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 98 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 98 of 140)
Font size
QR-code for this ebook


further discussion and consideration of various alternatives, the "MRC concluded
that any rental reductions at levels permissible to good business judgment sh.ould
be excluded on the basis of their minimal contribution to strategic objectives
when balanced against their undesirable revenue effect." The final decision
was to raise the CPU prices by no more than 8%. cut the purchase price of the
155-165 memory, and make no change to the rental price. (46) No combination
of actions could be found that would stop the competition in memory and still
be justified by IBM as anything other than an attempt to stop the competition in
memory.

TRAf's plan for the 370 Included core memories for the 155 and 165. semi-
conductor m(^mories using phase 2i or bii*olar technology for the 135 and 145.
and semiconductor memories using FET (field efi'ect transistor) tecbno'ogy for
the 125. The FET technology was cheaper than the bipolar but was not expected
to be availal)le initil 1975 in quantity. With improvements in memory tech-
nology by competitors, IBil accelerated its FET iirogram to the point where
FET memory could be delivered in 1973. (47) The 155 and 165 Greybooks
I extensive analysis of the machine, market, and strategy prepared for each
major product) prepared at the end of 1970 (after announcement but before
delivery of the machines) showed that both machines would he upgraded with
semiconductor memory and virtual memory ("relocate") capability during their
lifetimes. The memory enhancements were scheduled for delivery in Janunry.
19T4, \\ith ivlocate to follow in June, 1974. The initial assumption was that
relocate woidd be fitted retroactively to previously delivered machines. The 165
Greybook stated : "The relocate and midtiprocessing features were assumed to
be available to all customers equally. Thus, customers who purchased earlv
systems requiring extensive engineering changes to accomplish this feature
attachment will not pay a premium price." (48).

With the rapidly growing threat to memory during 1971, IBM abandoned the
idea of ui-grading the 155 and 165 and revived tlie strategv of introducing the
upgraded 155 and 165 as new machines with higher CPU prices and lower
memory prices. By delaying the introduction of virtual memory, IB:M cou'd
combine the virtual storage and semiconductor memory announcements. It could
then justify the higher CPU price through increased performance and the lower
nioi'iorv price through the reduced manufacturing costs of semiconductor
memory.

Besides price, the FET memory had many features which increased tlie proli-
lems of a potential replacement company, A late 1971 IBM engineering study of
the potential methods a PCM company m.ight use to attach memory to the IBM
CPU concluded that "the ability for competition to add to our minimum FET
memory configurations is a very large and maybe even an impossible task." (49)
The primary difficulty compared with core additions was that core memory was
in a separate box while the FET memory was inside the CPU sharing power
with other units. The CPU was so designed that if the competitor tried to
attach a separate box of memory, it would need to make changes in IBM's CPU
40-927—75 54



5678

logic. If the competitor tried to replace IBM's memory cards directly, be faced
jnajor problems because of differences between IBM specifications and standard
industry practice, and secret interface specifications and cable connections
which would not be published or sold. The study concluded : "In short, we may
be overreacting in our assumptions to reduce the $12K slope." (50) (to reduce
the FET memory price below the $12,000 per month per megabyte on the 155
and 165 core memory ) . In order to avoid losing money on a needless price cut,
IBM set up the Kenyon House Task Force to evaluate the conclusions of the
engineering study and reconsider the need for a price cut. The task force con-
cluded that the technological difliculties of adding to FET memories could be
overcome and the only solution to competition was to lower price enough to
prevent entry.

As a result of the studies, IBM announced tlie new versions of the 155 and
1G5 as the 158 and 168 on August 2, 1972, as part of the SMASH announcement.
The basic CPU price was raised 36% on the 165 (from $35,640 per month on the
165 to $48,600 per month on the 168) and 54% on the 155 (from $19,980 per
month on the 155 to $30,700 per month on the 158). The memory price was cut
57% from $12,000 per month per megabyte to $5200 per month per megabyte.
In addition, a minimum of one half megabyte on the 155 and one megabyte on
the 165 was tied to the basic CPU ; IB:M would not sell or rent the CPU alone.
The drastic increases in CPU price and decreases in memory price meant that
the total price of the new machines was higher in small configurations and lower
in large configurations than the original machines. The semiconductor memory
was not made available to the original 155 and 165, prohibiting customers from
buying the cheaper CPU and putting on the cheaper memory as well. The relo-
cate feature was made available, but at a high cost, $200,000 for the 155 and
$400,000 for the 165. (51)

There is some dispute as to whether or not IBM calculated the $5200 price
for memory as the price that would drive all competitors out of business. IB^I
analyst B. M. Hochfeld testified at the IBM-Telex trial (after leaving IBM)
that he viewed the $5200 price as low enough to prevent competition from
entering the 158-168 memory market. (52) However, IBM introduced charts
showing that the Kenyon House task force had concluded that companies could
be viable selling FET memory at $3000 per megabyte per month, enough to
allow a 42% discount over IBM's price. (53) Many industry observers assumed
that the prices were designed to eliminate competition. R. A. McLaughlin,
associate editor of Datamation wrote soon after the announcement : "This kind
of pricing can't do very miich for IBM revenues, but it may drive independent
memory suppliers * * * from several weeks before the aimounceinent, because the
substance of the announcement was revealed early through a Telex court attempt
to prevent it from being made. From July 6, 1972 to August 3, 1972 (the SMASH
announcement was made August 2), ISIemorex stock dropped 31% while Telex
stock dropped 28%. Memory maker Advanced Memory Systems dropped 25%
in the same period. Because of the early retirement of the 155 and 165, the move
was also a Mow to leasing companies who purchased the machines. Computer
lessors DPF dropped 22% and CIC dropped 11% over the .Inly 6-August 3 period.

The SMASH announcement was IBM's last significant move against the PCM
companies to date. The new products and prices announced August 2 combined
with previous actions effectively stopped plug compatible competition on the 370
machines. Some efforts continued but not enough to be a threat to IBM's revenue.
The focus of the PCM companies returned to the 360, where significant enhance-
ments were made by providing more and faster memory and other peripherals
than allowed by IBM. as well as cutting below IBM's prices. Enough 360's re-
main installed at present to provide a large market, but it is a deadend busi-
ness unless some way is found to add on to the new machines. The 360's with
PCM core and iierijvherals ai-e enough more powerful than the original 360's to
replace some 370 sales but the limited number of machines available has made
this a rc'titively unimportant problem for IBM.

Tlie 1970-1972 war between IBM and the PCM companies has been examined
in detail because it is an unusually clear example of a pui-e profit maximizer
taking advantage of the stmetural characteristics of the industry. If only the
record of price and product actions were available, as is generally the case in
industry studies, we could only say that IBM's actions were consistent witli the
profit maximization hypothesis given the structure of the industry. However,
because of the documents revealed in the Telex-IBM antitrust case, we can see
the information available to the decision makers at the time and the way they



5679

-analyzed it. This allows the unequivocal conclusion that IBM was practicing a
very spohisticated form of profit maximization which included explicit evaluation
-of the effects of its actions on current companies and potential entrants.

The basic structural characteristics that are important for imderstanding
IBM's actions are that there are high barriers to entry in the entire systems
market, but low barriers to entry in the peripherals market. The barriers that do
exist in the peripherals market are primarily related to brand loyalty to IBM
(including a fear of a mixed supplier installation) and the difficulty of produc-
ing exactly compatible products which will match the IB.M interface specifica-
tions. The latter problem causes a time lag between the time of IBM introduc-
tion of a product and the time it can be copied. This time lag reduces the rental
life of the competitive product and gives IBM some time after a new product in-
troduction without competition. Additional barriers to entry in the peripheral
market include the need for capital to finance a rental business and some econ-
omies of scale which give IBM a manufacturing advantage. It is important to
note that the economies of scale are small enough that the PCM companies could
make a satisfactory profit while undercutting IBil's existing price before IBM
took explicit action against them.

In any situation where a company possesses substantial market power in one
product and less market power in a complementary product, the best policy for
profit maximization is to tie the products together, refusing to sell the one with-
out the other. If the products are totally tied together, the company can choose
the prices for each combination to optimize profit. However, tying products to-
gether is illegal under the antitrust laws if it expands market power. A less
effective alternative is to increase the price of the product with barriers to entry
and price the product without barriers to entry at a competitive price. This
reduces the possibilities for pric^discrimination among various classes of users.
For example, if the CPU is priced high enough to gain as much revenue from
the intensive user with low peripherals prices as would be gained with a balance
■ of peripherals and CPU prices, then it is likely to be high enough to drive the less
intensive user out of the market because he is not compensated by the lower
peripherals prices.

IBM pursued a combination of both strategies, tying products together where
it could be technically justified, and raising CPU prices while reducing peripherals
prices where tying could not be defended legally. The integration of controllers
(IFA for the 2319 and ISC for the 3330), and the bundling of minimum memory
with the CPU were attempts to tie the market for peripherals and CPU's to-
geth<^r. The FTP on peripherals with its substantial price cut followed by in-
creases in CPU prices, and the 158-168 with substantially increased CPU prices
and reduced memory prices were reallocations of the revenue between CPU's and
peripherals to take account of the fact that IBM no longer had substantial market
power in peripherals.

A second aspect of market structure is the disruption caused by equipment in-
stallation and removal, even when compatibility is not a problem. This means
that extra revenue can be obtained when a price cut is necessary by requiring
an equipment exchange to take advantage of the price cut. This tactic was used
•effectively with both the 2314 disk drive and the 2420-7 tape drive. In both cases
IBM needed to make price cuts in order to remain competitive. In both cases,
the price cut products were introduced as new products with identical i^erform-
ance specifications, the 2314 as the 2319 and the 2420-7 as * * * involved, as
well as lack of information or lethargy on the part of some computer managers,
allowed IBM to have a competitively low price product while still receiving the
higher rent from many customers for some time after the price out.

The third structural characteristic that IBM took advantage of was the neces-
sary time lag between IBM introduction of a product and PCM introduction of a
replacement. In the early days of PCM competition, IBM overestimated this lag
and erroneously felt that it would make the PCM competition a minimal threat.
Later. IBM introduced the fixed term plan and more rapid minor product changes
in order to capitalize on the time lag. When a new product was introduced, cus-
tomers had little incentive not to accept it on the fixed term plan because no
competitive replacement was ready. The heavy penalties for early termination
under the FTP effectively reduced the PCM's prospects to those customers finish-
ing a lease. By making changes to the products rapidly (such as switching con-
trol functions between drives and control units), IBM was able to further reduce
the PCM's marketing effectiveness without reducing prices.



5680

Summary of Product Characteristics
DISKS drives

2311 Primary direct access storage devices for small scale 360 users, each

drive can access seven million bytes (characters) of data, transfers

data at 156,000 bytes/second.
2314 Primary direct access storage device for large scale 360 users, each drive

can access 29 million bytes of data, transfers data at 312,000 bytes/

second, must be usetl with a 2314 controller which can control up to

eight drives.
2319A Three 2314 drives put together in a single box, attaches to an integrated

file adapter on the 370 instead of a 2314 controller.
2319B Same as 2319A except it attaches to a motlified 2314 controller for use

on the 360 machines instead of only with the integrated file adapter.
3330 Primary direct access storage device for large scale 370 users, each drive

can access 100 million bytes of data, transfers data at 800,000 bytes/

second, originally used with a 3830 controller which could control up

to eight drives.
3333 3330 with some of the controller electronics contained in the same unit.

3830 Control unit for the 3330.

3830 II Control unit for the 3330 after August, 1972, controls up to 16 drives,

must be used with a 3333.

TAPE drives

729 Family of second generation tape di'j.ves, also used with the 3G0.

2401 Main familv of tape drives used with the 360, included six models rang-

ing from' the 2401-1 which recorded data at a density of 800 bytes/
inch and transferred data at 30,000 bytes/second to the 2401-6 which
recorded at 1600 bytes/inch and transferred data at 180,000 bytes/
second.

2420 New family of tape drives announced in 1968 for use with late 360 s and

370's, included 2420-5 which transferred data at 160,000 bytes/
second and the 2420-7 which transferred at 320,000 bytes/second.

3420 A slight revision of the 2420 family announced in 1970, the 3420-5 and

3420-7 are identical in performance to 2420-5 and 2420-7.

References

1 IBM "A Plug Compatible Manufacturer Overview," December 13, 1970 (Telex
v.' IBM, Plantiff's Exhibit 142), (further references to Telex v. IBM,
Piantiff's Exhibit are designated by PX) .

2. IBM, "Management Committee Minutes and Charts," July 15, 1968 (PX 3S4B-

3 IBM, "Management Committee Minutes," July 30, 19G9 (PX 385-039).

4. IBM, "Peripherals Task Force Flip Charts," (PX 52), quoted in "Telex\s

Proposed Findings of Fact and Conclusions of Law Relating to Telex's
Antitrust Claims," p. 50.

5. IBM, "Peripherals Study." June 25, 1970 (PX41).

6. IBM, "O. J. Hesner to H. E. Cooley," May 4, 1970 (PX 31) .

7. IBM, "Peripherals Study," June 25, 19(0 (PX41).

8. IBM, "Mallard Financial Analysis," September, 1970 (PX 135). p. 16.

9. IBM. "Plug Compatible Competition," December 31, 1970 (PX 162), p. 5.
10 IBM, "Summary Comparison — IBM — Telex," (PX 280).

11. IBM, "Fassig Charts," (PX 671), quoted in "Telex's Proposed Findings,"

12. IB:\i, "Fixed Term Plan Effect on Direct Access Storage Devices," (PX 290) »

1). 4.

13. IB:M, "Plug Compatible Competition," December 31, 1970 (PX 162), p. lo.

14. IBM, "Greybook — Tapes and Control Units," February 8, 1971 (PX 114), pp.

22, 48.

15. IMd. p. 17.

16. IBM, "Phase II Forecast, Aspen I, Monarch," June 15, 1970 (PX 272), p. 9.

17. IBM. "Greybook— Tapes and Control Units," February 8, 1971 (PX 114),

p. 22.

18. IBM, "Plug Compatible Competition," December 31, 1970 (PX 162), p. 33R.

19. IBM, "Greybook— Tapes and Control Units," p. 84.



5081

20. '"Telex's Proposed Findings, " p. 73.

21. IBM, "Quarterly Product Line Assessment," March 16, liiTl (PX 224),

p. 162.

22. ir.M. "PCM Control Book," September 28, 1971 (PX 344), p. 2.

23. IBM, "Plus Compatible Competition," December 31, 1070 (PX 1(52), p. 20.

24. IBM. "Management Review Committee Minutes," May 6, 1971 (I'X 391A-

054).
2o. IBM, "Lease Plan- Plug-Compatible Manufacturers," (PX 141A).

26. IBM. "Management Review Committee ]Minutes," May 20, 1971 (PX 391A-

0.57 ) .

27. Various articles, Datawaiion XVII (July 1, July 15, 1971).

28. IBM. "Fixed Term Plan Elffect on Direct Access Storage Devices," (PX 290),

pp. 7. 8. quoted in "Telex's Proposed Findings," pp. 95, 96.

29. IBM. "Reaction to FTP," July 19, 1971 (PX 331).

30. IBr^r. "Fixed Term Plan Effect." (PX 290). p. 5.

31. IBM, "Quarterly Product Line Assessment," December, 1971 (PX 129),

p. 106. quoted in "Telex's Proposed Findings," p. 96.

32. IBM. "FTP Decreased Order Rate by 447c," (PX 670).

33. Telex. "Telex Stock Value," (PX 681).

34. IBM, "F. A. Powell to D. P. Pliypers." August 5, 1971 (PX 1.52).

35. "Director Final Functional Specification," (PX 253), p. 11, quoted in

"Telex's Proposed Findings," p. 101.

36. IBM, "T. Y. Learson to P. J. Rizzo," April 20. 1970 (PX 472).

37. IBM, "Memorv Analysis," June 29, 1970 (PX 279).

38. IBM, "XS-370 Announcement," June 5, 1970 (PX 306), p. 27.

39. Il)i(J.. p. 28.

40. IBM. "A. :\I. Aldrich to J. R. Opel," June 5, 1970 (PX 306), p. 1.

41. IBM. "Grey book— System 370 Model 135," February, 1972 (PX 115), p. 36.

42. IbiiJ.. p. 41.

43. IB:M, "B. M. Hoclifeld to D. J. Perry." March 15. 1971 (PX 1-53), pp. 3. 4.

44. IBM. "B. M. Hochfeld to D. van Alderwerwelt." March 17. 1971 (PX 154),
4'5. IBM. "Management Review Committee Minutes," July 15, 1971 (PX S91A-

088), pp. 10, 13.

46. Ibid.

47. IBM, "IBM's Proposed Findings of Fact and Conclusions of Law Relating

to Telex's Antitrust Claims." p. 210.

45. IBM, "Greybook— System 370 Model 165." .Tanuary, 1971 (PX 118), pp. 24,

27, and "Greybook— System 370 Model 155," November 14, 1970 (PX 117),
p. 11.

49. "Plan 29B — PCM Memory Assumptions for Olympus, Pisces, and Phoenix,"

October 21, 1971 (PX 375).

50. Ibid.

51. R. A. McLaughlin. "IBM's Virtual Memory 370's," Datamation XVIII (Sep-

tember. 1972). p. 58-61.

52. Hochfeld transcript, p. 1762, quoted in "Telex's Proposed Findings,'' p. 114,

53. IBM, "IBM's Proposed Findings," p. 214.

54. R. A. McLaughlin, "IBM's Virual Memory 370's," p. 61.



Exhibit 3. — Paper by Gerald Broelc Re Structural Reorganization of
Computer Industry

Structural Reorganization of the Computer Industry : An Economic

Analyse

By Gerald Brock, Assistant Professor of Economics, University of Arizona,

February 22, 1974

[Note. — This paper was prepared for use of the Subcommittee on Antitrust and Monop-
oly under a contract between the author and the Committee on the Judiciary of the
United States Senate. It is not to be construed as the statement or position of the com-
mittee or any member thereof. ]

SUMMARY

This paper first reviews the known relationships between an industry's struc-
ture and its economic performance. These studies indicate that barriers to the
entry of new firms into an industry is a greater cause of poor economic per-



5682

formance than high concentration. The stimctiire of the computer is then
reviewed, showing that the industry is highly concentrated with one firm
controlling about 70% of the market, and that barriers to entry are extremely
high. The wealvnesses that exist in the performance of the industry are shown
to be a result of barriers to entry. Consequently, in order to optimize economic
performance, the proposed reorganization is designed to reduce barriers to entry
while avoiding either disruption of current customers or losses in the efficiency
of operations.

The proposed reorganization is to split IBM into four companies: (1) a
peripheral company containing all of IBM's production and research facilities
related to peripheral equipment, (2) a maintenance company containing all of
IBM's maintenance personnel, (3) a marketing, leasing, and consulting company
containing all of IBM's marketing personnel and consulting and application
programming capability, (4) a central processing unit company which would
produce CPU's and system control programming. Discussion of the proposed new
companies indicates how barriers to entry would be reduced and examines poten-
tial problems that could arise with the proposed reorganization.

/. Background Economics

In considering the possible reorganization of the computer industry, one of
the first questions is what economic science can tell us about the relationships
between an industry's structure and its economic performance. The confidence
with which the structure-performance links can be predicted is also a measure
of the confidence with which we can prescribe structural reorganization as the
cure for performance problems. If the structure-performance links are weak
or unpredictable then we cannot be sure what the results of any particular
reorganization will be ; if they are strong and regular, we can predict with
confidence the economic results of any proposed reorganization of the industry.

Economic theory gives the most definitive results for highly simplified market
structures such as perfect competition or pure monopoly. The computer industry,
as with most major U.S. industries, can be classified as an oligopoly. Theoretical
results relating oligopoly structure to an industry's economic performance are
very weak because of the many possible variations in oligopoly behavior. Although
economic theory cannot provide definitive relations between variations in the
structure of an oligopoly and variations in its performance, the theory can.
be used as a guide for gathering and analyzing data to formulate the structure-
performance links empirically.

Economic theory tells us that the conduct of an industry (price and product
strategy, research and innovation, advertising level, attempts to coordinate
actions with other firms in the industry) should be a function of the industry's
structure (number and size distribution of firms and barriers to entry). With
low concentration and low barriers to entry, the industry's conduct should ap-
proach that of a competitive industry, while with high concentration and high
barriers to entry, the conduct should approach that of monopoly. Structure
and conduct together should determine the economic performance of the industry
(allocative efiiciency, effects on income distribution, and progressiveness or
technological innovation ) }

Precise definition of the interrelationships among the variables mentioned above
would obviously be extremely useful to antitrust questions. If a reliable statisti-
cal estimate of the structure-conduct-performance links existed, then we would
predict the results of any proposed reorganization of an industry or restrictions
on its competitive analysis of the industry in question. Three problems stand in
the way of developing such an equation. The first is that of quantifying the
variables. In quantifying a variable such as "number and size distribution of
firms", one must use a measure such as the concentration ratio (the percentage
of the industry controlled by the largest four firms). This measure causes an
industry such as automobiles to appear more concentrated than the computer
industx-y because it combines the influences of number of firms and relative
sizes of firms. Brand lo.valty is a significant component of barriers to entry
but can only be quantified by such measures as the advertising/sales ratio or
a subjective index compiled by the researcher.^

1 For an extensive study of structnre-condnct-performance interrelationships, see F. M.
Scherer. "Industrial Market Structure and Economic Performance" (Chicago: Rand Mc-
Nally & Company. 1970).

2 For a study of harriers to entry and the problems in measuring them, see Joe S. Bain,



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 98 of 140)