Warranties Research

8/14/2019 Warranties Research

1/13

000

2009 by JOURNAL OF CONSUMER RESEARCH, Inc. Vol. 36 December 2009

All rights reserved. 0093-5301/2009/3604-0014$10.00. DOI: 10.1086/605298

Why Do Consumers Buy ExtendedService Contracts?

TAO CHENAJAY KALRABAOHONG SUN*

We examine purchases of extended service contracts, which are essentially insur-ance products, for electronic products in a retail setting. The primary insurance pur-chase determinants are perceived probability of loss, extent of loss, risk aversion,and amountof insurance premium. We examine howproductcharacteristics(hedonic/utilitarian, manufacturers warranty) and retailer actions (promotions, feature adver-tising) influence the purchase of extended service contracts. We also investigate theimpact of consumer characteristics (income, gender, and prior usage) on these in-surance purchase determinants. To test the predictions, we userevealedpreferences

from panel data of electronic purchases across several product categories.

Examining how consumers make decisions under un-certainty is a fertile field of research in several domains.The purchase of extended service contracts (ESCs) providesan excellent opportunity to understand consumer behaviorunder uncertainty in a field setting. Extended service con-tracts, also referred to as extended warranties or extendedservice plans, are usually offered by retailers and sometimesoffered by manufacturers to lengthen the coverage providedby the manufacturers basic warranty. Technically, ESCs are

insurance products that require consumers to pay premiumsup front for protection against possible failures or problemsin later periods. Usually the seller recommends the purchaseof the ESC immediately after the consumer buys a product.

Since their introduction by large electronics stores in thelate 1980s, ESCs have become a core product for manyretailers. Their cost usually ranges between 10% and 50%of a products original price (Business Week 2004). Typ-ically ESCs are priced at a category-specific level basedon product price tiers. For example, the ESC for televisionsbelow $199.99 costs $39.99, whereas that for televisions

*Tao Chen ([email protected]) is assistant professor of mar-keting at Robert H. Smith School of Business, University of Maryland,

College Park, MD 20740. Ajay Kalra ([email protected]) is professorof marketing at Jesse H. Jones Graduate School of Management, RiceUniversity, Houston, TX 77005. Baohong Sun ([email protected]) isprofessor of marketing at Tepper School of Business, Carnegie MellonUniversity, Pittsburgh, PA 15213. The order of the authorship is alpha-betized. The authors thank the editor, the associate editor, three reviewers,Uzma Kahn, and Erica Mira Okada for their helpful comments.

John Deighton served as editor and Tulin Erdem served as associate editor for this article.

Electronically published June 25, 2009

priced between $200 and $499.99 would be $59.99, andso on. Therefore ESC prices vary across price tiers butnot across products or brands within a tier. The terms typ-ically last from 1 to 4 years, depending on the productcategory. Generating approximately $15 billion dollars an-nually for retailers (Warranty Week2005), ESCs are extremelyprofitable. For example, even though they account for only3%4% of the revenue, in 2003 they contributed more than50% of Best Buys profit and almost 100% of Circuit Citys

profits (Business Week2004). Some analysts estimate that theaverage margin for the ESCs is 50%60%, or approximately18 times the margin for regular products (BusinessWeek2004).Thus, selling service contracts has become a dominant profitgrowth strategy for retailers of durable products (Business Week2005; Warranty Week 2005).

Although most consumer magazines and experts advocateconsumers not buy ESCs because they provide little value,it is intriguing that the demand for ESCs remains high.Consumers justify the ESC purchases because the plans pro-vide them with peace of mind (ABC News 2006). Giventhe financial stake that retailers have and the costs to con-sumers, understanding the factors that affect consumer pur-chase decisions of ESCs and whether the process can be

influenced is both theoretically and substantively important.Literature on ESCs is sparse. Early empirical studies

investigated correlations between attitudes toward ESCsand demographics (Day and Fox 1985). More recent re-search mainly has used an analytical paradigm. The lit-erature addresses the issue of why manufacturers shouldoffer extended warranties (Padmanabhan and Rao 1993)and examines the conditions, such as proportion of heavyusers or competition with third-party insurers, under whichthe ESCs are profitable (Lutz and Padmanabhan 1995; Pad-


2/13

000 JOURNAL OF CONSUMER RESEARCH

manabhan 1995). Although the stream of analytical re-search is rich, the mechanism(s) underlying consumer be-havior regarding the purchase of insurance products suchas ESCs in a retail environment are not well understood.

Insurance literature identifies the major determinants ofpurchase as being the probability of loss, the extent of loss,

the insurance premium charged, and buyers risk aversion(Mossin 1968; Padmanabhan and Rao 1993; Schlesinger1999). These determinants have been shown to influencepurchase of flood (Browne and Hoyt 2000), life (Hammond,Houston, and Melander 1967), and health (Cameron et al.1988; Feldman et al. 1989) insurance. With very few ex-ceptions, the empirical literature in insurance considers thesedeterminants to be objective. From the consumer decision-making perspective, however, these determinants can be sub-

jective. For example, different consumers may have verydifferent perceptions of the probability that the same modelof a TV will fail. Most insurance literature also assumesrisk aversion to be consumer-specific rather than product-or context-specific. Our focus in this article is on under-standing how the likelihood of ESC purchase can be influ-enced by product characteristics and the marketing actionstaken by retailers. In addition, consumer characteristics canaccentuate or attenuate the effect of these insurance purchasedeterminants. We examine how differences in consumercharacteristics affect ESC purchases.

More specifically, first, we examine the product categoriesin which consumers purchase ESCs. We use the categori-zation of hedonic versus utilitarian products and investigatewhether consumer purchase patterns differ based on thisproduct characteristic. Furthermore, there is evidence in theliterature that the product price and the length of a manu-facturers warranty coverage signal quality (Boulding and

Kirmani 1993). We examine whether the coverage lengthincreases consumers ESC purchase likelihood. Second, weexamine retail influences on perceptions of ESCs, becauseretail environments and retailer-controlled decisions may af-fect ESC sales. Retailers decide on price promotions andwhether to advertise these promotions, and we investigatewhether advertised and unadvertised promotions alter con-sumers risk attitudes and hence their propensity to purchaseESCs. Third, we examine whether consumers with certaindemographic variables (gender, income, and past experienceof ESCs) are more likely to purchase ESCs. To test ourpredictions, we use panel data provided by the electronicdepartment of a retailer on the consumer purchase historiesof ESCs across several categories. Thus, we test our pre-

dictions using consumers revealed preferences.To the best of our knowledge, our study is the first

attempt to examine the consumers purchase of ESCs in aretail environment using field data. We thus depart fromexisting research in several important ways. First, in con-trast to the empirical testing in the existing literature thataims to verify analytical models, we are interested in un-derstanding consumer decision making under uncertaintyand how ESC purchase propensities are influenced byproduct, retail, and consumer factors. Second, unlike prior

survey-based research, which generally covers only onecategory (e.g., automobiles), we use purchase history dataacross several categories. Third, unlike most previous re-search, which relies on self-reported risk attitudes and ob-

jective loss probabilities, we infer these factors from con-sumers revealed preferences. Fourth, whereas prior litera-

ture largely views the ESC purchase decision as exogenousand based primarily on consumers inherent risk aversionand usage patterns, we demonstrate that marketing actionscan influence purchase likelihood. Fifth, and finally, weprovide interesting insights into why the ESC purchaseprocess may differ across demographics (e.g., betweenhigh- and low-income consumers).

CONCEPTUAL DEVELOPMENT

Product Characteristics

Hedonic and Utilitarian Products. A robust findingin the insurance literature is that the relationship between the

monetary value of an object and the likelihood of purchasinginsurance is positive (Browne and Hoyt 2000). Therefore,given the same price of any two products and an equal prob-ability of breakdown of the two products, consumers shouldbe equally likely to insure both products. However, this ar-gument does not take into account that equally priced productsmay be valued quite differently by consumers. Products arecharacterized by a combination of utilitarian and hedonic at-tributes (Dhar and Wertenbroch 2000; Okada 2005). Utili-tarian products provide consumers with functional benefitsthat are useful, practical, and necessary, whereas hedonicproducts are associated with fantasy, fun, and pleasure. Dharand Wertenbroch (2000) show that equivalently priced he-donic products are valued more than utilitarian products in

forfeiture situations but that both are equally valued in ac-quisition decisions. This finding is qualified by Okada (2005),who shows that the purchase context affects the relative con-sumer valuations of hedonic and utilitarian products even inacquisition decisions. Specifically, where consumers evaluateproducts one at a time in acquisition decisions, hedonic itemsare valued higher than utilitarian items.

When purchasing durable products, consumers typicallydecide the category prior to visiting the store and then maketheir brand or model selection within the store. After thepurchase decision, a salesperson suggests the purchase of theESC. When deciding on the ESC purchase, the consumermust elaborate on the possibility of losing the product, whichis a state that approximates the forfeiture situation. Dhar and

Wertenbroch (2000) suggest that hedonic items are valuedmore than utilitarian products in forfeiture situations for tworeasons. First, because the attributes of hedonic products aremore sensory and image evoking, consumers are more likelyto imagine and elaborate on them (MacInnis and Price 1987).Second, forfeiture scenarios prompt upward prefactual think-ing that may cause consumers to minimize their emotionalloss by valuing the hedonic items more (Sanna 1999).

Given the same price of a product, consumers are morelikely to purchase a service contract for a product they


3/13

EXTENDED SERVICE CONTRACTS 000

value more as compared with a product that is ascribedlower value. In a series of interesting experiments, Hseeand Kunreuther (2000) manipulate the emotional value ofan identical object (e.g., a painting) by asking participantsto imagine that they fell in love with the product. Theyfind that, holding the level of expected compensation the

same, people are more willing to purchase insurance forproducts for which they develop greater fondness. Theyargue that, when valuation increases due to higher affectionfor the product, the pain of loss is higher and the com-pensation from the insurance serves to provide a conso-lation for the loss. A related line of reasoning is that thereare several instances of hedonic products, such as paint-ings, that are irreplaceable, but utilitarian products are usu-ally replaceable. In the electronic categories that we ex-amine, all products are replaceable. Even though theproducts are replaceable, because of their attachment andmeaning, consumers may anticipate keeping hedonic prod-ucts longer than utilitarian products.

Another set of arguments for why consumers are more

likely to purchase ESCs for hedonic products is that, ascompared with the purchase of utilitarian products, the pur-chase of hedonic goods is often seen as decadent and in-volving self-gratification, which can cause feelings of guilt(Khan, Dhar, and Wertenbroch 2005). Consumers then usecoping mechanisms to reduce or minimize their feelings ofguilt, such as finding reasons to support or justify the pur-chase. For example, consumers might consider the hedonicproduct a reward to themselves or they may categorize in-come they spend as obtained from a windfall (Strahilevitzand Myers 1998). The justification routes identified in theliterature mainly address the reasoning that consumers useprior to their purchase decision. Instead of rationalization,an alternative route to such post-purchase justification is to

take actions after purchase that serve the guilt reductionobjective. One such sequential action is insuring the productby the purchase of an ESC, a behavior that is very prudent.Additionally, and more directly, there is evidence that guiltis associated with risk aversion (Mancini and Gangemi2004). To summarize, given the same price, the extent ofloss associated with hedonic products is greater than thatfor utilitarian products. Moreover, because the purchase ofhedonic products induces more guilt, this may lead to greaterrisk aversion. We therefore predict that, all else being equal,consumers are more likely to purchase extended service con-tracts for products with relatively higher hedonic value thanfor those with relatively less hedonic value.

H1: Consumers are more likely to purchase ESCsfor products with relatively higher hedonicvalue than for products with relatively higherutilitarian value.

Manufacturers Warranty. Extended service contractstypically provide longer and more comprehensive coveragethan the manufacturers warranty. During the purchase de-cision, consumers do not have perfect knowledge about thereliability of the product and therefore of the likelihood of

product failure in the future. There is a large body of evidencethat, when decisions are made under uncertainty, consumerslook for cues that help them infer product reliability (Dodds,Monroe, and Grewal 1991). Prior research has indicated thatthe product price and the terms of the manufacturers basicwarranty are important sources of information regarding prod-

uct reliability. The literature on warranties shows that war-ranty terms affect consumers perceptions of product quality(Shimp and Bearden 1982) and purchase intentions (Bouldingand Kirmani 1993). Warranty terms include the length andscope of the warranty, while product quality dimensions coveroverall quality, probability of breakdown, and specific attrib-ute quality. A generalizable finding from this stream of re-search is that the length and the scope of the manufacturerswarranty indicate higher quality and, more specifically, lowerperceptions of product failure (Boulding and Kirmani 1993;Purohit and Srivastava 2001). Padmanabhan and Rao (1993)find that consumers who buy cars with shorter manufacturerswarranty coverage are more likely to buy ESCs. Since con-sumers are less likely to purchase ESCs for highly reliable

products, we predict that the likelihood of purchasing ESCslowers when manufacturer warranties are longer.

H2: Consumers are less likely to purchase ESCsfor products with relatively longer manufac-turer warranties.

Retail Factors

Promotions. Price promotions are ubiquitous, and re-tailers spend a considerable amount of money communi-cating the promotions. While there is considerable researchthat examines promotions (see Neslin [2002] for an ex-

cellent review), the relationship between price promotionsoffered on products and the purchase of service contractsis not obvious.

There are three lines of reasoning that suggest that therecould be a positive relationship between buying a producton promotion and buying an ESC. First, there is an incomeeffect through the saving accrued by buying a discountedproduct. There is considerable evidence that the incomeeffect resulting from a promotion can result in the purchaseof more expensive brands (Allenby and Rossi 1991; Blatt-berg and Wisniewski 1989). Second, it is also possible thatconsumers will transfer the saving on the promotion tospending on insuring the product because both these com-ponents belong in the same mental account (Thaler 1999).

Third, the fact that the product is being promoted couldlead to a negative impact on product quality evaluations(Dodson, Tybout, and Sternthal 1978; Guadagni and Little1983), though Raghubir and Corfman (1999) point out thatthe negative influence of promotions on quality perceptionsis moderated by factors such as category promotional prac-tices and the expertise of the consumers. Together, thesearguments imply that promotions could enhance the con-sumers willingness to pay for the ESC as well as increasethe perceived probability of product failure. Therefore, we


4/13


hypothesize that, when a product is purchased at a dis-count, consumers are more likely to purchase ESCs.

H3: Consumers are more likely to purchase ESCs forproducts they have purchased on promotion.

Unadvertised Promotions. Very often retailers offerprice promotions on products but do not advertise thosepromotions. The literature has identified two benefits of thisstrategy. First, using an analytical framework, Rao and Syam(2001) show that offering unadvertised price cuts is an equi-librium strategy for retailers as it helps them retain customerswho frequent the store. Heilman, Nakamoto, and Rao (2002)find that unexpected in-store coupons cause consumers toincrease their expenditures by making more unplanned pur-chases. One account for this behavior is that the increasedspending is caused by the consumers unexpected psycho-

logical increase in income. More important, they also arguethat the unexpected windfall gain makes people feel good,resulting in an elevated positive mood.

Mood can affect risk-taking behavior. Isen and her col-leagues have demonstrated that a positive mood can affectrisk taking. In the domain of gains or in low-risk tasks,positive moods lead to risk seeking. However, in the do-main of losses, such as insurance decisions, positive moodslead people to become more risk averse (Dunegan et al.1992; Isen and Patrick 1983). Isen and Patrick (1983) findrisk-averse behavior in lottery and gambling tasks when apositive mood is induced using gifts. Isen, Nygren, andAshby (1988) also show that positive mood makes peoplebecome more sensitive to losses and that they prefer to

minimize their losses rather than maximize their gains. Ina business decision-making context, Mittal and Ross(1998) also show that positive mood causes low risk taking.The underlying mechanism is that the positive mood af-fects the subjective utility value of the potential outcomes.As compared with a neutral mood, a positive mood causespotential losses to appear worse. The reasoning is that,besides the negative financial outcome, the loss also leadsto the demolishing of the positive mood. Interestingly, Isenand Geva (1987) find that people carefully elaborate aboutnegative outcomes when they are in a happy mood. Insummary, anticipating a loss is more unpleasant for thosewho are in a relatively happy mood as compared with thosewho are in a neutral mood.

As compared to consumers who expect a promotion, con-sumers who purchase a product on an unadvertised pro-motion are likely to have an elevated mood because of theunexpected positive surprise. When faced with the ESC pur-chase, this elevated positive mood, in turn, is likely to makethem more risk averse.

H4: Unadvertised promotions are more likely to in-crease consumers risk aversion and thereforeincrease their propensity to purchase ESCs.

Consumer Characteristics

We now discuss how consumer characteristics influencethe ESC purchase decisions. The consumer characteristicswe examine are gender, income, and prior experience withESCs.

Gender. The role of gender in risk taking has receivedconsiderable interest in both the economics (Dekel andScotchmer 1999) and psychology literatures (Magnan andHinsz 2005). In a meta-analysis of research that uses anexperimental paradigm, Byrnes, Miller, and Schafer (1999)conclude that women are less prone to risk taking than men.Empirical studies reinforce these results by finding thatwomen do less risk taking in their allocation of retirementsavings plans (Sunden and Surette 1998), selecting term lifeinsurance (Halek and Eisenhauer 2001), and buying healthinsurance (Feldman et al. 1989). As ESC is essentially aninsurance product, we also expect to confirm that womenare more likely to purchase these contracts.

H5: Women are more likely to purchase ESCs thanmen because they are more risk averse.

Income. A decreasing absolute risk aversion that im-plies income is negatively related to insurance purchase iswidely accepted in the theoretical analysis of insurancedemand (Mossin 1968). However, the empirical findingsin the insurance literature find mixed results, where somestudies show a positive relationship between income andpurchase of insurance (e.g., Browne and Hoyt [2000] forflood insurance), while others find a negative relationship(e.g., Cicchetti and Dubin [1994] for telephone line servicecontracts). Padmanabhan and Rao (1993) find that higher-

income consumers are more likely to buy ESCs in theautomobile category, but Padmanabhan (1995) did not findany correlation. We therefore view the effect of income asan empirical question.

Past Usage of ESCs. Consumers who purchase ESCsin any category consist of two segments post-purchase: thefirst segment does not experience product failure in the cat-egory purchased, and therefore does not avail of any benefitsfrom the purchase of the ESC. The second segment does facea problem with the product and has to exercise the contract.We argue that consumers who do face a product failure prob-lem, regardless of the category where product failure is ex-perienced, are more likely to view purchasing the ESC more

favorably on any subsequent purchase occasions. There isevidence that the purchase of insurance is prone to biases dueto past claims (Johnson et al. 1993). Prior claims can distortfuture product failure rates due to availability and vividnessbiases. The past product failure is likely to increase the po-tential perceived probability of future related and even un-related product category breakdowns. Therefore, if ESCs havebeen used in the past, consumers perceived probability offailure will increase, which, in turn, increases their propensityto purchase ESCs for other related or unrelated products.


5/13


TABLE 1

SAMPLE STATISTICS

Variable ExplanationMean or

frequencyStandarddeviation

Dijt Whether ESC was purchased .33 .47

Audio .38 .49Video .24 .43

Phone .31 .46Camera .45 .50Computer .34 .47

Game .30 .46Mobile audio .52 .50

ESCPricejt Price of ESC ($) 59.25 81.06ESCCoverjt Coverage period of ESC (years) 2.75 .90MCoverjt Coverage period of manufacturers basic warranties (years) .87 .33

PPricejt Price of the product ($) 340.09 657.00a

Hedonicj The difference between hedonic and utilitarian values of a product .49 1.61

Promijt Price promotion ($) 59.22 100.96Adpromjt Whether a promotion is advertised/featured .28 .45

Unadpromjt Whether a promotion is unadvertised/unfeatured .72 .45Incomei Income level on a 100-point index 64.48 23.63Genderi Whether the consumer is a male .62 .49

Usei Whether a previous purchase of ESC was availed .08 .27

NOTE.The sample contains 604 households that purchased 553 service plans for 1,676 products between November 2003 and October 2004.aThe standard deviation of price is relatively large because it reflects the variation across different product categories.

H6: Prior usage of a service contract increases con-sumers perceptions of product failure rates andtherefore makes it more likely that they purchaseESCs for other products.

In summary, we conceptualize that product characteris-tics, retailer actions, and consumer characteristics influence

the purchases of ESCs. The mechanisms of these influencesinclude altering the perceptions of the probability of loss,the extent of loss and risk aversion, and the psychologicalincome effect. We test the predictions of the outcomes usingdata on observed choices made by consumers and the as-sociated mechanisms permitted by the available data.

DATA DESCRIPTION

Our data come from the electronics department of oneretailer. Some retailers, such as Circuit City and Wal-Mart,offer consumers a menu of ESCs that vary in length andprice, whereas others, such as Best Buy and Target, provideonly one plan. Our focal retailer offers only one plan. We

have access to the consumer purchase histories of ESC plansof 604 households from November 2003 to October 2004;these include 1,676 purchases of products and 553 purchasesof ESCs. More specifically, the data consist of the completehistory of the households purchases of electronic durablesand service contracts during these 12 months. They containdetailed information about the products and service con-tracts, such as product type, product price, promotion offeredon the product, presence of a feature (i.e., advertised pro-motion), price and length of coverage of the service con-

tracts, and time and location of purchases. We also haveaccess to consumer characteristics, such as income, gender,and a dummy variable that identifies whether a consumerhas used an ESC in the past (even before the start of theobservation period). Since the number of product types isunmanageable, we follow the retailers practice and classifyproducts into seven categories on the basis of their generalfunction: video, audio, phone, camera, computer, game, andmobile audio.

We present the samples statistics in table 1, which revealsthat 31% of consumers purchase an ESC at least once duringthe observation period and that ESC purchases constituteapproximately 33% of all purchase occasions. Among theseven product categories, consumers are most likely to pur-chase service contracts for mobile audio and cameras andleast likely to purchase them for video and game systems.The average price of the products purchased is $340.09. Themost expensive product is a plasma television, with pricesranging between $4,000 and $5,000. The average price ofthe service contracts is $59.25, which means that they are

priced at roughly 17.42% of the product price. The averagelength of coverage is 33 months, varying from 24 to 48months. Among the product purchases, 35% included a pricepromotion, with an average promotion discount of $20.53.Of the price promotions, 28% had been advertised throughcirculated publications and local newspaper inserts. Finally,8% of consumers had used ESCs at least once during theobservation period. Sixty-two percent of the sample are maleconsumers. In the data set, income is provided as a 100-point index, the average being 64.48.


6/13


TABLE 2

HEDONIC CATEGORIES VS. UTILITARIAN CATEGORIES

Categories

Hedonicvalue

Utilitarianvalue

Mean Standarddeviation Mean Standarddeviation

Audio-visual receiver 3.92 .16 4.02 .15Boom box 4.63 .15 3.47 .20Car stereo 4.81 .16 4.38 .19CD recorder 2.93 .31 3.93 .29Cell phone 3.61 .31 5.36 .16Computer peripheral 2.89 .33 4.32 .23Desktop 2.50 .30 4.88 .23Digital camcorder 4.56 .15 4.42 .14DVD player 4.90 .16 4.34 .17High-end camera (1$400) 4.51 .25 3.06 .29High-end speaker (1$200) 4.11 .26 3.03 .31Home CD changer 4.17 .19 3.49 .18Home phone 1.75 .32 4.57 .31Home theater 4.66 .25 3.83 .30

Ipod 4.54 .24 3.23 .28Low-end camera (!$400) 4.72 .15 5.01 .12Low-end speaker (!$200) 2.53 .29 3.63 .29Notebook 3.61 .28 5.21 .16Pay phone 1.79 .29 4.40 .35PDA 4.11 .16 5.03 .14Portable CD player 4.91 .27 4.17 .30Portable DVD player 3.97 .23 2.76 .28Printer 2.25 .33 5.06 .21Projection TV 4.77 .28 3.21 .32Regular TV (!$1,000) 3.12 .25 5.23 .24Shelf stereo system 4.46 .16 3.99 .17Two-way radio 3.13 .20 4.15 .18VCR 2.65 .30 3.91 .29Video game console 4.53 .30 2.59 .31Video game controller 4.30 .31 2.26 .32Voice recorder 3.06 .18 4.46 .15

To determine the hedonic and utilitarian values of eachproduct, we conducted a survey on a sample of 107 adults,using the methodology of Okada (2005). Respondents wereasked to rate each product category on one hedonic scale,anchored between not at all hedonic (0) and extremelyhedonic (6), and one utilitarian scale, anchored between not

at all utilitarian (0) and extremely utilitarian (6). FollowingOkada (2005), hedonic items were described as fun/pleasant/enjoyable, whereas utilitarian products were described as use-ful/practical/functional. We report the results in table 2. Wecode as the difference between the hedonic andHedonicjutilitarian values of a category.

EMPIRICAL ANALYSIS

Model

We consider consumers, , who bought elec-ip 1, , Itronic products in product categories, , duringjp 1, , Jpurchase occasion t; we therefore use to representDijt

whether consumer i purchases an ESC for product j atoccasion t:

1 if consumer i buys ESC for product j at time t,D p (1)ijt {0 otherwise.

Perceived Failure Rate. Because consumers buy ESCsto cover the risk of future replacement or repair cost incurredby product breakdowns, purchase decisions should dependon the likelihood of product failure. However, at the timeof purchase, consumers face uncertainty regarding this pre-cise probability. According to the literature, consumers oftenuse the price of the product and the length of coverage of

the manufacturers warranty as quality cues for assessingproduct reliability (Boulding and Kirmani 1993; Erdem,Keane, and Sun 2008; Purohit and Srivastava 2001). Con-sistent with the literature, we model the consumers per-ceived product failure rate using the following function:

P MVp b + b + b Rprice + b Cover + e , (2)ij 0i 0j 1i j 2i j ij

where . This formulation is based on Erdem etVp V + eij ijijal.s (2008) empirical finding that price signals quality andBoulding and Kirmanis (1993) finding that the manufac-turers warranty signals quality. The variable rep-PRpricejresents the price of product j relative to the average priceof all products within its category. It reflects how much

more expensive or cheaper the price is relative to the pricesof other brands in a given category. The variable MCoverjis the length of coverage of the manufacturers basic war-ranty, which comes free with the product. We include thisvariable to control for the potential association of a longermanufacturers warranty coverage with lower failure rates,which would affect ESC purchase propensities. Finally, eijrefers to unobservable factors that affect consumers per-ceptions of breakdown rates.

The coefficient is a consumer-specific constant termb 0i

that captures unobserved factors that may affect perceptionsof failure rates across product categories. For example, con-sumers with children may perceive a higher probability ofproduct failure. Next, is a category-specific constant termb 0jthat captures different failure rates across categories. Forexample, product categories with the newest technology ormultiple sophisticated features may be perceived as morelikely to break down. Coefficient captures the effect ofb1iproduct price on perceived failure probabilities, and coef-ficient measures the effect of the length of the manu-b2ifacturers basic warranty.

Assuming that follows an independent and identicallyeij

extreme value distribution, the probability of product failureis given by a binary logit model:

Vijer p . (3)ij Vij1 + e

Expected Benefit of Purchasing Service Contract. Thedecision to purchase an ESC requires a comparison be-


7/13


tween the cost and the benefit of owning it. Whereas theprice represents the cost to the consumer, uncertainty sur-rounds the benefits, which accrue only if the product breaksdown. If the product does not fail, an ESC purchase in-volves only the cost; if the product fails, the ESC enablesconsumers to avoid out-of-pocket expenses to repair or

replace the product.At the time of purchase, consumers must assess the ben-

efits according to the expected replacement or repair coststhey may avoid by buying the ESC. For simplicity, we as-sume that the terms of the ESC include replacing the failedproduct with a similar product at a similar price. Becauseconsumers do not observe information about repair costs,the cost of replacement provides a reasonable surrogate; highlabor costs mean many consumers simply elect to replacerather than repair their durable products. In addition, repaircost usually is proportional to replacement cost.

Because consumers perceived failure rate is , the ex-rijpected cost of replacement is given by wherePr Price ,ij j

is the price for product j paid by consumer i. Thus,PPricejthe expected benefit of owning a service contract equals thereplacement price weighted by the perceived probability ofproduct failure. According to the special property of Ber-noulli distribution, the variance of the expected replacementcost is P 2r (1r )(Price ) .ij ij j

Purchase Decisions. Under uncertainty, consumersmake purchase decisions on the basis of their expected util-ities, given the perceived probability of failure , whichrijconsumer i develops for product j. We then assume that theexpected utility that affects the ESC purchase decisions isgiven by

PE[U Fr ]p a + a Hedonic + a r Priceijt ij 0i 1 j 2i ij j

P 2+ a r (1r )(Price ) + a Prom3i ij ij j 4i ij

ESC ESC+ a Price + a Cover + a PP5i j 6i j 7i it

P 2+ a r (1r )(Price ) #Hedonic (4)8 ij ij j j

P 2+ a r (1r )(Price ) # Unadprom9 ij ij j ij

P 2+ a r (1r )(Price ) #Adprom + ,10 ij ij j ij ijt

wh ere is th e difference between the h edonicHedonicjvalue and utilitarian value of product j. We include thisvariable to test whether a product with greater hedonicvalue makes consumers more likely to purchase ESCs.

The term refers to consumer is perceived failure raterijfor product category j, as described by equations 13,and is the expected cost of replacement. In ad-Pr Priceij jdition, equals the variance of the ex-P 2r (1 r )(Price )ij ij jpected cost of replacement, which contains informationabout the accuracy of consumers perceived failure rate.This formulation is consistent with Levy and Markowitz(1979), who allow the mean and variance to affect expectedutility functions. As in the standard marketing and econom-ics literatures, we interpret coefficient as sensitivity toa3i

uncertainty, a measure of consumer risk attitude (Erdem andKeane 1996). We include , the depth of the priceProm ijpromotion on product j that consumer i receives, to testwhether a price discount makes consumers more or lesslikely to purchase service contracts. The variable ESCPricejis the price of the ESC charged by the retailer, which rep-

resents the cost for consumers to buy the ESC. The variablerefers to the additional length of coverage providedESCCoverj

by the ESC over and above the length of the manufacturerswarranty. Finally, captures consumer is ESC purchasePPithistory. Conceptually similar to Guadagni and Littles(1983) loyalty index, is the weighted average of pur-PPitchases of ESCs prior to occasion t; it includes the purchaseof an ESC in any product category. Therefore,

J

PP p vPP + (1 v) D , (5)it it 1 ijt1jp1

where is the smoothing factor.v

To test whether purchasing a hedonic product makesconsumers more risk averse, we include the interactionterm between hedonic value and uncertainty (variance ofthe expected replacement cost), P 2r (1r )(Price ) #ij ij j

. Variable indicates an unadvertisedHedonic Unadprom j ijprice promotion, and its interaction with uncertainty,

, enables us to test theP 2r (1r )(Price ) # Unadpromij ij j ijpossibility that consumers become more risk averse witha surprise promotion. Similarly, indicates if theAdprom ijpromotion is advertised, and we again include its inter-action with uncertainty, P 2r (1r )(Price ) # Adprom .ij ij j ij

We use the vector to represent the parameters to beViestimated, that is, Vip (b0i, b0j, b1i, b2i, a0i, a1, a2i, a3i , a4i,a5i, a6i, a7i, a8, a9, a10) for all j. We define as

U p U ij ij ij

the deterministic part of the utility functions. Assuming theerror term is independently and identically extreme valueijdistributed, we obtain the probability of consumer choosingian ESC for product conditional on :j Vi

Fr ]E[Ui jt i jePr (D p 1FV )p . (6)ijt i Fr ]E[Ui jt i j1 + e

Consistent with the insurance literature, we allow consumersto make purchase decisions on the basis of perceived failurerates, expected replacement cost, and uncertainty associatedwith the future replacement costs.

Heterogeneity and Estimation. To measure the effectof consumer heterogeneity on responses to the main vari-ables (i.e., coefficients in the perceived failure rate and utilityfunctions), we assume each parameter is a linear functionof consumer demographic variables, such as income, gender,and experience with ESCs, and thereby characterize howthe coefficients in the utility function differ across consumerdemographics and past usage with ESCs:

V p g + g Income + g Male + g Use + e , (7)i 0 1 i 2 i 3 i i


8/13


TABLE 4

COMPARISON WITH SAMPLE STATISTICS

Sample Model

Predicted purchase probability:Audio .38 .34Video .24 .28Phone .31 .31Camera .45 .47Computer .34 .26Game .30 .39Mobile audio .52 .44

Hit rate .77Efrons R2 .23McFaddens R2 .18

NOTE.Hit rate p percentage of observations correctly predicted.

N 2 (y p)i iip1

2Efrons R p 1 ,N 2 (yy)iip1

where p p purchase probability predicted by the model.i

Log-likelihood(Model)Proposed2McFaddens R p1 .

Log-likelihood(Model )Intercept

TABLE 3

MODEL COMPARISON

Benchmarkmodel 1

Benchmarkmodel 2

Benchmarkmodel 3

Proposedmodel

Log-likelihood 963.79 943.80 937.19 869.15AIC 974.79 958.80 956.19 929.15

NOTE.Benchmark model 1: Proposed model without perceived probabilityof failure, expected replacement cost, uncertainty, hedonic value, three inter-action terms with uncertainty, and heterogeneity. Benchmark model 2: Pro-posed model without hedonic value, three interactions terms, and heteroge-neity. Benchmark model 3: Proposed model without heterogeneity. AIC pAkaike information criterion model.

where and refer to the income and gender of Income Ma lei iconsumer i and is a dummy variable indicating whetherUseiconsumer i has ever used a service contract. The coefficientsof these consumer-specific variables measure how the impactof these factors on perceptions of failure probabilities andESC purchase probabilities differ across consumers. Note

that, by modeling this heterogeneity, we introduce interac-tion terms among the three consumer demographic variablesand all coefficients from the perceived failure rate and utilityfunctions in a parsimonious way. The coefficient estimatesin the heterogeneity equation 7 reveal how consumer char-acteristics moderate all the main effects. For example, if

is estimated as positive and is estimated as negative,a g2i 1it implies that consumers are sensitive to expected replace-ment cost on average but that higher-income consumers areless so.

We assume that the unobserved part of equation 7 isdistributed as normal. We use a hierarchical Bayesian ap-proach to estimate the model. For identification, is fixed.b04Readers interested in the estimation procedure are referred

to Allenby and Rossi (1999) and Gelfand and Smith (1990)for details.

Results

To explore the fit of our proposed model, we estimate threebenchmark models. The first model is a logit model, in whichthe utility function is a linear function of the category con-stants, price of the product, promotion, price of ESC, andcoverage length of ESC. This is our proposed model withoutperceived failure rate, expected replacement cost, uncertainty,hedonic value, the three interaction terms with uncertainty,and heterogeneity. The second model matches our proposedmodel but without hedonic value, interaction terms, and het-

erogeneity. Finally, in the third model, we exclude just het-erogeneity from our proposed model. In table 3, we providethe log-likelihood and Akaike information criterion (AIC)model comparison results. As the results show, our proposedmodel fits the data better. By taking into account consumersperceived failure rate, the modifying effects of product char-acteristics, advertising, promotion, and consumer character-istics, our model explains the observed data better than thebenchmark models.

In table 4, we compare the predicted ESC purchase prob-

abilities with those from the sample for each product cat-egory. Given the discrete nature of the dependent variable,our estimated consumer model approximates the data rea-sonably well. To further establish model validity, we obtainthe available breakdown rates published by Consumer Re-ports and PC Magazine in the same period and comparethem with the perceived probabilities of breakdown inferredfrom the model. Holding the camera categorys failure rate

to be fixed at 12% (reported by Consumer Reports), ourmodel predicts consumers perceived rates of product failurefor video, games, and phone equipment to be 12%, 14%,and 22%, respectively. The rank order is quite consistentwith the actual breakdown rates of 9%, 18%, and 26%,respectively, for the three categories.

In table 5, we report the parameter estimates in the per-ceived probability function. As expected, the results suggestthat consumers tend to associate higher prices with a lowerprobability of product failure. We thus confirm the conven-tional wisdom that product price plays a significant role insignaling product reliability and reduces consumer concernsabout product failure. However, the length of the manufac-turers warranty has no significant impact on the perceived

probability of failure. This result is inconsistent with priorexperimental findings (Boulding and Kirmani 1993) and hy-pothesis 2. Unlike experimental research where the manu-facturers warranty length is manipulated, our data containmanufacturers warranties that are similar in length for mostelectronic categories. We do not believe that these resultscontradict existing theory; rather, the data may lack effectivewithin-category variation to provide results consistent withexperimental research.

In the purchase utility equation, the coefficient of the


9/13


TABLE 5

MODEL ESTIMATES

Parameter Symbol EstimateStandarddeviation

Perceived probability:Constant term: b0j

Audio 1.21* .33Video 2.46* .35Phone .65+ .37Camera FixedComputer 3.34* .30Game 1.5* .35Mobile audio .26 .41

(Relative) price of the product b1i .77* .24Coverage of manufacturer warranty b2i .04 .34

Utility function:Constant term a0i 4.75* 1.23Hedonic product a1 .38* .19Expected replacement cost a2i 16.97* 3.99Variance of expected replacement cost (uncertainty) a3i 11.79* 3.55Promotion a4i 1.31* .51Price of ESC a5i 19.11* 6.07

Coverage length of ESCcoverage length of manufacturers warranty a6i .61 .97

Past purchases of ESC a7i 3.14* .69Hedonic # Uncertainty a8i .88 .57Unadvertised promotions # Uncertainty a9i 2.26* .57Advertised promotions # Uncertainty a10i .97 .78

+Significant at the .10 level.*Significant at the .05 level.

dummy variable is positive and significantHedonicj( ). This result provides support of hypothesis 1,a p 0.381confirming that consumers are more likely to purchaseESCs for products with relatively higher hedonic valuethan for products with relatively higher utilitarian value.The parameter also is significant and positive, whicha 2iimplies that consumers are sensitive to the expected costof replacement and increase their probability of purchasing

ESCs when they estimate a higher replacement cost. Thisresult is consistent with existing findings in the insuranceliterature that consumers are more likely to purchase in-surance for products where the expected cost of replace-ment is high (Williamson, Ranyard, and Cuthbert 2000).The coefficient of the squared term of the expected re-placement cost, , also is positive and significant, whicha 3isuggests that consumers are risk averse and that greateruncertainty about the replacement cost increases their pro-pensity to purchase ESCs. The coefficient of price pro-motion is positive and significant, indicating that, if con-sumers receive a price discount, they are more likely topurchase an ESC. We find support for hypothesis 3. Asanticipated, consumers also are sensitive to the price (in-

surance premium) of the ESCs ( ), though wea p

19.115ifind no impact of the length of the ESC. This absence ofeffects may result from the lack of effective within-categoryvariation in the length of the ESC. Another reason mightbe that, in some categories, the length of the ESC offeredis too long relative to the product life cycle. Finally, thecoefficient for past purchase of ESC ( ) is sig-a p 3.147inificant, indicating that consumers who have purchasedESCs in the past are more likely to do so in the future.

The interaction of hedonic value and uncertainty is not

significant ( ). Therefore, we must reject the ex-a p 0.888iplanation that guilt following the purchase of hedonic itemsincreases risk aversion. Eliminating this explanation suggeststhat consumers may be more likely to buy ESCs because theyvalue hedonic items more, which is consistent with the ex-perimental results of Hsee and Kunreuther (2000).

In hypothesis 4, we predict that unadvertised promotionscreate a positive mood and increase risk aversion. The in-

teraction between unadvertised promotion and uncertaintyis positive and significant ( ). Thus, when a re-a p 2.269itailer uses unadvertised promotions, consumers risk aver-sion increases and this makes them more likely to buy ESCs.This supports hypothesis 4. The interaction between adver-tised promotions and uncertainty is not significant (a p10i

), which confirms that only unadvertised promotions0.97alter risk aversion levels.

We next examine the coefficients in the heterogeneityequations reported in table 6. These estimates describe howconsumers with different characteristics respond differentlyto our focal variables in the perceived product failure andutility equations. We discuss only the significant estimates.Please note that the estimates in the heterogeneity equations

are interpreted as interactions between consumer character-istics (income, gender, and prior usage) and the explanatoryvariables in the probability and utility equations. In otherwords, the estimates indicate how the explanatory variables,such as the impact of product price on the perceived prob-ability of failure, are moderated by consumer characteristics.For example, the coefficient of men (0.75) in the hetero-geneity equation for indicates that, relative to women,b1imen are less likely to associate higher product price withlower probabilities of failure. Said differently, women are


10/13


TABLE 6

HETEROGENEITY ESTIMATES

Parameter Symbol

Estimate

Income MenPrior

usage

Perceived probability:Constant term b0i .04

(.08)1.06+

(.58)1.18*(.41)

(Relative) price of the product b1i .31*(.13)

.75*(.23)

.04(.21)

Coverage of manufacturer b2i .01(.11)

.40(.55)

.41(.47)

Utility function:Constant term a0i .10

(.34)3.7*

(1.01)1.35

(1.93)Expected replacement cost a2i 1.50

+

(.84)14.64*(4.77)

1.82(7.07)

Uncertainty a3i 1.05(.98)

7.52*(2.84)

5.82(4.05)

Promotion a4i .53*(.16)

.62(.54)

.27(.99)

Price of ESC a5i .65(1.21)

6.28(3.94)

7.28(4.99)

Coverage length of ESCcoverage length of manufacturers warranty a6i .08(.24)

2.57*(.79)

1.40(.97)

Past purchases of ESC a7i .01(.16)

2.33*(.49)

.97(.74)

NOTE.Standard deviations are in parentheses.+Significant at the .10 level.*Significant at the .05 level.

more likely to believe that high-priced products are morereliable.

As indicated by the positive coefficient of income in theheterogeneity equation for (0.31), relative to high-incomeb1iconsumers, low-income consumers are more likely to inferproduct reliability from product price. Low-income consum-

ers are also more sensitive to the expected cost of replace-ment (1.50) relative to high-income consumers. When theproduct is on promotion, as compared to the high-incomeconsumers, the low-income consumers are also more sen-sitive to promotions (0.53) and more likely to purchaseESCs.

The overall effect implies that, given everything elseequal, lower-income consumers are more prone to buyingESCs. Two explanations account for this finding: first, lower-income consumers are more sensitive to the replacementcost, which is not very surprising since the more limiteddisposable income makes it much harder for them to re-purchase the product in event of product failure. Second,and more interesting, the impact of price promotions on the

purchase of ESCs is greater for low-income consumers. Theconjecture that low-income consumers are more prone toESC purchases is inconsistent with the findings of Pad-manabhan (1995) and Padmanabhan and Rao (1993). Theseresults can be accounted for by the differences in the au-tomobile category examined in the earlier studies and thecheaper electronic product categories. In the relativelycheaper electronic categories, higher-income consumershave the ability to self-insure their purchases, but lower-income consumers do not. Another possibility is that the

insurance objective for automobiles and electronics productsdiffers. In automobiles, the goal of purchasing the ESC isprimarily repair and maintenance, while for electronic prod-ucts, it is replacement. Since higher-income consumers havehigher time costs, they may be more willing to purchaseESCs for automobiles rather than for electronic products,

where product replacement is the primary goal.Gender also entails several differences. Compared with

women consumers, men are less likely to perceive that prod-ucts will fail (1.06) and less likely to rely on price to inferproduct reliability (0.75). Everything else being equal, menare more likely to purchase ESCs than women (3.70). Menare more sensitive to expected replacement costs (14.64).This is a puzzling result that suggests that there may besome veracity to the stereotype of men being more respon-sible for electronic and appliance product repairs and up-keep. Consistent with prior findings, men appear to be lessrisk averse than women (7.52). Men are less likely to beinfluenced by the length of the longer coverage (2.57) andare less influenced by past purchases of ESCs (2.33). How-

ever, it remains difficult to predict whether ESC purchasesrelate to gender overall. Although women are inherentlymore risk averse and perceive higher probabilities of productfailure, this is offset by mens lower price sensitivity to theprice of the ESCs as well as higher sensitivity to the re-placement costs. Interestingly, women are more likely torely on product price to infer product reliability. These re-sults are consistent with He, Inman, and Mittal (2008), whopropose that the relationship between gender and buyinginsurance can be moderated by beliefs of own-self capa-


11/13


bility. He and colleagues find that, in loss avoidance situationssuch as buying insurance, women are less likely to buy in-surance if their belief of own-self capability is high but thatissue capability has no impact on men. In the purchase ofESCs, high-priced products may increase womens percep-tions of self-efficacy. The raw data show that, across the

purchase occasions, 33.46% of men and 32.23% of womenpurchase ESCs. Thus, we cannot confirm hypothesis 5.

Finally, and interestingly, examining the column in table6 for prior usage, we see that, if consumers have previouslyexercised a service contract, they perceive a greater prob-ability of product failure, regardless of the product category(1.18). This could be due to the availability bias (Tverskyand Kahneman 1974), or perhaps past failures make con-sumers less confident about the way they handle products.This result confirms hypothesis 6 that past product failureincreases perceptions of future product failure and conse-quently has an impact on subsequent purchase of ESCs.

DISCUSSION

In this article, we use consumers purchase decisions ofextended service contracts to understand their decision mak-ing under uncertainty. The decision to purchase insurancerequires consumers to assess the probability of loss and themagnitude of loss and make a decision on the basis of theirperceptions of risk and the premium charged. Our resultssuggest that this decision is context specific and that it canbe influenced by marketing actions.

We conceptualize and demonstrate that the purchase ofESCs is influenced by three major factors: hedonic/utilitarianvalue of products, price promotions and advertising, andconsumer characteristics. We develop hypotheses about theimpact of these three factors on consumers perceptions of

product failure probability and their likelihood of purchasingan ESC. To test the predictions, we apply a consumer choicemodel to panel data containing consumer purchase historiesof service contracts from the electronics department of aretail store.

We hypothesize that consumers are more likely to pur-chase insurance provided by ESCs for products that haverelatively higher hedonic value than utilitarian value for tworeasons. First, given equal prices, consumers valuation ofhedonic products is higher, which increases the pain of loss.Second, the purchases of hedonic products also elicit feel-ings of guilt and heighten risk aversion. The results confirmthat the probability of purchasing ESCs is greater for prod-ucts with high hedonic value, although this is not necessarily

so because consumers feel guiltier about such purchases.The results reinforce the findings of Dhar and Wertenbroch(2000) that suggest that valuation of hedonic products ismore than that for utilitarian products. The results also sug-gest that consumers are willing to pay a premium to protectthe additional valuation.

Our results on the impact of retail environment demonstratethat retailer actions can influence ESC sales. Promotions in-crease the likelihood of purchasing ESCs because there maybe a psychological increased income effect realized from sav-

ings due to the price promotion. We also predict and findsupport that unadvertised promotions augment the effects ofpromotions. The unexpected gains from unadvertised pro-motions evoke positive moods in consumers (Heilman, Nak-amoto, and Rao 2002). The positive mood, in turn, increasestheir risk aversion and consequently the purchase of ESCs.

We confirm earlier empirical findings that product priceserves as a cue of quality (Erdem et al. 2008). Consumersuse high price as an indicator of product quality and assigna lower chance of the product failure if it has a higher price.The results do not confirm that the length of the basic man-ufacturers warranty is an indicator of quality. We suspect,however, that this is an artifact of the real world data wherethere is little variation in the length of warranty offeredacross brands. The results also suggest the state dependencein the purchase of ESCs. Consumers who have purchasedservice contracts in the past are likely to do so again inother product categories in the future.

The results from the heterogeneity analysis are intriguing.In contrast to prior empirical findings ( Padmanabhan 1995;Padmanabhan and Rao 1993), we find that, as compared tohigh-income consumers, low-income consumers are morelikely to purchase ESCs. The analysis reveals that they arelikely to do so because they are more sensitive to the re-placement costs in the event of product failure. Unlike inthe case of automobiles, where high-income consumers buyESCs to avoid maintenance because they have higher timecosts, in the electronics product category, low-income con-sumers buy insurance to hedge against the out-of-pocketcosts of replacing the product. Additionally, lower-incomeconsumers are also more predisposed to exercise the savingsobtained from promotions to purchase ESCs.

If the ESCs do, in fact, offer little value, the results imply

a perverse impact on consumer welfare. The lack of financialability of low-income consumers to replace products inducesthem to pay a potentially unnecessary and overpriced in-surance premium. High-income consumers, for whom prod-uct replacement is not a cause for anxiety, incur a lowertotal cost of product acquisition. These findings are some-what ironic in light of observations that poor sick patientswho are unable to afford health insurance pay the highestprices for drugs (e.g., Frank 2001). In the health domain,poor patients are unable to afford insurance in a categorywhere it is salubrious to buy it, but they opt for insurancein an area where the investment is inexpedient.

Although women are risk averse and perceive higherprobability of product failure, mens inherent propensity to

buy insurance is higher, and they are more sensitive to theexpected replacement costs. We find confirmatory evidencethat women are more risk averse than men. However, thehigher aversion does not necessarily translate into an in-creased likelihood of buying ESCs. According to the rawstatistics, the proportion of men and women who purchaseESCs is approximately the same.

Finally, the analysis confirms that consumers who haveencountered product failure and have utilized an ESC forthat product in the past increase their estimates of subsequent


12/13


product failures and are also less sensitive to the pricescharged for ESCs. We conjecture that perceptions of riskare higher for these consumers due to the availability bias.

With data based on revealed choices in a store environ-ment, we focus primarily on examining outcomes of thedecision process. Unlike experimental research, which can

pinpoint the processes by controlling all other factors, em-pirical research is more exploratory. The main advantage ofconducting empirical research is that we have revealed pref-erence data collected in a real setting. Thus, we can inves-tigate the simultaneous impact of many factors on the pur-chase decision. More important, it allows us to establish theexternal validity of the key effects. Finally, empirical re-search captures effects that may be difficult to replicate usinghypothetical experimental scenarios. For example, it maybe hard to simulate the positive mood evoked by an un-expected price promotion.

Consistent with the nature of our data, though, we cannottest for all underlying processes, which is an endemic lim-itation of empirical research. Although our primary objec-tive has been predicting outcomes based on theoretical rea-soning, when the data allow, we make the best possibleattempt to gain an understanding of the processes. It is im-portant that experimental research be conducted to confirmthe underlying mechanisms.

The mechanisms that influence perceptions of value, per-ceived probability of failure, and risk aversion require morein-depth analysis, including through experimentation. Forexample, we find that consumers are more likely to purchaseESCs in response to unadvertised promotions, but we cannotdetermine post hoc whether consumers actually were un-aware of these promotions, so research in an experimentalsetting should confirm this effect. In addition, our data in-

dicate that guilt following purchase of hedonic items doesnot account for ESC purchases. Future research should ex-amine our proposed explanation that high hedonic valuationprompts insurance purchase.

A key difference between ESCs and other insurance isthat the purchase process for the latter is more deliberate.For most ESC purchases, decision making likely takes placequickly and with less deliberation. These differences suggestpotential avenues for exploration. Such exploration alsomight consider the different ESC offers among retailers:some of them offer a single ESC option where the onlydecision a consumer faces is whether to buy or not buy.Other retailers offer menus of ESC contracts. An interestingissue for future research is to investigate how consumers

choose among a menu of ESC plans. We find evidence thatunadvertised promotions increase risk aversion. Other fac-tors that prime risk (e.g., Mandel 2003) in retail environ-ments are useful avenues for further investigation.

We adopted a static approach for our analysis, but furtherresearch might model consumer dynamics explicitly in thechoice of service contracts. Structural models in which con-sumers learn the benefits of ESCs, incorporate prior infor-mation to form their expectations of product reliability, de-velop expectations about future replacement costs, and are

forward looking about risk might help investigate the dynamicnature of consumer decision processes. Finally, additionalconsumer decisions, such as product returns, could be studiedin conjunction with the ESC purchase decision.

REFERENCES

ABC News (2006), Why You Shouldnt Buy a Warranty, http://www.abcnews.com, November 14, 2006.

Allenby, Greg M. and Peter E. Rossi (1991), Quality Perceptionsand Asymmetric Switching between Brands, Marketing Sci-ence, 10 (3), 185204.

(1999), Marketing Models of Consumer Heterogeneity,Journal of Econometrics, 89 (12), 5778.

Blattberg Robert C. and Kenneth J. Wisniewski (1989), Price-Induced Patterns of Competition, Marketing Science, 4 (8),291309.

Boulding, William and Amna Kirmani (1993), A Consumer-SideExperimental Examination of Signaling Theory: Do Consum-ers Perceive Warranties as Signals of Quality? Journal of

Consumer Research, 20 (1), 11123.Browne, Mark J. and Robert E. Hoyt (2000), The Demand forFlood Insurance: Empirical Evidence, Journal of Risk andUncertainty, 20 (3), 291306.

Business Week (2004), The Warranty Windfall, http://www.businessweek.com, December 20.

(2005), Watch Out, Best Buy and Circuit City, http://www.businessweek.com, November 21.

Byrnes, James P., David C. Miller, and William D. Schafer (1999),Gender Differences in Risk Taking: A Meta-Analysis, Psy-chological Bulletin, 125 (3), 36783.

Cameron, A. Colin, Pravin K. Trivedi, Frank Milne, and JohnPiggott (1988), A Microeconometric Model of the Demandfor Health Care and Health Insurance in Australia, Reviewof Economic Studies, 55 (1), 85106.

Cicchetti, Charles J. and Jeffrey A. Dubin (1994), A Microecon-ometric Analysis of Risk Aversion and the Decision to Self-Insure, Journal of Political Economy, 102 (1), 16986.

Day, Ellen and Richard J. Fox (1985), Extended Warranties, Ser-vice Contracts, and Maintenance Agreements: A MarketingOpportunity? Journal of Consumer Marketing, 2 (4), 7786.

Dekel, Eddie and Suzanne Scotchmer (1999), On the Evolutionof Attitudes towards Risk in Winner-Take-All Games, Jour-nal of Economic Theory, 87 (1), 12543.

Dhar, Ravi and Klaus Wertenbroch (2000), Consumer Choice be-tween Hedonic and Utilitarian Goods, Journal of Marketing

Research, 37 (1), 6071.Dodds, William B., Kent B. Monroe, and Dhruv Grewal (1991),

Effects of Price, Brand, and Store Information on BuyersProduct Evaluations, Journal of Marketing Research, 28(August), 30719.

Dodson, Joe A., Alice M. Tybout, and Brian Sternthal (1978),Impact of Deals and Deal Retraction on Brand Switching,

Journal of Marketing Research, 15 (1), 7281.Dunegan, Kenneth J., Dennis Duchon, and Sidney L. Barton

(1992), Affect, Risk, and Decision Criticality: Replicationand Extension in a Business Setting, Organizational Behav-ior and Human Decision Processes, 53 (3), 33551.

Erdem, Tulin and Michael P. Keane (1996), Decision-Makingunder Uncertainty: Capturing Dynamic Brand Choice Pro-cesses in Turbulent Consumer Goods Markets, MarketingScience, 15 (1), 120.


13/13


Erdem, Tulin, Michael P. Keane, and Baohong Sun (2008), ADynamic Model of Brand Choice When Price and Adver-tising Signal Product Quality, Marketing Science, 27 (6),111125.

Feldman, Roger, Michael Finch, Bryan Dowd, and Steven Cassou(1989), The Demand for Employment-Based Health Insur-ance Plans, Journal of Human Resources, 24 (1), 11542.

Frank, Richard G. (2001), Prescription Drug Prices: Why DoSome Pay More than Others Do? Health Affairs, 20 (2),11528.

Gelfand, Alan E. and Adrian F. M. Smith (1990), Sampling-BasedApproaches to Calculating Marginal Densities, Journal ofthe American Statistical Association, 85 (410), 398409.

Guadagni, Peter M. and John D. C. Little (1983), A Logit Modelof Brand Choice Calibrated on Scanner Data, Marketing Sci-ence, 2 (3), 20338.

Halek, Martin and Joseph G. Eisenhauer (2001), Demography ofRisk Aversion, Journal of Risk and Insurance, 68 (1), 124.

Hammond, J. D., David B. Houston, and Eugene R. Melander(1967), Determinants of Household Life Insurance PremiumExpenditures: An Empirical Investigation, Journal of Riskand Insurance, 34 (3), 397408.

He, Xin, J. Jeffrey Inman, and Vikas Mittal(2008), Gender Jeopardyin Financial Risk Taking, Journal of Marketing Research, 45(4), 41424.

Heilman, Carrie M., Kent Nakamoto, and Ambar G. Rao (2002),Pleasant Surprises: Consumer Response to UnexpectedIn-Store Coupons, Journal of Marketing Research, 39 (2),24252.

Hsee, Christopher K. and Howard C. Kunreuther (2000), TheAffection Effect in Insurance Decisions, Journal of Risk andUncertainty, 20 (2), 14159.

Isen, Alice M. and Nehemia Geva (1987), The Influence of Pos-itive Affect on Acceptable Level of Risk: The Person with aLarge Canoe Has a Large Worry, Organizational Behaviorand Human Performance, 39 (2), 14554.

Isen, Alice M., Thomas E. Nygren, and F. Gregory Ashby (1988),

Influence of Positive Affect on the Subjective Utility ofGains and Losses: It Is Just Not Worth the Risk, Journal ofPersonality and Social Psychology, 55 (5), 71017.

Isen, Alice M. and Robert Patrick (1983), The Effect of PositiveFeelings on Risk Taking: When the Chips Are Down, Orga-nizational Behavior and Human Performance, 31 (2), 194202.

Johnson, Eric J., John Hershey, Jacqueline Meszaros, and HowardKunreuther (1993), Framing, Probability Distortions, and In-surance Decisions, Journal of Risk and Uncertainty, 7 (1),3551.

Khan, Uzma, Ravi Dhar, and Klaus Wertenbroch (2005), Hedonicand Utilitarian Consumption, in Inside Consumption: Fron-tiers of Research on Consumer Motives, Goals, and Desire,ed. S. Ratneshwar and David Glen Mick, London: Routledge,14465.

Levy, H. and H. M. Markowitz (1979), Approximating ExpectedUtility by a Function of Mean and Variance, American Eco-nomic Review, 69 (3), 30817.

Lutz, Nancy A. and V. Padmanabhan (1995), Why Do We ObserveMinimal Warranties? Marketing Science, 14 (4), 41741.

MacInnis, Deborah J. and Linda L. Price (1987), The Role ofImagery in Information Processing: Review and Extensions,

Journal of Consumer Research, 13 (4), 47391.Magnan, Renee E. and Verlin B. Hinsz (2005), Mood, Gender,

and Situational Influences on Risk-Taking Advice for Others,Social Behavior and Personality, 33 (1), 110.

Mancini, Francesco and Amelia Gangemi (2004), Aversion toRisk and Guilt, Clinical Psychology and Psychotherapy, 11(3), 199206.

Mandel, Naomi (2003), Shifting Selves and Decision Making:The Effects of Self-Construal Priming on Consumer Risk-Taking, Journal of Consumer Research, 30 (June), 3040.

Mittal, Vikas V. and William T. Ross (1998), The Impact of Pos-itive and Negative Affect and Issue Framing on Issue Inter-pretation and Risk Taking, Organizational Behavior and Hu-man Decision Process, 76 (3), 298324.

Mossin, Jan (1968), Aspects of Rational Insurance Purchasing,Journal of Political Economy, 76 (4), 55368.

Neslin, Scott A. (2002), Sales Promotion, in Handbook of Mar-keting, ed. Barton A. Weitz and Robin Wensley, London:Sage, 31038.

Okada, Erica M. (2005), Justification Effects on Consumer Choiceof Hedonic and Utilitarian Goods, Journal of Marketing Re-search, 42 (1), 43346.

Padmanabhan, V. (1995), Usage Heterogeneity and ExtendedWarranties, Journal of Economics and Management Strat-egy, 4 (1), 3353.

Padmanabhan, V. and Ram C. Rao (1993), Warranty Policy and

Extended Service Contracts: Theory and an Application toAutomobiles, Marketing Science, 12 (3), 23047.

Purohit, Devavrat and Joydeep Srivastava (2001), Effect of Man-ufacturer Reputation, Retailer Reputation, and Product War-ranty on Consumer Judgments of Product Quality: A CueDiagnosticity Framework, Journal of Consumer Psychology,10 (3), 12334.

Raghubir, Priya and Kim Corfman (1999), When Do Price Pro-motions Affect Pretrial Brand Evaluations? Journal of Mar-keting Research, 36 (2), 21122.

Rao, Ram C. and Niladri Syam (2001), Equilibrium Price Com-munication and Unadvertised Specials by Competing Super-markets, Marketing Science, 20 (1), 6181.

Sanna, Lawrence J. (1999), Mental Simulations, Affect, and Sub-jective Confidence: Timing Is Everything, Psychological Sci-

ence, 10 (4), 33945.Schlesinger, Harris (1999), The Theory of Insurance Demand,in Handbook of Insurance, ed. Georges Dionne, Boston: Klu-wer, 13152.

Shimp, Terence A. and William O. Bearden (1982), Warranty andOther Extrinsic Cue Effects on Consumers Risk Perceptions,

Journal of Consumer Research, 9 (1), 3846.Strahilevitz, Michael and John G. Myers (1998), Donations to

Charity as Purchase Incentives: How Well They Work MayDepend on What You Are Trying to Sell, Journal of Con-sumer Research, 24 (March), 43446.

Sunden, Annika E. and Brian J. Surette (1998), Gender Differ-ences in the Allocation of Assets in Retirement SavingsPlans, American Economic Review, 88 (2), 20711.

Thaler, Richard H. (1999), Mental Accounting Matters, Journalof Behavioral Decision Making, 12 (3), 183206.

Tversky, Amos and Daniel Kahneman (1974), Judgment under Un-certainty: Heuristics and Biases, Science, 185 (4157),112431.

Warranty Week(2005), Extended Warranty Administrators: WhileAuto and PC Manufacturers Have the Top Spots, InsuranceCompanies and Third Party Administrators Grab the Bulk ofthe Pie, http://www.warrantyweek.com, January 19.

Williamson, Janis, Rob Ranyard, and Lisa Cuthbert (2000), RiskManagement in Everyday Insurance Decisions: Evidencefrom a Process Tracing Study, Risk, Decision and Policy, 5(01), 1938.

Warranties Research

Documents

Transcript of Warranties Research