0 ratings 0% found this document useful (0 votes) 873 views 22 pages Product Proliferation and Preemption Strategy Supplements
This paper attempts to extend the set of applications to a situation in which the visible hand of regulation was largely absent. The focus is the u.s. Ready-to-eat breakfast cereal industry, which is the subject of a major antitrust case brought by the federal trade commission.
AI-enhanced title and description
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content,
claim it here .
Available Formats
Download as PDF, TXT or read online on Scribd
Carousel Previous Carousel Next
Save Product proliferation and preemption strategy supp... For Later THE JOURNAL OF INDUSTRIAL ECONOMICS
Volume XXVIII ‘December 1979 No. 2
THE WELFARE ECONOMICS OF PRODUCT
VARIETY: AN APPLICATION TO THE
READY-TO-EAT CEREALS INDUSTRY
FM, Screrer*
1 INTRODUCTION
Ever since the publication of Chamberlin’s Theory of Monopolistic Competition
{6}, economists have known that how well a market economy solves the
three-way trade-off among product variety, economies of scale, and purity
of competition critically affects the level of economie welfare Following a
long period of neglect, there has recently been an explosion of war on the
theory of optimal product variety. Although unsettled questions remain,
the theory reveals that whether monopolistic competition yields the welfare-
inaximizing amount of product variety, too little, or too much, depends
Boon ne relationships among several. potentially observable. vaneble
Except for the rather special cases of television Programming and service
competition among publicly regulated airlines, there has been Iiele effort
‘0 apply the theory to real-world market situations. This Paper attempts to
extend the set of applications to a situation in which the visible hea of
regulation was largely absent.
The focus is the US ready-to-eat (RTE) breakfast ccreal industry, which
is the subject of a major antitrust case brought by the Federal Trade Com-
mission.? The FTC's complaint alleges that leading manufacturers
Kellogg, General Mills, Post (i.e. General Foods}, and Quaker, accounting
for approximately 90% of RTE cereal sales during the late 1960s, collec
ively ‘monopolized the industry and engaged in divece monopolistic
Practices. One charge is that the four respondents deterred new entry by
proliferating product varieties, thereby leaving insufficient room fre viable
Mrould ats Tn other words, they pre-empted the entry of what otherwies
fngts sns & Batel fo Ronald racutgam, Rickard Schalemee, John Begin
‘sHlan urn and an anonymous eee ipa sere e ,
Heading Le], Shae tnd Quen [oul Note a)" Douglas and Mae op
rs at of Reg ol PT decks n. Sy Fd and Dos ad Maer)
SRE ate Eovermment on mater he mon partly dea ae ga De
respondent,
“Tor the theory ofthis allegation, se Schmalensce [26}.
8 ng
Copyright © 2001. All Rights Reseved.ng FM, SCHERER
that last dificult step isthe aim of this paper. The paper begins by reformu-
rating the received theory of optimal product variety & mort empirically
sactable form, proceeding then to relate the available evidence to the
theory.
Dubjce we begin, an overview of the RTE cereal industry's sructise and
performance may be useful. The SIC industry of which it is the principal
Component, ‘cereal preparations’, is one of the 20 mos concentrated four-
digit US manufacturing industries, Among 412-451 four-digit industries,
cteeal preparations had in both 1967 and 1972 one of the fet lowest ratios
oa vats plus plant payroll costs to industry sales—52%/o in both years.
Sains had tess plant capital and inventory per dollar of sales than the
average manufacturing industry. Thus, in-plant costs were unusually low.
Obversely, advertising expenditures and profits were unusually high.
‘Among 324-329 four-digit manufacturing industries for which comparable
input-output statistics are available, cereal preparations had the second
highest ratio of advertising outlays to sales—23%/ in 1963 and 18°5% in
1967.6 Accounting profits after taxes plus intrest ‘averaged 19°8% of assets
se tne RTE operations of leading firms Kellogg, General Mills, General
Foods, Quaker, and Ralston between 1938 and 1970. For all manufacturing
corporations, the comparable figure was 8-70. Or when advertising expen-
Serres are capitalized and depreciated at an annual exponential decay
sate of 39% the adjusted 1958-70 afier-tax return on asset for the five
RTE firms was 17-89%) compared to 84% for all ‘manufacturing corpora-
tions.”
From the perspective of this article, an especially importa feature of
the RTE industry's recent history has been the introduction of numerous
ew product variants. Key innovations are widely considered to have been
presweetened cereals, with Post's introduction of Sugar Crisp in 195°
raring the first major success; nutritional cereals (with Kellogg's Special
K leading the way in 1996);® and natural cereals. (long available from
fringe sellers, but enjoying a demand surge in the early 19708). We concen-
fringe ere on the principal firms’ activities following the introduction of
Special K and preceding the natural cereal boom: ‘At the end of 1957, the
Be largest companies had a total of 38 RTE brands ip national distribution.
Se aren to the close of 1970, they carried 51 new brands beyond the
regional test stage while withdrawing 22, Teaving a total of 67 brands in
« Anayis of company data showed that the RTE subset of corel PERS Sp had an
eet Anais of company spbgrgy of the FTC hearing transcript, whch Wil subsequent
Ye gue TEs, cing Orsi 29) PP 17-8)
“ie 676-8, citing Ornstein (23) PP, 77-85),
CT 27.6768, St Oa er ae Te 193069 summarized in FTC cx-7o1.
(The prefix OX indicates that an exh
: les that an Or le company fromn whose records the document
FTO). A eed etter, irany, identifi the exhibit page)
ire, A ed ete Fe General Food 1967 ik fort uy en exhibit
Fefoyo. The same document oberes at p. 274 Hat “Good ideas for new cereals are
Rg ecroduced by complaint counsel (ie. the
Copyright © 2001. All Rights Reseved.‘THE WELFARE ECONOMICS OF PRODUCT VARIETY 15
national distribution in December 1970. Many of the new brands failed to
achieve viable market shares, and of the 51 brands launched between 1958
and 1970, only five—Kellogg’s Froct Loops, General Mills’ Total, Pos
Alpha Bits and Quaker’s Life and Captain Crunch—succeeded in obtaining
for a year or more a market share of at least 2%. Thus, the newer products
characteristically filed only small niches in the spectrum of consumer
demands. In contrast, older brands showed considerable staying power.
Only two pre-t958 brands were withdrawn from national dievibeany
during the 1960s, and the 29 leading brands of 1960, with 83% of total
industry pound volume then, retained 74% of the market in 1970 despite
the competition of numerous products. The key question is, was the coreal
‘makers’ proliferation of product variants carried beyond the point at which
€conomic welfare was maximized ?
11. THE THEORY
The problem of optimal product variety has been approached theoreti-
cally in several ways: using geographic spatial analogies following the lead
of Hotelling,? in an orthodox consumer wilty fanction framewerk.10 vith
Tancasterian characteristies preference models! and using demand frac,
tion and consumers’ surplus methods. The model employed here is a hybrid,
Adapting Spence’s arcas-under-the-demand-function focus (29), [30] to a
Loesch-Hotelling [19], [12] spatial approach, As a hybrid, it corsinecs
clegance to bring into clear focus relationships one can observe and attempt
to measure,
Figure 1 provides the simplest posible view. The horizontal axis represents
@ one-dimensional space over which a product's real or perceived characteris-
tics might vary—eg. in the degree of nutrient fortification, or sweetness,
Counchiness, or (with discrete rather than continuous variation) shape,
Consumers’ preferences for a given characteristic are deseribed by their
Jecation in product characteristics space. If, for example, the relevent
dimension is sweetness, a consumer who likes a relatively sweet product
Will have his or her preference located farther to the right, eg, at point B,
than a consumer (e.g. at A) preferring a less sweet product’ We aera
provisionally a uniform distribution and intensity of preferences over the
relevant spatial segment, ignore the problem of finite segment bounds act
Feume Jacome effects to be insignificantly small.® Adapting Spence’s
15g Hotelling [12], Lovell (18), Meade [20], Stern [33], and Hay [11).
12 See Usher (35), Dixit and Suite [6}, 4nd Lelana Poy
2 bamestcr (ig) [s5} (06), and Archibald and Reseltich (4),
48Since RTE. cereal pul
Eepreseat only s very small proportion of the average
NEERLES expendi fori Income eles Cakes na samen O he average
ig (99)
Copyright © 2001. All Rights Reseved.16 P.M, SCHERER
Totol Surplus
Producer's
Surplus
Product Characteristic ~
Fioure 1, Surplus Functions in Product Characteristics Space
whose preferences (or the relevant part of whose preferences!#) match that
point, given the prices of the product(s) purchased by those consumers
pad also the prices of al substitute and complement products. It is convenient
fo net out from this value the variable costs of production and distribution.
Thus, Figure 1 presents two sets of tent-like functions. ‘The upper (solid)
fanetion shows at any abscissa point the total surplus (consumers’ plus
producer's) realized from sales to consumers with preferences at any given
Point in product characteristics space. These ordinate values, it must be
Emphasized, are uniquely defined only fora specific set of products and pro,
duct prices, Our analysis must deal therefore only with incremental localized
product availability changes.
Figure 1 assumes the supply of only two products A and B. These products
have characteristics exactly matching the preferences of consumers located
at points A and B in characteristics space. By assumption, however, con-
Sumer preferences are distributed uniformly over space. Those whose pre-
ferences are imperfectly satisfied by A or B must either buy A or B or spend
more of their income on unrepresented (perhaps quite different) products.
The loss of utility experienced by a consumer with preferences at, s2Ys
boundary point AB from having to make do with imperfect substitute A or
B is analogous in a geographic spatial model to the transportation cost @
Conmumer incurs securing output from a geographically distant plant under
FOB pricing. Given downward sloping demand functions, a higher delivered
price in the uniform density geographic model means that less will be con-
Pimed by distant customers than by customers located near the plant, all
flee equal. In the variable product characteristics case, the loss of utility
from consuming less-than-ideal products leads analogously to a lower
2a There no reson hy single consumer cannot have multe Jocations, reflecting a
preference for variegated product
Copyright © 2001. All Rights Reseved.‘THE WELFARE ECONOMICS OF PRODUCT VARIETY 17
quantity demanded, the more distant the demander’s preferences lie from
the characteristics of available products. This, combined with the assumption
of @ uniform distribution of preferences over characteristics space, gives
re to the tent shape of the surplus functions, whose maxima nee at the
esans that consumer preferences in the neighborhood of C will be sxtttet
better, leading to an increase in consumption by thove consumers, and hence
iershaped functions with maxima above point C, Purchases wil, normally
be divided among products at boundaries (ue. AC and CB) where the total
surplus functions intersect, Assuming uniformly distributed preferences,
this implies an equal division of the market,
Whether or not product C will be offered depends critically upon the
fixed costs of launching it—e.g. the required investment in research and
A aC c ce 8
Fiount 2. The Consequences of Product Launching
development, plant and introductory marketing—and upon market structure,
To mesh these pieces of the problem, it is desirable to let the producers’
and consumers’ surplus functions measure discounted present vaio rather
than single-period flows. Market structure matters then in the following
sense. If the market is (at least locally) monopolized and entry at C by
Outsiders is blocked, the relevant surplus (or more accurately, discounted
Present value of quasi-rents) to the producer of A and B contemplating
introducing C is the diamond-shaped area consisting of Ts, To, Pa and
rane Producer's surplus increases. On the other hand, if entry if
* monopolistic competitor considering offering C in competion
Copyright © 2001. All Rights Reseved.118 F. M, SCHERER
with other firms’ products A and B and assuming, Cournot-like, that C
Js priced at the same level as the unchanged prices of 4 and B, perceives
itslf as gaining that diamond-shaped surplus plus Ke and Ke» which repre-
ieer amfers of quasi-rents from the producers of A and B to the producer
“ero Such transfers are often referred to as ‘cannibalization’ among business
people, and we shall name them ‘accordingly.!4 The monopolistic competitor
eoreerarned about whence Ke and Kp come views itself as gaining 9 larger
surplus through new product introduction than does the monopolist. It is
entirely possible that fixed launching costs exceed Ta + Ty+PotPo but
cee shan that sum plus K and Ky. Ifs0, product C will be introduced by
a snonopolistie competitor but not by a monopolist. This is the principle
tunderlying economists’ conclusion that greater product variety will emerge
tinder monopolistic competition than under pure monopoly.
Consider now a modification in the structural assumptions. If’ a mono”
polist supplies A and B but entry into Cis open, the ‘monopolist may choose
ee eer er C and pre-empt entry that would otherwise cause it to lose
Surplus Ka and Ky. If'so, as much product variety may emerge under such
cntry-deterring monopoly (or oligopoly) as under monopolistic competition,
other things (such as price) being held equal.
Tt remains to be seen whether the variety of products supplied under
monopolistic competition or entry-deterring monopoly is optimal. Assuming
that a dollar of ‘consumers’ surplus is valued equally with a dollar of
producers’ surplus, a new product should be introduced if its addition to
Freplus net of transfers—that is, pure producers’ surplus ineresst PatPo
plus consumers’ surplus gains S_+Sy—exceeds the fixed cosis of introduc
tion. That diamond-shaped arca may be either larger or smaller than the
wall-tent shaped surplus area guiding the decisions ‘of monopolistic com-
petitors or entry-deterring monopolists. As Figure 2 is drawn, the two arcas
Peete ar magnitude, and so decisions motivated by profit maximization
ar sor actual or potential monopolistic competition will tend also to advance
social welfare.
“This is not necessarily so, however. Figure 3 provides a counter-example.
‘The exten of cannibalization Ke plus Ky by product C is large relative to
‘he net inerease in social surplus Sq plus Sp. (Note that all producers! surplus
gains now come as transfers from what was consumers’ surplus when there
were only two products.) If fixed costs are smaller than Ka+KytTetTo
put larger than Sa-+Sp, excessive product proliferation will occur.
igure 3 differs from Figure 2 largely in the gentler slope of its surplus
functions, which in turn, given uniform preference density over characteris-
tice space, reflects greater substitutability of the products, ‘When products,
are relatively good substitutes for one another, as in Figure 3, not much
a4 See alo Schere:[25, pp. 392-4], where a distinction is drawn between new market and
marketsharing innovations.
acket sharing inne OR chibald and Rosenbluth [4], and Spence (29), (30]-
Copyright © 2001. All Rights Reseved.‘THE WELFARE ECONOMICS OF PRODUCT VARIETY 1g
a ac c cB 8
Fiourr 3, Launching of a Close-substitute Product,
demand is choked off when consumers have to make do with a product
that departs by some given amount from their ideal’ characteristics In the
Uniting case of perfect substitutes (reflected by horizontal surplus functions),
adding a new product has no demand-expanding effect, and so there i
nothing but cannibalization. Quite generally, the better substitutes existing
and new products are for one another, the more likely it is that additionat
Product launchings will add more to costs than to social benefits, all else
equal.16
Cannibalization is also apt to be high relative to surplus gains net of
transfers when the ratio of producers’ to consumers’ surplus is high. ‘This im
turn has two main determinants. First, as Spence has emphasized, when
feller maximize profits under the conventional assumptions of monope.
listic competition, producer's surplus is higher relative to consumers” surplus
with inverse demand functions that are concave downward relatively
Fat-topped) than with convex functions.1” Second, one would expect higher
Producers’ surpluses, the more sellers respect their interdependence in
Pricing and co-operate toward joint profit-maximizing policies rather than
viewing (as we have assumed thus far) the prices of spatially proximate
Products as insensitive to their own price and product launching choices,
Summing up, we have identified four potentially observable characteristce
associated with @ tendency toward excessive proliferation of product
variants: high substitutability among the variants, modest increases in
Producers’ surplus net of cannibalization or (assuming constant prices and
unit costs) in sales volume, a high ratio of producers’ to consumers’ surplus,
and (following from these three) extensive cannibalization, Also, fixed
8 Se FSL Tha ton Send con nd ceva mat an
Panay turpis i onstant one-half of produce’ tur epee Be oa
Parameter value, assuming conventional moncpoly profit wie echo
Copyright © 2001. All Rights Reseved.120 P.M, SCHERER
product launching costs must be substantial, but less than the gross FP
aersers' surplus (including cannibalization) realized by an added product
variant.
‘Four further complications must be noted. ‘Thus far, we have assumed
that the prices of products A and B remain fixed when product Cis introduced
sr the seme price. A new entrant could, however, be greeted by @ price
var, or the new entry might be accommodated by the raising of existing
product prices.!8 Assuming that pre-entry prices exceed marginal cost,
vraform price increases reduce total surplus and hence shift the total
Surplus functions downward; uniform price decreases have the opposite
tffect (unless and until marginal cost is undercut). For non-uniform or
Gieriminatory price changes, it is more difficult to state simple generaliza-
tions.
Second, our analysis has no way of determining what set of prices is
lobally optimal in either a first-best or second-best sense, It can only
saluace the desirability of marginal changes in the array of products
Given this limitation, the most that can be said is that if the prices of pre-
trusting products remain unchanged or increase in response to the introduc-
{ion of new product variant, the price effects of the introduction on non~
focal welfare will be either neutral or adverse incrementally. Thus, any
surplus of launching costs over loca! net surplus gains will not be ‘compensated
by surplus gains attributable to second-order price effects in other regions
of product characteristics space.
Thurd, consumer preferences may not be uniformly distributed and os
uniform intensity over product characteristics space. This, we shall see,
\will cause empirical difficulties
18 For the analogous cave of nev: plant entry into geographic space with lingar point ine
space demand functions, itcan readily be shown thatthe entry ofa new PIOr4 the boundary
reavontaats wile an inerese in prices ifthe operator of the okd and new
Bins maximize joint profs, The proof here follows Hay [11].
ants me ction for consumers located 2 kilometers from a plant be linear, ¢+
a qr=a—b (Pot IZ),
0 ve Pois the price FOB plant and ts the transport cost pe Llometer fora unit of product,
where Pes te Prep urnformly dense over gne-dimensional geographic space, Then fo &
apt sipping out to MF ‘Miprmeler in each direction from its location, the total quantity
® ant feo reine
= 2eM—2b PeM—bt M?,
With constant marginal cost ¢ and fixed costs of F, the plant's profit is
@) ‘w= (Po—<) (22M —2b PeM—Bt M?)—F.
If each plant has the same marginal production cost ¢ and transportation cof 250 ‘fall
Hae ae etiat profs, the optimal location of a new plant will be midway
‘between existing plants, all sellers will the same price, and conjectural variations can
berween existing Ptacse assumptions, profis are maximized by setting 9x/2Po= 0, from
which one obtains the relationshi
@) Pot al2b-+e/2— tM.
Cleary, as entey reduces the spacing between plant boundaries the joint profitemaxi-
izing price Pe* rises.
Copyright © 2001. All Rights Reseved.THE WELFARE ECONOMICS OF PRODUCT VARIETY 121
cannibalization will be high and the net growth of producers’ and consumers’
fils an important unserved niche in product characteristics space, the
obverse is more apt to be true. Extending this reasoning to the cone of very
fhroughout a preponderantly competitive economy will for the nant part
be units of consumption at the margins where price equals marginal cost,
and no appreciable surplus, consumer or producer, is foregone. If the new
Product yields a surplus over launching costs, welfare is undoubtedly in-
Greased, More perplexing is the situation in’ which product interacns
occur weakly over a Jarge number of dimensions within a world of mono-
Polies, to use Mrs. Robinson's phrase [24, p. 307], Then the market's
invisible hand may lead new product suppliers into incurring costs to swap
eng furplus for another, but it will be difficule or even imposible in Practice
(0 detect ‘that welfare has been reduced,
ML. THE CEREAL INDUSTRY EVIDENCE
Win iiieCGu ee ee RTE
Product launching, the extent to which sales gains were real as ‘compared
to cannibalistic, various aspects of pricing behavior, and an attempt to
cstimate the magnitude of added consumers’ surpluses
A. The Costs of New Product Launching
The earliest recognizable costs of launching a new cereal product into
the market are those associated with research, development, market research,
and test marketing. According to a survey of 21 new cereal products by
Buzzell and Nourse [5, pp. 914], research and development costs in the
1959-64 period averaged $122,000, market research costs $60,000 and
the excess of other marketing costs over product sale gross margins during
the period of test marketing approximately $740,000, These costs, like most
others, have undoubtedly risen over time with inflation. Therefore, we shall
to the maximum degree possible emphasize cost and sales data centered
on 1962 for purposes of comparison,
Investment outlays for the equipment needed to produce a new product
ot sn anc an hy ht the hater Prac coats a he regi
see Schmalensee [28].
*° For an extreme view, sce Triffin (34).
Copyright © 2001. All Rights Reseved.122 F. M, SCHERER
vary with the degree to which existing capacity can be used with minimal
Vebuilding, At one extreme were products such as Total or Sugar Frosted
Flakes, which merely required the addition of vitamin or sugar enrobing
equipment to standard wheat or corn flakes lines.2! At the other extreme
would be a product such as Quaker's Life, whose complex shape necessitated
oni temnilon dollar investment in nevily designed machinery. Appreciable
Startup costs might also be incurred in debugging production equipments
put no evidence on their magnitude is available.
For many consumer goods, and perhaps especially for ready-to-cst
cereals, expenditures on introductory advertising and other types of sales
promotion (such as special displays, free samples and coupet, campaigns)
are a major element of new product launching costs. Although considerable
ar a on ean be observed between products the need to attain a threshold
are of awareness in national ‘roll-out? campaigns imposes a rough lower
tee id om the level of introductory advertising outlays. A study of 25
raatively successful RTE cereals introduced between 1957 and 1967 showed
average advertising expenditures to have been $805,000 in the first two
serge pational distribution and $3°5 million over the first 14 months,
renwbich the average rate of spending tended to stabilize temporarily at
$115,000 per month.2 Whether in calculating introductory advertising
coste’one should truncate at 14 months with the stabilization of monthly
outlays is a difficult question. The same study showed that market shares
cateed at an average of 0-9% after two years-# At 1g62 industry levels
this implies annual sales on the order of $4°5 million per brand and, with
sverdzing outlays averaging $1716 million per year i the third yea", 4”
sa vertsing/sales ratio of 269%, 2° Since the comparable ratio for all products,
ed and new, was 17% this suggests that launching costs may have con
awed into the third year of national sales. On this and related issues more
will be said later.
‘We find then that average launching costs in the carly 1960s were at
131 Of course, if a new product increases yany demand for the products made on such
GE cous a net Rive bee tre initially for both Total and Sugar Frones Tite
machines, 25 29Pe2r ete new lines may be necesary. This iy not alan 0, Noort
an invent Ge both General Mi and’ Post emphasied the development of n=
preduge that could we exting exc capaci
oducts that could CP ching cigarcte brands see Seer tl (2740. 249)
5 Proms toy ater Ons Co, sy ined the tread ease record as OX-Q2177-
oF juzell and Nourse [5
Flat Bee ar, been included, the average market ghare would have Wee
aie ran dy reports that between rog8 and 1970, 76 brands were
much lower, Toe Quakes Pmariet rational dtripation and that onby 2445 age
introduce into cet yor Bary market sucre had some impact of nosey
red oe eee not grea deal OF the 25 brands analysed th Bt eh succesful,
Nertsing levels, but B03 Brac tye atthe end ofthe third year, had inal 94 Ten
wth an average mayke\ aon, Te ax least sucesl with an average share of 6347
2eeegin8 Clon on average during thei Fit 14 months.
Et 88 milion oo rabt share, which baed upon poundage, Gece 1962
indy ses, te product ale mate oh, An upward adjustment was made
act
7
industry sales, the produc 16 als mere priced Toughiy 30% higher per pound than old
cereals: those on the market by 1954- aa
Copyright © 2001. All Rights Reseved.THE WELFARE ECONOMICS OF PRODUCT VARIETY 193
least $4:4 million per product (ie. $122,000 +.60,000 + 740,000 +3'5
‘million) and perhaps, if new production facilites were needed and extercae
Promotional activity other than advertising was undertaken, a good deal
more.
B. Real Growth vs. Cannibalization
The cereal antitrust case record yields mixed and even contradictory docu-
TentarY and testimonial evidence on the degree to which new products
contributed to genuine market growth, as distinguished from cannibalizing
existing product sales and quasi-rents. In a 1966 speech to his field sale
Personnel, a Kellogg vice-president lamented the appearance of ‘so many
meaningless new products” and ‘the futility and the wastefulnese of having
i much of a good thing’.2® The same official observed three years lates
that “for the past several years, our individual company growth has come
out ofthe other fellow’s hide’? Analyzing the sources of anticipated patron,
age for its OKs product, introduced nationally in 1959, Kellogg marketing
staf listed a last of ive in importance ‘people who are not eating any ready-
Ceti e reals currently’. Brand-switching consumers of rival products
Cheerios and Alpha-Bits were listed first.28 A 1958 Post (i.e. General Foods)
cing Crandum observed similarly that ‘the pre-sweetened market is begin.
ning to show signs of inexpansibility” and that expected rival product,
introductions would ‘further fraction the category and make the marketing
of established pre-sweets more difficult in FY 19602 A marketing analysis
for a new pre-sweetened Post product introduced in the late 1960s estimated
that half the product’s sales came from cannibalization of an elder Pot
brand, 30% from cannibalization of a rival Kellogg product and the son
from all other RTE cereals Similarly, a 1971 General Mills analysis
found that in the highly fractionated, mature market for re-sweetened
cotcal, ‘much of the new product growth has come at the expense of ccc
ablished pre-sweets, which, as a category, have shown regular declines-s0
One can also find statements imputing to product innovation a favorable
impact on overall market growth. There appears to be something approach-
stimulating effect. A 1954 statement of Post cereal marketing philosophy
indicated that ‘Both new products, specifically pre-sweetened cereals, and
regular products have contributed to the industry growth’ 32 A 1965 General
BEBE {OG oC. H. Tornabene, CX-K5490 and R (empha in origina),
noe
et Gsferences to extensive cannibalization include CX-GF-4B,
< §25f) CXGE-Goab, CK-GE-igoab, CREE ays CRG aE
CS-GF 3008 2 155, CK 4098, CAKE] and OSG EE He
*OXCCF-4A, Bt compate py B ofthe tame document Set Sho GAGE go00 2 3
Copyright © 2001. All Rights Reseved.124 F, M, SCHERER
Mills memorandum said that ‘In the early 19508 the advent of pre-sweetened
cereals opened whole new avenues of volume and profit potential’ and that
“the sales of presweetened cereals have expanded the total cereals market
wee have not adversely affected sales of regular cereals’.#9 Discussing the
fortheoming introduction of a pioneering high-nutrition cereal, a Kellogg
creoutive predicted in 1956 that ‘with Special K, the possibilities are almost
cartmited for getting new users who seldom use any ready to eat cereal’
TTow does one reconcile these differing views, some of which are squarely
inconsistent? One variable apparently influencing company perspectives
vice the overall state of demand, Over the two decades following World War
TL, the RTE industry experienced nearly continuous pound sales growth
averaging 3.5% per annum. Decision-makers’ attitudes were buoyants
and many attributed at least a part of their good fortune to the concurrent
surge of new product introductions. However, from 1966 to 1970 the growth
rate dropped virtually to zero and continuing new product activity came to
Le uiewed more skeptically, with visibly more awareness of cannibalization.
fhlso, a more sanguine view of new products’ pure growth-inducing effect
aeaded to be taken by the executives of General Mills, which gained market
seine during the 1966s, than by officials of Kellogg, which only maintained
srethare until the the late 19608 and General Foods, which lost ground. It
1 amlear too that industry analysts devoted inadequate attention to the
sremtaneous movement of demographic variables and as a result, their
accounting of the sources of growth was often incomplete.
‘Kn attempt t0 take the fullest possible range of relevant variables inte
account wae made by Richard Schmalensee in an unpublished statistics!
seady of annual RTE cereal consumption per capita from 1953 10 1971.35
‘Among the variables included in his time series were the following:
Q Pound sales per capita.
Real disposable income per capita.
Dem Population weighted by an index of cereal consumption in various
age groups, divided by total population.
Using the Cochrane-Orcutt transformation (with p=0'575) minimize
Jnefficiency owing to autocorrelation, he estimated the following regression,
with t-ratios given in parentheses:
=164,
5-90 032 1 yr. pin
Log Q=-0.02+ (4 goyl08 Dem (4.74)!98 24% 0-974, D.W.
Both the income and weighted demographic variables were statistically
significant. When a measure of the number of brands nationally distributed
inveach year was added to the regression, its ratio was only 0°42, suggesting
io cathe mnereasing number of brands over time had no significant identifiable
38. OX-GMI-«11A. See also CX-GMI-2628 and CX-GMI-2634
34 Speech text of Mard Leaver, CX-K-360C. See also CXR S051.
3: Peesthadebted to Professor Schmalensee for providing these materials,
Copyright © 2001. All Rights Reseved.‘THE WELFARE ECONOMICS OF PRODUCT VARIETY 125
effect on consumption, Also insignificant was a variable measuring the
Proportion of working wives. Because the data were insufficient to estene
& Froperly identified simultaneous equation system and also because af
multicollinearity, relative price indices had implausible positive signs and
were therefore excluded.
The coefficients in Schmalensce’s regression can be interpreted as elastiie
ties, and the high value of the Dem coefficient is puzzling, Schmalensee’s
« Priori hypothesis was that its value should be near unity, One apparent
reason for the unexpectedly high value was that the income and dene,
graphic variables were strongly collinear up to 1965, after which er capita
income continued to grow while the growth of cereal consumption and the
demographic variable slackened—the latter because children are particu-
larly heavy RTE cereal eaters and the cohort of children 12 years old and
pier fell in size from 1965 to 1970. The demographic variable was thee,
fore forced to bear the burden of explaining the ost-1965 stagnation of
consumption growth, taking on an implausibly large coefficient walue ne
consequence.
A simpler but in some ways more revealing picture of these developments
by five-year intervals is provided by Table I. The first numerical column
Tame
CROWTH TRENDS IY US RTE CEREAL CONSLMPTION AND POPLLATION, 1950-708
RTE cereal
ay” Beraath, — Population 4 growth, Weighted % prowth,
Tear “Tooolty” “greats Poult Sates ion fotears
1950 609,890 150.607 a
1955 783,087 8-4 165,248 97 269 8
1 Sy0.5 20-1 180,670 93 erase 98
1965177 54 25-2 194,503 77 97343 uy
O"5 3
1970 1,183,100 5 204,879 3:3 204,779
pleigt, Cereal consumption, CX-101E, Population, US Buvagu of the Census, Current
[puation Reports, Series P25, Nos, 470 and 5193 and Slotiaign Wier Of the United States:
1957.
shows pounds of RTE cereal sold through grocery stores and the second
column gives five-year growth rates, There follow two sets of demographic
indices, one using a simple population count and the other population
HI baby boom are all evident. What stands out is that ia every five-year
Period but the last, the rate of cereal consumption growth is toons than
double the growth rate of the weighted population variable, ‘The would
at uni 1200 being the average forall consumers, the weights by Age group are: 12 years
id "023i 20-44, 0-699: and 45 and older, 0-973" The soaeee
Copyright © 2001. All Rights Reseved.126 F, M, SCHERER
appear to indicate that something more than mere demographic change—
apPrmyvably, the proliferation of pre-sweetened and nutritional cereal
variants—stimulated consumption growth.
vtnie i ail not the full story, however. Consumers evidently choose RTE
servings more on the basis of bull—.g. filling a standard 23 in? cereal bowl—
serving Tne basis of weight.” Pre-sweetening raises product weight without
vreasarily affecting bulk. Census data show that sugar use by the cereal
Jndustry increased from 67-8 million pounds in 1954 t0 420 million pounds in
1972. RTE cereal output meanwhile grew by 613 million pounds. Thus,
roughly 575% of the cereal output increase consisted of sugar. Assuming
Seen ae ce would otherwise have dipped commensurately into their
Sugar bowls in the absence of presweetening,** this part of the industry's
waeerded output growth is merely a transfer akin to vertical integration and
wot real growth. Applying the 57°5% transfer factor to the Table T data for
1955-70 (the period corresponding most closely to that for which Census data
1a able), one finds that total non-sugar growth of RTE consumption
see 21-79. This compares closely with the 23°9% growth of weighted
population, suggesting that most ifnot all ofthe industry's non sAger growth
corresponded to demographic changes. ‘Although other variables such as
increasing real incomes, a rise in the proportion ‘of working wives, a general
‘rend toward consumption of convenience foods, the appearanee of other
wrerM convenience breakfast foods, and relatively rapidly rising RTE cereal
prices were at work, this plus Schmalensce’s regressions suggest rather
Prongly that one cannot attribute much real grosth of RTE ‘cereal con
sumption to the proliferation of new product varietis, It follows that new
products must have cannibalized the actual and potential sales and surpluses
of older products extensively.
C. Profits and Prices
If new products cannibalized producers’ surpluses from existing products
eae adding to fixed costs, all else equal, one might expect to observe ®
sre in industry profitability. No such decline is evident. If anyshing,
«Re trend was toward rising accounting returns on assets Guring the 1608.29
oie reveal makers avoided a decline in returns by effecting general price
increases and by seeking on new products relatively high gross margins.
‘As a 1962 General Foods marketing plan observed, the newer products
27 See CX-GMI-279D and testimony at TR 14454 and 17-445
Fee ex ie Q debatable. Many presnsectened cereal consist of 40% more by
weg et sucrome—the equivalent of adding to and 2 half heaping \sPoety ‘of sugar toan
weight of erring of unsweetencd pul o flake. To dhe extent thay Consent ‘would not
average 14 0 2 sdded ar much sugar, the problem of unknwing overicss raises doubts
consciously ana ecepted here that consumers know what they are getting, On this, see
CX-GF-4039 2 7-
3 CXs701 CP.
Copyright © 2001. All Rights Reseved.‘THE WELFARE ECONOMICS OF PRODUCT VARIETY. 127
faried higher profit margins to make their introduction affordable.4o
Between 1952 and 1970, the average wholesale price per pound of Rik
cereals sold through grocery stores rose by 81% while the general wholesale
Price index for finished consumer food products rose by 20°4% 41
Ax example of the premium pricing strategy adopted for new products
gan De seen in Kelloge’s Sugar Frosted Flakes, the bestselling peeawect
in 1970. The product is obtained by applying a sugar frosting to the vener-
able corm flake, Following an increase in the price of corn flakes offerte
July 25th, 1970, the wholesale lst price of Kellogg’s Corn Flakes inan cee
Package was 34'44 cents/lb while the price of Sugar Frosted Flakes in a 40.0
package was 44°67 cents/Ib. The main physical difference between the ove
is that the sucrose content of the Corn Flakes is approximately 7-8%,
while it is 29°0% for Sugar Frosted Flakes.t? To make a pound of Sugar
Frosted Flakes, one in effect adds 0-23 Ib of sugar to 0-77 bof Cons Flakes.
In mid-1970, 0°25 lb of refined cane sugar could be purchased on the spot
Gurket for 2:55 cents. Thus, using 2°55 cents worth of sugar to comes
Corn Flakes into Sugar Frosted Flakes yielded additional resenue at 18.15
cents.*8 The variation of product characteristics accompanied by heavy
advertising (in this case, emphasizing brand ‘spokesman Tony the Tiger)
made it possible to segregate markets and practice price discrinvinavion
Jr is also possible, though less clearly supported, that prices on older
Products were raised more rapidly than corresponding increases in dete
cost. This is the reaction one would expect from joint profit-maximizing
sellers in geographic space when new plant entrants pack the whew
Pricing in the cereal industry was definitely more complex than our
inital theoretical exposition supposed. Nevertheless, three inferences appear
Warranted. The average level of prices rose in real terms as consequence
of new product introductions. With uniform demand, this should have
sore gramiat ect
mp races at tt
Ess Rien A
1S rece rot mle’ cond asin Rpt,
Copyright © 2001. All Rights Reseved.128, F. M, SCHERER
shifted total surplus functions downward on average, all else equal. But
‘here wae also a probable increase in the dispersion of prices. Since some
products (such as corn lakes, which experienced competitive preset from
Private labellers) remained available at relatively Tow prices, it is conceiv-
Five, although by no means certain, that demands of low reservation price
coneumers continued to be satisfied.” If dead-weight welfare losses were
cermuated ax a result, the downward shift of total surplus functions could
saeieen mitigated. Finally, for reasons considered in section TT and also
ae a consequence of price discrimination, it seems likely that for any given
wc of total surplus functions, the ratio of producer’ to consumers! surplus
fore relative to what it would have been under a regime of constant and
‘uniform prices. Substantial redistributions from consumers) to producers’
surplus must have occurred. And the relatively high gross margins secured
by the cereal oligopolists must have stimulated product proliferation beyond
what it would have been had there been more competitive pressure on
prices.¢8
D. Consumers? Surplus Increments
‘To advance farther, we must accept assumptions that cannot be conclusively
supported by the available evidence. ‘Opting to err on the side of overstating
benefits, let us assume that total surplus curves were not shifted downward
em cn the average as a result of the price increases attending product
proliferation, Consistent with the evidence that cannibalization was exten-
Fe (and hence that spatially proximate differentiated products were fairly
se ampstitenes) and with the related demographic evidence that product
proliferation did not significantly increase physical consumption, we assume
that producers’ surplus gains were transfers from what otherwise would have
‘been consumers’ surplus. In other words, the situation is better described
by Figure 3, with non-cannibalized producers’ surplus gains Ta and Ts
reflecting transfers from pre-existing consumers’ surplus, than by Figure
ty in which producers’ surplus gains Pa and Pp come from the growth of
consumption without offsetting old consumers’ surplus reductions. To the
cianrvat these assurnptions are valid, the increases in welfare net of
exter before lnunching costs are deducted must be associated solely
‘with the consumers’ surplus increments Sa and So.
‘The are not measurable in terms of the analysis presented thus far.
However, they have an exact counterpart in the consumers’ surplus arcs
Howe vethat might be called the ceeris paribus demand function—that is,
untvee demand curve drawn under the assumption that all other pro
sects but the one in question continue to be available, so that what i
Guets eel is the consumers’ surplus gain from having the additional
SSE Seber (06 p94 as, a40) and Adams and Yellen (2}, [3]
Copyright © 2001. All Rights Reseved.‘THE WELFARE ECONOMICS OF PRODUCT VARIETY 129
$ / Pound
1.20
110
EMR D
6 7 8 9 10
Millions of Pounds per Year
Fiovre 4. Demand Function for an Additional Product
Product.«® There is reason to believe that these demand functions, at least
for most of the newer cereal products, tend to be concave downward, as
illustrated in Figure 4.8 The absence of a steep slope at low quantities is
(oyalty. Thus, a Kellogg memorandum addressed to the Federst Trade
Commission observed that ‘the fickleness of consumers toward Rite cereals
mrcans {hat manufacturers must continually remind consumers of then
Products or run the risk of being forgotten when the buying decision is
fad peane Spence (25, p.219). This approach accep ner alia the assumption of
defn dependent cored De SPR SE ai cereals the
adveta
among other thing stress ‘helping sis the cea (CXEr ogc) Tsim
contrived
‘traditional consumer choice
interdependence of preference, violating a basic assumptios
a
9
Copyright © 2001. All Rights Reseved.130 ¥, M. SCHERER
made’! Similarly, an internal General Foods planning. memorandum
Characterized the cereal business in 1966 as exhibiting » ‘multiplicity of
brands and categories with little, f any, brand loyalty’! On the other
hand, there may be a few consumers with intense loyalty to particular
new products. Their existence would give rise to the convex-downward
see eg curve extension above the question mark in Figure 4°
Tosight into the function's curvature in the vicinity of prevailing prices
is provided by the testimony of cereal company executives and internal
decision making memoranda. For products that are similar to one another,
tc lee clove substitutes than, say, Kellogg's Corn Flakes and Post Toasties,
retail shelf prices can apparently diverge by ‘a few cents @ pack’ without
having much impact on volume.%¢ But wider differentials were said in 1967
va eeome a potentially ‘significant factor in people switching’, oF move
produets into ‘the trouble area’, or even, in the case of Jow-volume products,
‘greatly weaken our efforts’ to keep them in business.S5 Cereal makers
generally chose not to maintain differentials wider than 4 cers Pet pack
fon similar products, which means that they must have considered their
oe nd functions to become quite elastic at prices exceeding those of the
nearest substitutes by more than that magnitude.
Tn view of this, we estimate that the average consumer surplus 0” a ty pical
11 oz package was 7 cents in 1970. This overstates the surplus for marginal
consumers and understates it for especially loyal and/or price insensitive
infra-marginal consumers. The implied consumers’ surplus per pound is
0-2 cents, OF 20% of average 1970 wholesale sales revenvc Pet pound.¢
We seek to compare such consumers’ surpluses on new brands, represeOt-
ing the social gain net of transfers, with the average launching cost of a
brand, estimated conservatively at S4r4 million for 1962, We assume that
the ratio of incremental consumers’ surplus to wholesale sales is 0-2. If
manufacturer sales volume associated with a percentage point of market
share in year ¢ is V(t), the market share achieved in year ¢ is M(t), and the
‘Hiscount rate is 7, we want to know whether the estimated discounted pre-
51 CX-K-10goZa
82 CX-GF-48526,
Bee aera son of a referee for this paper is apparenily, ote such exception.
Ignoring such exceptions leads to fismeasurement, but she C0 probably offset by the
Tenor t Brarplusecreasing price elects and the fixed investment necessitated by new
product launching,
eluet launching ogg executive R. F. Nicholls at TR 12,547 and 12.00%
Beran ey ak General Mills executive Robert Bodcau at TR 10009 along with
eee CES Mabo, See also General Foods new product pricing anal CX-GF
1g7M and OX-GF-1412J.
M and OX or viral rate is fairly high in comparison to the four leading cereal pro
Se tnt a rom eta on ales, which was approximately 1574 In A904 Tes ‘National
ducer’ before tas po fanketing (2ty P, 206]. The rae & low gompate 0 Bro eg
(qveraging about 50%) sae he ptice minus variable manufacturing cost margin (Averaging
about 62%).
Copyright © 2001. All Rights Reseved.‘THE WELFARE ECONOMICS OF PRODUCT VARIETY 131
sent value of consumers’ surplus increments:$?
(0-2) M(t) Vie)
se 2, (At
Market share data for products introduced between 1958 and 1970 were
available through 1975 at best, and there are gaps in many of the series
Internal gaps were filled by linear interpolation. When the available data
covered fewer than 20 years of product life, market shares were estimated
through year 20 by linear extrapolation and a remainder term for years 21
to infinity was computed, assuming conservatively a continuation of year
20's observed or estimated share.S® A discount rate of 0-09 was used, reflect
ing the average of values employed by Kellogg and Quaker in investmena
decisions.6° This may well exceed the economy-wide return on investment,
but it is below the marginal rate confronting a typical consumer buying on
installment credit. Because the launching cost estimate is centered on 1462,
the product sales data must be benchmarked accordingly, even though
new Products were introduced more or less continuously between 1958 and
1970: with inflation of both launching costs and sales figures occurring. as
time passed. This was taken into account by making two alternative eles
Per market share point assumptions: a $1-0 million value associated with
1962 sales levels and a $1-46 million value associated with 1967 sales."
Inequality (1) was estimated for all new RTE cereals introduced between
1958 and 1970 experiencing any appreciable success on the market, Taking
4962 sales as the benchmark, 17 brands satisfy the inequality-—that is, they
ielded estimated net consumers’ surplus in excess of estimated average
launching costs With 1967 sales as the benchmark, the lst is expanded vo
22 brands. The other ag brands known to have been launched nationally
clearly fail the test, having cost more to launch than they added to cones.
a) 5 $4-4 million,
oe Note that launching costs are not incurred precisely at time ¢=0, Of the etimated
Sed milion, $922,000 i incurred over several years prior to =e and the cent inde aes
doa towing 40. Any error introduced by centering the cost elias ee ae ne
ingeubtedly small. Fit year sales were discounted at midyear ie: with Sera gh foro
incrementing from that value for subsequent years,
3 CX-434,
are any Products were withdrawn from the market before year 20, in which case market
hares subsequent to withdrawal were uniformly sete
° See GX-Q-188B and GX-K-goz0
fatal by sae ake share data are denominated in pounds, the sales figures have been
of nae pY 28% Pelaive to total industry values to reflect the higher averagetonee ar ped
Sharpe tea Baal Benchmarks for years later that 196) webid wee pe eaah
ghe,rcpults becatue all the brands at the margin of sasiyeag inegeae ny ed
efclning mari shares over tine an all Suuenpner odie bya.
are Coroa Crispics, Product 10, Bran Buds, Froot Loops, Anple Joxe; Cocoa
Puffs, Frosty Os, Lucky Ch Ain her Tae rate Unluding Forth! Oar ake),
Grispy. Critters, Honeycomb, Al ife, QuispiQuake' (launched jointy), requlay
Capitin Crunch and Gorn Che ”
c dats or bee ON All Stats, Twinkles, Country Corn Flakes and Team, Although
ihe Sa Beans acing relly nai
ery ee pay narlctshare point by as much as 50% would appear to add ae me as
very few brands to the lst of me
Copyright © 2001. All Rights Reseved.132 FM. SCHERER
mer’ surplus, And although launching costs were presumably much lower,
mrmlarly adverse verdict must be rendered for at least 25 new products
o aene intvoduetion was terminated after test marketing but before nation-
we Simibution commenced. Thus, the analysis of consumers’ surplus
etrements suggests that product proliferation was carried well beyond the
point at which marginal social benefits equalled launching cos's
“Among the numerous doubts one might raise concerning these caleulations;
only one will be added here. The “fixed” launching cost catimate is quite
oon srvative in excluding plant investment and start-up costs and in termina
ting the accumulation of introductory advertising outlays at 14 months.
Sunaining advertising costs—ic. those incurred indefinitely after the Pro-
duct is Inunched—pose peculiarly difficult welfare accounting problems.
If'a potential new product were not introduced, sustaining advertising costs
eee re mat be incurred. Neither would the product's production and physi-
wn distribution costs. But given the apparently high degree of cannibaliza-
fon, production costs avoided if a new product is not introduced will be
offest by additional costs of turning out more of the old products. That the
crt ofsetting will occur with respect to advertising costs is much less
saecin. ‘There are threshold effects in sustaining as well as introductory
SGvertsing,#! and it seems reasonable to assume that with fewer products
sereritotal advertising outlays would be required to have each product
vee eh its threshold level of effectiveness. Also, many of the advertising
Tressages concerning spatially adjacent products are mutually cancelling,
Tiafering a net benefit on neither producers nor consumers. 1n view of the
generally high advertising/sales ratios prevailing in the cereal industry,
Gerermental increases in sustaining advertising and/or mutual cancellation
wrefaime owing to brand proliferation entail costs whose present valve
could be large relative to the value of initial product launching costs. Te
coun teat that this is true, substantially langer values of in inequality (1)
may have been needed to offset the total social costs of product proliferation.
E, Conclusion
In drawing conclusions at the end of such an analysis, one recognizes it is
not only the cereal industry's produets that can be flaky. It seems undeniable
aoe eee ready-to-eat product innovations have yielded net consumers)
plus producers’ surpluses welll in excess ‘of launching costs. But equal
Prearly, conditions in the industry exhibit several hallmarks of the scene
seericch the supply of product variety is overstimulated: high substitu,
tability, extensive ‘cannibalization, concave-downward inverse demand
functions, ample producers’ surplus margins (plus the monopoly powel '@
sarin them), and substantial product launching costs. In view of these
See Comanor and Wilson [95 pp: 49-53]; Lambin [13, pp. 94-8 127-9), Ackoff and
Emshoif [1] and Scherer et al. (27, PP. 249-50]
Copyright © 2001. All Rights Reseved.THE WELFARE ECONOMICS OF PRODUGT VARIETY 133
qualitative features and the crude quantitative findings, it appears probable
that product proliferation has, at least at the margin, cost more than ic
was worth,
NORTHWESTERN UNIVERSITY ACCEPTED NOVEMBER 1978
REFERENCES,
U1] Aexorr, Russet L. and Ensuorr, Jaucts R., ‘Advertising Research at Anheuser.
Rusch, Inc. (1965-68)', Sloan Management Review (Winter 1975); pps int 5.
[a] Abas, W. J and Yetizn, Janer L., ‘Commodity Bundling snd tre Bucien of
Monopoly’, Quarterly Jounal of Economics (August 1976), pps 475-08,
[5] “What Makes “Advertising Profitable”, Exonomis Jioat (September
1977)» PP. 427-49.
[4] Ancumato, G. C. and Rosexsuurx, Gozox, “The “New"" Theory of Consumer
Banand and Monopolistic Competition’, Quarry Joural of Ezmomics (Novena
ber 1975), pp. 569-90.
(5) Buzzett, Rovent D. and Nourse, R. E., Product Innovation, the Product Life
(Goee and Competitive Behszior in Sclecled Food Processing Indusrien 1947-64 (Arthur
D. Little, Inc., Cambridge, 1966).
(6) Guaunenuix, Eowano H., The ‘Theory of Monopolstic Competition (Harvard
University Press, Cambridge, 1993).
{7} Gouaxon, Wrttan S. and Wisox, THowas A. Advertsing and Market Pocer
(Harvard University Press, Cambridge, 1974)
(6) Doar, Avisast K. and Snicurrz, Josers E., ‘Monopolistie Competition and
HPGmum Produet Diversity", American Economic Reiew June 1977) pe, 297-308.
[9] Dovotas, Geonor W. and Musser, Janes C., Ili, Boon? Regulation of
Domestic Air Transport (Brookings, Washington, 1974).
{0} Greewser, Metts L.A Tacory ofthe Firm in Beomortic Space (Appleton-Century-
Crofis, New York, 1970).
(11) Hay, 'D. Ay ‘Sequential Entry and Entry-deterring Strategies in Spatial
Competition’, Oxford Economie Papers (July 1976); pp. 240-37,
[12] Horetiixe, HaRoto, ‘Stability in Competition, Eemaore Journal (March
1929), Pp. 41-57.
[13] Lamon, JeawJacours, Advertising, Competition and Market Conduct in Oligopoly
over Time (North-Holland, Amsterdam, 1976),
[14] Lancaster, Ketvis, *A'New Approach to Consumer Theory", Journal of
Political Economy (April 1966), pp. 132-37.
U5] ‘Socially Optimal Product Differentiation’, American Economie Review
(September 1975), pp. 567-85,
(16] -—‘Monopolistic Competition as a Perfect Market Structure’, processed
(October 1977).
[17] Fatanp, Havne E., ‘Quality Choice and Competition’, American Economie
Review (March 1977), pp. 127-37.
(:8] Lovers, Mrciiate C., ‘Product Differentiation and Market Structure’, Western
Economic Journal (June 1970), pp. 120-43.
[9] Lorsen, Avoust, The Bzonomics of Location, and edn, trans. W. Hi. Woglom
(Yale University Press, New Haven, 1956).
{20] Meaoe, J. E., “The Optimal Balance between Economies of Seale and Variety
of Products’, Economica (November 1974), pp. 359-67.
[1] Namioxat Consiox ox Foon ManszTinc, Studer of Orgonization and Come
Potton in Grocery Manafactring, Technical Suady No, 6 (Washington, June 1960),
Copyright © 2001. All Rights Reseved.134 F, M, SCHERER
[22] Nowe, Rooen G., Prox, M. J. and McGowan, J. J Eeonomie Aspects of Tele-
vision Regulation (Brookings, Washington, 1973)-
[ag] Onestets, Staxtey I, Industrial Concentration and Advertsing Intensity (American
Enterprise Institute, Washington, 1977).
[24] Rommsos, Joax, The Ezonamics of Imperfect Competition (Macmillan, London,
1934).
[25] Sennen, F. M,, ‘Research and Development Resoures Allocation under
Rivalry", Quarterly Journal of Economics (August 1967), PP- 330°94-
26) a Industrial Market Structure and Economic Performance (Rand “McNally, Chicago,
1979)-
far] Serenen, F. My Broxzssress, Anas, Kauren Enact and Mente RD,
sci Romie of Mult plant Operation: An International Comparisons Stay (Harvard
University Press, Cambridge, 1975).
{28} Scmatenses, Rictaro, ‘Entry Deterrence in the Ready-to-Eat Breakfast
Geel Tedustry", Bell Journal of Economies (Autumn 1978), PP- 905-327:
{29) Srexce, Micitatt, ‘Product Selection, Fixed Costs and Monopolistic Competi-
Hon’, Review of Economic Studies (June 1976), PP. 217-35:
90] Product Diflerentiation and Wellare’, American Economie Reving (May
1976), pp: 407-14
[gt) Srence, Micitazt and Owex, Bruce, ‘Television Programming: Monopolistic.
Competition and Welfare’, Quarterly Jounal af Economics (February 1977)
pp. 103-26.
{ga} Sremen, Peter O., ‘Program Patterns and Preferences and he Workability of
Gmpetion in Radio Broadcasting’, Quarterly Journal of Economies (May 1952),
pp. 194-223-
{ga} Brann, Ncuo.as, “The Optimal Size of Market Areas’ Journal of Economic
Theory (April 1972), PP. 154-73
{a4) Trees, Rosenr, Monpolstic Competition and General Egaibrium Theory (Harvard
University Press, Cambridge, 1960).
{g5] Unven, Dax, “The Welfare Eeonomics of Invention’, Exononica (August 1964),
pp. 279-87.
[6] Wuntie, Rosear D., ‘Consumer's Surplus without Apology’, Anericon Economic
Review (September 1976), PP- 589-97-
Copyright © 2001. All Rights Reseved.