THE JOURNALOF INDUSTRIALECONOMICS
VOLUMEXLIV
March 96
0021-1821
No. 1
CUSTOMER RETURN POLICIES FOR EXPERIENCE
GOODS*
YEoN-Koo CHE
This paper studies the economic rationalefor customer returnpolicies, by
focusing on the "experience goods" aspect of many products. Return
policies allow consumersto defer their purchasingdecisions until after they
gain some experiencewith goods. In so doing, they insureconsumersagainst
ex post loss, which allows a monopoly seller to chargemore than otherwise.
It is shown thatthe seller adoptsthe returnpolicy when consumersarehighly
risk averseor retailcosts are high. Consumersare strictlybetteroff underthe
returnpolicy, but there is too little adoptionof the policy in equilibrium.
I. INTRODUCTION
MOSTSTORES allow free return of merchandise. Many electronic discount stores
customarily allow return of their products within 30 days of purchase, and some
catalog clothing marketers, like L. L. Bean, even accept returns years after the
purchase. While retail superstores, such as Wal-Mart and Toys 'R' Us, are best
known for their generous return policies, the types of retailers adopting return
policies are much broader. According to one US survey, more than 95% of
retailers interviewed allow some form of returns.' Clearly, consumers value the
return policies. As many as 20% of personal computers sold to home buyers are
known to be returned, for example.2
The standard economic rationale for return policies is that of warranties.
Return policies insure customers against defective products, making risk-averse
customers willing to pay more for a product than if there is no return policy. This
warranty argument explains the use of return policies for products that are subject
to random malfunctions. But return policies are used for a wide variety of
products, for which defects are not an issue. Many return policies, for example,
do not require customers to provide evidence or an explanation regarding the
malfunction of the returned good. Rather, a customer's not liking a product is
* I am gratefulto SeverinBorenstein,JinsookCho, Ian Gale, Don Hausch,LawrenceWhite andtwo
anonymousreferees for helpful comments and suggestions.
1 "Small Store Survival," a recent study of Illinois retailers conducted by the Illinois Retail
MerchantsAssociation (IRMA) and ArthurAnderson & Co. reveals that, of the retailerssurveyed,
78% give cash refunds with a receipt, 59% give merchandise credit with a receipt, 44% give
merchandisecreditwithout a receipt, 32% give cash refundswithout a receipt. Twenty-threepercent
limit the returnperiod,while fewer than five percent say all sales are final. (See "Returnto Seller,"
Sales & MarketingManagement,August 1994, p. 21.)
2 See "CorporateFocus: PackardBell Prospersdespite PC IndustryShake-Up;ComputerMaker's
Secret of Success: Focusing on the ConsumerMarket,"WallStreetJournal, June 14, 1994.
? Blackwell PublishersLtd. 1996, 108 Cowley Road, Oxford OX4 1JF,UK and 238 Main Street,Cambridge,MA 02142, USA.
17
18
YEON-KOOCHE
often sufficientfor stores to accept the return.The "no-questions-asked"full refund
policy is customarywith many big retailers.
For these latterproducts,a more compelling rationalefor returnpolicies has to do
with their "experiencegood" nature:Customersdo not fully know theirpreferences
for the productsuntil afterthey gain some experiencewith them. Such an experience
good naturemay reflecta simple psychologicalreaction,like "buyer'sremorse,"but
it may also result from the fact that the purchaserof a good may not be its final
consumer.For example, a shirtbought for a family memberor as a gift may run into
the problems of wrong color, wrong size, or wrong style. Return policies allow
customers essentially to defer their purchasingdecisions until after gaining some
experiencewith the products.A consumerwho has learnedthat he does not like a
productcan nullify his purchaseby simply returningit.
This paper explores the consumer learning implications of return policies, by
developing a model where customers realize idiosyncraticvaluations of the good
aftertheirpurchase.The analysis focuses on the following trade-offfor a monopoly
seller: on the one hand, the returnpolicies insure consumers against ex post loss,
allowing the seller to charge a higher price than otherwise;on the other hand, the
seller can never induce consumersto buy at a price above their ex post valuations,
which she could do, for some consumers,with a no-returnpolicy. It is shown that
the returnpolicy is optimalif the consumersare sufficientlyrisk averseor retailcosts
are high. Superiorrisk sharingmakes consumersstrictlybetteroff underthe return
policy, but the seller'sfailureto internalizethis benefit leads to too little adoptionof
the returnpolicy in equilibrium,relativeto the socially efficient outcome.
I am not awareof any literatureon consumerreturnpolicies. Pasternack[1985]
and Marvel and Peck [1994] study the manufacturers'return policies toward
retailers,with a special focus on channel coordinationin the presence of aggregate
demand uncertainty. The focus on consumer learning and delayed purchase
distinguishesthe currentpaper.
The paper is organizedas follows. The model is describedin Section II. Section
III studies equilibriumadoptionof returnpolicies in the monopoly setting. Section
IV explores its welfare implications. Section V concludes.
H. MODEL
A monopoly seller (retailer)faces a unit mass of consumers,each of whom desires at
most one unit of a good. The seller is risk neutraland maximizes expected profits,
and she incursretail costs of C E [0, i) for each unit that she carries,which includes
the paymentsto a manufacturer.
The consumers'preferencesfor the good are unknownat the time of purchase,but
they are leamed after purchase. A consumer'spreference is parameterizedby a
"valuation" v that is drawn randomly from Lv,v'], 0 <v < v, by a distribution
function F(-), which has a well-defined positive density functionf(*). Note that
customersare ex ante identical,and that theirex post realized valuationsare purely
idiosyncratic. I consider a simple returnpolicy that provides cash refunds for a
(j
Blackwell Publshers Ltd. 1996
19
CUSTOMERRETURN POLICIESFOR EXPERIENCEGOODS
return and henceforth refer to it as "the returnpolicy." Under the returnpolicy,
consumerscan returnthe good after learningtheir valuations,at zero cost.3
The von-Neumannutility function when a consumer purchasesthe good at the
price of p and realizes v is U(v - p), where U(-) is strictly increasing, (weakly)
concave, and exhibits constantabsolute risk aversion(CARA). The utility from no
purchaseis normalizedto be zero: U(O)= 0. The degree of risk aversionis measured
by the certainty equivalent of the valuation, Vce (i.e., U(Vce)= E[U(v)]). When
consumersare risk neutral,Vce = E[v], while vce= v if they are infinitelyrisk averse.
III.
EQUILIBRIUM ADOPTION OF THE RETURN POLICY
I firstconsiderthe case where the seller adoptsthe no-returnpolicy. For any pricep,
a consumerpurchasesthe good if and only if E[U(v - p)] > 0. Given this, the seller
sets the highestp such that E[U(v - p)] > 0, if that price covers the retail costs, or
else the seller does not sell the good. Clearly, the optimal price, Pn, satisfies
E[U(v - pn)] = 0. The following lemma shows thatthe optimalprice is the certainty
equivalentof the valuation:
Lemma 1. With constant absolute risk aversion,Pn =
Vce.
= Ofor some real numberk.
Proof. I show thatE[U(v- Vce)] = kE[U(v)-U(Vce)]
The second equality holds by the definition of the certaintyequivalent. The first
equality is shown as follows. Let z U(v) - U(Vce). Then, U(v - Vce)=
(z) U(U (z + U(Vce)) - Vce). It suffices to show that 0(.) is linear, which
follows since +"(z) is proportionalto
[U"(U-1(z
[U'(U
1(z
+
U(Vce))
-
Vce)
U"(U-1(z
+
U(Vce)))]
+
U(Vce))
-
Vce)
U'(U
+
U(Vce)))
1(z
-O
(Q.E.D.)
-
It follows from Lemma 1 that, under the no-returnpolicy, the seller earns
(1)
7rn(Vce, c)
-max{vce
- c,
0).
Note that the seller's profit is inversely related to the degree of consumer risk
aversion.Under the no-returnpolicy, consumersbear the entirerisk associatedwith
their uncertainex post valuation.As risk aversion increases, the seller must lower
her price to compensateconsumersfor the risk.
Suppose now thatthe seller adoptsthe returnpolicy. Then, all the consumerswill
attemptto purchasethe good initially and decide whetherto returnit, after learning
theirvaluations.Given a price,p, a consumerwith ex post valuationv will returnthe
good if and only if v <p. Thus, only [1 - F(p)] consumerswill eventuallykeep the
good. Given this, the seller's optimal strategyis to hold an inventoryof precisely
q = 1 - F(p). This limited inventory leads to an initial rationing of some
consumers, but eventually all consumers whose valuations exceed p will obtain
3In practice, even a free returnpolicy requiresa trip to the store, so returninga good may not be
costless. A positive returncost does not alter the results of this paper qualitatively,however.
C) Blackwell Publshers
Ltd. 1996
YEON-KOOCHE
20
the good. Such a result will arise when the retaileradopts the policy of reselling
returned goods to other consumers, as many discount superstores do.4 Finally,
the seller sets p to maximize her profit: (p - c)[l - F(p)]. Assuming that
1 - F(v) - vf(v) is decreasingin v, there exists a unique optimalpricepr(c). Under
the returnpolicy, the seller thereforeearns
(2)
7rr(C)
(Pr(C) - c)[
-
F(pr(C))].
Inspectionof (2) reveals several featuresthat distinguishthe returnpolicy. First,
the seller's profit does not depend on the degree of consumer risk aversion. The
returnpolicy eliminatesa consumer'srisk of paying more thanhis ex post valuation,
so the marginal consumer is not adversely affected by risk aversion. Second, the
returnpolicy opens up a screening opportunity:the seller can charge a high price
and sell only to high-valuationconsumers.5Clearly,this opportunityis relatively
more valuablewhen the retail costs are high, since the seller can maintainher profit
margin by selling only to high-valuationconsumers. Such screening is impossible
under the no-returnpolicy
Intuition therefore suggests that consumer risk aversion and high retail costs
would favor the adoption of the returnpolicy, as is confirmedby the following
proposition:
Proposition 1. There exists c E (v, E[v]) such that the returnpolicy is optimal for
the seller, regardless of Vce, if and only if c > c. For c < c, there exists
v(c) E [max{V,c}, E[v]] such that the seller prefersthe returnpolicy if and only if
vce< v(c), where v(c) is non-increasingin c.
c) denote the profit difference between the
Let P(vce, c) _tr(C)-tn(Vce,
two regimes. T(-, *) is continuous.6 Furthermore,T(-, c) is decreasing for all
c < vce, and P(vce, ) is non-decreasingfor all vce > c. To see the latter, use the
Envelope theoremto get
Proof.
W2(Vce, c) =
-[1 -F(pr(C))
The first statementfollows since
T(E[v],
= (PrO-[1
> 0
-(-1)
P(Vce,
> 0 for all Vce(< E[v]), and
E[v]) = 7rr(E[v])
-
-F(pr(V)]
rV
=
-lJJV
JPr(i)JV
for all vce > c.
-pr(IrdF(V
)
-
{E[v]
-
v}
d
r~~~~Pro
V
J[v-dF(v)
< 0.
Clearly,c must satisfy P(E[v], c) = 0. For any c < c, define v(c) E [max{v, c}, E[v]]
to satisfy T(v(c), c) = 0; v(c) exists since T(max{v, c}, c) > 0 for all c > 0,
4 Such a practiceis common in many stores, which use mark-downprices for returnedgoods. Some
retailersappearto go furtherby selling them as new merchandise.Recently, Wal-MartStores Inc. and
Toys 'R' Us Inc. were accused to routinely misrepresentand sell as new merchandisethat has been
returned,and sometimesdamaged(see "TwoMajorChainsareAccused in Suit of Sales Deception," Wall
StreetJournal, April 30, 1993).
5 This opportunityis absent under the no-returnpolicy, since consumersare ex ante identical.
6 7r(*) is continuousby the Theoremof the Maximum(Debreu [1959]).
C) Blackwell PublishersLtd. 1996
CUSTOMERRETURN POLICIESFOR EXPERIENCEGOODS
21
T(E[v], c) < 0 for all c < c, and T(., c) is continuous.It is also unique since T(., c)
is strictlydecreasing.The second statementthen follows from the monotonicityof
(Vce,
Q.E.D.
0).
The propositionimplies, in particular,that the returnpolicy can never be optimal
if consumersare risk neutraland retail costs are sufficientlysmall. The intuitionfor
this result is clear. The no-returnpolicy essentially implements the outcome of
"selling the firm to the agent," well known in the principal-agentliterature.7The
seller (principal)transfersthe entire risk to the consumers(agent) and by doing so
generatesthe highest expectedprofitsif consumersare risk neutral.By contrast,the
returnpolicy eliminatesthe consumers'downside risk, which means that the seller
does not extractthe full consumer surplus.8As the consumersbecome more risk
averse, the no-returnpolicy becomes less attractive,however,since the seller must
lower her price to compensate consumers for the risk. Such risk compensationis
unnecessary under the returnpolicy. Likewise, the presence of high retail costs
favorsthe returnpolicy, relatively:the seller can protecther profitmarginby selling
only to high-valuationconsumersunderthe returnpolicy, whereasthe seller has no
such option under the no-returnpolicy.
IV WELFAREIMPLICATIONSOF THE RETURN POLICY
The analysis in the previous section yields several welfare implications. First, the
return policy leads to better risk-sharing between the seller and risk-averse
consumers, by eliminating the downside risk of the consumers.9 This effect
unambiguouslybenefitsthe consumers.Second,the returnpolicy typicallyresultsin
screeningsome low-valuationconsumers.The welfare implicationsof this effect are
ambiguous.If retailcosts are small, then the screeningwill resultin the loss of some
consumersurplus.If retail costs are large, however,the screening of low-valuation
consumers will result in an efficiency gain, since the return policy will simply
reallocate a good from a consumer with low valuation to a consumer with high
valuation,above the retail costs.
The combined effect is unambiguously positive, at least for the consumers.
Consumersare alwaysbetteroff when the seller adoptsthe returnpolicy. Under the
returnpolicy, the consumers are protected from any loss, so they receive strictly
positive expected utility.10By contrast,the consumersreceive zero expected utility
under the no-returnpolicy (see Lemma 1).
7 I thank a referee for suggesting this interpretation.
8 The consumer learning requiresindividualrationalityto be met in ex post terms, which imposes a
strongerconstrainton the seller than ex ante individualrationalitythat holds underthe no-returnpolicy.
See Maskinand Tirole [1990] and Gale and Holmes [1993] for furtherdiscussion on the effect of learning
on the agent's individualrationalityconstraint.
9 The return policy does not eliminate the upside risk of the consumers. If pr(c) = v, there is no
downside risk to share, so the returnand no-returnpolicies are welfare-equivalent.
10Underthe returnpolicy, consumersreceive E[U(max{v - pr(c), 0})]), which is strictlypositive since
Pr(C)
?
< v.
Blackwell PublishersLtd. 1996
22
YEON-KOOCHE
The consumers'strict preferencefor the returnpolicy means, however, that the
seller does not fully internalizethe social gains from adoptingthe returnpolicy. The
consequence of the latter is too little adoption of the returnpolicy in equilibrium,
relative to the socially efficient outcome. The next proposition formalizes this
observation,using the total certaintyequivalentof the seller- and consumer-surplus
as a welfare criterion.11It shows that the returnpolicy dominates the no-return
policy in welfare for a larger set of parametervalues than in profit.
Proposition2. The returnpolicy is socially desirablewheneverthe seller adoptsit.
The converse may not hold. Formally, the return policy is socially desirable,
regardless of
vce,
for c > c for some c < c; and, for c < c, it is desirable if Vce<iv(c)
for some v(c) > (C).12
Proof. Under the no-returnpolicy, the total certainty equivalent is simply the
seller's profit: Wn(Vce,c) 3tn(Vce, C). Under the return policy, the total certainty
equivalent is Wr(Vce,c) U- 1(E{ U(max{v - pr(c), O})) + r(C). The first term of
WAVce, c) is strictly positive (since pAc) < v), and it is continuous in (Vce,c).
Therefore, Wr(Vce, c) > Wn(Vce,c) whenever irAc)> ln(Vce, c). The proof is completed by noting that WAVce, c) - Wn(Vce,c) is continuous in (Vce,c). 13
Q.E.D.
V
CONCLUSION
In this paper, I have studied the economic rationalefor a consumer returnpolicy,
from the perspectiveof consumerlearningaboutgoods. The resultsof this paperdo
not just applyto the returnpolicy for experiencegoods, but, to a limited degree,they
also apply to various other returnpolicies and other retail practices that promote
customerlearningof products.Widely used retailpracticessuch as the money-back
guarantee and limited-periodfree trial of a new products can promote customer
learningof the productsat low risk and can play a role similarto that of the return
policy studied in this paper.14Likewise, part of the rationale for the in-store
customer service and, to some extent, call-in technical support for products like
personal computerscan be understoodin a similarvein.
1"
Expectedtotal surplusis not an appropriatewelfare criterion,since it does not capturethe consumer
risk aversion. Since the CARA utility function displays no wealth effects, however, the certainty
equivalentsof the total surpluscan measurethe level of welfare.This criterionis used by Holmstromand
Milgrom [1990, 1991] in their principal-agentmodels.
12 The propositiondoes not containthe "only if" part, since the social welfare differencebetween the
two regimes cannot generallybe shown to be monotonic in (- vc, c).
13 Continuityof W(vce,c) in vce is not immediate since vce affects WX(ve, c) indirectlythrough the
curvature of U(s). With CARA utility functions, vce uniquely determines U(s), through the risk
coefficient. Continuitythen holds since both vce and U(s) vary continuouslywith the risk coefficient.
14 The money-backguaranteeand free trial offer are common for many newly-introducedgoods and
services. For example, many credit card companies make such offers when introducingtravel agency,
diner'sclub and catalogueshoppingservices. Free trialoffer is also used for new machines.For example,
PerSeptive,a biotech lab instrumentmaker,used a sales strategy:Pay nothing for it now and returnit later
if you don't like it (see "Biotechnology: Biotech Company is Questioned about 'Try It Out' Sales
Strategy,"WallStreetJournal, November 8, 1994).
?
Blackwell PubhshersLtd. 1996
CUSTOMERRETURN POLICIESFOR EXPERIENCEGOODS
23
This paper represents an exploratory study of a largely unexplored - yet
practicallyimportant- issue, so the model capturesonly the most salient featureof
the returnpolicy. A more comprehensivestudy should attemptto extend the model
in several ways. First, I have considered only two choices for the firm: the return
policy and the no-retur policy. In practice,firms offer a variety of returnpolicies,
differingin termsof returnrequirementsand returnperiods.This varietypresumably
allows finrs to vary the difficulty of returnand, more importantly,fine tune the
amount of consumer learning prior to the purchasingdecision. In some cases, a
manufacturer's
qualitydecision, such as the size of clothes, can also affect the actual
return.15Second, consumermoral hazardmust be incorporated.The possibility of
consumersabusingreturnpolicies is an importantconsiderationthatlimits theiruse.
Some of the well-known cases include returnof TV sets after Super Bowl Sunday,
returnof camcordersaftera daughter'swedding, or "borrowing"of partydresses for
special occasions.16 Return policies must be designed to mitigate these kinds of
moral hazard problem. For example, prohibitingcash refimd (i.e., return only in
kind) is effective against some of the above moral hazardproblems. Third,a more
detailedspecificationof consumerpreferencescan be introducedto add realism.For
example, one can think of a specificationof consumerpreferencesthat involves ex
ante diversity as well as ex post diversity.Finally, oligopoly competitionmust be
introduced.Some of the returnpolicies may be introducedas a resultof competition
among retailersratherthanbecause of the profitand efficiency gains thatwe studied
in this paper These extensionswill revealnew dimensionsin the adoptionof return
policies and warrantfirther studies.
YEON-KOOCHE,
Departmentof Economics,
Universityof Wisconsin-Madison,
WI53706, USA
ACCEPTEDSEPTEMBER1995
REFERENCES
DEBREu,G., 1959, Theoryof Value:An Aomatic Analysis of Economic Equilibrium(John
Wiley & Sons, New York).
GALE, 1. and HoLMES, T, 1993, 'Advance-Purchase Discounts and Monopoly Allocation of
Capacity,' American Economic Revew, 83, pp. 135-46.
HOLMSTROM,B. and MILGROM,P., 1990, 'Regulating Trade Among Agents,' Journal of
Institutionaland TheoreticalEconomics, 146, pp. 85-105.
B. and MILOROM,P., 1991, 'MultitaskPrincipal-Agent Analyses: Incentive
HOLMSTROM,
Contracts, Asset Ownership, and Job Design,' Journal of Law, Economics and
Organization, pp. 24-52.
'5 Catlog marketersoften make the sizes of theirclothes fullerthan storesizes becausethey have found
that shoppers are less likely to returngoods that are slightly too large - while they will almost always
returnclothes tat are too small. See "Fashion:Go Figure:Same ShopperWearsSize 6, 8, 10, 12," Wall
StreetJournal, Noveember11, 1994.
16See "Unfair RetUns From Retailers PlagUe Many EleCtronicsMakers," Wall Street joUral,
September26, 1994, and "Retailing:At Clothiers,Its ManyUnhappyReturnsWhen BuyersTurnOut To
Be 'Borrowers,"'WallStreetJournal, June 2, 1992.
? Blackwel
Publishers
Ltd.1996
YEON-KOOCHE
24
E. and TiROLE, J., 1990, 'The Principal-Agent Relationship with an Informed
Principal: The Case of Private Values,' Econometrica, 58, pp. 379-410.
MARVEL, H. P. and PECK, J., 1994, 'Demand Uncertainty and Returns Policies,' mimeo., The
Ohio State University.
PASTERNACK, B. A., 1985, 'Optimal Pricing and Return Policies for Perishable Commodities,'
Marketing Science, 4, pp. 166-76.
MAsIUN,
?
Blackwell Publshers Ltd. 1996