Was Mr. Hewlett Right? Mergers, Advertising and the PC Industry Michelle Sovinsky Goeree 1 Preliminary, please do not cite. March, 2005 (First Version June 2002) Abstract In markets characterized by rapid change, such as the personal computer industry, con- sumers may not know every available product. Failing to incorporate limited information and the strategic role of informative advertising into merger analysis may yield misleading results regarding industry competitiveness. This is of particular importance when accessing the welfare impact of mergers. I use the parameters from a model of limited consumer in- formation to (1) estimate the effect on profits and consumer welfare from mergers and (2) to examine the role of advertising as it relates to market power and the implications for an- titrust policy. The methodology used to evaluate the impact of mergers follows Nevo(2000), but incorporates limited information and strategic choices of advertising. I simulate post- merger price and advertising equilibria for the Compaq-HP merger and for a hypothetical merger. I decompose the change in prices into changes due to increased concentration and changes due to the influence of advertising. The results indicate advertising can be used to increase market power when consumers have limited information, which suggests revisions to the current model used to access the impact of mergers in antitrust cases. JEL Classification: L15, D12, M37, D83 Keywords: merger analysis, informative advertising, discrete-choice models, product differ- entiation, structural estimation 1 This paper is based on various chapters from my 2002 dissertation. Special thanks to my dissertation advisors, Steven Stern and Simon Anderson, for their guidance. I am grateful to Gartner Inc. for making the data available, and specifically to Sandra Lahtinen. I also gratefully acknowledge financial support from the University of Virginia’s Bankard Fund for Political Econ- omy. Address for correspondence: Claremont McKenna College, 500 E. Ninth Street, Claremont, CA 91711 (email: [email protected])
27
Embed
Was Mr. Hewlett Right? Mergers, Advertising and the PC ... I/I.C... · I examine both the HP-Compaq merger and a hypothetical merger between IBM and Dell.5 Considering a merger between
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Was Mr. Hewlett Right?Mergers, Advertising and the PC Industry
Michelle Sovinsky Goeree 1
Preliminary, please do not cite.
March, 2005
(First Version June 2002)
Abstract
In markets characterized by rapid change, such as the personal computer industry, con-sumers may not know every available product. Failing to incorporate limited informationand the strategic role of informative advertising into merger analysis may yield misleadingresults regarding industry competitiveness. This is of particular importance when accessingthe welfare impact of mergers. I use the parameters from a model of limited consumer in-formation to (1) estimate the effect on profits and consumer welfare from mergers and (2)to examine the role of advertising as it relates to market power and the implications for an-titrust policy. The methodology used to evaluate the impact of mergers follows Nevo(2000),but incorporates limited information and strategic choices of advertising. I simulate post-merger price and advertising equilibria for the Compaq-HP merger and for a hypotheticalmerger. I decompose the change in prices into changes due to increased concentration andchanges due to the influence of advertising. The results indicate advertising can be used toincrease market power when consumers have limited information, which suggests revisionsto the current model used to access the impact of mergers in antitrust cases.JEL Classification: L15, D12, M37, D83Keywords: merger analysis, informative advertising, discrete-choice models, product differ-entiation, structural estimation
1 This paper is based on various chapters from my 2002 dissertation. Special thanks to mydissertation advisors, Steven Stern and Simon Anderson, for their guidance. I am grateful toGartner Inc. for making the data available, and specifically to Sandra Lahtinen. I also gratefullyacknowledge financial support from the University of Virginia’s Bankard Fund for Political Econ-omy. Address for correspondence: Claremont McKenna College, 500 E. Ninth Street, Claremont,CA 91711 (email: [email protected])
1 Introduction
OnMay 7, 2002 Hewlett-Packard (HP) Company launched the new Hewlett-Packard with an
ad titled “We are Ready.” The new Hewlett-Packard is a result of a merger with Compaq
Computer Corporation, the largest ever in the information technology sector. The $19
billion deal has drawn a lot of media attention for a number of reasons. Investors and rival
firms are interested in its impact on shares and profits. Consumers are interested in the
effect on prices. Regulators are interested in its implications for competition in an already
concentrated industry.
Originally proposed in June 2001, the merger prompted a bitter battle between Hewlett
and Packard family interests and corporate executives. It was ultimately approved by a slim
majority of shareholders (only 3%). Many HP shareholders opposed the deal because they
thought the time lost in absorbing Compaq and incorporating cost synergies would distract
from winning new orders at a time when the market was slowing. Walter Hewlett, whose
father cofounded HP, launched a court battle against HP arguing the merger would result
in lost profits in the long run. As further evidence of his conjecture, Hewlett pointed to his
competitors, “We believe that HP stockholders should be concerned when competitors, like
SUN, Dell, and IBM don’t object to a transaction that is supposed to add value to HP.”
Meanwhile, the Federal Trade Commission (FTC) voted unanimously to approve the merger.
Likewise, the European Commission approved it without placing any conditions on the two
companies, saying “A careful analysis of the merger. . . has shown that HP would not be in
a position to increase prices and that consumers would continue to benefit from sufficient
choice and innovation.”2
The analysis used by antitrust authorities to evaluate the impact of mergers is based on
a model which assumes consumers are aware of all products for sale when they make their
purchase decision. However, it is reasonable to think consumers may have limited informa-
2 “HP-Compaq Merger Wins European Approval,” NewsFactor Network, Feb.1, 2002.
1
tion regarding the products available, especially in markets characterized by a high degree
of change such as the PC industry. Price elasticities calculated under the assumption of full
information may yield misleading results regarding the degree of competition in the market.
Incorrect conclusions regarding the competitive effects of a merger could lead antitrust au-
thorities to approve a merger which has negative consequences for consumers. Goeree (2002)
showed that assuming consumers are aware of all products in a market characterized by rapid
change generates estimates of product-specific demand curves that are biased towards being
too elastic. This is of particular importance when considering the welfare impact of mergers,
one of the primary stated interests of the FTC when examining mergers.
The goal of this empirical work is twofold: (1) to estimate the effect on profits and
consumer welfare from mergers when consumers have limited information and (2) to examine
the role of advertising as it relates to market power and the implications for antitrust policy.
The methodology used to evaluate the impact of mergers is based on previous work,3 but
allows for limited information and strategic choices of advertising. I use the estimated
parameters from a structural model of informative advertising4 to simulate post-merger
equilibrium price and advertising levels. I calculate the effect of mergers on the profits of
merging and non-merging firms as well as the cost-synergies necessary to offset losses. I
decompose the change in prices and markups into the changes due to increased concentration
and the changes due to the influence of advertising. Perhaps surprisingly, the results suggest
the latter effect is the strongest. The post-merger equilibrium results indicate advertising
can be used to increase market power, which suggests revisions to the current model used
by the FTC to determine market power in antitrust cases.
I examine both the HP-Compaq merger and a hypothetical merger between IBM and
Dell.5 Considering a merger between these two firms is of interest for two reasons. First,
3 For example, see Baker and Bresnahan, 1985; Berry and Pakes, 1993, Hausman, Leonard and Zona,1994; and Nevo, 2000.
4 See Goeree (2002) for more detail.
5 While the merger is supposed, IBM and Dell entered into a $16 billion cross-licensing agreement in
2
IBM is the world leader in sales of non-PC’s, while Dell is the world leader in sales of PC’s.
They focus on different aspects of the computer industry. As a result, cost synergies are
expected to be lower than those between HP and Compaq. Secondly, and more directly
related to the topic of this research, IBM and Dell have very different ad-to-sales ratios.
IBM is a high-advertising intensity firm, while Dell is a low-intensity firm. In contrast,
HP and Compaq are thought to have more cost synergies and are closer in their ad-sales
concentration. Comparison of the outcomes from the two very different mergers yields
insight into the role that advertising plays in this industry and how firms use it when industry
concentration increases.
The remainder of the paper is organized as follows. In the next section, I present an
overview of the PC industry and discuss the data used in estimation. In section 3, I describe
the model used in the counterfactual merger simulations. In section 4, I discuss the impact
of mergers on welfare and profits, the role that advertising plays as concentration increases,
and the implications for antitrust policy.
2 Trends in the PC industry and Data
The PC industry, as we know it today, has been growing since 1971 when Intel introduced the
first microcomputer. It utilized the 4004 microprocessor, an integrated circuit able to process
four bits of data at a time (hence the ‘4’s). Intel was the first to imbed all components of a
computer — central processing unit (CPU), memory, and input-output controls — on a single
chip. The advantage was that, simply by changing the external program, the same device
could be used for a multitude of projects. The first generally available microcomputer, the
Altair 8800, was on the market only four years later. It retailed for $439, which made it
affordable, but it was not easy to use. The kit required assembly and software was not
available. Two young hackers tackled the second problem and began the process of writing
software. The language was the Beginners All-purpose Symbolic Instruction Code, BASIC,
1999. The agreement will last until 2006 and calls for broad patent cross-licensing between the two firmsand collaboration in the development of product technology.
3
and the hackers were William Gates and Paul Allen.
Three years later Steve Jobs and Steve Wozniak introduced the Apple II. The first PC
that was both affordable and usable. It had 4K RAM, built-in BASIC, color graphics, and
sold for about $1300. Apple controlled the market until IBM introduced the (modestly
named) “PC.” The PC featured a modular design so that pieces could be added easily.
It soon surpassed the Apple II in popularity and allowed IBM to dominate the market for
several years. In 1984, Apple introduced the first generation Macintosh during a SuperBowl
commercial. The new Macintosh came with a graphical user interface and a mouse, which
made it easy to use. As a result, sales of the Macintosh boomed.
Competition and technological improvement has continued to spur innovation, and the
PC industry has seen numerous product introductions in the past years. Due to the fre-
quency with which new products are brought into the market, consumers may not be aware
of all products offered. Indeed every year over 200 new PCs are available from the top 15
firms alone. Due to the large number of brand introductions and the competitive nature of
the industry, it is has become increasingly important for firms to advertise to inform con-
sumers about their products. As Figures 1 and 2 illustrate, prices dropped from an average
of close to $2700 in early 1997 to under $1600 in 1999, while advertising expenditures grew
by over $0.5 billion.
Advertising has been an important dimension of competition in this industry since its
beginnings. Between 1995 and 1999, advertising expenditures grew by nearly 100% to $2.3
billion. In 1998 over 36 million PCs were sold, generating over $62 billion in sales — $2 billion
of which was spent on advertising. There is much variation in advertising expenditures across
firms and across media. For instance, in 1998, fifty percent of the industry expenditures are
by IBM, resulting in an ad-to-sales ratio of over 19 percent. While the top two firms (in
terms of market share), Dell and Compaq, have much lower ad-to-sales ratios of under 3%.
The data are from 1996-1998, a period of tremendous growth in the PC industry. During
4
this time there were a number of mergers. In 1996 Packard Bell6 was a 4.5 billion com-
pany and its 15% market share made it the largest PC manufacturer in the US. Compaq
passed Packard Bell in mid 1996 and price pressure from Compaq and eMachines along with
poor showings in consumer satisfaction surveys made it difficult for the company to remain
profitable.7 In 1997, 3 mergers occurred: Packard Bell, NEC, and ZDS; Acer and Texas
Instruments; and Gateway with Advanced Logic Research.
After this period, there was a slowdown in the PC market. Demand in the home market
sector (as well as other sectors) declined. In part because there was not as much of a need
to upgrade as often since the PCs were so well made and in part due to the slump in the
economy. This data is from 1996-1998, a period of tremendous growth in the PC industry.
Immediately after this period, Compaq merged with DEC, which proved to be a mistake
for Compaq in that they never fully recovered their pre-merger market position. It is this
Compaq with which HP merged in 2002. In this sense the results of this study are not as
insightful as they could be given more recent data. However, these results indicate what
would have happened as a result of the merger during a period of incredible growth. Even
during this period of tremendous growth, I find that the merger would increase profits for
the merged companies, but not by a lot more than it would for the other companies. It
is perhaps not surprising then, that the HP-Compaq merger as it currently stands is not
working so well. I also simulated the effect of a 3 way merger between Compaq, DEC,
and HP. Finally, the in-sample mergers are used to test the predictions of the model, that
is I can examine how well the model predicts new prices and advertising outcomes for the
mergers that occurred over the period of the data.
The data come from three primary sources. The product-level data are from Gartner
Inc. and consist of prices, market shares, and product characteristics from the first quarter
of 1996 to fourth quarter of 1998. The Gartner data detail sales across sectors: home,
6 Packard Bell was an American radio manufacturer (and has no association with Hewelett Packard).
business, educational institutions, and government. I use the home market data, which
accounts for over 30% of all PCs sold, to estimate the model. The product level data
are combined with advertising data obtained from Competitive Media Reporting’s LNA/
Multi-Media. Advertising data are quarterly expenditures by firm across media8 with some
brand level information.9 The final dataset is from Simmons Market Research and includes
consumer-level purchases across manufacturers, information on media habits, and consumer
characteristics for about 20,000 households annually. I restrict my attention to the top 10
firms (based on market share) and to 5 others to make full use of the micro-level data.10
These 15 “included firms” account for over 85% of PC sales to home users. Details of the
data and its construction are given in Appendix 1 and in Goeree (2002).
3 Model
The econometric model follows those found in recent studies of differentiated products, such
as Berry, Levinsohn, and Pakes (1995, 2004) and Nevo (2000) and is presented in more detail
in Appendix B and Goeree (2002).
3.1 Consumer Behavior
Individual i = 1, ..., N chooses from j = 1, ..., J products at time t = 1, ..., T . Gartner
collects information on five main PC attributes: manufacturer (e.g. Dell), brand (e.g. Lat-
itude LX), form factor (e.g. desktop), CPU type (e.g. Pentium II), and CPU speed. A
product pertains to a specific PC model defined as a manufacturer-brand- CPU type-CPU
8 The expenditures are reported for 10 media, from which I construct 4 main categories: newspaper,magazine, television, and radio.
9 In the PC industry, it is common for manufacturers to advertise groups of products simultaneously. Forexample, in 1996 one of Compaq’s advertising campaigns involved all Presario brand computers (of whichthere are 12). See Appendix 1 for more detail regarding the treatment of group advertising.
10 The included manufacturers are Acer, Apple, AST, AT&T, Compaq, Dell, Epson, Gateway, HP, IBM,Micron, NEC, Packard Bell, and Texas Instruments.
6
speed-form factor combination.11 The indirect utility consumer i obtains from product j
at time t is given by
uijt = α ln(yit − pjt) + x0jβit + ξjt + ijt (1)
The characteristics of product j are represented by (pjt, xj, ξjt), these are price, non-price
observed characteristics (such as CPU speed, laptop and Pentium dummies, firm fixed ef-
fects, and a constant), and unobserved (to the econometrician) characteristics, respectively.
Income is represented by yit, ijt is a mean zero stochastic term which is assumed to be in-
dependent and identically distributed across products and consumers, and βit are individual
specific components.
The individual specific components are random coefficients where
βit = β +ΠDit + Σνi, νi ∼ N(0, Ik)
where mean preferences for observable product attributes are captured by β, the matrix of
coefficients, Π, measures how tastes vary with these attributes and Σ is a scaling matrix.
Characteristics not observed by the econometrician that may influence tastes are captured
by the νi.
Consumers may decide not to purchase any of the products. The indirect utility provided
from purchasing the “outside” option is ui0t = α ln(yit) + ξ0t + i0t, where the price of the
outside good is normalized to zero.
I assume the consumer purchases at most one good per period, that which provides the
highest utility from all the goods in her choice set.12 The set of variables that results in the
purchase of good j : Rjt ≡ (yit,Dit, νi, ijt) : Uijt ≥ Uirt ∀r 6= j. If a consumer has full-
information regarding the products for sale and, assuming ties occur with zero probability,
11 The data allow for a very narrow model definition. For example, Compaq Armada 3xxx Pentium150/166 laptop and Compaq Armada 4xxx Pentium 150/166 laptop are two separate models, as are anApple Power Macintosh Power PC 604 180/200 desktop and deskside.
12 This assumption may be questionable in markets where multiple purchase is common. However, it isnot unreasonable to restrict a consumer to purchase one computer per quarter. Hendel (1999) examinespurchases of PCs by businesses and presents a multiple-choice model of PC purchases.
7
the market share of product j is
sjt =
ZRjt
dF (y,D, ν, ) =
ZRjt
dFy,D(y,D)dFν(ν)dF ( ) (2)
where F (·) denotes the respective known distribution functions. To derive the market share
of product j, I integrate over the observed joint distribution of (yit,Dit) and the assumed
distribution of (νi, ijt) in the population, where the second equation follows from indepen-
dence assumptions. I assume the are distributed i.i.d. type I extreme value in order to
obtain simple expressions for choice probabilities.
In industries where introductions of new products are frequent (like the PC industry),
the assumption that consumers are aware of all products for sale is not an innocuous one.
As in Goeree (2002), I develop a model of random choice sets: the probability consumer
i purchases product j depends upon the probability she is aware of product j, the other
products competing with j of which she is aware, and the probability she would buy product
j given her choice set.
Let Cj be the set of all possible choice sets that include product j. Assuming consumers
are aware of the outside option with probability one, the (conditional) probability that
consumer i purchases product j is given by
sijt =XS∈Cj
Yl∈S
φiltYk/∈S(1− φikt)
(yit − pjt)α expx0jβit + ξjt
yitα +P
r∈S(yit − prt)α expx0rβit + ξrt(3)
where φijt is the probability consumer i is informed about product j, the outside sum is over
the different choice sets that include product j, and the yitα in the denominator is from the
presence of an outside good.
Advertising affects demand through the information technology function, φijt which de-
scribes the effectiveness of product j advertising at informing consumer i. The information
technology is a function of medium-specific advertising, observed consumer attributes, and
(unobserved) consumer-advertising- medium specific effects.13 The information technology
13 For more details regarding the specification of the information technology see Appendix 2 and Go-eree(2002).
8
approaches one as advertising increases, but may be positive even when no advertising oc-
curs. The latter flexibility allows for the possibility that a consumer may be informed, even
if she hasn’t seen an advertisement. For instance, she may have received information by
word-of-mouth, through experience with the product, or exposure to other non-advertising
media coverage.14
3.2 Firm Behavior
I assume there are F firms in an oligopolistically competitive industry and that they are non-
cooperative, Bertrand-Nash competitors. Each firm produces a subset of the J products,
Jf . Suppressing time notation, the profits of firm f are
Xj∈Jf
(pj −mcj)Msj(p, a) +Xj∈Jf
Πnhj (p
nh, a)−Xm
mcadjm(Xj∈Jf
ajm)−Cf (4)
where sj is the vector of home market shares, which is a function prices and advertising for all
products; mcj is the marginal cost of production; Πnhj is the gross profit (before advertising)
from sales to the non-home sectors; mcadjm is the marginal cost of advertising in medium m;
ajm is the number of medium m advertisements; and Cf are fixed costs of production. The
potential market size, M , is given by the number of US households in a given period, as
reported by the Census Bureau.
Given their products and the advertising, prices, and attributes of competing products,
firms choose prices and advertising media levels simultaneously to maximize profits. PC
firms may sell to non-home sectors (such as the business, education, and government sectors).
I assume marginal costs are constant, which implies pricing decisions are independent across
sectors.15 Any product sold in the home market sector will have prices that satisfy the
14 See Anand and Shachar (2001).
15 There are reasons to believe that pricing decisions may not be independent across sectors. See Go-eree(2002) for more discussion.
9
following first order conditions
sj(p, a) +Xr∈Jf
(pr −mcr)∂sr(p, a)
∂pj= 0 (5)
In vector form, the J first-order conditions are
s−∆(p−mc) = 0
where ∆j,r = −∂sr∂pj
Ij,r with Ij,r an indicator function equal to one when j and r are produced
by the same firm and zero otherwise. These FOC’s imply marginal costs given by
mc = p−∆−1s (6)
An advertisement intended to reach a home consumer may affect sales in other sectors.
Optimal advertising choices must equate the marginal revenue of an additional advertisement
in all sectors with the marginal cost. Optimal advertising medium choices ajm must therefore
satisfy
MXr∈Jf
(pr −mcr)∂sr(p, a)
∂ajm+mrnhj (p
nh) = mcadjm (7)
where mrnh is the marginal revenue of advertising in non-home market sectors.
4 Post Merger Equilibrium
This section presents the methodology used and the results from two mergers in the PC
industry. I consider a merger between HP and Compaq and of a hypothetical merger
between IBM and Dell. All post-merger computations use the estimates from the model of
demand and supply presented in the above section. These parameter estimates are given in
Tables 3-5 and discussed in more detail in Goeree (2002).
To simulate the effects of a merger, I follow the same strategy as Nevo (2000). First,
I estimate the parameters of the (pre-merger) model and use the implied price elasticities
to recover the marginal costs of production. I use these marginal costs and the post-merger
10
industry structure to simulate a new price and advertising equilibrium. I simulate post-
merger equilibria under two assumptions of post-merger behavior on part of the firms. First,
I compute the new price equilibrium under the assumption that advertising choices remain at
pre-merger levels. That is, I do not allow firms to reoptimize over advertising choices. This
is used as the benchmark case. Notice that this benchmark case will provide an accurate
picture of the post-merger industry only if firms do not change their advertising strategy or
if advertising does not impact demand. I compare the benchmark case to the new price and
advertising equilibrium that arises post-merger allowing firms to reoptimize over both prices
and advertising.
For the first two parts of the analysis I use estimated costs and, hence, assume the cost
structure is the same both before and after the merger. Firms may experience decreased
costs as the result of a merger. Therefore, in the final part of the analysis (following Nevo,
2000), I determine the magnitude of cost savings that would be necessary to return to the
pre-merger price and advertising equilibrium.
I assume post-merger market conduct is the same as pre-merger, and use the estimated
pre-merger parameters to simulate the (counterfactual) post-merger equilibrium. The pre-
dicted post-merger equilibrium price ppost solves
ppost = cmc+∆post(ppost)−1s(ppost) (8)
where the matrix ∆post is constructed to reflect post-merger ownership structure, cmc are the
predicted marginal costs and ppost is the vector of post-merger predicted equilibrium prices.
The predicted post-merger advertising levels solve (7) under the new ownership structure,
holding marginal costs constant at their estimated levels. The post-merger equilibrium
levels of prices and advertising are simulated jointly using the data from the last quarter of
1998.
11
4.1 Advertising and Market Power
Hewlett-Packard Co.’s PC division more than doubled in size with the acquisition of Compaq
Computer Corp., instantly making the company the top PC supplier.
Table 6 presents firm and industry level changes in prices, advertising, markups, and
variable profits after the mergers. The first column presents average percentage changes in
prices under the assumption that post-merger advertising levels are unchanged. Predicted
price changes are higher for the merging firms, especially for the Compaq-HP merger. In
addition all firms increase their prices under the HP-Compaq merger. We might expect
prices to increase for all firms in the industry, as the industry is more concentrated allowing
firms to exercise more market power. However, counter to intuition, under the Dell-IBM
merger some firms charge lower average prices and overall industry prices experience only a
small increase in price (0.2%). This unexpected result may be due to the assumption that
firms leave advertising expenditures unchanged.
The second and third columns present results based on the assumption that firms choose
new prices and advertising levels after the merger. As the second column shows, the pricing
outcome is more intuitive. All firms increase their prices as industry concentration grows.
The overall price increase in the industry is 9% and 6% under the HP-Compaq and Dell-IBM
mergers, respectively. Surprisingly, Dell-IBM do not raise prices as much as Compaq under
the Dell-IBM merger. The first column of Table 7 provides a breakdown of price changes
for selected products. As the table indicates, price increases are greatest for the HP and
IBM products under their respective mergers. One possible reason for this distribution is
that both HP and IBM have lower pre-merger prices than their merging counterparts. As
the third column of Table 6 indicates, prices are 2% and 6% higher under the HP-Compaq
and Dell-IBM mergers respectively relative to when firms don’t reoptimize over advertising.
The fourth column presents percentage changes in advertising. All firms choose to
advertise more in both post-merger environments. Advertising increases by 3.3% under the
HP-Compaq merger and 1.8% under the Dell-IBM merger. Prior to merging HP-Compaq
12
have a combined ad-to-sales ratio of 7%, while Dell-IBM’s is on the order of 14%. HP-
Compaq increases their advertising by 10% (or $75 million) after the merger, yielding a
post-merger ad-to-sales ratio of 20%.16 Dell-IBM increases their advertising by 0.9% (or
$12 million) post-merger, yielding a post-merger ad-to-sales ratio of 22%. It is interesting
that post-merger ad-to-sales ratios are similar for the merging firms.
Column two of Table 7 provides an idea of how these advertising changes are distributed
across the products offered by the merging firms. The table shows that advertising increases
are larger for HP under the first merger — the more advertising intensive pre-merger firm.
While under the Dell-IBM merger, advertising increases are greatest for the Dell products —
the less advertising intensive pre-merger firm. One thing is clear from these results. Had
we considered only one merger we would have come away with a partial picture of the role
of advertising in this industry.
The fifth column of Table 6 presents a breakdown of post-merger markup percentages,
which include advertising costs. Pre-merger summary statistics are presented in Table 8.
The industry average markup is higher (24%) relative to pre-merger markups (10%) and
post-merger price only markups (20%). The percentage markups for Compaq-HP is about
50% higher than pre-merger levels, while for Dell-IBM it is roughly equivalent.
These results shed light on the implications for informative advertising in the PC in-
dustry. The results suggest that firms will charge higher prices in conjunction with more
advertising. Markups are about 10% higher on average when firms are permitted to reopti-
mize over advertising choices relative to optimizing over price only. Part of the post-merger
higher markup is due to industry concentration, and part is due to the effect of advertising
in this industry. If the firms were prohibited from changing their advertising, the industry
would look much more competitive. The results indicate that advertising is used to reduce
competition in the market.
The above analysis assumes costs do not change after the mergers. We may expect
16 This is calculated using the post-merger shares, which are lower than pre-merger combined shares.
13
there to be some synergies between companies resulting in decreased post-merger costs. I
calculated the percentage change in costs that would induce post-merger prices to equal pre-
merger prices. I do not make any assumptions regarding which products of the acquiring
firms would enjoy cost reductions. The median percentage changes are given in Table 11.
The results indicate that a cost savings of 4% for HP-Compaq is enough to reach pre-merger
equilibrium prices. The cost savings would be mostly on HP products. Even smaller costs
savings, 2.7%, are enough to keep prices constant at pre-merger levels for the merger of
Dell and IBM. Under the Dell-IBM merger, cost savings are spread more evenly over the
products of both firms, relative to the HP-Compaq outcome, but are still more intense for
Dell products. In addition, the costs savings are not of an unattainable magnitude. HP
and Compaq or Dell and IBM may experience a reduction in costs of close to 4% and 3%
respectively after merging. (more analysis of the results forthcoming...)
4.2 Consumer Surplus and Profits
Remarking on the HP-Compaq merger, a research fellow at Gartner noted, “There’s nothing
good for the consumers. They’ve eliminated one of two fiercely competitive brands in the
market. And what that means generally is higher prices and less choice.” The previous
subsection uses the structural estimates to produce a counterfactual of what the market
equilibrium would look like under the merger. In this section, I examine the impact of the
mergers on consumer surplus and profits to provide further insight into the social welfare
effects of these mergers, the primary stated interest of the FTC.
Consumers face different prices and (possibly) different choice sets as a result of a merger.
Evaluating the change in utility generated by differences in choice sets is possible since
preferences are defined over characteristics. I use compensating variation to measure changes
in consumer welfare. Compensating variation is the amount of money an individual would
need to be compensated at the new equilibrium prices and choice set to be as well off as
they were under the pre-merger equilibrium. If household welfare is improved as a result of
14
the merger, the expected compensating variation measure is negative.
Compensating variation for individual i, cvi, is implicitly defined by
Maxj∈Jprei
U(yi − pprej ) = Maxj∈Jposti
U(yi − ppostj − cvi, ) (9)
where the superscripts ‘pre’ and ‘post’ are used to distinguish the original versus the new
conditions associated with the merger. Recall, consumers may not know all products avail-
able to them, Ji denotes the choice set facing consumer i. 17 The pre- and post- merger
choice set may differ because the equilibrium level of advertising may change. Under assump-
tions of full information, the effects on consumer welfare (as measured by the area under the
Hicksian or Marshallian demand curves) will be understated.
The resulting compensating variation is a random variable, which will depend upon
in a nonlinear fashion. I am interested in the expected value of this random variable,
E(cv). Since marginal utility of income varies with income and prices there is no closed
form solution for E(cv). I simulate E(cv) using a procedure developed by McFadden (1999).
The simulation requires sampling from the underlying distribution of errors and employing a
numerical algorithm to solve for the implicitly defined cv as given in equation (9). I compute
cvi for R draws from the underlying distribution of errors. Mean compensating variation
for an individual is the average over these draws.18 I average across the sample to obtain
the mean compensating variation for the population. The change in welfare is computed by
comparing consumer and producer welfare across the two economic environments. For each
17 The choice set for an individual is determined by comparing the value of the advertising technology,φij , to a draw from a uniform distribution. The advertising technology is evaluated at the estimated valueof the parameters and equilibrium advertising levels. If the uniform draw is larger, the product is not inthe choice set. If the uniform draw is smaller the product is in the choice set.Specifically, the pre-merger choice set is constructed as follows: given φij(
bθ, apre) and a draw from auniform, uij , construct a J dimensional Bernoulli vector, bi. This defines the choice set, where the jthelement is determined according to
bij =
½1 if φij > uij0 if φij < uij
analogously for the post-merger choice set.
18 McFadden (1999) demonstrates in a monte-carlo experiment that the number of iterations required toobtain a 5% root mean squared error when are distributed extreme value 755. For this reason I chooseR = 755.
15
merger, I calculate two compensating variations: one in which advertising is not a strategic
variable and one in which it is. I compare consumer welfare to pre-merger welfare under
both environments. (Consumer surplus and profit results are forthcoming...)
References
Ackerberg, Daniel (2003) “Advertising, Learning, and Consumer Choice in ExperienceGoods Markets: A Structural Empirical Examination,” International Economic Review44:1007-1040.
Anand, Bharat and Ron Shachar (2001) “Advertising, the Matchmaker,” HarvardBusiness School Working Paper No. 02-057.
Anderson, Simon, A. de Palma, and J.F. Thisse (1989) “Demand for DifferentiatedProducts, Discrete Choice Models, and the Characteristics Approach,” Review of Eco-nomic Studies 56: 21-35.
Anderson, Simon, A. de Palma, and J.F. Thisse (1992) Discrete Choice Theory ofProduct Differentiation, Cambridge: MIT Press.
Baker, J. and T. Bresnahan (1985) “The Gains from Merger or Collusion in ProductDifferentiated Industries,” The Journal of Industrial Economics, 33(4): 427-444.
Berry, Steven (1994) “Estimating Discrete Choice Models of Product Differentiation,”Rand Journal of Economics 25(2): 242-262.
Berry, Steven, James Levinsohn and Ariel Pakes (2004) “Differentiated Products De-mand Systems from a Combination of Micro and Macro Data: The New Car Market,”Journal of Political Economy 112(1,1): 68-105.
Berry, Steven and Ariel Pakes (1993) “Some Applications and Limitations of RecentAdvances in Empirical Industrial Organization: Merger Analysis,” American EconomicReview 83(2): 247-252.
Bresnahan, T. and S. Greenstein, (2000) ”Technological Competition and the Structureof the Computer Industry,” Journal of Industrial Economics
Bresnahan, T., S. Stern, and M. Trajtenberg (1997) ”Market Segmentation and theSources of Rents from Innovation: Personal Computers in the Late 1980s,” Rand Jour-nal of Economics 28(0): S17-S44.
Chamberlain, G. (1987) “Asymptotic Efficiency in Estimation with Conditional Mo-ment Restrictions,” Journal of Econometrics 34: 305-344.
Erdem, Tulin and Michael Keane (1996) “Decision-making Under Uncertainty: Cap-turing Dynamic Brand Choice Processes in Turbulent Consumer Goods Markets,”Marketing Science 15(1): 1-20.
16
Genakos, Christos (2004) “Differential Merger Effects: The Case of the Personal Com-puter Industry,” mimeo, London Business School.
Goeree, Michelle S. (2002) “Advertising and the US Personal Computer Industry,”Claremont Colleges Working Paper 2004-09.
Gourieroux, Christian, et al. (1987) “Generalized Residuals,” Journal of Econometrics,34: 5-32.
Grossman, G. and Carl Shapiro (1984) “Informative Advertising with DifferentiatedProducts,” Review of Economic Studies 51: 63-82.
Hausman, J., G. Leonard, and J. Zona (1994) “Competitive Analysis with Differenti-ated Products,” Annales D’Economie et de Statistique 34: 159-180.
Hendel, I. (1999) “Estimating Multiple-Discrete Choice Models: An Application toComputerization Returns,” Review of Economic Studies 66(2): 423-46
Imbens and Lancaseter (1994)
McFadden, Daniel (1999) “Computing Willingness-to-Pay in Random Utility Models,”in J. Moore, R. Riezman, and J. Melvin (eds.), Trade, Theory and Econometrics:Essays in Honor of John S. Chipman, Routledge.
Nevo, Aviv (2000) “‘Mergers with Differentiated Products: The Case of the Ready-to-Eat Cereal Industry,” Rand Journal of Economics 31(3): 395-421.
Nevo, Aviv (2002) “New Products, Quality Changes and Welfare Measures Computedfrom Estimated Demand Systems,” forthcoming, Review of Economic Studies.
Pakes, Ariel and David Pollard (1989) “Simulation and the Asymptotics of Optimiza-tion Estimators,” Econometrica 57(5): 1027-1057.
Petrin, Amil (2002) “Quantifying the Benefits of New Products: The Case of theMinivan,” Journal of Political Economy 110(4):705-729.
Shapiro, C. (1996), “Mergers with Differentiated Products,” Antitrust, 10(2), 23-30.
Small, Kenneth, and Harvey Rosen (1981) “Applied Welfare Economics with DiscreteChoice Models,” Econometrica 49: 105-130.
Werden, G.J. (1997) “Simulating the Effects of Differentiated Products Mergers: APractictioners’ Guide,” Proceedings of the NE-165 Conference.
Werden, G.J. and L.M. Froeb (1994) “The Effects of Mergers in Differentiated Prod-ucts Industries: Logit Demand and Merger Policy,” Journal of Law, Economics, andOrganization, 194, 407-26.
Willig, R.D. (1991), “ Merger Analysis, Industrial Organization Theory, and theMergerGuidelines,” Brookings Papers on Economic Activity, Microeconomics, 281-332.
17
Appendices
A DataGartner collects information on five main PC attributes: manufacturer (e.g. Dell), brand(e.g. Latitude LX), form factor (e.g. desktop), CPU type (e.g. Pentium II), and CPU speed.I define a model as a manufacturer, brand, CPU type, CPU speed, form factor combination.Even though I do not have data on some product attributes, the richness of the Gartner datastill allows for a very narrow model definition. For example, Compaq Armada 3xxx Pentium150/166 laptop and Compaq Armada 4xxx Pentium 150/166 laptop are two separate models,as are an Apple Power Macintosh Power PC 604 180/200 desktop and deskside. Treating amodel/quarter as an observation, the total sample size is 2112.19
In the PC industry, it is common for manufacturers to advertise groups of productssimultaneously. For example, in 1996 one of Compaq’s advertising campaigns involved allPresario brand computers (of which there are 12). I construct “effective” product advertisingexpenditures by adding observed product-specific expenditures to a weighted average of allgroup expenditures for that product, where the weights are estimated. Let Gj be the setof all possible product groups that include model j. Let adH be total effective advertisingexpenditures for H ∈ Gj. Define adH ≡ adH
|H| . Then effective advertising expenditures forproduct j are given by
adj =XH∈Gj
γadH + πad2
H
where the sum is over the different groups that include product j. The parameters γ and πare estimated if the product is advertised in a group, otherwise γ is restricted to one and πto zero.
Finally, I use data from the Consumer Population Survey in combination with theaggregate product level data to define the distribution of consumer characteristics. Moredetail about the various datasets, their construction, and descriptive statistics can be foundin Goeree (2002).
B Model and Estimation DetailsThe market share for product j is obtained by integrating over the relevant distributions inthe population
sjt(p, a) =
ZsijtdF (y,D)dF (ν)dF (κ) (10)
and is a function of prices and advertising of all products, where sijt is given in equation (3).The advertising technology for product j for consumer i is given by
φijt(θφ,Υ) =exp(τ ijt)
1+exp(τ ijt)
τ ijt = eD0itλ+ a0jmt(ϕm + ρmajt +Ψf +ΥmD
sit + κi) lnκi ∼ N(0, Im)
19 This is the sample size after eliminating observations with negligible quarterly market shares.
where θφ = λ, γ, π, ϕ, ρ,Ψ and Υ are parameters to be estimated. The m-dimensionalvector a is the number of advertisements in each medium, eD is a subset of consumer charac-teristics, and κ are unobserved consumer heterogeneity with regard to advertising mediumeffectiveness.
Under the assumption that the data are the equilibrium outcomes, the parametersθ, β, η, ηAD are estimated by generalized method of simulated moments. The demand-side parameters are β and θ = α, σ,Π, θφ and the supply-side parameters are η and ηAD,where ηAD is the vector of parameters associated with advertising medium choices.20
The first two sets of moment conditions are analogous to those in BLP. I restrict themodel predictions for product j’s market share to match the observed market shares. Usinga contraction mapping suggested by Berry (1994) I compute the vector δ (θ), which is theimplicit solution to Sobs
t − st(δ, θ) = 0, and use it to solve for the demand side unobservable,ξjt.
I assume marginal costs of production, mcj, are composed of unobserved (ωjt) andobserved characteristics (wjt), where ln(mcj) = w0jη + ωj. I use the pricing FOCs to solvefor the cost side unobservable, ωjt.
Marginal costs of advertising in medium m are a function of the average price of anadvertisement and unobserved product-specific components,
ln(mcadjm) = ψ ln(padm) + υj υj ∼ N(0, σ2υIm)
Some firms choose not to advertise some products in some media. To allow for cornersolutions, I use the marginal costs associated with advertising and the interior FOCs toconstruct a tobit maximum likelihood function,
$(Ω) =Y
j:ajm>0
1
συφNµlnhj − ψ ln padmt
συ
¶ Yj:ajm≤0
1− Φ
µlnhj − ψ ln padmt
συ
¶where φN is the standard normal pdf, Φ is the cdf, and hj is the marginal revenue ofadvertising (the left-hand side of equation 7).
For ease of exposition rewrite the advertising medium FOC as
lnhjm(ajm)− ψ ln padmt = υjm
where the latent variable, a∗jm is its implicit solution. We observe a∗jm if there is an interior
solution and zero otherwise. Notice the errors, υjm, are linked with the latent variable. Sincethey depend on unobserved variables, they cannot be used to construct moment conditions.
Gourieroux et al.(1987) suggest an alternative method: replace the errors by theirbest prediction conditional on the observable variables and use these to construct momentconditions. They show the moment conditions then express an orthogonality between thegeneralized residuals, eυjm(bΩ) = E[υjm(bΩ) | ajm], and the instruments, where bΩ is the maxi-mum likelihood estimator of Ω. Let T be the vector formed by stacking the resulting samplemoments over media and over products.
The micro-moment conditions use the Simmons data. Let Bi be an F × 1 vector offirm choices for individual i. Let bi be a realization of Bi where bif = 1 if a brand producedby firm f was chosen. Define the residual as the difference between the vector of observedchoices in the data and the model prediction given (δ, θ). The residual for individual i,denoted Gi(δ, θ), can be written as
bi − EνE[Bi | Di, ν, δ, θ] = Gi(δ, θ)
20 The additional parameters of the model, Υ, are estimated separately using maximum likelihood, seeGoeree (2002).
19
The population restriction for the micro moment is E[Gi(δ, θ) | (X, ξ)] = 0. Let G(δ, θ) bethe vector formed by stacking the residuals Gi(δ, θ) over individuals.
I use generalized method of simulated moments to find the parameter values thatminimize the objective function
Λ0ZA−1Z 0Λwhere A is an appropriate weighting matrix which is a consistent estimate of E[Z 0ΛΛ0Z] andZ are instruments orthogonal to the composite error term
Λ =
∙1Jξ(δ, β) 1
Jω(δ, θ, η) 1
JT (δ, θ, ηAD) 1
NG(δ, θ)
¸0Chamberlain (1987) shows the optimal instrument for any disturbance-parameter pair
is the expected value of the derivative of the disturbance with respect to the parameters(evaluated at the true value of the parameters). Product and cost characteristics are op-timal instruments for the demand side parameter β and cost side parameter η, respectively.The optimal instruments for the other parameters are functions of either price or advertis-ing. Due to endogeneity of price and advertising, the instruments for the other parametersare not valid. Similar to the approach taken in Berkovec and Stern (1991), I form exoge-nous instruments for the disturbance-parameter pairs by simulating the expectations of thedisturbance-parameter pairs and regressing the simulated value on exogenous regressors.
20
C Figures and Tables
Industry Trends
$1,500
$1,700
$1,900
$2,100
$2,300
$2,500
$2,700
1996
1997
1998
1999
2000
Ave
rage
Pric
e (9
8$)
-1
1
3
5
7
9
11
13
Tota
l Uni
ts (i
n m
illio
ns)
Average Price Total Units SoldFigure 1
Advertising Expenditures
$1,250
$1,500
$1,750
$2,000
$2,250
$2,500
1995 1996 1997 1998 1999
Mill
ions
of 9
8 D
olla
rs
Figure 2: Advertising Trends
21
Manufacturer Advertising Total Ad to Sales Ad$ per MarketExpenditures Market Share Ratio Share Point
IBM $1,079 7.30% 19.55% $147.82Hewlett-Packard $466 9.62% 10.28% $48.44Gateway $358 15.07% 5.99% $23.75Dell $227 16.02% 2.28% $14.17Compaq $240 16.10% 2.56% $14.91Apple $181 8.88% 4.90% $20.37Total for PC market $2,068 3.34%Note: Dollars are in millions, Market Share is dollar market share of all sectors (home,business, education, and government)
Figure 1: Table 1: 1998 Advertising Expenditures for Selected Firms
Manufacturer Percentage Dollar Share Percentage Dollar Share All Sectors Home Market Sector
Total Sales $58,515 $66,636 $62,118 $16,529 $18,610 $17,673Notes: Total sales are in millions of dollars. In 1997 three mergers occurred:Packard Bell, NEC, and ZDS; Acer and Texas Instr.; Gateway and Advanced Logic Research
Table 2: Market Share for All Sectors and Home Sector
22
Variable Means Standard Coefficient estimates for interactionsDeviation
utility coefficients
interactions with demographic variableshousehold size income > $100,000 30<age<50 white male
Dell-IBM Merger:Dell Dimension 4% 1% 18%Dell Latitude* 4% 10% 40%IBM Aptiva 8% 0% 18%IBM Thinkpad* 8% 0% 92%Notes: Percentage changes are averages over all quarters.
Table 7: Post Merger Results for Selected Products
25
Median Percentage MarkupMedian Price Ad to Sales Ratio over Marginal Costs including ad costs
Total Industry $2,239 3.34% 19% 10%
Top 6 firm $2,172 8.66% 22% 12%
Apple $1,859 4.90% 19% 10%Compaq $2,070 2.56% 24% 16%Gateway $2,767 5.99% 12% 9%Hewlett Packard $2,203 10.28% 17% 11%IBM $2,565 19.55% 17% 10%Packard Bell $2,075 19.55% 18% 12%Note: Ad to sales ratios are from LNA and include ad and sales across all sectors. Percentage markups are the median(price-marginal costs)/price across all products. The last column is percentage total markups per unit after including advertising. These are determined from estimated markups and estimated effective product advertising in the home sector.
Table 8: Summary Statistics for Prices, Advertising, and Markups