currency of search: how spending time on search is not the same as spending money

13
Journal of Retailing 85 (3, 2009) 245–257 Currency of Search: How Spending Time on Search is Not the Same as Spending Money Ashwani Monga a,,1 , Ritesh Saini b,1,2 a Moore School of Business, University of South Carolina, 1705 College Street, Columbia, SC 29208, United States b College of Business, University of Texas at Arlington, 701 S. West Street, Arlington, TX 76019, United States Abstract Search theories suggest that a decline in search costs increases search behavior. This relationship has been well supported by prior experimental research but not by studies conducted in retail settings. Our review of the literature suggests that this discrepancy might be driven by the fact that prior experiments typically involve money-based search whereas actual search in retail settings is usually time-based. We argue that the currency of search plays a moderating role. We find that when participants spend money on search, a decrease in search costs has a significant effect on search decisions but, when they spend time on search, a decrease in search costs either has a relatively weak effect (Experiment 1) or no effect at all (Experiment 2). Furthermore, this insensitivity in time also emerges for search payoffs (Experiment 3). We also offer evidence for the processes underlying these effects. Our results provide a new lens to examine inconsistencies in the search literature, and present a view of search that is more applicable to the retail context. © 2009 New York University. Published by Elsevier Inc. All rights reserved. Keywords: Consumer search; Time versus money; Judgment and decision making Search is a key component of the decision process in retail settings (Miller 1993), and involves seeking informa- tion to resolve purchase uncertainties (Moorthy, Ratchford, and Talukdar 1997). Consumers make search decisions in either sequential (Schotter and Braunstein 1981, Zwick et al. 2003) or non-sequential (Burdett and Judd 1983) settings. Sequential search is open-ended and the search decision pertains to when search will be terminated. For example, a consumer trying to buy a couch at the lowest price could keep on visiting furni- ture stores, and decide to stop searching as soon as she finds a satisfactory price. In non-sequential search, consumers make search decisions even before commencing search. For exam- ple, the consumer trying to buy a couch could make an a priori decision about the number of stores to visit in order to check prices. What is common to both these kinds of search, however, is a tradeoff between costs and payoffs. Search costs have to be incurred (e.g., spending time to visit stores) in order to achieve potential search payoffs (e.g., finding a lower price). The will- Corresponding author. Tel.: +1 803 777 5918; fax: +1 803 777 6876. E-mail address: [email protected] (A. Monga). 1 The authors contributed equally to this article. 2 Tel.: +1 817 272 2876; fax: +1 817 272 2854. ingness to search refers to the number of stores that one decides to visit. This paradigm follows from Stigler’s (1961) seminal paper in which he analyzed search as an optimization-under- constraints problem; greater search leads to a higher likelihood of success but involves greater costs as well. Although search entails costs, it can lead to a better payoff, such as a lower price for the good. Therefore, a decline in search costs or an increase in search payoffs should increase consumers’ willingness to search. We propose that these fundamental relationships are moderated by the currency of search: time versus money. We experimen- tally show that willingness to search is less sensitive to changes in costs and payoffs when search is conducted by spending time rather than money. Our findings have direct relevance for retail theorists and practitioners. As we discuss in the next section, the effect of search costs on search decisions has been well supported by prior experimental research but the evidence from retail settings is not supportive. Our review of the literature suggests that this discrep- ancy might be driven by the fact that prior experiments typically involve money-based search whereas search in retail settings is usually time-based. We argue that search is likely to occur differently in settings in which money is spent (e.g., paying a real-estate agent to search for home buyers) than in retail settings in which search involves spending time (e.g., visiting different 0022-4359/$ – see front matter © 2009 New York University. Published by Elsevier Inc. All rights reserved. doi:10.1016/j.jretai.2009.04.005

Upload: ashwani-monga

Post on 05-Sep-2016

222 views

Category:

Documents


1 download

TRANSCRIPT

A

rposaum©

K

rtTsossbtaspdpiip

0d

Journal of Retailing 85 (3, 2009) 245–257

Currency of Search: How Spending Time on Search is Not the Same asSpending Money

Ashwani Monga a,∗,1, Ritesh Saini b,1,2

a Moore School of Business, University of South Carolina, 1705 College Street, Columbia, SC 29208, United Statesb College of Business, University of Texas at Arlington, 701 S. West Street, Arlington, TX 76019, United States

bstract

Search theories suggest that a decline in search costs increases search behavior. This relationship has been well supported by prior experimentalesearch but not by studies conducted in retail settings. Our review of the literature suggests that this discrepancy might be driven by the fact thatrior experiments typically involve money-based search whereas actual search in retail settings is usually time-based. We argue that the currencyf search plays a moderating role. We find that when participants spend money on search, a decrease in search costs has a significant effect onearch decisions but, when they spend time on search, a decrease in search costs either has a relatively weak effect (Experiment 1) or no effect at

ll (Experiment 2). Furthermore, this insensitivity in time also emerges for search payoffs (Experiment 3). We also offer evidence for the processesnderlying these effects. Our results provide a new lens to examine inconsistencies in the search literature, and present a view of search that isore applicable to the retail context.2009 New York University. Published by Elsevier Inc. All rights reserved.

ing

itpcoefsWbtir

p

eywords: Consumer search; Time versus money; Judgment and decision mak

Search is a key component of the decision process inetail settings (Miller 1993), and involves seeking informa-ion to resolve purchase uncertainties (Moorthy, Ratchford, andalukdar 1997). Consumers make search decisions in eitherequential (Schotter and Braunstein 1981, Zwick et al. 2003)r non-sequential (Burdett and Judd 1983) settings. Sequentialearch is open-ended and the search decision pertains to whenearch will be terminated. For example, a consumer trying touy a couch at the lowest price could keep on visiting furni-ure stores, and decide to stop searching as soon as she findssatisfactory price. In non-sequential search, consumers make

earch decisions even before commencing search. For exam-le, the consumer trying to buy a couch could make an a prioriecision about the number of stores to visit in order to checkrices. What is common to both these kinds of search, however,

s a tradeoff between costs and payoffs. Search costs have to bencurred (e.g., spending time to visit stores) in order to achieveotential search payoffs (e.g., finding a lower price). The will-

∗ Corresponding author. Tel.: +1 803 777 5918; fax: +1 803 777 6876.E-mail address: [email protected] (A. Monga).

1 The authors contributed equally to this article.2 Tel.: +1 817 272 2876; fax: +1 817 272 2854.

sesaiidri

022-4359/$ – see front matter © 2009 New York University. Published by Elsevier Ioi:10.1016/j.jretai.2009.04.005

ngness to search refers to the number of stores that one decideso visit. This paradigm follows from Stigler’s (1961) seminalaper in which he analyzed search as an optimization-under-onstraints problem; greater search leads to a higher likelihoodf success but involves greater costs as well. Although searchntails costs, it can lead to a better payoff, such as a lower priceor the good. Therefore, a decline in search costs or an increase inearch payoffs should increase consumers’ willingness to search.e propose that these fundamental relationships are moderated

y the currency of search: time versus money. We experimen-ally show that willingness to search is less sensitive to changesn costs and payoffs when search is conducted by spending timeather than money.

Our findings have direct relevance for retail theorists andractitioners. As we discuss in the next section, the effect ofearch costs on search decisions has been well supported by priorxperimental research but the evidence from retail settings is notupportive. Our review of the literature suggests that this discrep-ncy might be driven by the fact that prior experiments typicallynvolve money-based search whereas search in retail settings

s usually time-based. We argue that search is likely to occurifferently in settings in which money is spent (e.g., paying aeal-estate agent to search for home buyers) than in retail settingsn which search involves spending time (e.g., visiting different

nc. All rights reserved.

2 of Ret

ripIlletWpnIisCcnsba

Gblsotsp

tipWmttr

sttewt1oZrn2ceu

r(dSa

eoDmawapaieoi

dm(bcrSa(RwassoiaR

ldfiagteif1c2

46 A. Monga, R. Saini / Journal

etailers or e-tailers). This has consequences for retailers try-ng to be in the consideration set of consumers searching forroducts and services. One interesting implication relates to thenternet context, which has been widely studied in the retailingiterature (Grewal and Levy 2007). Given that search costs areower in the electronic world than in the physical world, retail-rs worry that the Internet leads to higher consumer search and,herefore, more intense competition (Lynch and Ariely 2000).

hile these worries are perfectly understandable, it seems thateople might not be searching very extensively over the Inter-et (Brynjolfsson and Smith 2000). We argue that, because thenternet reduces search costs of time (not money), the increasen search activity will not happen as would be expected fromearch models that have been supported in monetary settings.onsequently, retailers need to be less fearful of lower searchosts on the Internet, and more enthusiastic about the opportu-ities offered by online environments (Lynch and Ariely 2000),uch as the potential to reduce product performance uncertaintyy using various communication practices (Weathers, Sharma,nd Wood 2007).

Implications also arise for store location models (Achabal,orr, and Mahajan 1982). Our results imply that consumers wille relatively more sensitive to monetary aspects of stores (theower price offered by an outlet mall relative to the neighborhoodtore) than to temporal aspects (the time required to visit theutlet mall). We do not suggest that time does not matter; theime of travel will indeed be a cost to consumers. What weuggest instead is that consumers will react more strongly torice differences than to time-of-travel differences.

We next present the literature that motivates our inquiry intohe currency of search. Then, we offer a prediction about its rolen moderating the effect of search costs on search behavior andresent supporting evidence from two laboratory experiments.e then extend our theorizing to search payoffs and find a similaroderating effect in a third experiment; people are less sensitive

o changes in payoffs when the currency of search is time ratherhan money. Finally, we conclude with the implications of ouresults for the theory and practice of retailing.

Time versus money as currency of search

Search is frequently conducted by spending time. Peoplepend time searching inside stores, in traveling from one storeo another, and in searching over the Internet. The prevalence ofime-based searching is evident from field research. When wexamined the retail situations that are studied in this literature,e found that they overwhelmingly relate to expenditures of

ime. When consumers search for automobiles (Moorthy et al.997; Punj and Staelin 1983; Srinivasan and Ratchford 1991)r generally for products in the marketplace (Pratt, Wise, andeckhauser 1979), they usually spend their time. And when

esearchers study the effects of lower search costs on the Inter-et relative to conventional markets (Brynjolfsson and Smith

000), the costs refer to the time that consumers spend. Thisonsideration of the costs of time rather than money is also inher-nt in the measures that are used. In field research, researcherssually measure search costs via questions that directly assess

idrs

ailing 85 (3, 2009) 245–257

espondents’ own valuation of the time required to searchSrinivasan and Ratchford 1991) or that indirectly assess respon-ents’ opportunity costs of time from other indicators (Punj andtaelin 1983). That is, field studies seem to consider search asn activity that involves expenditure of time.

In stark contrast to the field research, our review of searchxperiments revealed an overwhelming reliance on the currencyf money. Barring rare exceptions (Smith, Venkatraman, andholakia 1999), search costs are operationalized in terms ofoney. This is true for experimental economics research (Cox

nd Oaxaca 1989; Kogut 1992; Schotter and Braunstein 1981) asell as experimental consumer research (Diehl 2005; Srivastava

nd Lurie 2001; Zwick et al. 2003). The use of money is appro-riate because it enables easy quantification of search costss researchers focus on the phenomena that they are study-ng. However, from the perspective of ecological validity, thesexperiments seem disconnected from the reality of consumersften spending their time rather than money in order to searchn retail settings.

This disconnect is especially consequential because, as weiscovered from a comparison of results from several experi-ents (manipulating monetary search costs) and field studies

measuring temporal search costs), there is an inconsistencyetween the two. The theoretical prediction of lower searchosts leading to higher search behavior (Stigler 1961) has beenepeatedly demonstrated in experimental studies (Kogut 1992;chotter and Braunstein 1981, see Davis and Holt 1993 forreview). In contrast, the support from field research is rare

Moorthy et al. 1997). Consider the findings of Putrevu andatchford (1997). Although search costs had a significant effecthen they were measured in terms of opportunity costs from

n economic perspective (e.g., wage rate), their effect was notignificant when they were measured in terms of felt time pres-ure, which represented the psychological cost of time. Even inther field studies, the effect of search costs on search behav-or has been found to be either only marginally significant (Punjnd Staelin 1983), or completely non-significant (Srinivasan andatchford 1991).

Our review of the literature on price dispersion further high-ights the discrepancy in results. Theoretically, if search costsecrease, the increase in search behavior should deter sellersrom offering discrepant prices and, therefore, price dispersionn the market should decrease. For example, because search costsre believed to be lower over the Internet, search models sug-est that price dispersion on the Internet should be lower thanhat in comparable conventional markets (Bakos 1997). Thisffect on price dispersion is evident from experiments involv-ng money (Cason and Friedman 2003) but, once again, notrom field research (Brynjolfsson and Smith 2000; Pratt et al.979). For instance, the price dispersion in online markets isomparable to that in offline markets (Brynjolfsson and Smith000).

We clearly recognize that this inconsistency between exper-

mental and field results may be driven by the numerousifferences between the two settings and not just by the cur-ency of search (i.e., money in experiments and time in fieldtudies). However, these results do underscore the importance

of Ret

ohc

pI(i(fatccpcoFcjTv

erw–ffwntaritimcpsl(

dftFetstnfi

aottbpSvBitgmothe

nWetiat

pttdasimletfassco

Howwtr

tc

A. Monga, R. Saini / Journal

f isolating the effect of currency and investigating whether andow this variable can influence the relationship between searchosts and search behavior.

Currency of search will moderate the effect of search costs

Currency of search need not be an issue if one considers theerspective of time and money being economically equivalent.n his economic theory about the allocation of time, Becker1965) equates the value of time to its opportunity cost, whichs usually assumed to be one’s wage rate. Similarly, Graham1981) discusses time as a straight line from the past into theuture that can be separated into discrete units, and allocated justs money is. In other words, time is believed to be a resourcehat is akin to money. However, other researchers suggest a moreomplex view of time. From the perspective of economics, timean indeed be conceptualized as an objective resource; from theerspective of sociology, it refers to a social construction in aultural context in which it is valued; and from the perspectivef psychology, it is a subjective perception (McDonald 1994).or instance, Marmorstein, Grewal, and Fishe (1992) found thatonsumers’ subjective value of their time is influenced by notust wage rates, but also the perceived enjoyment of shopping.hese insights have spawned further research on how consumersiew the notion of time, especially in comparison to money.

Research on time–money differences has not studied theffect of search costs and payoffs, as we do in the currentesearch. However, in our prior work (Saini and Monga 2008),e have employed the context of search to study how heuristicsrules of thumb that simplify decision making – are used dif-

erently in time than in money. Because our experiments wereocused on examining heuristics rather than the search process,e had held constant the magnitude of relevant search determi-ants (i.e., search costs), but manipulated unrelated informationhat could be used as a heuristic (e.g., value of an arbitrarynchor). We demonstrated that when decisions related to timeather than money, people displayed a higher use of the anchor-ng heuristic; they were more prone to considering the value ofhe irrelevant anchor. So, if time makes people more sensitive torrelevant information (e.g., anchors), does it suggest that it will

ake people more sensitive to even relevant information (e.g.,hanges in search costs)? Or, does it suggest that time will makeeople less sensitive to relevant information? A host of researchuggests that the latter possibility is more likely; people are lessikely to respond to changes in information that relates to timevs. money).

Although the disutility from spending money is usually airect function of the magnitude of the money spent, it has beenound that the disutility from spending time is less sensitive tohe magnitude of the time spent (Elster and Loewenstein 1992;redrickson and Kahneman 1993). In general, when experiencesxtend over time, people are not adept at judging them based onhe duration of those experiences. Apart from rare situations,

uch as when individuals can easily compare magnitudes ofime (Ariely and Loewenstein 2000), people exhibit a durationeglect. They judge an experience based on aspects such as thenal intensity of an experience, rather than its duration (Varey

Ttpm

ailing 85 (3, 2009) 245–257 247

nd Kahneman 1992). Ebert and Prelec (2007) extend the ideaf duration neglect to time intervals that extend from now intohe future, and find that choices are not sufficiently sensitive toime. In other research by Okada and Hoch (2004), time haseen labeled as being more ambiguous than money. For exam-le, although the value of an hour might seem low on a leisurelyunday afternoon but high on a hectic Monday morning, thealue of a dollar does not change as easily across situations.ecause of such ambiguity, Okada and Hoch (2004) find that

t is much easier for people to rationalize their expenditures ofime, but not of money. Finally, research by Soman (2001) sug-ests that most people are not good at accounting for differentagnitudes of time because they do not routinely keep account

f time the way they keep account of money. He demonstrateshat people are not very sensitive to the amount of time theyave spent in the past, even though they carefully consider pastxpenditures of money.

It is important to note that the above stream of research doesot suggest that expenditures of time are completely ignored.hen people spend time, they do notice the presence of that

xpenditure. What they ignore, however, is the magnitude ofhat expenditure, leading to a lack of sensitivity toward changesn the magnitude. So while people respond to small versus bigmounts of money in a relatively precise fashion, their responseo changes in time is not as precise.

The above arguments have important consequences for theremise that a decline in search costs increases search, a premisehat has been long established in the theoretical and experimen-al literature on search (Davis and Holt 1993; Stigler 1961). Asiscussed earlier, established search theories do not considerny role for the currency in which search is conducted, andearch experiments supporting these theories have overwhelm-ngly relied on the currency of money, implicitly assuming thatoney can be a surrogate for all expenditures. However, the

iterature overviewed in the previous paragraphs suggests thatven though people consider monetary information carefully,hey disregard the magnitude of temporal information. There-ore, if search involves spending time instead of money, peoplere likely to be less sensitive to changes in the magnitude ofearch costs. That is, the extent to which they are willing toearch (e.g., go to different stores to find a lower price) will nothange much with the time that it takes to search. This leads tour first hypothesis:

1. Currency of search will moderate the effect of magnitudef search costs on people’s willingness to search. Specifically,hen the currency is money, lower (vs. higher) search costsill result in higher willingness to search. When the currency is

ime, this effect of search costs on willingness to search will beelatively weaker.

Following our earlier discussion, H1 is predicated on the ideahat people are relatively indifferent to the magnitude of searchosts, when search involves spending time rather than money.

his proposed process is consistent with research on prospec-

ive duration judgments, in which it has been demonstrated thateople show a lack of attentional focus toward temporal infor-ation (Block and Zakay 1997; Zakay 1998). Therefore, our

2 of Ret

nu

Hpsmb

euiptdaaa

O

ssstPn1

brbnmoubrir

rtdtwraoaEaa2

itrtedt2mtp

D

ohns

P

istiotppr

rpocospedfa

wec

tpnp

48 A. Monga, R. Saini / Journal

ext hypothesis offers a direct test of the process that we believenderlies lack of sensitivity to changes in search costs.

2. Currency of search will determine the extent to whicheople rely on search costs as a basis for their willingness-to-earch decisions. Specifically, when the currency is time (vs.oney), people will be less likely to report search costs as a

asis for their willingness-to-search decisions.

We test the above hypotheses in two experiments. The firstxperiment tests the focal effect (H1) in a type of set-up that issed in experimental economics research. In the second exper-ment, we test the focal effect (H1) as well as the underlyingrocess (H2) in a service context; participants are provided withhe situation of choosing a moving company and asked to makeecisions. Later on, we extend our theorizing to search payoffsnd test it in a third experiment in a product context; participantsre provided with the situation of buying a digital camera andsked to make decisions.

Experiment 1

verview

In this experiment, we test our core prediction (H1) thattates that the currency of search will moderate the effect ofearch costs on willingness to search. We consider a sequentialearch setting in which search is open-ended and the decisiono terminate it is contingent on results from prior search effort.articipants keep on searching for the best payoff but can termi-ate search at any stage in a Bingo-based game (Cox and Oaxaca989).

The experimental apparatus was a Bingo cage that containedalls numbered from 1 to 50. In each round, a number would beandomly drawn. The payoff would be the highest number drawnefore the subject decides to stop (i.e., if x were the highestumber drawn, $x would be the earnings). The costs would be aultiple of the number of rounds played (i.e., if y were the cost

f one round, n × y would be the cost of n rounds). This set-p represented the relationship between search costs and searchehavior very well. If one chose to search more (i.e., play moreounds), the likelihood of a higher payoff (i.e., higher number)ncreased but so did the search costs (i.e., cost of playing theounds).

The cost of playing each round was either monetary or tempo-al. In the case of money, search costs referred to a dollar amounthat one would need to pay ($1 or $4 in the two search-cost con-itions) whereas, in the case of time, search costs referred tohe amount of time that one would have to spend on data-entryork (5 min or 20 min). Therefore, the implicit hourly wage

ate that we used in our manipulations was $12 (5 min ∼ $1nd 20 min ∼ $4). This wage rate was very similar to the ratef $12.50 used by Okada and Hoch (2004) to maintain equiv-lence between the time and money conditions (e.g., in their

xperiment 1, 4 h of data-entry work was the time condition,nd $50 was the money condition). Our pretest also revealedsimilar wage rate (M = $11.8; SD = 5.6); this involved asking7 participants, who were not part of the main experiment, to

cpdh

ailing 85 (3, 2009) 245–257

ndicate the minimum dollar amount that they would be willingo accept in return for one hour of simple data-entry work. Byelying on the notion of a wage rate, we ignore other factorshat might influence the subjective value of time (Marmorsteint al. 1992). But, we do maintain economic parity between con-itions, and remain consistent with prior time–money researchhat has relied on wage-rate equivalence (e.g., Okada and Hoch004). This equivalence is, however, not critical to our experi-ent because our prediction is not about a main effect between

ime and money, but about an interaction effect: how sensitiveeople are to changes in time versus changes in money.

esign

A between-subjects design was used in which both currencyf search (time vs. money) and magnitude of search cost (low vs.igh) were manipulated. The dependent variable was the willing-ess to search—the number of rounds played before terminatingearch.

rocedure

Sixty-three undergraduate students participated in this exper-ment in exchange for partial course credit and a chance to winome money. The study was conducted in multiple sessions onhe same day. Each session began with the experimenter explain-ng how the modified Bingo game would be played. The conceptsf payoff and cost were explained via a PowerPoint presenta-ion. In order to make participants have a real stake in the searchrocess, it was announced that at the end, one randomly chosenarticipant would actually pay the costs of playing rounds andeceive the payoff from those rounds.

Participants were told that if they chose not to play anyounds, there would be no payoffs and no costs. If they chose tolay one or more rounds, their payoff would be the dollar amountf the highest number that is drawn in those rounds, and theirost would be the cost of each round multiplied by the numberf rounds they chose to play. Each participant was told that if/he were randomly chosen in the end, s/he would receive theayoff but would also have to incur the cost. After the verbalxplanation, the experimenter demonstrated how balls would berawn and how earnings might change with each round. Apartrom making the workings of the game clear, the purpose waslso to communicate that no artifice was involved.

The experimenter then handed out the questionnaires. Theritten instructions repeated what the experimenter had already

xplained and provided information specific to the experimentalondition that the participant was randomly assigned to.

Before the first round, the experimenter gave participantshe option to terminate search (i.e., with zero costs and zeroayoffs). Those who decided to do so returned their question-aires. Then one number was drawn from the bingo cage andarticipants wrote, in two separate columns, their earnings and

osts after the first round. After the first round, some morearticipants terminated search and the experimenter continuedrawing a Bingo ball in successive rounds till all participantsad terminated search and returned their questionnaires. While

of Ret

rtrpIp

R

sccstator

5cmd(tiFitth(

c5tc5tm

Fc

cc

D

isasWtttmttstsacIam

O

fwsba

A. Monga, R. Saini / Journal

eceiving the questionnaires, the experimenter checked themo ensure that earnings and search costs accumulated till thatound were honestly reported. Finally, one randomly chosenarticipant actually received the payoffs and incurred the costs.n addition, all participants received partial course credit forarticipation.

esults

Our experimental manipulations ensured a 75% decline inearch costs such that the money and time costs in the lowondition ($1 or 5 min) were one-fourth of those in the highondition ($4 or 20 min). However, we expected this decline inearch costs to have a greater impact in the money conditionshan the time conditions. To test this prediction, we employedbetween-subjects analysis in which willingness to search was

he dependent variable, and currency of search and magnitudef search cost were the independent variables. Fig. 1 depicts theesults of the Analysis of Variance.

There was a main effect of magnitude of search cost (F(1,9) = 91.8, p < .001) such that those in the low search costondition (M = 5.9) were willing to search more (i.e., playore rounds) compared to those in the high search cost con-

ition (M = 2.6). There was also a main effect of currencyF(1, 59) = 5.1, p < .05) such that those in the money condi-ion (M = 4.7) were willing to search more compared to thosen the time condition (M = 3.9). However, as is clear fromig. 1, these main effects are qualified by the second-order

nteraction that is pertinent to our prediction. The magni-ude × currency interaction (F(1, 59) = 4.7, p < .05) confirmedhat the willingness-to-search difference between the low and theigh search cost conditions was significantly greater for moneyM = 4.1) than for time (M = 2.6).

Planned contrasts revealed a significant increase in the moneyondition such that willingness to search was higher (F(1,9) = 65.8, p < .001) when search cost was low (M = 6.7) ratherhan high (M = 2.6). An increase also emerged in the time

ondition such that willingness to search was higher (F(1,9) = 28.9, p < .001) when search cost was low (M = 5.2) ratherhan high (M = 2.6). However, as indicated by the significant

agnitude × currency interaction, the increase in the money

ig. 1. Willingness to search (for high-value bingo chip) as a function of searchurrency and search cost.

w(tdca

m(opittuamobc

ailing 85 (3, 2009) 245–257 249

ondition was significantly larger than the increase in the timeondition.

iscussion

In this experiment, we employed a sequential search settingn which search was open-ended and participants terminatedearch at the point beyond which they did not want to searchny longer. The results support the prediction that currency ofearch moderates the effect of search costs on search decisions.

hen we experimentally induced a 75% decline in search costs,he increase in willingness to search was much weaker in timehan in money. It is important to note that, in the time condi-ion, even though participants showed insensitivity to different

agnitudes of search costs, they did treat time as a cost ratherhan as something that they can freely spend. If participants hadhought of time as simply being less valuable than money, theyhould have chosen to search more when they were spendingime than when they were spending money, but they did not. Inum, our results show that people do treat time as a cost justs they treat money as a cost. However, the magnitude of thoseosts matters less in the case of time than in the case of money.n the next experiment, we examine this idea once again, as wells the underlying process, in a service context: searching for aoving company.

Experiment 2

verview

In Experiment 1, we employed a sequential search setting andound support for our core prediction (H1). In this experiment,e examine whether this effect replicates in a non-sequential

earch setting in which people make search decisions in advance,efore commencing search. In addition, this experiment alsoims to examine the process (H2). Specifically, we examinehether people are less likely to consider search costs of time

vs. money) when they decide on the extent of their search. Toest this, we asked participants to explain the reasons for theirecision. If time (vs. money) participants attend less to searchosts, they are less likely to mention the use of search costs inrriving at their decision.

The situation that we used was a modified version of aoving-company scenario that we have used in earlier research

Saini and Monga 2008). There, we had manipulated a menuf different choice options because our focus was on the com-romise heuristic rather than on search costs. Given our focusn the current research, we manipulated search costs instead. Inhe scenario, one needed to spend either time or money in ordero receive estimates from different moving companies. The sit-ation clearly represented the relationship between search costs

nd search behavior. If one chose to search more (i.e., invite moreoving companies for estimates), the likelihood of a desirable

utcome (i.e., finding a cheaper moving company) increasedut so did the search costs (i.e., cost of inviting the movingompanies).

2 of Retailing 85 (3, 2009) 245–257

D

ovwt

P

ept

ucscApttc

R

R

8

Fc

chdaatr

soamtwcsmfil(ipt

cuw9tcb(

R

tt

50 A. Monga, R. Saini / Journal

esign

A between-subjects design was used in which both currencyf search (time vs. money) and magnitude of search cost (lows. high) were manipulated. The dependent variable was theillingness to search—the number of moving companies that

he participants wanted to invite.

rocedure

Ninety-seven undergraduate students participated in thisxperiment in exchange for partial course credit. In the followingaragraphs, underlining highlights the four conditions to whichhe participants were randomly assigned.

Imagine that you are moving to a different location and havedecided to hire a moving company. Because you have nevertried movers before, you ask a knowledgeable friend for help.He picks up the Yellow Pages and tells you the names of50 moving companies that provide a good level of service.He adds that the prices could vary a lot—anywhere from$500 to $1000 for the amount of stuff you have. He thereforesuggests that you randomly choose some of these 50 movingcompanies, get them to your apartment so that they can giveyou their price estimates, and then pick the one that is thecheapest.

The problem that you now face is to decide how many youshould get home for an estimate. To get the cheapest rate, thebest thing to do would be to let each of 50 companies visityour apartment and give you an estimate so that you may pickthe cheapest one. However, there is a cost involved in termsof the amount of money (time) you spend on this activity.Each moving company will charge $5/$40 (take 30 min/4 h)to inspect your stuff and provide an estimate.

Out of the 50 moving companies that you are considering,how many are you going to get over for price estimates?(Please provide one number, not a range.)

I will ask moving companies to come over and providetheir price estimates.

For these manipulations, a wage rate of $10 per hour wassed to achieve round numbers. Therefore, the time and moneyonditions are equivalent (i.e., 30 min = $5; 4 h = $40) and wetudy how participants respond to an identical decrease in searchosts (i.e., 30 min is one-eighth of 4 h; $5 is one-eighth of $40).fter participants had indicated the number of moving com-anies, they were asked to turn to the next page and then givenwo minutes to explain the thought process that led to the answerhey wrote. Finally, they were debriefed and given partial courseredit for participation.

esults

esults for H1As is clear from the search costs we imposed, there was an

7.5% reduction in costs from the high condition to the low

tsts

ig. 2. Willingness to search (for cheap moving company) as a function of searchurrency and search cost.

ondition. However, we expected this decline in search costs toave a higher impact in the money condition than the time con-ition. To test this prediction, we employed a between-subjectsnalysis in which willingness to search was the dependent vari-ble, and currency of search and magnitude of search cost werehe between-subjects independent variables. Fig. 2 depicts theesults of the Analysis of Variance.

There was no effect of currency (F(1, 93) = .02, p > .80)uch that, on average, the willingness to search (i.e., numberf moving companies) was comparable in the money (M = 4.6)nd the time (M = 4.5) conditions. There was a main effect ofagnitude of search cost (F(1, 93) = 17.9, p < .001) such that

hose in the low search cost condition (M = 5.8) had a higherillingness to search compared to those in the high search

ost condition (M = 3.4). This main effect was qualified by theecond-order interaction that is pertinent to our prediction. Theagnitude × currency interaction (F(1, 93) = 6.7, p = .01) con-rmed that the willingness-to-search difference between the

ow and the high search cost conditions was greater for moneyM = 3.8) than for time (M = .9). That is, even though the declinen search costs was exactly 87.5% for both currencies, partici-ants in the money condition were more sensitive to this declinehan those in the time condition.

Planned contrasts revealed that willingness to search signifi-antly increased with a decline in monetary search costs, but wasnaffected by a decline in temporal search costs. Specifically, theillingness to search in the money condition was higher (F(1,3) = 23.1, p < .001) when search cost was low (M = 6.5) ratherhan high (M = 2.7) but the willingness to search in the timeondition was not statistically different (F(1, 93) = 1.4, p > .24)etween the low search cost (M = 5.0) and the high search costM = 4.1) conditions.

esults for H2To gain additional insights into the process leading to the

ime–money differences we observed, we analyzed the cogni-ive responses—the comments that the respondents wrote about

heir thought process. Our prediction about lower search-costensitivity in time (vs. money) is based on the underlying processhat people disregard relevant information when the currency ofearch is time rather than money. If this is true, respondents’

of Ret

cttitjrTiaaiscwm

(lactccdcst

D

oosiIOmnfcm1feit

twidttft

awm

isamlaftctoatri

tatsopwtets

losowpTclpw(rt1ttetct

A. Monga, R. Saini / Journal

omments should reveal a greater tendency to ignore informa-ion about search costs when they are making decisions relatedo spending time rather than money. Therefore, we asked twondependent judges who were blind to the experimental condi-ions to code participants’ responses. It was explained to theudges that search-cost processing is reflected if participantsely on the search costs that are mentioned in the situation.hey coded some responses as “0,” indicating that the partic-

pant had not taken search costs into consideration (e.g., “Find 2nd pit them against each other,” or “Didn’t feel it necessary tosk a lot of companies to give an estimate.”), and some as “1,”ndicating that the participant had taken search costs into con-ideration, even if only to a limited degree (e.g., “30 min × 5ompanies = more than enough” or “I chose 10 because thatould be $50 for estimates—I wouldn’t want to spendore.”).Out of the 97 responses, the two judges agreed on 93

r = .91, Cohen’s κ = .91). They then discussed the four dissimi-ar responses and arrived at a consensus. We used the mutuallygreed list of responses to analyze 49 participants in the timeondition and 48 participants in the money condition. Consis-ent with our theorizing, the proportion of participants who hadonsidered search costs was only 34.7 % (17 out of 49) in the timeondition but 58.3% (28 out of 48) in the money condition. Thisifference was statistically significant (z = −2.33, p < .01), indi-ating that participants were indeed more prone to disregardingearch-cost information if search involves spending time ratherhan money.

iscussion

The current experiment employed a different setting from thene used in Experiment 1 but yielded the same result: currencyf search moderates the effect of search costs on search deci-ions. When we lowered the search costs by 87.5%, the increasen willingness to search was much weaker in time than in money.n fact, the increase was significant only for money, not for time.ne discrepancy from the earlier experiment was that Experi-ent 1 revealed a main effect for time but Experiment 2 did

ot. Although we cannot be sure, this might be due to the dif-erent settings that we employed in the two experiments. As aonjecture that future research could examine, maybe people areore sensitive to time in sequential-search settings (Experiment

) than in non-sequential ones (Experiment 2). But, given theocus of the current research, the absence or presence of a mainffect of time does not deny the key interaction that we foundn both experiments—that sensitivity to search costs is lower inime than in money.

As is true for Experiment 1, our results cannot be explained inerms of people treating time as less valuable than money. If thisere true, participants ought to have searched, on average, more

n time than in money, which they did not. What participants didiffer on was the extent to which they responded to changes in

he costs of time and of money; they expressed insensitivity inhe case of time, but not money. We also demonstrated supportor the hypothesized process. When the currency was time, par-icipants did not report the use of search costs in decision making

r

mu

ailing 85 (3, 2009) 245–257 251

s much as they did when the currency was money. This explainshy participants in time (vs. money) were less influenced by theagnitude of search costs.We conducted a follow-up experiment to gain further insights

nto the underlying process. Specifically, if the process for insen-itivity to temporal search costs is just as we propose, can welter that process to increase the sensitivity and make time seemore like money? We earlier argued that responses to time are

ess precise because time is ambiguous (Okada and Hoch 2004)nd people are unable to do accounting for time the way they door money (Soman 2001). If this is true, then making the value ofime more concrete would make it easier for people to account forhanges in its magnitude. For instance, Soman (2001) showedhat although people do not usually consider past expendituresf time, they do account for them when they are first informedbout the value of time in terms of a wage rate. Consistent withhis, our follow-up experiment primed participants with a wageate to see whether that makes people more sensitive to changesn search costs of time.

We used the same low and high search-cost conditions ofime as used in the main moving-company experiment. We alsodded two more conditions of time that were identical excepthat participants responded to a prime before responding to thecenario. Specifically, participants were told that, according tone estimate, students in the United States get paid around $10er hour of work. Then, on a 7-point scale, they respondedhat they themselves thought (1 = students earn a lot lower

han $10; 4 = students earn about $10 per hour; 7 = studentsarn a lot higher than $10). After indicating their responses tohis question (M = 4.5), they proceeded to the moving-companycenario.

The analysis involved a between-subjects design with twoevels of magnitude of search cost (low vs. high) and two levelsf wage prime (absent vs. present), which we tested with 110tudent participants. An ANOVA revealed that the main effectf prime was not significant (F(1, 106) = .08, p > .77). Thereas a main effect of magnitude of search cost (F(1, 106) = 13.5,< .001) that was qualified by the key second-order interaction.he magnitude × prime interaction (F(1, 106) = 4.5, p < .05)onfirmed that the willingness-to-search difference between theow and the high cost conditions was greater when prime wasresent (M = 2.4) rather than absent (M = .6). Specifically, theillingness to search in the prime-present condition was higher

F(1, 106) = 16.2, p < .001) when search cost was low (M = 5.6)ather than high (M = 3.2), but the willingness to search inhe prime-absent condition was not statistically different (F(1,06) = 1.2, p > .25) between the low search cost (M = 4.8) andhe high search cost (M = 4.2) conditions. Therefore, time par-icipants behaved more like the money participants of the mainxperiment, when they were first primed with the value of theirime. That is, when the value of time was primed to be moreoncrete rather than ambiguous (Okada and Hoch 2004), par-icipants did engage in accounting of time (Soman 2001); they

esponded differently to different search costs.

This follow-up experiment, coupled with the results of theain experiment, provides robust evidence for the process that

nderlies the moderating influence of currency of search on the

2 of Ret

ems

parmrtpacstwesd

amspsbtppipctt

Howwtb

escBcmnu

Hp

tmb

O

sseamtFtewli

dcattastahHw

tsbttifwt

D

wd(To

52 A. Monga, R. Saini / Journal

ffect of search costs on search decisions. We now examine theoderating influence that currency can have on the effect of

earch payoffs on search decisions.

Currency of search will moderate the effect of searchpayoffs

In Experiments 1 and 2, we kept the payoffs constant (i.e.,otential benefits from search were the same in all conditions)nd demonstrated that people are relatively insensitive to aeduction in search costs, when the currency is time rather thanoney. By analyzing participants’ written responses, we further

evealed a lack of focus on search costs. It is possible that, ratherhan attending to this relevant information about costs, partici-ants attended to more unrelated information in order to arrivet their decisions. For instance, a decision to invite three movingompanies might have been based on a heuristic of always con-idering three options (Saini and Monga 2008). Irrespective ofhe type of other information that participants were relying on,hat is clear is that they were ignoring information about a rel-

vant determinant of search decisions: search costs. However,earch costs are not the only relevant determinants of searchecisions.

Search theory suggests that people treat a search situation asn optimization-under-constraints problem in which they try toaximize the potential for search payoffs while minimizing the

earch costs (Stigler 1961). Therefore, when our experimentalarticipants show disregard for a change in the magnitude ofearch costs, they are also displaying a disregard for the tradeoffetween search costs and payoffs. Given that this cost-payoffradeoff can be altered not only by varying costs but also payoffs,eople are also likely to demonstrate insensitivity to changes inayoffs when the currency of search is time. That is, if searchnvolves spending time instead of money, the extent to whicheople are willing to search (and incur search costs) will nothange much with the size of the search payoff (e.g., whetherhe payoff is getting a product that is only slightly better, or onehat is a lot better). This leads to our third hypothesis:

3. Currency of search will moderate the effect of magnitudef search payoffs on people’s willingness to search. Specifically,hen the currency is money, higher (vs. lower) search payoffsill result in higher willingness to search. When the currency is

ime, this effect of search payoffs on willingness to search wille relatively weaker.

As detailed in our discussion about search costs, and asvidenced by the results from our earlier experiments, peoplehow a lower consideration of relevant information (e.g., searchosts), when the currency of search is time rather than money.y the same token, we predict that people will not adequatelyonsider the relevant information of search payoffs when theyake decisions regarding willingness to search in time. Our

ext hypothesis offers a direct test of the process that we believenderlies lack of sensitivity to changes in search payoffs.

4. Currency of search will determine the extent to whicheople rely on search payoffs as a basis for their willingness-

ic

a

ailing 85 (3, 2009) 245–257

o-search decisions. Specifically, when the currency is time (vs.oney), people will be less likely to report search payoffs as a

asis for their willingness-to-search decisions.

Experiment 3

verview

In Experiments 1 and 2, we observed that the currency ofearch moderates the impact of search costs on willingness toearch. In the current experiment, we examine whether this mod-rating impact of currency extends to search payoffs as well (H3nd H4). Examining changes in payoffs requires a different for-at from the one we employed in our first two experiments. In

hose experiments, the potential for payoffs was probabilistic.or instance, inviting more moving companies only increased

he chance of getting a cheaper moving company; it did notnsure it. To manipulate payoffs in a more definitive manner,e make them deterministic in that greater search necessarily

eads to a better payoff, and then test whether a fixed increasen payoff changes the willingness to search.

We borrow this approach of deterministic payoffs fromecision-making research that studies consumer search in theontext of relative savings. Thaler (1980) suggests that peoplere more willing to extend search for a $5 saving on a $25 radiohan on a $500 TV because, as a proportion of the product price,he saving is higher on the former than the latter. Kahnemannd Tversky (1984) found empirical support for this relative-avings idea using the classic “jacket and calculator” problem;hey showed that, given a potential saving of a specific dollarmount, people are more eager to save it on a low-priced (vs.igh-priced) product. Our next experiment is set in this tradition.owever, rather than focusing on the effect of relative savings,e focus on the effect of absolute payoffs in the search process.The situation used was that of a customer visiting a store

o purchase a 1.3 MP (Mega pixel) camera. Upon reaching thetore the customer learns that there is a possibility of receiving aetter camera of the same brand. Participants are asked to stateheir maximum willingness to search for the better camera inerms of either money or time. Our prediction was that a higherncrease in search payoffs – from 1.3 MP to 5.1 MP rather thanrom 1.3 MP to 2.4 MP – would lead to a weaker change in theillingness to search when the search was being conducted in

he currency of time rather than money.

esign

To test H3 and H4, a between-subjects design was used inhich participants were randomly assigned to one of four con-itions. The two manipulated factors were currency of searchtime vs. money) and magnitude of search payoff (low vs. high).he dependent variable was the willingness to search which wasperationalized as the amount of time or money that the partic-

pants were willing to spend in order to receive the payoff of aamera upgrade.

Although our design was a between-subjects one, we alsodded a within-subjects element to test H3. We asked those in

of Ret

tscl

sawctmtwwetttbwcb(tc

P

ic

ewlnttp

R

R

otcesmdstlt

t(tprior research (Okada and Hoch 2004), we transformed dollarsinto z-scores by subtracting the mean of the entire money datafrom each data point and then dividing by the standard deviationof the entire money data. In a similar fashion, the minutes were

A. Monga, R. Saini / Journal

he low-payoff condition to later indicate their willingness toearch for the high-payoff condition, and those in the high-payoffondition to later indicate their willingness to search for theow-payoff condition.

Although the between-subjects analysis would have beenufficient to test our predictions, the additional within-subjectsnalysis was useful because of two reasons. The first reasonas the kind of dependent variable that we used. In the time

onditions, willingness to search referred to the amount ofime that participants were willing to spend whereas, in the

oney conditions, the same measure referred to the moneyhat participants were willing to spend. As we explain later,e did create standardized scores for time and money, in lineith prior time–money research (Okada and Hoch 2004). How-

ver, the within-subjects analysis helped us employ a measurehat is already standardized: percentages. Specifically, givenwo responses from the same participant, we could look athe percentage change in willingness to search brought abouty a change in search payoffs. Another reason for having theithin-subjects design was that, in spite of randomization across

ells, the between-subjects results might have been influencedy individual differences in how time is subjectively valuedMarmorstein et al. 1992). The within-subjects analysis wouldell us whether, even for the same individual, sensitivity tohanges is lower in time than in money.

rocedure

Eighty-four undergraduates participated in this experimentn exchange for partial course credit. The scenario for the twoonditions of time was as follows:

You have decided to buy a Digital Camera and you knowthe specific brand you want. You go to your neighborhoodelectronics store and consider buying the brand’s 1.3 MP (1.3Mega Pixel) camera. Just when you are about to pick up the1.3 MP camera for purchase, the salesperson tells you thatyou might want to consider the 2.4 MP (5.1 MP) camera of thesame brand. This better-quality camera is priced the same butyou would need to wait in the store while the salesperson goesto the warehouse to get it. As he heads toward the warehouseto get the 2.4 MP (5.1 MP) camera, you are wondering howmuch more you would be willing to wait, over and abovethe time you have already spent on shopping for the 1.3 MPcamera.

The maximum amount of additional time that I am willingto wait in the store in order to receive the 2.4 MP (5.1 MP)camera is Min.

The scenario for the two conditions of money was as follows:

You have decided to buy a Digital Camera and you knowthe specific brand you want. You go to your neighborhoodelectronics store and consider buying the brand’s 1.3 MP (1.3

Mega Pixel) camera. Just when you are about to pick upthe 1.3 MP camera for purchase, the salesperson tells youthat you might want to consider the 2.4 MP (5.1 MP) cameraof the same brand. This better-quality camera is, of course,

Fc

ailing 85 (3, 2009) 245–257 253

priced higher. As he checks into his computer to find out theprice of the 2.4 MP (5.1 MP) camera, you are wondering howmuch more you would be willing to pay, over and above themoney that you would be paying for the 1.3 MP camera.

The maximum amount of additional money that I am will-ing to pay in order to receive the 2.4 MP (5.1 MP) camera is$ .

After participants had indicated their willingness to search inither time or money, they were asked to turn to the next pagehere they were given 2 min to explain the thought process that

ed to the answer they wrote. They were then asked to turn to theext page where they indicated their willingness to search forhe other payoff condition. Finally, they were debriefed abouthe real purpose of the study and given partial course credit forarticipating in the experiment.

esults

esults for H3 (between-subjects)Given the baseline of 1.3 MP, search could lead to a payoff

f either 2.4 MP or 5.1 MP. Given our theorizing, we predictedhat the increase in payoff would have a lower impact when theurrency is time rather than money. To test this prediction, wemployed a between-subjects analysis in which willingness toearch was the dependent variable and currency of search andagnitude of search payoff were the between-subjects indepen-ent variables. As can be seen from Fig. 3, the pattern of meanseemed consistent with H3. Specifically, for money, willingnesso search seemed higher when search payoff was high rather thanow, but, for time, willingness to search seemed similar acrosshe two conditions.

An Analysis of Variance cannot however be conducted withhe raw means presented in Fig. 3 because the money measurei.e., dollars) has different scale properties of mean and variancehan the time measure (i.e., minutes). Therefore, consistent with

ig. 3. Willingness to search (for camera upgrade) as a function of searchurrency and search payoff.

2 of Retailing 85 (3, 2009) 245–257

ttV

eowapegewmi

ntswh(te(Ti

R

rtIwwncpttt

7sfis

wtn%sspwtF

Fu

R

tasuwttiiicstesan

(iacwscdisr

D

it

54 A. Monga, R. Saini / Journal

ransformed separately into z-scores. This process yields datahat does not violate any of the assumptions of the Analysis ofariance.

Analysis using the z-scores revealed that there was no mainffect of either magnitude of search payoff (F(1, 80) = .50, p > .4)r of the currency of search (F(1, 80) = .001, p > .90). Whatas most pertinent to our prediction was the second-order inter-

ction. The magnitude × currency interaction (F(1, 80) = 4.43,< .05) confirmed that the willingness-to-pay (z-score) differ-nce between the low and the high search payoff conditions wasreater for money (M = .61) than for time (M = −.30). That is,ven though the low (2.4 MP) and high (5.1 MP) search payoffsere exactly the same for both currencies, participants in theoney condition were more sensitive to this increase than those

n the time condition.Planned contrasts revealed that willingness to search sig-

ificantly increased with an increase in search payoffs whenhe currency of search was money, but was unaffected byearch payoffs when the currency was time. Specifically, theillingness to search (z-score) in the money condition wasigher (F(1, 80) = 3.95, p < .05) when search payoff was highM = .32) rather than low (M = −.29), but the willingnesso search in the time condition was not statistically differ-nt (F(1, 80) = .98, p > .32) between the high search payoffM = -.14) and the low search payoff (M = .16) conditions.hese results are consistent with the prediction that we made

n H3.

esults for H3 (within-subjects)We then proceeded to examine how participants changed their

esponses when they were asked to respond to one payoff condi-ion after having already responded to another payoff condition.n other words, we wanted to examine if our results also holdithin subjects. For this analysis, a percentage-increase measureas created for each participant by first subtracting the willing-ess to pay of the low payoff condition from that of high payoffondition, then dividing by the willingness to pay of the lowayoff condition, and finally multiplying by 100. This measureherefore indicated the degree to which a respondent was willingo search more for an upgrade to the 5.1 MP camera rather thanhe 2.4 MP camera.

The total sample size for the within-subjects analysis was9 rather than 84 that we used for the between-subjects analy-is because the percentage change could not be calculated forve participants (e.g., because they had failed to respond to theecond willingness-to-pay measure).

Given our prediction, we expected the percentage increase inillingness to search to be higher in the money condition than

he time condition. In line with this, the change in the willing-ess to search was higher for the money condition (M = 125.9) than the time condition (M = 50.4 %). This difference was

ignificant (F(1, 75) = 7.69, p < .01). Additionally, there was noignificant interaction because of order effects (F(1, 75) = .16,

> .6). That is, the percentage change was not influenced byhether the first question was about the low payoff condi-

ion or the high payoff condition. The results are depicted inig. 4.

oram

ig. 4. Percentage increase in willingness to search (for higher vs. lower camerapgrade) as a function of search currency.

esults for H4As in Experiment 2, to gain some additional insights into

he process leading to the observed time–money differences, wenalyzed the cognitive responses. Our prediction about lowerearch-payoff sensitivity in time (vs. money) is based on thenderlying process that people disregard relevant informationhen the currency of search is time rather than money. If this is

rue, respondents’ comments should reveal a greater tendencyo ignore information about search payoffs when they are mak-ng decisions related to spending time rather than money. Twondependent judges coded some participants’ responses as “0”,ndicating that the participant had not taken search payoffs intoonsideration (e.g., “I am a busy person,” or “I’m not sure, it justeemed like a good amount to pay.”), and some as “1,” indicatinghat the participant had taken search payoffs into consideration,ven if only to a limited degree (e.g., “The 5.1 MP camera doeseem like a much better deal, so I’d be willing to wait a fairmount of time for it,” or “5.1 MP and 1.3 MP are very differentumbers.”).

Out of the 84 responses, the two judges agreed on 81r = .92, Cohen’s κ = .92). They then discussed the three dissim-lar responses and arrived at a consensus. We used the mutuallygreed list of responses to analyze 42 participants in the timeondition and 42 participants in the money condition. Consistentith our theorizing, the proportion of participants who had con-

idered search payoffs was only 14.3 % (6 out of 42) in the timeondition but 42.9 % (18 out of 42) in the money condition. Thisifference was statistically significant (z = 2.90, p < .01) indicat-ng that participants were indeed more prone to disregardingearch-payoff information when search involved spending timeather than money.

iscussion

These results showing insensitivity to one kind of searchncentive (increase in search payoffs) complement those ofhe first two studies that showed insensitivity to another kind

f search incentive (decrease in search costs), when the cur-ency is time rather than money. Using both a between-subjectsnd a within-subjects design, we found that currency of searchoderates the effect of search payoffs on search decisions. Fur-

of Ret

torb

S

vsimtudctct2eiplcdcmafiip

osCepar(esd

ttcwtsiimT

bepset

I

ttobrobicrt

bfcistia3mi

Rtrirft

I

tmbtstio

A. Monga, R. Saini / Journal

hermore, in the currency of time (vs. money), considerationf search payoffs was found to be lower in participants’ verbalesponses, just as the consideration of search costs was found toe lower in Experiment 2.

General discussion

ummary

In this research, we commence our inquiry with the obser-ation that the effect of search costs on search behavior is wellupported by prior experimental research but not by field stud-es conducted in retail settings. We argue that this discrepancy

ight be driven by the reliance on money in the former case andhe reliance on time in the latter case and, consequently, try tonderstand the moderating effect of currency. We experimentallyemonstrate that in the currency of money, a decrease in searchosts has a consistent and significant effect on the willingnesso search but that, in the currency of time, a decrease in searchosts has a significantly weaker effect in the context of sequen-ial (Experiment 1) as well as non-sequential search (Experiment). Furthermore, we find that this relative insensitivity in timemerges for not only search costs, but also search payoffs (Exper-ment 3). Participants’ verbal responses in Experiments 2 and 3rovide direct evidence for why this happens: People are moreikely to ignore information about costs and payoffs when theurrency is time rather than money. Finally, the follow-up studyetailed after Experiment 2 revealed that even time participantsan be made to behave like money participants, if they are firstade aware of the value of their time. Overall, we show why

nd how search occurs differently in time than in money. Wend that the willingness to search is less influenced by search

ncentives (lower search costs or higher search payoffs) wheneople search by spending time rather than money.

One limitation of the current research is that it is focusedn situations in which participants make decisions based on theearch costs they expect to incur, rather than actually experience.ould the neglect in the currency of time disappear if costs arexperienced, the way they are in real life? Although this is aossibility, it seems unlikely given that related phenomena suchs duration neglect seem to occur irrespective of whether expe-iences are hypothetical (Varey and Kahneman 1992) or realFredrickson and Kahneman 1993). Moreover, even if actualxperience increases sensitivity to search costs, it is likely to doo for both time and money. If that happens, the time–moneyifferences that we demonstrate will persist.

A limitation related to the above is that we examine costshat people expect to incur sometime in the future rather thanhe present (especially in studies 2 and 3). Could our resultshange if people incur costs in the present? Extending priorork by Soman (1998), Zauberman and Lynch (2005) show

hat time costs are discounted more than money costs becauselack (i.e., perceived surplus) for time is perceived to be higher

n the future than the present. This could suggest that the time-nsensitivity that we demonstrate might reduce when people are

aking decisions about spending time and money in the present.his possibility is worth exploring further. However, it needs to

eI(

ailing 85 (3, 2009) 245–257 255

e noted that we observe differences between time and moneyven when we kept their values equivalent. We find that partici-ants are not simply more liberal in spending time; they are lessensitive to changes in temporal costs than to changes in mon-tary costs. These results afford interesting implications for theheory and practice of retailing.

mplications for theories of retailing

A fundamental premise of search theories (Stigler 1961) ishat a decrease in search costs increases consumers’ willingnesso search. We offer a refinement of this premise; it holds wellnly for money, not for time. This finding provides a plausi-le explanation for prior inconsistencies between experimentalesults and field results from retail settings. Given the numer-us differences between the two styles of research, it cannote argued that the currency of search is the sole reason for thenconsistency in results. However, our results do suggest thaturrency might be one of the culprits. Therefore, search theo-ists in retailing could consider including the currency of searcho their models.

We also add to prior research that has examined searchehavior in the context of savings that occur if one goesrom one store to another. In the classic “jacket and cal-ulator” study (Kahneman and Tversky 1984), people arenfluenced by relative savings—they are more willing topend time when the saving is on a low-priced producthan when it is on a relatively high-priced product. Thats, consumers use a psychophysics-of-price heuristic (Grewalnd Marmorstein 1994). Given the results from Experiment, we suggest that, when people spend time rather thanoney to search, they are less sensitive to absolute differences

n potential benefits, such as the payoffs in a search process.Implications also arise for search-related emotions.

eynolds, Folse, and Jones (2006) argue that retailers oughto reduce search regret, which is a post-search dissonance thatesults from an unsuccessful prepurchase search. If people arensensitive to the amount of time they spend on search, as ouresults suggest, their regret in retail settings might only be aunction of success or failure in search; it might depend less onhe amount of time invested in the process.

mplications for practice of retailing

Because increased search by consumers heightens competi-ion and forces marketers to lower prices, Kuksov (2004) urges

arketers to defend themselves from a decline in search costsy differentiating their products. Given our results, this prescrip-ion is valid if consumers search by spending money but, if theyearch by spending their time (as they do in most retail situa-ions), retailers need not fear that a decline in search costs willncrease search behavior; they need not invest precious resourcesn product differentiation.

The lowering of search costs is also a matter of concern for-retailers. They worry that because it is easier to search on thenternet, competition will intensify and margins will be lowerLynch and Ariely 2000). The underlying assumption, of course,

2 of Ret

iGpsiibemsolaagnstttd

(dpttattctttc

rtacss

Uv

A

A

B

B

B

B

B

C

C

D

D

E

E

F

G

G

G

K

K

K

L

M

M

M

M

O

P

56 A. Monga, R. Saini / Journal

s that lower search costs always lead to higher search behavior.iven that search across stores and over the Internet encom-asses expenditures of time rather than money, the effect ofearch costs on search behavior is likely to be minimal. Fornstance, in the pre-Internet era, a consumer might have vis-ted two brick-and-mortar bookstores to check prices of a bookefore buying it from the cheaper bookstore. In this Internetra, even though temporal search costs are lower, a consumeright still visit only two websites – amazon.com and barne-

andnoble.com – even though one can potentially visit dozensf other similar websites. Consequently, retailers need to beess fearful of decreasing search costs, and more enthusiasticbout the opportunities offered by online environments (Lynchnd Ariely 2000; Weathers et al.+ 2007). That said, this sug-estion regarding insensitivity to search costs on the Interneteeds to be considered tentative. For example, consumers do notearch endlessly on the Internet, suggesting that search costs ofime do matter to them. So, retailers still need to watch out forhe challenges in the world of Internet commerce. Perhaps fur-her research using field studies can examine this aspect moreirectly.

Finally, implications also arise for store location modelsAchabal et al. 1982). How much of a deterring effect does theistance to a store have on its success? Our results suggest thateople will be less sensitive to changes in the time of travel thano changes in price. For instance, the willingness of consumerso drive to far-flung outlet malls might be due to the fact that theyre much more sensitive to the lower price that it offers, than tohe additional time that it takes to visit the outlet mall rather thanhe neighborhood store. Once again, this suggestion needs to beonsidered tentative. If people were completely insensitive toime, they would go to outlet malls for every shopping trip, buthey do not. So even though people are less sensitive to costs ofime than of money, one should not infer that consumers do notare about temporal costs.

In conclusion, search is an integral part of the retail expe-ience. The current research establishes that one’s willingnesso search is not simply dictated by costs and payoffs, butlso by the currency in which search is conducted. In fact,hanges in costs and payoffs seem to have little influence onearch decisions when people spend time rather than money toearch.

Acknowledgement

The authors would like to thank seminar participants at theniversity of Texas at San Antonio and at George Mason Uni-ersity for valuable feedback on a previous version of this article.

References

chabal, Dale D., Wilpen L. Gorr and Vijay Mahajan (1982), “MULTILOC:A Multiple Store Location Decision Model,” Journal of Retailing, 58 (2),

5–25.

riely, Dan and George Loewenstein (2000), “When Does Duration Matter inJudgment and Decision Making?,” Journal of Experimental Psychology:General, 129 (4), 508–23.

P

ailing 85 (3, 2009) 245–257

akos, Yannis (1997), “Reducing Buyer Search Costs: Implications for Elec-tronic Marketplaces,” Management Science, 43 (12), 1676–92.

ecker, Gary S. (1965), “A Theory of the Allocation of Time,” Economic Jour-nal, 75 (299), 493–517.

lock, Richard A. and Dan Zakay (1997), “Prospective and Retrospective Dura-tion Judgments: A Meta-Analytic Review,” Psychonomic Bulletin Review, 4(2), 184–97.

rynjolfsson, Erik and Michael D. Smith (2000), “Frictionless Commerce? AComparison of Internet and Conventional Retailers,” Management Science,46 (4), 563–85.

urdett, Kenneth and Kenneth Judd (1983), “Equilibrium Price Dispersion,”Econometrica, 51 (4), 955–69.

ason, Timothy N. and Daniel Friedman (2003), “Buyer Search and PriceDispersion: A Laboratory Study,” Journal of Economic Theory, 112 (2),232–60.

ox, James C. and Ronald L. Oaxaca (1989), “Laboratory Experiments witha Finite Horizon Job-Search Model,” Journal of Risk Uncertainty, 2 (3),301–29.

avis, Douglas D. and Charles A. Holt (1993), “Experimental Economics,”Princeton, NJ: Princeton University Press.

iehl, Kristin (2005), “When Two Rights Make a Wrong: Searching Too Muchin Ordered Environments,” Journal of Marketing Research, 42 (3), 313–22.

bert, Jane E.J. and Drazen Prelec (2007), “The Fragility of Time: Time-insensitivity and Valuation of the Near and Far Future,” ManagementScience, 53 (9), 1423–38.

lster, Jon and George Loewenstein (1992), “Utility from Memory and Antic-ipation,” in Choice over Time, Loewenstein George and Elster J., eds. NY:Russell Sage.

redrickson, B.L. and Daniel Kahneman (1993), “Duration Neglect in Ret-rospective Evaluations of Affective Episodes,” Journal of Personality andSocial Psychology, 65 (1), 45–5.

raham, Robert J. (1981), “The Role of Perception of Time in ConsumerResearch,” Journal of Consumer Research, 7 (4), 335–42.

rewal, Dhruv and Michael Levy (2007), “Retailing Research: Past, Present,and Future,” Journal of Retailing, 83 (4), 447–64.

rewal, Dhruv and Howard Marmorstein (1994), “Market Price Varia-tion, Perceived Price Variation, and Consumers’ Price Search Deci-sions for Durable Goods,” Journal of Consumer Research, 21 (3),453–60.

ahneman, Daniel and Amos Tversky (1984), ““Choices, Values and Frames,”American Psychologist, 39, 341–50.

ogut, Carl A. (1992), “Recall in Consumer Search,” Journal of EconomicBehavior and Organization, 17 (1), 141–5.

uksov, Dmitri (2004), “Buyer Search Costs and Endogenous Product Design,”Marketing Science, 23 (4), 490–9.

ynch, John G. and Dan Ariely (2000), “Wine Online: Search Costs AffectCompetition on Price, Quality, and Distribution,” Marketing Science, 19 (1),83–103.

armorstein, Howard, Dhruv Grewal and Raymond P.H. Fishe (1992),“The Value of Time Spent in Price-Comparison Shopping: Surveyand Experimental Evidence,” Journal of Consumer Research, 19 (1),52–61.

cDonald, William (1994), “Time Use in Shopping: The Role of PersonalCharacteristics,” Journal of Retailing, 70 (4), 345–6.

iller, Harvey (1993), “Consumer Search and Retail Analysis,” Journal ofRetailing, 69 (2), 160–92.

oorthy, Sridhar, Brian T. Ratchford and Debabrata Talukdar (1997), “Con-sumer Information Search Revisited: Theory and Empirical Analysis,”Journal of Consumer Research, 23 (4), 263–77.

kada, Erica M. and Stephen J. Hoch (2004), “Spending Time versus SpendingMoney,” Journal of Consumer Research, 31 (2), 313–2.

ratt, John W., David A. Wise and Richard Zeckhauser (1979), “Price Differ-ences in Almost Competitive Markets,” Quarterly Journal of Economics, 93

(2), 189–211.

unj, Girish N. and Richard Staelin (1983), “A Model of Consumer InformationSearch Behavior for New Automobiles,” Journal of Consumer Research, 9(4), 366–80.

of Ret

P

R

S

S

S

S

S

S

S

T

V

W

Z

Zsity to Discount Delayed Investments of Time versus Money,” Journal of

A. Monga, R. Saini / Journal

utrevu, Sanjay and Brian T. Ratchford (1997), “A Model of Search Behaviorwith an Application to Grocery Shopping,” Journal of Retailing, 73 (4),463–86.

eynolds, Kristy E., Judith Anne Garretson Folse and Michael A. Jones (2006),“Search Regret: Antecedents and Consequences,” Journal of Retailing, 82(4), 339–48.

aini, Ritesh and Ashwani Monga (2008), “How I Decide Depends on What ISpend: Use of Heuristics is Greater for Time than for Money,” Journal ofConsumer Research, 34 (6), 914–22.

chotter, Andrew and Yale M. Braunstein (1981), “Economic Search: An Exper-imental Study,” Economic Inquiry, 19 (1), 1–25.

mith, Gerald E., Meera P. Venkatraman and Ruby Roy Dholakia (1999), “Diag-nosing the Search Cost Effect: Waiting Time and the Moderating Impactof Prior Category Knowledge,” Journal of Economic Psychology, 20 (3),285–314.

oman, Dilip (1998), “The Illusion of Delayed Incentives: Evaluating FutureEffort-Money Transactions,” Journal of Marketing Research, 35 (4), 427–3.

(2001), “The Mental Accounting of Sunk Time Costs:

Why Time is Not Like Money,” Journal of Behavioral Decision Making, 14(3), 169–85.

rinivasan, Narasimhan and Brian T. Ratchford (1991), “An Empirical Test of aModel of External Search for Automobiles,” Journal of Consumer Research,18 (2), 233–42.

Z

ailing 85 (3, 2009) 245–257 257

rivastava, Joydeep and Nicholas Lurie (2001), “A Consumer Perspective onPrice-Matching Refund Policies: Effect on Price Perceptions and SearchBehavior,” Journal of Consumer Research, 28 (2), 296–307.

tigler, George J. (1961), “The Economics of Information,” Journal of PoliticalEconomy, 69 (3), 213–25.

haler, Richard (1980), “Toward a Positive Theory of Consumer Choice,” Jour-nal of Economic Behavior and Organization, 1 (1), 39–60.

arey, Carol and Daniel Kahneman (1992), “Experiences Extended AcrossTime: Evaluations of Moments and Episodes,” Journal of Behavioral Deci-sion Making, 5, 169–85.

eathers, Danny, Subhash Sharma and Stacy L. Wood (2007), “Effects ofOnline Communication Practices on Consumer Perceptions of PerformanceUncertainty for Search and Experience Goods,” Journal of Retailing, 83 (4),393–401.

akay, Dan (1998), “Attention Allocation Policy Influences Prospective Tim-ing,” Psychonomic Bulletin Review, 5 (1), 114–8.

auberman, Gal and John G. Lynch Jr. (2005), “Resource Slack and Propen-

Experimental Psychology: General, 134 (1), 23–37.wick, Rami, Amnon Rapoport, Alison King Chung Lo and A.V. Muthukr-

ishnan (2003), “Consumer Sequential Search: Not Enough or Too Much,”Marketing Science, 22 (4), 503–19.