Laurie Simon Bagwell, B. Douglas Bernheim

American Economic Association
Veblen Effects in a Theory of Conspicuous Consumption

Author(s): Laurie Simon Bagwell and B. Douglas Bernheim
Source: The American Economic Review, Vol. 86, No. 3 (Jun., 1996), pp. 349-373
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/2118201
Accessed: 23-04-2015 22:09 UTC
REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/2118201?seq=1&cid=pdf-reference#references_tab_contents
You may need to log in to JSTOR to access the linked references.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The American Economic
Review.
http://www.jstor.org
This content downloaded from 139.80.123.42 on Thu, 23 Apr 2015 22:09:57 UTC
All use subject to JSTOR Terms and Conditions
Veblen Effects in a Theory of Conspicuous Consumption
By LAURIE SIMON BAGWELL AND B. DOUGLAS BERNHEIM*
We examine conditions under which "Veblen effects" arise from the desire to
achieve social status by signaling wealth through conspicuous consumption.
While Veblen effects cannot ordinarily arise when preferences satisfy a "single-
crossing property," they may emerge when this property fails. In that case,
"budget" brands are priced at marginal cost, while "luxury" brands, though
not intrinsically superior, are sold at higher prices to consumers seeking to ad-
vertise wealth. Luxurybrands earn strictly positive profits under conditions that
would, with standard formulations of preferences, yield marginal-cost pricing.
We explore factors that induce Veblen effects, and we investigate policy impli-
cations. (JEL D1i, D43)
In his celebrated treatise on the "leisure erature, "Veblen effects" are said to exist
class," Thorstein Veblen (1899) argued that when consumers exhibit a willingness to pay
wealthy individuals often consume highly a higher price for a functionally equivalent
conspicuous goods and services in orderto ad- good.2
vertise their wealth, thereby achieving greater Anecdotal evidence suggests that Veblen
social status. Veblen's writings have spawned effects may be empirically significant in mar-
a significant body of research on "prestige" kets for luxury goods. According to one mar-
or "status" goods.' In the context of this lit- keting manager, "Our customers do not want
to pay less. If we halved the price of all our
* Respectively, Department of Finance, J. L. Kellogg products, we would double our sales for six
Graduate School of Management, Northwestern Univer- months and then we would sell nothing."3
sity, 2001 Sheridan Road, Evanston, IL 60208-2006, and Indeed The Economist (1993) emphasizes
Departmentof Economics, Stanford University, Stanford, that "[r] etailers can damage a glamorous
CA 94305-6072. The work reported here was supported good's image by selling it too cheaply." A
by the National Science Foundation through Grant No.
SES-9110211, and through a Presidential Young Investi- recent article in the Wall Street Journal noted
gator Award. The first author would also like to that "a BMW in every driveway might thrill
acknowledge the supportof the Lynde and HarryBradley investors in the short run but ultimately could
Foundationand the Hoover Institution.The authorswould dissipate the prestige that lures buyers to these
like to thank Kyle Bagwell, Scott Brandwein, R. Preston
McAfee, and the anonymous referees for helpful com-
luxury cars."4 Econometric evidence also cor-
ments and discussions. Previous drafts of this paper were roborates the existence of Veblen effects.5
circulated under the titles "Conspicuous Consumption, Recent incarnations of Veblen's theories
Pure Profits, and the Luxury Tax: Some Surprising Con- simply proceed from the premise that price en-
sequences of Perfect Competition" (draftdated November hances utility (see, for example, Leibenstein,
1991) and "Conspicuous Consumption, Pure Profits, and
the Luxury Tax" (draft dated September 1992).
1950; Braun and Wicklund, 1989; or Creedy
' See, for example, Harvey Leibenstein (1950), and and Slottje, 1991). Yet Veblen himself did not
more recently Robert H. Frank (1985), Kaushik Basu
(1987), Yew-Kwang Ng (1987), R. L. Basmann et al.
(1988), OttmarL. Braun and Robert A. Wicklund (1989),
John Creedy and D. J. Slottje (1991), and Norman Ireland 2 The following passage typifies modem discussions of
(1992). More generally, other recent studies, including prestige goods: "Conspicuous consumption, or Veblen ef-
George A. Akerlof (1980), Steven R. G. Jones (1984), fects, are said to occur when individuals increase their
Timothy Besley and Stephen Coate (1990), B. Douglas demand for a good simply because it has a higher price"
Bernheim (1994), Harold L. Cole et al. (1992), Chaim (Creedy and Slottje, 1991).
Fershtmanand Yoram Weiss (1992), and Amihai Glazer 'Quoted in The Economist (1993 p. 96).
and Kai A. Konrad (1992) explore the impact of status 'See Timothy Aeppel (1992 p. B4).
consciousness on economic behavior. ' See Creedy and Slottje (1991).
349
350 THE AMERICANECONOMICREVIEW JUNE 1996
endorse the view that the price of an object of conspicuous consumption would generate
affects utility directly, or that individuals seek Veblen effects. There is no particular reason
to pay high prices for the sheer pleasure of to believe that wealth is most effectively sig-
being overcharged. Rather, he proposed that naled by paying excessive prices for conspic-
individuals crave status, and that status is en- uous goods. Instead, one might prefer to
hanced by material displays of wealth. Ac- purchase a larger quantity of conspicuous
cording to Veblen, "In order to gain and to goods at a lower price, or a higher quality of
hold the esteem of men, wealth must be put in conspicuous good at a higher price.6
evidence, for esteem is awarded only on evi- This paper investigates the conditions under
dence" (p. 24). By social custom, the evidence which Veblen effects, defined as a willingness
consists of unduly costly goods that fall into to pay a higher price for a functionally equiv-
"accredited canons of conspicuous consump- alent good, arise from the desire to signal
tion, the effect of which is to hold the con- wealth. We examine a model in which each
sumer up to a standardof expensiveness and individual's status depends upon perceptions
wastefulness in his consumption of goods and of his wealth among social contacts. Consum-
his employment of time and effort" (p. 71). ers have private information about the value
Thus,in a theoryof conspicuousconsumption of their assets, and attempt to signal their
thatis faithfulto Veblen's analysis,utilityshould wealth by consuming a conspicuous good.
be defined over consumptionand status, rather The sellers of this good have access to iden-
thanover consumptionand prices. Althoughthe tical production technologies, and compete
prices that one pays for goods may affect status under conditions that would yield marginal-
in equilibrium,this relation should be derived, cost pricing under standard formulations of
not assumed. Moreover, since Veblen argued preferences. The model does not constrain
that individuals engage in conspicuous con- consumers to signal wealth by overpaying for
sumption to advertiseand provide evidence of visibly labeled conspicuous goods: it is also
wealth, the equilibriumrelation between price possible to signal by consuming large quanti-
and status should resultfrom signaling. ties of the good at a lower price, and/or by
The details of Veblen's argumentsnaturally selecting higher quality. Thus, to the extent
invite the interpretationthat conspicuous con- Veblen effects are present, they must be gen-
sumptionreflectssignaling.In particular,Veblen erated endogenously.
distinguishedbetween two motives for consum- We show that Veblen effects do not arise
ing conspicuous goods: "invidious compari- when the model satisfies the standard"single-
son" and "pecuniary emulation." Invidious crossing property" (which, in this context,
comparison refers to situations in which a states that the marginal cost of consuming the
member of a higher class consumes conspicu- conspicuous good is higher for individuals
ously to distinguish himself from members of with lower wealth, so that the indifference
a lower class. Pecuniaryemulationoccurs when curves of consumers with different levels of
a member of a lower class consumes conspic- wealth cross at most once).' However, when
uously so that he will be thought of as a mem-
ber of a higher class. In modem terms, these 6 Certain social customs in Thailand illustrate the prac-
motives are the essence of the incentive com- tice of advertising wealth through quantity, rather than
patibilityconditions that form the basis for sig- price. According to Philip Shenon (1991), "It is consid-
naling. Members of higher classes voluntarily ered acceptable, even by some Western-educated Thai
incur costs to differentiate themselves from women who would otherwise describe themselves as fem-
members of lower classes (invidious compar- inists, for a man to take one or more mistresses and even
to be seen with them in public, so long as all of the women
ison), knowing that these costs must be large and their children are provided for financially ... Mis-
enough to discourage imitation (pecuniary tresses are to some degree a demonstrationof wealth, and
emulation). as a rule, the more mistresses, the wealthier the man. A
Once the need to derive an equilibrium handful of Bangkok's flashier millionaires are said to have
10 or more extramaritalcompanions" (p. A4).
signaling relation between price and utility 7 Use of the single-crossing (or Spence-Mirrlees sort-
(through status) is acknowledged, it is nat- ing) condition is common in models with asymmetric in-
ural to wonder whether any plausible model formation. See, for example, David M. Kreps (1990).
VOL. 86 NO. 3 BAGWELLAND BERNHEIM: VEBLENEFFECTS 351
the single-crossing property fails in a particu- sympathetic to the role of any of the factors
lar way, and preferences satisfy a "tangency considered here) suggests that Veblen effects
property," Veblen effects may emerge. In par- are naturally rationalized within a signaling
ticular, the resulting equilibria are character- context.
ized by the existence of "budget" brands The existence of Veblen effects in the con-
(sold at a price equal to marginalcost), as well text of our model has some provocative im-
as "luxury" brands (sold at a price above plications for public policy. Since supranormal
marginal cost). Luxury brands are purchased profits result from the characteristics of de-
by consumers who seek to signal high levels mand rather than from the nature of strategic
of wealth. It is importantto emphasize that, in interaction among firms, evidence of high
equilibrium, the luxury brands are not intrin- profitability does not necessarily support in-
sically superior to the budget brands-they ferences of either collusion or oligopolistic
are simply goods of identical quality, sold at forbearance. This observation also has impli-
a higher price. The manufacturers of these cations for tax policy. Within our model, the
brands earn strictly positive economic prof- equilibrium prices of luxury brands are de-
its, even under conditions that would, with mand driven, ratherthan supply driven-that
standard formulations of preferences, yield is, luxury brands are sold at the consumer's
marginal-cost pricing, and despite the ability preferred price, which is tax inclusive, and
of firms to vary both price and quality. does not vary with the tax rate. Thus, as long
The theoretical plausibility of Veblen ef- as the tax per unit does not exceed the differ-
fects therefore depends upon the plausibility ence between the consumer's preferred price
of the single-crossing property.In the simplest and marginal cost, an excise tax on luxury
models of conspicuous consumption (for ex- brands amounts to a nondistortionary tax on
ample, Ireland, 1992), the single-crossing pure profits.
property is satisfied. Since consumption of This observation is of particularinterest in
conspicuous goods reduces expenditures on light of the Omnibus Budget Reconciliation
other goods, declining marginal rates of sub- Act of 1990, which, for a time, established
stitution imply that conspicuous consumption substantialfederal taxes on the sale of various
is more costly for households with less wealth. conspicuous goods, including expensive au-
As a result, overpayment for these goods does tomobiles, yachts, jewelry, and aircraft. One
not arise in equilibrium. Thus, one reading of should not conclude from our analysis that
our results (one that is based on the premise these taxes were nondistortionary;whether the
thatthe single-crossingpropertyholds) suggests demand for luxury items is characterized by
that Veblen effects are difficult to rationalize. Veblen effects is a question that can be settled
However, we also exhibit several slightly only through empirical analysis.
more elaborate models in which the single- However, it should be noted that several
crossing property fails, and where this failure predictions of our model are consistent with
gives rise to Veblen effects. For example, we anecdotal evidence. First, many individuals
demonstrate that the tangency property is sat- appear to consume conspicuous goods to ad-
isfied in the presence of bankruptcy con- vertise affluence. According to Daniel Piette,
straints. This follows from the fact that the vice-president of LVMH (a French conglom-
marginal cost of conspicuous consumption is erate that owns Louis Vuitton, Moet et Chan-
inversely related to wealth at low expenditure don, and Christian Dior perfumes), for many
levels, but positively related to wealth at high individuals buying luxury goods "is all about
expenditure levels. Remarkably,when Veblen demonstration."8 According to The Econo-
effects emerge, bankruptcyconstraints do not mist, "the most famous (example) was Ralph
bind in equilibrium, and so appear to be irrel- Lauren, whose Polo brand was positioned to
evant, despite the fact that Veblen effects appeal to American yuppies pretending to be
would not exist without them. Two other fac-
tors that can produce the requisite breakdown
of single crossing are also examined. Thus, a
second reading of our results (one that is 8 Quoted in The Economist (1993 p. 97).
Edwardian toffs" (p. 92). Indeed, The Econ- The paper is organized as follows. We
omist concludes that "price is ... a powerful describe the model in Section I. Section II
signal of exclusivity" (p. 98). This motivates examines signaling equilibria under the as-
marketing strategies that appeal to status consumption that the single-crossing property is
sciousness. For example, a recent Jaguar ad- satisfied. Section III analyzes cases where the
vertisement reads: "If you could drive one car single-crossing property fails to hold. The
to your high school reunion, this would be it. plausibility of the assumptions required to
As you swing into your alma mater in a beau- generate Veblen effects is discussed in Section
tiful new Jaguar XJS convertible, you can al- IV. Policy implications and other conclusions
most see the heads turnas your classmates ask, are considered in Section V.
'Isn't that ...?'
Second, brand-name producers apparently I. The Model
charge high premia on many status goods. In
some cases, these premia persist even though This section presents the model. Sections A,
the good is easily imitated.9As a result, man- B, and C describe the choices to be made by
ufacturersof status goods tend to earn supra- households, social contacts, and producers,re-
competitive returns.'0 This is most clearly spectively. The sequence of decisions in the
illustrated by cases in which nearly identical game and the conditions for equilibrium are
versions of the same good are sold at vastly presented in Sections D and E, respectively.
different prices. Marshall Schuon (1993) Section F defines the single-crossing property
notes that less expensive cars are often "vir- in the context of our model.
tually identical" clones of pricier models. For
example, "If you don't mind a different grille A. Households
and headlights, opting for the long-wheelbase
Bentley Brooklands at $152,400 ratherthan its Consider a household that must allocate re-
twin, the $178,200 Rolls-Royce Silver Spur sources over two types of consumption goods.
III, can save $25,800" (p. 20). One type is "conspicuous," in the sense that
Third, there is some evidence that the tax- its characteristics,as well as the quantity con-
inclusive pricesof certainluxurygoods were un- sumed, are publicly observed. The character-
affected by the luxury tax. Specifically, Rolls istics of the conspicuous good include quality,
Royce, Jaguar,and BMW have each run pro- q, where q E [q, q]. The second type of good
motionalcampaignsin which they offered to re- is "inconspicuous," in the sense that it is
imburse customers for the full amount of the consumed privately, and not observed by oth-
luxurytax. A 1991 advertisementforRolls Royce ers. Because of our assumptions about ob-
reads: "If the luxurytax is all that separatesus, servability, only conspicuous consumption
it's time to talk. From today throughDecember can potentially serve as a signal of wealth.
31, 1991, Rolls-Royce Motor Cars Inc. will re- The inconspicuous good is assumed, for sim-
imburse you for the full amount of the federal plicity, to be of fixed quality. We will use the
luxury excise tax incurredwhen you purchase inconspicuous good as the numeraire.
or lease a new Rolls-Royce or Bentley." This The household is endowed with resources,
offer was still in effect through1993." R, which it allocates to the consumption of the
conspicuous and inconspicuous goods. Let
x(q) denote the quantitypurchasedof the con-
'Ireland (1992) describes an interesting case of this spicuous good with quality q, and let s denote
involving a very expensive brand-namedbasketball shoe:
"The shoes became so much a passport to social success
among poor urbanteenagers that a campaign to limit their
commercial promotion and advertising was initiated"
(p. 2). your Jaguar dealer and we'll send you a reimbursement
' A number of examples are cited in The Economist check equal to the luxury tax based on the manufacturer's
(1993). suggested retail price." It is interesting to note that Jaguar
" Similarly, Jaguar advertised: "Now you can have did not offer this deal on the XJ-6, which is its least ex-
Jaguar luxury,free of the luxury tax ... Just buy or lease pensive automobile. An advertisement for BMW read:
a new 1990 or 1991 Sovereign, Vanden Plas or XJ-S from "We will pay the luxury tax on any new BMW..."
total conspicuous expenditures. Since the in- utility function, and write the utility of a type
conspicuous good serves as the numeraire,we i household as
use z to denote both the total inconspicuous
expenditure and the quantity purchased. (3) Wi (x, s, p) = Ui ( x, 'y(s, Ri ), p).
There are two types of households (H and
L), differing according to their levels of re- It will be analyticallyconvenientto assumethat
sources (RH and RL, with RL < RH). The thereis an absolutebound,s, on totalconspicuous
associated population frequencies are a and expenditures.This assumptioncan be modified
(1 - a), respectively. Each household knows or relaxedat the cost of additionalanalyticcom-
its own type, but cannot observe the type of plexity, withoutalteringour centralfindings.
any other household.
Each household of type i must respect the B. Social Contacts
following resource constraint:
Veblen's theory of conspicuous consump-
(1) Z ' y(s,Rj) tion is based on the premise that those who put
wealth "in evidence" are rewardedwith pref-
where, ordinarily,oy(s, Ri )Ios < 0 and oy(s, erential treatmentby social contacts. Our ob-
Rj)IR > 0. This generalized form of the re- ject here is not to explore the validity of this
source constraint subsumes several special premise, but rather to identify the conditions
cases considered later in this paper. Note that under which it gives rise to Veblen effects.
the standardbudget constraint, z < R - s, is Consequently, we adopt a treatmentof social
a special case of equation (1). contacts which, though highly stylized, cap-
Each household cares about its total quality- tures the essence of this premise.
weighted conspicuous consumption,defined as We assume that the payoff of the representa-
tive social contact is O(R, p), where R is the
resourcesof the householdwith which the social
(2) x f /(q)x(q)dq, contact interacts. We also assume that social
q
contactscannotobserve a household'sresources
directly,butmust insteadformconjecturesbased
where ,u(q) are weighting parameters (com- upon the household'sobservedactions.Contacts
mon to all households), assumed to be in- then choose p to maximize the expected value
creasing in q. We assume that households also of 4. Let r(7r) denote the value of p that maxi-
care about (i) consumption of the inconspic- mizes the expected value of 4 given R, and
uous good, z, and (ii) an action p taken by the given a subjectiveassessmentthatthe household
representative social contact. Total utility for is type H with probability7r. We will assume
type i = H, L is given by Ui(x, z, p). We that r(-) is strictlyincreasing so that, in partic-
assume that Ui is strictly increasing and con- ular, r(1) > r(O). Since Ui is increasingin p,
tinuous in each of its arguments. the properinterpretationof this monotonicityas-
Note that, with this formulation of utility, sumption is that social contacts would, given
higher quality is a perfect substitutefor greater perfect information,treat wealthier households
quantity. Yet in terms of the workings of the better. Define PH= r(1) and PL = r(O); pi is
model, there will be an important difference theninterpretedas the act of treatinga household
between quality and quantity,in that firms will as if it is of type i.
determine the set of available qualities, while
consumers will exercise any discretion that C. Producers
may exist with respect to quantities.
Since Ui is increasing in z, and since in- The conspicuous good can be produced by
conspicuous consumption is, by definition, not a large number of firms.'2 These firms are
observable, each household will certainly con-
sume z up to the point where equation (1)
holds with equality. Therefore, we can substi- 12 We have assumed that households' resources consist
tute the binding resource constraint into the of the inconspicuous good. Consequently, we abstract
dividedinto two groups.The firstgroupcon- froman entrant.In the contextof conspicuous

sists of F incumbents, indexed f E [1, ... , F]. goods, this assumptionis naturalsince incum-
The restof the firmsarepotentialentrants.All bentsmarketrecognizedbrands.We will also
firmscan producethe samerangeof qualities, resolve consumerindifferenceby assuming
q E [q, q] and conspicuous goods are other- thateach householdacquiresthe conspicuous
wise identical.Themarginalcost of producing good froma single vendorwheneverit is op-
a conspicuous good of quality q is given by timal to do so, ratherthan spreadingpur-
c(q), and the production technology exhibits chases among several vendors. This too is
constantreturnsto scale. naturalas long as there is some cost associ-
We will also assumethatconspicuousprod- ated with consummatingeach transaction.
ucts of some quality level q? are always avail- Finally, when householdsare indifferentbe-
ablefroman alternative(not modeled)source tween the offerings of several different in-
ata price,p?, thatis "prohibitive"in thesense cumbents,customerswill be allocatedso that
that, with perfect information,households aggregatequality-weightedvolume (that is,
would not purchaseany of the conspicuous units of x, ratherthan units of x(q)) is split
good at this price.1"The existence of this al- equally between these incumbents.'4 Note
ternative source simplifies the analysis of that these conditionswould yield marginal-
equilibria,but is not essentialto our analysis. cost pricingwith any standardformulationof
Note that,in any first-bestallocation,firms preferences.
will only producequalitylevels thatminimize
the expressionc(q)/lu(q). For simplicity,we D. Timing
will assumethatthereis a uniquequalitylevel
that satisfies this condition;we refer to it The gameunfoldsas follows. First,eachin-
henceforthas the first-bestqualitylevel, qF. In cumbentf announcesa qualitylevel, qf, and
addition,we will use CF to denote the mini- a price,pf, for the conspicuousgood. Second,
mizedvalue of c(q)l/p(q). potential competitorsobserve these quality
Eachfirmproducesa single product,which levels andprices, andthen decide whetherto
is "branded"(labeled) so thatsocialcontacts enter.If a firmchooses to enter,it announces
caneasily identifythe manufacturer. Branding a qualitylevel anda pricefor the conspicuous
does not affect utility directly, and in any good.5 Third, consumers observe all an-
ordinary(inconspicuous) context, branding nouncedqualitylevels and prices, and deter-
would be irrelevant.If insteadfirmswere al- mine the amountsof the conspicuousgood to
lowed to choose betweenlabelingandnot la- be purchasedfromeach firm.Eachconsumer
beling, some would label in equilibrium,and carriesoutthesetransactions, spendingin total
the outcomewouldbe unchanged. the amounts andacquiringin totalthe quality-
All consumersand social contactsobserve weighted volume x. Residualresourcesare
the pricesannouncedby all firms.Since social usedforinconspicuous consumption, z. Fourth,
contactsalso observebrandlabels, quantities, social contacts observe each household's
and qualities,they can infer any household's brandedconspicuous consumptionbundles,
totalexpenditureon conspicuousproducts,as
well as totalquality-weighted volume.
We endow incumbentswith the following
minoradvantageoverentrants:consumersbuy "'These assumptionsare not essentialfor the existence of
the productfrom an incumbent,unless they Veblen effects, but affect other featuresof the equilibrium.
s Since one can take F, the number of incumbents, to
can strictly improve their utility by buying be large, allowing for furtherentry in the second stage may
seem superfluous. One might therefore be inclined to de-
lete this stage. We do not believe that this would alter our
results. However, it would renderthe analysis more com-
from the processes by which that good is produced and plex, as the currentstructureallows us to ignore problem-
allocated. atic subgames (for example, if all firms name a very low
13 The existence of such a price follows, for example, or very high price). The reader should bear in mind that
if one assumes that OU,(O,z, p)l9x is finite. As motivation, potential entry in the second stage would only serve to
one can think of the alternativesource as "custom" work. strengthen competitive pressures.
VOL 86 NO. 3 BAGWELLAND BERNHEIM: VEBLENEFFECTS 355
form inferences about each householdi' s It is useful to describe the second-phase

wealth, and react accordingly (by choosing equilibria in a bit more detail. Let Q denote
p). The payoff to each household is given by the set of quality levels named by firms in the
Ui (x, z, p). Social contacts receive payoffs of first phase. For each q E Q, let P(q) denote
4(R, p), where R is the household's actual the set of prices announced by firms for con-
resources. Firm's payoffs are given by profits spicuous products of quality q. Define p(q)
(revenues minus costs). min P(q), andff(q) max P(q). For a con-
spicuous good of quality q and price p, we
E. Equilibrium Conditions define the "quality-weighted" price, pl,u(q);
this corresponds to the price of a quality-
Our game is divided into two main phases. weighted unit of conspicuous consumption
In the first phase (stages 1 and 2), firms com- (that is, a unit of x, rather than of x(q)).
pete by naming prices and qualities. In the sec- In light of equation (2), households would,
ond phase (stages 3 and 4), households select in the absence of informational imperfec-
consumption bundles and social contacts draw tions, choose the conspicuous good with the
inferences about households' characteristics. lowest quality-weighted price. Let p
The second phase is recognizable formally as minqE Qp(q)l,u(q), and -maxq EQp(q)I
a "signaling game," in the sense of Jeffrey S. ,u(q). In other words, p(pf) is the lowest
Banks and Joel Sobel (1987) or In-Koo Cho (highest) quality-weighted unit price quoted
and Kreps ( 1987). In particular,the household in the first phase.
is the "sender," and possesses private infor- Note that,unlessp = p an individual'squality-
mation concerning its type, R. The ex ante weighted conspicuousconsumption,x, does not
probability distributionover ( RL,RH I is com- uniquelydeterminehis totalconspicuousexpen-
mon knowledge, and is summarizedby a. Af- diture,s. Depending upon which brandshe se-
ter the household learns its type, it sends a lects, he may spend as little as px, or as much
''message," in this case (x, s), to the social as p-x.In fact, for any s satisfyingpx < s < p-x,
contacts, who play the role of "receivers." In it is possible to purchasex quality-weightedunits
response to this message, the social contacts of the conspicuous good for exactly s.18
select a "response," p. The payoffs to the No rational consumer would ever spend
sender (household) and receiver (social con- more thanthe minimalamountneededto ac-
tact) depend upon the sender's type (R), the quire a given quantity of the inconspicuous
message (x, s), and the response (p).'16 good. However, a consumer may be willing to
We reduce the set of equilibria through the spend more than px to acquire x units of the
use of a refinement that is similar in spirit to conspicuous good. Since others can observe
subgame perfection: for any outcome of the his level of consumption, x(q), his selection
first phase, actions and inferences constitute a of brands,brandprices, and brandqualities,
separating equilibrium in the second phase, they can infer his total expenditure,s, and
and this equilibrium satisfies the Cho-Kreps quality-weighted consumption,x.
(1987) "intuitive criterion," which is equiv- Formally,a separating equilibrium consistsof
alent to equilibrium dominance.'7 Given this totalconspicuousquality-weighted quantityand
selection criterion for the second phase of the expenditure choices(XL, SL, XH, SH), with (XL,
game, we look for Bertrand-styleequilibria in incentivecompatibility,
SL) * (XH, SH), satisfying
the first phase.
(4) WL(XL, SL, PL) 2 WL(XH, SH, PH)
16 For the receiver, dependence on the message itself is (5) WH(XH, SH, PH) 2 WH(XL, SL, PL)
degenerate,as is often the case in applied signaling models.
7 Equilibriumwith complete pooling can be ruled out
with the intuitive criterion. There do exist equilibria with
imperfect separation which survive the intuitive criterion
and strongerrefinements.Considerationof these equilibria 8 Specifically, the individualcould purchase q7xquality-
does not materially alter the analysis, so for simplicity we weighted units at F and (1 - q)x quality-weightedunits at
focus exclusively on full separation. p, where tq = [(slx) - p]I(p - p).
356 THE AMERICANECONOMIC REVIEW JUNE 1996
andfeasibility, ample,s, the absolutelimiton conspicuousex-

penditures,is sufficientlylow, then it may be
SH - impossible to deter imitationby consuming
(6) P_ - p high quantityor quality.For the moment,we
XH
will assumethat sufficientlyhigh quantityor
qualitydoes sufficeto deterimitation,so that
(7 ) psL-c the existence or nonexistenceof Veblen ef-
- XL
fects dependsupona comparisonof the desir-
abilityof signalingthroughprice,quantity,or
Moreover,social contactsform beliefs, rep- quality,ratherthanon the feasibilityof deter-
resented by a function *- mapping the message rence.We returnto this issue in SectionII.B,
(x, s) to a subjective probabilitythat the where we discuss factorsthat affect the pos-
householdis typeH. Thisfunctionmustsatisfy sibility of deterringimitation throughhigh
the restrictionsthat*(XL, SL) = 0 and *(XH, quantityor quality.
SH) = 1. Finally,the choices (XL, SL) and(XH, To statethis assumptionformally,we must
SH) must be optimal (for type-L and type-H first develop some additionalnotation, and
householdsrespectively)giventherelationbe- state one preliminaryresult.Define XL*(p) as
tweeninferencesandactionsthatis impliedby the solutionto maxXWL(x, px, PL) (where,for
the combinationof * and r(r). simplicity, we assume that this solution is
Ourdescriptionof a separatingequilibrium unique), and let W* (p) denote the corre-
is incompletein the followingsense:although sponding optimized value of the objective
we have specified total conspicuous con- function.We thenhave the following.
sumption(XL and XH) as well as total con-
spicuous expenditure(SL and SH), we have LEMMA 1: For any separating equilibrium
not indicated which brands are purchased. of the subgame beginning in stage 3, all
Theremay well be an infinitenumberof con- type-L households purchase xL*(p) quality-
spicuousconsumptionbundlescontainingX, weighted units of the conspicuous good at a
quality-weightedunits, and requiringan ex- total cost of PXL*(P).
penditureof exactly si (i = L, H).'9 Fortu-
natelythis is immaterial,since consumersdo The argumenthereis a standardone. In the
not care aboutbrandselection,except insofar separatingequilibrium,the L's are correctly
as it affects total cost. Indeed,consumersare identified.Therefore,they cannot (in a se-
completelyindifferentbetween all conspicu- quentialequilibrium)inducesocialcontactsto
ous consumptionbundlescontainingthe same reducep below PLby deviatingfromtheirpre-
totalnumberof quality-weighted unitsthatre- scribedchoice.Theiroptimalchoiceis thento
quirethe sametotalexpenditures.Thoughwe select feasible levels of consumptionwhich
havemadeseveralassumptionsto resolvecon- maximizetheirintrinsicutility.Consequently,
sumerindifference,ourresultsdo not depend if there is anotherx ? 0 and feasible s that
on the specificsof these assumptions. raisesthe valueof WL(X,S, PL), it mustmake
It shouldbe noted that,given our assump- thembetteroff. This contradictsthe supposi-
tions so far,Vebleneffects mnight arisesimply tion thatan equilibriumprevails.
becauseit is impossibleto deterimitationby We arenow preparedto providethe follow-
type-L householdsexcept by paying inflated ing sufficientconditionforthefeasibilityof de-
prices for the conspicuousgood. If, for ex- terringimitationwithoutpayinginflatedprices
for the conspicuousgood:for allp 2 CF,
(8) WL(S/P, S, PH) < WL(P).

9 Suppose,for example, thatthereare threebrands,A, B,
and C, of qualitiesqj (j = A, B, C), sold at pricespj, where This conditionstates that type-L households
and suppose that PA'
pAI/(qA) < pB14(qfi) < pcI4(qc),
I'(qA) < si xi < pcIL(qc). Then there is an infinitenumber
would choose not to imitate type-H house-
of bundles,(XA, XB, XC), satisfying PAXA + PBXB + PCXC = Si, holds if type-Hhouseholdspurchasedenough
XA4(qA) + xA(qB) + xc/I(qc) = xi, and xj 0O. of the conspicuousgood at quality-weighted
Definition: Preferences satisfy the single-

crossing property if, for any feasible (x, x', s,
s', p) with 0 < x < x', 0 - s < s' < s, and
P E [PL, PH], WL(X, St, P) ? WL(X, s, P)
implies WH(x', s', p) > WH(X, s, p).
Example: Suppose that household utility
is additively separablein x, z, and p, and that
H~~~~~~~' the resource constraint is given by z < R - s.
Then
(9) Wi (x, s, p) = u (x)
FIGURE 1. SINGLE CROSSING AND THE ABSENCE

+ v(Ri - s) + w(p).
OF VEBLEN EFFECTS
In that case, WL(x', s', P) 2 WL(x, s, p) im-
plies u(x') - u(x) 2 V(RL - S) - V(RL -
s'). But as long as v is strictlyconcave, v(RL -
s) - v(RL - s') > v(RH - s) - V(RH - s ).
prce p. Consequently, type-H households Combining these statements gives us WH(x',
could, if desired, differentiate themselves by S', p) > WH(X,s, p). Consequently, this ex-
selecting high quantity at the lowest available ample satisfies the single-crossing property.
price, rather than by overpaying for a lower Section II demonstratesthat Veblen effects
quantity. cannot emerge in our model when the single-
crossing propertyholds. Section III establishes
F. The Single-Crossing Property that, in contrast,Veblen effects do emerge un-
der alternativeassumptionsconcerning house-
As discussed in the introduction,our objec- hold preferences.
tive is to determine whether Veblen effects
emerge as a consequence of signaling. As we II. Equilibriumwith the Single-
will show, the existence of Veblen effects CrossingProperty
hinges on the propertiesof the households' in-
difference curves in the (x, s) plane. In this section, we characterizethe separat-
A condition known as the "single-crossing ing equilibria of this model when the single-
property""often plays an important role in crossingpropertyholds. Section A demonstrates
models with asymmetric information.Figure 1 that the model cannot generateVeblen effects.
depicts indifference curves in the (x, s) plane SectionB considerstherobustnessof this finding.
for type-L households (IL, corresponding to a
constant value of WL(X,s, p)) and for type-H A. Analysis of Veblen Effects
households (IH, corresponding to a constant
value of WH(x, s, p)) that satisfy this property. The following result demonstrates that
Note that the curves IH and IL cross only once, Veblen effects cannot arise when the single-
and that, at the crossing, the slope of IH is crossing propertyis satisfied.
steeper than the slope of IL. To appreciatethe
economic content of this property, define the THEOREM 1: Suppose that the single-
"benefit ratio" as the ratio of the utility gains crossing property holds, and that, entering
associated with another unit of the conspicu- stage 3, p 2 c'. Then every equilibrium for
ous good, to the utility losses associated with the continuation game has the property that
another dollar of conspicuous expenditure. all households purchase the conspicuous
When the single-crossing property holds, the good at the quality-weighted price p. Fur-
benefit ratio is always higher for households thermore, on the equilibrium pathforthe en-
with greater resources. Formally, the single- tire game, p = c F, and only conspicuous
crossing property is defined as follows. goods of quality level qF are produced.
A formal proof of Theorem 1 appearsin the good consumed (as in Ireland, 1992), rather
Appendix. The intuitionfor this result is as fol- than by the prices or qualities of the brands
lows. The fact that type-L households buy the chosen.20
conspicuous good at the quality-weightedprice
p follows directlyfrom Lemma 1. It is certainly B. Robustness
possible for type-H households to discourage
imitationby choosing(XH, SH) with SH > PXH. We now establish the robustness of the re-
But in that case, type-H households could also sult obtained in Section II.A by examining the
consider signaling by purchasing a larger roles of several potentially important, and
quality-weighted quantity at a lower quality- possibly objectionable, assumptions. The first
weighted price. If the single-crossing property of these concerns the nature of competition
is satisfied, then the benefit ratio is higher for among the producersof the conspicuous good.
type-H households, which makes it possible to From the statementof Theorem 1, it should be
choose an increasein quantityand a decrease in evident that the absence of Veblen effects does
effective price that makes type-H households not depend upon the nature of competition
betteroff, while leaving type-L imitatorsworse among firms. Note in particularthat, in every
off. The intuitive criterionthen guaranteesthat, continuation game, all households purchase
upon observing this deviation, social contacts the conspicuous good at the lowest available
would infer thatthe deviatorwas of type H, and quality-weighted price, p. A different model
respondaccordingly.But in that case, the equi- of competition among producers of the con-
librium with SH> PXH would be undermined. spicuous good might produce an equilibrium
The second half of the theorem (conspicuous price other than CF, but would not alter the
goods of quality qF are available at price p = nature of any continuation game.2' Thus, re-
cF) then follows from standardBertrand-style gardless of the process generating price and
arguments. quality choices, when faced with a choice be-
The central idea of the theorem is illustrated tween a higher-priced brand of the conspicu-
graphically in Figure 1. If type-H households ous good and a lower-priced brandof the same
choosesome (XH, SH) with SH/XH > cF, single- quality, households will always choose the
crossing implies that the shaded area is lower-priced brand.
nonempty. Thus, it is possible for type-H In deriving Theorem 1, we have also as-
households to increase their utility, without in- sumed that the quantity of the conspicuous
ducing imitation, by purchasing more of the good is variable. This assumption is appropri-
conspicuous good at a lower price. More gen- ate even if conspicuous goods are discrete ob-
erally, it should be evident from this figure that jects that are used one at a time, such as
efficient signaling is ordinarily inconsistent watches, cars, and silk ties. Someone possess-
with prices in excess of marginal cost unless ing many expensive watches can wear a Rolex
there are no points above IL, below IH, and on Monday, a Patek Phillipe on Tuesday, a
above the line s = cFx, which would require Cartieron Wednesday, and so forth. Similarly,
the indifference curves to be tangent at (XH, there are well-publicized examples of wealthy
SH). This observation anticipates the main re-
sult of the next section.
Although Theorem 1 does not establish the
existence of an equilibrium satisfying the in- 20 With other specifications of utility, type-H house-
tuitive criterion where SH = C FXH, this follows holds might signal wealth through their choices of both
quantity and quality. However, as long as the single-
from standard arguments. Thus, our analysis crossing propertyholds, they would never choose to over-
establishes that, when the single-crossing pay for a conspicuous good.
propertyholds, there can be no Veblen effects: 21 While we have constructed a Bertrand-stylegame in
no household would choose to pay a higher which equilibriumprices are driven to marginalcost, other
price in order to enhance its status. To the ex- models of the competitive process might well produce
higher prices (for example, when entry barriers are high
tent signaling distorts the choices of type-H and producers set capacity prior to choosing prices, as in
households, these households differentiate Kreps and Jose Scheinkman [1983] and Carl Davidson and
themselves by the quantity of the conspicuous Raymond Deneckere [1986]).
celebrities who accumulate "stables" of ex- and

pensive automobiles. Although these individ-
uals wear only one watch at a time and drive (11) WH([t(qH), PH, PH)
only one car at a time, they manage to display
quantity through ostentatious variety. More- 2 WH(4(qL), PL, PL).
over, even when variety across time cannot be
observed (for example, if social contact is in- Moreover, PL and PH must be prices actually
frequent), the ability to select an arrayof con- named by firms in the first phase of the game,
spicuous goods serves the same role as and qLand qHmust be qualities actually named
variable quantity. If ownership of an Armani by firms in the first phase of the game.
suit does not suffice to differentiate the Define q4 as the solution to maxqWL(4(q),
wealthy from pretenders,those who wish to sig- c(q), PL), and let WLdenote the correspond-
nal may add a Rolex watch; if that fails to do ing maximized level of utility. We will assume
the trick, they may also brandisha Mont Blanc for simplicity that qLis unique, and that WL>
fountainpen, and a Hermestie or scarf. As long WL(O, 0, PH), so that type-L households will
as the single-crossing propertyis satisfied, ar- actually purchase the conspicuous good. We
gumentsanalogousto those used in the proof of will also assume that24
Theorem 1 imply that households will preferto
signal with greater variety (quantity of differ- (12) WL(,U(O, C(), PH) < WL.
entiated conspicuous items), rather than by
overpayingfor less variety. Condition (12) takes the place of condition
Finally, as we now demonstrate, even if (8) (for the case of variable quantity) as
quantity and variety are both fixed, Veblen ef- a sufficient condition for the feasibility of
fects still cannot arise in equilibrium,as long as deterring imitation without paying inflated
firms are free to vary quality.22We modify the prices for the conspicuous good.25In particu-
model by assuming that each household pur- lar, type-L households will not imitate type-H
chases either one unit of the conspicuous good households if type-H households purchase the
from a single firm at a single price and quality, highest quality conspicuous good at a price
or none. A separatingequilibriumfor the sec- equal to marginalcost. With these changes, we
ond phase of the game consists of (qL, PL, qH, have the following theorem.
PH) with (qL, PL) * (qH, PH) suchthat2
THEOREM 2: Suppose that the single-
(10) WL(/1(qL),PL, PL) crossing property holds, and that each house-
hold buys at most one indivisible unit of the
- WL(/I(qH), PH, PH) conspicuousgood. ThenPL = c(qL) andPH =
c(qH).
A proof of this result is contained in the Ap-

22
pendix. The intuition is similar to that given
It is importantto realize that this conclusion does not for Theorem 1. The fact that there is no
follow directly from Theorem 1. Even though quality is,
in our model, a perfect substitute for quantity from the markup over marginal cost on units of the
perspective of households, control over available qualities conspicuous good sold to type-L households
resides with firms, rather than households. Since house-
holds must choose among the products that are actually
offered, the case of fixed quantityand variablequality give
households much less discretion than the case of variable 24Technically, with indivisibilities, we must also as-
quantity. sume that c(q < s-, so that it is feasible to purchase one
23 It is conceivable that one might also construct a sep- unit of a conspicuous product with quality q at a price
arating equilibrium in which some households purchase equal to the marginal cost. This is not restrictive, since, in
none of the conspicuous good. This would necessitate the quantity-constrainedmodel, we could simply define q
modification of the incentive constraints (10) and (11). as C-S
However, subsequent assumptions will rule out the pos- 25 In some sense, condition (12) is less restrictive than
sibility that such an equilibrium would survive the appli- condition (8), in that it only requires the inequality to be
cation of the intuitive criterion. satisfied when price equals marginal cost.
follows from standard Bertrand-style argu- Let q = 1 denote the single feasible level of
ments; competition among producers will quality. Once again, a standardBertrand-style
make quality qL available at price c(qL). It is argument implies that some firm will sell the
certainly possible to construct an equilibrium conspicuous good at price p = c( 1), and that
in which type-H households discourage imi- type-L households will purchase this brand.
tation by choosing (qH, PH) with PH > C(qH) - Define PHas the solution to27
But in that case, the single-crossing property
guaranteesthe existence of an alternativeoffer- (13) WL(,U(1), C(1), PL)
ing with higherqualityanda lower markup,such
that type-H households would strictly pre- = WL(,U(1), PH, PH)-
fer this alternative, while type-L households
would strictly prefer (qL, PL). The intuitive It is easy to verify that, in any equilibriumsat-
criterion then implies that, if this alternative isfying the intuitive criterion, some firm will
was offered, and if a household selected it, so- offer the conspicuous good at price PH, and
cial contacts would infer that the deviating type-H households will purchase this brand,
household was of type H, and respond accord- despite the availability of an equivalent prod-
ingly. Consequently, if the alternative is of- uct at the lower price, c( 1). Thus, the fixed-
fered, type-H households will select it. But this quantity,fixed-qualitycase gives rise to Veblen
undermines the equilibrium, since an entrant effects even in equilibrium. This is the es-
would then have an incentive to offer this sence of the signaling equilibrium described
alternative. by Wolfgang Pesendorfer (1995). It should
Notice that Theorem 2 is weaker than The- be evident, however, that Veblen effects,
orem 1, in the sense that it only describes when generated in this way, are highly fragile
behavior on the equilibrium path.26In equi- and depend on a variety of tenuous assump-
librium, households never purchase an exces- tions, including (i) households cannot vary
sively priced conspicuous good, so Veblen quantity (either by changing the amount used
effects do not arise as an equilibrium phe- at one time, or by using distinguishable units
nomenon. However, off the equilibrium path, at different points in time), (ii) firms cannot
type-H households may well elect the more vary quality, (iii) there are no other conspic-
expensive of two equivalent brands, for the uous goods with either variable quantity or
simple reason that no other alternatives are variable quality, and (iv) there exist rela-
available, and acquisition of the less expen- tively few categories of conspicuous goods,
sive brand would not be sufficiently costly to so that households cannot display wealth
deter imitation by type-L households. Al- through ostentatious variety.
though Veblen effects can therefore arise out
of equilibrium, it is importantto reiterate that III. EquilibriumWithoutthe Single-
even this is a fragile result. The existence of CrossingProperty
even a single conspicuous good with flexible
quantity, or the existence of a large variety of In the last section, we demonstrated that
conspicuous goods (such as watches and Veblen effects cannot arise when the single-
cars) each with fixed quantity, effectively crossing property holds if quantity, quality,
brings us back to the variable-quantity case, or variety is variable. However, signaling
and eradicates Veblen effects both on the equilibria may exist even when the single-
equilibrium path and out of equilibrium. crossing propertyis violated. This observation
For completeness, it is worth mentioning the suggests that it may be possible to generate
case where both quantityand quality are fixed. Veblen effects under less standard assump-
tions. In this section, we focus on cases in
which the violation of the single-crossing
26 It should be noted that this is not attributableto the
difference between conditions (12) and (8) described in

the preceding footnote, but rather follows from the fact 27 A sufficient condition for the existence of PHis that
that firms, not households, choose available qualities. WL(/1(1), CO), PL) 2: WL(/,(1), S, PH)-
separationis feasible).28 In equilibrium, some

S~~~~~~~~~ firms market "budget" brands (which are
sold at a price equal to marginal cost), while
others sell "luxury" brands (which are sold
IH at a price above marginal cost, despite the fact
s*(x) that producers are perfectly competitive).
Luxury brands are not intrinsically superior
to budget brands but are purchased by con-
sumers who seek to signal high levels of
wealth. Section B examines the robustness of
x
this finding.
FIGURE 2. THE TANGENCY PROPERTY A. Analysis of Veblen Effects

To analyze the characteristics of equilibria
when the tangency propertyholds, we require
propertytakes a particularform, which we re- some additional notation. Specifically, let x*
fer to as the tangency property. At this point, denote the solution (if any) to the equation
we introduce this property as a technical con-
dition, and explore its implications. In Section (14) WL(X*, S*(X*), PH) = WL(CF).
IV, we identify more primitive conditions
from which this property can be derived. We illustrate this solution in Figure 3, where
Definition: Preferences satisfy the tan- we have superimposed the indifference curve
gency property if there exists a continuous described by the equation WL(X, S, PH) =
function s*(x) such that for any (x, x', s') WL(CF) on s *(x). The point x* corresponds
(with s' * s*(x)), WL(X', S', PH) = WL(X, to the intersection of this indifference curve
S*(X), PH) implies WH(X',S', PH) < WH(X, and s*(x). It is, of course, possible that s*(x)
S*(X), PH)- could lie everywhere below, or everywhere
We illustrate this property in Figure 2. No- above, the indifference curve of interest, in
tice that the indifference curves IL and IH which case there would be no intersection. For
(which presuppose p = PH) are tangent to each our central result, we will simply assume the
other where they cross the function s*(x). existence of a unique intersection.29In Section
This could occur, for example, if the benefit IV, we discuss conditions under which this as-
ratio (defined in Section I.F) is higher for sumption is satisfied.
households with greater resources when total It is easy to verify that the indifference curve
conspicuous expenditures are less than s*(x), of interest must lie strictly above the line s =
and lower when conspicuous expenditures ex- Fx at the pointX*(CF) .3 Thus,dependingon
ceed s*(x).
When the tangency property is satisfied,
mutual nonimitation may be infeasible. We
28Veblen effects may arise when the single-crossing
therefore use the following additional condi-
propertyfails to hold, even when the tangency propertyis
tion to guarantee the existence of separating not satisfied. However, the analysis of equilibria is partic-
equilibria. ularly tractable when the tangency property holds.
Definition: Separation is feasible if there 29Once existence is assumed, it is easy to establish that
exists (XH,SH) * (XL (CF), CFx (CF)) with SH/ the intersection must be unique. Specifically, if a type-L
indifference curve crosses s*(x) more than once, it must
XH2 C such that(XL(C), CL (C),XH, SH) be tangent to the same type-H indifference curve at each
satisfies expressions (4) and (5) (incentive crossing (otherwise type-H indifference curves would
compatibility). cross). But this contradicts the strict inequality in the def-
The remainder of this section is organized inition of the tangency property.
Consider s solving WL(XL(C), S, PH) = WL(X
as follows. Section A demonstrates that S*(X*), PH). Since WL(X*, S*(X*), PH) = WL(X(CF),
Veblen effects can emerge in our model when CFX4(CF), PL) WL(CF), it follows from PH > PLthat s>
the tangency property holds (assuming that CFX*(CF).
the location of s*(x), it is certainly possible

that s *(x * ) would exceed cFX *,3 as shown 1; . . ; 0 . . '
in Figure 3. The following result demonstrates

that Veblen effects do arise in equilibrium for
this case.
THEOREM 3: Suppose that the tangency

property holds, that separation is feasible,
and thats*(x*)/x p**> cF. Then(i) (XL, ~~' - x
SL) = (XL*(CF), CFXL*(CF)) and (XH, SH) =

(x*, s*(x*)), (ii) only conspicuous goods of
quality q F are purchased, (iii) type-H house-
holds only patronize incumbents, who earn FIGURE 3. VEBLEN EFFECTS WITH
strictly positive profits, and (iv) effective THE TANGENCY PROPERTY
quality-weighted prices, (PL, PH), and aggre-
gate profits are independent of F (the number
of incumbents).
reduces price to some p between CF and p*,
A formal proof of Theorem 3 appearsin the then type-H householdswill be able to in-
Appendix. The intuition for part (i) is closely crease utility holding p constant by selecting
related to the intuition given for Theorem 1. a point in the shadedarea of Figure 3. But
The fact that type-L households buy the con- type-L householdspreferevery point in the
spicuous good at the quality-weighted price CF shaded area to (x4(cF), cFx*(cF)); conse-
follows from Lemma 1 and Bertrand-style quently, they would imitate this choice. Since
competition among producers. It may be pos- type-H households wish to avoid imitation, no
sible for type-H households to discourage im- luxury brand producer can profitably attract
itation by purchasing a sufficiently large additionalbusiness by cutting price. Under our
quantity of the conspicuous good at the price assumptionsconcerningthe resolutionof con-
C'. But in that case, type-H households would sumer indifference, the resulting profits are
do well to consider signaling by purchasing a shared equally by the incumbent firms. More-
smaller quality-weighted quantity at a markup over, since the equilibriumprice PH is demand
over cost. Given the assumed failure of the driven (that is, it is the "optimal" price for
single-crossing property, it is possible to type-H households) ratherthan supply driven,
choose a decrease in quantity and a markup the key features of equilibrium are indepen-
that makes type-H households better off, while dent of the number of incumbents.
leaving type-L imitators worse off. The intui- Severalaspectsof Theorem3 deserveem-
tive criterion then guarantees that, upon ob- phasis. With the tangency property,Veblen ef-
serving this deviation, social contacts would fects are apparent: type-H households pay
infer that the deviator was of type H, and re- p*i(qF) for the conspicuous good, even
spond with PH. But in that case, the equilib- though a qualitatively identical brand is avail-
rium would be undermined. able at a priceof c(qF) < p*ii(qF). No firm
In contrast, suppose that type-H households can attract type-H purchasers by lowering
attempt to distinguish themselves from type price, since, with a lower-priced brand, it is
L's by purchasingx* units at the inflated price more costly in total to deterimitationby type-L
p * > cF. As illustrated in Figure 3, no other households (recall that a higher quantity
point preferredby type-H households is con- would be required). Although signaling dis-
sistent with the deterrence of imitation by torts the price and quantity choices of type-H
type-L households. If a luxury brandproducer households, it does not distort their quality
choices. Moreover, firms continue to select
first-best quality. Finally, positive profits pre-
3' For example, if x* is sufficiently close to x*(cF), it vail despite the fact that we have assumed
necessarily follows that s*(x*) > cFx*. homogeneous goods, free entry, a constant
VOL 86 NO. 3 BAGWELLAND BERNHEIM: VEBLENEFFECTS 363
returns to scale technology, and Bertrand- cific choice of qF is a somewhat special

style pricing-conditions that would yield consequence of assumptions about functional
marginal-cost pricing (even when indiffer- forms. However, the finding that firms will not
ence is resolved in favor of incumbents) with dissipate profits throughquality competition is
ordinary formulations of preferences. quite general, and holds even when quality is
not a perfect substitute for quantity. To clarify
B. Robustness this point, consider the equilibrium described
in Theorem 3, where type-H households pur-
We now establish the robustness of Veblen chase x*/,(qF) units of a quality qF conspic-
effects when the tangency property holds by uous good at a total cost of s*(x*). Now
examining the roles of several potentially im- suppose that some firm can, through some
portant assumptions, including those high- costly activity (for example, advertising or in-
lighted in Section II.B. novation), enhance the value of its product. If
Once again, it is evident that different as- it undertakes this action without changing
sumptions about the nature of competition price or quality, it will not succeed in luring a
among producers of the conspicuous good single buyer away from its competitors. The
would not qualitatively alter our result. Our reason is that the activity would also enhance
analysis of signaling equilibria for the second the value of the productfor type-L households,
phase of the game implies that, regardless of who would then prefer this enhanced package
the process generating p and pf, type-H house- to (xL(c'), cFxL(cF)). Thus, the enhanced
holds will, in a wide range of situations, select package would not function as a signal of
a higher-priced brand over a qualitatively wealth, and thereby attracttype-H households,
equivalent lower-priced brand. unless the producer simultaneously raised
Veblen effects are also robust with respect price and/or lowered quality.32
to the alternative assumption that quantity is A final robustness issue concerns our as-
fixed (at one unit). For the same reasons as in sumption that there are only two types of
Section II.B, competition among producers households. When the single-crossing prop-
will make quality 4L available at price c(lL) erty is satisfied, the extension of our model to
(where the definition of 4L is unchanged). De- an arbitrary number of types is completely
fine q * as the solution to standard.Although standardargumentsdo not
apply when the tangency property holds, a
(15) WL(,O(q*), s*(,I(q*)), p) = WL. somewhat stronger version of this condition
permits the introduction of additional types
The term q * serves a role analogous to that of with little added complexity.33Suppose in par-
x* from the variable-quantitycase (see equa- ticular that there are I + 1 types of households
tion ( 14) in Section III.A). Using an argument indexed i = 0, 1, 2, ..., I, with resources Ri
completely analogous to that given in the (Rj < Rk for j < k). For each type i > 0, we
proof of Theorem 3, it is possible to show that, will assume that there exists a continuous
as long as s*(,iu(q*)) > c(q*), type-H house- function s* (x) such that for any (x, p, x', s',
holds purchasequalityqH = q * at pricePH = p') with s' * s (x), Wo(x', s', p') = (x
s *( t( q *)), and that the other features of The- s (x), p) implies Wi(x', s', p') < Wi(x,
orem 3 are unchanged. Naturally, Veblen ef-
fects also arise if both quantity and quality are
fixed. 32 In an earlier version of this paper (Bagwell and
Another feature of equilibriumwith variable Bernheim, 1992), we considered a model with advertis-
ing in which advertising enhanced utility, but, unlike
quantity described in Theorem 3 is that all quality in the current model, did so in a way that was not
firms choose to produce the first-best quality, a perfect substitute for quantity. We demonstrated that
qF. Thus, despite the imposition of assump- firms would not dissipate profits through competition in
tions that would yield marginal-cost pricing advertising.
" This analysis extends that of Bernheim (1991), who,
under standard formulations of preferences, in a much different context, analyzed signaling equilibria
finns will not dissipate excess profits by com- under a failure of the single-crossing property somewhat
peting in quality. As in Section II.B, the spe- analogous to the tangency property.
364 THE AMERICANECONOMIC REVIEW JUNE 1996
si*(x), p).34 We use xi* to denote the solution Example 1: Personal Bankruptcy. Con-
to Wo(x, s(4), Pi) = W *(cF) (where sider a modified version of the game, in which
W (cF) is definedanalogouslyto W L (CF)). consumers must borrow to finance conspicu-
To constructthe equilibrium,we assign type-0 ous expenditures, and where loans to house-
households the bundle (x*(cF), cFX*(cF)) holds are potentially risky, since households
(analogously to type-L households in the two- may choose to default. The game unfolds in
type case), and type-i > 0 households the bun- five stages rather than four, with the set of
dle (x, s* (x)). In otherwords,we select agents expanded to include a large number of
the bundle for each type-i > 0 household to potential creditors, who can borrow and lend
deter imitation by the lowest type, ratherthan money at the riskless rate r* (their opportunity
(as in the standard case) by its next lowest cost of funds). First, each incumbent f an-
type. With a pair-wise analog of feasible sep- nounces quality qF and price, pf, for the con-
aration between type 0 and each type i, this spicuous good, and each creditor n announces
assures us of mutual nonimitation for each an interest rate, r,. Second, potential conspic-
(0, i) pair. Note that W0(x , s * (xi4), pi ) = uous good producers observe these qualities,
W (cF) = Wo(x s>(xp),p)for each (i, j) prices and interest rates, and then decide
pair (with i, j > 0). It follows immediately whether to enter. If a firm chooses to enter, it
from this observationand the modifiedtangency announces a quality and price for the conspic-
condition that mutual nonimitationis satisfied uous good. Third, consumers observe all an-
for (i, j) (stricdy if s* (4) SJ(xJ)). nounced qualities, prices and interest rates,
and determine the amounts of the conspicuous
IV. Are VeblenEffectsPlausible? good to be purchased from each firm. Since
income is not received until stage 5, these pur-
Our analysis has isolated theoretical condi- chases must be financed through borrowing.35
tions that allow us to rationalize the existence We assume for simplicity that creditors can-
or nonexistence of Veblen effects. However, not monitor the total indebtedness of any cli-
the fact that it is possible to produce Veblen ent. Thus, households can obtain any desired
effects under some appropriateset of assump- loan at the prevailing rate market of interest,
tions does not necessarily imply that one is r ? r*. Fourth, social contacts observe house-
likely to observe these effects in practice. In- holds' branded conspicuous-consumption
deed, the conditions required to generate bundles, form inferences about households'
Veblen effects may strike the reader as im- resources, and react accordingly (choose
plausible. In this section, we examine the p).36 Social contacts do not observe house-
plausibility of the tangency property. We holds' choices of lenders. Fifth and finally,
then discuss in greater detail the effective- income (R) is received and loans mature.
ness of conspicuous expenditures as a signal Each household has the option to default on
of wealth. its loan, in which case creditors receive an
A. Justificationsfor the Failure of the

Single-Crossing Property " Nothing would change if households received a por-
tion of their incomes prior to stage 3. The extreme as-
In this section, we demonstratethat the tan- sumption that all income is received in stage 5 is made for
gency property can be derived from more simplicity only.
36 Note that inferences are drawn in stage 3, prior to the
primitive assumptions. Three separate exam-
time at which bankruptcymay be declared. We have there-
ples are considered. fore assumed implicitly that a declaration of bankruptcy
would not affect status (possibly it is unobservable). How-
ever, this assumption is not essential. Even when a dec-
laration of bankruptcy negatively affects status, the
marginal cost of conspicuous spending will be higher for
" Note that, under this assumption, the locus of tan- higher-income households past the point where the bank-
gencies between indifference curves is independent of p. ruptcy constraint binds for lower-income households.
This propertyarises naturallywhen, for example, utility is Consequently, with some technical qualifications, our re-
separable in p and (x, s). sults will be qualitatively unchanged.
enforceable claim against the household's S

I, ~~~IH
income. Since bankruptcy protection allows
the household to retain resources of the
amount z, independent of type, a declaration
of bankruptcy leaves the household with re- (R -Z)/(1+r) ___ _ ,_I_
sourcesequal to max I z, R - (1 + r)s }. We (RL- Z)/1 + r)___ _ _______
assume that the household's conspicuous
consumption is unaffected by bankruptcy (in
other words, the conspicuous good cannot be
repossessed) ." All residual resources are
spent on the inconspicuous good at the end x
of stage 5. As in the example in Section I.F,
utility is additively separable, Ui (x, s, p) = FIGURE 4. INDIFFERENCE CURVES
WITH PERSONAL BANKRUPTCY
u(x) + v(s) + w(p).
Once creditors have selected interest rates,
the game is merely a special case of the model
considered in Section III. The utility of the Under relatively mild conditions, incum-
household is described by equation (3), where bent producers name quality qF and price
p *,((qF), creditors name the interest rate r*,
(16) y(s, R1) = max{z,R -(1 + r)s}. entrants
namequalityqF andpricec(qF), type-L
householdschoose (XL (c ), CFXL (CF)), type-H
Indifference curves in the (x, s) plane for households choose (x *, s *), and no consumer
type-L households, IL, and for type-H house- defaults on any loan. With more than two
holds, IH, are depicted in Figure 4. It is easy to types of households, each i >-1 would spend
verify directly that, given appropriatechoices exactly the same amount (s *), but those with
of CF and s-, the tangency property holds for greater resources would purchase a smaller
this model (including the stronger version of volume at a higher price.39
the condition introduced in our discussion of The absence of default (in equilibrium) de-
multiple types), as shown in Figure 2. In ef- serves emphasis. Type-L households spend the
fect, the marginal cost of conspicuous con- amountCFX4*(CF) on conspicuousconsump-
sumption is higher (and the benefit ratio is tion. Since, by assumption, s* > cF4L(cF),
lower) for type-L households as long as total their bankruptcy constraint does not bind.
expenditures do not exceed s * = (RL - Z) Since type-H households spend the amount
(1 + r). However, for higher levels of con- S* = (RL - Z)/(1 + r*) on conspicuouscon-
spicuous spending, the marginalcost is higher sumption, their bankruptcyconstraintdoes not
(and the benefit ratio lower) for type-H house- bind either. Hence, the casual observer would
holds.38Since, in this example, s* is indepen- conclude that bankruptcyconstraints are irrel-
dent of x, we abbreviate s*(x) as s*. evant. But this is certainly not the case-with-
out the bankruptcy constraint, Theorem 1
" This assumptionrequiressome justification. Between applies, and no Veblen effects emerge. The
the purchase of a conspicuous good and a declaration of key to this apparent puzzle is that type-L
bankruptcy,some or all of that good may be consumed or households are influenced by the fact that they
depreciated. Moreover, bankruptcycourts often allow in-
dividuals to retain expensive possessions (see, for exam-
ple, KirstinDowney, 1991; or RichardHylton, 1991). One
could, at the cost of introducing considerable analytic
complexity, instead assume that, in the event of bank- Since s*(x) = sj(X*7) = s* for i, j> 0, it does not
ruptcy, a household consumes some fraction X (0 < A < follow that the mutual nonimitation constraints hold
1) of its conspicuous possessions, while the remainder, strictly. Indeed, while each type i > 0 strictly prefers to
(1 - X)x,is repossessed. It does not appearthat this would differentiate itself from type 0, all other incentive com-
disturb the qualitative features of our analysis. patibility conditions hold with equality. If income was un-
38For s < (RL- Z)/(1 + r), the analysis is the same as certain as in Bagwell and Bernheim (1992), s (x4) would
for the example in Section I.F. For s > (RL- Z)/(1 + r), vary with i, and the nonimitation constraints between i,
the benefit ratio of the type-L household is infinite. j > 0 would hold strictly.
would confront a binding bankruptcy con- version introduced in our discussion of mul-
straint if they attempted to imitate the type-H tiple types).
households. Furthermore, with the introduc- Example 3: Intrinsic Utility. Suppose that
tion of income uncertainty,default does occur utility is given by the expression
with positive probability, while the other
features of the equilibrium remain qualita- (18) Ui(x, z, p) = Oiu(x) + v(z) + w(p)
tively unchanged (see Bagwell and Bernheim,
1992). This is consistent with evidence that for i = L, H, where OL > OH > 0. In other
many wealthy individuals are at significant words, assume that type-L households receive
risk of bankruptcy(see, for example, Michael greaterintrinsic utility from consuming luxury
Allen, 1991), and that this risk is often asso- goods, perhaps because of "novelty" value,
ciated with the acquisition of costly, conspic- or because of personal satisfaction derived
uous possessions." from "acting rich." Assume also that the
Example 2: Taxation. Suppose that, as in household faces the budget constraintz = R -
Example 1, utility is additively separable.Sup- s. For simplicity, assume that u (x) = x. It is
pose also that some conspicuous expenditures then possible to verify that, under appropriate
(such as lease payments on luxury automo- choices of cF and s-, the tangency property
biles, or mortgage payments on expensive holds if v'(RH)/V' (RL.) < OH/9L, and if
homes) are deductible expenses for the pur- V'(RH-- s)/v'(RL- s) declines with S.41 This
pose of computing personal income taxes, and final property requires v"' < 0 over the rele-
that marginal personal income tax rates de- vant range.
cline with taxable income over some range
(we note that the U.S. tax code has satisfied B. WhySignal with Prices?
this condition since 1986 due to the "take
back" of personal exemptions and itemized An implicit assumption in the analysis of
deductions). More specifically, suppose that Sections II and III.A concerns the resolution
taxable income is given by Y = R - s, that the of indeterminacy. It is well recognized that
first Y dollars of taxable income are taxed at equilibria such as those considered in this pa-
the rate Tr, and that additionaltaxable income per are characterized by "money burning,"
is taxed at the rate T2, with TI > T2. Assume where an action imposes the same cost on all
furtherthat RL> Y. Then agents regardless of type (see Paul Milgrom
and John Roberts [1986] or the literature on
(17) y(s, R) = ( - -ri)(R - s) signaling in financial markets). Agents must
waste a certain amount of money to sustain the
+ (T1 -2)max{O,R-s- Y}. equilibrium, but, as a formal matter, they are
indifferent between overpaying for conspicu-
In this case, s * = RL - Y (which is indepen- ous goods and other dissipative activities, such
dent of x, as in Example 1). It is easily verified as literally burning money. We recognize that
that, as long as either the curvatureof v is not individuals throw money away in a variety of
too great, or RH- RLis relatively small, then, forms-witness, for example, the phenome-
under appropriatechoices of cF and s-, the tan- non of heavy tipping by "high rollers." Even
gency property holds (including the stronger so, we would argue that, in practice, most
methods of burning money are inferior to the
conspicuous consumption of expensive, dura-
ble goods.
4 One reporterrecently summarizedthis relation as fol- To signal wealth effectively, the act of burn-
lows: "In the 1980s, people lived out their materialistic ing money must be observed readily by large
dreams, overspending on BMWs, huge boats, Caribbean numbers of people (even if these people make
vacations and dream condos, bankruptcy lawyers say.
Then came the real estate crash and the job layoffs. Now,
lawyers say, their clients are using their credit cards for
basic necessities, including food and children's clothes" " This follows from the fact that the slope of a type-i
(BarbaraCarton, 1991 p. 41). indifference curve is given by /ilv'(Ri- s).
little or no attempt to observe it), and it must remove price tags from their conspicuous pos-
be interpreted as evidence of substantial re- sessions. Obviously there are other important
source dissipation. Thus, one can certainly de- considerations that influence the choice of a
stroy substantial resources by burning dollar signal; completely transparentexhibitions of
bills on one's front lawn, but relatively few wealth seem socially unacceptable. An exam-
people will observe this activity since it is both ple of these kinds of concerns has been de-
brief and localized. It is also possible to waste scribed by TheEconomist (1993 p. 98): "Men
substantialresources in numerous small incre- flocked to the likes of Patek Philippe and Ebel,
ments during the normal course of social in- apparentlybecause watches are the one sort of
teraction, for example by tipping excessively luxury good that has what the marketers at
at every opportunity.However, since other in- Cartier smilingly describe as a 'functional
dividuals observe this behavioronly at isolated alibi'.'
moments, tipping serves as an effective signal
only if others believe that current behavior is C. The Observability of Price Concessions
representative. In contrast, expensive durable
goods, including one's automobiles, jewelry, Another assumption implicit in the analysis
and clothing are all observed regularly by of Sections II and III.A is that firms cannot
numerous other individuals during the nor- make a secret price concession to any given
mal course of social interaction, and pro- buyer. If secret concessions are possible, then
vide durable emblems of substantial resource the equilibrium described in Section III.A will
dissipation.42 break down: type-H agents prefer to buy the
The predisposition for signaling wealth conspicuous luxury good at a lower price, as
through the purchase of expensive, conspicu- long as they still get credit for purchasingit at
ous, durable goods does not completely re- the higher price.43
solve the issue of indeterminacy. While we There are many possible solutions to this
have focused on the case of a single conspic- problem. Each luxury brand producer clearly
uous good (with many brands), the extension has an incentive to commit himself to a policy
of our analysis to an arbitrarynumber of con- of making no secret concessions. Indeed,
spicuous goods is straightforwardand yields the argument in the preceding paragraphim-
essentially identical conclusions, except that plies that the signaling value of conspicuous
the distributionof demand across conspicuous consumption is present only if the producer
goods is indeterminant.Thus, our theory does has made a credible commitment of this sort.
not explain which durable,conspicuous goods One approachis to rely on intermediaries.By
households will choose as signals. Given the selling productsto intermediaries(for example,
diversity of actual behavior, we regard this as car dealerships) at publicly observable prices
a virtue, ratherthan a difficulty. that exceed marginal cost, the manufacturer
One might also wonder why households do places a lower bound on secret price conces-
not simply publish tax returnsor audited asset sions (equilibrium prices cannot be less than
statements. If one takes our theory literally, it "dealer invoice"). Another possibility is that
is also difficult to understandwhy consumers manufacturerswill rely on reputations.Once a
luxury brand acquires a reputation for being
sold at heavy discounts, the "snob value" as-
sociated with its purchasemay be eroded.
42The observation that agents can burn money in a va-
In practice, manufacturersof luxury goods
riety of different ways is troublesome in other contexts,
such as financial market signaling. Since investors have
do indeed adopt these kinds of mechanisms in
strong incentives to learn all relevant information about order to maintain high prices. According to
stocks, the firm need not undertakedissipative activities The Economist (1993 p. 98):
that investors will observe with little or no effort, nor need
they seek durable emblems of dissipation (since potential
investors can always review financial records). Thus, one
can entertain general doubts about models with money 43 This motivates manufacturersto thwart the distribu-
burning without being skeptical about the specific model tion of imitations of their products, such as fake Rolex
examined in this paper. watches and Polo shirts.
Recently a fight broke out between big eral luxury taxes created by the Omnibus
perfume houses and a clutch of British Budget Reconciliation Act of 1990 (OBRA).
retailerswho, having bought such brands In particular, this Act imposed a 10 percent
as Chanel No. 5 on the "grey market," excise tax ( t = 0.1 ) on the portion of the retail
were then reselling them for 60% below price of certain items that exceeds a product-
the recommended price. For some com-
panies, like Vuitton, the solution is to re- specific threshold (k). The thresholds were:
strict sales outside their own boutiques. $30,000 for automobiles, $100,000 for boats
Those that use specialist distributors and yachts, $250,000 for aircraft,and $10,000
must monitor them closely. Cartier has for jewelry and furs.45
one person in Paris whose sole respon- It is worth emphasizing that traditional
sibility is to keep tabs on its watches modes of analysis would produce highly mis-
once they leave the workshop. leading conclusions within the currentcontext.
Suppose that the inconspicuous good, z, is
V. PolicyImplicationsand Conclusions nontaxable. Assuming that the planner's ob-
jective is efficiency, rather than distribution,
The existence of Veblen effects in the con- the preceding analysis implies that only luxury
text of our model has some provocative im- brands should be taxed (as long as revenue
plications for public policy. Since supranormal requirements are not too high). In contrast,
profits result from the characteristics of de- since the demand for luxury brands in our
mand rather than from the nature of strategic model is highly price elastic, the application
interaction among firms, evidence of high of traditional Ramsey-style optimal tax for-
profitability does not necessarily support in- mulas would suggest that the government
ferences of either collusion or oligopolistic should raise a significant fraction of its reve-
forbearance. nue by taxing budget brands.46
The implications for tax policy are equally We mentioned in Section IV.B that, in a
controversial. Within our model, the equilib- model with many conspicuous goods, the dis-
rium prices of luxury brands are demand tributionof demand across conspicuous goods
driven, ratherthan supply driven-that is, lux- is indeterminant. This implies that the re-
ury brandsare sold at the consumer's preferred sponse to an increase in the luxury tax rate on
price, which is tax inclusive, and does not vary some specific conspicuous good is also inde-
with the tax rate. Thus, as long as the tax per terminant. It is possible, for example, that
unit does not exceed the difference between prices could simply adjust so that profits shift
the consumer's preferred price and marginal to a more lightly taxed industry (with side-
cost, and as long as the tax does not fall on payments among firms, one might even expect
budget brands, an excise tax on conspicuous this to occur). Thus, there may be advantages
goods amounts to a nondistortionary tax on to adopting a reasonably broad-based luxury
pure profits. tax, such as the one originally envisioned in
Consider, for example, tax-price relations of OBRA.
the form r(p) = max { 0, t(p - k)}, where the Our results on luxury taxation stand in sharp
parameterst and k satisfy 0 < t < 1 and k- contrast to those of Ireland (1992) and Ng
c(qF) (so that, in equilibrium, the tax falls (1987). Ireland considers a signaling model
only on luxury brands).' It is easy to verify of conspicuous consumption that satisfies the
that equilibrium quantities and prices are in- single-crossing property (and which, there-
variant with respect to t, and that the tax is fore, does not give rise to Veblen effects). In
therefore equivalent to a nondistortionarylevy his model, a tax on luxury goods does dis-
on pure profits. This family of tax schedules is
of particularinterest, since it includes the fed-
45 The 1994 threshold for automobiles was $32,000,
while the luxury tax on all other goods has been repealed
'The analysisproceedssimilarlyas long as r(c(qF)) = effective January 1993.
O,p > c(qF) + r(p) forp slightly greaterthancF as well as 46 See Bagwell and Bernheim (1992) for a detailed
forsomep 2 p*A(qF), andr(p) 2 0 foral p. derivation.
tort behavior. However, since equilibrium SH, PH) in particularto their equilibrium out-
involves excessive consumption of the con- come, (XH, SH, PH). Thus, the candidate equi-
spicuous good, the tax may actually be wel- librium fails the intuitive criterion. This
fare improving. Ng studies optimal taxation of establishes that (Al) must hold.
a special class of commodities, which he labels Now we define x as the maximum value of
"diamondgoods," for which consumers' pref- X E [XH, TIp] satisfying
erences are defined over the amount of money
spent on acquisition, ratherthan over amounts (A2) WL(X, PX, PH) = W L (P)-
consumed. Since a change in the price of a di-
amond good does not alter the utility received Step 2: x is well defined,x > XH,andpS <
by consumers, the optimal rate of taxation for T. First, note that sT2 SH> PXH,SOXH< slp,
a diamond good is infinite. The contrast be- implying thatthe feasible set is nonempty.Next,
tween our results and those of Ireland and Ng note that
therefore reveals the sensitivity of policy im-
plications to the particularformulation of the (A3) WL(XH, PXH, PH)
demand for luxury goods.
> WL(XH, SH, PH) = W L (P)
APPENDIX
and
PROOF OF THEOREM 1:
To prove the first part of this theorem, we (A4) WL(T/P, S, PH) < W L(P)
begin by applying Lemma 1, which tells us
that all type-L households purchase X*(p) (where the last inequality follows from (8)).
quality-weighted units at p. Next, we claim (A3), (A4), continuity, and the intermediate
that type-H households will also purchase all value theorem give us existence of S. x * XH
conspicuous units at the quality-weighted (and therefore x > XH)follows from (A3).
price p. We establish this claim by supposing Finally, since x < slp (x * slp follows from
that, on the contrary,SH/XH > p, and inducing (A4)), px < s.
a contradiction. The argument leading to the Step 3: WH(X, PX, PH) > WH(XH, SH, PH).
contradictionconsists of four steps. This follows from the single-crossing prop-
Step 1: The incentive compatibility con- erty, since x > XH and WL(X, PX, PH) =
straintbinds, that is, W L (P) = WL(XH, SH, PH) (which in turn im-
plies pS > SH, since x > XH).
(Al ) WL(XH, SH, PH) = WL(P)
Step 4: There exists xH, sH with sH =
pxH such that WL(xH, SH, PH) < WL(p) and
Suppose on the contrary that this is not the WH(XH,SH, PH) > WH(XH,SH,PH). To see this,
case. Then the right-hand side of (Al) must let xH = X + s and sH = pxH. For s > 0
exceed the left-hand side (otherwise incentive sufficiently small, (xH, sH) necessarily satisfies
compatibility would be violated). Consider the requiredproperties(recall that x is defined
(x H, s H), with sH slightly less than SH, and as the largest value of x satisfying (A2)).
XH = XH, such that sH/Hx' p, and such that Now we assertthat the candidateequilibrium
incentive compatibility still holds with strict fails the intuitive criterion. This follows from
inequality. Then type-L households prefer step 4 through an argumentcompletely analo-
their equilibrium payoff, W * (p), to (xH, sH, gous to thatgiven in step 1 above. Thus,we have
p) for all p E [PL, PHI In contrast, there are the desired contradiction,and we conclude that
somevaluesof p E [PL, PHI (specifically,PH) type-H households must purchaseall conspicu-
for which type-H households prefer (xH, sH, ous units at the quality-weightedprice p.
p) to their equilibrium outcome, (XH, SH, PH) The second part of the theorem is an im-
Thus, by the logic of the intuitive criterion, mediate corollary of the first part. The free
social contacts must infer R = RH upon ob- entry assumption, combined with the usual
serving (xH, sH), and respond by selecting Bertrand-style argument, implies that some
P = PH. But type-H households prefer (xH, firms must, in equilibrium, choose quality qF
and sell their products at price c(qF) = choose this alternative regardless of social
i(qF)cF. Clearly, only conspicuous goods contacts' inferences, while type-H households
of quality qF will be available at quality- would choose this alternative for some infer-
weighted price CF, so only these will be ences, for example, PH) . Consequently, if this
produced. alternative is offered, type-H households will
select it. Knowing this, it is in the interest of
PROOF OF THEOREM 2: an entrant to offer (q', pH), which under-
We begin by establishing that, in equilib- mines the equilibrium.
rium, some firms offer, and type-L households Now define qcas the maximum value of q E
purchase, a conspicuous product of quality qL [qH, q-] satisfying
at price C(qL). Supposeon the contrarythat
(qL, PL) * (4L,
C(qL)). Since firms cannot (A7) WL(j(q), c(q), PH) = WL
lose money in equilibrium, either qL * qLand
PL 2 c(qL), or qL = qL andPL > c(qL). In Step 2: qcis well-defined, and q-> q?> qH
either case, Since, by supposition, PH > c(qH),
(A5) WL(j(qL), pL, PL) (A8) WL(j(qH), c(qH), PH)
< C(qL) + ?, PL) > WL(f(qH), PH, PH) = WL

WL(8(L),
for some small s. Suppose then that an entrant Existence of qcfollows from (A8), (12), con-
offered qLat price C(qL) + s. For any house- tinuity, and the intermediate value theorem.
hold electing the entrant'soffering, social con- q * qH (and therefore q > qH) follows from
tacts must respond with p 2 PL. Therefore, by (A8). qc * q- (and therefore qc < q) follows
(A5), all type-L households will purchase the from (8).
conspicuous good from the entrant. This im- Step 3: WH (I(j ), C ( ), PH) > WH(/(qH),
plies that the entrantwould earn a profit,which PH, PH). This follows from the single-crossing
contradictsthe supposition that an equilibrium property, since 8t(qc) > f(qH) and WL(b(qc),
prevails. c(q), PH) = WL = WL(/(qH), PH, PH)
Now we establish thatPH = c (qH). Suppose Step 4: There exists q H, P'Hwith pH >
on the contrary that PH > c(qH) (this is the c(q ) such that WL(/.t(qH), pH PH) < WL and
only other possibility, since, in equilibrium, WH(y(qH), pH, PH) > WH(/(qH), PH, PH). To
price cannot be less than marginal cost). We see this, simply take q' slightly larger than qj,
induce a contradiction through four steps (the and pH slightly greater than c (qH)
argumenthere is analogous, but not identical, The intuitive criterion implies that, if (qH,
to that used in the Proof of Theorem 1). pH ) was offered by some new entrant, and if
Step 1: The incentive compatibility con- a household selected it, social contacts would
straintbinds, that is infer that the deviating household was of type
H, and respond with PH (type-L households
(A6) WL(/(qH), PH, PH) = WL would not choose this alternativeregardless of
social contacts' inferences, while type-H
Suppose on the contrary that this is not the households would choose this alternative for
case. Then the right-hand side of (A6) must some inferences, for example, PH). Conse-
exceed the left-hand side. Consider (qH, quently, if this alternativewas offered, type-H
pH,),withq = qHandpH = PH - e for some households would select it. Knowing this, it is
small s, such that the incentive constraint still in the interest of an entrant to offer (qH,
holds with strict inequality. The intuitive cri- pH), which underminesthe equilibrium. Thus,
terion implies that, if this alternative was of- we have the desired contradiction.
fered by some new entrant,and if a household
selected it, social contacts would infer that the PROOF OF THEOREM 3:
deviating household was of type H, and re- Using the same arguments given in the
spond with PH (type-L households would not proof of Theorem 1, it is easy to show that, in
a separating equilibrium, (i) type-L house- some values of p E [PL, PH] (for example PH)
holds purchase X*(p) quality-weighted units for which type-H households prefer (xH, sH,
of the conspicuous good at price p (the lowest p) to their equilibrium outcome, (XH, SH, PH)
quality-weighted price announced in the first Thus, by the logic of the intuitive criterion,
phase), and (ii) p = cF. We prove the rest of social contacts must infer R = RH upon ob-
the theorem through the following five steps. serving (xH, SH), and respond by selecting
Step 1: If p = CF, then the intuitivecri- P = PH. But type-H households prefer (xH,
terion requires that, in a separating equilib- SH, PH) in particularto their equilibrium out-
rium, the type-H choice (XH, SH) must solve come, (XH, SH, PH). Thus, the candidate equi-
librium fails the intuitive criterion.
(A9) max WH(X,S, PH) Step 2: If p- 2 p*, then (x*, s*(x*)) is
x's
the unique solution to maximizing (A9) sub-
subject to ject to (AIO) and (All). By the tangency
property, it is immediate that (x*, s*(x*)) is
the unique solution to maximizing (A9) sub-
(AIO) CF c- c p ject to (All). Since it is assumed that p- 2
x
p* > cF, (x*, s*(x*)) satisfies (AIO), and
and therefore continues to be the unique solution
to maximizing (A9) subject to both (AlO) and
(Al1l) WL(X, S, PH) WL(CF).
L (A1l).
Step 3: (XL(CF), CFXL(CF), X*, S*(X*))
That is, type-H households maximize their satisfies incentive compatibility (expres-
utility, buying from a combination of different sions (4) and (5)). Note that, with (XL, SL) =
firms offering prices between CF and -, subject (L(C), C XL(C )), expression (4) is
to the constraint that the type-L households equivalent to (Al l), and expression, (5 ) is
must choose not to mimic them. equivalent to
Suppose on the contrary that the type-H
choice, (XH, SH), does not solve this problem. (A14) WH(XH, SH, PH)
Then there is some feasible (x, s) such that
WH(X, S, PH) > WH(XH, SH, PH) and WL(X, s, 2 WH(XL(C ), C cXL(C), PL)
PH) < W L(cF). This in turnimplies that there
is some feasible (xH, sH) such that (x*, s*(x*)) satisfies (All) by the definition
of x*. Assume (contraryto the claim) that (x*,
(A12) WH(XH,SH, PH) > WH(XH, SH, PH) s*(x*)) violates expression (A14). Since
(x*, s*(x*)) maximizes (A9) subject to
and (Al l), it is necessarily the case that (A14) is
violated for all (XH, SH) satisfying (Al l). But
(A13) WL(XH,SH',PH) < W L (CF).
this contradicts the assumption that separation
is feasible.
To see this, suppose first that slx < -. Then Step 4: In equilibrium, p- 2 p *. Suppose
we can simply take xH = x and sH = s + s for that, on the contrary,p- < p *. Then an entrant
some small s > 0. If, on the other hand, could name qF and price p*,I(qF), implying a
slx = p-, let xH= x + s andsH = P-XIfor some quantity-weighted price of p *. Our previous
small s > 0; since p- 2 p01(q0) (where p0 analysis (steps 1, 2, and 3) implies that all
was defined as the prohibitive price at which type-H households would purchase exactly
a conspicuous good of quality q? is available X*/ja(qF) units from this entrant,who would
from some alternative source), the desired earn profits of x*(p* - cF) > 0. This con-
conclusion follows. Now we argue that (XH, tradicts the assumption that an equilibrium
SH) cannot satisfy the intuitive criterion. By prevails.
(A13), type-L households prefer their equilib- Step 5: In equilibrium, the purchases of
rium payoff, W*(CF), to (XI, SI, p) for all the type-H households are divided equally be-
P E [PL, PH]. In contrast, by (A12), there are tween the F incumbents, each of which names
quality qF and price p *s (qF). To demonstrate Allen, Michael. "Personal Bankruptcy Claims
this result, we exploit our assumptions con- New Victims: Wealthier Americans." Wall
cerning the resolution of household indiffer- Street Journal, 26 June 1991, p. Al.
ence. First, we argue that at least one firm must Bagwell,LaurieSimonand Bernheim,B. Douglas.
quote the quality-weighted price p *. Suppose "Conspicuous Consumption, Pure Prof-
on the contrary that no firm quotes this price. its, and the Luxury Tax." National Bu-
Since each household resolves indifference in reau of Economic Research (Cambridge,
favor of making all purchases from a single MA) Working Paper No. 4163, Septem-
firm, an entrant could profitably undertake a ber 1992.
deviation to quality qF and price p*bt(qF). Banks,JeffreyS. and Sobel,Joel. "Equilibrium
Second, we argue that at least one incumbent Selection in Signaling Games." Economet-
must name p *. Suppose on the contrary that rica, May 1987, 55(3), pp. 647-61.
no incumbent names p *. Then, by the preced- Basmann,R. L.; Molina,D. J. and Slottje,D. J.
ing argument, at least one entrant must quote "A Note on Measuring Veblen's Theory
p *. Since indifference is resolved in favor of of Conspicuous Consumption." Review of
making all purchases from a single firm, in- Economics and Statistics, August 1988,
cumbents make no sales and earn zero profits. 70(3), pp. 535-40.
Since indifference is also resolved in favor Basu, Kaushik. "Monopoly, Quality Uncer-
of incumbents, any incumbent could earn tainty and 'Status' Goods." International
x*(p* - cF) > 0 by deviating to quality qF Journal of Industrial Organization, Decem-
and pricep*lu(qF), which establishesa con- ber 1987, 5(4), pp. 435-46.
tradiction. It follows that type-H households Bernheim,B. Douglas."Tax Policy and the Div-
only purchase the conspicuous good from in- idend Puzzle." Rand Journal of Economics,
cumbents. Third, we argue that all incumbents Winter 1991, 22(4), pp. 455-76.
must quote p*. Suppose on the contrary that . "A Theory of Conformity." Jour-
N < F incumbents quote p*. Since type-H nal of Political Economy, October 1994,
households only purchase conspicuous output 102(5), pp. 841-77.
at the quality-weighted price p *, no incumbent Besley, Timothy and Coate, Stephen. "Under-
earns a profit unless it quotes p *. Suppose that standing Welfare Stigma: Resentment and
the (N + 1)th incumbent deviates to quality Statistical Discrimination." Mimeo, Prince-
qF and price p*It(qF). Type-H households ton University, 1990.
will be indifferentbetween purchasingtheirx * Braun, Ottmar L. and Wicklund, Robert A.
quality-weighted units from this incumbent "Psychological Antecedents of Conspicu-
and from the other N incumbents quoting the ous Consumption." Journal of Economic
quality-weighted price p*. Demand would Psychology, June 1989,10(2), pp. 161-87.
then be split evenly between these incumbents, Carton,Barbara."BankruptcyGets Personal."
and the deviator would earn x*(p* - CF)/ Boston Globe, 28 August 1991, p. 41.
(N + 1) > 0. Fourthand finally, we argue that Cho, In-Koo and Kreps, David M. "Signaling
each incumbent chooses quality qF. This fol- Games and Stable Equilibria." Quarterly
lows from the fact that incumbent f earns Journal of Economics, May 1987, 102 (2),
x*(p* - c(qf)/A(qf))/F, which is maxi- pp. 179-221.
mized at qf = qF* Cole,HaroldL.;Mailath,GeorgeJ. andPostelwaite,
Andrew. "Social Norms, Saving Behavior,
REFERENCES and Growth." Journal of Political Economy,
December 1992, 100(6), pp. 1092-125.
Aeppel,Timothy."BMW, Despite Success, Is Creedy,John and Slottje, D. J. "Conspicuous
Acting Like It's Under Siege." Wall Street Consumptionin Australia." Research Paper
Journal, 19 February 1992, p. B4. Number 307, University of Melbourne,June
Akerlof,GeorgeA. "A Theory of Social Cus- 1991.
tom, of which Unemployment May Be One Davidson,Carland Deneckere,Raymond."Long-
Consequence." Quarterly Journal of Eco- Run Competition in Capacity, Short-Run
nomics, June 1980, 94(4), pp. 749-75. Competition in Price, and the Cournot
Model." Rand Journal of Economics, Au- Kreps, David M. and Scheinkman, Jose. "Quan-
tumn 1986, 17(3), pp. 404-15. tity Precommitment and Bertrand Com-
Downey, Kirstin. "Antonelli's Lifestyle Sur- petition Yield Cournot Outcomes." Bell
vives Bankruptcy." Washington Post, 14 Journal of Economics, Autumn 1983,
December 1991, p. A12. 14(2), pp. 326-37.
TheEconomist."The Luxury-Goods Trade." 8 Leibenstein,Harvey. "Bandwagon, Snob, and
January 1993, pp. 95-98. Veblen Effects in the Theory of Consumers'
Frank,RobertH. "The Demand for Unobserv- Demand." Quarterly Journal of Econom-
able and Other Nonpositional Goods." ics, May 1950, 64(2), pp. 183-207.
American Economic Review, March 1985, Milgrom, Paul and Roberts, John. "Prices and
75(1), pp. 101-16. Advertising Signals of Product Quality."
Fershtman,Chaim and Weiss, Yoram. "Social Journal of Political Economy, August
Status,Cultureand Economic Performance." 1986, 94(4), pp. 796-821.
Mimeo, Tel-Aviv University, 1992. Ng, Yew-Kwang."Diamonds Are a Govern-
Glazer, Amihai and Konrad, Kai A. "A Sig- ment's Best Friend: Burden-Free Taxes on
nalling Explanation for Private Charity." Goods Valued for their Values." American
Irvine Economics Paper No. 90-92-35, Economic Review, March 1987, 77(1), pp.
University of California, Irvine, August 186-91.
1992. Pesendorfer,Wolfgang."Design Innovationand
Hylton,Richard."Sometimes Those Bare Ne- Fashion Cycles." American Economic Re-
cessities Include Yacht Docking." New view, September 1995, 85(4), pp. 771-92.
York Times, 25 August 1991, Section 4, Schuon,Marshall."Twins, Triplets and Other
p. 5. Siblings." New YorkTimes, Sunday, 7 No-
Ireland,Norman."On Limiting the Market for vember 1993, Section 8, p. 20.
Status Signals." Mimeo, University of War- Shenon, Philip. "Irate Wife, Talky Mistress,
wick, February 1992. Shellshocked General." New YorkTimes, 1
Jones, StephenR. G. The economics of confor- August 1991, p. A4.
mism. Oxford: Basil Blackwell, 1984. Veblen, Thorstein. The theory of the leisure
Kreps, David M. A course in microeconomic class: An economic study of institutions.
theory. Princeton:PrincetonUniversityPress, London: Unwin Books, 1899; reprinted
1990. New York: Dover Publications, 1994.

Laurie Simon Bagwell, B. Douglas Bernheim

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Laurie Simon Bagwell, B. Douglas Bernheim

Uploaded by

Copyright:

Available Formats

American Economic Association

Veblen Effects in a Theory of Conspicuous Consumption

You may need to log in to JSTOR to access the linked references.

By LAURIE SIMON BAGWELL AND B. DOUGLAS BERNHEIM*

dividedinto two groups.The firstgroupcon- froman entrant.In the contextof conspicuous

form inferences about each householdi' s It is useful to describe the second-phase

andfeasibility, ample,s, the absolutelimiton conspicuousex-

(8) WL(S/P, S, PH) < WL(P).

Definition: Preferences satisfy the single-

(9) Wi (x, s, p) = u (x)

FIGURE 1. SINGLE CROSSING AND THE ABSENCE

celebrities who accumulate "stables" of ex- and

A proof of this result is contained in the Ap-

difference between conditions (12) and (8) described in

separationis feasible).28 In equilibrium, some

FIGURE 2. THE TANGENCY PROPERTY A. Analysis of Veblen Effects

the location of s*(x), it is certainly possible

in Figure 3. The following result demonstrates

THEOREM 3: Suppose that the tangency

SL) = (XL*(CF), CFXL*(CF)) and (XH, SH) =

returns to scale technology, and Bertrand- cific choice of qF is a somewhat special

A. Justificationsfor the Failure of the

enforceable claim against the household's S

(A5) WL(j(qL), pL, PL) (A8) WL(j(qH), c(qH), PH)

< C(qL) + ?, PL) > WL(f(qH), PH, PH) = WL

You might also like

SL) = (XL(CF), CFXL(CF)) and (XH, SH) =