View original document

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

fh'ilurO

Uiu^cu,

DISTRIBUTION OF TAX PAYMENTS BY
INCOME GROUPS: A CASE STUDY FOR 1948




by
R. A. MUSGRAVE, J. J. CARROLL,
L. D. COOK, AND L. FRANE

Reprinted from
NATIONAL TAX JOURNAL
March, 1951, Volume IV, Number 1

National Tax Journal
V o l u m e IV, N o . 1

March 1951

DISTRIBUTION OF TAX PAYMENTS BY INCOME
GROUPS: A CASE STUDY FOR 1948 *
R. A. MUSGRAVE, J . J . CARROLL, L . D. COOK, A N D L . F R A N E

absence of information. There is some
empirical evidence to work with, and
some conclusions may be reached deductively. Moreover, the economist
can be useful in making explicit the
assumptions underlying a particular
conclusion. His informed judgment, to
say the least, should prove a better basis
of policy than random guesses or political slogans. Thus there is no need to
apologize for the tentative nature of
our results.
In section I the conceptual difficulties
and limitations underlying a study of

I. T H E M E A N I N G O F I N C I D E N C E

T

HE SEARCH for quantitative estimates of tax incidence leads the explorer through a wasteland of muddled
concepts, inadequate theory, and lack
of data. Nevertheless, the question
who pays the taxes must be answered
if taxes are to be raised in accordance
with the public's ideas of distributional
justice and the maintenance of sound
economic conditions. Although available evidence is discouragingly scarce,
the economist cannot plead complete
* This study is a product of a seminar in public
finance at the University of Michigan, conducted
during the fall term of 1949 under the direction of
Professors R. S. Ford and R. A. Musgrave. Members
of the seminar who contributed to the study were
J. J. Carroll, L. D. Cook, L. Frane, H . Frisinger,
H . S. Gordman, C. T. Hardwick, J. Hurwitz, H.
Jaroche, and J. W. Reher.
The revision, expansion, and assembling of the
work was undertaken by the present authors, who
alone should be blamed for its shortcomings.
More than on any other source, this study draws
upon the findings of the Survey of Consumer Finances, prepared by the Survey Research Center of
the University of Michigan in conjunction with the
Board of Governors of the Federal Reserve System.
Our thanks are due to members of the staff of the




Survey Research Center, especially George Katona
and Kent Winter, for their help in supplying and
interpreting much of the underlying data. Also, we
are indebted to the Executive Board of the Horace
H. Rackham School of Graduate Studies of the
University of Michigan for a grant from the Faculty
Research Fund.
EDITOR'S NOTE: Because of the importance

of

this

subject and the convenience to users of having the
whole study published in a single issue, we have departed from customary restrictions on length of articles. A limited number of reprints are available
from Secretary of National Tax Association, P. O.
Box 1799, Sacramento 8, California, at a price of
50 cents each {45 cents each for 20 or more).
1

2

NATIONAL TAX

tax incidence are given brief consideration. In section II sources of data, assumptions, and our procedures with
regard to particular taxes are reviewed.
In section III we examine the results for
the "standard case," based upon what
appears to us the most likely set of assumptions. In section IV the significance of alternative assumptions is explored, and our results are compared
with those of earlier studies. Some of
the underlying data are explained further in the appendix.
This study, like other attempts at
quantitative estimation of tax incidence,
inevitably involves the use of grossly
simplifying assumptipns. Nevertheless,
it is important to consider first how the
measurement of incidence should be
formulated if all empirical difficulties
could be overcome. Unless we are clear
as to what it is we wish to measure, we
are in no position to appraise the merits
or shortcomings of our statistical results. The impatient reader may want
to skip the somewhat theoretical discussion presented in the initial section; but
if he does he should be warned that
"incidence" is an ambiguous and far
from self-explanatory term.
1. Taxation and Surrender of Resources
Let us start out from an economy in
which there is no public budget. Certain public services are now to be introduced. For this purpose, resources
previously employed to meet private
demand must be transferred to public
use. The government may resort to
two methods of doing this: It may conscript or, as done usually, it may go into
the market and purchase. If the resources or products are to be purchased,
various means of finance are available.
Finance by fees or special assessments is
feasible for only a small part of public




JOURNAL

[VOL. I V

services. The bulk of public services
must be tax- or loan-financed. Consider the case of tax finance. The government must raise a tax yield equal to
the purchase price of the required resources or products. 1 To simplify, let
us suppose that the economy is limited
to two factors of production, a and b,
and to two taxpayers, X and Y. Also,
let us assume that X is primarily a consumer of a and Y primarily a consumer
of b. Suppose now that the government is in need of a. It is evident that
this resource must be surrendered
largely by X, but it does not follow that
X rather than Y must bear the major
burden. Let us suppose that the government wishes to impose the major
burden upon Y. The revenue may
then be obtained from an income tax
on Y, whose demand for b will decline
as a result. X will find that the price
of a has risen because of additional public demand for it, while that of b has
fallen because of the shrinkage of Y's
demand. Accordingly, he will substitute b for a in his consumption pattern.
Thus he will recoup from Y in the form
of b most of what he has surrendered
to the government in the form of a.
As is apparent from this illustration,
the party surrendering the resources
required for public use need not be the
party bearing the major part of the cost
of the public service. It follows that
the formulation of tax policy involves
an inevitable decision as to whose income should be cut. Considerations of
income distribution are the heart of tax
policy.
2. Nature of Tax Burden
Suppose now that it has been decided
^•We are here concerned with exhaustive or real
expenditures only. Transfer payments, throughout
this discussion, should be thought of as negative taxes.

No. 1]

TAX PAYMENTS BY INCOME GROUPS

how much various members of the community should contribute to the given
yield total. There remains the question
how this desired distribution can be implemented or how the cost of public
services will in fact be distributed under
any given revenue structure. This,
broadly speaking, is the problem of incidence. But before considering how
the distribution of this cost or burden
can be determined, let us examine more
closely how the burden concept is to be
defined.
It is evident that the term " burden of taxation " applies to only one
side of the revenue-expenditure process.
Whether a particular revenue-expenditure combination results in an ultimate
net gain or loss to the community depends upon the balance between the
gains incident to the public use of resources and the losses incident to their
withdrawal from private use. The concept of burden, as here used, refers only
to the losses incident to the withdrawal
of resources. Therefore, it does not imply the existence of a social burden,
which results only where additions to
the revenue-expenditure total result in
a net loss on balance.
Let us begin with the concept of
money burden. Given the level of income before tax, every tax dollar collected by the government will reduce
someone's income after tax by a similar
amount. This is evident as regards personal taxes because they are simply deducted from income before tax to obtain income after tax. The case is a
little more difficult for taxes paid by
consumers in the form of higher prices.
To simplify matters, we shall translate
such taxes into personal taxes and define
a person's income net of tax to equal
his money income received, minus his




3

3

personal taxes I>aid, and minus indirect
taxes imputed to him.
We may then say that the money
burden of prevailing taxes upon any
particular taxpayer equals the difference between his actual money income
net of tax and the money income which
he would receive in the absence of taxes.
And if we assume money income before
tax to be the same with and without
taxes, it follows that the aggregate
money burden carried by taxpayers as
a group is equal to the tax yield.2 But
if income before tax differs in the two
cases, the total money burden will exceed or fall short of the tax yield.
Money burden thus defined is the
basic concept of this study, but it has
serious shortcomings. It may, for various reasons, be a poor measure of the
welfare loss involved, and the measurement of welfare loss, after all, is the
raison d'etre for the development of
any burden concept. Discrepancies between
money burden " and " true
burden " may arise for several reasons.
Two situations may involve the same
total money burden; yet the true burden may differ, depending on the distribution of the money burden among
members of the community. Or, as between two situations involving similar
distributions of an equal money burden,
the total true burden may differ depending upon what taxes (e.g. personal
taxes or excises) are used. For similar
reasons, the distribution of the true
burden among members of the community may differ from the distribution of
the money burden. Some individuals
may, in fact, be subject to no money
2
The reader will note that it is absurd to assume
income to be unchanged by tax repeal, if, at the
same time, public expenditures are held constant.
See below.

4

NATIONAL TAX JOURNAL

burden, yet carry a share of the true
burden. Consider, for instance, taxpayer X in the above illustration. Suppose that the price increase in commodity a is matched by a proportionate
decrease in the price of b. In this case
X will suffer no money burden; but if
he prefers a over b> his economic position will be worsened nevertheless.
For purposes of this study, we shall
disregard these rather elusive issues and
consider only changes in welfare as reflected in corresponding changes in
money income. This omits interesting
issues but reduces the problem to measurable form. The burden of taxation
applicable to any particular taxpayer,
accordingly, is defined as the difference
between his actual money income net of
tax and what this income would be in
the absence of the tax. This difference,
we repeat, is defined to include differences in money income before tax, differences in personal taxes paid, and differences in excise tax payments imputable to the particular taxpayer.
3. Impact and Adjustment
Tax Changes

to

In determining the distribution of
this money burden, two extreme approximations suggest themselves. One
is that each tax dollar rests with whoever carries the statutory liability for
its payment; the other is that, no matter where the initial payment is made,
the tax liability will eventually be
shared alike by all members of the community. Both propositions are untenable. The statutory taxpayer will attempt to recover part of his payments
by adjusting the terms on which he
trades in the market. Depending upon
the type of tax and the market setting,
he will be more or less successful in




[VOL. I V

doing so. The truth, it appears, lies
somewhere between the extremes of the
impact and diffusion theories.
We
must try, as best we can, to determine where the burden of particular
taxes lies.
As will be shown presently, it is impossible to determine the burden distribution for any particular tax system
without reference to the burden distribution under an alternative system or
without including the effects of the expenditure side of the public budget.
But let us avoid this difficulty for the
time being and consider the consequences of substituting a new tax structure for a prevailing structure of equal
yield.
This change will disturb the prevailing equilibrium (or alter the prevailing
disequilibrium) and call forth an extensive chain of adjustments throughout
the economy. The chain of adjustments will extend much beyond the
points of impact, where old taxes are
removed and new taxes are imposed.
The manufacturer who pays a new tax
will raise his price to the wholesaler and
the wholesaler to the retailer who, in
turn, will raise the price charged to the
consumer; consumers will shift to products the prices of which have risen less
or have fallen; wages and profits will
rise in industries facing increased demand and decline in industries facing
reduced demand. These changes will
bring about further adjustments in the
budgets of wage and profit recipients
and so forth. This succession of induced adjustments, radiating out and
branching off in all directions, will continue until the economy has reached a
new position of equilibrium. 3
3
The distinction between new and old taxes is
clear-cut in a setting in which tax changes provide

No. 1]

TAX PAYMENTS BY INCOME GROUPS

3

In the course of this process the in- income before tax is changed in the
come positions of individuals will be af- process. But even if income before tax
fected in various ways. Changes in is assumed constant, the sum of losses
Mr. Jones' income due to direct changes will typically exceed T, since taxpayers
in his tax payments are only one among will incur losses while avoiding tax paymany factors. This is frequently over- ment. Thus the producer, who is sadlooked, as the shifting process, which is dled with an excise, will raise his price
but a different term for the process of to curtail the burden; but even though
adjustment just described, is thought of the tax, as reflected in the price rise, is
in terms of a spreading network of in- imputed to the consumer, the producer
come losses, coming to an end wherever is still likely to suffer a decline in
any particular part of the tax burden profits. At the same time, the very
stays put. Thus, a seller who is con- shift in demand which leads to a decline
fronted with an additional excise yield- of profits and wages in the taxed indusing a total amount T raises his price but try, also leads to a corresponding insells less.4 Accordingly, he finds his crease in profits and wages somewhere
profits reduced (suffers a loss) by an else. The tax adjustment produces a
5
amount L s . The consumer, in turn, series of gains as well as losses. Denotfinds the price increased. He reduces ing all such gains with G, we have
his purchase but suffers a remaining
T = L, + L0 + LW — G.
loss Lc equal to the increment in price If income before tax is unchanged, the
times the new quantity bought. The excess of income losses over tax yield
workers in the taxed industry may find will be offset by gains in another part
their wage receipts cut and suffer a loss of the system.
Lw, and so forth. If we interpret the
conventional concept correctly, the 4. Incidence and Effects
The difference in the distribution of
total money burden T is thought of as
the sum of these losses, while its distri- the money burden under the new and
bution is taken to be split according to the old revenue structures may now be
the amounts L8, Lc, and Lw. But this is determined by comparing the initial
equilibrium with that which prevails
not correct.
The sum of L8 -f- Lc + Lw will equal after adjustment to the new tax system
T only under quite exceptional condi- is completed. Unless total income betions. Obviously, there will be no such fore tax has changed, total income net
equality in the general case, where the of tax will be the same in both cases.
The distribution of income, however,
the only disequilibrating factor. In a dynamic econwill differ. This change in distribution,
omy, which is continuously in a process of change,
as determined by comparing the two
the distinction between old and new taxes is blurred;
even old taxes remain part of the economic map and
hence a factor in the continuous process of adjustment.
4
To facilitate presentation, the argument in this
and the following paragraph is in terms of a particular tax, and hence absolute incidence, rather than in
terms of a comparison between two taxes. However,
the same line of reasoning applies with regard to differential incidence.




5

Other gains might result as well. Thus X might
benefit from a decline in the price of b, brought
about by the decline in Y's demand due to an increase in Y's income tax. In order to allow for this
type of gain or loss, the above definition of change
in income net of tax must be amended to allow for
all changes in the prices of consumer goods bought,
not only for price changes of commodities which
have been directly subject to an excise change.

5

NATIONAL TAX JOURNAL

6

equilibria, will reflect all the previously
mentioned gains and losses brought
about in the course of the adjustment
process. A comparison of the two equilibria does not permit us to single out
any particular chain or link in the adjustment process and to determine its
result separate from that of the remaining adjustments. Yet, this is precisely
what most authors, employing the timehonored distinction between " incidence " and " effects " attempt to do.6
This separation, usually appears to
aim at a distinction between changes in
income brought about " directly" by
tax payments (i.e., changes in personal
and excise taxes imputed to the particular individual) and changes in income
brought about by other phases of the
adjustment process (e.g., changes in income before tax or changes in the prices
of products bought, which do not directly reflect indirect tax payments).
As far as we can see, such a separation
serves no analytical purpose. For the
taxpayer as well as the government,
what counts is how income positions
are changed on balance by the entire
tax adjustment. A decline in a taxpayer's wages, brought about by reduced demand for his services is no less
painful than a decline in his wages
brought about by an increase in his personal tax. Both changes are part of the
same equilibrating process and can be
determined only simultaneously.
If there is any meaningful way in
which the end result of the tax adjustment may be divided, it appears to be
6

Compare, for instance, E. R. A. Seligman, The
Shifting and Incidence of Taxation, p. 14, and Otto
y. Mering, The Shifting and Incidence of Taxation,
p. 3. For a healthy skepticism regarding the difference between incidence and effects, see, however,
Cannan, Edgeworth, and Black, as noted in D.
Black, The Incidence of Income Taxes, p. 123.




[VOL. I V

between changes in the level of income
before tax and changes in the distribution of income net of tax. Both are
important outcomes of the same tax adjustment, but they may be usefully distinguished and be measured separately.
This distinction, perhaps, contains the
kernel of usefulness in the dichotomy
between " incidence" and " effects."
Resulting changes in the level of income may be referred to as "effects,"
while resulting changes in distribution
may be referred to as ff incidence " of
taxation. But if such a distinction is
made, it must not be forgotten that
both parts are the result of the same adjustment process and that both encompass all links and chains in this process;
the distinction merely applies to different features of the end result. 7
5. Absolute vs. Differential

Incidence

The reader may wonder why, after
having defined the concept of burden
in section 2 in terms of absolute money
burden, we conducted the discussion in
section 3 in terms of changes in the distribution of the money burden which
results when one tax system is substituted for another. Would it not have
been simpler to compare income net of
tax under the given tax structure with
income net of tax in the absence of
taxes and thus determine the absolute
burden distribution of the given tax
structure?
Unfortunately, this cannot be done,
for the simple reason that taxes are the
counterpart of public expenditures. 8
7
In addition to effects on distribution of income
and level of employment, one might distinguish f u r ther effects, such as effects on capital formation or
the rate of growth, which are also separately measurable.
8
Note that we start with a situation in which
government expenditures are tax financed. Suppose
now that we start with a situation in which in order

No. 1]

TAX PAYMENTS BY INCOME GROUPS

In a situation where, to begin with, we
have public expenditures as well as
taxes, we cannot <c think the existing
taxes away"—to borrow Wicksell's
terms—without either (a) thinking
public expenditures away as well, or
(b) substituting some alternative means
of finance.9
(a) If we choose the first approach,
we aim at determining the combined
incidence of taxation and public expendituresh That is, we compare the
situation under the prevailing budget
with a hypothetical situation in which
there are neither taxes nor public expenditures. If the problem is formulated in this way, two aspects of
expenditure incidence need to be distinguished. First, we must account for
changes in the income position of individuals which are brought about by the
expenditure side of the budget; second,
we must account for the benefits derived frohi public services which are enjoyed free of charge.
to maintain full employment expenditures are credit
financed. We have two lines of reasoning.
One is to argue that, in the absence of public expenditures, there would be unemployment of resources; hence the public employment of resources
involves no opportunity cost in alternative private
use. While there is some validity to this line of
thinking, it is rather unsatisfactory: in a rationally
conducted economic system, the allocation of resources for any use, public or private, should always
be on the assumption of full employment; policies
aimed at maintaining full employment should (and
can) be handled as such and so as to be neutral
with regard to allocation.
The other line of reasoning is to argue that in
the absence of the given expenditure program, the
maintenance of full employment would have been
secured by alternative expenditures, say transfer payments, proportional to income. In this case the burden of public services might be obtained by comparing the actual situation with one of proportional
transfer payments.
• See Knut Wicksell, Finanztbeoretische XJnterzuchungen und das Steuerwesen Scbwedens (Jena,
1896), p. 6 if.




3

The first problem offers no conceptual difficulty. Changes in money income due to public expenditures may
be in the form of direct income subsidies through transfer expenditures; or
they may take the form of changes in
factor income due to changes in the
over-all pattern of demand. By comparing the distribution of income in the
pre- and post-budget economies, such
gains and losses are determined together
with those stemming from the tax side
of the budget. If money income before
tax is the same in both situations, the
gains and losses thus determined for the
combined tax-expenditure process will
show a burden (decline in income net
of tax) equal to the amount of tax
yield or expenditure. 10 The component changes in individual income positions will show how the total money
burden, incident to the withdrawal of
resources from private use, is distributed. But note that the result thus obtained is one of budget incidence, not
of tax incidence. We cannot obtain
two separate and additive results relating to the tax and expenditure sides
respectively.
The second aspect of the expenditure
problem is more difficult to handle.
Unless one is willing to follow Adam
Smith's rule of thumb that " benefits
received are proportional to income received under the protection of the
state," there is no simple way in which
the benefits from general public services may be imputed to particular members of the community. And without
this, we do not have a complete picture
of the distributional effects of the public budget. But even though this comThis refers to goods and service expenditures.
In the case of transfer payments there will be an
offset of gains and losses and a zero burden.

7

8

NATIONAL TAX

plete picture may be impossible to obtain, we may at least derive an absolute
measure of the distribution of money
burden by combining the income
changes which result from both the tax
and expenditure side of the budget.
(b) While it is evident that a complete analysis of incidence must include
both sides of the budget, it nonetheless
remains of interest to consider the tax
side as such. Except for certain minor
cases of clear-cut benefit taxation, the
two sides of the budget are the result of
separate legislation and involve quite
distinct issues. In expenditure determination, the objective is to render certain services and to do so efficiently; the
distributive effects of public demand
for resources are secondary. In tax
determination, the distribution of the
money burden is of primary importance. But, if we wish to study the
incidence of taxation as such, it must be
recognized that this incidence cannot be
thought of in absolute terms; an alternative means of finance must be substituted when the prevailing taxes are
" thought away," and this means that
the result will necessarily be in differential terms.
The concept of differential incidence
is not necessarily one of differential tax
incidence. Suppose, for instance, that
bank credit is substituted for tax
finance. If full employment would
have prevailed in the case of tax finance,
inflation is bound to result. The comparison, then, will be between the burden distributions for inflation and for
tax finance. This, to be sure, is not a
particularly useful comparison.
A
more useful approach to differential incidence is to follow the procedure applied in section 3 above and to compare
the incidence of the prevailing tax




JOURNAL

[VOL. I V

structure with that of an alternative tax
structure. In this case, the comparison
is between the prevailing equilibrium
and that which would result if the new
tax structure were substituted.
While this permits a comparison of
incidence under any two tax structures,
a more general measure permitting
comparison between more than two tax
structures might be desirable. For this
purpose, each of the given tax structures might be compared with some
easily definable reference case, say a tax
structure in which tax payments are
proportional to income before tax.
Whatever is done, it appears that a
meaningful concept of incidence must
either be formulated in differential
terms (in which case public expenditures and their incidence may be held
constant) or it must be a combined
measure of tax and expenditure incidence.
6. Empirical Difficulties
We now proceed to a quite different
level of argument and consider the empirical difficulties which prevent statistical implementation of a conceptually
proper estimate of tax incidence. The
limitations are appalling. Economists
can rarely experiment. They cannot
switch tax systems or repeal the budget; and even if they could, they would
be unable to " hold other things constant " while equilibrating adjustments
took place. Thus our reasoning is
largely deductive. Occasionally a major
deductive conclusion may be buttressed
by empirical observation of the adjustments that result from a minor change
in the tax structure. Beyond this, empirical studies of tax shifting are nonexistent. In view of these limitations,
let us now consider how the issues

No. 1]

TAX PAYMENTS BY INCOME GROUPS

raised in the preceding pages are handled in this study.
(a) We do not attempt to trace incidence by individual taxpayers but
limit ourselves to a study of incidence
by spending unit income groups.
(b) We do not adopt a general impact or diffusion theory, but attempt
—in light of accepted theories of market behavior—to make reasonable assumptions regarding the point in the
income-expenditure stream from which
the tax is drawn. By applying these assumptions to our data, we attempt to
estimate the final money burden for
each tax by spending unit income
brackets. The deductive reasoning underlying certain important assumptions,
especially those relating to the shifting
of the corporation tax, is by no means
conclusive; in such cases alternative assumptions are formulated and the results compared.
(c) In considering the equilibrating
adjustment to any given tax change, it
is not feasible to trace out the entire
pattern of gains and losses. We must
be content to record only items of loss
associated directly with the sale (or
purchase) of the taxed commodity or
service. Items of loss incurred in adjustment to the tax, such as reduced
profits due to reduced output, are not
accounted for, and gains and losses
occurring elsewhere in the system are
wholly neglected. For example, the
rule of thumb is used that an excise tax
should be imputed to consumers in the
same proportion in which they consume
the taxed commodity. Changes in
business profits or in returns to factors
before tax are disregarded, as are consumer gains and losses that result from
changes in the relative prices of taxed
and nontaxed commodities.




3 9

From the preceding discussion, it is
evident that such simplifications can be
defended only because they are essential
in order to reduce the problem to empirically manageable terms. They cannot be defended by arguing that the
resulting concept of money burden is
endowed with analytical significance
which it does not possess. There is a
question of how seriously these oversimplifications affect the quantitative
results. Undoubtedly the resulting distortions would be serious if we were to
measure the incidence of the tax system
with respect to each individual taxpayer; the factors here omitted may
well be decisive for changes in the income position of any one person. Since
we are concerned with the distribution
of the money burden by income brackets only, the distortion will be less serious. We may presume that a transfer
of profits from one seller to another or
an inter-industry change in wages will
be distributed more or less in line with
the existing income distribution; and
the same may be expected to hold for
consumer gains and losses from a change
in the prices of commodities not taxed.
Restated in these terms, the diffusion
principle has some merit.
(d) This study does not take into
account the expenditure side of the
budget but deals only with the distribution of the tax burden; and, as shown
in the preceding discussion, this concept
can be formulated meaningfully only in
terms of differential incidence.
Yet, the results of this study, as presented in subsections 1-4 of section III,
trace the absolute tax yield under the
given revenue structure to its resting
place by income brackets. This provides us with the type of data presented
in earlier incidence studies; but it ap-

10

NATIONAL TAX

pears to record incidence in absolute
terms, which, according to our preceding discussion, is not meaningful.
Rather, the incidence of the given tax
structure must be measured in terms of
its difference from an alternative structure. For a reference case we may consider a hypothetical structure which
provides for a proportional distribution
of the money burden. In subsection 5
of section III an attempt is made to
provide such a measure of differential
incidence.
Admittedly this is little more than a
gesture of respect to our conceptual discussion. The distribution of tax payments for the proportional system case
is arrived at by proportionately allocating the given yield total on the basis of
the same distribution of income before
tax existing under the impact of the
given 1948 tax structure. That is to
say, we disregard the fact that the basic
distribution of income before tax would
have differed had a proportional tax system been applied. Again, this is a procedure which can be defended only on
grounds of empirical limitations.
(e) The study, finally, makes no allowance for possible changes in the level
of money (and real) income which may
result from the tax adjustment. In the
terminology of incidence and effects
suggested above, the implicit assumption is that the entire result of the tax
adjustment is measurable in terms of
incidence, effects being zero. If we
consider our results in terms of differential incidence, this involves the additional assumption that the level of income before tax (as well as its distribution) would be the same under both tax
structures.
( / ) To this list of shortcomings
others must be added which arise from




JOURNAL

[VOL. I V

faulty assumptions or deficient data.
These will be considered in detail in the
following section. At this point, the
reader may despair of the usefulness of
undertaking a quantitative study of this
kind. Certainly, a healthy skepticism
is in order. It should not, however,
be carried too far. We must remember
that distributional considerations are of
major importance in the determination
of tax policy. Legislators must choose
between alternative taxes, and it is better that they be provided with part of
the answers than with none. And,
where empirical evidence cannot be obtained, it is important at least that assumptions be made explicit and be acceptable on deductive grounds.
II. SOURCES, M E T H O D S , A N D
ASSUMPTIONS

In this section we present a brief
summary of the methods by which our
distribution of tax payments has been
derived. No detailed calculations are
presented, but we lay our cards on the
table and give the reader an opportunity to familiarize himself with the underlying data, procedures, and assumptions. Only thus will he be in a
position properly to appraise the reliability of our results.
1. Basic Data: Distribution of Income
Calendar year 1948 was chosen for
this study as the latest period for which
data were available. The first step was
to obtain a distribution of income by
income brackets to which tax payments
might be allocated. The distribution of
income by spending units, appearing in
the " 1949 Survey of Consumer Finances," was most appropriate for our
purpose. Prepared by the Survey Research Center of the University of

TAX PAYMENTS BY INCOME GROUPS

No. 1]

Michigan for the Board of Governors
of the Federal Reserve System, this distribution accounts for 50.4 million
spending units and a total money income of $176 billion. This income is
distributed among seven income brackets, ranging from "under $1,000 " to

3 11

tive tax rates are higher than they
would be had we used the larger income base. We decided to use the
S.R.C. income total because the entire
study is based upon the S.R.C. distribution of income, the distribution of the
difference between the two income

TABLE 1
BASIC DISTRIBUTIONS FOB TAX ALLOCATION

(In per cent)
Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1
(1) Money income
Consumer expenditures:
(2) Total
(3) Durables
House expenditures:
(4) Owners
(5) Renters
(6) Liquid asset holdings
Consumer expenditures:
(7) Tobacco
(8) Auto transport
(9) Other transport
(10) Recreation
(11) Food
(12) Alcoholic beverages
(13) Housing
(14) All other
(15) Dividends
(16) Wage and salary income . . . .
(17) Wages covered by p.r. tax ..
(18) Rental income

1-2

2-3

3-4

4-5

5-7.5

Total

2.0

7.6

16.1

19.5

14.5

17.3

23.0

100.0

3.9
2.2

9.3
5.6

18.1
18.2

20.7
20.0

14.4
15.8

16.1
17.0

17.5
212

100.0
100.0

2.7
3.2
4.0

8.8
13.6
7.0

16.9
24.0
14.0

21.2
22.2
14.0

13.9
13.7
11.0

15.0
14.5
18.0

21.5
8.7
32.0

100.0
100.0
100.0

3.9
3.4
1.0
1.8
4.1
9
4.1
4.2
1.1
2.5
1.1
4.7

11.4
6.0
10.1
5.0
11.4
8.9
9.6
8.7
2.3
9.2
8.7
8.0

21.1
12.1
20.5
13.4
20.5
17.2
18.4
17.6
3.4
19.5
23.7
9.9

22.9
20.1
23.3
20.0
21.8
22.5
22.1
19.2
4.3
23.6
30.6
11.5

14.3
16.5
13.3
15.1
14.2
12.7
14.7
14.2
3.7
17.5
20.2
8.5

14.1
18.8
16.5
20.5
14.6
16.4
14.5
16.8
7.3
16.0
12.8
12.2

12.2
23.1
15.3
24.1
13.3
21.4
16.6
19.4
77.9
11.7
2.9
45.0

100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0
100.0

Note: Details do not necessarily add to totals because of rounding.
Source: Items (1) to (6) were obtained from data of Survey Research Center, University
of Michigan. Items (7) to (14) were estimated on basis of B.L.S. data. Items (15) to (18) were
estimated on basis of various other sources. For details see appendix.

"over $7,500." The basic distribution
is shown in line (1) of Table 1 .
.

|

£

t '11*

r 11

The income total ot $176 billion tails
short of the Department of Commerce
estimate. ,of/ personal income for 1948 by
1
, , 1 .11'
-xr
1
i i r
nearly $36 billion. More than half ofe
this amount may be accounted for by
conceptual differences, the remainder
being due to statistical discrepancies.1
«
1
1
.
1
j .
Since the level ofe income here used is
lower, the tax to income ratios or effec-




inc
° r m e d a t a ^ not include an estimated
$13.0 billion of income in kind and imputed income
or a n estimated $7.0 billion of income to institutional
and transient population. This leaves an unexplained

of

*

T o have used t h e

sonal income figure of the Department of Commerce,

and t o have d i s t r i b u t e d it b y income groups accord _

ing to the S.R.C. income distribution, would have
meant making the unwarranted assumption that the
$36 billion difference was

distributed in the same
manner as the S.R.C. income. In view of the nature

,

.

.

,

,

of the major items that account for the difference,
there is no reason to assume this.

12

NATIONAL TAX

totals being unknown. Moreover, the
underlying income total is relatively
unimportant for our purpose, which is
primarily to estimate the distribution of
tax payments. Whereas the general
level of " effective tax rates" (i.e.,
taxes contributed as a percentage of income received) shown in section III depends upon the income total used, the
distribution of the tax bill does not.
However, in comparing our results with
those of other studies, an adjustment to
the Department of Commerce level of
income will be made.
The S.R.C. distribution provides no
breakdown of income above $7,500.
Nor is it possible to obtain consumption data for income brackets above
this level. Hence no attempt is made
to determine the allocation of tax payments among income groups above
$7,500. While a further allocation of
tax payments among the higher income
groups would be of considerable interest, there are no adequate data on which
to base such estimates. Also, it is evident that tax policy in a high-taxation
economy, such as lies ahead, will be concerned increasingly with the middle
and lower level income range here
studied.
2. Basic Data: Distribution of
Expenditures
In order to allocate the yield from
excise taxes and such fraction of other
taxes as is assumed to be shifted forward to consumers, it was necessary to
obtain a distribution of consumer expenditures by income brackets. This
information was obtained from two
sources.
A distribution of total consumer expenditures was derived as a residual
from S.R.C. data providing distribu-




JOURNAL

[VOL. I V

tions of income, income tax liability,
and savings. The resulting distribution for total consumer expenditures is
shown in line (2) of Table 1. Most
excise taxes and the forward-shifted
parts of some other taxes, however, do
not apply to consumer expenditures in
general but to specific expenditure
items. Expenditures on specific items,
in many instances, are distributed in
a quite different way than are total expenditures. In order to allocate excises
as much as possible in accordance with
the distribution of specific types of consumption expenditures, a more detailed
pattern of consumer expenditures was
needed. The S.R.C. data provide such
distributions for consumer durables as
well as for housing expenditures of
house owners and renters. They are
shown in lines (3), (4), and (5) o£
Table 1.
Expenditure distributions for a number of further items were derived from
a series of consumer budget studies conducted by the U. S. Department of
Labor from 1946 through 1948. These
data, which cover a number of selected
cities only, had to be reworked to present a national picture. Also, they had
to be assembled into a breakdown of
items most useful for the purpose of
tax allocation. The results of this
rather involved and unsatisfactory procedure are shown in lines (7) to (14)
of Table 1.
3. Basic Data: Revenue Sources and
Yield Total
Total tax revenue collected in 1948
by all levels of government amounted
to $57.3 billion. A breakdown of this
total between levels of government and
among revenue sources is shown in
Table 2.

No. 1]

TAX PAYMENTS BY INCOME GROUPS
TABLE 2

4. Tax
Tax

TAX COLLECTIONS IN 1948

(In millions of dollars)
Tax

Collections

Federal

(1) Personal income tax
(2) Corporation income tax
(3) Excises and customs, total . .

(4) Alcohol
(5) Tobacco
(6) Automotive
(7) Amusement
(8) Transportation
(9) Communication
(10) Luxury
(11) Electrical
(12) Food
(13) Other excises
(14) Customs
(15) Estate and gift
(16) Payroll
(17)

Total

18,045
12,920
7,912

2,177
1,312
1,279
640
610
512
463
265
115
129
410
900
4,015
43,794

State and Local

(18) Personal income tax
(19) Corporation income tax

499
699

(20) Excises, total

6,093

(21) Automotive
(22) General sales
(23) Alcohol
(24) Tobacco
(25) Insurance
(26) Amusement
(27) Public utility
(28) Other excises
(29) Property taxes
(30) Inheritance taxes

1,989
1,899
545
365
202
184
267
642
6,064
197

(31)

Total

13,552

(32)

Grand total, all levels . . . .

57,344

Sources: Lines (1), (2), (14), (19), (29), and
(30) from Survey of Current Business, July,
1949, Table 8. Lines (3) through (13) from
Bureau of Internal Revenue, Release 7, February 23, 1949. Line (16), Survey of Current
Business, July, 1949, Tables 34 and 35, and
Social Security Agency, Annual Report (adjusted to calendar-year basis). Line (18), State
Finances, 1949, U. S. Bureau of the Census,
August, 1949. Lines (20) through (28), U. S.
Bureau of the Census, State Tax Collections in
1949 (G-SF49, No. 4, August, 1949); fiscal year
figures converted to calendar-year basis by
averaging fiscal 1948 and 1949. No serious distortion is involved in this averaging, since the
change during the period was less than 7 per
cent.




Allocation:

3 13

Personal

Income

In prior studies one of the most serious difficulties arose in the allocation
of personal income tax payments by
family or spending unit income brackets. A distribution of income tax
payments by brackets of taxable income may be readily obtained from the
Treasury's Statistics of Income, but
there is no simple way of moving from
a distribution of this type to a distribution by family or spending units. Income splitting and other tax adjustments result in quite different patterns
for the two, distributions. Fortunately
the need for constructing a bridge between the two has been obviated by the
5.R.C. data which provide us with a
percentage distribution of Federal income tax liabilities of spending units
by income brackets. 2
A distribution of state and local income tax payments was not available
from the S.R.C. data. As state income
tax exemptions are considerably higher
and rates less progressive, the Federal
distribution could not be used in allocating the state taxes. While there are
distinct differences among the income
taxes of various states, it was not feasible to estimate separate distributions
for each of the thirty-five state income
tax laws. As an approximation, the
distribution of the Wisconsin income
tax yield by spending unit income
brackets was estimated and this distribution was used to allocate the entire
yield of state income taxes.
The resting place of the entire personal income tax is assumed to be with
the individual upon whom the statutory liability is imposed. Following
2

This distribution is shown in Table 7.

14

NATIONAL TAX

JOURNAL

[VOL. I V

generally accepted practice, we apply earned by the marginal firm. Since the
an impact theory to the incidence of price is set by the marginal firm which
this tax. Yet, it is quite conceivable pays no tax, the tax cannot be reflected
that income tax payments in some in- in a higher price. Under monopolistic
stances may result in a rise in income conditions the firm which sets the price
before tax. Reward for professional does pay a tax. But since a tax on
services, the supply of which may be profits does not alter the price at which
adjusted in small increments and which profits are maximized, the seller will do
are highly specialized, may be a case in best to remain at his prevailing (optipoint. Collective bargaining demands mum) position and to absorb any tax
may be directed at take-home rather imposed on profits.
than gross pay. Business income deThis line of reasoning has been acclared under the personal income tax cepted consistently by the profession
may be subject to the same shifting and has been objected to with equal
considerations as apply to the corpora- consistency by the " practical m a n "
tion tax. Nevertheless, we can be who stoutly maintains that he does add
fairly certain that such adjustments are his corporation tax to his price. Some
the exception rather than the rule. reconciliation is provided by certain
Moreover, it seems impossible to estab- obvious exceptions to the academic rule.
lish any working hypothesis as to how First, the tax base may not be pure
much might be passed on at various in- profit in the economist's sense but may
come ranges. In the absence of an include wages of management, return
alternative formula, it seemed best to for risk taking, interest on equity capretain the conventional assumption that ital, and other nonrent incomes. To
the tax stays where it is imposed.
the extent that this is the case, the tax
is a cost item and will be reflected in
5. Tax Allocation: Corporation Income price. Second, imperfections in the
Tax
tax law inevitably lead to differential
In determining the allocation of the burdens among various enterprises and
corporation income tax yield a number industries. And where this is the case
of different problems arise. First, it (i.e. the tax is not truly general), price
must be decided whether the tax is adjustments must follow. Third, corshifted; if so, by how much and in poration management may not charge
what direction. Next it must be de- the highest price which the market will
cided how to distribute the part of the bear. This restraint may reflect an eftax which is assumed to fall on profits. fort to maintain consumer good will or
Finally the shifted parts of the tax must a notion that profits should not exceed
a tc fair " level (the modern version of
be allocated.
Shifting
Assum ptions.—Economists the old concept of just price); or
with few exceptions have held that the prices may be held down merely to apcorporation tax cannot be shifted. pease anti-trust authorities.
Under pure competition there are no
The extent to which such adjustprofits and hence no tax is paid. Or, if ments are possible will depend on
profits arise from efficiency differentials market conditions, but the widely held
between firms, no such profits are view that there will always be more




No. 1]

TAX PAYMENTS BY INCOME GROUPS

shifting in a seller's market is not as
obvious as it appears.8
In addition to the situations described above where forward shifting is
possible, there are situations in which
backward shifting may occur. Consider for instance a situation where the
spoils of profit exacted from the consumer are shared by the seller and a
strong union. In this case wage demands depend upon profits net of tax,
and a change in tax may well be reflected in a change in wages. In all
these cases the theorist will readily admit that the imposition of a corporation
income tax may lead to price or wage
adjustments so that the no-shifting
rule does not apply.
Some other lines of argument need
concern us less. One is the practical
man's contention that " the pricing
process does not conform to economic
theory."
The businessman, we are
told, will treat the tax as if it were a
cost. That is, he will include the tax
in his computation of average costs to
which then is added a " customary"
profit (net of tax) margin in order to
obtain the price. What gives rise to
the " custom " is, of course, the crux
of the problem. Its existence either reflects the presence of voluntary restraint as noted above, or it must be
that such a margin is a measure, however imperfect or lagging, of optimum
profits. If the latter is true, the argument breaks down. The customary
margin, sooner or later, will be adjusted to a change in market data so as
to maximize profits. Sellers will find
3 To argue that there will be more shifting in a
seller's market it must be shown that the tax has
more bearing on price policy, not merely that it is
easier to raise prices. The situation may be the reverse if the case for shifting is based on the " restrained use of monopoly power " argument.




3

15

out, sooner or later, that adding the tax
does not pay and the margin will be adjusted accordingly.
A final argument refers to shifting in
the long run and is frequently combined with the no-shifting rule for the
short run. While the tax is held to fall
on profits in the first instance, it is suggested that this will result in a decline
in capital formation. The decline in
capital formation in turn will result in
lower productivity and hence bear upon
the standard of living of the entire
community, including wage as well as
profit recipients. Thus the resulting
burden is distributed more broadly in
the long run, even though the shortrun incidence of the tax is on profits
only.
In evaluating this argument, one
would have to determine first whether
the level of capital formation would be
affected significantly by a corporation
tax of the 1948 rate level, given the
economic conditions which then prevailed. But let us suppose that capital
formation does decline. It follows that
the community's real income in the
future will be lower than it otherwise
would be. On the assumption that the
lower level of capital formation will
not be permitted to reduce the level of
employment, a larger fraction of the
economy's output will come to be consumer goods. This result is extremely
difficult to appraise; we have no way by
which to match the loss of real income
due to lower productivity with such
gains in satisfaction as might be derived from a higher level of immediate
consumption. Welfare theory, as it
now stands, does not permit us to define
an optimum rate of capital formation
in this sense.
Nor is it an easy matter to assess the

16

NATIONAL TAX

distributional effects of the change toward a higher consumption economy.
The productivity of labor will be less
than it would have been had there been
more capital formation. The productivity of capital will be higher than it
would have been had the capital stock
been larger. The lower productivity of
labor is applied to the same labor supply as would have existed otherwise,
whereas the higher productivity of capital is applied to a smaller capital stock.
There may be a presumption that the
profits share in national income will
have increased as a result of these adjustments, but no firm conclusion can
be drawn on a priori grounds. The results will depend upon the particular
circumstances of the case. It seems that
the long-run argument is rather inconclusive and that not much is lost by
neglecting it in this study.
By and large, it may be concluded
that the academic view remains essentially correct. But certain exceptions should be allowed for, especially
under conditions of a seller's market
such as prevailed in 1948. Since this is
a controversial matter and the corporation tax is of great quantitative importance, the testing of a number of
shifting hypotheses seemed desirable.
Thus we may determine the extent to
which the results depend upon any particular assumption, and the reader who
disagrees with our judgment will be in
a position to choose the assumptions
best suited to his own appraisal of the
problem.
Accordingly, the allocation of corporation tax payments by income
groups was estimated on the basis of
three shifting assumptions. First, it
was assumed that one-third of the tax is
shifted forward to consumers and one-




JOURNAL

[VOL. I V

eighth of the tax is shifted backward to
the wage earners, the remainder falling
on profits. This is considered the
standard case. While these particular
ratios are more or less arbitrary, a
choice of what seemed the most likely
set of assumptions had to be made. In
addition, two limiting cases are considered. One involves the assumption
that the entire tax is shifted forward to
consumers and the other the assumption that the entire tax remains on
profits. A third limiting case of complete backward shifting did not seem to
warrant investigation; it is not sufficiently realistic, and the results would
not differ greatly from those obtained
with complete forward shifting.
Treatment of Retained Earnings.—
Were it not for the problem of retained
earnings, the part of the tax which is
assumed to fall on profits could without further discussion be imputed to
shareholders. The fact that corporations retain a substantial part of their
profits complicates matters. It cannot
be assumed that taxes falling on profits
will be reflected entirely in reduced
dividends; rather they must be spread
between dividends and retained earnings. The part imputed to dividends
may be traced to shareholders, but the
allocation of the remainder requires
closer consideration.
Two possibilities suggest themselves:
We may either disregard the part of the
tax which falls on retained earnings or
we may impute it to the shareholders,
along with the tax component which is
taken to fall on dividends. The former
procedure has the disadvantage that a
considerable part of the total tax yield
is not accounted for; the latter has the
disadvantage of implying the rather
unrealistic assumption that taxes paid

No. 1]

TAX PAYMENTS BY INCOME GROUPS

out of retained earnings are, in effect,
taxes paid by shareholders. Since something may be said for either assumption, the corporation tax was allocated
on both grounds. For purposes of the
standard case the tax allocated to retained earnings was imputed to shareholders, so that the total yield may be
accounted for.
Six Cases Distinguished.—Our
two
assumptions regarding the treatment of
the tax on retained earnings may be
combined now with the above three assumptions with regard to shifting.
This gives us six distinct cases for the
allocation of the corporation tax. The
assumptions applicable to these cases
are summarized in lines (6) to (9) of
Table 3.
In all cases we begin with the data as
reported for 1948 and shown in lines
(1) to (5) of the table. Profits for
the year amounted to $34.7 billion.
Income taxes collected from corporations amounted to $13.6 billion, leaving
profits net of tax of $21.1 billion. Of
this, $7.9 billion was paid out in dividends and $13.2 billion was retained.
Following the shifting assumptions
of Case A, $1.7 and $4.5 billion of the
yield are imputed to wage earners and
consumers respectively. The remainder,
or $7.4 billion, is taken to fall on profits.
Of this, $2.2 billion is assumed to be
reflected in reduced dividends and $5.2
billion in reduced retentions; that is, in
the absence of a corporation tax, distributions and retentions would have
been correspondingly higher. Allocation between the two is based on observation of past movements of profits,
dividends, and retained earnings which
suggest that changes in profits will be
apportioned among dividends and retained earnings on a 3 to 7 basis.4
Since the stockholders are thus deb-




3

17

ited with tax payments made for them
by the corporation, they must be credited on the income side of their ledger
with a corresponding amount. Thus
$2.2 billion of tax allocated to dividends must be imputed to shareholders
because, in effect, this tax constituted
part of dividend receipts before tax.
In Case A, where retained earnings are
to be accounted for, we must further
impute the tax of $5.1 billion which
was allocated to retained earnings as
well as the remaining retention of $13.2
billion. The total imputed income to
be added to stockholders' money income
thus amounts to $20.6 billion. Such
imputed income is not allowed for in
the S.R.C. distribution of money income. The latter must be adjusted accordingly to permit us to obtain a
meaningful ratio of taxes paid to income received. The adjusted distribution is shown in lines (10) and (11) of
Table 6 below.5
4
There are two bases for breaking down the tax
between the part falling on undistributed profits and
the part falling on dividends. One basis is according
to distribution of profits between undistributed
profits and dividends in 1948. This breakdown was
62 per cent to undistributed profits, 3 8 per cent to
dividends. Another basis is according to distribution
of marginal profits between undistributed profits and
dividends. Additions to profits were examined year
by year to see in what proportion they were divided
between retained earnings and dividends. If changes
in dividends are compared with changes in profits
for those years between 1934 and 1948 when profits
rose, it may be noted that changes in dividends are
in most cases between 14 per cent and 25 per cent
of changes in profits. This suggests a marginal rate
of distribution of about 20 per cent. The change
between 1947 and 1948 was exceptional, however, the
change in dividends being 43 per cent of change in
profits. In the light of this information, 30 per cent
of the tax was assumed to fall on dividends. We
thus adopt a ratio which does not differ greatly
from the 1948 allocation. Some allowance is made
for the fact that between 1947 and 1948 the marginal rate of distribution of dividends was higher
than it usually has been.
5
The adjusted distribution, strictly speaking, is
not applicable to the unadjusted bracket limits. In-

NATIONAL TAX

18

Case B involves the same shifting assumptions as Case A, but retained earnings ' and the tax thereon are disregarded. Shareholders, accordingly are

[VOL. I V

JOURNAL

departure. This raises certain difficulties, as there is an implicit assumption
that underlying economic conditions
differ from those assumed to prevail in

TABLE 3
ASSUMPTIONS FOR CORPORATION INCOME TAX

(Money amounts in billions of dollars)
Cases
A*
Reported data
(1) Profits before tax
(2) Tax collected from corporations ..
(3) Profits after tax
(4) Dividends
(5) Retained profits
Shifting assumptions
(6) Per cent absorbed by profits
(7) Per cent absorbed by wage earners
(8) Per cent absorbed by consumers ..
(9) Are retained earnings allowed for?
Allocation of tax (amounts)
(10) To dividends
(11) To retained earnings
(12) To wage earners
(13) To consumers
(14) Total accounted for
Imputed income (amounts)
(15) From tax on dividends
(16) From tax on retained profits
(17) From retained profits
(18)

Total imputed

B

C

D

E

34.7
13.6
21.1
7.9
13.2

34.7
13.6
21.1
7.9
13.2

34.7
13.6
21.1
7.9
13.2

34.7
13.6
21.1
7.9
13.2

34.7
13.6
21.1
7.9
13.2

34.7
13.6
21.1
7.9
13.2

54.2
12.5
33.3
yes

54.2
12.5
33.3
no

100.0

....

100.0

....
yes

....
no

100.0
yes

100.0
no

2.2
5.2
1.7
4.5

....

2.2

4.1
9.5

....

....
....

....
....

1.7
4.5

4.1

....

13.6

....

13.6

ma

8.5
2.2
5.2
13.2

....

4.1
9.5
13.2

....

....

....

20.6

~22

26.8

~41

~13L2

....

2.2

4.1

13.2

•Standard case.
Note: Details do not necessarily add to totals because of rounding.
Source: Survey of Current Business, July, 1949, Tables 8, 17, 19, 20, 21.

debited with a tax of only $2.2 billion
(i.e. the part of the tax on profits that
is taken to be reflected in reduced dividends), and their income is credited by
a similar amount.
In Case C the entire tax is taken to
fall on profits, but the reported data
for 1948 are again used as the point of
stead of referring to " spending units in the, say,
$4,000 to $4,999 group " we should say " spending
units which fall between the unadjusted bracket
limits of $4,000 to $4,999."




Case A. Otherwise, since the entire tax
is now assumed to fall on profits, the
gross profit figure could not be the
same as in Case A where part of the tax
was shifted. 6 Proceeding in a manner
analogous to our reasoning in Case A,
6
For any given economic situation, gross profits
will differ, depending on whether part of the tax is
shifted or not. Since gross profits in Table 3 are
taken to be the same for all shifting cases, it must
be that underlying economic conditions are assumed
to differ. Differences in the burden distribution for
the different cases, therefore, reflect differences in

No. 1]

TAX PAYMENTS BY INCOME GROUPS

we now debit shareholders with tax
payments of $13.6 billion and credit
them with imputed income of $26.8
billion. The data for Cases D to F are
set up in a similar fashion, as may be
seen from Table 3.
underlying economic conditions as well as differences
in shifting.
An alternative procedure might be followed which
would remove the obstacle of implied differences in
basic economic conditions, so that the measure of
differential incidence would not have to be impeded
by extraneous differences in the general economic
situation. Suppose the standard case is taken to be
the true one, i.e., that one-third of the tax is shifted
forward and one-eighth back. Observed gross profits
are $34.7 billion. Deducting shifted taxes, we find
that profits net of shifted tax are $28.5 billion. The
picture for the standard case is exactly the same for
both methods. Now we work out the other five
cases with the same shifting and retained earning assumptions shown in the table. But instead of starting from given gross profits, as in the text method,
we take net profits of $28.5 billion as the common
starting point for all cases. With this method, taxes
would be distributed between groups exactly as with
the previous procedure. But income additions would
differ in cases C, D, E, and F. These differences in
income would be traceable to the fact that retained
profits would turn out to be different, and so would
dividends. In Case C, for example, gross profits as
well as net profits would be $28.5 billion since it is
assumed that there is no shifting. Under the text
method, gross profits in Case C were $34.7 billion,
to equal the observed figure. The difference in
profits net of shifted tax leads to a fall in retained
profits of $2.1 billion from the figure applicable
under the text method, and a fall in dividends of
$3.1 billion. Thus income additions under the alternative method will fall short of those under the text
method by $5.2 billion.
The difficulty with this procedure, however, is that
we do not know to begin with which shifting assumption—and hence which interpretation of gross
profits—is the correct one. And the results under
the alternative method will differ, depending upon
which case is chosen as the starting point. The procedure used in the text has the advantage that the
results, in each case, are based on the reported figures.
It shows how the burden of the tax, would have been
distributed had one or the other shaking assumption
(and the implicit interpretation of ffefit position in
the absence of a tax) been correct. However, in
comparing the results under the various cases we
must be careful to note that differences not only
reflect different shifting assumptions but also different
interpretations of the underlying economic situation.




3

19

Allocation of Tax Components.—We
may now consider how the various components of the corporation tax are
distributed by spending unit income
brackets.
The part of the tax allocated to
shareholders is distributed according to
an estimated distribution of dividend
income, and the allocation of imputed
income is carried out on the same basis.
The distribution of dividend receipts
was derived from a distribution of dividends by taxable income brackets, available in the Treasury's Statistics of Income for 1947. In the absence of
better information, it was assumed that
thii 1947 distribution applies for 1948
as well. The distribution by taxable
income brackets had then to be translated into a distribution by spending
units, comparable to the S.R.C. data.
The results of this procedure, which involves a good deal of difficulty and possible error, are shown in line (15) of
Table 1.
The forward-shifted part of the corporation income tax was treated similarly to excises, as discussed below.
Where possible the forward-shifted part
of the corporation tax was allocated to
specific items of consumer expenditure,
about 12 per cent of the total being
treated in this fashion. The remainder
was distributed among income brackets
in accordance with the distribution of
total consumer expenditures shown in
line (2) of Table 1.
The backward-shifted
part of the
corporation income tax was allocated in
accordance with a distribution of wage
and salary incomes by spending units.
This distribution was estimated again
from S.R.C. and Statistics of Income
data and is shown in line (16) of Table
1. As in the case of dividends, the

20

NATIONAL TAX

derivation of the wage and salary distribution is subject to a considerable
margin of error.
State Corporation Taxes.—In determining the allocation of the corporation
income tax the share levied at the state
level is treated as though it were
distributed between income brackets
exactly like Federal taxes. This introduces some inaccuracy, since in all
probability state corporation taxes are
less progressive than Federal taxes;
items other than net income are frequently included in the state tax base.
The above procedure was adopted because it was not feasible to submit each
of the state taxes to separate treatirfent.
6. Tax Allocation: Excises Taxes 7
Total excise and sales taxes are included in this study, no exception being
made for such " benefit taxes " as state
gasoline levies. It might be argued
that such earmarked taxes are tied to
specific expenditure purposes and that
the concept of differential incidence
cannot be applied, the only feasible approach being in terms of a combined
tax-expenditure incidence. While there
is some merit to this view, the line is
difficult to draw; and it was decided to
include all taxes without distinction.
Excise and sales taxes of all kinds
are assumed to be entirely shifted forward to the consumer. As noted in
section I, this disregards the resulting
interindustry shifts in profits and
wages. Short of a much more elaborate analysis, there is no way in which
such adjustments can be allowed for.
And, as noted before, it does not seem
unreasonable to assume that such shifts
7

The term excise is here used loosely to include
all kinds of taxes assessed largely or wholly on gross
receipts, sales, or production costs.




JOURNAL

[VOL. I V

will produce no major changes in the
distribution of income by brackets.
The yield of indirect taxes, therefore,
is allocated by spending unit income
brackets in accordance with the distribution of consumer expenditures.
Wherever possible, excises are assigned
to specific expenditure items and distributed in accordance with the specific
expenditure patterns shown in lines (7)
to (14) of Table 1. Where no such
assignment was possible, as in the case
of customs duties, the allocation is made
according to the general distribution of
consumer expenditures shown in line
(2) of the table.
In certain cases the entire tax can be
assigned to a specific category of expenditures. Thus, proceeds from liquor
and tobacco taxes can be distributed
wholly in accordance with the respective distributions of consumer expenditures. In other cases, e.g. automotive
taxes, it was necessary to estimate a
breakdown of the yield between that
part reflected directly in automotive
expenses of the consumer and another
part entering into the costs of operating business in general. The former
was allocated in accordance with automotive expenditures of consumers; the
latter was assigned according to the distribution of total consumer expenditure. A summary of the amounts of
excises allocated to specific expenditure
categories is shown in Table 4, which
also provides similar information for the
forward-shifted part of the corporation
income tax.
The distfbution for total consumer
expenditu®s shown in line (2) of
Table 1 was obtained directly from the
S.R.C. data. The specific distributions
shown in lines (7) to (14), on the

TAX PAYMENTS BY INCOME GROUPS

No. 1]

other hand, involved extensive adjustments in the B.L.S. data. Therefore the
advantage gained by allocating excises
according to expenditure patterns for
specific items may be cancelled by the
lesser reliability of these distributions.
TABLE 4
ALLOCATION OP EXCISE TAXES BY EXPENDITURE
PATTERNS

(In millions of dollars)
Excises *
Allocated
According
to Table 1:
Line
Line
Line
Line
Line
Line
Line
Line
Line

Forwardshifted
Part of
Corporation
Tax t

Federal

State
and
Local

2,173
1,312
601
110
576
197
2,177
203
463

1,890
409
994
....
165
784
583
134
134

3,561
39
232
305
32
250

7,912

6,093

4,540

(2)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)

Total

121

* For explanations, see appendix, Table 19.
f Standard case.

An alternative allocation of excise taxes
was prepared in which all specifically
allocated taxes were redistributed according to the pattern for total consumer expenditures. As shown in section IV, the results do not differ greatly
from those obtained by the more detailed method.
7. Tax Allocation: Property Tax
Allocation of the property tax raises
numerous difficulties. Since this tax is
of great importance, especially for the
distribution of the tax burden among
the lower income brackets, these difficulties must be set forth in some detail.
Allocation of Tax by Category.—
Notwithstanding the scarcity of data,




3 21

it seemed desirable first to determine the
principal categories into which the
property tax yield should be divided in
order to permit us to apply reasonable
shifting assumptions to the component
parts.
Accordingly, the total yield was divided into fourteen component parts.
The first division is between taxes paid
on real estate and taxes paid on personal
property. The tax on real estate is subdivided into taxes paid on owner-occupied homes, leased residential property,
business property, and farm property.
In each case a further division is made
between tax on assessed land value and
tax on assessed value of improvements.
The tax yield allocated to personal
property is divided between property
held by farmers, businesses, and nonfarm individuals. In each case, a f u r ther distinction is drawn between tax
on tangibles and tax on intangibles.
The estimated distribution of the
property tax yield according to these
categories is shown in column (1) of
Table 5. The reader who is at all familiar with the paucity of data in this
area will be aware that this implies a
great deal of conjecture. While there
is fairly adequate information to support the over-all allocation between real
estate and personal property, information on which to base the further allocation of real estate taxes is more
limited, and there is little if any basis
on which to estimate the breakdown of
the yield total that is derived from personal property. The detailed breakdown, shown in Table 5, therefore,
should not be taken to imply that we
are in a position to defend each figure
shown. In some instances, it is merely
a matter of making our assumptions
explicit. Changes in the distribution of

22

NATIONAL TAX

property tax payments which might result from alternative yield allocations
are explored in section III.
Shifting Assumptions.—We now proceed to consider what shifting assump-

[VOL. I V

JOURNAL

housing by home owners, as shown in
line (4) of Table 1. The tax on business property is distributed in accordance with the distribution of dividend
payments shown in line (15) of Table

TABLE 5
ALLOCATION OP PROPERTY TAX YIELD BY TYPE OF PROPERTY *

(In millions of dollars)

Case A t

Case B

Case C

Allocated
According
to Table 1,
line:

434
217
322
643
412
823
670
1,330

434
217
506
759
374
561
670
1,330

434
217
133
532
307
1,228
670
1,330

(18)
(2)
(15)
(2)
(18)
(5)
(4)
(4)

4,851

4,851

4,851

275
274
57
58
275
274

175
175
57
58
225
525

375
375
57
58
105
245

1,213
6,064

1,213
6,064

1,213
6,064

Yield
Type of Property
Real Estate:
(1) Farm, land
(2) Farm, improvements
(3) Business, land
(4) Business, improvements
(5) Rental, land
(6) Rental, improvements
(7) Owner-occupied residential, land
(8) Owner-occupied residential, improvements
(9) Total, real estate
zrsonal Property:
(10) Business, tangible
(11) Business, intangible
(12) Farm, tangible
(13) Farm, intangible
(14) Non-farm, tangible
(15) Non-farm, intangible
(16)
(17)

!

Total, personal property
Total, real and personal

(2)
(2)
(2)
(2)
(3)
(6)

....

•For sources and explanation see appendix,
t Standard case.

tions to apply to each of the fourteen
parts of the property tax and what pattern of allocation to use. These patterns are also summarized in Table 5.
To the extent that the real estate tax
is assessed upon the value of land, it is
assumed to rest on the owner. This is
in accordance with the principle that a
tax on land cannot be shifted, since rent
is not a determinant of price. The tax
on owner-occupied residential property
is allocated in accordance with the
S.R.C. distribution of expenditures for




1. For leased residential and farm
property the distribution of the tax
corresponds to the distribution of rental
income shown in line (18) of Table 1.
To the extent that the tax on real
estate is assessed upon the value of improvements y it may or may not be
shifted depending upon the type of
property. In the case of owner-occupied residential property, this part of
the tax is again taken to rest on the
owner. As in the case of land value,
the tax is distributed on the basis of

No. 1]

TAX PAYMENTS BY INCOME GROUPS

housing expenditures of owners. In the
case of leased residential property, the
tax on improvements is assumed to be
shifted to the tenant since taxes on improvements will enter into cost of production. The tax thus shifted to the
tenant is allocated by income brackets
on the basis of housing expenditures of
renters, as shown in line (5) of Table
1. In the case of business property, the
tax on improvements is treated as an excise entering into the general cost of
doing business. It is assumed to be
shifted forward and is allocated according to the pattern of total consumer
expenditures. The same procedure is
applied to improvements on farm real
estate, it being assumed that the assessed
value of improvements in farm real estate reflects business rather than residential facilities. In the absence of information permitting a split between
these two types of farm improvements,
this appears to be the more reasonable
assumption.
Turning now to the tax assessed on
personal property, let us begin with the
tax on tangibles. The tax on tangible
property owned by individuals may be
expected to rest with the owner. It is
allocated according to a distribution of
consumer expenditures on durables, obtained from the S.R.C. and shown in
line (3) of Table 1. The tax on tangible property owned by businesses and
farmers is considered a cost of production. Shifted forward, it is allocated
according to the distribution of total
consumer expenditures. . In the case of
property held by businesses and farmers, the tax falling on intangibles is
treated similarly to that on tangibles.
In the case of individuals other than
farmers, the assumption is again that
the tax on intangibles is not shifted.




3 23

The allocation, in this case, is based
upon the S.R.C. distribution of liquid
asset holdings shown in line (6) of
Table 1.
8. Tax Allocation: Social Security Taxes
Payroll taxes, including the contribution of employers and employees, are
covered in this study along with other
taxes. There may be some doubts
whether such taxes should be included
in a study of this kind. First, it may
be argued that payroll tax receipts are
linked contractually to benefit payments, so that our concept of differential incidence cannot be applied. Second, it may be argued that payroll taxes
are traded on a quid pro quo basis
against benefit payments, the entire
process being more in the nature of an
insurance purchase than a tax. While
there is some validity in both points,
neither is conclusive.
The link between payroll tax receipts
and benefit payments is rather legalistic,
and alternative methods of finance
might well be considered. Moreover,
the assumption of a quid pro quo relationship between contributions and
benefit payments is subject to considerable qualification. Not only are the
contributions mandatory, but there are
substantial redistributional elements in
the benefit formulas. The very notion
of an employer contribution is incompatible with the quid pro quo interpretation. From the employer's point
of view the payroll tax is no less a cost
of doing business than any other tax
levied upon the use of resources. A
possible solution might be to include
only the employer contribution, but
this presumes that the latter is not
shifted to the employee. While the
whole problem of expenditure treat-

24

NATIONAL TAX

ment is particularly acute in connection
with the payroll tax, it seemed preferable to exclude it. The reader who
wishes to consider the results net of
payroll tax may do so readily from the
data presented in section III.
Shifting Assumptions.—Of total payroll taxes collected in 1948, about twothirds were contributed by employers
and one-third by employees.
Depending upon circumstances, the
employer contribution may be reflected
in reduced profits, lower wages, or
higher prices. The presumption for a
competitive system, however, is that the
entire burden is borne by wage earners.
Imposition of a payroll tax on the employer raises labor costs; the gross wage
rate (including tax) is raised without
there being a corresponding increase in
productivity. Unless workers are willing to accept a corresponding cut in
wages, producers will substitute other
factors of production and less labor will
be employed. On the assumption that
the supply of labor is inelastic, wage
rates will fall by the amount of tax.
While it is conceivable that the tax may
in some instances result in reduced
profits, the profit effect is likely to be
one of redistribution from more to less
labor intensive industries.
But this is an over-simplified view.
Once a more realistic view is taken, we
may point to a number of reasons why
complete backward shifting is unlikely.
Suppose for a moment that production
requires the combination of all factors
in fixed proportions. In this case, the
producer cannot avoid the payroll tax
by substitution of untaxed factors.
The tax, in effect, becomes a general excise assessed on the total cost of production, and as such we may expect it to
be shifted forward. While the com-




JOURNAL

[VOL. I V

bination of factors is not fixed in practice, this assumption is less absurd than
it may appear. Since the payroll tax
applies fairly generally to all types of
labor, only a general substitution of
capital for labor can be resorted to.
This is not possible in the short run.
In the longer run such substitution may
occur, but we may expect that replaced
labor will be reabsorbed in capital goods
industries. And public policy, in any
case, is not likely to permit prolonged
unemployment to result; absorption of
such labor through an upward adjustment of prices is the more likely outcome.
Unions, moreover, may cite the " intent of the law " in resisting employer
attempts to recoup their contribution
by wage reduction. Considerations of
these and other institutional factors
suggest that forward shifting to the
consumer is more likely than backward
shifting to the wage recipients. In view
of the uncertainties involved, we again
make use of alternative assumptions.
Three cases are considered. In Case A
the employer is taken to shift one-third
of his payments to wage earners and
two-thirds to consumers. This is the
standard assumption. The two other
cases present limiting assumptions, Case
B providing for complete backward and
Case C for complete forward shifting.
The employee contribution is assumed to fall on wage earners in all
three cases. This follows the usual
practice, although it is by no means
evident why the incidence of this part
of the tax should differ from that of the
employer contribution. A different
treatment of the employee contribution
may be explained only on psychological
grounds. Just as the unions may cite
the " intent of the law " in their sup-

No. 1]

TAX PAYMENTS BY INCOME GROUPS

port when resisting backward shifting
of the employer contribution, the public mores are against unions that attempt to recover their own contributions through higher wages. While it
seems desirable, in principle, to apply
alternative assumptions to both parts
of the payroll tax, the additional results do not justify the effort. The
burden distributions resulting from allocation on the basis of consumer expenditures are rather similar to those
obtained by allocation on the basis of
wage receipts.
Allocation of Tax.—The part of the
tax which falls upon wage earners is
allocated according to an estimated distribution of covered wage income by
spending unit income brackets. In order to obtain this distribution, income
received by spending units had to be
divided into wage and other income;
and wage incomes, in turn, had to be
divided into covered and uncovered
wages. The resulting estimate shown in
line (17) of Table 1 again involves a
good deal of conjecture.
The part of the tax which falls upon
consumers was allocated according to
the distribution of total consumer expenditures. An attempt to assign part
of the tax to specific expenditure items
accounted for too small a fraction of
the total to be justified.
9. Tax Allocation: Estate, Inheritance,
and Gift Taxes
Available data did not permit a
breakdown of estate, inheritance, and
gift tax payments among income brackets above $7,500. In view of the high
level of exemptions permitted under
both Federal and state succession taxes,
it seemed reasonable to assume that all
but a negligible part of the yield should




3 25

be imputed to income recipients in the
top bracket*
The allocation of the yield of succession duties to high income people as
a group avoids considerable difficulties
which arise when an attempt is made to
allocate the incidence of death duties to
particular tax payers. 8 Suppose that
Mr. Jones is a typical taxpayer with an
income of $50,000 and that an average
taxpayer at this level of income is endowed with an estate of, say, $400,000.
If Mr. Jones had died this particular
year, his death duties would have
amounted to about $115,000. But,
being an average taxpayer, Mr. Jones
did not die during the year under consideration. Yet, as the average taxpayer, he must look forward to a corresponding tax liability which will arise
at the time of his death. This may be
allowed for by imputing to him a death
tax payment equal in amount to the
premium payable on an annuity sufficient to discharge his death duty at
the time of his death. This premium
may be estimated at about 7 per cent of
income.
No such computation had to be made
here. Since it is reasonable to assume
that succession duties were paid wholly
by spending units with incomes in excess of $7,500, this yield may be related
to the income of this group and it may
be concluded that spending units in this
group paid about 2 per cent of their
incomes in such taxes. For taxpayers
as a group, the two methods are, however, quite compatible. If instead, we
had computed the premiums (as defined above) payable by all taxpayers
8
See Nicholas Kaldor, " The Estimation of the
Burden of Death Duties," in G. F. Shirras and L.
Rostas, The Burden of British Taxation (New York,
1943), pp. 80-90.

26

NATIONAL TAX

in this group, the totals of such premiums would, under certain simplifying assumptions, have to equal the actual yield total obtained.
III. D I S T R I B U T I O N O F T A X P A Y M E N T S
U N D E R STANDARD

ASSUMPTIONS

Keeping in mind the conceptual and
statistical limitations set forth in the

JOURNAL

[VOL. I V

1. Distribution of Total Tax Payments
The over-all distribution of tax payments, combining total taxes paid at all
levels of government, is shown in Table
6 and Chart I. Lines (1), (2), and
(3) of Table 6 give the dollar amounts
of total tax payments contributed by
spending units within the various in-

TABLE 6
DISTRIBUTION OP TAX PAYMENTS BY INCOME GROUPS: SUMMARY*

(Money amounts in millions of dollars)
Spending Unit Income Brackets (Thousands of dollars)
Item

fyaenrd

Total

6,754
2,061
8,815

17,010
3,542
20,552

43,794
13,552
57,344

11.6
12.9
11.9

15.4
15.2
15.4

38.8
26.1
35.8

100.0
100.0
100.0

19.3
6.7
3.5
3.2
26.0

21.1
6.4
3.4
3.0
27.6

30.1
6.3
3.4
2.8
36.3

22.3
6.9
3.6
3.3
29.2

26,283 31,953 56,542
13.4
28.8
16.3

196,619
100.0

Under 1

1-2

2-3

3-4

4-5

5-7.5

Amounts
(1) Federal
(2) State and local
(3) All levels

620
433
1,052

2,243
1,123
3,366

5,392
2,131
7,523

6,682
2,512
9,194

5,079
1,755
6,834

Per cent of yield total
(4) Federal
(5) State and local
(6) All levels

1.4
3.2
1.8

5.1
8.3
5.9

12.3
15.7
13.1

15.3
18.5
16.0

16.2
8.1
3.9
4.2
24.3

18.6
7.3
3.7
3.7
25.9

19.0
7.1
3.7
3.5
26.1

Per cent of income
16.5
(7) Federal
11.6
(8) State and local
5.8
а. State
5.8
б. Local
28.1
(9) AH levels
Addenda
(10) Income received . . . 3,747
1.9
(11) Per cent of income t
(12) Per cent of spending
12.2
units'

13,850
7.0

29,037 35,207
14.8
17.9

17.7

22.9

20.1

11.6

10.2

7

5.3

100.0

•Standard assumptions throughout.
t Includes income imputed under standard corporation assumption.
Note: Details may not add to totals because of rounding.

preceding chapters, let us now turn to
the quantitative results of the study.
In this section we examine the results
obtained on the basis of our standard
assumptions.
Other results derived
from alternative assumptions will be
considered in the final section.




come brackets. Lines (4), (5), and
(6) show the corresponding percentage
distributions of total tax payments, and
lines (7) to (9) give the dollar
amounts as percentages of income received in the respective brackets. The
latter ratios, which relate taxes paid to

No. 1]

TAX PAYMENTS BY INCOME GROUPS

income received, provide the most interesting part of the evidence.1 Com1
Our concern is with differences between effective
rates at different levels of income rather than with
the average level of effective rates as shown in the
last column. While the latter reflects the total level
of income and tax payments, the former depends
upon their distribution only. While there are some
doubts about the proper income total to be used,

3 27

monly referred to as " effective rates of
tax," these ratios show whether the tax
structure is proportional, i.e. tax payments as a percentage of income are
the same at all income levels; progresthese doubts do not apply to the underlying distribution.

8
in
<r
\

\

X
<
H

co
f - H

is

O Lj
>
hO
LLI

LI.
L_
L
LU

-J
<
o
o
-J
Q
2
<
L
J
1<
h_J
<
t
r
L
U
Q
LI
L
U.

*<

-J
<
1-

^

^ — ^ —

^

\

</)
L
U
X

tn
L
U

<n X
L
U
X

\

1
1
•
1
1
•
— V —

£

_J
<
c
r
L
U
Q
L
U

<
1<
o
o

-J
C
O
L
U

h-

<
ho
I- Li- J c/>

©® ©




CM

(3IN00NI JO 1N30 U3d SV XVI) 31VU 3AIJL03dd3

28

NATIONAL TAX

sive, i.e. tax payments are an increasing
percentage of income when moving up
the scale; or regressive, i.e. the tax to
income ratio declines with rising income.2
As will be seen from line (9) of
Table 6, the tax-to-income ratio for
1948 is regressive at the bottom of the
income scale and then remains more or
less proportional up to about the $5,000
income level. It becomes progressive
thereafter. Or, looking at the percentage distribution of tax payments, we
find that the lowest 30 per cent of
spending units, which receive approxi-*
mately 9 per cent of the income, contribute about 8 per cent of the yield
total, the tax-to-income ratio moving
regressively over this range. The next
54 per cent of spending units, which receive about 46 per cent of the income,
contribute some 41 per cent of the yield
total. The tax-to-income ratio, over
the middle range, remains more or less
proportional. The top 15 per cent of
spending units, finally, receive 45 per
cent of the income and contribute 51
per cent of the total yield, the tax-toincome ratio being decidedly progressive over this range.
Special caution is called for in interpreting the picture for the lowest income bracket. There is reason to believe that spending units in this bracket
2
The concepts of " progressive," " regressive," and
" proportional" as defined above, are unambiguous,
and for the greater part of our discussion these concepts are all that will be needed. When reference is
made to the degree of progression or regression, matters are more complicated. These terms may then be
defined in a number of different ways. Where reference is made in the present paper to the "degree"
of progression, it is measured as percentage change
of tax divided by percentage change of income. This
is the concept of liability progression as defined in
R. A. Musgrave and Tun Thin, " Income Tax Progression, 1929-48," Journal of Political Economy,
LVI (December, 1948), 504.




JOURNAL

[VOL. I V

are composed, not only of genuinely
low income units, but also of others
which have incurred temporary losses
but are normally in a higher bracket.
This results in a higher rate of expenditures or dividend receipts—and hence
greater tax payments—for the group as
a whole than is likely to apply to low
income units more permanently within
this bracket. To some extent, this argument applies to other brackets as
well, but the resulting distortion may
well be more pronounced in the bottom
bracket. This calls for care in interpreting the appearance of regressivity
at the bottom of the scale.
The reader who has followed our exposition of the difficulties inherent in
the preparation of such estimates will
not be overly impressed with minor
fluctuations in the estimated tax-toincome ratio. Also, he will bear in
mind that the implications of alternative assumptions remain to be tested.
Nevertheless, the general pattern shown
in line (9) of Table 6 should be reasonably accurate. The over-all tax
structure is by no means as progressive
as is generally surmised, at least not as
far as the lower 90 per cent of the taxpayers are concerned. father, the
effective rate curve follows a U-shaped
pattern with regression at the lower
end, a proportional range over the middle and progression at the upper end of
the scale. Over a wide range of incomes, including 90 or more per cent of
spending units, the progressive elements
of the tax structure appear to be balanced or outweighed by others which
are proportional or regressive. As will
be seen in section IV, this result follows
the general pattern of other studies.
The important thing for the individual taxpayer, as well as for public

No. 1]

TAX PAYMENTS BY INCOME GROUPS

policy, is the distribution of payments
(and hence the tax-to-income ratios)
under the composite tax structure and
not the separate distribution of payments under one or another particular
tax. But, only by considering how tax
payments for the various components
of the revenue structure are distributed,
can it be understood how the over-all
pattern of distribution comes about and
what may be done to adjust it.
2. Distribution of Federal vs. State
and Local Tax Payments
The first breakdown to be considered
is between levels of government. A
comparison of lines (7) and (8) of
Table 6 indicates a distinct difference in
the distribution of Federal and state and
local tax payments*
The degree of initial regression in the
distribution of state and local payments
is pronounced and is subject to little
doubt. The distribution of state and
local payments continues to be regressive throughout.
Some degree of regression at the bottom of the income scale appears for the
Federal as well as for the state and local
system. However, the decline of the
Federal ratio is very slight and falls well
within the margin of error which must
be applied to these results. This is followed by a distinct element of progression around the $2,000 income level,
where the personal income tax makes
itself felt. Moving up the scale, we
then find a slightly progressive distribution of payments up to the $5,000
range, with distinct progression following at the top of the scale. The progression in the top brackets, of course,
reflects almost entirely the personal income tax.
Looking again at the percentage distribution of tax payments, as shown in




3 29

lines (4) and (5) of Table 6, we find
that both the Federal and state-local
revenue structures derive about 27 to
28 per cent of their yield totals from
incomes in the $4,000 to $7,500 range.
But, whereas the Federal Government
derives 39 per cent from incomes above
$7,500 and 33 per cent from incomes
below $4,000, the state-local system
draws 46 per cent from incomes under
$4,000 and 26 per cent from the over
$7,500 group. And, while the Federal
Government obtains only 6.5 per cent
of its yield from incomes under $2,000,
the share of state-local revenue contributed by these brackets amounts to
11.5 per cent.
It is thus evident that the progressive element in the combined revenue
structure is provided by Federal taxes,
whereas the distribution of state-local
tax payments is proportional or regressive. While no separate study was
made of the distribution of state as distinct from local tax payments, a rough
estimate of this kind is shown in lines
(8a) and (8b) of Table 6? It appears
that the two patterns are quite similar.
However, regression in the distribution
of state taxes is more distinct at the beginning of the range and not so great at
the upper end of the scale as in the distribution of local taxes. As shown below, this reflects differences in the distribution of excise and property tax
payments.
3. Distribution of Federal Tax
Payments by Type of Tax
The distribution of Federal tax payments by type of tax is shown in Table
7. The distribution for some major
3
In breaking down line (8) into (8a) and (8b),
the distribution of state excise and property tax payments is assumed to be the same as the distribution
of similar tax payments to local governments.

30

NATIONAL TAX

taxes is also shown in Chart II. As
shown in lines (13) to (18) of Table
7, the major progressive element is the
personal income tax. This tax provides
nearly one-half of the yield total and
results in a progressive distribution of

JOURNAL

[VOL. I V

Little need be said in explanation of
the personal income tax distribution as
shown in line (13) of Table 7. The
reader will recall that the tax-to-income
ratios shown apply to the income of
spending rather than taxpaying units.

TABLE 7
DISTRIBUTION OF FEDERAL TAX PAYMENTS BY INCOME GROUPS *

(Money amounts in millions of dollars)
Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1
Amounts
(1) Personal income tax ..,
(2) Corporation income tax .
(3) Excises
(4) Payroll taxes
(5) Estate and gift taxes ...
(6) Total
Per cent of yield total
r (7) Personal income tax ..
(8) Corporation income tax
(9) Excises
(10) Payroll taxes
(11) Estate and gift taxes ..
(12) Total
Per cent of income
(13) Personal income tax ..
(14) Corporation income tax
(15) Excises
(16) Payroll taxes
(17) Estate and gift taxes ..
(18)

Total

1-2

2-3

3-4

4-5

5-7.5

7,5 a n d

over

Total

9
273
226
112

469
712
714
348

1,543
1,331
1,394
1,133

2,328
1,582
1,700
1,072

2,201
1,166
1,117
595

3,537
1,466
1,292
459

...

7,958
6,388
1,467
297
900

18,045
12,920
7,912
4,015
900

620

2,243

5,392 6,682

5,079

6,754

17,010

43,794

0.1
2.1
2.9
2.8

2.6
5.5
9.0
8.7

8.5
10.3
17.6
28.2

12.9
12.2
21.5
26.7

12.2
9.0
14.1
14.8

19.6
11.3
16.3
11.4

1.4

5.1

12.3

15.3

11.6

15.4

44.1
49.4
18.5
7.4
100.0
38.9

100.0
100.0
100.0
100.0
100.0
100.0

0.2
7.3
6.0
3.0

3.4
5.1
5.2
2.5

5.3
4.6
4.8
3.9

6.6
4.5
4.8
3.0

8.4
4.4
42
2.3

11.1
4.6
4.0
1.4

...

14.1
11.3
2.6
.5
1.6

9.2
6.6
4.0
2.0
&

21.1

30.1

22.3

...

...

. • • ...

...

...

...

...

...

...

...

...

...

...

16.5

16.2

18.6

19.0

19.3

...

•Standard assumptions throughout.
Note: Details do not necessarily add to totals because of rounding.

payments over the entire range to
which it applies. Excise taxes are regressive throughout, as may be expected.
The corporation income and payroll
taxes show a fluctuating pattern up to
the $5,000 range, above which the former becomes distinctly progressive and
the latter distinctly regressive. The
estate and gift taxes are highly progressive but do not carry much weight in
the total yield picture.




Because of the presence of more than
one income recipient in the typical
spending unit, the splitting of income
between spouses, and other factors, the
pattern of income tax payments is less
progressive when allocated by spending
units than when allocated by taxpaying
units. Also, it is evident that our data
understate the degree of progression applicable to high incomes, since there is
no breakdown of incomes above $7,500.

(3IAIOONI JO 1N30 d 3 d SV XVI) 31VH
3AI103JJ3




32

NATIONAL TAX

The indicated distribution of corporation tax payments seems puzzling
at first sight. As shown in line (14)
of Table 7, the distribution is distinctly
regressive up to the $2,000 to $3,000
range, remains more or less proportional
up to the $7,500 level, and becomes
progressive only thereafter. In order
to understand this pattern, the assump-

[VOL. I V

JOURNAL

retained earnings after tax plus the
share of the tax falling on profits are
similarly imputed to shareholders' income. An explanation of the over-all
pattern of corporation tax payments
may be found by examining the respective patterns applicable to the various
shifting components. This is shown in
Table 8.

TABLE 8
DISTRIBUTION OP FEDERAL CORPORATION TAX PAYMENTS FOR SEPARATE
SHIFTING COMPONENTS*

(Money amounts in millions of dollars)
Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1

1-2

157
39
77

403
148
161

779
314
238

900
381
301

623
284
259

695
259
511

748
190
5,451

4,306
1,615
6,998

(4) Total
273
Per cent oj income
(5) Forward-shifted part . . . . 4.2
(6) Backward-shifted part .. 1.0
(7) Shareholders' part
2.1

712

1,331

1,582

1,166

1,466

6,388

12,920

2.9
1.1

2.7
1.1
&

2.6
1.1
.9

2.4
1.1
1.0

2.2
.8
1.6

1.3
.3
9.6

22
&
3.5

(8) Total
Per cent of yield total
(9) Forward-shifted part . . . .
(10) Backward-shifted part ..
(11) Shareholders' part
(12) Total

71

5.1

4.6

4.5

4.4

4.6

11.3

6^5

3.0
2.4
1.1
2.1

9.4
9.2
2.3
5.5

18.1
19.4
3.4
10.3

20.9
23.6
4.3
12.2

14.5
17.6
3.7
9.0

16.1
16.0
7.3
11.3

17.4
11.8
77.9
49.4

100.0
100.0
100.0
100.0

Amounts
(1) Shifted forward
(2) Shifted backward
(3) On shareholders

12

2-3

3-4

4-5

5-7.5

Total

•Standard case.

tions underlying this allocation of corporation tax payments must be recalled.
Line (14) of Table 7 is based on our
standard assumptions which specify
that one-third of the tax is shifted to
consumers and one-eighth to wage earners, the remainder falling on profits.
This entire remainder, including the
part coming out of retained earnings
as well as the part coming out of dividends, is imputed to shareholders; and




'the distribution of the forwardshifted part of the corporate tax is based
largely on the general distribution of
consumer expenditures. Accordingly,
it is sharply regressive at the bottom of
the scale and continues regressive
throughout the income range. As may
be seen by comparison of line (5) of
Table 8 with line (15) of Table 7, this
distribution differs somewhat from the
distribution of excise payments, which

No. 1]

TAX PAYMENTS BY INCOME GROUPS

rests more largely upon the expenditure
pattern for specific consumption items.
However, the difference is not substantial.4
The backward-shifted part of the tax
is allocated in accordance with wage income. As shown in line (6) of Table
8, it follows a more or less proportional
pattern up to the higher brackets, where
it becomes regressive. This is as may be
expected, since capital income tends to
become a rising fraction of total income when moving up the scale. The
weight of the backward-shifted part in
the total is slight.
The distribution of the shareholders'
part may appear somewhat surprising.
As will be seen from line (7) of Table
8, this part of the tax is distinctly regressive at the bottom of the scale.
Since the allocation of this component
is based upon the estimated distribution
of dividend receipts, it reflects a decline
in the ratio of dividend to other income
when moving from the first to the second bracket. 5 Regression at the lower
end of the scale appears in the part falling on profits as well as in the forward4
Had the forward-shifted part been distributed in
accordance with the pattern of Federal excises as
given in line (9), Table 7, the effective rates for
the first three brackets as shown in line (15) of
Table 7 would have been 3.2, 4.3, and 2.6 per cent
respectively. For the higher brackets the results
would have differed little from those shown in line
(5) of Table 8. See also Table 14.
5
Dividend receipts as percentage of spending unit
income are estimated at 2.3, 1.3, 0.9, 1.4, 1.1, 1.8,
and 10.9 per cent for the seven brackets respectively. This result, it should be noted, is not peculiar
to our data but appears in even stronger form in the
distribution of dividend receipts by taxable incomes
as shown in the Preliminary Statistics of Income for
1947, Part I, Treasury Release 32171, November 25,
1949, p. 8. As noted previously, an explanation may
be found in the hypothesis that the bottom group
includes a large number of rentiers and of normally
higher bracket people who have suffered temporary
losses.




3 33

shifted part of the tax. While the
former element is more sharply regressive at the bottom, regression in the
latter extends further up the scale and
weighs more heavily in the total. Both
contribute to the [/-shaped form of the
effective rate curve for total corporation tax payments. Distributions of
corporation tax payments for alternative assumptions will be considered below.6
The allocation of excise tax payments
is shown in lines (9) and (15) of Table
7. While the general pattern is more
or less regressive, there are a few progressive components. The distribution
of payments by components allocated
6
A further complication of the corporation tax
distribution .arises from the interrelationship of corporation and personal income tax liabilities. The
complication applies only to the unshifted part of the
corporation tax.
When an additional corporation tax is imposed, the
additional liability incurred by the shareholder is
equal to the additional corporation tax to be imputed
to him, minus such reduction in his personal income
tax liability as results from curtailment of his dividend income because of the increased corporation tax.
These personal income tax savings will be the higher,
per dollar reduction of dividend income, the higher
is the dividend recipient's income bracket and hence
his marginal rate under the personal income tax.
The incremental corporation tax is thus regressive
per dollar of reduced dividend income. This offsets,
in part, the progressive tendency of the corporation
tax which results because (with the exception of the
bottom range) dividends tend to increase as a share
of income as we move up the income scale.
This personal income tax offset does not enter into
determining lines (3) and (7) of Table 8 as in lines
(3) and (7) we are concerned with the allocation of
a given yield total including corporation and personal
income taxes. At the same time, line (7) of Table
8 does not accurately reflect the additional liability
which would result when adding the corporation tax
to the other taxes; there would, in effect, be a partial
offset due to lowered personal income taxes.
The difficulty might be overcome, for present purposes, by considering the sum of lines (13) and
(14), Table 7, as the joint personal and corporation
tax liabilities. In considering the joint result of the
two taxes, we avoid the difficulties that arise when
considering either of the taxes separately.

NATIONAL TAX JOURNAL

34

according to various spending patterns
is shown in Table 9, details of the components being described in Table 4
above.
Among the major elements in the
excise tax picture, the distribution of
tobacco taxes is distinctly regressive
through the entire income scale. The
distribution of liquor taxes, on the contrary, is progressive at the bottom of

[VOL. I V

Note also the sharp initial element of
regression in the tax component allocated according to automotive expenditures. As is apparent from these data,
various excises show distinctly different
payment patterns.
The distribution of payroll tax payments as shown in line (16) of Table
7 follows an irregular pattern up to the
third income bracket but becomes dis-

TABLE 9
DISTRIBUTION OF FEDERAL EXCISE TAX PAYMENTS FOR SEPARATE SHIFTING COMPONENTS:
TAX PAYMENTS AS PERCENTAGE OF INCOME

Allocated According
to Expenditures on:
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)

Tobacco
Auto transport
Other transport
Recreation
Food
Liquor
Rent
All other
Total consumer expenditures
Total

Total
Amount
Allocated
(Millions
of dollars)

Spending Unit Income Brackets (Thousands of dollars)
Under 1 1-2

5-7.5 7.5 and
over

2-3

8-4

4-5

2
a

1.0
.3
.1
.3
.1
1.3
.2
.3

0.9
.3
.1
.3
.1
1.4
.2
.3

0.7
.4
.1
.3
.1
1.1
.2
.3

0.6
.4
.1
.4
.1
1.1
.1
.2

1,312
601
110
576
197
2,177
303
463

1.4
.5

2,173

1.8

1.2

1.1

1.0

1.0

.9

.5

.9

7,912

~5S

19

Z5

4J6

To

3S

Z5

3.8

*

.3
.2
.5
.3
.5

1.1
.3
.1
.2
.2
1.4

Total

0.3
.2
*

.2

*

.8
.1

2

0.7
.3
.1
.3
.1
1.1
.2

2

* Less than 0.05 per cent.

the scale; it becomes more or less regressive thereafter, but less so than most
other components.
The component
which is not allocated to specific expenditure categories, but is distributed
in accordance with the general pattern
of consumer expenditures, shows the
sharpest degree of regression at the bottom of the scale. This reflects the
occurrence of dis-saving at low levels of
income, characteristic of available data
on income-expenditure relationships.7
7
This remains a rather puzzling matter. As noted
before, the expenditure-income relationship at the
lower end of the scale, and hence the distribution




tinctly regressive thereafter. This is
as may be expected, since only the first
$3,000 of wage income for individual
earners within any spending unit was
subject to payroll tax. The element of
regression when moving from the first
to the second bracket and the subseof excise payments, may well be distinctly different
for spending units which are permanently in the
lower income range than for others who find themselves in the low income position due to temporary
misfortune and who maintain more nearly customary
standards of living by drawing on their savings. T o
the extent that this is the case, it should be allowed
for in interpreting our excise data for the low income
range.

TAX PAYMENTS BY INCOME GROUPS

No. 1]

quent occurrence of progression when
going from the second to the third
bracket must again be explained in
terms of the distributions applicable to
the various shifting components. These
are shown in Table 10.
The distribution of payments is again
based on our standard assumption,
which in this case postulates that twothirds of the employer contribution is
shifted to consumers while the remain-

3 35

4. Distribution of State and Local Tax
Payments by Type of Tax
A detailed analysis of state and local
tax payments by type of tax is shown
in Table 11. The personal income tax
again provides the progressive element,
but its contribution to the yield total is
too slight in this case to leave a significant impact on the combined pattern of
tax payments. The corporation income

T A B L E 10
DISTRIBUTION OF PAYROLL TAX PAYMENTS FOR SEPARATE SHIFTING COMPONENTS t

(Money amounts in millions of dollars)
Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1 1-2
Amounts
..
(1) Part falling on wages
(2) Part falling on consumers . ..

2-3

3-4

4-5

5-7.5 7.5 and
over

Total

50
62

200
148

846
287

745
327

366
229

204
255

20
277

2,431
1,584

,
(3) Total
Per cent of income
(4) Part falling on wages
(5) Part falling on consumers . ..

112

348

1,133

1,072

595

459

297

4,015

1.3
1.7

1.4
1.1

2.9
1.0

2.1
0.9

1.4
0.9

0.6
0.8

0.5

12
0.8

..

1u)

2.5

3.9

3.0

^3

Ti

~0J5

2.0

(6)

Total

•

tCase A (standard).
* Less than 0.05 per cent.
Note: Details do not necessarily add to totals because of rounding.

ing one-third, as well as the entire employee contribution, falls on wage earners. As will be seen from Table 10, the
forward-shifted part is distributed according to the general expenditure pattern and hence is regressive throughout
the scale. The part falling on wages is
progressive up to the $3,000 income
level, because covered wages are an increasing fraction of income up to this
level. The wage share subsequently becomes regressive, since covered wages
decline as a fraction of income received.




tax distribution is identical with that at
the Federal level, but its quantitative
importance is much less* The distribution of excises is generally similar to
that shown at the Federal level, although the degree of regression at the
bottom of the scale is more marked.
This is as may be expected, since a
larger part of state and local excises are
of the sales tax variety, and taxes on
luxury and semi-luxury items are less
important than in the Federal structure.

36

NATIONAL TAX

The new and most interesting factor
in the distribution of state-local tax
payments is the property tax.
As
shown in line (16) of Table 11 and in
Chart II, the ratio of property tax payments to income is distinctly regressive
at the bottom of the scale and con-

[VOL. I V

JOURNAL

The item carrying the greatest
weight is the part of the tax falling
on owner-occupied residences. It is
slightly regressive throughout and reflects the declining importance of housing expenditures when moving up the
scale. Next in importance is the tax

T A B L E 11
DISTRIBUTION OP STATE AND LOCAL TAX PAYMENTS BY INCOME GROUPS

(Money amounts in millions of dollars)
Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1 1-2
Amounts
(1) Personal income taxes
(2) Corporation income taxes..
(3) Excises
(4) Property taxes
(5) Inheritance taxes
(6) Total
Per cent of yield total
(7) Personal income taxes
(8) Corporation income taxes..
(9) Excises
(10) Property taxes
(11) Inheritance taxes
(12) Total
Per cent of income
(13) Personal income taxes
(14) Corporation income taxes..
(15) Excises
(16) Property taxes
(17) Inheritance taxes
(18)

Total

39
545
539

9
72
1,056
994

32
86
1,267
1,127

37
63
886
769

91
79
1,003
888

...

331
346
1,121
1,547
197

499
699
6,093
6,064
197

433

1,123

2,131

2,512

1,755

2,061

3,542

13,552

2.1
3.6
3.3

5.6
8.9
8.9

1.8
10.3
17.3
16.4

6.4
12.3
20.8
18.6

7.3
9.0
14.5
12.7

18.3
11.3
16.5
14.6

3.2

8.3

15.7

18.5

12.9

15.2

66.3
49.5
18.4
25.5
100.0
26.1

100.0
100.0
100.0
100.0
100.0
100.0

*

.3
3.9
3.9

.2
3.6
3.4

0.1
.2
3.6
3.2

0.1

.4
5.8
5.4

...

0.3
.2
3.1
2.8

...

0.6
.6
2.0
2.7
.3

0.3
.4
3.1
3.1
.1

7.3

7.1

6.7

6.4

6.3

6.9

1L6

...

...

...

...

...

5-7.5

F,RAND
over

3-4

15
217
201

4-5

7

2-3

...
2

3.4
2.9

Total

* Less than 0.1 of 1 per cent.
Note: Details do not necessarily add to totals because of rounding.

tinues to be somewhat regressive
throughout the income scale. This
over-all pattern again must be explained in terms of the distributions of
payment applicable to the various components. This is shown in Table 12
where various parts of the tax are combined according to their underlying
patterns of allocation.8




assessed on personal property, which is
regressive, especially at the beginning of
the scale. As shown in Table 5, this
component combines various elements
allocated according to different patterns. However, it appears that the
initial element of regression may be
8

For explanation of these components see Table S.

TAX PAYMENTS BY INCOME GROUPS

No. 1]

traced to the shares falling on business
and farm intangibles, which are taken
to be reflected in production costs and
allocated according to consumer expenditures. The share of the tax falling on business land reflects the pattern
of dividend distribution, while that on
farm and rental land reflects the pattern of rental income distribution.

37

with the exception of the part falling
on rental improvements show some degree of regression at the lower end of
the scale. Moreover, all components
with the exception of the part falling
on business, farm, and rental land value
continue to be regressive from the
second bracket up. The pattern, it appears, is fairly uniform. Except for

T A B L E 12
DISTRIBUTION OF PROPERTY TAX PAYMENTS FOR SEPARATE COMPONENTS *

(Yield in millions of dollars)
Percentage of Income
Item

Total
Yield

Spending Unit Income Brackets (Thousands of dollars)
Under 1 1 - 2

2-3

3-4

4—5

5-7.5

and
over

7.5

Total

846
(1) Land: farm and rental . . .
322
(2) Land: business
(3) Improvements: business
and farm
860
(4) Improvements: rental
823
(5) Owner-occupied residences 2,000
(6) Personal property
1,213

1.07
.11

0.49
.05

0.29
.04

0.28
.04

0.27
.05

0.32
.07

0.68
.44

0.43
.16

.88
.72
1.44
1.15

.58
.81
1.27
.69

.54
.68
1.16
.72

.51
.52
1.21
.65

.47
.43
1.06
.64

.43
.38
.94
.64

.27
.13
.76
.46

.44
.42
1.02
.62

(7) All components

57~

3IT

3.4

6,064

2S~

2T

3T

•Standard case.
Note: Details do not necessarily add to totals because of rounding.

Both items are regressive at the bottom
of the scale. Both become progressive
above the $4,000 range; the tax allocated to business land becomes highly
progressive at the upper end of the
scale. The two remaining items,
finally, are both regressive throughout
the entire scale. The distribution of
tax payments falling on business and
farm improvements reflects the excise
pattern, while that of tax payments
falling on improvements on rental real
estate is based on the pattern of housing expenditure.
In short, all the shifting components




the impact of the tax on land upon the
higher income ranges, it appears that
the combined pattern shown in line
(16) of Table 11 may be considered
fairly reliable. This conclusion is supported by the testing of alternative
property tax assumptions in the following section.
5. A Measure of Differential

Incidence

In the introductory section we concluded that a proper determination of
incidence could be in the form of either
(1) an absolute measure of the com-,
bined incidence of both the revenue

38

NATIONAL TAX

and expenditure sides of the budget, or
(2) a differential measure of tax incidence. We are here concerned with
tax incidence only, but up to this point
have approached the problem in absolute rather than differential terms.
In order to translate our results into a
measure of differential incidence the

[VOL. I V

JOURNAL

a whole; but as noted before, lines (7)
through (9) of that table present only
a rough approximation to differential
incidence. The results are based on the
assumption that the same income before
tax applies for both tax systems, and
this is hardly an acceptable simplification.

T A B L E 13
MEASURE OF DIFFERENTIAL INCIDENCE : ALL LEVELS OF GOVERNMENT COMBINED *

Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1 1-2
Estimated Actual Distribution
(1) Tax payments in millions
1,052
of dollars
(2) Percentage of tax pay1.8
ments
(3) Tax payments as per cent
28.1
of income
Hypothetical Proportional
Distribution
(4) Tax payments in millions
of dollars
1,093
(5) Percentage of tax pay1.8
ments
(6) Tax payments as per cent
29.2
of income
Differential Distribution
(7) (1) - (4)
(8) (2) - (5)
(9) ( 3 ) - ( 6 )

- 41
-0.1
-1.1

2-3

3-4*

4-5

5-7.5

3,306

7,523

9,194

6,834

8,815

20,552

57,344

5.9

13.1

16.0

11.9

15.4

35.8

100.0

24.3

25.9

26.1

26.0

27.6

36.3

29.2

4,039

8,469

10,268

7,665

9,319

16,490

57,344

7.0

14.8

17.9

13.4

16.3

28.8

100.0

29.2

29.2

29.2

29.2

29.2

29.2

29.2

- 9 4 6 -1,074
1.9
- 1.7 3.1
- 3.3 -

-831
- 1.5
- 3.2

-504
- 0.9
- 1.6

+ 4,062
+ 7.0
+ 7.1

-673
- 1.1
- 4.9

over

Total

•Standard assumptions throughout. Lines (1) to (3) are taken from Table 6. Line (4)
is obtained by allocating ihe total of $57,344 million according to distribution patterns shown in
line (11) of Table 6. A similar procedure may be applied to obtain the respective hypothetical
distributions for the Federal or state-local taxes.
Note: Details do not necessarily add to totals because of rounding.

estimated absolute incidence of the
actual tax structure may now be compared with that of an alternative structure. For reference purposes, let us
consider a structure under which the
same yield total is raised but the money
burden is distributed proportionally to
income before tax. Table 13 provides
such a comparison for the tax system as




Lines (1) through (3) repeat the estimated actual distribution of tax payments as shown in lines (3), (6), and
(9) of Table 6. Lines (4), (5), and
(6) show what the pattern of tax payments would have been had the same
yield been distributed proportionally.
Lines (1) and (4) show the dollar
amounts of the tax; lines (2) and (5),

No. 1]

TAX PAYMENTS BY INCOME GROUPS

the percentage distribution of the tax;
and lines (3) and (6), the tax as a
per cent of income. Lines (7), (8),
and (9) indicate the differential between the two systems, obtained by
deducting the reference distribution
from the actual distribution.
It is apparent that spending units
with incomes up to $7,500 would bear
a heavier tax load under a proportional
tax system than under the present distribution. Spending units in the top
income bracket, on the other hand,
would bear a smaller tax load under a
proportional system.
IV. T E S T I N G T H E S I G N I F I C A N C E O F
ALTERNATIVE ASSUMPTIONS

The distribution of tax payments
presented in the preceding section was
based on what appears to us the " most
likely " set of assumptions. However,
these assumptions involve highly speculative judgements and we do not wish,
in any way, to be dogmatic about them.
In order to obtain some notion of the
weight carried by the choice of one or
another hypothesis and to accommodate
the reader whose ideas of " most likely "
differ from ours, we now turn to the
distribution of tax payments under a
number of alternative hypotheses.
1. Corporation Income

Tax

The testing of alternative assumptions is most important and interesting
with regard to the corporation income
tax. Not only does this tax contribute
a substantial portion of the yield, but
its incidence has also been subject to
much controversy. As shown in Table
3, six cases are considered. They combine three shifting assumptions with
two treatments of retained earnings and
the share of tax falling thereon.




3 39

The assumptions applying to each of
the six cases are listed in Table 3, and
the resulting distributions of corporation tax payments are shown in Table
14. In lines (1) to (6) of Table 14
corporation tax payments are given as
a percentage of income. As shown in
the last column, total corporation tax
payments as a percentage of total income differ for each case. Whereas
cases A, C, E, and F account for the
entire tax yield, the part of the tax
falling on retained earnings is disregarded in cases B and D. These cases,
therefore, show a lower average level of
rates. Further differences in average
rate level arise because taxes paid out of
retained earnings, as well as retained
earnings after tax, are imputed to the
shareholders' income in cases A, C, and
E, but not in cases B, D, and F. When
comparing the tax distributions for the
various cases, the reader should keep in
mind that there is a difference in the
average level of effective corporation
tax rates, as well as in the degree of
progression or regression over various
parts of the income range.
For this reason, a comparison of the
results obtained under the various assumptions is difficult. It is facilitated
in some respects by reference to the
corresponding percentage distributions
of corporation tax payments as shown
in lines (7) to (10) of Table 14.1
1
As will be seen from Table 14, the distributions
are similar for cases C and D and for E and F. This
is obvious in the case of E and F, because the entire
tax is assumed to be shifted forward. Distributions
are similar for C and D because the additional tax,
accounted for in C, is distributed in the same way
the tax on dividends is distributed in both C and D .
The distributions are dissimilar for A and B. While
the same shifting assumptions apply in both cases, the
inclusion of the tax on retained earnings in A raises
the fraction of the total that is allocated according
to the dividend pattern over that of case B.

NATIONAL TAX JOURNAL

40

These distributions may be compared
without reference to differences in the
various yield totals accounted for;
however, the comparison thus provided
is somewhat defective because it disregards the imputed income aspect of
the problem.

[VOL. I V

for these three cases. Since the average
effective rates shown in the last column
of Table 14 are rather similar, the comparison may be drawn largely in terms
of effective rates.
In Case C the total payment is assumed to come out of profits. The ef-

T A B L E 14
DISTRIBUTION OF CORPORATION TAX PAYMENTS FOR ALTERNATIVE ASSUMPTIONS

Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1 1-2
Corporation Tax Payments as
Per Cent of Income *
(1) Case A (standard)
(2) Case B
(3) Case C
(4) Case D
(5) Case E
(6) Case F
Per Cent of Corporation Tax
Payments Contributed *
(7) Case A (standard) . . . .
(8) Case B
(9) Cases C and D
(10) Cases E and F
Total Tax Payments as
Per Cent of Income t
(11) Case A (standard)
(12) Case B
(13) Case C
(14) Case D
(15) Case E
(16) Case F

2-3

3-4

4-5

5-7.5

over

Total

7.3
6.2
3.7
1.2
13.5
13.5

5.1
4.5
2.1
.7
8.8
9.0

4.6
4.1
1.5
.5
8.1
82

4.5
4.0
1.6
.5
7.7
7.9

4.4
3.8
1.8
.6
7.2
7.3

4.6
3.6
2.9
.9
6.6
6.8

11.3
6.1
16.3
6.9
4.4
5.5

6.6
4.5
6.4
2.1
6.8
7.3

2.1
2.7
1.1
3.7

5.5
7.5
2.3
9.3

10.3
14.5
3.4
18.1

12.2
17.1
4.3
20.9

9.0
12.3
3.7
14.5

11.3
13.8
7.3
16.1

49.4
32.1
77.9
17.4

100.0
100.0
100.0
100.0

28.1
28.1
24.0
22.7
34.6
36.0

24.3
24.2
20.9
20.1
28.4
29.0

25.9
25.9
22.5
22.0
29.8
30.3

26.1
26.1
22.9
22.3
29.7
30.2

26.0
25.9
23.0
22.4
29.1
29.7

27.6
27.6
25.5
24.6
30.1
31.1

36.3
39.2
39.8
38.9
31.9
40.0

29.2
29.3
28.3
26.5
30.3
32.6

* Includes Federal corporation tax only.
t Includes all taxes and all levels of government. Based on standard payroll tax assumptions.
Note: Details do not necessarily add to totals because of rounding.

1. Let us first compare the results of
the two alternative shifting assumptions with those of the standard assumption used in Case A. For the purpose, let us consider only those cases in
which retained earnings and the tax
thereon are allowed, that is, cases A, C,
and E. Lines (1), (3), and (5) show
tax payments as percentage of income




fective rate curve follows the U shape
found to prevail in Case A under the
standard assumption. Indeed, the initial element of regression is somewhat
stronger than in the standard case,
especially when moving from the
second to the third bracket. 2 This is
2
According to our definition (see note 2, section
III), the initial element of regression in line (3) of

No. 1]

TAX PAYMENTS BY INCOME GROUPS

so because dividend receipts as a percentage of income fall off more sharply
at the lower end of the income scale
than do expenditures as a percentage of
income. As in Case A, there follows a
proportional middle range, but progression seems to make a somewhat earlier
appearance. Most important, the degree of progression at the upper end of
the scale is considerably sharper than in
the standard case. Whereas the effective rate paid in the highest bracket
under Case C is much above that paid
in the standard case, the effective rate
paid by all other brackets is considerably less. The same picture is obtained
also by comparing lines (7) and (9).
Whereas spending units in the top
bracket contribute 78 per cent of the
yield total in Case C, they contribute
only 50 per cent in Case A.
Case E, where the total corporation
tax payment is assumed to fall entirely
on consumers, follows the excise pattern
and is regressive throughout. Accordingly, the effective rate is much higher
at the bottom and much lower at the
top of the income scale than in Case A.
The contribution of spending units in
the top bracket, in this case, is reduced
to 17 per cent, as against 50 per cent
under the standard assumption.
It is apparent from these comparisons
that the application of alternative shifting assumptions results in drastically
different distributions of the tax payment. The corporation tax, it appears,
is an altogether different proposition
when one shifting assumption is made
instead of another. 8 Moreover, since
the corporation tax carries considerable
weight in the yield total, the resulting
Table 14 is stronger than in line (1) because the
ratio of 3.7 to 2.1 is larger than the ratio of 7.3 to
5.1.




3

41

distributions for the revenue structure
as a whole are also significantly different for the different assumptions.
This may be seen by comparing lines
(11), (13), and (15) of Table 14.
While the general U shape of the effective rate schedule continues to apply
in all cases, it is somewhat more marked
in Case C than, for example, in Case E.
Whether the reader will consider
these divergent results as seriously invalidating the picture presented in the
preceding section depends on his views
regarding shifting. Even though he
may not agree with the assumptions
underlying our standard case, the reader
should bear in mind that Cases C and E
represent extreme and unlikely assumptions.4
3
It is evident from the distributions shown in
lines (2) and (16) of Table 1 that differences in
assumptions with regard to forward as against backward shifting are of rather minor importance. The
important factor is how much of the tax is shifted
and how much falls on profits.
4

The reader may wish to consider alternative
shifting assumptions as follows:
Case 1.—Let us assume that he wishes to consider
retained earnings and the tax on these, as in the
standard case, but wishes to change shifting assumptions. He may then redistribute the total shown in
the last column of Table 8 to suit his preferences
and then allocate the new dollar totals by brackets
on the basis of lines (9), (10), and (11) of Table
8. He may then obtain a new line (4) in Table 8
by adding vertically.
This new line (4) may then be expressed as a percentage of the income shown in line (10) of Table
6. The results obtained will not be quite correct,
however, as these income figures provide for imputations which will differ from those properly applied under other assumptions.
In order to obtain the appropriate income, the
reader may take the difference between his allocation
into line (3) of Table 8, added horizontally, and the
standard case allocation of $6,998 million. This difference will then be allocated by income brackets in
accordance with the percentage distribution shown in
line (11) of Table 8. If the allocation to dividends
is greater than in the standard case, the bracket distribution thus obtained equals the additional imputed
income and is added to the standard case income dis-

42

NATIONAL TAX

JOURNAL

[VOL. I V

2. Let us now consider the differences may also be seen by comparing lines (7)
which result from varying our assump- and (8), which show the percentage of
tions regarding the treatment of re- yield contributed by spending units in
tained earnings and of the tax thereon. the top bracket to be considerably
To bring out this aspect of the prob- smaller in Case B than in Case A.
Cases. C and D, where the entire tax
lem, let us hold the shifting assumption
is assumed to fall on profits, involve
constant and compare Cases A and B,
sharply different yield totals and hence
C and D, and E and F.
For Cases A and B, i.e. the standard sharp differences in the average level of
shifting assumption, the results are effective rates. Since the percentage
shown in lines (1) and (2) of Table distribution of tax payments as shown
14. Since the yield total accounted for in line (9) is the same for both cases,
in Case A is larger than in Case B, the the difference in the pattern of proaverage level of effective rates is accord- gression and regression is due entirely
ingly higher. However, the degree of to the difference in the imputation of
regression shown at the beginning of income. Leaving aside the difference in
the scale is rather similar in the two the average level of effective rates, we
cases, as is the proportional middle find the general pattern of rates to be
range. The only significant difference very similar for the two cases. Tlje
in the distribution of payments appears only significant difference is the lesser
to be the flattening of progression at degree of progression at the top of the
the upper end of the income scale in income scale which arises in Case D.
Case B. This reflects disregard of the This may be expected since disregard
tax on retained earnings. The same of the retained earnings reduces tax liabilities by a greater fraction than income before tax.
tribution shown in line (10) of Table 6. If the allocation to dividends is smaller than in the standard
Cases E and F, finally, allow for comcase, the bracket distribution must be deducted.
plete forward shifting. Here the averCase 2.—Now assume that the reader wishes to
age levels of effective rates are more
disregard retained earnings and the tax on them. In
nearly comparable. Again the differthis case only a smaller yield total will be accounted
for. Suppose that $6,920 million is assumed to be
ences in effective rate pattern are due
shifted forward or backward and $6,000 million to
entirely to differences in income imfall on profits. Using the 3:7 ratio between diviputation, the percentage distributions
dends and retained earnings, $1,800 million of this
$6,000 million falling on profits is taken to fall on
of tax payments being the same for
dividends and is substituted for $6,998 million in
both cases. As will be seen from a comline (3) of Table 8, the total being reduced correparison of lines (5) and (6) of Table
spondingly from $12,920 million to $8,720 million.
On this basis a new line (4) of Table 8 is obtained
14, both cases result in a regressive rate
as in Case 1.
structure throughout the income scale,
Again this may be related to income as shown in
the degree of regression being slightly
line (10) of Table 6, but this will be quite unsatisfactory. In order to obtain the correct income, line
more pronounced in Case E, where re(10) of Table 6 must be adjusted. This line allows
tained earnings are imputed.
for imputation of $13,200 million of retained earnOn the whole, it appears that the difings, plus $6,988 million of tax on profits. Imputation in the present illustration should equal $1,800
ferences in the pattern of effective rates
million. Thus the difference, or $18,398 million,
arising from the change in the treatshould be allocated according to line (11) of Table
ment of retained earnings and of the
8 and be deducted from,line (10) of Table 6.




No. 1]

TAX PAYMENTS BY INCOME GROUPS

tax thereon are not very significant;
certainly, they are much less significant
than those arising from the variation of
shifting assumptions. The difference is
largely one of average rate level and
merely expresses the fact that different
yield totals are accounted for under the
two procedures.
2. Excise Taxes
As noted in section II, the allocation

3

43

ber of B.L.S. surveys. Since the latter
derivations may involve a considerable
margin of error, it seemed desirable to
compare the results obtained by use of
the detailed data with the distribution
which results if all excises are allocated
in accordance with the pattern of total
consumer expenditures. The alternative results, in this case, reflect differences in estimating procedure rather

T A B L E 15
DISTRIBUTION OF EXCISE TAX PAYMENTS FOR ALTERNATIVE SPENDING PATTERNS:
TAX PAYMENTS AS PERCENTAGE OF INCOME

Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1

1-2

2S

3-4

4-5

5-7.5

7.5 and
over

Total

6.0
5.8

5.2
3.9

4.8
3.6

4.8
3.6

4.2
3.4

4.0
3.1

2.6
2.0

4.0
3.1

. 1L8
Total
General Expenditure Pattern f
8.2
Federal
6.3
State and local

91

8£

SA

7£

7\2

4~6

71

5.3
4.1

4.9
3.8

4.6
3.6

4.3
3.3

4.0
3.1

2.4
1.9

4.0
3.1

14.5

9.4

87

83

77

71

7.3

71

Detailed Expenditure Pattern5¥
Federal
State and local

Total

* Allocated according to pattern shown in Table 4. This allocation used for excise distribution in all other tables.
t Allocated wholly according to line (2), Table 1.
Note: Details do not necessarily add to totals because of rounding.

of excises, wherever possible, was made
in accordance with expenditures on specific consumption categories. Only that
part of the excises which could not be
allocated in this fashion was distributed
in accordance with the pattern of total
consumer expenditures. This general
pattern, as shown in line (2) of Table
1, was obtained directly from the S.R.C.
data, whereas the specific expenditure
patterns had to be derived from a num-




than in shifting assumptions. The results are shown in Table 15, where the
two excise tax distributions are compared. It appears that the element of
regression at the bottom of the scale is
somewhat sharper if the general expenditure pattern is used throughout.
However, the over-all distribution of
excise payments is very similar for the
two cases. Similar conclusions apply
under either method.

44

[VOL. IV

NATIONAL TAX JOURNAL

3. Payroll Taxes
The distribution of payroll tax payments, shown in section II, was based
on the assumption that two-thirds of
the employer contribution is shifted
forward to consumers, whereas the remainder, as well as the entire employee
contribution, is taken to fall on the
wage earner. In Table 16 the estimated

While these differences in the distribution of payments are obviously significant, the payroll tax contribution to the
yield total was relatively slight in 1948,
so that the over-all picture is not
affected greatly by the variation of payroll tax assumptions. This may be seen
in lines (7) to (9) of Table 16. With
the growing weight of payroll taxation
in the future, uncertainty regarding

T A B L E 16
DISTRIBUTION OP PAYROLL TAX PAYMENTS FOR ALTERNATIVE ASSUMPTIONS

Item

Payroll Tax Payments as
Per Cent of Income
(1) Case A (standard)
(2) Case B
(3) Case C
Percentage of Payroll Tax
Payments Contributed
(4) Case A (standard)
(5) Case B
(6) Case C
Total Tax Payments as
Per Cent of Income *
(7) Case A (standard)
(8) Case B
(9) Case C

Spending Unit Income Brackets (Thousands of dollars)
~
Under 1 1-2
2-3
3-4
4-5
5-7.5 7 f a n c l
Total

3.0
2.2
4.6

2.5
2.6
2.7

3.9
4.4
2.5

3.0
3.6
2.4

2.3
2.4
2.2

1.4
1.1
2.0

0.5
.1
12

2.0
2.0
2.0

2.8
2.1
3.9

8.7
8.9
9.3

28.2
32.0
18.1

26.7
31.5
20.6

14.8
16.0
14.4

11.4
8.7
16.1

7.4
.7
17.5

100.0
100.0
100.0

28.1
27.3
29.3

24.3
24.4
24.5

25.9
26.4
24.5

26.1
26.7
25.4

26.0
26.2
25.9

27.6
27.2
28.2

36.3
35.9
37.1

29.2
29.2
29.2

* Standard corporation tax assumptions.
Note: Details do not necessarily add to totals because of rounding.

distribution of payments is shown for
two alternative assumptions. It is assumed in Case B that the entire tax falls
on wages and in Case C that the entire
tax falls on the consumer.
As would be expected, the distribution is more regressive at the lower end
of the income scale in Case C than in
the standard case. The distribution is
progressive at the lower end of the income scale in Case B, with regression appearing only above the $4,000 level.




payroll tax shifting will be of increasing importance. In time it may create
a difficulty comparable to that presented
by the corporation tax.
4. Property Tax
Certain alternative assumptions for
the property tax remain to be examined.
As set forth in Table 5, these relate to
the allocation of the yield total among
different shifting components, rather
than to the shifting assumptions appli-

TAX PAYMENTS BY INCOME GROUPS

No. 1]

cable to each of these components. In
the allocation of the yield total for Case
B, doubtful allocations were resolved in
each instance so as to place as large a
part of the burden upon the upper income groups as appeared reasonably
possible, while the opposite policy was
followed in determining the allocations
for Case C. 6

3 45

is more pronounced than in the standard case. This reflects the allocation of
a smaller fraction of the total yield to
the land component. As a result, the
fraction distributed according to the
pattern of dividend receipts and rental
income is reduced, while the fraction
distributed according to the excise pattern is increased.

T A B L E 17
DISTRIBUTION OF PROPERTY TAX PAYMENTS FOR ALTERNATIVE ASSUMPTIONS

Spending Unit Income Brackets (Thousands of dollars)
Item
Under 1 1-2
Property Tax Payments as
Per Cent o] Income *
(1) Case A (standard)
(2) Case B (progressive) . . .
(3) Case C (regressive) . . . .
Percentage oj Property Tax
Payments Contributed
(4) Case A
(5) Case B
(6) Case C

5.36
5.28
5.47

3.89
3.71
4.18

2-3

3-4

4-5

5-7.5

over

Total

3.42
3.25
3.63

3.20
3.07
3.34

2.93
2.84
3.01

2.78
2.76
2.82

2.74
3.02
2.40

3.08
3.08
3.08

3.31

8.89

16.39

18.58

12.68

14.64

25.51

100.0

3.26
3.38

8.46
9.54

15.56
17.39

17.82
19.40

12.28
13.05

14.52
14.85

28.09
22.38

100.0
100.0

28.08
28.00
28.18

24.30
24.12
24.59

25.91
25.74
26.12

26.11
25.99
26.26

26.00
25.91
26.09

27.59
27.57
27.63

36.35
36.63
36.01

Total Tax Payments as
Per Cent oj Income t
(7) Case A
(8) Case B
(9) Case C

29.16
29.16
29.16

* Includes state and local property tax payments.
t Includes all taxes and all levels of government. Based on standard corporation and payroll
tax assumptions.
Note: Details do not necessarily add to totals because of rounding.

The results are shown in Table 17.
As may be seen from a comparison of
lines (1) to (3), resulting differences
are very slight. The degree of regression at the lower end of the income
scale is about the same for all cases.
The standard case, as noted before, is
regressive throughout. In Case B some
slight progression appears at the upper
end of the scale. In Case C regression
5
For detailed assumptions see appendix notes to
Table S.




Except for the appearance of some
degree of progression at the upper end
of the income scale in Case B, it appears
that variations in property tax payments, introduced by the use of alternative shifting assumptions, are very
slight. The inadequacy of the data on
which to base the allocation of the total
property tax yield among specific components does not seem to do serious
damage to the reliability of the estimated distribution of property tax pay-

46

NATIONAL TAX

ments. As shown in lines (7) to (9)
of Table 17, the distribution of total
payments is not seriously altered by the
use of alternative property tax assumptions.
This concludes our analysis of alternative assumptions. The reader may
now combine the effective rate schedules for particular taxes obtained under
any of the assumptions in Tables 14 to
17 and thus derive alternative over-all
schedules of tax payments comparable
to those shown in Tables 6 and 7.
Drastic differences in the over-all picture, however, will arise only from
rather sharp changes in the shifting assumptions with respect to the corporation tax. The experimentation with
alternative methods, except for this tax,
appears to strengthen rather than
weaken the confidence which may be
placed in the general pattern of tax distribution under the standard case.
5. Comparison with Earlier Studies
Finally, our results may be compared
briefly with the earlier Colm-Tarasov
estimate for 1938-39 and with the Adler estimate for 1946-47.® A comparison of effective rates for the combined
Federal, state, and local tax systems for
the three studies is shown in Chart III
and Table 18.
(a) Differences in Level of Effective
Rates.—The average effective rates for
the three studies are 20.2 per cent for
the Colm-Tarasov, 24.2 per cent for the
Adler, and 29.2 per cent for the present
study (standard assumptions). This
6
Helen Tarasov, Who Does Pay the Taxes? (Supplement IV, Social Research, 1942). Reference is to
the revised figures for 1938-39 as given on p. S.
Also, see John Adler, " The Fiscal System, the Distribution of Income, and Public Welfare," in Kenyon
E. Poole (ed.), Fiscal Policies and the American
Economy (New York: Prentice-Hall, Inc., 1951).




JOURNAL

[VOL. I V

wide divergence results from the use of
both different tax totals and income
bases. The difference between our average effective rate and that of the
Colm-Tarasov estimate needs no detailed explanation, as it reflects largely
the higher level of tax rates in the later
period. But a reconciliation with the
Adler ratio is required.
Our tax total amounts to $57.3 billion as against $49.5 billion for the Adler study. This is due in part to differences in the yield total for the two time
periods and in part to the fact that the
tax total used in this study includes
accrued income taxes whereas the Adler
figure is based on collections. Part of
the difference in the income base is
again due to the difference in time
period, but the remainder is due to a
difference in the underlying sources of
income estimate. The Adler study uses
the estimate for personal income by the
Department of Commerce, whereas the
present study uses the S.R.C. income
total. In order to facilitate a comparison with the Adler study, our effective
rate schedule was recomputed, using the
Department of Commerce income estimate. On this basis, the average effective rate for our standard case is reduced from 29.2 per cent to 24.7 per
cent. In Chart III and Table 18 our
result for the standard case is shown on
both the S.R.C. and the Department of
Commerce income base. As far as imputations to income are concerned,
there appear to be no major differences
between Adler's method and ours.
(b) Difference in Shape of Effective
Rate Curves.—The shape of the effective rate curve, as noted previously, depends considerably upon the shifting
assumptions made with regard to the

No. 1]

TAX PAYMENTS BY INCOME GROUPS

corporation income tax. As the other
studies assume that there is no shifting
of this tax, they should be compared for
statistical compatibility with our noshifting assumption (Case C) rather
than with our standard case. Accordingly, Case C, adjusted to the Commerce income base, is also included in

3

47

as the two periods considered and the
two tax structures involved are rather
similar. The source material and the
description of methods presented in the
Adler study do not permit a detailed
and careful comparison, but it appears
that the differences in the pattern for
the tax system as a whole are due to

T A B L E 18
COMPARISON OF EFFECTIVE TAX RATES FOR VARIOUS, STUDIES

(In per cent)
Spending Unit Income Brackets (Thousands of dollars)
Study
Under 1 1-2
Colm-Tarasov *
Adler t
Case A, unadjusted
Case A, adjusted to commerce
income
Case C, adjusted to commerce
income

2-3

3-4

4-5

5-7.5

7 5and

IT ~
over

Total

18.0
19.6
28.1

17.5
15.1
24.3

17.4
17.3
25.9

17.7
17.7
26.1

18.2
22.9
26.0

18.7
24.2
27.6

32.7
36.3
36.3

20.2
24.2
29.2

23.6

20.3

21.6

21.8

21.7

23.1

31.7

24.7

20.2

17.5

18.8

19.1

19.3

21.4

35.1

24.0

* For 1938-39. See Helen Tarasov, Who Does Pay the Taxest (Supplement IV, Social Research, 1942), p. 5.
tFor 1946-47. See John Adler, "The Fiscal System, the Distribution of Income, and Public
Welfare," in Kenyon E. Poole (ed.), Fiscal Policies arid the American Economy (New York:
Prentice-Hall, Inc., 1951).

the chart and table. The average effective rate for Case C, adjusted to Department of Commerce income, is 24.0
per cent, or about the same as Adler's.
A study of Chart III and Table 18
shows a fair degree of similarity between the patterns of Case C and the
Colm-Tarasov study. The fact that
progression in our case begins sooner reflects the downward extension of the
personal income tax; the higher level of
rates through the middle income brackets similarly reflects the increased severity of personal income taxation.
Adler's effective rate pattern, on the
other hand, shows marked difference
from that of Case C. This is surprising




numerous differences in the distribution of particular taxes. One major
source of divergence appears to result
from differences in the consumption
patterns used. The Adler study appears
to rely on a smaller sample of B.L.S.
data. Also our study uses a general
pattern of consumer expenditure, derived from the S.R.C. data, while the
Adler study appears to be based on
B.L.S. data only. This may well account for the fact that the pattern
shown in the Adler study is distinctly
more regressive at the bottom of the
scale. The greater progressivity of the
Adler schedule over the $3,000 to $6,000 income range may reflect the alio-

0

48

X
<
LL.
O

\

V
-A-

E
o

T- J»
o
/
>
C ;=• to <
O
> 2
C ^ ^
C I?
= ° °
O I
< < o
C -= ~ 0) 0) 4)v
O P $ C w >
O
O T5 o O O
o o < o u o
CL

cn <D
LU

<
cr (r
<

o
LU

ro
ro

o
ro

C
M

'V-

NATIONAL TAX JOURNAL

C
M

(3IAI00NI JO 1N30 U3d SV XVI) 31VU 3AI103JJ3

^

[VOL. I V

o
o
Q:

m

g
o
o

to
N-

O
O

8
O
O
o
<0

o
o
o

lO

o
o
?

o
o
o

P
O

o
o
o

C
M

O

o
o

UJ
o

o

s

o

z

\tt




No. 1]

TAX PAYMENTS BY INCOME GROUPS

cation of $4 billion of " miscellaneous "
taxes on a proportional basis. Many of
these taxes are of a regressive nature and
have in our study been allocated according to specific expenditure patterns.
6. Conclusion
In view of the above results, we may
reappraise briefly the validity of the incidence picture as presented in section
III. The estimated distribution of personal income tax liabilities should be
rather accurate. While the assumption
that there is no shifting of such liabilities may be open to debate, such shifting as occurred was hardly on a large
scale; moreover, there is no basis on
which to appraise the possible magnitude of such adjustments. The excise
distribution is based on expenditure
patterns which in turn had to be derived from rather incomplete B.L.S.
data; however, the results obtained by
using the more reliable S.R.C. pattern
for total consumer expenditures are
rather similar. This suggests, although
it does not prove, that the errors involved in the excise allocation are not
of major importance.
The distribution of corporation tax
payments provides much the greatest
difficulty; and it is evident that the
shifting assumptions chosen will affect
considerably the shape of the final pattern of tax distribution. A further

3

49

difficulty arises from the inadequacy of
data regarding the distribution of dividend income. The property tax problem is complicated by the dearth of data
for the allocation of shifting components. However, experimentation with
alternative allocations shows less difference in results than might be expected.
Hence, the outcome seems fairly reliable. The allocation of payroll taxes
again involves difficulties with regard
to shifting assumptions; but the differences resulting from alternative assumptions are less serious than for the
corporation tax, and the over-all weight
of the payroll tax is much less. Finally,
a major difficulty of interpretation arises
at the bottom of the income scale, owing to the heterogeneous composition of
the low spending unit income brackets.
At the upper end of the scale no detailed breakdown is possible, since no
adequate data for detailed income
brackets in excess of $7,500 are available.
While these difficulties must be kept
in mind, it appears that the results are
fairly adequate, if we are satisfied to
accept the simplified framework of this
analysis. But this is a big i f . The results become much less satisfactory
when we recognize that any statistical
measure of tax incidence is subject to
the conceptual limitations set forth in
the introductory section.

APPENDIX
1. Notes on Table 1

"When the data shown are carried to more decimal

Data for lines (1) through (6) were derived from
the " 1948 Survey of Consumer Finances," by the
Survey Research Center, University of Michigan.
Most of the data were published in the Federal Reserve Bulletin, June through September, 1949, but
some unpublished data were obtained from the S.R.C.

P l a c e s t h a n a P P e a r i n t h e Published sources, such
additional information was obtained directly from
the S.R.C.
Line 1: F.R.B., July, 1949, p. 780.
Line 2: Determined as a residual. The S.R.C. percentage distributions for money income, savings, and




JO

NATIONAL TAX

Federal personal income tax liability (S.R.C.) were
converted into dollar distributions, using totals of
$176.0, $12.0, and $16.0 billion respectively
(S.R.C.). Dollar savings and tax liabilities for each
income bracket were deducted from dollar income,
and a further deduction was made for state personal
income tax liabilities estimated on the basis of the
Wisconsin law. The residual was then reconverted
into a percentage distribution.
Line 3: A percentage distribution of expenditures
for the purchase of automobiles and a percentage
distribution of expenditures for the purchase of all
other durable goods were obtained directly from the
S.R.C. These were combined into a weighted average, the weights being equal to $11.5 billion for
autos and $7.5 billion for other durables (F.R.B.,
June, 1949, pp. 647 and 650). The S.R.C. distributions were furnished by deciles of spending units
and had to be converted into distributions by spending unit income brackets as used in this study.
Line 4: The S.R.C. data do not show a percentage
distribution of housing expenditures by home owners.
They do show housing expenditures of home owners
as a percentage of their incomes (F.R.B., September,
1949, p. 1048). Since this information was given
only for three broad income brackets, the first step
was to estimate a corresponding distribution for our
seven income brackets. Step number two was to
estimate the dollar income distribution of home
owners. Use was made of S.R.C. data showing the
percentage of spending units in each income bracket
who own their homes (F.R.B., September, 1949, p.
1040). By applying these percentages to the dollar
distribution of money income (see explanation to
line 2 above), a dollar distribution of money income
by home owners was obtained. Step three was to
multiply this money income in each income bracket
by the percentage of income spent on housing, and
thus to obtain a dollar distribution of housing expenditures by home owners from which the percentage distribution shown in line 4 could be derived.
Line 5: The method of derivation was identical
with that for line 4, except that data on renters
rather than home owners were used. See the F.RJ$.,
September, 1949, p. 1040, for the percentage of
renting families in each income bracket, and p.
1050 for the percentage of income spent for housing
by renters. Note that for both lines 4 and 5 the
original data are given for family units. Since ninetenths of the family units are single family spending
units, the uncorrected application of this data to a
spending unit distribution involves no significant
error.
Line 6: F.R3., August, 1949, p. 910.
Lines 7 through 14: To properly allocate the forward-shifted portions of various taxes it was necessary to determine what proportion of expenditures
for certain consumption items was made by each income class. The expenditure pattern developed in
lines 7 through 14 is based on consumer budget




JOURNAL

[VOL. I V

studies conducted by the U. S. Department of Labor
during the years 1946 through 1948. The original
data appear in (a) " Consumer Spending: Detroit,
Denver, and Houston, 1948," Monthly Labor Review,
December, 1949; (b) "Family Income and Expenditures in 1947, Washington, Richmond, and Manchester," Monthly Labor Review, April, 1949; and
(c) " Survey of Prices Paid by Families, Milwaukee,
Savannah, and Scranton," an unpublished report by
the U. S. Department of Labor. These studies conducted during the years 1946 through 1948 were
selected in preference to the more comprehensive
studies undertaken in 1941 because they are more
likely to accurately reflect the changed price and
income relationships of the postwar years. Rural
families are not included in the postwar studies.
Some comparison of rural and urban spending habits,
however, reveals no systematic differences for which
adjustments could be readily made.
No attempt was made to correct for the discrepancy arising out of the fact that this study is
conducted in terms of spending units whereas the
consumer budget studies surveyed family units. Nor
was any attempt made to correct for the fact that
although single person spending units are included in
the S.R.C. distribution of spending units, the B.L.S.
provides useful information only about family units
of two or more persons. Any error introduced by
this nonidentical treatment was not deemed to be
sufficiently significant in this context to warrant
working out a complicated conversion system.
Since a single consumption pattern for the entire
United States is required, the data from these budget
studies were consolidated into a single pattern. For
this purpose the size of each of the cities was determined according to the 1940 census. Each city
pattern was then given a weight equal to the weight
carried by a city of such size in the consumer price
index prepared by the B.L.S. Since the size distribution of cities included in the B.L.S. surveys provided
a fairly representative coverage of the range of cities
included in the index, this weighting system appears
to be fairly adequate. It is evident, however, that
the estimated national expenditure pattern suffers
from the fact that it was based on a rather small
sample.
Line 15: No direct information on the distribution of dividend income by spending units is available. An attempt was made to translate the 1947
dividend distribution by taxable income brackets
(Treasury Press Service Release S-2171, November
25, 1949) into a similar distribution by spending
unit income brackets. Since the Treasury distribution for 1948 was not available at the time of writing, it was necessary to assume that the distribution
for 1948 would be similar to that for 1947.
Step number one was to allocate the money income received in each spending unit income bracket
to the various brackets of taxable income in which
it comes to be declared. This involves making allow-

No. 1]

TAX PAYMENTS BY INCOME GROUPS

ance for income splitting and other factors and is to
a considerable extent a matter of judgment.
The second step was to obtain from this crossclassification the total spending unit income falling
into each taxable income bracket, and to derive for
each taxable income bracket a percentage distribution
of the total by spending unit income brackets.
Step three was to distribute dollar dividend payments by taxable income brackets in accordance with
the 1947 pattern. The total of dividend payments
going to each taxable income bracket was then distributed among spending unit income brackets contributing to the taxable income bracket in accordance with the pattern developed in step two.
The final step was to aggregate for each spending
unit income bracket the dividend payments thus allocated to that bracket and from this to derive a
percentage distribution of total dividend receipts by
spending unit income brackets.
Line 16: On a basis of Statistics of Income data
(Treasury Release S-2171) it is estimated that between 85 per cent and 90 per cent of income below
$5,000 represents wage and salary payments, and
that 67 per cent and '37 per cent respectively would
be appropriate estimates for the two upper income
brackets used in this study. It was assumed, therefore, that 87.5 per cent of incomes up to $5,000
consisted of wages and salaries. An estimate of wage
and salary income by income brackets was obtained
by multiplying these percentages by the money income for each bracket. The figure for each bracket
was then converted to a percentage of the total of
all brackets.
Line 17: The percentage of spending units in each
income class that derive their income from employment covered by social security legislation was estimated using Low Income Families and Economic Stability, by the Joint Committee on the Economic
Report, 1949, p. 86. Those units whose principal
earner was clearly not covered by social security
(e.g., retired, farmers, etc.) were eliminated. The
percentage of units remaining multiplied by the wage
and salary income in each income bracket (see line
16) resulted in a distribution of covered wage and
salary income by income brackets. The figure for
each bracket was then converted to a percentage of
total for all brackets.
In view of substantially higher rates applicable to
railroad retirement and Civil Service retirement plans
these were treated separately and assigned as though
the unit contained a single earner. Government
employees were assumed to bear full burden of Federal retirement plans.
Line 18: Rental income was determined by translating the 1947 rental income distribution by taxable
income brackets (Treasury Press Service Release S2171, November 25, 1949) into a similar distribution by spending unit income brackets. The method
was analogous to that used in the determination of
line 15.




3 51

2. Notes on Table 4
Excise taxes listed in Table 2 were assigned to
specific consumption pattern items. The amounts so
assigned were then allocated to spending units according to the consumption pattern developed in lines
7 through 14 of Table 1. The assignment of excises
according to the consumption pattern classification
is summarized in Table 19. Taxes such as those on
food, alcohol, or tobacco are assigned directly to their
corresponding items in the consumption pattern.
Others involve more difficulty. For example, the
automotive excises cannot simply be assumed to be
borne by consumers in proportion to their expenditures on automotive transportation. A large part of
the tax involves commercial vehicles. This portion
of the tax becomes a cost of production to firms
that use motor vehicles in their business. Thus, a
part of the automotive excises appears in the price
of nonautomotive products. It was estimated, after
some consultation with the research departments of
large automobile manufacturers, that 50 per cent of
the automotive excises are subject to this sort of diffusion. The same ratio was used in the assignment
of excises on electrical goods, insurance companies,
and public utilities. The assignment of communication excises is the same as is made in studies by Mabel
Newcomer (Annals of the American Academy of
Political and Social Science, November, 1949, p. 56).
These same studies suggested a basis for the apportionment of transportation excises. General sales
taxes were distributed among appropriate consumption
items in proportion to the relative importance of
those items as a fraction of total consumer expenditures on taxed items. Customs duties, on the other
hand, were distributed according to the general consumption pattern. It was assumed that 10 per cent
of amusement taxes were paid by business firms. Although there is no evidence to support this, it seemed
reasonable to recognize entertainment as a business
expense.
3. Notes on Table 5
Let us first consider Case A, the standard case.
Of total property taxes of $6,064 million, 80 per
cent or $4,851 million, was assigned to real estate
(line 9), and 20 per cent or $1,213 million was
assigned to the personal property classification (line
16). This division is justified by the fact that the
Bureau of Census shows that in 1940 the assessed
value of real property was 80 per cent of the assessed value of all property (2950 Economic Almanac, NICB, p. 484); and that in 1948 in thirtyseven large cities the assessed value of real property
constituted 81 per cent of the total (Large City
Finances in 1948, City Finances; 1948, No. 3, Washington, D . C., September, 1949, p. 10, Table V ) .
Further, the Department of Agriculture estimates
that in 1948 farm real estate taxes were 85 per cent

NATIONAL TAX JOURNAL

52

of total farm property taxes (1950 Economic Almanac, NICB, p. 190).
The real estate component of the property tax,
$4,851 million, was further divided into those parts
falling on farms (lines 1 plus 2), business property
(lines 3 plus 4 ) , rented homes (lines 5 plus 6), and
owner-occupied homes (lines 7 plus 8).
Of the $4,851 million tax on real property, $651
million was assumed to be a tax on farm land. This

[VOL. I V

remainder, $217 million, was assigned to improvements (line 2).
After appropriate parts of the $4,851 million tax
on real property had been assigned to farm, rented
homes, and owner-occupied homes, there remained
$965 million which was allocated to business property. This remainder of $965 million is simply a
residual; no data are available for determining it directly. It was assumed that one-third of the real

T A B L E 19
ALLOCATION OF EXCISE TAXES BY CONSUMPTION PATTERN ITEMS

(In millions of dollars)
Assignment to Consumption Pattern Items
Excises by.
Industry

To- J ^ 1 0 £ t h e r Recrebacco T p ™ n f
ation

, T.
F o o d Ll( uor
*

Hous- All
ing other

General
Con
-

™+ ,
Total

Pattern
Federal Taxes
Alcohol
Tobacco
Automotive
Amusement
Transportation
Communication
Luxury
Electrical
Food
Other
Customs
Total
State and Local
General sales
Automotive
Alcohol
Tobacco
Amusement
Insurance
Public utility
Other
Total

2,177

i29
410

1,312
548

10

53

ioo

82
170
iis

1,312

601

44
.•

994

365

110

576

2,177

784

303

38

i32

463
1,033

01

19
101
133
642

1,899
1,989
545
365
184
202
267
642

1, 34

1,890

6,093

995

165
134

409

165

784

583

7,912

463

545

estimate is based on Department of Agriculture data
(1950 Economic Almanac, NICB, p. 190). It was
further assumed that two-thirds of the real estate
tax on farm land falls on land and one-third on improvements. This is identical with the Tarasov estimate, and is substantiated by data in Studies in Income and Wealth, vol. 12, (NBER, 1946), p. 196.
On this basis $434 million of the $651 million falling on farms was assigned to land (line 1), and the




197

2,173

639
64
457
342

576

i33

2,177
1,312
1,279
640
610
512
463
265
115
129
410

134

estate tax on nonfarm land falls on land itself and
that two-thirds falls on improvements. This again
is identical with the Tarasov estimate. Thus, $322
million was assigned to business land (line 3), and
$643 million was assigned to improvements on business property (line 4 ) .
Of the $4,851 million, $1,235 million was assumed
to fall on rented homes. This estimate is consistent
with S.R.C. data, which show that somewhat lest

No. 1]

TAX PAYMENTS BY INCOME GROUPS

than 50 per cent of all homes are rented (Federal
Reserve Bulletin, September, 1949, p. 1040), and that
therefore the tax on rented homes should be less
than on owner-occupied residences. It is also reasonable to presume that rented homes are less valuable than owner-occupied homes, since the former
tend to be occupied by spending units in the lower
income brackets {ibid,).
The $1,23J million was
then divided so that one-third or $412 million fell
on land (line 5), and two-thirds or $823 million fell
on improvements (line 6).
On the basis of an S.R.C. estimate, $2,000 million
of the real property tax was assigned to owner-occupied homes (F.R.B., September, 1949, p. 1046).
Again one-third or $670 million was assumed to fall
on land (line 7), and the remainder, $1,330 million,
was assumed to fall on improvements (line 8).
The division of the tax on personal property
among its several components presented the same sort
of problem as did the division of the tax on real
property. The task was complicated by the fact that
even less information was available on which to base
estimates of the proper breakdown.
Of the $1,213 million representing the tax on personal property, the Department of Agriculture estimates that $115 million was paid by farmers (1950
Economic Almanac, NICB, p. 190). There was no
basis for determining how much of this $115 million
was properly a tax on tangibles, and how much on
intangibles. In the absence of information, an equal
break of this amount between tangibles and intangibles seemed the best procedure.
The remainder of the $1,213 million tax on personal property was arbitrarily divided equally between business and individuals, $549 million being assigned to each. It was again assumed in each case
that half the tax was on tangibles and half on intangibles. Thus $275 million was assigned to each
of these categories (lines 10, 11, 14, and 15). The
breakdown between tangibles and intangibles in the
case of the tax on business firms is not important,
since the entire tax is assumed to be shifted forward
to consumers. On the other hand, the breakdown
between business firms and individuals is important.
Of the tax falling on individuals the part assigned to
tangibles was assumed to be distributed on the same
basis as the ownership of durable goods and the part
assigned to intangibles to be distrbuted the same as




3 53

liquid assets. This is a more progressive allocation
than that for the portion of the tax assumed to fall
on business firms.
Cases B and C, the progressive and regressive cases
respectively, differ from the standard case in the assumptions regarding the distribution of the real property tax between rented homes and business property
and in the distribution of the personal property tax
between business firms and individuals. In addition,
different assumptions are made regarding the amounts
of real property tax assignable to land and improvements and in the amounts of the personal property
tax assignable to tangibles and intangibles.
In the progressive case, Case B, $93 5 million of
the property tax was assumed to fall on rented
homes, and $1,265 million was assumed to fall on
business property. In both instances 40 per cent was
assigned to land and 60 per cent to improvements.
Thus, of the $93 5 million falling on rented homes,
$374 million was assigned to land (line 5) and $561
million to improvements (line 6). Of the $1,265
million falling on business property, $506 million was
assigned to land (line 3) and $759 million to improvements (line 4 ) .
Instead of dividing the residual part of the personal property tax equally between individuals and
business firms as in the standard case, in the progressive case $3 50 million was allocated to business
firms and $750 million was allocated to individuals.
Of the $3 50 million falling on business firms, onehalf or $175 million was assumed to fall on tangibles and one-half on intangibles (lines 10 and 11).
Of the $750 million falling on individuals, 30 per
cent or $225 million was assumed to fall on tangibles (line 14), and the remainder, $525 million, to
fall on intangibles (line 15).
In the regressive case $1,535 million of the real
property tax was assumed to fall on rented homes,
while only $665 million was allocated to business
property. In both instances only 20 per cent was
assumed to be assignable to land and 80 per cent
to improvements. The personal property tax was
divided so that $750 million fell on business firms
and only $3 50 million on individuals. Again, the
part paid by business firms was divided equally between tangibles and intangibles. Of the $3 50 million
paid by individuals, 70 per cent was assumed to fall
on tangibles and 30 per cent on intangibles.