United States. Congress. Senate. Committee on the.

The Industrial reorganization act. Hearings, Ninety-third Congress, first session [-Ninety-fourth Congress, first session], on S. 1167 (Volume pt. 7) online

. (page 100 of 140)
Online LibraryUnited States. Congress. Senate. Committee on theThe Industrial reorganization act. Hearings, Ninety-third Congress, first session [-Ninety-fourth Congress, first session], on S. 1167 (Volume pt. 7) → online text (page 100 of 140)
Font size
QR-code for this ebook


generally adopt all IBM standards including interface specifications with the
CPU, channel, or controller to which they are attached, and design their
products to be compatible with IBM stoftware. Both the switching cost and
the uncertainty about performance are eliminated, leaving a clear decision based
on price and the customer's degree of confidence that the competitor can per-
form according to his promises. The minicomputer segment of the market pro-
duces relatively simple products which are more easily compared than larger
computer systems. The CPU's and i)eripherals are often sold separately and
thus each can be compared with its competition rather than having to compare
the entire package. E'xtremely rapid growth in the minicomputer segment of the
market has meant that competition has been primarily for new customers with
little emphasis on luring away the established customers of another
manufacturer.

High capital requirements form a barrier to entry because of the differential
cost of borrowing funds between new and established companies, putting the
new company at a competitive disadvantage. Besides the higher cost of capital
for a new company, the necessary capital may not be available at all. Higli
capital costs are a particularly important barrier to entry when combined with
product differentiation. Product differentiation increases the time between
initial production and profitability while the new company tries to overcome
the customers' preference for established firms. This both increases the amount
of capital necessary (to pay for extensive losses as well as production facilities)
and makes it more difficult to raise capital because of the adverse effects on the
company's stock price of showing a loss for several years.

In the computer industry, capital requirements are correlated with economies
of scale and product differentiation. In the minicomputer and plug compatible
peripheral segment of the market where software and product differentiation
are of relatively minor importance, the initial capital requirements for entry
are extremely low. The two minicomputer leaders. Digital Equipment and Data
General, each began with initial capital of under $100,000, and then raised more
with public stock offering for expansion. IB^NI's analysis of a hypothetical new
entrant into the plug compatible memory business concluded that a viable com-
pany could be established using $75,000 initial capital from the founders, a
private placement the second year, and a public stock offering the third year.'^
Although companies in this category may experience difficulty raising needed
expansion funds on the stock market from time to time, the capital requirements
for entry are so moderate that they cannot be considered a barrier to entry.

In the business systems market, capital requirements are an almost insur-
mountable barrier to entry. A firm must offer a variety of products including
CPU's and a vride range of peripherals, together with extensive software and
support to be taken seriously in the business systems market. The necessity
of offering many products combined with the economies of scale that exist in
software makes a large firm necessary for efficiency. In addition, the new entrant
finds a reluctance to buy his product because of the conversion cost and poten-
tial disruption to the customer's operation if the new company should fail to
deliver as promised or go out of business. Consequently, he must offer large
discounts over the established companies in order to attract new business, and
incur substantial losses until his reputation is established. Because the total
capital costs are dependent upon how long it tcikes to achieve respectability in

"For a (lisonssion of the problem of computer standards, see Gerald Brock, "Competi-
tion. Standards, and Self-Re?ulatinn in the Computer Indnstr.v." in Caves and Roberts
(ed.). "Reffuhiting the Firm" (Washington. D.C. : The Brooliings Institution, forth-
cominir). . , . „ -r «,s

^2 IBM. :Memorandum of P. C. Vilandre to L. S. Halperin, "Memory Analysis" .Tune 29,
1970, p. 8, Plaintiff's Exhibit 279 in Telex vs. IBM.



5689

the eyes of consumers rather than just the cost of a plant and equipment, no
single figure for capital costs can be given. However, a good indication of their
magnitude can be gained from the companies who have entered.

RCA seriously entered the computer industry with the RCA 501 in June,
1959 and never achieved sustained profitability. The company took a pre-tax
write-off of $490 million on its computer operations after its exit in September,
3971. Robert Saruoff, RCA chairman, estimated that an additional $500 million
investment would have been required to achieve profitability, suggesting total
capital requirements of more tnan a billion dollars over a period of several
years for effective entry.^'' Similarly, G.E. entered the industry in July, 1959,
with the G.E. 210 and exited (through merger with Honeywell) in 1970 without
achieving profitability. G.E.'s actual investment in the computer business is not
available but from tiie market share figures, was probably as large as RCA's.
Although Uuivac brought out the first commercial electronic computer, the com-
pany did not achieve sustained profitability until 19GG, causing a long term
drain on capital to finance production facilities and losses. Assuming the figure
of $1 billion for effective entry is of the right order of magnitude, entry is
closed to all but the largest established corporations. A new company could not
hope to raise that amount of capital, especially if it was showing continuous
losses for five or more years, and even a large established corporation would
have to be concerned with the impact of an investment of that magnitude on
its financial strength.

The analysis of entry barriers is confirmd by the record of entry. Over a
hundred firms have entered the computer field in minicomputers and plug com-
jiatible peripnerals in the last ren years, as well as several hundred more in
other easy entry areas such as consulting and programming services, and leasing.
In contrast, there have been no major entrants into the business systems market
in the same time period. Of the top eight companies in 1973, only Control Data
and Digital Equipment were not already in the computer market with vacuum
tube computers before 1960, and neither company has a large share of its busi-
ness in the business systems area. There has oeen a small amoant of entry into
the business systems market in recent years through established companies ex-
panding their product line. For example, Xerox Data Systems has moved from
a scientific orientation to mixed scientific and business, and Digital Equipment
has expanded from minicomputers into business systems, but movements of this
sort account for a very tiny fraction of the business systems market.

Business conduct is the focal point of most antitrust investigations and trials.
However, from an economic point of view, structure is the more important
variable. From analysis of an industry's structure and economic reasoning, one
can predict an industry's conduct. Only if an examination of actual conduct
reveals substantially different patterns than those predicted do we need to
be concerned with conduct as a policy varialtle. If conduct in the industry follows
the expected patterns given the structure, then it is reasonable to expect conduct
to change in predictable M-ays with changes in the industry's structure. If conduct
changes are imposed without changing the structure that led to the previous
conduct patterns, the firms in the industry will have an incentive to violate the
spirit if not the letter of the conduct decree. Consequently, effective conduct
changes without structure changes require continuous court or administrative
agency supervision, with the attendant possibilities for abuse, mistakes, and the
stifling of creative effort. While structural change appears at first sight to be a
more drastic interference with the free market than conduct restrictions, in
reality structural change causes less interference because after one change the
firms are allowed to operate without restrictions.

For purposes of this summary, it is suflScient to state that conduct in the
computer industry is in conformity with what would be expected given the
industry's structure. Documentation of this fact requires extensive analysis of
price and product actions in the industry and is given elsewhere."

Economic performance consists of allocative efficiency, income distribution
effects, and progressiveness or technical progress. One measure of allocative
inefficiency is the deviation of a firm's long run average profit rate from the
cost of capital in the economy. Table 3 shows IBM's profits for the years 19.58-
1971. The 17.6% return of after tax profit on stockholder equity is slightly higher
than the 16.4% average return in industries with high concentration and very
high barriers to entry found in the Mann study." This profit rate combined with

" W. David Gnrrlner, "Curtain Act at RCA." Datamation XVIII (March. 1972), p. 34-41.
" For an analysis of conclnet in thfi computer industry, see Brociv, "The United States
Computer Industry 1956-1973," pp. 150-363.
^ Mann, op. cit.



5690

IBM's sales level can be shoAvii to cause a loss of approximately $100 million per
year to the economy through allocative inefficiency.^^ This is not money received
by IBM but pure loss to everyone in the economy because of lower efficiency in
the allocation of resources than would exist in a competitive economy. In addi-
tion to the pure loss of $100 million, there is a transfer of over $1 billion per
year from IBM's customers to IBM before tax profits, compared with the prices
that would exist for IBM to earn an average return on stockholder equity.
Forty-eight i>er cent of this transfer goes to the government as corporate profits
tax while the remainder is an addition to the wealth of IBM stockholders. Income
distribution questions cannot be answered without reference to personal value
judgments, but it is unlikely that the transfer of income from computer users
to IBM could be considered good performance on the income distribution scale.

TABLE 3.— IBM PROFITS 1958-71
[Dollar amounts in millions]





Stock-


After


Profit






Stock-


After


Profit






holder


tax


rate






holder


tax


rate




Year


Equity


profit


(Percent)


Sales


Year


Equity


profit


(Percent)


Sales


1958


$720


$126


17.5


$1,172


1965


2,578


476


18.5


3,573


1959


844


141


17.3


1,310


1965


3, 323


526


15.8


4,248


1960_-


973


168


17.3


1,436


1967 -


3, 832


651


17.0


5,343


1961


1,185


207


17.5


1.694


1968


4, 569


871


19.1


6,888


1962


1,381


241


17.5


1,925


1969


5.277


934


17.7


7,167


1963...


1,592


290


18.2


2.060


1970


5.947


1,018


17.1


7,504


1964


2,254


431


19.1


3,239


1971


6, 642


1,078


16.2


8,274



Note: Average profit rate, 17.6.
Source: IBM annual reports.



In oi'der to consider improving the industry's performance in allocative effi-
ciency and income distribution through structural reorganization, it is necessary
to be sure that IBM's high profit rate is a result of barriers to entry and not a
result of IBM's superior efficiency. The question cannot be answered definitively
because of the difficulties of exactly measuring computer performance discus.sed
above. However, three pieces of evidence suggest that the high profits are due to
entry barriers and not efficiency. First, in areas where entry barriers are low,
IBM has been forced to cut its prices to compete, particularly with plug com-
patible peripherals. IBM has also taken action to tie its peripheral products to
its CPU's and protect them from competition, suggesting that at lea.«t in periph-
erals high profit margins were not a result of superior efficiency. Second, the
estimated economies of scale are not great enough to account for the difference
between normal manufacturing profits and IBM's profits. And finally, a regres-
sion equation relating computer' price and performance (as calculated by a
weighted average of internal characteristics) shows that IBM computers have
a statistically signicant higher price for equivalent performance than non-IBM
computers." The statistical result is not definitive in itself because of the varia-
tions in software and service provided by the different manufacturers, but it
helps to confirm the hypothesis that IBM's profits are not solely a result of
efficiency.

Because of the many opportunities for technological progress in the computer
industry, progressivenes.s is a most important criterion of economic nerformance.
If the rapid rate of technical progress that has existed in the industry could be
shown to be a result of its concentrated structure, then tlie relatively ooor per-
formance that has existed on the allocative efficiPiny and income distribution
criteria could be accepted as a necessary price for continuing progress. If. on
the other hand, progressivenes.s has been retarded by the industry's structure,
then changes need to be made in the structure to improve the performanr^e. An
evaluation of progressiveness in the computer industry cannot be made by simply
observing that technical progress has been more rapid than in most industries ;
it is necessary to know how technical progress would have differed with a
differing structure. In the absence of an opnortunity to directlv observe pro-
gressiveness in an industry with the same technical opportunities as the computer
industry but a different sfnirtnrp. we can sul.stitute observations on the propor-
tions of innovations introduced by firms of var.ving sizes in the industry.

18 Brock. "The Unitpcl States Computer Indiistrv 1956-1973," pp. 426-467
^'^ Ihid. pp. 164-169.



5691

An analysis of many innovations in tlie industry indicates that nearly all com-
panies have made substantial contributions to technical progress."* xUthough IBM
has contributed the most iunovatiims of any single company, its proportion of
contributions has been far below its market share. Actual introduction of an
innovation requires both the technical ability to produce the product and the
market situation that makes it profitable. IBM generally has an advantage in
technical ability because of its extensive research labs, but the smaller lirms have
more incentive to rush an innovation on to the market. A major innovation is one
method of overcoming IBM's product reputation advantage because it allows the
innovator to offer either a much better product or a much lower price than the
proudcts without the innovation. Probably the njost important single innovation in
the industry was the switch from vacuum tubes to transisvtors as the basic circuit
component. Recognizing a major opportunity, four new companies (I'hilco,
Autonetics. G.E., and RCA) entered the computer industry with transistor com-
puters to market against IBIM's vacuum tube models before IBM delivered the
transistor 7090 in late 1959. All innovations imply risk taking. It is generally to
IBM's advantage to avoid taking the risk of an untried concept. A new company
or one trying to expand its market share must take the risk in order to prove the
superiority of its products. If the innovation is successful or appears to be a
threat to IBM's market position, IBM can usually deliver a competitive product
before incurring a substantial loss.

Besides risk, it is in IBM's advantage to avoid taking the lead because of
technical standards. For example, even though the IBM 7090 was delivered after
the Philco 2000 system for large scale users, the 7090 soou became the dominant
machine in that segment of the market throvigh IBM vacuum tube computer
upgrades. Consequently, Philco found itself at a disadvantage in competing for
ui)grades because the Philco 2000 was incompatible with the IBM 7090. IBM was
more protected from Philco competition by the incompatibility of the 7090 and
2000 than if IBM had delivered the 7090 earlier and allowed Philco to copy some
of its specilicatious. The same problem exists with the plug compatible peripheral
nianufacturers. Because they are dependent upon compatibility with IBM, they
can make only minor improvements over the announced IBM specifications. Even
if the technology is available to them, they cannot make major advances such as
the change from 2314 type disc drives to 3330 type drives without waiting for IBM
to define the specifications which must be followed.

From the record of innovations in the computer industry, it appears that
maximum technical progress occurs with a wide variety of firm sizes. A perfectly
competitive industry of very small firms would lack the financial resources for
very expensive development work such as complex operating systems. However,
domination by one firm also slows progress because of the lack of incentives for
that firm to lead in putting innovations into marketable products. An industry
with some very small firms to introduce high risk products or innovations passed
over by the larger ones, some medium scale firms of the size of IBM's current
major competitors to provide sustained technical progress, and possibly some
larger firms could be expected to encourage technical progress somewhat more
efficiently than the current structure.

///. Proposed Reorganization of the Computer rndii-strii

In considering methods of improving the performance of the computer indus-
try, it is necessary to evaluate the problems leading to poor performance. Taa'c
situations can be distinguished. The first is poor economic performance based on
the criteria outlined in the previous section. Poor performance in allocative effi-
ciency and income distribution arises from barriers to entry in computers. If there
were no barriers to entry, excessive profits Avould he competed away by new
firms or expanding old ones and thus could not remain for long. The estimated
poorer than necessary performance on the progressiveness criteria is also due to
barriers to entry. Without barriers to entry, new innovations from any source
could be expected to rench the market more rapidly than at present. Concentration
per se is not a problem of economic performance except insofar as it contributes to
barriers to entry. It is indirectly the cause of at least part of the poor economic
performance, because it is IBM's dominance of the industry which allows the
creation of barriers to entry.

The second situation is poor performance on criteria that are not strictly
economic. Jlore dispute is likely among well intentioned observers in evaluating
performance on these criteria because they relate to one's own value system and
political beliefs. Chief among the non-economic criteria would be the distribu-

" Ibid. pp. 364-~425 for a detailed discussion of innovations in tlie computer industry.



5692

tion of power. Concentration is tlie liey problem here and not barriers to entry.
Even with relatively free entry, IBM would have substantial control over con-
ditions in the computer industry with present levels of concentration. Because
of the dependence of business and government operations upon computers, and
the close ties of computer users to manufacturers, IBM maintains tremendous
power over the functioning of modern economy.

These cases must be separated because emphasis on one or the other leads to
different remedy proceedings. Because the author of this paper is an economist,
the assumption taken here is that optimal economic performance in the industry
is the primary goal of the proposed reorganization. This is not meant to imply
that other goals are unimportant, but simply that other goals are less well de-
liued and less generally accepted than the goal of good economic performance.
The economic approach allows one to be objective about the costs in terms of
economic performance of reaching other goals as well. To illustrate, suppose it
could be shown that economies of scale existed in a given industry up to a scale
of operations equal to 50% of the entire market.

On purely economic grounds, a proposal for reorganization of that industry
could not have more than two firms, because otherwise inefficiency in production
would result. However, if one believed that power in the economy must be dis-
sipated, lie would propose more than two firms in this hypothetical industry, ex-
plicitly balancing the loss to the economy from less eflicient production against
the gain from reduction in the concentration of economic power. Alternatively,
if one believed that power must be controlled, he could propose a government
regulatory agency to oversee the industry's operations, a path that has been fol-
lowed with many industries which have too high economies of scale for effective
competition. Because regulatory agencies incur some economic costs Cat the
very least, the salaries of the regulators and their counterparts in the industry
who fill out forms, but often much more substantial losses because of disin-
centives to efficiency), the regulation approach is also a trade off between eco-
nomic and non-economic goals.

Both brand loyalty and capital costs are increased through the vertical in-
tegration of the major manufacturers. The large capital cost comes from the
necessity of setting up a vertically integrated organization to compete with the
existing one. A potential entrant into the business systems market cannot manu-
facture just CPlT's ; he must introduce a complete line of CPU's and peripheral
equipment, as well a.'* software, maintenance, and other support services in order
to compete for customers mth the established companies. Consequently, capital
costs are much higher than they would be if entry was possible into only one
segment. Vertical integration increases brand loyalty because it contributes to
the difficulty of evaluating equipment or switching among manufacturers. An
integrated organization has no incentive to make the component parts of its
computer compatible with other parts produced b.v different manufacturers.
Consequently, a user must compare entire systems rather than specific devices,
causing a decrease in the accuracy of the comparison. If the user decides that a
competitive system is superior to his present one, he must replace the entire sys-
tem rather than only the superior devices, causing increased costs.

The above analysis leads to the conclusion that the most improvement in eco-
nomic jierformance would come about through a division of IBM into several
companies by function, peripherals, CPU's, maintenance, etc. Although each com-
pany would dominate its market, barriers to entry would be low. Many variants
of a functional reorganization plan are possible. The final plan chosen should be
the one with the least disruption to customers or to IBM operations. If two
operations are ordinarily handled together and there is no strong economic
reason for separating them, they should be left together to minimize disruption.
Much of the information on IBM's current internal operations that would be
necessary to formulate an exact efficient plan is not availal)le to the author.
The plan presented below should be treated as an example of the type of re-
organization that would improve performance with minimal disruption. The prin-
cipal of functional reorganization can be firmly established with current informa-
tion from economic analysis, but the final details of which division should go to
which new company cannot be determined without more complete information and
analysis of the problem by persons in the relevant disciplines other than econ-
omies. The reorganization proposal has four objectives: (1) eliminate or dras-
tically reduce barriers to entry, (2) minimize disruption or confusion on the
part of computer users, (3) preserve the current operating efficiency of the IBM
organization, and (4) avoid conduct restrictions that will require continuous



5693

court supervision or put unreasonable restraints on the freedom of the new
companies. No proposal can fully meet all four objectives because of the basic
technology of the industry. For example, the existence of economies of scale
in the software segment indicates that software ought to be left as a single com-
pany to meet objective 3, but leaving software together means that substantial
barriers to entry remain at least in that segment. The author has used the



Online LibraryUnited States. Congress. Senate. Committee on theThe Industrial reorganization act. Hearings, Ninety-third Congress, first session [-Ninety-fourth Congress, first session], on S. 1167 (Volume pt. 7) → online text (page 100 of 140)