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 94 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 94 of 140)
Font size
QR-code for this ebook


Consequently, eliminating integrated systems buying would make
it much easier to measure exactly what you are getting. The problem
of transfer costs would be eased by the standards that would come
about naturally, I believe in the normal evolution of the marketplace.
The standards necessary for using equipment of different manufactur-
ers together would also improve the compatibility of programs and
data with different systems.

So that is one barrier.

Now look at the second barrier, that of economies of scale. There
would be relatively little change in that.

The third barrier is the capital cost. There would be a substantial
improvement because it requires much less capital to enter any one
segment of the market than all segments together. So long as there
are major companies existing independently in each one of the seg-
ments, anyone who wants to enter any one of the segments can enter
only that one segment and pay the capital cost of entering that seg-
ment, and not have to build an integrated systems business.

Right now. if you want to make a special kind of peripheral unit
you must find someone else to supply the CPU to go with it because
there is nobody making just CPU's So that is what I meant by barriers
to entry that result from integrated systems selling.

Mr. Granfield. I am sure we could go on forever with the barriers
to entry, but let me finally conclude with the statement that I was
delighted to see your concluding statement where you say the lack
of competitiveness is not due to concentration in itself. I think that is
a very fine conclusion. Thank you, Mr. Chairman.

Senator Hart. Professor, thank you very much.

Dr. Brock. Thank you.

Senator Hart. I want to thank all who participated in this Senate
hearing on computers. We will I'ecess, to resume Tuesdaj'' at 10 o'clock
in this room.

Mr. CiiuMBRTs. May I just ask one question before we leave. I know
we will be getting a lot of questions on when we may be resuming
on IBM. Is there any indication from staff; 2 months. 3 months,
1 montli?

Mr. O'Leary. We just don't know at this time.

Mr. Chumbris. Thank you. That is a good answer, and what I
would like for the record so they won't bother us on the telephone
from all over the country. Thank you very much.

Senator Hart. Thank you.

fWliei'eupon. at ?> :?>5 )).m. the Subcommittee adion.rned. to rocmivone
Tuesday, July 30, 1974, at 10 a.m., in room 222S, Dirksen Senate Office
Building, on the subject of the communications industry.]

[The following material was received for the record:]



5657

MATERIAL RELATING TO THE TESTIMONY OF GERALD BROCK

Exhibit 1. — Prepared Statement of Mr. Brock

Statement of Gerald Brock, Assistant Professor of Economics, University

OF Arizona

The computer industry occupies a uni(iue position in the American economy
because of its size and importance. In the hist twenty years, computers have
chnnged from a curiosity of use only in specialized scientific problems to an
indispensible tool without which no large corporation, university, or govern-
ment agency can survive. The computer industry is of special interest to academic
economists such as myself because of its unusual market structure and rapid
technological progress. I have been studying the industry for the past three
years in an attempt to bring economic analysis into the debate surrounding
potential government action toward the computer industry. My statement will
concentrate on how the industry's price and product policies are related to the
industry structure.

TABLE l.-IBM MARKET SHARES

Census Honeywell IBM interna

Year:

1955.. 56.1

1956 73.1 75.3 .........

1957 .._ 78.5

1958 71.2 77.4 ..'..'.'.'..'.

1959 74.5

I960.. 70.7 71.6 '

1961.. _ 69.3

1962 _.__ _ 70.4 70.0

1963.. 74.5 69.8

1964 _... 72.5 68.3 75.9

1965 66.7 65.3 74.8

1966 69.7 66.2 74.2

1967 74.3 68.1 74.3

1968 74.6 73.8

1969.. .._. 71.0

1970 70.6 70.0

1971 67.4

1972

1973 __ _ ■"..^^[Ilim "



Source: Census figures from "Computers and Automation" and "Diebold Automatic Data Processing Newsletter".
Honeywell figures from "Findings of Fact, Conclusions of Law and Order for Judgment", Honeywell vs. Sperry-Rand,
4-67 Civ. 138, district of Minnesota, p. 157. IBM Internal figjrasfrom "IBM Quarterly Product Line Assessment", Novem-
h^x, 1968, p. 13 and March, 1971, p. 19, listed as planitiff's exhibits 123 and 127 in Telex vs. IBM, 72-0-18 and 72-C-89.
northern district of Oklahoma.

By industry structure, I mean primarily the concentration of firms and barriers
to new firms entering the market. IBM's market share, as computed by three dif-
ferent sources, is shown in Table 1. The first column, marked "Census', was com-
puted from information on numbers of machines and average monthly rental given
in Computers and Automation and Diebold Automatic Data Processing Newsletter.
It is probably the least reliable of the three because of the secrecy attached to
the figures by the manufacturers involved, and the resultant estimations used
by (he editors of the respective periodicals. The second column, marked "Honey-
well." is taken from the Honeywell vs. Sperry Rand patent infringement case.
Tiie thii-d column, marked "IBM internal," is taken from IBM's Quarterly Prod-
uct Line As'^essment and is the market share listed for "Systems and Periph-
erals." The three computations each have somewhat different l)ases and the
figures do not exactly agree. However, they are close enough that we may con-
clude with confidence" that IBM's market share has been in the 65-75% range.
With less confidence we can conclude that IBM's share has been dropping slowly.
The IBM internal figures show a steady drop from 1964 to 1970. the Honeywell
deci.sion figures show a cyclical movement around product generations with a
sliirht downward trend, and the census figures show a cyclical movement around
product generations with no downward trend.



5658

Table 2 shows the market shares of the top seven non-IBM companies. The
figures for 1955-1967 are taken from the Honeywell decision, while the figures
for 1968-1971 are computed from the Computers and Automation censuses..
Roughly speaking, the major competition to IBM has been the seven companies^
listed with two to ten per cent of the market apiece. The seven were reduced
to five with the Honeyr^'ell-Geueral Electric merger in 1970 and the exit of RCA
in 1971. In recent years, Digital Equipment has expanded its offerings to the
point where it should be considered a major competitor. Besides the companies-
listed, there are a large number of companies with under 1% of the market apiece,
largely competing in specialty areas such as minicomputers rather than the main^
general purpose computer market.

TABLE 2.— MARKET SHARES TOP 7 NON-IBM COMPANIES











General








Sperry-Rand


Honeywell


Control data


Elect! ic


Buiroughs


NCR


Year:














1955


38.5 -

18.6 ..

16.3










0.3


1956








4.4
3.9


.1


1957 .-.


.3 .






.06


1958


16.3


1.0 .




.2


3.3


.04


1959


17.8


1.2 .




.9


4.2


.12


1960


16.2


.9


1.0


2.8


3.4


.4


1961


15.5


2.0


2.2


3.4


2.6


.7


1962


12.4


2.3


3.1


3.7


2.2


1.9


1963


11.2


1.8


4.0


3.5


2.6


2.7


1964 .


11.8


2.5


4.4


3.3


3.1


2.8


1965


12.1


3.8


5.4


3.3


3.6


2.9


1956


11.3


5.2


5.3


3.5


3.0


2.4


1967


10.6


4.7


4.7


3.0


2.9


2.5


1958


5.6


4.1


3.9


3.2


2.1


2.2


1969


3.2


4.8


7.3


3.1


3.4


2.3


1970


4.4


7.6


7.7 ...




4.1


2.5



Source: 1955-67 figures from Honeywell v. Sperry-Rand Decision, p. 157. 19S8-71 figures from "Computers and Auto-
mation".

As part of the discovery proceedings related to the IBM-Control Data antitrust
suit, the Minnesota District Court authorized IBM to conduct a census of the
computer industry. For purposes of the census, IBM defined the industry very
broadly to include electronic switching systems, teletype machines, and applica-
tions software. (1)* Using IBM's definition of the market, American Telephone
and Telegraph, which makes no general purpose computers, became the second
largest "computer industry" participant. IBM computed its share of the market
at 35.1% for 1970.(2) Economically, there is no single correct definition of the
market, but it must include only those products which are in competition with
each other. IBM's definition is so broad as to be meaningless for analyzing
competition.

Barriers to entry are anything which makes it difficult or impossible for a
new firm to enter the industry. For the computer industry, barriers to entry
conisist primarily of raising capital, economies of scale, and brand loyalty. All
three problems are much greater in the integrated systems portion of the market
than in individual parts of the computer market such as software companies,
or companies which make only input-output equipment.

Economies of scale are of minor importance in the actual manufacturing stage,
but pose a significant barrier to new competition when considering production
of software. Producing software is like writing a book. The first copy is very
exi)ensive but subsequent copies have a low marginal cost. A very small well
managed firm would experience only slightly higher unit costs than IBM when
producing tape drives, memory units, or central processing units, but would
have an extreme cost disadvantage if it tried to compete with IBM's systems soft-
ware because almost all the cost is in producing the first copy. Although applica-
tions software has the same theoretical economics of scale as systems software,
in practice they are not so significant because modifications are often required
for each user of application software. Consequently, economies of scale have not
been a significant barrier to new competition for companies producing competi-
tive input-output equipment, minicomputers (which generally utilize relatively

* References may be found at the end of Mr. Brock's prepared statement. See p. 5664.



5659

simple systems software), and sei-vice and consulting organizations. Economies
of scale have increased the diflit-uiry of entering the integrated business systems
marliet where complex and extensive software support is required for success.

Two factors account for the very great brand loyalty found in the computer
industry. The first is the difficulty of making rational computer selections. Be-
cause the major manufacturers are integrated, and parts of one manufacturer's
system generally cannot be used with another manufacturer's system, the com-
puter user must choose between complete systems, including the central proc-
essor, various kinds of input-output equipment, systems software, and various
levels of consulting help and applications programs. It will seldom be true that
one system dominates the others on all points.

Generally the user will have to balance, say, a better operating system on one
machine against better input-output equipment on another. To compliance mat-
ters further, the user generally expects various enhancements over the life of
the machine, either in software or hardware, which are not known in detail
at the time of the selection. The user also is often dependent upon the manu-
facturer for help in defining his computing requirements and choosing the proper
equipment to meet them. As a result of all these factors, the decision of which
computer to purchase is seldom a perfectly rational one, but instead is heavily
dependent upon the user's judgment about the future actions of the various
manufacturers. Equally qualified managers faced with the same information are
likely to come to different decisions about which computer best meets their
needs (3). As in amy situation where clear choices are difiicult, the established
manufacturers with good reputations have a great advantage over new com-
panies even if the new companies have equal or sui)erior products.

The second factor accounting for brand loyalty is the lack of compatibility
among systems. In an extreme case, if all programs were written in assembly-
language and the two computers had different instruction sets, it would prob-
ably not be profitable to switch regardless of the price of the new computer be-
cause the cost of transferring all the programs would be equal to the value of
the machine. In the more normal case, with most programs in a language such as
COBOL or Fortrain, the basic programs are usable but some changes are required
either in the programs themselves or the job control language. Switching com-
puters is conceivable in this situation, but the new company must offer sub-
stantial price discounts in order to induce the customer to pay the cost of
conversion.

The compatibility and evaluation problems are largely the result of selecting
complete systems and of the incompatibility of various manufacturers' specifica-
tions. However, when both of these problems are eliminated, some residual brand
loyalty remains. Probably the most straightforward decisions regarding com-^
puter equipment are those made between IBM and competitive plug compatible
peripheral equipmeait. Only a single piece of equipment is considered at a time,
so systems trade-offs are not involved, and the competitors adopt IBM specifica-
tions so that compatibility is not an issue. The only uncertain issue is the reliabil-
ity of competitive claims for their product versus those of IBM. Even in this
simplified situation, an IBM study showed considerable brand loyalty. IBM asked
a sample of its disk customers what the maximum discount was that a competitor
could offer for a replacement product aiud the customer still remain with IBM.
The results are shown in Table 3.





TABLE 3.-


-IBM


DISK CUSTOMER LOYALTY
[In percent]












IBM customer remaining with IBIVI




Competitive discount




Overall 2319B users 2319A users


3330 users


Over 20..

16 to 20






-


31 24 23
46 42 37
70 60 58
92 86 95
97 92 99


22
36


11 to 15






64


6 to 10..






88


lto5






94











Source: IBM, "DASD Survey Summary," Mar. 31, 1971, Telex v. IBM, plantiff's exhibit 81.

As can be seen from the table, only a small proportion of IBM customers were
potential targets at discounts of less than 10% and 31% would remain with IBM
even with discounts over 20%. If the table is accurate, bramd loyalty is a sub-,
stantial barrier to entry even in the plug compatible peripheral market.



5660

The qnautity of capital required for entry iuto tlie computer industry depends
heavily upon the segment entered. Service bureaus, consulting groups, and soft-
ware liouses can have very low capital requirements of a few tliousand dollars
for initial salaries and office rental before payments begin. Of course, capital
requirements may be far larger if rapid expansion is planned or large complex
projects are contemplated. Plug compatible peripherals and minicomputers have
capital requirements beyond the ordinary range of private financing, but still
moderate in comparison with many industries. In minicomputers, the most suc-
cessful new entrant in recent years. Data General, began with $50,000 in 1968,
but raised .$20 million more on the stock market between 1969 and 1971 to become
an established company. In analyzing potential entry into the memory business,
IBM estimated that a new company could begin with $75,000 but would need
to invest $15-20 million over a four to six year period before reaching the break-
even point (4). The capital re(iuirements for entry imto the integrated systems
business are huge. When RCA left the computer market in September, 1971, it
took a $490 million write-olf. In addition, the company estimated that it would
have needed a $500-700 million new investment over the next five years in order
to attain profitability (5), suggesting total capital requirements of over one
liillion dollars for a competitive systems company. The amount of capital required
is so large that it is unlikely that any new company could raise it, amd it is beyond
the financing capabilities of all the largest corporations.

Taken together, the capital costs, brand loyalty, and economies of scale form
•an almost insurmountable barrier to new entry in the integrated systems busi-
ness. In spite of tremendous market growth and extensive technological changes,
no new companies have successfully entered the integrated systems business since
1960. In contrast, there has been continuous entry in peripherals, software, and
minicomputer companies, as would be expected from their very low barriers to
entry.

Tlie extensive brand loyalty in the computer industry means that all nrnjor
manufacturers, not only 1I3M, have market power. Market power is defined as the
ability to raise prices above the cost of production, including a normal return to
capital, without being driven out of the market. If no manufacturer makes a
specific attempt to attack the customer base of another manufacturer through
compaibility and/or conversion aids, then all can enjoy high profits and stable
market shares. This is equivalent to the practice of price leadership in other in-
dustries which allows firms to coordinate their practice and prices without explicit
agreements. It has not been the pattern observed so far in the computer' industry
but appears to be becoming more important. So long as growth in the industry
was extremely rapid and the non-IBM manufacturers had a relatively small cus-
tomer rental base, it was in their advantage to aggres.sively attempt to expand
their market share. Hovv'ever, xirice competition is very expensive to companies
with large rental bases, because a price cut means not only lower revenues on the
new customers attracted but also lower revenues from machines placed earlier.
Consequently, there is a tendency to concentrate more on upgrading one's own
customers than on taking customers away from competitors as the rental base
grows larger. IBM's evaluation of competitors for the 370/135 concluded :

"It appeared that many competitors were focusing on protecting and growing
their own inventory bases and were not prepared in the near term to get around
the M135 in pursuit of IBM's lease base." (6) IBM has been the best example of
this "grow your customers" approach to computer marketing, but not the only
one. IBM has made a strong effort to develop close relationships with its customers
through emphasizing rental rather than sales, and before 1970 by emphasizing a
full range of service^ in addition to the hardware for the basic price. IBM does
not provide conversion aids from competitive machines and makes no attempt to
directly undercut any given systems manufacturer.

Because of the high cost of switching between incompatible machines, a firm
that wants to expand its share of the market cannot simply cut the price of its
machines. It must either provide a lower price together with compatibility or
provide a technological advance wliich increases the capability of the machine.
Both strategies have been used regularly in the industry. If a sxilistantial increase
in the cfipability of a computer can be made through technical innovation, the
innovating company can attract new customers who either are not using com-
puters or who need the new capability enough to pay the necessary conversion
costs. Two examples are the Control Data 6600 and the Burroughs Master Control
Prosrram. When the Control Data 6600 was delivered in 1064. its capacity wns far
beyond anything else on the market. Consequently, for very large scale computing
needs, such as atomic energy research, the 6600 could easily induce users to



5661

switch from previous computers even witliout compatibility. Burrouglis' Master
Control Program for the 6-5000 system and related AOSP system for the military
oriented D-S25 system put Burroughs well ahead of other companies in develop-
ing effective multiprocessing capability.

Multiprocessing is the ability to tie several central computers together under
a single oiierating system. Its primary advantage is reliability; if one of the
processors should fail, the computer system can continue operating with reduced
capability rather than being incapacitated while repairs are being made. One
application of the D-825 limited the system to four hours of downtime over a ten
year period, an impossible requirement without multiprocessing. (7) Burroughs'
early development of multiprocessing and virtual memory capability allowed the
company to attract customers needing those capabilities in spite of compatibility
problems.

The innovation approach to increasing market share is only successful if the
innovation is accomplished successfully without too great cost to the company
and if it cannot be easily duplicated by competitors. Transistor computers were
probably the greatest single advance in the industry so far. Philco was the first
company to introduce a large scale transistor computer, but was followed so
closely by the IBM 7090 that Philco gained little advantage from its innovation.
Time-sharing has been a very important development since the mid-1960's. How-
ever, General Electric made early large investments in time sharing technology
but failed to capitalize on its innovations because the costs and technical diflS-
culties were greater than expected.

The third possible strategy is to market directly against a particular computer
manufacturer through copying design specifications in order to achieve a high
degree of compatibility. This is the most direct method of expanding market share
and also the most dangerous for the firm, because it is equivalent to starting a
price war. Once compatibility is achieved, customers can more easily leave the
aggressive company as well as come to it. The price cutting company must be
concerned both with competitive response and the timing of his attack in the
product cycle. If the competitive product is introduced late in the target firm's
product cvcle. it must also be competitive with the target firm's next generation
or the rental life will be too short for profitability. The target firm may choose
to cut prices as a result of the competition, but it must be concerned for the effect
on its ovn\ rental base. If only a small proportion of the rental machines are likely
to move to the competition, then the firm would be hurting itself with an across
the board price cut to restrain competition.

Two of the best examples of the compatibility strategy are the Honeywell H-200
and the RCA Spectra 70 series. During the early 1960's, the IBM 1401 was the
most popular computer on the market. Honeywell announced the H-200 in
December. 19G3, for first delivery in July, 1064, together with a program called
the Liberator which would convert IBM 1400 series programs into H-200 pro-
grams. The H-200 was a great improvement over the 1400 series for comparable
prices, and strong enough to remain competitive with IBM's replacement for the
1401, the 860/30. The success of the H-200 was increased by incompatibility
liet^y'een the 1401 and the 360/30, making it easier for a customer to convert to
Honeywell than to upgrade to IB]M. The PI-200 success l)rought Ploneywell into
an earlv position of stability and profitability in the computer industry.

A less successful example of the compatibility strategy was the RCA Spectra 70
series. Soon after the IBM System/360 was announced in 1964, RCA announced
the Spectra 70, a series of four computers each designed to he compatible with,
but to outperform its 360 counterpart. The instructions, formats, and character
codes were identical to those on the 360. One estimate placed the Spectra prices
at 40% under the corresponding 360 prices, but accurate comparison is difficult
because of the importance of software and service. (8) In spite of compatibility,
the Spectra series did not sell well against the 360. RCA was trying to sell against
a largely undefined, rapidly changing target as various enhancements were added
to the hardware and software of the System/360, making close comparison difficult
for customers to make. In addition. RCA could not offer the same kind of clear
price advantage that Honeywell did against the 1401 because RCA was bringing
out its machines at approximately the same time as IBM, while Honeyw:4l had
the advantage of several years additional technical development. RCA's market
share remained at the 2..5-3.r>% level of its pre-Spectra days. To make matters



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 94 of 140)