector of the variables in the VAR at time t.
The VAR system can be written as

Equation A-2 can be used to evaluate E t 1rt + I'
By definition, ~1rt+1 = EtTIt + 1
Et-ITIt+I'
Updating A.2 by one, TIt + I = S7TA(L)zt + S7T Ut+ 1
so that EtTI t + I = S7TA(L)zt· Similarly, Et-ITI t + I is
equaltos7TAoEt_lzt + S7T A IZt-l + S7TA 2Zt-2'"
It follows that

(A. I)

where A(L) is a 6x6 matrix of polynomials in the lag

Et1Tt + I -Et S7TAO (Zt
= S7TAOUt

11T t + I

(A.3)

where Vt = GUt is the orthogonalized vector of
disturbances obtained from the VAR residual vector
A(L)L) IG-I. Using the
Ut , and B(L) = (I
selection vector S7T to pick out the equation for 1T and
Si to pick out the equation for the nominal rate, the ex
post real rate can be expressed as

E t - IZt)

Thus, the revision to the inflation forecast is equal to
a linear combination of the errors made in forecasting all the elements of Zt. Because E t7r t + I depends
on the one-step ahead forecast errors (u t) and coefficients from the VAR (A o), it is easily calculated
from the estimated system.

(A.4)

siB(L)v t - s7TB(L)v t + I

kqJkb~IVjt_q - kgJkb(;;;lvjt+ l-q
~

Parte
where b(ij) is the coefficient on the ith lag of the jth
shock in sxB(L), x = i, 1T. The contribution of the
jth shock to exrt is equal to

The decomposition of the ex post real rate into
components attributable to the various underlying
shocks is based on the moving average representation of the VAR system given in equation A.I:

k(b(i)
- b(7T»)
vJ-q
t
q
qj
qj
(A.3)

The orthogonalized shocks were obtained using a

Choleski decomposition based on the following
ordering of the variables in the VAR: government
purchases and defense expenditures were ordered
first in their respective VARs,Jollowed by real GNP,
Ml, the nominal interest rate, the relative price of
fuel, and the rate of inflation. When the deficit was
used, the ordering was real GNP, Ml, the nominal
interest rate, the relative price of fuel, the deficit and

the rate of inflation.
For each ordering, the hypothesis that a given
variable Granger-caused a variable ordered before it
could be rejected. Note that when the monetary or
fiscal shock is set equal to zero, the predicted path of
the money supply or the fiscal variable will still vary
endogenously in response to movements in the other
variables in the system.

FOOTNOTES
1. See the discussion of real rate targeting in Walsh
(1983).

9. See Litterman and Weiss (1985).
10. Dummy variables for 19800lland 19800111 were also
included to capture the effects of the credit controls in
effect at that time. The use of defense expenditures as a
proxy for fiscal policy is discussed in Section III; the
general conclusions in this section were not affected when
other proxies were used.
11. From equation 5, tt is just equal to 1t
Et'ITt +1. This
also implies that var (1t) = cov (1t, tt) + cov (1t, EtTIt+ 1)·
The fraction of nominal rate innovation variance associated
with real rate innovations can then be estimated by cov
(1t, tt) / var (tt)· The fraction of var (1t) associated with revisions in expected inflation is thus 1
cov (1t, tt) /
var (1t) = COV (ft, EtTIt+1) / var (1t). Since var (1t) = var
(tt) + var (EtTIt+1) + 2 cov (ft, EtTIt+1), the measure
used to estimate the fraction of var (1t) associated with tt is
not equal to var (ft) / var (1t) unless cov (tt, EtTIt+1) = o. In
fact, cov (1t, ft) = var (ft) + cov (tt, EtTIt +1)' so the measure
used here is equal to var ([t) / var (1t) + cov
(ft, EtTIt+1) / var (1t)·

2. By a random walk component I mean that a variable xt
can be written as Yt + Zt where Yt is a stationary random
variable and Zt = Zt _ 1 + lOt when £t is a stationary process. Realizations of lOt have permanent effects on Zt and
Xt·

3. The test statistic does not have a standard t-distribution,
but the appropriate critical values are given in Fuller
(1976).
4. For nontechnical introductions to real business cycle
theories, see Walsh (1986,1987).
5. If all changes in a variable Xt are permanent, then the
variance of xt + k - xt is equal to k times the variance of xt +1
- Xt. If all changes in Xt are temporary, then the variance of
Xt+k
Xt should tend to zero for large k. Thus the ratio
~ • var (Xt+k -

Xt)

var (Xt + 1 - Xt)

12 Based on a VAR estimated using quarterly data from
1949011 to 1983011, Litterman and Weiss (1985) report that
a i-percent innovation to the nominal rate was, on average,
associated with a .56 percent real rate innovation and a .44
percent expected inflation innovation, exactly the same
division reported in Table 2 for the 196101-19840111
period. Note that, while the actual estimation period runs to
198401V, one observation is lost in calculating the realized
future rate of inflation needed to form the ex post real rate.
13. For a discussion of this view, see Barro (1984, Chapter
15). Some empirical evidence is presented in Motley
(1987).

is a measure of the relative importance of the random walk
component. The ratio Uk equals 1 for a pure random walk
and zero if all changes are transitory. Cochran's method is
evaluated in Campbell and Mankiw (1987).
6. Because the ex post real rate used to construct the
measure of persistence is equal to the ex ante real rate plus
a serially uncorrelated inflation forecast error, Cochrane's
measure will yield a value of Uk less than one for the ex post
rate even if the ex ante rate is a pure random walk.
7. For discussions of the use of the nominal interest rate in
the conduct of monetary policy, see Sargent and Wallace
(1975), McCallum (1986), and Goodfriend (1987).

14. The estimated effects of the 1980 credit controls have
been subtracted out of the real rate series plotted in both
Chart 2 and Chart 3.

8. From the properties of conditional expectations,

REFERENCES
Antoncic, Madelyn, "High and Volatile Real Interest Rates:
Where Does the Fed Fit In?," Journal of Money, Banking and Credit, Vol. 18, No, 1 (February 1986),
Barro, Robert J, Macroeconomics, New York: John Wiley
and Sons, 1984,
Barsky, Robert B, "The Fisher Hypothesis and the Forecastabilityand Persistence of Inflation," Journal of
Monetary Economics, Vol. 19, No, 1 (January 1987),
Blanchard, Olivier and Lawrence Summers, "Perspectives
on High World Real Interest Rates," Brookings Papers
on Economic Activity, 1984:2,
Campbell, John Y, and N, Gregory Mankiw, "Permanent
and Transitory Components in Macroeconomic Fluctuations," American Economic Review, Vol. 77, No, 2
(May 1987),
Cecchetti, Stephen G, "High Real Interest Rates: Can They
be Explained?," Economic Review, Federal Reserve
Bank of Kansas City, September/October 1986,
Cochrane, John H, "How Big is the Random Walk in
GNP?", mimeo, February 1986,
Dickey, David A. and Wayne A. Fuller, "Likelihood Ratio
Statistics for Autoregressive Time Series with a Unit
Root," Econometrica, Vol. 49, No, 4 (July 1981),
Engle, Robert F, and C,WJ, Granger. "Co-Integration and
Error Correction: Representation, Estimation and
Testing," Econometrica, Vol. 55, No, 2 (March 1987),
Fama, Eugene, "Short-Term Interest Rates as Predictors of
Inflation," American Economic Review, Vol. 65 (1975),
Fuller, Wayne A. Introduction to Statistical Time Series,
New York: John Wiley and Sons, 1976,
Goodfriend, Marvin, "Interest Rate Smoothing and Price
Level Trend Stationarity," Journal of Monetary Economics, Vol. 19, No, 3 (May 1987),
Granger, C, WJ, "Developments in the Study of Cointegrated Economic Variables," Oxford Bulletin of Economics and Statistics, Vol. 48, No, 3 (August 1986),
Hendry, David F. "Econometric Modelling with Cointegrated Variables: An Overview," Oxford Bulletin of
Economics and Statistics, Vol. 48, No, 3 (April 1986),

Huizinga, John and Frederic S, Mishkin, "Monetary Policy
Regime Shifts and the Unusual Behavior of Real Interest Rates," in Carnegie-Rochester Conference Series
on Public Policy, 24 (Spring 1986),
Jenkins, Paul and Carl E, Walsh, "Real Interest Rates,
Credit Markets, and Economic Stabilization," Journal
of Macroeconomics, Vol. 9, No, 1 (Winter 1987),
Litterrnan, Robert B, and Laurence Weiss, "Money, Real
Interest Rates, and Output: A Reinterpretation of Postwar U,S, Data," Econometrica, Vol. 53, No, 1 (January
1985),
McCallum, Bennett T, "Some Issues Concerning Interest
Rate Pegging, Price Level Determinacy, and the Real
Bills Doctrine," Journal of Monetary Economics, Vol,
17, No, 1, January 1986.
Motley, Brian, "Ricardo on Keynes: Does Government
Debt Affect Consumption?", Economic Review,
Federal Reserve Bank of San Francisco, Winter 1987.
Nelson, Charles R, and Charles I. Plosser, "Trends and
Random Walks in Macroeconomic Time Series: Some
Evidence and Implications," Journal of Monetary Economics, Vol. 10, No, 2 (Septernber 1982).
Rose, Andrew K, "Is the Real Rate Stable?" mimeo, April
1987.
Walsh, Carl E, "Should the Federal Reserve Establish a
Real Interest Rate Target?", Economic Review,
Federal Reserve Bank of Kansas City, June 1983,
Walsh, Carl E. "New Views of the Business Cycle: Has the
Past Emphasis on Money Been Misplaced?", Business Review, Federal Reserve Bank of Philadelphia,
January/February 1986,
Walsh, Carl E. "Real Business Cycles," Weekly Letter,
Federal Reserve Bank of San Francisco, January 30,
1987,
Wilcox, James A. "Why Real Interest Rates Were So Low in
the 1970s," American Economic Review, Vol, 73,
No, 1 (March 1983),

Christopher James*
Commercial loan sales and the issuance of Standby Letters of Credit
(SLCs) involve the separation of many of the services associated with
lending, such as credit risk evaluation and underwriting, from the
funding of a loan. These activities are shown to provide banks a way of
issuing collateralized debt claims. This ability can induce banks to
undertake profitable loan opportunities they would not undertake if
restricted to deposit financing. Moreover, the incentives to issue collateralized claims increase when capital requirements are raised. Empirical
evidence suggests that loan sales and SLCs are not important determinants of bank risk.
Over the past decade, there has been a dramatic
increase in what is called "off-balance sheet" banking. Examples include the issuance of Standby
Letters of Credit (SLCs) and commercial loan sales.
These activities have the common feature of separating many of the services associated with lending,
such as credit risk evaluation and underwriting,
from the funding of a loan. By separating the
funding of a loan from these other activities, a bank
earns fee income without putting an asset or corresponding liability on its balance sheet.
This paper examines two questions pertaining to
commercial loan sales and the issuance of SLCs.
The first concerns the regulatory and other economic factors that induce a bank to separate the
funding of a loan from the other services associated
with lending. The most frequently cited explanation
for the growth of these activities is that they provide
banks a way of avoiding reserve requirements and
bank capital adequacy requirements.
While these regulations may provide incentives to
go "off-balance sheet," nonregulatory factors are
also important. In particular, as Benveniste and
Berger (1986) show, SLC-backed loans and commercial loan sales have payoff characteristics that
are similar to secured or collateralized debt. This
observation suggests that the incentive banks have

to sell loans or to issue SLCs may be similar to the
incentives other financial as well as nonfinancial
firms have to issue secured debt. How bank regulation affects the incentives to issue collateralized
debt and the linkage between capital requirements
and off-balance activities are also explored.
A second and related question concerns the effect
of loan sales and SLC issues on the default risk of
deposits (borne by uninsured depositors and/or the
FDIC). The effects of off-balance sheet activities on
the risk of deposits depends on the reasons banks
undertake those activities. For example, one explanation for the growth of off-balance sheet banking is
that it is a manifestation of a moral hazard problem
that is endemic to a system of fixed rate deposit
insurance pricing. Because SLCs and certain loan
sales are not subject to capital requirements, these
contingent liabilities provide a way for a bank to
increase leverage. By increasing leverage, a bank
can generate or enhance subsidies arising from
deposit insurance. This argument implies that offbalance sheet activities increase the risk of deposits.
An alternative explanation, examined in this
paper, is that loan sales and SLC issues permit banks
to engage in lending that they would find unprofitable to undertake if they were restricted to funding
loans through deposit financing. This argument
implies that off-balance sheet activities may
enhance bank profitability and reduce bank risk.
The effect of loan sales and SLC issues on bank
risk therefore is an empirical issue. To address this

* Visiting scholar Federal Reserve Bank of San
Francisco. Thanks to Bill Robertson for research
assistance.
21

issue, the relation between the interest rate paid on
bank large CDs (greater than $100,000) and bank
asset risk, financial leverage, the volume of SLCs,
and loan sales is examined. The results reveal that
the risk premium of large CDs increases with asset
risk and financial leverage. However, no significant
relation is found between the rate paid on CDs and
either the volume of SLCs issued or loans sold. This
suggests that SLCs and loan sales are not an impor-

tantdeterminant of bank risk as perceived and
priced by large depositors.
The paper is organized as follows. In Section I, I
describe the market for SLCs and commercial loan
sales. In Section II, the reasons for the use and
growth of SLCs and loan sales are discussed. In
Section III, the effects of loan sales and SLCs on
bank risk are examined empirically. Section IV
provides a summary.

The Market for Commercial Loan Sales and SLCs

Commercial loan sales involve the sale of newly
originated commercial loans. Most commercial
loan sales are structured contractually as participations so that the selling bank maintains a creditordebtor relationship with the borrower. 1 This means
that the selling bank continues to be responsible for
servicing the loan, enforcing covenants, monitoring
the financial condition of the borrower, and handling workouts and other problems that might arise in
the event of default. In exchange for performing
these services, the selling bank is compensated
through a "spread." The spread represents the difference between the rate paid by the borrower to the
bank and the return promised the purchaser of the
loan. An average spread of 15 basis points was
reported on commercial loan sales in the June 1,
1987 Senior Loan Officer Lending Practices Survey
(LPS).
Current bank regulations require that loans sold
with recourse (that is, with an issuing bank's guarantee against default) be treated as assets when
calculating capital requirements. Moreover, the proceeds of loans sold with recourse are subject to
reserve requirements. As a result of these tegulations, commercial loans are rarely sold with
recourse. 3
While loans sold without recourse avoid reserve
requirements and capital requirements, they raise
concerns with the purchaser regarding both the
quality of loans sold (an adverse selection problem)
and the diligence with which the selling bank will
monitor the borrower after a sale (a moral hazard
problem). One technique used to provide the purchaser a credible assurance of quality is for the

selling bank to maintain or fund a portion ofthe loan
sold. A second technique involves selling shortterm "strips" oflonger term loans. While the buyer
of the strip is exposed to default risk in the short-run,
before the maturity date of the strip, the originating
bank retains exposure to default in the longer run if it
is committed to re-financing the loan. Finally,
because most commercial loans sold are short-term
and selling banks return repeatedly to the market,
"reputational" capital (that is, the value of future
earnings, which depend on honest dealing) may
provide a mechanism for assuring quality (Gorton
and Haubrich, 1987, make this argument).
Information on the volume of commercial loans
sold indicates a dramatic increase in sales over the
past few years. Information on the volume of loan
sales comes primarily from two sources: Schedule L
of the Call Report and periodic Senior Loan Officer
Opinion Surveys on bank lending practices (LPS)
conducted by the Federal Reserve System. Information from the Call Report indicates total loans sold
increased from $23 billion in 1983 to $111 billion in
1986; an increase of 326 percent. A similar pattern
of growth is observable in the LPS survey data. In
November 1984, LPS respondents reported less
than $5 billion in sales. By March 1987, sales of
37.5 billion were reported. Chart 1 shows the
growth in loan sales (as reported in the Call Report)
relative to commercial and industrial (C&I) loans
outstanding over the 1984 through 1986 period.
Loan sales have grown from about 7 percent of C&I
loan volume to almost 25 percent of C&I loan
volume over this period.

Chart 1
Commercial Loan Sales Have Grown
as a Percent of Total C & I Loans

credit information without funding the loan.
As discussed in the next section, the payoff or
cash flow characteristics of an SLC-backed loan are
identical to a loan sale with full recourse. However,
unlike loan sales, SLCs are not, under current
capital regulation, considered when calculating
capital requirements. Under risk-based capital standards recently proposed by the Federal Reserve
Board however, SLCs would be treated the same as
loans sold with recourse when calculating capital
adequacy requirements. 4
Like the commercial loan sales market, the volume of SLCs has grown rapidly in recent years. For
example, since 1980, SLCs outstanding have grown
at an annual rate of 20 percent, from $47 billion to
$169 billion in 1986. Chart 2 shows the growth of
SLCs relative to C&I loans over the 1980 through
1986 period. As Chart 2 reveals, SLCs have grown
faster than C&I loans during the 1980s.

Percent

25
20
15
10
5

1983

1984

1985

1986

Source: Report of Condition

Payoff Characteristics
It is useful when analyzing the reasons for loan
sales and SLC issues to begin by evaluating the
payoff or cash flow characteristics of these activities
because they determine when a bank will undertake
those activities. Specifically, the focus is on factors
that affect the cash flows received by the purchaser
of a loan or the beneficiary (lender) in an SLC-

The SLC Market
SLCs are similar to loan sales in that they also
involve a separation of many of the services associated with a commercial loan from the funding of the
loan. An SLC is a guarantee by a bank to pay one
party (called the beneficiary) if the bank's customer
(called the account party) fails to repay a loan or
perform some other contractual obligation (for a
description of the SLC market see Bennett, 1986, or
Koppenhaver, 1987). Because the bank's obligation
is contingent on the default or nonperformance of
the account party, most SLCs expire unused. For
example, a special survey conducted by the staff of
the Federal Reserve Board found that defaults by
account parties constituted only 2.03 percent of
SLCs outstanding in 1978. (Bank losses were much
smaller because 98 percent of payments made were
recovered from account parties.)
The majority of SLC issues are used to back
financial contracts such as commercial paper, municipal bonds and direct loans. Because the issuing
bank assumes the credit risk associated with an
SLC-backed loan, the bank has the same incentive
to evaluate and monitor the credit risk of the borrower as if it had funded the loan. SLCs therefore
provide a bank the opportunity to realize its comparative advantage in obtaining and processing

Chart 2
SLCs Have Grown Faster
Than C & I Loans
Ratio SLC/C&I

.50
.45
.40
.35
.30
.25
.20
.15
.10
.05
.00 +--,-----,,---,---r-...,---,----,
1980

1981

1982

1983

Source: Report of Condition

1984

1985

1986

backed loan. The promised rate the purchaser or
beneficiary will require depends on the expected
cash flows, and determines the profitability of a loan
sale or SLC issue for the bank.
The payoff characteristics of SLC loans and loan
sales with recourse, as Benveniste and Berger
(1986) have shown, are similar to the cash flow
characteristics of secured or collateralized debt. 5
Consider first a loan sale made with recourse. (The
effect of removing the recourse provision on the
cash flows will be discussed later.) The loan sold is
the primary source of cash flows to the purchaser. In
the event of a default on the loan, the purchaser still
receives the contracted payment as long as the
selling bank does not fail. SLC-backed loans operate in a similar fashion. The primary source of cash
flows is the loan funded. The lender receives less
than the contracted rate on the loan only when the

II.

borrower defaults and the bank fails. If the bank
could issue uninsured deposits secured by a specific
loan,precisely the same factors would determine
cash flows to secured depositors. Specifically, the
secured depositor would receive less than the contracted payment only when the bank failed and the
cash flows on the loan serving as collateral were less
than the contracted payment due on the debt.
Loans sales with recourse and SLC-backed loans
are therefore functionally equivalent to secured debt
and should have the same expected rate of return in a
competitive market. The payoff characteristics of
loan sales without recourse depend solely on the
cash flows of the underlying loan because the selling
bank issues no guarantee in the event of default.
Because banks are generally prohibited from issuing collateralized deposits or debt, loan sales and
SLCs provide effective substitutes. 6

Reasons for SLCs and Loan Sales

The reasons for the growth and use of loan sales
and SLCs can be divided into two groups: regulatory and non-regulatory explanations (although the
explanations are not mutually exclusive). Regulatory explanations focus primarily on the incentives
capital adequacy requirements, reserve requirements, and deposit insurances provide for issuing
SLCs or selling loans. The nonregulatory explanations focus on why these activities might take place
even in the absence of bank regulation and deposit
insurance. Both sets of explanations explain why
banks have begun increasingly to unbundle funding
from other lending activities.

bank above what nonbank institutions must pay.
Therefore, while a bank may have a comparative
advantage in originating and servicing loans, regulatory taxes prevent it from profitably funding
certain types of loans.
This argument implies that the cost of bank
regulation exceeds the benefits banks receive from
access to deposit guarantees and the Federal
Reserve's discount window, that is, that deposit
insurance is overpriced when regulatory taxes are
considered.
While reserve requirements unambiguously
increase the cost of bank funds, the effect of capital
requirements is not well understood. In a world
without taxes and transaction and agency costs (the
costs of controlling potential conflicts between
bondholders and stockholders as well as between
managers and outside investors), the cost of financing a loan would be independent of a bank's capital
structure (the mixture of debt and equity used). This
is the famous Modigliani and Miller (1958) proposition concerning corporate capital structure. Introducing agency costs, personal and corporate taxes
have been shown to yield an optimal capital structure (see, for example, Miller, 1977, and Bamea,
Haugen and Senbet, 1981).

Regulatory Motives
Two hypotheses have been made concerning how
regulation affects off-balance sheet banking: the
regulatory tax hypothesis and the moral hazard
hypothesis.
The regulatory tax hypothesis (see Pennacchi,
1987, and Pavel and Phillis, 1987), argues that loan
sales and SLCs are responses to burdensome regulatory taxes. In particular, it says that the cost of
holding noninterest-earning reserves, the need to
meet capital requirements, and the level of deposit
insurance premiums raise the cost of funds for a

Given these costs, if the maximum debt to equity
ratio set by the regulators is below what banks
would hold in the absence of regulation (and deposit
insurance), then capital requirements might serve to
raise the cost of bank financing above that of nonbank institutions. However, because the debt to
equity ratio for banks generally exceeds that of other
financial institutions (and nonfinancial firms), it is
unclear in what way capital requirements impose a
tax on banks.
An alternative hypothesis concerning how regulation and deposit insurance influence off-balance
sheet activities is the moral hazard hypothesis. This
hypothesis focuses on the incentives a bank has to
increase asset risk and financial leverage when
deposit insurance is provided at a fixed price. By
increasing leverage or the risk of its assets, a bank
can generate or enhance the subsidies associated
with fixed rate deposit guarantees. Under existing
regulation SLCs are excluded from capital requirements. Therefore, by issuing SLCs, a bank can
increase its financial leverage and enhance whatever
subsidies it receives from the deposit insurer.
As Pyle (1985) and others have argued, fixed rate
deposit insurance together with capital requirements provide incentives to undertake "off-balance
sheet" activities that increase financial leverage.
Moreover, by selling relatively low risk loans and
maintaining riskier loans in its portfolio, a bank can
increase risk and therefore raise the subsidy deposit
insurance provides. The moral hazard hypothesis
predicts that off-balance sheet activities will
increase bank risk.
While regulation may enhance incentives for a
bank to engage in off-balance sheet activities, it is
unlikely that bank regulation is solely responsible.
Several institutional facts support this conjecture.
First, nonbank financial institutions, which are not
subject to the same regulatory taxes and do not issue
insured deposits, are active participants in the loan
sales and financial guarantee markets. For example,
General Motors Acceptance Corporation (GMAC)
sold over $7 billion in loans during 1986. 8 In
addition, insurance companies issue financial guarantees that compete directly with bank-issued
SLCs. The volume of these guarantees, according to
Hirtle (1987), has grown at approximately the same

rate as bank-issued SLCs (that is, 20 percent per
year since 1980).
Moreover, according to recent LPS surveys, a
significant portion of loans sold are purchased by
other commercial banks. For example, in 1985,
approximately half of the loans sold were purchased
by other domestic commercial banks (and 36.5
percent were purchased by banks with assets of over
$1 billion). The 1987 Survey indicates that 35
percent of commercial loans sold were purchased by
other domestic banks. Because most banks (and all
banks over $1 billion in assets) are subject to the
same marginal reserve requirement on deposits and
money center banks (the primary sellers of loans)
generally hold less capital than do other banks, it is
unclear why the regulatory tax burden should be
higher for financing the same loan for money center
banks than for regionals. 9
Nonregulatory Motives
Nonregulatory motives may also provide incentives for separating funding from other lending
activities. SLC issues and loan sales facilitate interest rate risk management and loan portfolio diversification (see, for example, Pavel and Phillis, 1987,
and Koppenhaver, 1987). SLCs permit banks to
separate the interest rate risk from the credit risk
associated with a loan. With an SLC issue, a bank
can underwrite the credit risk while the beneficiary
or purchaser bears the risk of any change in the
value of the loan caused by unanticipated changes in
interest rates. Loan sales also permit banks to invest
in and diversify across a different set of loans than
they originate and service.
A problem with this set of explanations is that it is
unclear why bank stockholders would reward bank
management for these activities when presumably
they can diversify their own portfolios or hedge
interest rate risk themselves. 10
Collateralization as a Motive
A second set of nonregulatory explanations for
loan sales and SLC issues is that such activities
permit banks to issue a collateralized claim. As
shown in Section II, loan sales and SLC-backed
loans have payoff characteristics similar to secured
debt. Therefore, banks may sell loans and issue

determine when a bank will find it advantageous to
engage in off-balance sheet activities requires
knowledge of when the cost of financing a new loan
will be less using a loan sale or an SLC than using
deposits. In addition, to determine how off-balance
sheet activities affect the risk of bank deposits
requires examining how these activities affect the
types of loans a bank will make.
To abstract from the effects of bank regulation
and deposit insurance, these questions are examined
first in the context of a deregulated environment
without· deposit insurance; examination of the
effects of insurance and regulation follows.

SLCs for the same reasons nonfinancial as well as
nonbank financial corporations issue collateralized
claims. Moreover, as discussed later, fixed price
deposit insurance and capital requirements can
increase the incentives for banks to issue collateralized claims.
Stulz and Johnson (1985) have analyzed the economic rationale for secured debt issues by nonfinancial firms. One hypothesis they studied argues that
firms can effect a wealth transfer from unsecured
debtholders to the firms' shareholders by increasing
the default risk of unsecured debt. This shift occurs
when a firm unexpectedly issues secured debt using
existing assets as collateral. By providing secured
debtholders a higher priority claim to the cash flows
from some of the firm's existing assets, the remaining unsecured claimants are worse off. (This is
similar to the moral hazard hypothesis discussed
earlier.) As Stulz and Johnson point out, if this were
the primary reason secured debt is used, unsecured
debt would contain convenants prohibiting secured
debt issues. II
An alternative hypothesis, referred to by Stulz
and Johnson as the underinvestment hypothesis, is
that the ability to issue secured debt can affect a
firm's investment policy and therefore the size of the
firm's cash flows as well as how cash flows are
distributed among claimants. In particular, the ability to issue collateralized debt may enable a firm to
undertake profitable new investment opportunities
that it would pass up if constrained to issue
unsecured debt. This can occur when the firm has
risky debt outstanding that pays a contractually
fixed rate of interest. The promised payment on new
unsecured debt will reflect the uncertainty concerning the cash flows associated with the firm's existing
assets as well as the newly acquired asset. However,
if secured debt were used to finance a new project,
the contracted rate would primarily reflect the
uncertainty concerning the cash flows associated
with the new investment opportunity. If the new
investment were relatively low ri:;k, the cost of
secured debt would also be lower than the cost of
unsecured debt. Therefore, the firm may undertake a
project using secured debt that it would pass up if
constrained to issue unsecured claims.
A similar argument can be made for the use of offbalance sheet activities by commercial banks. To

Cost of Financing
The effects of collateralization on the cost of
financing is analyzed in the context of a two-period
model. A bank makes a new loan at time t = 0 and
realizes cash flows at time t = 1. The new loan has
payoffs denoted as a2(s), that is, cash flows are
contingent on the state of the world, "s", at time
t = 1. For simplicity, the new loan is assumed to have
a face value of $1. The bank is also assumed to have
"booked" loans with payoffs at time t= 1 of a](s).
Booked loans have been financed with a mixture of
deposits and equity.
If the new loan were financed by issuing deposits
promising a payment of r d at time t = 1, then the
realized payment to new depositors in any state
would be
(1)

where Ld equals the sum of contracted payments to
depositors. New depositors will receive the contracted payment, rd , if the bank does not default; in
the event of default, they receive a proportion of the
cash flows from the bank's assets.
Suppose that instead of deposit financing, the
new loan is sold with recourse. The contracted or
promised payment to the purchaser of the loan is rsr
(this represents 1 plus the contracted interest rate).
This contracted payment will generally be less than
the rate the bank charges on the loan sold - the
difference representing the bank's "spread." The
realized payments to the purchaser of the loan will

depend on the promised rate or rSf and the payments
the purchaser receives in the eventof default on the
loan sold and the selling bank's failure. Denoting the
cash flows associated with the underlying loan sold
as a2(s), the realized payoff to the loan purchaser for
a given state is

Specifically, when investors are risk neutral, r Sf will
be less than rd if and only if (I) there is some positive
probability of bank failure with deposit financing
(that is, deposits are risky) and (2) in the event of
default, the cash flows collateralized debtholders
receive are larger than the cash flow new depositors
would receive in the event of bankruptcy. That is,

(2)
(5)
where L Sf equals the sum of the contracted payments
to the loan purchaser and depositors. In words,
when no default occurs, the loan purchaser receives
the contracted payment rSf ' In the event of the bank's
failure, the purchaser receives r Sf when the cash
flows from the loan are sufficient to meet the contracted payment. When not, the purchaser receives
the cash flows from the loan sold, a2(s), plus a
proportion of the cash flows from the bank's other
assets. As discussed in Section 11, expression 2 also
describes the cash flows to the lender in the case of
an SLC-backed loan and the cash flow characteristics of a secured deposit claim.
The expression for the realized cash flows for a
loan sold with partial recourse is more complex. Let
a represent the proportion of the loss guaranteed.
The realized cash flow to the purchaser is
min [rSf' a2(s)

+ min {a[rSf - a2(s)],

or, for a nonrecourse loan sale: 12

The left hand side of expression 5 is what secured
depositors receive, and the right hand side is what
uninsured depositors would receive in the event of
default. Intuitively, if investors were risk-neutral,
the expected return on deposits and secured debt
must be equal. For the contracted rate on coll'l-teralized debt (that is, the promised payments investors receive when default does not occur) to be less
than the rate paid on deposits, collateralized debtholders (or loan purchasers) must expect higher
payments in the event the bank fails. This condition
is expressed in 5 or 6.
Note that the above discussion suggests that the
difference between r Sf and rd will be greater the
higher the probability of bank failure (that is, the
riskier a bank's deposits) and the lower the risk of
the collateral (that is, the default risk of the new
loan). This suggests that collateralization provides
the greatest benefit for high-risk banks investing in
low-risk loans (that is, investment grade credits).

(3)

a rSf - ais) al(s)}]
L Sf
In words, the purchaser receives either the contracted payment r Sf or, in the event of default, the
cash flows from the loan plus either reimbursement
for losses or a proportional claim on the bank's other
assets. When no recourse is provided, a equals zero
and the realized cash flows become simply

(4)

Types of Loans
How does issuing SLCs or selling loans affect a
bank's investment policy, that is, the types of loans
it will make? The effect can be illustrated by the
following example. Suppose a bank has a portfolio
of risky loans and has risky deposits outstanding
(deposits with a positive probability of default).
Ignore deposit insurance for the moment. The con-

In this case, the rate paid by the purchaser depends
solely on the characteristics of the new loan sold.
By comparing the payoff characteristics of new
deposits (expression I) to the payoff characteristics
of a loan sale or SLC-backed loan (expressions 2 or
3), one can determine when the rate paid on collateralized debt will be less than the rate on deposits.

tracted rate the bank must pay on deposits will
reflect the risk of default.
Assume that the bank has an opportunity to invest
in a .new loan that has a positive net present value
and yields a safe or certain return. If the new loan
were financed by issuing additional deposits, investing .in the loan would reduce the risk of existing
deposits (since they receive a proportional claim in
the cash flows of the new loan). If the existing
deposits pay a contractually fixed interest rate, the
market value of the deposits would increase because
the new loan lowers the likelihood of bankruptcy
and increases the level ofthe bank's cash flows. This
outcome implies that old or existing depositors
gain. Moreover, these depositors' gain lowers the
return bank shareholders receive from making the
new loan. The lower return to shareholders reduces

their incentives to make new relatively low risk
loans. This transfer is illustrated numerically in the
box (Case 1).
Selling loans or issuing SLCs provides a bank an
incentive to undertake low risk loans by reducing
this transfer. Recall that the promised rate on colIateralized debt, rsr (or loan sales and SLC-backed
loans), will be lower than the rate paid on new
deposits when the payoffs to secured debtholders are
larger in the event of bank failure (that is, expression
5 or 6 holds). However, this implies that, in the event
of failure, existing depositors would receive less
than if new deposit claims were issued. Therefore, if
the contracted rate on secured debt is less than the
rate on deposits (that is, expression 5 holds), the
gain existing depositors realize is less, and the
return to shareholders is larger when secured debt is

issued. Case 2 in the Box provides a numerical
illustration.
An alternative and perhaps more intuitive explanation for how issuing collateralized claims affects
investment policy is that the rate on an SLC-backed
loan or loan sale will reflect primarily the riskof the
new loan (in the case of a nonrecourse loan sale, the
rate reflects only the risk of the new loan). The rate
paid on uninsured deposits will reflect the average
risk of the bank's loan portfolio. Therefore, the cost
of financing a relatively low risk loan will be less
with loan sales or SLCs than with deposit claims.

Implications
Thecollateralization argument points out an
important aspect of off-balance sheet activities and
restrictions on bank financial policy generally: The
financing techniques available to a bank affect its
investment policy and therefore its overall profitability. This implication is similar to an implication
of the regulatory tax hypothesis in that both imply
that off-balance sheet activities may permit banks to
engage in investment opportunities that they might
pass up if constrained to use deposit financing.
Moreover, unlike the moral hazard hypothesis
which predicts off-balance sheet activities increase
bank risk, the collateralization hypothesis implies
that the risk of deposits does not necessarily
increase with off-balance sheet activities because
even though leverage may increase, profitable loan
opportunities of lower risk are undertaken 13. In
addition, while the collateralization hypothesis is
consistent with the regulatory tax hypothesis discussed earlier, it suggests that even if regulatory
taxes were eliminated (or extended to off-balance
sheet activities), banks would still have an incentive
(albeit reduced) to engage in loan sales or to issue
SLCs.
The collateralization hypothesis yields important
implications concerning the types of loans sold or
backed by SLCs and the effect of off-balance sheet
activities on the default risk of bank deposits. First,
the collateralized debt argument suggests that relatively low-risk loans will be sold or backed by an
SLC. LPS surveys on loan sales indicate that currently loan sales are concentrated primarily in loans
to investment-grade credits. For example, the 1986
LPS survey indicated that two-thirds of the loans
sold by respondents were obligations of investmentgrade borrowing. Second, the riskier a bank's existing deposits (and therefore the higher the rate the
bank must pay on new uninsured deposits), the more
likely it will be to engage in off-balance sheet
activities. Finally, the collateralization hypothesis
indicates that SLC and loan sales may not adversely
affect the risk of deposits (the same reasoning
explains why unsecured creditors of nonfinancial
firms permit secured debt issues).

Deposit Insurance
The above discussion is intended to show why a
bank might issue SLCs and sell loans even in the
absence of regulation and deposit insurance. Introducing deposit insurance does not affect the basic
conclusions as long as the rate paid on deposits,
including insurance premiums and regulatory taxes,
exceeds the risk-free rate. With fixed rate deposit
insurance, the rate on existing deposits will not
adjust fully to reflect the marginal contribution of
the new loan to the overall risk of the bank. Indeed,
with complete, that is, 100 percent, insurance, the
cost of deposits does not adjust at all to changes in
asset risk. Therefore, a bank with risky deposits
outstanding will tend to underinvest in relatively
low risk loans and overinvest in high risk loans. This
phenomena is referred to as the underinvestment
problem (see Myers, 1977).
It is important to point out that capital requirements can exacerbate the underinvestment problem
and enhance incentives to go "off balance sheet."
By increasing the amount of equity required to
finance new loans, the gain both existing uninsured
depositors and the FDIC receive from a bank that
undertakes a new low risk loan increases (because
increased capital requirements lowers the risk of
new loans to depositors). Therefore, as capital
requirements are raised (as they were in 1981) loan
sales and SLC issues would be expected to increase.

III.

EmpiricalEvidence

The various explanations for why banks issue
SLCs and sell loans have different implications for
the effect of these activities on bank risk. The moral
hazard hypothesis predicts loan sales and SLCs
increase bank risk, while the regulatory tax and
collateralization hypotheses predict that these
activities do not necessarily increase bank risk.
Ideally, to determine the effect of these activities
on the risk of deposits one would examine the
relation between the risk exposure of the FDIC and
uninsured depositors and a bank's use of SLCs and
loan sales. While the FDIC's risk exposure is not
directly observable, one can obtain a measure of the
risk premium on a bank's uninsured (or partially
insured) deposits. Assuming uninsured depositors
behave as if they are not implicitly fully insured, as
recent evidence by Hannan and Hanweck (1987)
suggests, the moral hazard hypothesis predicts a
positive relation between the risk premium on uninsured CDs and the volume of SLCs and loan sales.
However, if one motive for loan sales or SLC is to
avoid an underinvestment problem or regulatory
taxes, existing depositors as well as bank stockholders may be better off given SLCs and loan sales.
Therefore, finding no significant relation (or a negative relation) between the risk premium on bank
CDs and the volume of SLCs and/or loan sales is
consistent with the collateralization hypothesis and
inconsistent with the moral hazard hypothesis.
To examine the effect of SLCs and loan sales on
bank risk the relation between the interest cost on
large CDs (deposits in excess of $100,000), the
volume of SLCs and loan sales, and a set of variables designed to act as proxy for other factors
affecting bank risk is examined.
The interest cost of large CDs is estimated from
information contained in the Consolidated Report of
Condition and Income. The average rate paid on
CDs is estimated by dividing the total interest paid
on large domestic CDs during a quarter by the
average dollar value of domestic CDs outstanding
during the quarter. A problem with this measure,
noted by previous researchers (see Baer and Brewer,
1986) is that it fails to account for differences in the
maturities of CDs outstanding. 14 However, the large

bank supplement to the Report of Condition contains information on the maturity structure of CDs
outstanding. From this information, a weighted
average maturity of a bank's CDs can be computed. ls
The interest cost on large CDs in a quarter is
assumed to be a function of several factors: (l) the
average maturity of the CDs outstanding, (2) the
general level of interest rates as measured by average yield on ninety day Treasury bills over the
quart~r, (3) the leverage of the bank, (4) the default
or credit risk ofthe bank's loan portfolio, and (5)the
interest rate risk of the bank.
Month-end quotes for the yield on 90-day Treasury bills in the secondary markets are used to
calculate the average yield on Treasury bills during
each quarter. Financial leverage is estimated as the
ratio of total assets of the bank (or bank holding
company) to the market value of total bank capital.
The total market value of capital is estimated as the
sum of the book value of subordinated debt and
preferred stock of the bank or bank holding company and the market value of common stock of the
bank or bank holding company. The market value of
common stock outstanding is calculated by multiplying the number of shares outstanding at the
beginning of the quarter by the price of the bank's
stock at the beginning of the quarter.
Two variables are used to measure the risk of a
bank's asset portfolio. The first measure is th~ provision for loan and lease losses in each quarter divided
by the end of the quarter total of loans and leases
outstanding. A second measure is the variance of
the bank's or bank holding company's monthly
common stock returns for the twelve months prior to
the end of each quarter. The variance in stock returns
is multiplied by the square of the ratio of the asset to
market value of equity. This adjusted variance measure provides an estimate of the variance. of the
bank's asset returns. 16
The interest rate risk of the bank is measured by
the maturity mismatch. b.etween the •bank's assets
and liabilities. A measure of maturity mismatch,
identical to the one used in Flannery andJames
(1984), is constructed from the Call Report. This

measure, denoted as "Short" represents the absolute value of the difference between dollar value of
assets subject to repricing within one year and the
dollar value of liabilities subject to repricing within
the same period, divided by the book value of
equity. 17

sheet. (Only bank holding companies with the lead
bank constituting 75 percent or more of the holding
companies' assets in 1986 were included in the
sample. For the holding companies in the sample,
the assets of the lead bank averaged 90 percent of the
holding company assets.)
Quarterly data over the period 1984 through 1986
were used to test the model. This period was chosen
because the first full year loan sales were reported in
the Call Report is 1984.

Data
The empirical analysis is based on a sample of
fifty-eight banks. Banks were included in the sample if they met the following criterion: (I) information for the bank or bank holding company was
contained in the Compustat Quarterly Bank File
during the period 1984 through 1986, and (2) a lead
bank was identifiable in the case of a multibank
holding company.
Only banks contained in the Compustat Quarterly
Bank File were included because Compustat is used
to obtain monthly stock prices and balance sheet
information for the bank holding companies. Only
bank holding companies with an identifiable lead
bank were included in the sample so balance sheet
items obtained from the lead bank's Call Report will
adequately reflect the holding company's balance

Empirical Results
Table 1 provides descriptive statIstics for the
banks in the sample. It is interesting to note that
SLCs and loan sales constitute a sizable proportion
of the total capital of the bank holding company.
Total SLCs, (the sum of SLCs issued from foreign
and domestic offices) average 95 percent of total
capital, with a maximum value of 12 times total
capital. The average ratio of loan sales to total
capital is 24 percent.
Because the empirical analysis is based on an'
assumption that the rate paid on CDs reflects a

default risk premium, the first step was to investigate the relation between average CD rates and the
measures of bank leverage and asset risk described
in the previous section. Two models were estimated.
One model relates the average rate paid on CDs to
balance sheet measures of credit risk, interest rate
risk, and financial leverage. The second model
relates CD rates to the adjusted variance in the
bank's stock returns over the preceeding 12 months
(which should reflect both interest rate risk and
credit risk) as well as financial leverage.
The results of this analysis are reported in Table 2.
The first column of Table 2 contains the results of an
OLS regression relating the rate paid on CDs to
Treasury bill rates, the average maturity of the
bank's CDs and balance sheet measures of risk. The
second column presents the results of an OLS
regression in which the adjusted variance in the

monthly return on the bank's common stock is used
as proxy for asset risk.
The results in Table 2 are generally consistent
with the hypothesis that CD rates reflect a default
risk premium. With both models, a positive and
statistically significant relation is found between the
interest cost on CDs and the ratio of assets to total
capital of the holding company. Moreover, the
coefficients on the loan loss variable and on Short
(which measures interest rate risk) are positive and
statistically significant. This result is consistent
with the view that CD rates reflect both the credit
risk and interest rate risk of the issuing bank. In the
second column, the. coefficient on the adjusted
variance in monthly stock returns is positive and
statistically significant. 18
To investigate whether "off-balance sheet"
activities affect the risk premium on large CDs, the

regressions reported in Table 2 were re·estimated
with two additional independent variables: (l) the
ratio of SLCs outstanding to total market value of
capital and (2) the ratio of loan sales to total market
value of capital. If the volume of SLCs outstanding
or loans sales relative to total capital were to
increase the risk borne by uninsured depositors (and
the FDIC), a positive relation would be expected
between CD rates and SLCs outstanding as well as
loan sales. No significant relation would be
expected under the underinvestment or regulatory
tax hypothesis.
The results of this analysis are reported in Table 3.
No statistically significant relation is found between
the rate paid on CDs and either SLCs outstanding or

loan sales during the quarter. Moreover, using an
F-test, one cannot reject at the .10 level the hypothesis that the coefficients on SLCs and loan sales are
jointly equal to zero in either model. The results
presented in Table 3 are therefore inconsistent with
the moral hazard hypothesis that SLCs and loan
sales increase bank risk.

Summary and Conclusion
The growth of loan sales and SLCs in recent years
has raised concerns over the effect of these offbalance sheet activities on bank risk. How these
activities affect bank risk depends on the reasons
banks undertake them.

In this paper, I show that one motive for selling
loans and issuing SLCs is that they permit banks to
make relatively low-risk loans that would be
unprofitable to finance with deposits. This suggests
that off-balance sheet activities are not motivated
solely by the incentives created by deposit insurance
to increase leverage or asset risk through "off bal-

ance" sheet activities.
The empirical evidence from the CD market. is
consistent with this conclusion. Specifically, loan
sales •and SLCs do not appear to be important
determinants of bank risk as perceived and priced by
large uninsured depositors.

FOOTNOTES
1, Loan sales structured as participations differ from what
(1986) show that this result is quite sensitive to the assumphas traditionally been referred to as a participation in the
tions concerning forebearance (that IS, closure rules) of the
banking literature, The older form of participation is better
FDIC,
described as a syndication or assignment, and involves a
8, See Leonard Sloane, "New Securities Tied to Assets",
lead bank negotiating for each bank in the syndicate,
New York Times, July 20, 1985, and Lowell Bryan, "The
Selling of American Loans", Wall Street Journal, October
However, each of the banks in the syndicate make a
20,1986,
separate loan to the borrowing firm, Recent loan sales,
9, Money center banks hold less capital relative to assets
structured as participations, involve the creation of a new
than most regional or smaller banks, Therefore, it is unclear
contract betwe,en the bank and the purchaser of the loan,
The purchaser s co~tract IS With the originating bank and
why the cost of financing the same loan should be lower for
not With the bank s loan customer. See Gorton and
regional and small banks who are the primary purchasers
Haubrlck (1987) for a diSCUSSion of the contractual
of loans, An explanation for this pattern is provided below
when the motives for collateralization are discussed,
aspects of loan participations,
2, See Melvin (1986) for a description of the regulatory
10 Another explanation for the use of SLCs and loan sales
treatment of loan sales, If a depOSitory Institution sells a
wiih recourse is provided by Benveniste and Berger
loan and agrees to be responSible for 75 percent or less of
(1986), They show that the ability to issue securitized
claims can improve the allocation of risk-sharing among a
the losses from the loan, then under present regulations,
the proceeds from the sale are not reservable, "
bank's debtholders and depositors, In particular, securitiz3, Information on loan sales With recourse IS difficult to
ation provides some bank claimants a senior claim to
obtain, Available evidence suggests that loans sold With
certain assets, If investors were to vary in their degree of
full recourse are rare, The 1985 LPS Survey indicated 13
risk-aversion, securitization may result in a lower cost of
percent of loans sold had a put option, allowing the purfunds by providing richer risk-sharing opportunities, Benchaser to sell back the loan, In addition, loans sold With full
veniste and Berger's model is based on an assumption that
investors' risk-sharing opportunities outside the bank are
recourse were reported to be only $11 million while nonrecourse loan sales totaled $26 billion,
limited, Their model does, however, yield implications
4, Under the Federal Reserve Board's guidelines,
similar to the model developed in this paper.
11, A similar argument can be made for banks with uninreleased for comment February 12, 1986, capital requiresured deposits and subordinated debt outstanding, Morements for SLCs would vary depending on the account
party and use of the SLC, For SLCs backing commercial
over, the largest issuers of SLCs and sellers of loans are
paper or loans to nongovernment entities, the capital
money center and large regional banks with the largest
proportion of uninsured (or partially insured) deposits,
requirements proposed are identical to those that apply to
"booked" loans and loans sold With recourse,
12, See James (1987) for a formal proof of this proposition,
5, The payoff characteristics of collateralized debt as w~1I
13, Selling loans or issuing collateralized debt can of
as the effect of collateralized debt Issues on a bank s
course increase the risk of deposits, This will occurwhen
investment poliCY are derived formally In James (1987),
existing assets are sold or collateralized (that is, when
6, Under Federal law and regulation (12 USC 90 and 12
cash flows to depositors are reduced) or when new loans
CRF 7,7410) national banks may pledge assets against
are sold when the activity would have been profitable with
public deposits, However, in 1934, the U,S, Supreme Court
deposit-financing,
14 Another problem with using the average interest cost
ruled that national banks may not pledge assets against
private depOSits, (Texas and Pacific Ry Co, vs, Portorff,
of 'large CDs calculated from the Call Report data as a
291, U,S, 245,1934),
, ,
, p r o x y for the rate paid on newly issued CDs is that the
7, The regulatory tax hypotheSIS IS based on the propOSIaverage interest cost reflects rates paid on CDs issued in
tion that depOSit Insurance IS overpriced (when regulatory
previous quarters as well as newly issued CDs, To determine whether the average interest cost of CDs is a reasontaxes are included), The results of empirical studies on the
under- or overpricing of deposit guarantees are mixed, For
able proxy for the rate offered on CDs in a given quarter, I
obtained the Innerline survey of rates paid on newly issued
example, Marcus and Shaked (1984) find that for large
banks, guarantees are overpriced, Ronn and Verma

CDs by 300 banks for the first quarter of 1985. Thirty-nine of
the banks in my sample reported rates to Innerline. The
average rate reported by these banks in the Innerline
survey was 9.28 percent. The average interest cost of CDs
from the Call Report is 9.12 percent for that quarter. The
difference in rates is not significantly different from zero.
Moreover, the correlation between the two series is .60.

16. This calculation is based on a simplifying assumption
thatthe variance of the return on debt is zero.
17. A larger value of Short implies a greater maturity
mismatch between bank assets and liabilities. The absolute value of the difference between short term assets and
liabilities is used to account for the fact that earnings
variability induced from interest rate changes can arise
through either short-term assets exceeding short-term liabilities or the converse. Reporting requirements necessitated using a one-year dividing line between Short and
long-term assets. See Flannery and James (1984) for a
description of how Short is constructed.

15, The dollar volume of time deposits of $1 00,000 or more
is reported for six maturity categories: one day, 3 months or
less, over 3 months to 6 months, over 6 months to 12
months, over 1 year to 5 years, and over 5 years. The
weighted average maturity is calculated in months, with
deposits with a maturity of 3 months or less assigned a
maturity of one month. For the remaining categories, the
maturity of CDs is assumed to be the longest maturity in
that category. For deposits over 5 years, a maturity of 60
months was assigned.

18. The coefficient on the T-bill rate variable of less than
one may appear puzzling. However, note that for this
sample Qf banks the average maturity of CDs outstanding
exceeds 90 days (see Table 1). The results in Table 2 may
reflect the fact that short-term rates (that is, 90 T-bill rates)
are less vQlatile than long-term rates.

REFERENCES
Baer, Herbert and Elijan Brewer. "Uninsured Deposits as a
Source of Market Discipline: Some New Evidence,"
Federal Reserve Bank of Chicago Economic Perspectives, 1986, 10,23-31.
Barnea, Amir, Robert Haugen and Lemma Senbet. "An
Equilibrium Analysis of Debt Financing Under Costly
Tax Arbitrage and Agency Problems," Journal of
Finance, 1981,36.
Bennett, Barbara. "Off Balance Sheet Risk in Banking: The
Case of Standby Letters of Credit," Economic Review,
Federal Reserve Bank of San Francisco, NO.1.
Benveniste, Lawrence and Allen N. Berger. "Standby Letters of Credit: Benefits of Financing Loans Off a Bank's
Balance Sheet," Research Papers in Banking and
Financial Economics, Board of Governors of the
Federal Reserve System, 1986.
Flannery, Mark and Christopher James. "The Effect of
Interest Rate Changes on the Common Stock Returns
of Financial Institutions," Journal of Finance, 1984,39.
Goldberg, Michael and Peter Lloyd-Davis. "Standby Letters of Credit: Are Banks Over Extending Themselves?" Journal of Bank Research, 1985, 16,
Gorton, Gary and Joseph G. Haubrich. "The Loan Sales,
Recourse and Reputation: An Analysis of Secondary
Loan Participations," Working Paper, University of
Pennsylvania, 1987.
Gurel, Eitan and David Pyle. "Bank Income Taxes and
Interest Rate Risk Management: A Note," Journal of
Finance, 1984,39.
Hannan, Timothy and Gerald Hanweck. "Bank Insolvency
Risk and the Market for Large Certificates of Deposit,"
Journal of Money Credit and Banking, forthcoming.
Hirtle, Beverly. "The Growth of the Financial Guarantee
Market," Quarterly Review, Federal Reserve Bank of
New York, 1987, 12.
James, Christopher. "An Analysis of the Use of Loan Sales
and Standby Letters of Credit by Commercial Banks,"
mimeo, 1987.

Koppenhaver, Gary. "Standby Letters of Credit," Economic Perspectives, Federal Reserve Bank of Chicago, 1987, 11.
Marcus, Alex and Israel Shaked. "The Valuation of FDIC
Deposit Insurance Using Option Pricing Estimates,"
Journal of Money Credit and Banking, 1984, 16.
Melvin, Donald. "A Primer for RMA Staff on Legal and
Regulatory Concepts and Standards in the Securitization of Loans." Robert Morris Associates: Philadelphia, PA, 1986.
Meyers, Stewart. "Determinants of Corporate Borrowing,"
Journal of Financial Economies, 1987,5.
Miller, Merton. "Debt and Taxes," Journal of Finance,
1977,32.
Modigliani, Franco and Merion Miller. "The Cost of Capital,
Corporation Finance and the Theory of Investment,"
American Economic Review," 1958, 48.
Pavel, Christine. "Securitization," Economic Perspectives,
Federal Reserve Bank of Chicago, 1986, 10.
Pavel, Christine and David Phillis. "Why Commercial
Banks Sell Loans; An Empirical Analysis," Economic
Perspectives, Federal Reserve Bank of Chicago,
1987,11.
Pennacchi, George. "Loan Sales and the Cost of Bank
Capital," mimeo, 1987.
Ronn, Ehud and Avinash Verma. "Pricing Risk-Adjusted
Deposit Insurance: An Option-Based Model," Journal
of Finance, 1986, 41.
Pyle, David. "Discussion of Off Balance Sheet Banking" in
The Search for Financial Stability: The Past Fifty Years,
Federal Reserve Bank of San Francisco, 1985.
Salem, GeQrge. "Selling Commercial Loans: A Significant
New Activity for Money Center Banks," Journal of
Comme.rcial Bank Lending, 1986.
_ _. "Loan Selling: A Growing Revolution That Can
Affect Your Bank," Journal of Commercial Bank Lending, 1987.
Stulz, Rene and Herb Johnson. "An Analysis of Secured
Deb!," Journal of Financial Economics, 1985, 14.

Randall Johnston Pozdena*
In the past decade or so, corporations have been using debt increasingly
to finance their activities. This is manifested both in a generally rising
trend in corporate leverage and in the growing use oflow-grade ("junk")
bond financing. This article discusses the theory of the choice of corporate .financial structure and the role that tax policy plays in that choice.
The findings suggest that tax policy has contributed importantly to the
observed trends and that recent changes in federal tax policy make it
likely that the preference for debt financing will continue.
edness occur, certain conventional avenues of
finance operate inefficiently "closing off" access to
financial capital and depressing economic activity.
This argument has been employed, in fact, to argue
that the Great Depression was a phenomenon of a
credit rather than monetary system failure. 4
The purpose of this article is to explore the
reasons behind the rise in the use of debt by U.S.
nonfinancial corporations. In particular, theory suggests that personal and corporate income tax policy
influence the corporate use of debt. Using historical
data on corporate leverage, debt and equity issuance
activity, and federal tax policy, we find that changes
in tax policy are indeed related to the changes
observed in corporate financial policy over the last
century in a manner generally consistent with theory. Moreover, a review of the major features of the
1986 Tax Act reveals that it is an unusually strong
potential source of stimulus to the corporate use of
debt.
The remainder of the paper is structured as follows. First, the theory of corporate financial structure is reviewed briefly. Then, in Section II, a study
designed to detect the influence of tax policy on
financial structure is presented. In Section III, the
data employed to test these notions is introduced
and the empirical findings summarized. The paper
concludes with a summary of the findings and a
discussion of policy implications, focussing on
recent Federal tax reforms.

Corporations in the United States appear to be
financing their activities increasingly through the
use of debt (bonds, loans and other liabilities) rather
than equity (corporate stock). Indeed, available data
suggest that the ratio of corporate debt to equity
outstanding has increased by two-and-one-half
times since the 1960s. I In addition, the issuance of
low-grade debt obligations by corporations ("junk"
debt) has increased significantly in recent years. The
quarterly issuance of corporate bonds with below
investment-grade ratings has climbed from less than
$1 billion in 1982 to over $32 billion in 1986. 2
The growing use of debt financing by corporations may have a number of important implications.
First, everything else being equal, highly leveraged
finance makes the profitability and solvency of
individual corporations more susceptible to fluctuations in income. Some observers have expressed
concern for the welfare of investors in corporate
bonds should the corporations involved suffer unexpectedly low earnings. Second, a widespread
decline in corporate earnings might also be a
destabilizing force for the financial system as a
whole. The argument, made most cogently by Bernanke 3 , is that when widespread defaults on indebt-

* Assistant Vice President, Federal Reserve Bank of
San Francisco. The author wishes to thank William
Robertson for his valuable and diligent research
assistance during the course of this study.
37

Determinants of Corporate Financial Structure
the discount rate applied in computing the present
value of the flows) are the same and constant for
investors and the firm.

Corporations finance their activities in two basic
ways. First, they issue debt in the form of bonds,
notes, and other primary securities, and take on
liabilities in the form of loans from individuals and
financial institutions. Bonds and other primary
securities of the corporation may be sold into active,
organized markets or placed directly with the ultimate investor. The second way they obtain funds is
through the sale of shares in the equity of the
corporation and the retention of earnings.
The theory of corporate financial structure - that
is, the mixture of debt and equity finance - has
been a subject of interest to finance economists for
many years. Conventional theory of the firm
assumes that the goal of management is to maximize
the present value of the profits of the corporation,
which is tantamount to maximization of the value of
the firm's shareholder equity. It is assumed that this
same goal motivates the selection of corporate
financial structure.

Effects of Taxes
Although there may not be an optimal debt-equity
ratio for an individual firm in a perfect market, such
may not be the case in a world in which taxes,
transactions costs, and other "imperfections" exist.
Moreover, financial structure for the corporate sector in the aggregate may be influenced by these
imperfections even if individual firm conduct is not.
Tax'policy is a particularly likely source of influence
on financial structure since there are important differences in the treatment of debt and equity
securities in the U.S. tax structure.

Corporate Tax Treatment of Interest and Dividends
Let us tum first to the treatment of debt and equity
under the corporate income tax system. Corporations have been subject to an income tax in the
United States since 1908. The tax is paid on income
net of deductible expenses. The interest paid by a
corporation on its debt is one such deductible
expense; the dividends it pays to equityholders are
not deductible. Thus, everything else being equal,
the income of a firm financed by debt is partly
"shielded" from taxation whereas that of an allequity firm is not.
The present value of the tax-saving represented
by this shield is a potential source of additional
value to the firm. The tax-saving varies in direct
proportion to the corporate tax rate and the size of
the debt shield. For perpetual debt, the value of the
shield is teD where te is the corporate tax rate and D
is the amount of debt outstanding. 7 Thus, the value
of the shield rises with the tax rate and increased use
of debt. Other influences aside, therefore, corporate
tax policy biases financial structure toward an alldebt configuration.

The Notion of Irrelevance
In 1958, Franco Modigliani and Merton Miller
first put forth the notion that, in fact, the value of the
firm may be independent of its financial structure at least in simple financial environments. This
notion, known as the Modigliani-Miller (MM) or
"irrelevance" theorem, can be motivated in a number of ways. The simplest is the argument that the
value of the firm is determined fundamentally by the
firm's assets and the cash flow generated by those
assets. Partitioning those cash flows into payments
to equityholders versus payments to debtholders
does not have any obvious influence on the present
value of the cash flows and, hence, does not change
the value of the firm. s Put differently, the value of
the firm is independent of its financial structure, and
there is no optimal debt-equity ratio for an individual firm.
This fundamental point also can be demonstrated
formally using the capital-asset pricing model
(CAPM).6 The key assumptions, however, are that
the value of the partitioned cash flow is the same to
each type of investor (investors in debt or investors
in equity) and that the costs of borrowing (and hence

Personal Taxation ofDebt and Equity Income
Corporations make financing decisions not on
their own behalf, but on behalf of investors. Seen

from this perspective, the objective of corporate
finance decisions is to maximize the present value of
the income of its investors after all taxes. If these
investors face a personal income tax, the earlier
conclusion that corporate taxes bias financial structure in favor of debt may not be unambiguous.
Personal income in fact has been taxed in the
United States since 1913. In addition, for most of
that time, income from equity has been taxed differently from so-called "ordinary income", including interest income. Income from equity is in the
form of dividend payments and capital gains that
accrue to equityholders as the result of earnings
retention. Although dividends historically have
been taxed at the same rate as ordinary income, the
accrual of earnings in the form of capital gains is
treated favorably. In particular, the gains typically
are not taxed until they are realized (that is, until the
appreciated equity is sold), and realized capital
gains typically have been taxed at a favorable rate.
The fact that the firm can elect to retain, ratherthan
distribute, its net earnings, and that the tax obligation on the resultant capital gains is delayed, is one
source of the preferential treatment afforded equity
income. In addition, even realized capital gains
have been taxed at a rate lower than ordinary
income. s
If the tax rate applied to interest income is tb and
the perceived effective tax rate on equity income is
te , the combined effects of corporate and personal
income taxation policy can be described succinctly.
A dollar of corporate income paid out as interest
expense avoids corporate income taxation but is
taxed at the personal level, yielding an after-tax
income of (1 - tb ) dollars for the investor. If paid out
as equity income, in contrast, a dollar of corporate
income implicitly must bear both the corporate and
personal tax burdens, yielding (l - tc)(l - te ) dollars to the investor.
Debt finance will be preferred, therefore, when it
offers after-tax income to the investor that exceeds
that offered by equity income, that is, if
(l - tb »(1- tc)(l

same in both cases. Note that preference for debt
financing increases with (a) lower tax rates on
normal income, (b) higher tax rates on corporate
income, and (c) higher tax rates on equity income.
Note also that if the personal tax rate faced by debt
and equity income were the same (that is, tb equaled
te ), debt finance unambiguously would be preferred.
In essence, the effect of taxes is to make the value
of the firm dependent upon how the cash flow from
the firm is partitioned. That is, the conventional
theory's assumption that the values of these flows
are the same regardless of the way in which they are
partitioned proves not to be false when tax policy is
considered.
The Determinacy ofAggregate Leverage
This simplified view of corporate capital structure
does not address important practical issues about
corporate debt policy. Perhaps most important of
these issues is the model's implication of an "eitheror" nature to the corporate debt decision. That is,
depending upon the tax structure, the model implies
that investors are likely to prefer either an all-debt or
all-equity structure. In reality, a mixture of debt and
equity is observed both in the aggregate and among
individual firms in our economy.
A number of explanations have been offered for
the observation of a determinate amount of debt and
equity in the aggregate. Two are particularly relevant to the analysis pursued in this paper. The first is
the notion offered by Miller that debt and equity
"clienteles" exist because of differences in the tax
rates faced by individuals in the economy. That is,
investors facing low personal marginal tax rates will
tend to prefer debt, and those with sufficiently high
tax rates will prefer equity. The relative wealth of
these different "clienteles" thus makes the aggregate balance of debt and equity observed in the
economy determinate. Individual firms, however,
remain indifferent on the margin between the two
avenues of finance because the equilibrium prices of
debt and equity must satisfy all of these clienteles or
the clienteles would continue to shift. 9
A second explanation recognizes that a progressive income tax structure creates incentives to
exchange corporate securities to achieve maximal
after-tax income for investors. Individuals in high

te )·

Equity finance will be preferred when the opposite
is true. Investors will be indifferent between the two
modes of finance when their after-tax income is the

tax brackets, for example, will tend to be willing to
trade debt holdings for equity holdings to obtain the
preferred tax treatment afforded income from the
latter. This process of trading corporate securities
affects both the equilibrium prices of debt and
equity, and the equilibrium effective marginal tax
rate. Indeed, it can be shown that in equilibrium,
effective marginal tax rates will be equilibrated
across households. 10 In such an instance there are
no clienteles as such and firms are indifferent
between debt and equity financing, but the aggregate amount of debt and equity in the economy
remains determinate.

both the leverage of the individual firm and firms in
the aggregate determinate.
The treatment of non-debt related shields in the
U.S. Tax Code also can influence leverage. In
particular, depletion and depreciation allowances
are long-lived deductions whose value as a deduction for tax purposes is fixed at the time of the
relevant investment. If the corporation subsequently
is exposed to general price inflation (including
inflation in the price of its own product) and its
nominal income rises, the effectiveness of the
depletion or depreciation shield implicitly declines.
Thus, by inference, inflation can have the effect of
increasing the attractiveness of additional debt
shields, everything else being equal, and thereby
increase the degree of leverage observed in the
corporate sector.
Finally, restrictions on the ability of households
to borrow also may influence observed levels of
corporate leverage. This point relates to the notion
that exchange of corporate securities is a strategy to
achieve maximal after-tax income for households
holding such securities. The strategy may require,
however, that certain households borrow (issue personal debt) to acquire corporate equity.
If there were limits on short-selling and deductibility of interest expenses, or other limitations on
borrowing by households, then the use of a corporation as a "tax intermediary" would become more
important and may make corporate financial policy
relevant. Say, for example, that interest expenses are
deductible by corporations but not by households. If
households were able to purchase shares in a highly
leveraged corporation, they could possibly sidestep
such restrictions on personal leverage and provide
an incentive for an increased corporate use of debt.
(In effect, the corporate securities would be used to
arbitrage the differences in personal and corporate
tax treatment of debt.) Models of the use of corporate securities in such "tax minimization" strategies
show that household borrowing restrictions can
influence corporate leverage. 12 This finding is relevant because, as we shall see in Section IV, recent
tax reform limits personal borrowing.
In the long literature on optimal corporate financial structure, many other factors have been discussed as possible influences on the observed finan-

The Determinancy of Individual Firm Leverage
The observed variation in debt and equity held by
individual firms also has a number of explanations.
Miller has argued that since individual firms' financial structure is irrelevant, it is not costly for firms to
pursue what they feel is a value-maximizing financial structure. This argument implies that the
observed variation is serendipitous.
A second explanation recognizes the fact that
firms enjoy shields against income taxation generated by sources other than debt. These "non-debt
shields" include such things as depreciation and
depletion allowances and the investment tax credit.
If these shields were large enough relative to the
income of the firm, the interest deduction could be
completely redundant as a tax shield, and, in effect,
make the marginal corporate tax rate (tc ) zero,
reducing the incentive to prefer debt-financing.
Even if the non-debt shields were not a complete
offset to income, however, there is some probability
that a debt shield will be redundant in a stochastic
income environment.
Thus, the existence of non-debt shields reduces
the expected shield benefit of additional debt. That
is, the contribution to the value of the firm of an
additional unit of debt is not constant, but declines
with expanded debt usage because it increases the
probability that the debt shield will be redundant for
any given amount of non-debt shielding.
DeAngelo and Masulis have argued that, in the
presence of these non-debt-related shields, individual firms can lose their indifference to financial
structure. 11 The loss of indifference would make

cial structure of corporations. Factors such as
bankruptcy costs 13 , differences between the preferences of managers ("inside equity holders") and
other equity holders l4 , and the influence of intangible assets, all have received some attention in the
literature as sources of determinacy in the amount of
leverage observed in the corporate sector.
While it may be likely that these and other nontax considerations play some role in determining
corporate financial structure, their influence is difficult to study empirically as these aspects of the
economic environment are difficult to quantify. 15
The empirical work presented in this paper focuses,
therefore, on the influence of tax policy on corporate
financial structure. As we shall see, a significant
fraction of the variation in observed corporate leverage over time appears to be associated with changes
in tax policy.

assessment of new (and perhaps existing) debt.
Tax policy also may directly influence the preference for debt of low quality, however. A bond that is
risky will contain a compensatory premium in its
yield. The higher interest payments associated with
risky debt implicitly provide a larger tax shield (for
a given amount of debt) than less risky, lower yield
debt. The higher default probability of the risky
debt, of course, means that its tax shield effects have
a higher probability of going unused. Zechner and
Swoboda have argued, however, that because of
peculiarities in the way in which tax law treats the
obligations of corporations in a bankrupt state, the
present value of the implied tax shield effects can
nonetheless be greater for riskier debt. 16
Another, possibly offsetting influence of tax policy on risky debt is the probability of a high-risk
bond becoming a low-risk bond (as the corporation
that issued the obligations evolves into a corporation
with strong earnings and a growing net worth). Such
an event would, in effect, confer a capital gain on
the holders of the (formerly) risky debt that is
treated favorably for tax purposes. Everything else
being equal, had the corporation issued high-quality
debt to begin with, there would be no prospect for
such gains. (In essence, low-grade debt has some
"equity-like" characteristics.) This suggests that,
unlike debt in general, the issuance of high risk debt
may be retarded by increases in the capital gains
rate. I?

Taxes, leverage and Debt Quality
We tum now to the association between tax policy
and the quality of corporate debt. Specifically, if a
change in tax policy stimulates an increase in a
corporation's leverage, it likely will result in the
deterioration of the quality of its debt on the margin.
One direct reason for this effect is that increased
leverage simply reduces the capital buffer against
default and thereby reduces the risk of a corporation's default on its debt. Debt-rating agencies and
the marketplace in general would respond to
increased default risk by downgrading the quality

II.

Corporate Financial Structure and Tax Policy:
Measurement Issues
much of that variation may be inherently endogenous. I8 In addition, using tests on aggregate time
series data to determine the influence of taxes on
leverage offers the opportunity to discover that
influence whether it operates at the firm level or only
at the level of the corporate economy overall.

In this study, we examine empirically the influence of tax policy on the financial structure of
corporations. We use aggregate data on the tax
treatment of corporations and corporations' financing behavior over time. This longitudinal approach
has a number of advantages over a study design that
relies on examining the behavior of firms in the
cross-section. For example, the considerable variation in tax policy and corporate financing behavior
over time permits forging a statistical association
between the two. There is less variation in the tax
treatment across firms at a given point in time, and

Measuring Corporate Financial Structure
A major empirical issue in our analysis concerns
the measurement of corporate financial structure.
The theory discussed earlier suggests that the relative stocks of debt and equity outstanding in the

increase the desired degree of leverage to

economy as a whole (and, perhaps, for individual
firms) may be influenced by features of the tax
system and other variables. Accurate measurement
of these stocks (to create a leverage measurement or
some other summary statistic) requires estimates of
the market values of the debt and equity of all firms.
Unfortunately, a long time-series of such data is
not available for the corporate sector as a whole and
is difficult to construct from generalized indices.
For equities, value-weighted indices of share prices
such as the Standard and Poor's 400 and 500 exist,
but their coverage is limited and has changed over
time, and the indices themselves have been
"rebased" at various times. 19 Also, these indices
cover only companies with traded equity which,
arguably, may behave differently from other corporations. An even more serious problem exists for
'measurement of the market value of corporate debt
in the aggregate, since no single value-weighted
index exists. The result is that market value debt and
equity estimates constructed from some indices are
of questionable value to empirical work.
Book-value measures of total corporate assets
and total corporate liabilities, in contrast, are available in a reasonably consistent form. 20 They have
been reported to the Internal Revenue Service (for
firms with and without tax liability) for about 50 of
the 80 years that corporate income has been subject
to tax in the United States. While not the ideal
measures, they may nevertheless approximate market measures reasonably closely in the aggregate if
corporate asset and liability portfolios tum over
sufficiently rapidly.
The market values of net issuance of corporate
debt and equity also are observable. Net issuance is
the market value of new gross securities issued
minus the value of retired securities. While net
issuance activity is not ideal data for examining the
leverage process directly, it can offer some assistance. Specifically, if a change in tax policy were
likely to induce additional corporate leverage, relatively more debt than equity should be observed to
be issued. For example, suppose that the ratio of
debt (D) to equity (E) initially is
L

L' = (D

dD)/(E

dE)

> L,

where dD and dE are, respectively, the net issuance
of debt and the net issuance of equity. If L exceeds
L,
I

(dDID) > (dElE).
That is, the percentage change in outstanding debt
must exceed the percentage change in outstanding
equity. Computing the percentage change accurately would require accurate measures of the stocks
of debt and equity. However, as long as D is less than
or equal to E (as it is for the aggregate of all U.S.
nonfinancial corporations), net issuance of debt in
excess of net issuance of equity will be associated
with an increase in leverage.
In the empirical work in this paper, leverage
measures and issuance activity are both employed to
test the relationship between tax policy and corporate financial structure.

Measuring Tax Policy
Measurement of the tax policy environment also
raises conceptual and practical issues. Miller's
notion of debt and equity "clienteles" implicitly
suggests that the degree of leverage observed in the
economy as a whole will depend upon the wealth of
groups in various tax brackets. This, in tum, suggests that wealth-weighted relationships between
personal and corporate taxes might be an appropriate measure of ambient tax policy. The view of
corporate securities as devices to arbitrage such tax
differences, however, argues that such clienteles do
not exist in equilibrium.
In either case, the outcome will be driven by the
clientele with the highest individual wealth who, in
tum, might be assumed to face the highest ex ante
marginal tax rates on ordinary personal income. It is
this rate that is used in our study, and it ranges from
7 percent (in 1913, the first year that personal
income was subject to taxation) to a peak of 94
percent in 1944 and 1945.
The measurement of the tax rates applicable to
income from equity also poses conceptual and
empirical problems. Income from equity takes the
form of dividends and capital gains. Given the

DIE

and that the reaction to a change in tax policy is to

nondeductibility of dividend payments from gross
income at the corporate level, it is something of a
conundrum to financial economists that corpora
tions pay dividends at all; it would appear preferable
in all cases for firms to retain earnings and convert
current income to capital gains for its security
holders. 21 In addition, since the timing of the reali
zation of capital gains can be controlled by the
investor in most cases, the argument has been made
that investor behavior will result in effective avoid
ance of the tax and thus that the effective capital
gains rate is zero. 2 2
This debate will not be resolved here, but it seems
reasonable to assume that some differential treat
ment of income from holding corporate equity
occurs and certainly that the ex ante rate of taxation
of capital gains (which is what will influence
security holding behavior) is nonzero and differs
from the rate applied to ordinary income. Once
again, we will measure changes in the taxation of
capital gains using the highest statutory rate. Since
no distinction was made between equity income and
ordinary income until 1922, these two statutory
rates correspond for the first 9 years of taxation of
personal income.
Fewer problems exist in defining and measuring
the corporate tax rate. Over most of its history, the
corporate income tax in the United States has been a
simple proportional tax. That is, a single tax rate,
with exceptions to that rate only for very small
corporations, has been employed. In the analyses
that follow, therefore, the corporate tax rate has been
measured as the primary (maximum) statutory rate
on corporate income. The taxation of corporate
income began in 1908 and the primary rate has
ranged from 1 percent in that year to a peak of 52.8
percent in 1960.
The other feature of tax policy examined in this

IIL

Chart 1
Tax Rates and Shield
Percent
(Tax Rates)

Percent
(Tax Shield)

Note: Tax rates are maximum rates

research is the influence of nondebt-related tax
shields. Depletion and depreciation allowances and
the investment tax credit are the major nondebt
sources of shields to net income. Unfortunately, it is
not possible to measure these features of tax policy
using a single parameter, making them difficult to
characterize ex ante in a consistent empirical man
ner. In the analyses that follow, a measure of the
actual use of these shields is used in lieu of a policy
parameter. Specifically, the ratio of nondebt-related
deductions to total deductions actually claimed by
nonfinancial corporations is employed. This ratio
can be interpreted as a measure of the likelihood that
interest deductions would be redundant. In contrast
to the other tax parameters examined, therefore, the
nondebt-related tax shield is measured using real
ized (or ex post) data. 2 3
Chart 1 presents the tax rate and shield values
employed in this study and displays the consider
able variation exhibited by these policy parameters
over the last century.

Data Description and Econometric Evidence
policy parameters and its association with corporate
financial behavior can be studied. Financial corpo
rations are excluded from the study on the grounds
that special regulatory factors likely influence their
behavior and would confound the effects of tax
policy.

This section contains simple econometric evi
dence of the relationship between tax policy and ( 1 )
aggregate corporate leverage, (2 ) the aggregate net
issuance of corporate debt and equity in the econ
omy, and (3) the gross issuance of low quality debt.
The study employs data, where possible, from 1900
to the present so that the maximum variation in tax
43

eters, measured in level terms. Two formulations
were made with the first employing the individual
tax parameters entered directly. The second uses the
relative size of certain tax rates to others rather than
the tax rates themselves. This procedure represents a
simple attempt to recognize the notion that the
relation of the corporate tax rate to the personal tax
rate and the relation of the capital gains tax rate to
the personal tax rate may be more relevant to leveraging decisions than the individual levels of tax
rates.
In both formulations, a variety of other specifications involving both complete and incomplete sets
of the tax variables and the use of lagged as well
as contemporaneous - measures of the independent variables also were employed. 25 The results of
these complex variants are not reported here
because the coefficients on the tax rate variables (the
corporate, personal, and capital gains tax rates)
appear quite insensitive to the model specification.
The parameters of the two basic regression formulations are presented in Table 1. In both cases,
the signs on the tax parameters are those expected
from the earlier theoretical discussion. Leverage
appears to be positively associated with the corporate tax rate, the capital gains tax rate, and increases
in the inflation rate; it is negatively related to the
personal tax rate and the prevalence of use of
nondebt-related tax shields. Consequently, as is
indicated in the second regression, leverage is
positively associated with increases in the difference between the corporate tax rate and the
personal tax rate (the "tax differential") and with
the difference between the maximum capital gains
tax rate and the tax rate on ordinary personal
income.
The Durbin-Watson statistics for both regressions
suggest a moderate degree of correlation among the
residuals of the regression and, hence, the possibility of imprecision in the estimates of the standarderrors of the coefficient. Correcting this problemwith simple techniques yields essentially
similar results. 26 The consistency of the signs with
that suggested by theory and the relative robustness
of the finding with respect to specification of the
regression is encouraging. The few tax variables
(and the inflation rate variable) alone explain up to

Tax Policy and Aggregate
Corporate Leverage
The theoretical discussion above suggested that
leverage may be positively associated with higher
corporate tax rates, higher tax rates on equity
income relative to ordinary income, lower personal
tax rates, lower non-debt related shields, except in
the instance that tax rates on ordinary and equity
income were identical, in which case a pure preference for debt would be exhibited regardless of the
level of tax rates. 24
Chart 2 presents a measure of leverage derived
from the book value of total liabilities and total
assets reported to the Internal Revenue Service and
its predecessor agencies. Only data for manufacturing corporations is represented to extend the data
series back in time as far as possible while keeping
consistent measures. A simple tax differential (the
corporate tax rate minus the personal tax rate) also is
presented in Chart 2. From the discussion above,
leverage should be positively associated with this
differential. From Chart. 2, it is apparent that the
association is, indeed, seemingly positive, and linear.
We examined the statistical association between
leverage and the tax differential and other representations of the tax parameters using ordinary least
squares regression techniques to create a linear
representation of the relationship between contemporaneous measures of leverage and the tax param-

Chart 2
Corporate Leverage and Tax Policy
Percent

Tax Differential

85 percent of the observed variation in aggregate
leverage in the manufacturing sector over the last 50
years.

Chart 3 presents one measure of the excess of debt
issuance over equity issuance along with the simple
tax differential variable. Although there is considerable volatility in the measure of debt minus equity
issuance, the pattern of corporate security issuance
seems to be positively and quite consistently related
to the tax parameters. In Table 2, regression results
thatrelate the issuance measure to the complete set
of tax parameters are presented. As might be
expected given the much greater volatility of the
issuance measure than the direct leverage measure
(and the theoretically less straightforward link
between issuance and tax policy), the empirical
findings are less consistent than those using the
leverage measure directly.
In particular, while the coefficients on the corporate and personal tax rates and the capital gains tax
rate have the anticipated signs, the sign on the

Tax Policy and the Issuance
ofDebt and Equity
The disadvantages of direct study of leverage are
apparent from Chart 2. Conceptual problems of
measurement aside, data are available consistently
only from the 1930s. We are therefore unable to test
the effects of the greater variation in tax policy that
characterized the first part of this century. For the
reasons stated earlier, however, the trend in the
excess of debtover equity issuance also may provide
information about leverage trends. In contrast to the
leverage measure, data on net debt and equity issuance are available in market value terms from the
first decade of the century to the present.

"nondebt shield" variable is the opposite of what
was expected. In addition, the sign on the inflation
variable in the second regression is inconsistent
with expectations, although both sets of coefficients
are not statistically different from zero. The essential relationships between the issuance activity and
the tax policy variables, however, have the signs
predicted by the simple model of the leverage decision presented earlier.

Chart 3
Debt vs. Equity Issuance and Tax Policy
Percent

Index

40
35
30
15
10
5

20
10
0
-10
-20
-30
-40
-50

...

Tax Policy and the Issuance
of Low-Grade Debt

-5
-10
-15

We turn now to an examination of the influence of
tax .policy on a particular type of debt: belowinvestment grade or "junk" debt. In recent years,
the issuance of debt below investment-grade debt
has increased sharply. This phenomenon usually has

- 2 0 -lnmmnmmmrmmnmnmmmrlTlT1Tl1mnmmmrmmnmrmrmmrl--1 00

..
Excess of Bond Over

...... Stock Issuance*

1900 1910 1920 1930 1940 1950 1960 1970 1980
*Normalized by GNP

-70
-80
-90

been ascribed to a variety of nontax factors. One
explanation, for example, is that recent declines in
interest rates have made investors generally mOre
reluctant to seek high-risk investments to obtain the
high yields to which they have become accustomed.
Related to this explanation is the claim that investment bankers and brokers only recently have discovered untapped investor interest in high-yield,
high-risk instruments. A second conventional
explanation is that improvements in information
technology now make it economical to evaluate
investments in smaller and high-risk firms, whose
debt typically would be of lower grade.
Combined with the growth of investment portfolios of sufficient scale to permit diversified holdings of low-rated debt, the factors cited are seen as
making the issuance of junk bonds more feasible.
Indeed, the factors may be contributing to the recent

growth inthe use oflow grade debt by U. S. corporations,except that the first argument is ad hoc and
difficultto verify empirically. The second explana"
tion,emphasizing technological change, is at variance with the history of the use of low-grade debt.
As we shall see, low-grade debt was usedextensivelyearlyin this century. Indeed,the· highest
volumes of junk debt were issued in the "lowctech"
decades of the century.
In this context,. it is interesting to examine the
influence of tax factors alone on junk debt activity.
Unfortunately, a single continuous body of data on
the outstanding volume of junk debt does not exist.
All that is available is the data on the gross flow of
new issues of debt that are below investment
grade. 27 This is a biased estimate, of course, of net
issuance of this type of debt, since neither retirements of outstanding low-grade debt nor the effects

of changes in the rating of outstanding debt issues
are incorporated in this data. Finally, there is a
potential problem in the consistency of even the
available data over time since there has been no
single source for the data over the long time period
of interest.
Although the various authors that have produced
estimates of debt issuance have attempted to employ
consistent standards and sources, there easily may
be "drift" in effective debt rating criteria over time.
Additionally, the rating of directly placed debt is
usually not available, and analysts have had to apply
proxies (such as the rating of publicly issued debt of
a corporation) in making inferences about the
quality of private placements. 28
Despite these serious difficulties, the statistical
relationship between tax policy and junk debt issuance activity displays rough correspondence with
that suggested by the theories of issuance of highrisk debt. We used the same regression models
employed in the analysis of aggregate debt issuance
with a time series on junk debt issuance assembled
from the available sources. As with the aggregate
debt series, the issuance volume is expressed relative to current gross national product as a simple
means of expressing the dimension of the activity
relative to the aggregate "size" of the economy. 29
The results of the regression analysis are presented in Table 3. Most of the signs on the tax
parameters are the same as those found earlier in the
analysis of aggregate corporate leverage relationships and the aggregate issuance of debt and equity.
There are some statistical problems with the
estimates, however. In particular, the low DurbinWatson statistics suggest that there is a strong serial
correlation among the residuals. This problem
likely follows from the exclusion of important
explanatory variables and thus is not likely to be
redressed by simple statistical treatment of the correlated residual problem. Nonetheless, in equations
2 and 4, the results of regressions employing a
Cochran-Orcutt specification are reported for the
two basic specifications of the model. Qualitatively,
the effects of the corporate tax rate and the personal
tax rate on ordinary income are unaffected by the
specification and confirm the notion that junk bond
issuance is positively related to the corporate tax
rate and negatively related to the personal tax rate.

However, the possibility remains that the equations
are misspecified in an important way.
From the results of regressions land 2, the effect
of the capital gains tax - which is potentially
theoretically ambiguous for reasons cited earlier appears to be such that an increase in this tax rate
decreases junk bond issuance. However, this finding
is not confirmed by the alternative specifications
represented by regressions 3 and 4. Similarly, the
sign on the nondebt-related shield variables in equations I and 3 is consistent with the theoretical
expectation that increases in such shields decrease
the use of debt generally and low-grade debt specifically; when the Cochran-Orcutt specification is
employed, however, the coefficient is indistinguishable from zero, although it has a positive sign.
Despite these difficulties, the general concordance of these results with those found earlier
provides at least weak confirmation of the notion
that factors that increase leverage generally also
tend to increase the use of "junk" debt. In Chart 4,
the predicted and actual junk debt issuance volume
is displayed. Changes in tax variables alone appear
clearly to be associated not only with the high
volumes of junk debt issuance early in the century,
but also the recent resurgence in low-grade debt use.

Chart 4

Junk Bond Issuance
Percent

15
10

Predicted Using
Tax Policy"

...

o
-5

-lrrmnmmiTl1'lTlTrrmmnmmrmrmrrmmnmm'l1TlTTl1llT1T11Trrmmmn
1900 1910 1920 1930 1940 1950 1960 1970 1980
Note: issuance Is par amount of offerings
relative to GNP
"From Table 3, equation 1 .

IV.

Summary and Policy Implications

The theory of corporate financial structure was
altered three decades ago by the notion that the
mixture of debt and equity used by a firm to finance
its <tssetswas iITelevant to the value of the firm. The
"irrelevance theorem", first advanced by Modigliani and Miller, stimulated a largeliteraturethat tries
to explain the apparently contradictory empirical
evidence of capital structure to firms. Several
strands ofthis literature emphasize the role of the tax
treatment()f corporations and households as the
mechanism by which the total amount of debt and
equity in the economy - if not that of individual
firms as well - becomes determinate. This paper
has attempted to test for the influence of tax policy
using long time series on various indicators of
corporate financial structure.
Specifically, trends in the aggregate leverage displayed by U.S. manufacturing corporations, the
relative volume of net debt and equity issuance, and
the volume oflow-grade debt issuance were studied.
They were examined for evidence of a relationship
to four important tax parameters: the marginal corporate tax rate, the marginal personal tax rate, the
personal tax rate applied to capital gains, and the
relative importance of nondebt related shields at the
corporate level. Although such a study faces a
number of theoretical and measurement problems,
simple regression analyses reveal essential consistency between the relationships posited by a simple
theoretical model of tax influence on corporate
structure and actual behavior of these various measures.
Everything else being equal, an increase in the
corporate marginal tax rate or the tax on capital
gains increases the use of debt generally and lowgrade (risky or "junk") debt specifically. In contrast, increases in the personal marginal tax rate, or
the availability of nondebt related shields (such as
depreciation and depletion allowances and the
investment tax credit) appear to reduce debt use by
corporations.

that tax reform legislation altered significantly the
relationship between personal and corporate •tax
rates, the tax treatment of capital gains ,and the
ayailability of nondebt related. tax. shields for the
corporation. The corporate tax rate, for exatnple,
will be 34 percent while the marginal personal tax
rate paid by the highest income households will be
only 28 percent. 30 The resiltt is that, for thefirsttime
in almost 80 years, the corporate tax rate will exceed
the personal marginal tax rate.
In addition, the tax preference afforded long-term
capital gains is to be eliminated. In terms of the
discussion above, the elimination is tantamount to
an increase in the rate at which income from capital
is taxed relative to ordinary income. The Tax Act
alters the availability of nondebt-related shields in a
significant way as well. The allowed period over
which various assets may be depreciated is shortened significantly and the investment tax credit that had been in existence in some form for most of
the post-war period - is eliminated. 31 Finally,
some restrictions have been imposed on household
borrowing through the limitation of deductibility of
consumer debt.
If the discussion and results of this paper were
correct, all of these changes bias the balance
between debt and equity toward increased use of
debt. Using the estimated coefficients from the
regression models presented earlier and the tax
parameters implied by the 1986 tax reform, significant increases in the use of debt generally and junk
debt in particular can be projected. For leverage in
manufacturing corporations, for example, this projection implies an increase in the debt-equity ratio
from the 1.3 observed in the last year for which data
are available (1982) to a ratio of 1. 9. Similarly, the
excess of debt over equity issuance is projected to
increase by 200 percent over its 1982 level, and junk
bond issuance by 150 percent over its 19851eveJ.32

Policy Implications
From a broader policy viewpoint, these developments may have undesirable implications. Higher
levels of corporate leverage make the corporate
sector more susceptible to adverse changes in their
income. Thus, an unanticipated economic downturn

Implications oUbe 1986 Tax Act
With these findings, the changes in federal tax
law made with the passage of the 1986 Tax Act take
on special importance. Among other provisions,

would have a more deleterious effect on U. s. corporations.
This result is troublesome in and of itself to
investors (including the banking sector) that hold
debtand equity shares in American corporations.
But even more serious is the prospect raised by
Bernanke that widespread loss of confidence in the
liabilities of U. S. corporations could have a
d~pressing systemic on economic activity that
exceeds the aggregate of the individual losses that
might confront firms. Others have pointed out that
precarious financial circumstances in the corporate
sector make it more difficult for a nation's central

bMktopursue a tight money policy, if that should
be desired, for fear of precipitating a recession.
The tax treatment of corporations - specifically,
the relatively high tax rates to which U. S. corporations are now exposed -long has been guided by a
concernthat corporations "pay their fair share" of
federal government revenue requirements. If the
links between tax policy and corporate leverage
discussed in this paper were realistic, and the link
between corporate leverage and economic fragility
is as important as some have suggested, then requiringcorporations to pay relatively high tax rates
could prove to be a very costly political stance.

FOOTNOTES
taxed at the same rate as ordinary income, can be converted easily into capital gains outside the firm if an investor borrows optimally to finance share ownership. See M.
Miller and M. Scholes, "Dividends and Taxes," Journal of
Financial Economics, December 1978, pp. 333-364.
Indeed, if this were not the case, it is unclear why firms
would ever pay dividends given the preferential treatment
afforded capital gains.

1. The ratio of debt to equity in manufacturing firms was
.55 in 1960 and 1'25 in 1982. The source of this data is the
Internal Revenue Service, Statistics of Income: Corporation Income Tax Returns, annually, and its predecessor
publications.
2. Data SourceVIDD Information SerVices, Inc.
3. Ben S. Bernanke, "Nonmonetary Effects of the Financial
Crisis in the Propagation of the Great Depression," American Economic Review, June 1983, pp. 257-276.

9. See, M. Miller, "Debt and Taxes," Journal of Finance,
May 1977, pp. 261-275.

4. Bernanke, ibid, and Hyman Minsky, "A Theory of Systemic Fragility," in Altman and Sarnetz, eds, Financial
Crises: Institutions and Markets in a Fragile Environment,"
New York: Wiley-International, 1977.

10. A simple graphical presentation of this point is available in V. Aivazian and J. Callen, "Miller's Irrelevance
Theorem: A Note," Journal of Finance, March 1987, pp.
169-179.

5. Another variant of the same argument is that any advantages of leverage achieved at the corporate level can be
undone by investors. For leverage to cause corporate
share values to be higher, investors must find it more costly
to achieve leverage privately (that is, by issuing their own
debt). If, in contrast, firms and households face the same
borrowing and lending opportunities, such "homemade"
leverage will be able to undo any effects of corporate
leverage. Once again, therefore, leverage at the corporate
level will be irrelevant, although the aggregate of corporate
plus household debt and equity outstanding could be
determinate. See, F. Modigliani and M. Miller, "The Cost of
Capital, Corporation Finance and the Theory of Investment," American Economic Review, June 1958, pp.
261-297.

11. H. DeAngelo and R. Masulis, "Optimal Capital Structure under Corporate and Personal Taxation, "Journal of
Financial Economics," March 1980, pp. 3-29.
12. Aivazian and Callen, op cit.
13. See, for example, N. Baxter, "Leverage, Risk of Ruin,
and the Cost of Capital," Journal of Finance, March 1967,
pp. 395-403.
14. M. Jensen and W. Meckling, "Theory of the Firm:
Managerial Behavior, Agency Costs, and Ownership
Structure," Journal of Financial Economics, October 1976,
pp.305-360.
15. The difficulties involved are typified by the study of
bankruptcy cost by Warner. See, J. Warner, "Bankruptcy
Costs: SOme Evidence," Journal of Finance, May 1977, pp.
337-348.

6. See, R. Bresley and S. Myers, Principles of Corporate
Financf'!, MCGraw-Hili, 1984, appendix to Chapter 17.
7. This follows from the fact that the periodic shield is the
corporate tax rate times the debt coupon, or tcDr, where r is
the coupon interest rate on the perpetual debt. The present
value of such a perpetual stream of shields is (tcDr)!
r
teD.

16. Zechner and Swoboda, "The Critical Tax Rate and
Capital Structure," Journal of Banking and Finance, (10)
1986, pp. 327-341.

8. Income from equity consists of appreciation of capital
shares and payment of dividends. Appreciation of capital
shares is treated favorably because the tax liability can be
postponed until the gains are realized and because these
gains typically have been taxed at a lower rate than
ordinary income. Dividend income, although nominally

18. Specifically, firms may have selected their industrial
activities or their form of organization to obtain the most
generous tax treatment. Conglomerate organization, for
example, allows use of nondebt-related shields by firms
whose other activities would not normally generate them,
for example.

17. I am indebted to Chris James for suggesting this
effect.

19. See Cohen, Zinbarg and Zeikel, Investment Analysis
and Portfolio Management, Richard D. Irwin, Inc, 1977, p.
127 for a description of the construction of the Standard
and Poor's value-weighted stock indices.

26. Both first differences and Cochran-Orcutt specifications were employed.
27. Data is available from Hickman for the period
1900-1943, Atkinson for the period 1944-1965, and Altman
and Namacher and 100 Information Services, and the
Board of Governors of the Federal Reserve System for
1970 to the present. Data permitting separation of public
and private placements is not always available, so the
regressions reported below are based on total (that is,
public and privately placed) debt. Junk debt is considered
debt issued with a Moody's rating below Baa or equiva.lfi!nt
plus unrated corporate debt. Tests were conducted on
sub-periods Of the data to explore the sensitivity of the
findings in this paper to the definition of junk debt.

20. The aggregate of corporate liabilities and assets is
reported annually by industry in Internal Revenue Services, Statistics of/ncome: Corporation Tax Returns and its
predecessor publications. See also, L. Tambini, "Financial
Policy and the Corporation Income Tax," in A. Harberger
and M. Bailey, eds., The Taxation of Income from Capital,
Brookings Institution, 1969, pp. 185-222.
To the author's knowledge, the IRS data is the only consistent source of data both on liabilities and assets of U.S.
corporations that includes the pre-war period. However,
the flow-of-funds data of the Board of Governors of the
Federal Reserve System provides quarterly estimates of
the book value of outstanding debt for the nonfinancial
corporate sector from 1952 to the present, and annual
balance sheets for nonfinancial corporations are available
in "Balance Sheets For the U.S. Economy," Board of
Governors of the Federal Reserve System, Washington,
D.C., from 1947. An attempt is made in this publication to
assign market values to debt and equity, but the estimates
likely suffer the handicaps cited in the text. Regressions
run with these measures, however, generally conform to
those presented here.

See W. Hickman, Statistical Measures of Corporate Bond
Financing Since 1900, Princeton University Press, 1960, T.
Atkinson, Trends in Corporate Bond Quality, Columbia
University Press, 1967, and E. Altman and S. Nammacher,
Investing in Junk Bonds: Inside the High- Yield Debt Market, John Wiley and Sons, 1987, for additional discussion
of the market in bonds of various quality ratings.
28. See the individual data sources cited in the previous
note for details on the treatment of privately placed and
unrated debt.
29. As was noted above, the conceptually appropriate
treatment of these flow measures would require that the
difference in the percentage changes of debt versus
equity be stUdied. In the absence of accurate measures of
outstanding stocks of debt and equity, expression of the
issuance flow data relative to the gross national product
may be justified as the basis that GNP may move in
proportion to total corporate assets.

21. At best, dividend policy is considered irrelevant by
most economists. See, for example, F. Black and M.
Scholes, "The Effects of Dividend Yield and Dividend
Policy on Common Stock Prices and Returns," Journal of
Financial Economics, May 1974, pp. 1-22.
22. M. Miller, op cit.
23. The tax shield variable is constructed as follows. Total
deductions for depletion, depreciation and interest costs
and total investment tax credits taken by all nonfinancial
corporations is reported annually in the Internal Revenue,
Statistics on Income: Corporate Income Tax Returns and
its predecessor publications for the study period. The
"deduction equivalent" of the investment tax credit (ITC) is
computed using the current primary corporate income tax
rate. A variable called "Nondebt Shield" is computed by
taking the ratio of depletion, depreciation, and the deduction equivalent of the ITC to total deductions.

30. In fact, the effective marginal tax rate for middle
income individuals can be as high as 33 percent because
of a provision that phases out exemptions as gross income
rises. In either case, however, the highest marginal personal income tax rate is lower than the corporate rate.
31. By lengthening the allowable life of depreciable assets
for tax purposes and by eliminating the investment tax
credit, the availability of nondebt shields relative to debtrelated shields is reduced.
32. In these projections, a corporate tax rate of 34 percent
and a personal tax rate of 28 percent are assumed. In
addition, the difference between the capital gains tax rate
and the tax rate on ordinary income is set equal to zero
(that is, the tax on equity income is assumed equal to the
tax on ordinary income) and the nondebt shield variable is
set to 0.5. Finally, the three percent change in the inflation
rate is assumed. The projections are generated by the first
regression in Tables 1, 2 and 3; the numbers cited in the
text are approximations derived from those simulations.

24. Since the ability to convert dividend income to capital
gains and to delay payment of taxes on capital gains exists
even if the statutory rate of tax on capital gains is the same
as that on ordinary income, no attempt is made to imbed
this condition in the regression analyses presented below.
25. A first differences formulation and various simple and
polynomial distributed lag structures on the coefficients of
the independent variables were examined as well. There
was no evidence of significant lagged effects or qualitative
differences among the performance of the simple regressions, the lagged representations, and first difference
representations.

Full text of Economic Review (Federal Reserve Bank of San Francisco) : Fall 1987, Number 4

FRASER