Skeptical Shoppers

8/12/2019 Skeptical Shoppers

http://slidepdf.com/reader/full/skeptical-shoppers 1/53

1

The Skeptical Shopper:

How Default Options Affect Choice By What They Tell Consumers

CHRISTINA L. BROWN

ARADHNA KRISHNA*



2

*Christina L. Brown is Assistant Professor of Marketing and Aradhna Krishna isProfessor of Marketing at the University of Michigan. The authors would like to thankFred Feinberg, Carl Mela, Joe Urbany, Carolyn Yoon; members of the marketingdepartment at the University of Michigan and the Fuqua School of Business; and

members of the University of Michigan Decision Consortium for comments on thismanuscript. We would also like to thank Mary Wagner for her help in coding Study 2.Correspondence should be directed to the first author ([email protected]).



3

Abstract

A “default” option is the choice alternative a consumer receives if he/she does notexplicitly specify otherwise. In this article we argue that consumers may treat default

designations as though they contain relevant information about the value of the product;i.e., when our preferences are uncertain, we look to defaults to tell us something aboutthem. More specifically, defaults can invoke a consumer’s “marketplace metacognition”(Wright 2002), his/her social intelligence about marketplace behavior (for example whyand how marketers attempt to manipulate the behavior of consumers). We demonstratethat how the consumer responds to the default depends on whether this social intelligenceis invoked and how it changes the meaning of the default . This “metacognitive” accountof defaults leads to different predictions than accounts based on cognitive limitations orendowment: in particular, it predicts that on some occasions low (less expensive) defaultswill have a greater positive impact on choice than high (more expensive) defaults.Moreover, it predicts the possibility of negative or “backfire” default effects. We report

the results of three experiments that show that how the consumer responds to the defaultwill depend on 1) the ordinal position of the default, 2) whether or not a consumer’smarketplace metacognition is readily accessible, and 3) whether or not a consumer ismotivated and able to use marketplace metacognition to interpret the meaning of thedefault.



4

You are purchasing a computer system on-line. There are several options

for monitors: 15”, 17”, or 19”. The standard system comes with 15” (theleast expensive option), but if you want, you can switch to the 17” or 19”

monitor (for more money).

You are purchasing a computer system on-line. There are several options

for monitors: 15”, 17”, or 19”. The standard system comes with 19” (themost expensive option), but if you want, you can save money by switching

to the 17” or 15” monitor.

What is the marketer up to? Why do you think the marketer did what he

did? Does your answer affect which computer you buy?

Almost every consumer has encountered “defaults” when choosing among

alternatives: the “default” option is the one the consumer will automatically receive if

he/she does not explicitly specify otherwise. Although default options are not restricted to

the Internet, the operational advantages of the Internet have made them more pervasive

over the past few years. For example, on the Dell Computer website, computer systems

come in a pre-packaged combination of RAM memory, hard drive size, disk-drive

options, monitor, software, etc. However, if the consumer prefers a bigger hard drive, or a

small monitor, she may adapt the package to her preferences by selecting her preferred

attribute option from a menu. More notoriously, almost every reader of this article will

have encountered the “opt-out” e-mail list, in which the consumer condemns herself to a

lifetime of promotional emails unless she takes the time to “deselect” the “send me

emails” box.

Most existing research on the effects of default options concludes that default

options affect choice by taking advantage of consumers’ cognitive and processing

limitations. For example, Johnson, Lohse, and Bellman (2002) have shown that

consumers who are obliged to “opt out” of an e-mail list are as much as twice as likely to



5

participate in the list than consumers who can explicitly choose between receiving and

not receiving the emails, as though they had not noticed the difference. Similarly, Park,

Jun, and MacInnis (2000) find that consumers presented with a “fully-loaded” car and

given the opportunity to remove optional features for a cost savings end up with a more

expensive set of features than those presented with a basic model and given the

opportunity to add features for more money. Park et al.’s explanation is related to

endowment effects or loss aversion (Kahneman, Knetsch and Thaler 1991): they suggest

that people presented with a default alternative shift their reference point to include the

features of the default product, framing the forgoing of these features are a loss. In both

cases, the implication is that defaults affect choice because consumers are unintentionally

manipulated by the marketer’s choice of default. In other words, defaults cause consumer

choices to deviate from their true preferences.

In this article we argue that consumers may also treat default designations as

though they contain relevant information about the value of the product (Prelec,

Wernerfelt, and Zettelmeyer 1997; Wernerfelt 1995); i.e., when our preferences are

uncertain, we look to defaults to tell us something about them. In particular, in some

important circumstances a consumer might conclude that the default option is the one the

marketer prefers to sell. In other words, the default is a marketing tactic that is

interpreted by a consumer’s “marketplace metacognition” (Wright 2002), his/her social

intelligence about marketplace behavior (for example why and how marketers attempt to

manipulate the behavior of consumers). We demonstrate that how the consumer responds

to the default depends on whether this social intelligence is invoked and what it means to

the consumer once invoked. Specifically, we show that how the consumer responds to the



6

default will depend on 1) the ordinal position of the default (i.e., whether or not a less-

expensive or more-expensive alternative is the designated default), 2) whether or not a

consumer’s marketplace metacognition is readily accessible, and 3) whether or not a

consumer is motivated and able to use marketplace metacognition to interpret the

meaning of the default.

Importantly, our view of “defaults as a source of marketplace information,”

filtered through a consumer’s prior knowledge of the marketplace, suggests different

empirical results than would default effects generated by inattentiveness on the

consumer’s part (Johnson et al. 2002) or by processing biases such as endowment,

anchoring-and-adjustment (Park et al. 2000), and focus-of-comparison effects (Dhar and

Simonson 1992). A story based on inattentiveness, anchoring, or focus-of-comparison

should apply equally to low (less expensive) and high (more expensive) defaults. A story

based on endowment suggests that a high default will have a greater effect on choice than

a low default.

Previous researchers are undoubtedly correct in their assessment that incomplete

or biased processing helps make consumers more likely to choose default items in many

circumstances. However, when defaults are information-based, we will show that low

(i.e., less expensive) defaults can sometimes have more positive effects than high (more

expensive) defaults, a result that cannot be obtained from existing theories . Even more

remarkably, information-based defaults may create negative or “backfire” effects: if a

consumer believes a marketer is suggesting a default that appears to be in the firm’s best

interests but not the consumer’s, consumers may be less likely to choose an alternative

when it is the default than when no default is indicated.



7

In our story about default effects, then, the consumer is not inattentive but

skeptical and alert, reflecting that defaults can invoke the marketplace metacognition of

the consumer to help him/her better predict the value of each alternative (Friestad and

Wright 1994, Wright 2002). We present three studies to support our information-based

explanation for default effects. In Study 1, we confirm the generally positive effect of

default designations on choice among alternatives, and suggest circumstances in which

low (less expensive) defaults may have greater positive effects on choice than high (more

expensive) defaults. In Study 2, we show that the strength and direction of the default

effect depends on whether or not “marketplace metacognition” is clearly invoked, and

whether or not subjects see the marketer as acting for or against the self-interest of the

consumer. In Study 3, we show that information-based default effects depend on whether

the consumer is motivated and able to use his/her marketplace metacognition to interpret

that information.

THEORETICAL FRAMEWORK

What Is a Default and What Does It Do

We define a default as the alternative the consumer receives if s/he does not

explicitly request otherwise. Defaults are most prominent in cases where a product

consists of a standard configuration of features, which a consumer may add to, eliminate,

or change (as in our opening Internet example). However, defaults are not restricted to

the Internet; all consumer decisions—even innocuous ones or decision not to act—can be

said to involve some form of default: at McDonald’s they will put ketchup on your burger

unless you tell them not to; the personnel department will assume you do not want to



8

make a contribution to the company pension fund unless you tell them otherwise

(Schweitzer 1994), etc.

We are interested in “default effects” – the change in likelihood that a particular

alternative is chosen when designated as the default versus a control condition when no

default is designated. It is possible, after all, that defaults exist but do not change the

likelihood of a particular alternative being selected. For example, a computer marketer

might identify one system configuration as the default because it is the one they expect to

be most popular—this default would make the choice easier, but without changing which

computer the consumer ends up with.

However, there is substantial evidence that defaults can change the likelihood that

a particular item is chosen (Johnson, Bellman, and Lohse 2002; Park, Jun, and MacInnis

2000). The supposed positive effect of defaults on choice has been of great public-policy

interest since it suggests that consumers can be made to unintentionally deviate from their

“true” preferences by the identification of a particular option as the default. The

explanations offered by existing research by and large support the characterization of

consumers as accommodating to the perceptual biases imposed by the default frame.

These existing explanations fall into two overall categories: attention-based default

effects and defaults due to biased processing. We will briefly discuss these to show how

they contrast to our own framework.

Attention-Based Default Effects. Consumers may use defaults heuristically, to

reduce the amount of cognitive effort required to reach a decision (Johnson et al. 2002).

In extreme cases the consumer receives the default option because s/he simply has not

noticed that a choice is being asked of him(her). Less extremely, the default diverts



9

attention toward the designated alternative and away from thorough consideration of the

undesignated alternatives. This explanation is supported by Johnson et al’s finding

(Johnson et al. 2002, Study 1) that consumers are far more likely to choose the default

when faced with a checkbox that already contains a checkmark (that may be then

removed) than when faced with an empty checkbox (that may then be checked). We will

refer to default effects created by lack of thorough cognitive effort or inattentiveness on

the part of the consumer as “attention-based defaults.” Note that if defaults “work”

merely by drawing attention to the default option at the expense of careful consideration

of its alternatives, this would affect choice regardless of whether the default option is

comparatively inexpensive or expensive. Attention-based explanations for default,

therefore, do not provide a clear explanation for why the strength of the default might

vary according to its rank.

Default Effects Due to Biased Processing . Park et al. (2000) suggest that

consumers anchor on the default option and simply fail to adjust sufficiently away from

the anchor and toward their stable underlying preference. As evidence for this, they show

that subjects presented with a fully loaded car and allowed to eliminate unwanted options

ended up with a significantly more expensive car than subjects presented with a stripped-

down model and allowed to add options. The difference in price paid over the two

conditions is attributed to an anchoring-and-adjustment process (Chapman and Johnson

1999). A related mechanism for such anchoring might be a “focus-of-comparison” effect

(Dhar and Simonson 1992), in which a consumer processes more information and more

favorable information about the focal item than about the alternatives it is being



10

compared to. In other words, identification of an alternative as the default encourages

consumers to treat it as the focus of comparison.

As Park et. al. (2000) suggest, this anchoring on defaults might also be attributed

to endowment or loss aversion (Johnson et al. 2002; Park et al. 2000; Schweitzer 1994;

Sen and Johnson 1997). In an endowment-based account, consumers imagine possessing

the default option, thereby reframing all lesser alternatives as losses (Kahneman,

Knetsch, and Thaler 1991; Sen and Johnson 1997). A key feature of the endowment

mechanism is that subjects are less willing to give up the features of their reference

option than they are willing to pay to add features to a lesser option (“willingness to

accept>willingness to pay”). Thus, endowment clearly predicts that high-end defaults

will be stronger positive effects on choice than low-end defaults (Park et al. 2000).

Information-Based Defaults

Our premise is that attention-based and processing-bias explanations omit an

important characteristic of defaults in the real-world marketplace: defaults may contain

information about the value of the choice alternative, which is interpreted in light of a

consumer’s market knowledge.

Wernerfelt and colleagues (Prelec, Wernerfelt, and Zettelmeyer 1997; Wernerfelt

1995) have shown, for example, that consumers can extract information about the

preferences of other consumers from the composition of the choice set. If a consumer is

uncertain about his decision, but knows something about his computer preferences

relative to the preferences of other consumers, this can affect his choice. Prelec et al.

(1997) give the example of museum-goers choosing from among rain ponchos of three

ordered sizes. Subjects choose either from a 32” 34” and 36” set, a 34-36-38” set, a 36-



11

38-40” set, or a 38-40-42” set. Even if a subject had no idea how long a poncho typically

is, each presumably knew whether he/she was short, medium, or tall relative to the

general population. Thus, the rank or ordinal position of the choice alternative was more

predictive of choice than the absolute length of the poncho. The authors refer to this

phenomenon as “rank-order inference.”

In the real-world marketplace, the signal offered by the default designation may

represent not just the will of other consumers but also the will of the marketer. If the

consumer comes to believe that the default is what the marketer prefers to sell , it may

activate his/her social knowledge about marketplace tactics and how to respond to them

(his/her “marketplace metacognition,” Wright 2002).1 When marketplace metacognition

is invoked, the default means something new: it is an attempt to persuade that must be

considered more skeptically. Thus, the perceived cause of the default will have a

powerful impact on the value a consumer imputes to the designated alternative.

We propose that whether or not this marketplace metacognition (MM) is invoked

in a default context may depend on which item is identified as the default. Specifically,

the ordinal position of the default (its expense relative to its alternatives) may have an

impact: items that are the most or least expensive in the set of alternatives are more likely

to invoke MM and thus generate information-based effects. Consider our opening

example in which a consumer may choose among three different monitors

1 Borrowing from Wright (2002), we make a distinction here between “marketplace metacognition” and“persuasion knowledge.” Persuasion knowledge is defined as socially focused market intelligence such asnaïve theories and beliefs about marketers’ motives, strategies, and tactics (see also Moreau, Krishna andHarlam 2001); and the folk psychology of persuasion. “Marketplace metacognition” includes not only suchsocially focused metacognition but also self-focused market-related metacognition such as self-knowledgeand self-control, i.e., “knowledge about one’s own knowledge” (Alba and Hutchinson 2000). Our claim isthat both self- and socially-focused market intelligence will affect how one responds to market events suchas defaults.



12

(small/inexpensive to large/expensive) in making a computer purchase on-line. The

marketer has three choices when it comes to identifying a default: the least expensive, the

middle option, or the most expensive. Previous research (Wegener and Petty 1995, 2001)

has suggested that extremeness affects evaluative inference (Stapel and Koomen 2000),

and that people are more likely to try to correct a bias if it is blatant rather than subtle

(Stapel, Martin, and Schwartz 1988). Thus we argue that if a marketer persistently

chooses one of the extreme options, this extreme behavior may be more likely to be

noticed, identified as a bias, and require an explanation. If the most expensive alternative

is the default, the consumer is likely to become skeptical of the marketer’s intentions. In

contrast, if the least expensive item is the default, the marketer may appear to be acting

counter to expectations and in the consumer’s self-interest rather than her own. She is

more likely to accept the low default but discount the information offered by the high

default. Thus, in cases with many options, low defaults will be more effective at changing

choice than high defaults:

Study 1 thus seeks a simple empirical confirmation of the impact of default

designations on choice. We expect to re-affirm the positive effect of default designation

on choice among multiple alternatives relative to these same alternatives without a

default designation. We also explore conditions under which “low” defaults might be

more positive drivers of choice than “high" defaults, to support our contention that

consumers treat defaults as though they contain information about value. We expect the

strength or power of the default to affect choice may depend on which alternative is

identified as the default , because identifying lowest or highest-priced options as the

default sends different information to the consumer.



13

H1: Default designations will have a positive impact on the choice

of the designated alternative (relative to the same alternative

without such a designation).

H2: A default designation for the lowest-priced alternative will

have a more positive impact on its choice than a default on thehighest-priced alternative (relative to the same items without

such a designation).

STUDY 1

Design and Procedures

One hundred and seventy-eight undergraduates participated in study 1 for course

credit. Subjects completed a questionnaire described as a study of mass customization on

the Internet, involving the specification of six attributes from the websites of (fictional)

manufacturers in three product categories— the number of musical voices and styles for

music keyboards, the monitor size and hard drive size for computers, and meal options

and transportation options for vacation packages. (See Table 1 for a complete description

of the attribute levels.) These product attributes were chosen because they are typical of

those customizable via the Internet (see www.dell.com, and www.libertytravel.com for

examples), and because they were expected to be somewhat familiar to our subject base.

The specific attribute levels were chosen from a web search of these product categories.

Order of presentation of the categories was counterbalanced. Each product attribute

offered three different ordered levels (low, middle, high). For each of the six attributes,

subjects chose one of the three options – low, middle or high attribute level.

---------------------------------------------

Table 1 about here.

---------------------------------------------

http://www.dell.com/

http://www.libertytravel.com/

http://www.libertytravel.com/

http://www.dell.com/



14

Subjects were randomly assigned to one of four between-subjects experimental

conditions—no default, low option indicated as default, middle option as default, or high

option as default. Thus, the design of the study was mixed, with default type (none, low,

medium, high) as a between-subjects factor and the six product attributes a within-subject

factor. Subjects read a brief description of each of the product attributes, then were asked

to indicate their choice among the three alternatives for each of the six attributes.

Subjects in the control “no-default” condition were asked to indicate their preference by

circling the option of their choice.2 “Default” subjects saw either the low, middle or high

option indicated as the default for each of the six attributes. Their instructions reported

that “the manufacturer has indicated a “default” attribute—the value you will get unless

you indicate otherwise by circling a different option.”—but no other reason for this

default attribute was given. Relative prices were expressed as the difference in dollars

between each pair of options (i.e., a non-default option might be described as “30$ more”

or “30$ less” than the default option; see figure 1).

---------------------------------------------

Figure 1 about here.

---------------------------------------------

2 This control condition allows us to eliminate many competing explanations for the effects of default onchoice, by allowing us to compare choice of an item designated as the default to its exact peer without sucha default designation– the same product, with the same competitors, at the same difference in price, presented in the same order. Thus, our results cannot be explained by differences in basic preferences fortrading off increased quality and price, by effects of order of presentation, or by the core tendency ofsubjects to choose a compromise option.



15

Results

We first performed a simple test of difference in proportions to determine whether

default had a significant effect on the choice probabilities of the three alternatives.3

Overall choice probabilities are given in table 2. A chi-square test confirms that default

affected choice probabilities (χ2(6) = 30.20, p ≤ .0001); in each case, choice likelihood

increased versus the control condition when the item in question was the default: 39.77%

versus 30.43% for the low option; 58.71% versus 49.28% for the middle option, and

25.76% versus 20.29% for the high option. Figure 2 illustrates these effects graphically.

---------------------------------------------

Table 2 and Figure 2 about here.

---------------------------------------------

We then performed a fixed-effects conditional logit analysis (McFadden 1974) to

determine whether this overall effect of default on choice was significant, and whether it

occurred regardless of whether the designated default is the low, middle, or high option in

an ordered choice set. (Results are reported in table 3.) The conditional logit model treats

each available alternative as a separate observation, thus including three observations for

each subject for each choice problem. The dependent variable of the model was whether

or not the particular alternative was selected by the subject.

The independent variables included two indicator variables representing the

ordinal position of the option (middle=0,1, high=1,0). These position variables indicate

3 For Studies 1, 2, and 3, we ran each of these models including interactions between the attributes and thecondition variables; none of these interactions were significant, suggesting that the manipulations hadessentially the same effect for the six attributes. We therefore present the results pooled over the sixattributes for each of the three studies.



16

whether ordinal position affected choice probability. If, for example, the middle (high)

indicator is positive and significant it implies that people are more likely to choose the

middle (high) option compared to the low option (regardless of default condition). A

second set of indicator variables allowed us to test for default effects while controlling for

differences in underlying preferences for each of the three options (by comparing choice

of a particular alternative when designated as the default to the same alternative without

such a designation). In model 1, this ‘set” consisted of a single variable indicating

whether or not the alternative was designated as the default. In model 2, we include three

separate indicators for the three different default types (low, medium and high) to

determine if there were any significant differences in default effects based on the ordinal

position of the default option.

---------------------------------------------

Table 3 about here.

---------------------------------------------

Hypothesis 1: Overall Effects of Default (Table 3, Model 1). We first look at the

likelihood of choosing the low, middle and high options when none is specified as the

default . The indicator variable for the middle alternative is positive and significant ( z =

7.769, p ≤ .001; see table 3), whereas the indicator variable for the high alternative is

negative and significant ( z = -3.598, p ≤ .059), showing that subjects tend to prefer the

middle option over the low or high options (when no option is specified as the default).

This is consistent with a well-documented tendency for people with uncertain preferences

to favor compromise options (Simonson 1989). The effect of the default variable is



17

significant and positive, confirming hypothesis 1: a default designation increases the

likelihood an alternative is chosen ( z = 4.968, p ≤ .000).

Hypothesis 2: Low versus High Default (Table 3, Model 2). In a second analysis

we replaced the single default variable with three default indicator variables (i.e., low,

medium, high). The base case is therefore the control (no default) condition. The low

default indicator, for example, shows the effect of the low default on choosing the low

option (vs. the control condition). There are significant positive default effects in the case

where the low item is the designated default ( z = 3.731, p ≤ .000; see figure 2 and table

3), and where the middle item is the default ( z = 2.083, p ≤ .037), but the default effect

for the high option does not reach significance ( z = 1.243, p ≤ .214). Thus, default effects

appeared to be slightly weaker for higher-cost alternatives. However, a test of a model

allowing low and high defaults to differ did not quite outperform a model restricting the

coefficients for low and high defaults (χ2(1) = 2.40, p ≤ .122). Thus, hypothesis 2

received only directional support.

DISCUSSION

Study 1 confirms a significant positive effect of a default designation on choice of

a particular alternative over and above the same alternative without such a designation.

Furthermore, study 1 suggests conditions exist under which a low (less expensive) default

might have a greater positive effect on choice than a high (more expensive) default. This

result could not be derived from an attention-based or focus-of-comparison mechanism,



18

and is in conflict with an endowment mechanism.4 Our results are directionally consistent

with our specific claim that a high default may be discounted by the sense that it might be

the result of a marketer’s attempt to persuade. Although the overall positive effect of

default in study 1 was quite strong, study 1 suggests that ordinal position alone may not

invoke MM, or at least may not invoke it for all subjects. Furthermore, a “low>high”

effect would be at odds with previous research, specifically Park et al. 2000, which

showed high defaults that had a greater positive effect on choice than low defaults.

In study 2, we address the question of why high defaults may be quite positive in

some cases (e.g., Park et al. 2000) and weak or even negative in others. We focus on a

critical difference between the paradigm offered in our study 1 and that used by Park et

al. In Park et al. (2000) and related research such as Dhar and Simonson (1992), the

experimental context did not involve an attempt by a marketer to persuade the subject,

which plays a critical role in our explanatory framework. Our “information-based”

explanation for defaults suggests that it is precisely the perception that the marketer is

attempting to persuade the consumer that changes what the default “means” to him/her.

These researchers have shown that outside a persuasive context biased processing will

lead to positive effects of default in general and high default in particular. We argue that

when consumers feel that marketers are trying to bias their (consumers’) choices, they

4 Note that a similar result is given even if consumers are informed only about other consumers (and notabout marketers), if we also assume that individual preferences correlate with product experience. Consider

the heuristic “buy the least expensive item of acceptable or better quality.” Assume that most consumersare not quite certain about what “acceptable quality” might be. Following the reasoning given by Prelec atal. (1997), if a consumer considers his/her needs to be modest relative to other consumers’, s/he will purchase at or below the default. If he considers his needs to be average, he will purchase at the default.High end users, however, are likely to be more experienced and more certain about their preferences(Coupey, Irwin, and Payne 1998), and therefore relatively unaffected by defaults. This explanation impliesthe same aggregate result hypothesized in hypothesis 1: Low defaults will be more powerful than highdefaults, because low-end and medium users will respond to low default, only middle-of-the-road users will





20

his/her own interests) recommends the less expensive item and skepticism, resulting in a

reduced or negative effect, when the marketer recommends the more expensive item.

H3: There will be an interaction between the effect of ordinalposition of the default and accessibility of marketplace

metacognition on consumer choice:

a) When a more expensive alternative is the default and

marketplace metacognition is not accessible, there will be a

significant positive “default effect:” i.e., subjects will be more

likely to choose the expensive item relative to its control.

b) When a more expensive alternative is the default and

marketplace metacognition is accessible, the “default effect”

will be weaker or even negative.

c) Marketplace metacognition will have no effect on the choice

likelihood of a low (less expensive) default.

The Role of Counterarguing in Creating Default Effects

Skepticism about the marketer’s intentions or abilities may not cause one’s

overall attitude toward the market to change, but simply change the reasons that are

brought to bear on the decision. Work by Wegener and Petty (the “Flexible Correction

Model;” 1995, 2001) has suggested that if subjects perceive a bias to be present in an

argument, they will be more likely to counterargue (to resist the bias) than if they do not

perceive such a bias. Thus, when subjects recognize a high default as a biased attempt to

persuade (biased by the marketer’s interest in greater sales), they are more likely to argue

against choosing the high item than when it is not the default. If MM is not accessible the

default may not be recognized as a persuasive attempt, thus counterarguments would be

less likely. In contrast, recommending a low default is less likely to be perceived as an

attempt to bias one’s behavior, so the accessibility of MM should not have an effect.



21

MM-driven skepticism might instead lead consumers to choose the less-expensive

option because it involves less financial risk (i.e., if I must make a purchase from a

marketer whose intentions are questionable, the least expensive will get me into less

trouble). If MM operates in this fashion, there should be a main effect of the accessibility

of MM on choice of the high option. Alternatively, subjects might conclude that the

marketer is not really competent to provide advice; in this case there should be no real

effect of default designation at all.

Thus in study 2, we will also investigate the mechanisms through which MM

affects choice by asking subjects explicitly for the reasoning behind each of their choices.

Subjects who are discounting the information present in the default should demonstrate

this by giving counterarguments, i.e., reasons to reject the default item (Brown and

Carpenter 2000). Limiting the alternatives each subject sees to only two will also make it

easy to determine whether subjects are taking a supporting (arguing for the choice) or

counterarguing approach (arguing against the unchosen item). If MM is driving choice

effects by making subjects more skeptical in the presence of a perceived bias, then

counterarguing should mediate the effects of default on choice.

H4: When a more expensive item is the default, and subject’s

marketplace metacognition is made more accessible, the

subject will be more likely to report counterarguments

(reasons to reject the unchosen item) than when marketplace

metacognition is not explicitly made accessible. However, when

the less expensive item is the default, accessibility will have no

effect on the likelihood of counterarguments.

H5: The likelihood of counterarguments will mediate the effects of

default and accessibility of marketplace metacognition on

choice.



22

STUDY 2

Manipulations

We reasoned that giving subjects more information about the status of the

marketer’s business would make subjects’ marketplace metacognition more accessible.

By reminding them of the marketer as a business with its own goals and motives

(Campbell and Kirmani 2000), we expected to enhance subjects’ motivation to “hold

valid agent attitudes” and concern about manipulation (Friestad and Wright 1994, p. 9).

With this in mind, we created a simple description of “Allbuy.com,” an Internet retailer.

Subjects in the “limited information” condition saw a single paragraph:

Allbuy.com is an Internet-based retailer of a wide variety of products— from electronic goods such as computers and keyboards to consumer

services such as vacation packages. They do not make the products

themselves but offer a wide variety of products that are availableelsewhere. Allbuy allows the customer to choose among several models for

each product by specifying its attributes: for example, you can customizewhat type of monitor and how large a hard drive comes with the computer

you order, or how many meals come with your vacation package.

In the “enhanced information” condition, we added a second paragraph describing

the strengths and weaknesses of Allbuy.com as a business:

Allbuy has become known for its large and well-trained staff dedicated

to selecting a high-quality assortment for presentation on its website. Allbuy has given large discounts in the past but is currently

experiencing great cash-flow difficulties, and there is some chance it

will go out of business in the next few months if customers cannot beconvinced to spend more money very soon.

Note that if the additional information merely serves to lower subjects’ overall

opinions of Allbuy.com or make them seem less trustworthy, the effect of the information

on default would be to reduce the size of the default effects regardless of which item is



23

identified as the default. Thus, the effect of the information would be to reduce the size of

default effects regardless of which item is identified as the default. In other words,

consumers should fully discount a low or a high default because the marketer is not a

credible source of information. Our hypotheses, however, suggest that marketplace

metacognition will reduce effects on choice for high defaults but that positive default

effects will persist for low defaults regardless of MM (hypothesis 3 above).

Ideally, therefore, the additional text in our “enhanced information” condition

should not result should in lowering overall opinions of Allbuy.com. It should include

both positive and negative features, so as to be as close to affectively neutral as possible.

In this way, subjects motivated to think well of the marketer can do so, but subjects

motivated to question the marketer’s intent may do that instead. A pre-test comparing the

limited and enhanced descriptions of Allbuy.com showed no significant differences in

attitude (good/bad; favorable/unfavorable; negative/positive) between the versions.

Design and Procedures

Ninety-six undergraduates participated in study 2 for an eight dollar cash

payment. Subjects in study 2 saw the same six problems as those in study 1 (hard drive

and monitor size for computers; number of musical styles and voices for electronic

keyboards, and meal and transportation options for vacation packages). To simplify

interpretation of results and to reduce the number of experimental cells, each subject saw

only two options per attribute – called “low” and “high.”6 Thus, the design of the study

was mixed, with default type (none, low, or high) and information (limited or enhanced)

as between-subjects factors and the six product attributes as a within-subjects factor.

6 The low and middle attribute levels from Study 1 were retained and the high attribute level was dropped.



24

Subjects read the introductory paragraphs (containing the information manipulation) and

were told that their product choices would help us assess the likely mix of products

Allbuy.com would sell over the next few months. Then, subjects saw all six problems in

counterbalanced order.

To determine whether the likelihood of counterarguments might be affecting our

default results, subjects were asked to write a brief sentence explaining why they made

the choice they did after each of the six problems (each response was coded as primarily

supportive of the choice or counter to the unchosen item). Finally, subjects answered

several seven-point scale questions assessing overall attitude towards Allbuy

(good/favorable/positive), opinions of Allbuy’s product quality, price, stability,

knowledgeability, sincerity, trustworthiness, reliability, and whether or not the firm was

“trying to serve their own needs” rather than “trying to serve my needs”.

Results

Hypothesis 3: Choice Results. Table 4 shows the choice results for each cell. A

chi-square test confirms that the overall likelihood of choosing the high or low option

differed by experimental cell (χ2(5) = 11.72, p ≤ .02). Over the entire dataset, 51.0%

chose the low option and 49.0% chose the high option. Table 4 reports the cell means and

figure 3 illustrate the results graphically.

---------------------------------------------

Figure 3 and Table 4 about here.

---------------------------------------------



25

We used a logit model with robust standard errors to determine the effects of the

manipulation on choice (see table 5). Since there were just two options to choose from,

the dependent variable was “choice of the high option”. Independent variables were

indicators for the low and high defaults, an indicator variable for the information

condition, and interactions of defaults and information. 50.5% of subjects chose the high

option in the limited information condition versus 47.5% in the enhanced information

condition ( z = 1.11, n.s.). Thus it appears that the information manipulation did not

encourage subjects to dislike Allbuy or see it as more risky, thereby inclining them

towards the less-expensive choice. The low default increased choice of the low option

(57.8%, vs 50.0% in the control condition; hypothesis 3c), while the high default

increased choice of the high option (55.0% vs. 50.0% in the control condition).

Consistent with our expectation that only high defaults would be interpreted

differently when MM was accessible, the high default* information interaction was

significant ( z = -2.22, p ≤ .026) but the low default* information interaction was not ( z = -

.12, n.s.). Thus, hypotheses 3a, 3b, and 3c are supported.

In the “enhanced information” condition (see figure 3), the high default had a

slightly negative effect on choice (47.9% vs. 54.4%, n.s.). The pattern was reversed in the

“limited information” condition: the high default increased choice of the high option

(62.1% vs. 45.8%, z = 1.643, p ≤.051). The low default had a slightly positive effect on

choice in the enhanced information condition (59.4% vs. 45.6%, z = 1.353, p ≤ = .09), but

it had little effect on the choice of the low option in the limited information condition

(56.2% vs. 54.2%, n.s.). This confirms hypothesis 3c and previous research (Park et al.

2000). Thus, we find that although expensive defaults may have a positive effect on



26

choice in the limited-information condition, they can have a limited or even negative

effect in the enhanced information case.

Finally, we measured attitude as the mean of a subject’s response to the

good/favorable/positive scale questions. Mean attitude toward Allbuy was positive (2.447

on a 7-point zero-centered scale; significantly different from 0, t(93) = 6.233, p ≤ .0001).

An ANOVA showed no significant effect of either manipulation (information, default

type) or their interactions on good/favorable/positive scale questions; thus, any choice

effects may not be attributed to a generally more negative attitude toward the

manufacturer. Also, neither the default manipulation nor the information manipulation

significantly affected the perceived price, quality, or trustworthiness of the marketer, nor

did the manipulations affect the perception that the marketer was serving his own needs

rather than the customer’s.

---------------------------------------------

Table 5 about here.

---------------------------------------------

Hypothesis 4: Likelihood of Counterarguments. We now turn to the idea that MM

works by making counterarguments more likely in the high default case. Reasons for

choice provided by subjects were coded by two judges. Each judge coded the reasoning

as either a “support argument” or a “counterargument”. A “support argument” was a

positive reasoning strategy in favor of the chosen option (for example, “bigger is better”).

A “counterargument” was a negative reasoning strategy against the unchosen option (for

example, “I don’t really need a big hard drive.”). Inter-rater agreement after initial coding

was 90.81% (kappa = .8128, z = 19.26, p ≤ .001); conflicts were resolved by discussion.



27

Across all conditions, 45.3% of subjects offered counterarguments in explanation

for their choice. A chi-square test confirmed that the percent of counterarguments

differed significantly by experimental cell (χ2(5) = 13.56, p ≤ .02). Subjects in the

enhanced information condition were not more likely to counterargue than subjects in the

limited information condition (45.3% vs. 45.3%), suggesting that the information

condition itself did not make subjects more skeptical. Furthermore, subjects exposed to a

default (either high or low) were slightly less likely than control subjects to counterargue

(39.3% vs. 46.3%, although this difference was not significant), suggesting also that the

mere presence of a default did not make subjects skeptical. However, as expected,

subjects in the high default/enhanced condition were more likely to report

counterarguments (46.8%) than were subjects in the high default/limited condition

(33.0%, z = 1.952, p ≤ .025). Information condition had no effect on counterarguing for

subjects exposed to a low default (51.6% vs. 48.4%, n.s.). This is consistent with

hypothesis 4, which proposed that making marketplace metacognition more accessible

would increase counterarguing against the more expensive option (as subjects attempt to

correct for perceived bias) but that MM would have no effect on counterarguments when

the less expensive option was the default.

Hypothesis 5: Mediating Effects of Counterarguing. Hypothesis 5 proposed the

likelihood of counterarguing would mediate the effects of default and information

conditions on choice. Table 5, model 1 shows the effects of information and default on

the likelihood of choosing the higher option. Table 5, model 2 shows the effects of these

same conditions on the likelihood of counterarguments, confirming that subjects in the

high default/enhanced condition were more likely than subjects in the high default/



28

limited condition to counterargue ( z = 2.74, p ≤ .02). Finally, as shown in table 5 (model

3), when “counterargument” was added as a covariate to our original analysis, it had an

extremely powerful effect on choice ( z = 12.81, p ≤ .001), and the effects of default as

well as the default*information condition become statistically insignificant ( z = -.06,

n.s.). In other words, the change in likelihood of counterarguments fully mediated the

effects of default, information, and their interactions on choice, confirming hypothesis 5.

DISCUSSION

Study 2 confirms that subjects treated the default as though it contained legitimate

information about the marketplace. One way to understand the difference between study

1 and study 2 effects is that the mere presence of a default designation may not by itself

convey information about preferences to a consumer unless it is interpreted by his/her

marketplace metacognition. If one has no theory about why a marketer chose to designate

a particular item as the default, the designation can only affect behavior through

attention-based or biased-processing mechanisms described by previous researchers.

Making marketplace metacognition more accessible by reminding subjects of the

marketer’s business status changed the meaning of the default: instead of being an

unbiased recommendation, it was evaluated in the context of the consumer’s expectations

that businesses will act in their own financial self-interest by attempting to increase

sales.. MM had this effect not by because it made subjects pessimistic about the quality

of the marketer’s advice, or by changing overall attitudes toward the marketer, but

because it gave subjects a motivation to draw a particular conclusion and making

motivation-consistent reasons more accessible. When a firm in trouble offers a high



29

default, it serves to make counterarguments (reasons against the default) more accessible.

In contrast, when a firm in trouble offers a low default, consumers do not view it

skeptically and do not attempt to counterargue its effects.

Explanations based on discounting of the marketer’s abilities or based on

differences in perceived financial risk imply a fairly involved cognitive process in which

subjects review the information provided by the default, draw a conclusion about the risk,

trustworthiness, or competency of the marketer, and correct their initial inclinations

according to this conclusion. Our explanation is much simpler and less cognitive

demanding of the consumer – skepticism in the presence of a high default merely changes

the relative accessibility of reasons for and against choice. It does not require the

consumer to draw a thoughtful or permanent conclusion about the marketer’s abilities and

intentions. Although she might do so, our theory does not require it and study 2 does not

show that she has.

In study 3, we provide additional evidence that marketplace metacognition is the

construct underlying information-based defaults. In particular, we propose that

individuals differ in the degree to which MM is well-formed and thus accessible

(Bearden, Rose, and Hardesty 2001). Secondly, we distinguish MM from category

expertise. Although the two are likely to be correlated, they will generate different default

effects. Consumers extremely high in category-level expertise are likely to have

developed well-defined preferences (Coupey, Irwin, and Payne 1998). Expert consumers

may in fact have plenty of marketplace metacognition but no motivation or need to access

it. They know what they want. In contrast, novice consumers have a motivation seek



30

outside information in defaults, and to rely heavily on external sources of advice or

information.

However, extracting information from defaults requires some degree of naive

theory about the marketer's intentions. Thus, novice consumers who are also low in MM

will not be able to access theories to distinguish between the signal provided by low,

rather than high, default effects. Default effects for these novice/low-MM consumers will

resemble the high>low pattern created by positively biased processing and established by

Park et. al 2000. In contrast, novice consumers who are high in MM will become

skeptical only of high defaults, but not low, creating the low>high effect established in

study 2. In other words, in both studies 2 and 3 the accessibility of MM creates

information-based defaults; in study 2 this is done via experimental manipulation,

whereas in study 3 this is a measured individual-difference variable.

In short, information-based default effects require that consumers be both

motivated and able to access marketplace metacognition to interpret market signals.

Thus, expertise and MM should interact to create the effect. Thus, study 3 will show both

convergent and discriminate validity by showing individual-difference effects that are

inconsistent with expertise alone and require MM to explain them.

H6: Subjects with low levels of expertise will be more subject to default

effects than subjects with high levels of expertise.

H7: Subjects with low levels of expertise and low levels of marketplace

metacognition will be more positively affected by more expensive default

than less expensive defaults, relative to their controls.

H8: Subjects with low levels of expertise and high levels of marketplace

metacognition will be more positively affected by less expensive defaults

than more expensive defaults, relative to their controls.





32

Results

Sixty undergraduate subjects participated in study 3 for a $6 cash payment.

Marketplace metacognition scores were calculated by summing across all 31 self-

confidence measures for each subject. Subjects were then divided into low/high MM

groups via a median split based on their mean response to the scales (with negative items

reverse-scored). Subjects were also divided into novices/experts via a median split on the

expertise scales. A chi-square test confirms that there were significant differences in

choice probabilities based on experimental conditions (χ2(11) = 18.56, p≤ .069). A logit

analysis with robust standard errors used “chose high” as the dependent variable, and

indicator variables for low default, high default, novice status, and low/high MM status,

and their interactions. Table 6 shows the choice probabilities for all subjects broken out

by novice/expert status and by low/high MM. Figure 4 shows the results graphically.

---------------------------------------------

Table 6 and Figure 4 about here.

---------------------------------------------

Hypothesis 6: Novices versus Experts. There was a negative main effect of novice

status on choice of the high option (i.e., experts prefer expensive products; z = -2.15, p ≤

.016): Novices in both the low-default condition (53.5% vs. 49.0%) and the high-default

condition (54.7% vs. 51.0%) showed an increased likelihood of choosing the default

relative to its control. In contrast, experts showed slight decreases in the likelihood of

choosing the default, although these decreases were not by themselves significant (35.6%

vs. 37.6% control for the low default condition, and 60.2% vs. 62.5% control for the high



33

default condition). Taken together, these results show a significantly higher default

effects for novices than for experts ( z = 3.998, p ≤ .001). Thus, hypothesis 6 is supported.

Hypotheses 7 and 8: Low versus High-MM Novices. H7 and H8 together proposed

that low-MM novices would not become skeptical, but that high-MM novices would.

Thus default effects for low-MM novices should be positive, particularly for high

defaults, consistent with the low-information cells for study 2 and with Park et al. 2000

(hypothesis 7); in contrast, default effects for high-MM novices should be positive for

low defaults and weak or negative for high defaults, consistent with the enhanced-

information cells for study 2 (hypothesis 8). The logit analysis showed a significant three-

way interaction between high default, high MM, and novice status consistent with these

hypotheses, such that low-MM novices responded positively to high defaults but high-

MM novices responded negatively ( z = -2.38, p ≤ .01). Similarly, there was a modest

three-way interaction between low default, high MM, and novice status, such that low-

MM novices did not respond to low default but high-MM novices responded positively ( z

= -1.39, p≤ .082).

Table 6 and Figure 4 illustrate these differences clearly. Low-MM novices were

significantly more likely to choose the high default versus its control (55.9% vs. 40.0%).

They were slightly more likely to choose the low default versus its control (61.9% vs.

60.9%), although this difference was not by itself significant. Thus, hypothesis 7 was

supported. In contrast, high-MM novices were somewhat less likely to choose the high

default versus its control (52.8% vs. 62.5%), and somewhat more likely to choose the low

default versus its control (45.5% vs. 37.5%). Thus, hypothesis 8 was confirmed.7

7 Figure 4 also illustrates an interesting if unanticipated effect: High-MM experts (who are confident thatthey want the high option specifically) reacted negatively to low defaults: 31.0% of high-MM experts



34

GENERAL DISCUSSION

We have presented a series of three experiments supporting our explanation that

under some circumstances consumers treat defaults as though they contained information

about the relative value of the alternatives (Prelec et al. 1997). More specifically, they are

consistent with our claim the consumer treats defaults as though they tell her something

about the intentions of marketers themselves. To interpret this information, the consumer

calls on her social intelligence of the marketplace, her “marketplace metacognition”

(Wright 2002). Thus, conditions that make marketplace metacognition more accessible

also serve to moderate default effects.

However, these effects of default are not uniformly positive. Selecting a high

(more expensive) item as the default may even have a negative (backfire) effect. These

results are not explainable by mechanisms proposed in previous research: attention-based

mechanisms make no prediction about the strength of default based on relative expense,

and loss aversion or endowment-based explanations predict the opposite result (that high

defaults will be stronger than low defaults). We do not mean to suggest that these

mechanisms are not operative, but argue that they cannot fully explain our data. Thus, our

results add to the explanations provided for default effects in the literature.

When defaults are information-based, the meaning of the default to the consumer

changes based on what metacognitions are invoked (Loewenstein 2002). In Study 1, we

suggested that manipulating the “extremeness” (ordinal position) of the default may make

choose the low item when it was the default, versus 39.9% when it was not. This suggests perhaps that thecontent of the MM of experts differs from that of novices, for example it may include the notion thatmanufacturers endorsing inexpensive goods are trying to pawn off low-quality items. We will discussdifferences in the content of MM in the concluding discussion of this paper.



35

consumers more skeptical of high defaults and more accepting of low defaults. However,

Study 1 results implied that information-based defaults would not be particularly

powerful unless the potential information contained in the default were interpreted ; i.e.,

unless the consumer brought to bear a theory or theories about why a marketer might

make a particular item the default. In other words, a default is not informative unless the

consumer has an idea what it means.

In study 2, we confirm this idea that the ordinal position of the default and a

consumer’s marketplace metacognition interact to create information-based default

effects. In doing so, we also shed some light on why previous researchers have found a

positive effect of expensive defaults (Park et al. 2000), whereas we have found negative

effects in some cases. Moreover, study 2 provided some support for the cognitive process

underlying MM effects: Rather than directly making consumers more pessimistic about

the marketer’s abilities, the accessibility of MM appears to have changed subjects’

motivation to draw a particular conclusion, making motivation-compatible reasoning

more accessible (Brown and Carpenter 2000; Kunda 1991; Wegener and Petty 1995).

Thus, counterarguments (arguments against accepting the default) were highly accessible

in the high default case (when the marketer presumably was acting in its own self-interest

and against the consumer’s). In contrast, the same accessibility made support arguments

highly accessible in the low default case (when the consumer was more favorably

inclined toward the marketer because it appeared to be acting against its own self-interest

and for the consumer’s). The reasons—specifically the counterarguments--made

available via marketplace metacognition fully mediated the effects of the accessibility



36

manipulation on choice, providing strong evidence for our proposed theoretical

mechanism.

In study 3, we showed additional support for information-based or

“metacognitive” defaults in the form of convergent and discriminant validity. We saw

that the accessibility of marketplace metacognition varies across individuals, resulting in

default effects similar to those generated by direct manipulation in study 2. Furthermore,

we saw that the effects of MM on default effects are empirically distinct from the effects

of category expertise. Novices in study 3 with high levels of marketplace metacognition

were able to discriminate between what a low default and what a high default might mean

to them. Low-MM novices, however lacked the knowledge base to allow them to

discriminate between low and high defaults, while expert subjects lacked the motivation

to determine a marketer’s intent (presumably because these experts already knew what

they want). This complex relationship supports the importance of MM rather than

expertise alone as a key moderator of default effects.

A practical extension of this work would be to other forms by which a default

may be designated: i.e., via drop-down menus, via pre-configured bundles of attributes

that might then be customized, or via order of presentation. Although our research

spanned three diverse consumer categories (computers, electronic keyboards, and

vacation packages), default effects in other categories may of course differ. We would

expect, based on our theory, that product categories regarding which most consumers are

highly confident and familiar to be relatively immune to information-based defaults. Our

product categories involved fairly expensive purchases; product categories with

extremely low levels of involvement might be less likely to generate backfire effects



37

driven by skepticism. Both these boundary conditions suggest that cognitive capacity

and/or cognitive load may have a powerful effect on defaults. Moreover, marketplace

metacognition seems not only content-specific but culturally specific: cultures that have

different knowledge and expectations about the marketplace should respond differently to

defaults. These would be fruitful grounds for future research.

From a managerial perspective, our results suggest that some care should be taken

in the selection of defaults. To the extent that defaults are also driven by attention-based

mechanisms, manufacturers may continue to see defaults have a positive effect on choice

as is typically the case for post-purchase or post-registration opt-out mailing lists

(Johnson et al. 2002). However, firms that consistently propose the most expensive

product as the default may lose their credibility and be subject to a “backfire” effect.

AN OPTIMISTIC VIEW:

FUTURE DIRECTIONS FOR THE STUDY OF METACOGNITION AND PRIOR

KNOWLEDGE IN DECISION MAKING

Defaults are pervasive in modern consumer life; they are not limited to the

Internet but have merely been made more relevant by its arrival. To our minds, however,

the primary significance of default effects is not that they reflect a new technology, but

that they are carriers of meaning – they transmit information about the marketplace to the

consumer. They are one of a broad array of marketer-generated techniques that may

change in value to the consumer based on how he/she interprets them.

Factors Affecting the Use of Marketplace Metacognition. Our studies use a text-

based description of the marketer and the presence or absence of a default designation to

provoke information-based effects. But marketplace metacognition could be primed in



38

many other ways, just as many other features of a shopping environment may invoke

skepticism. Will a high number of alternatives suggest the marketer is indifferent among

them and make consumer neglectful of the marketer’s default recommendations? Does

the presentation of unordered categorical alternatives (rather than the ordinal alternatives

presented in our studies) ever invoke skepticism? These and similar questions would be

worth exploring in future research

The Content of Marketplace Metacognition. Our “low>high” effect rests on one

very simple assumption about the content of marketplace metacognition: a shared notion

that marketers prefer to sell more expensive rather than less expensive products. This

notion does not require complex market knowledge—for example it is not necessary for

consumers to ‘know’ that expensive products are also high-margin products, although

they may in fact believe this. It is only necessary for marketers to appear to act in a self-

interested way (according to whatever naïve theories the consumer might have), thus

generating skepticism and greater access to counterarguments. Consumers with different

theories about where marketer’s self-interests lie may well generate different effects than

we propose here; in fact, study 3 offers a tempting hint of such behavior: High-MM

experts were actually less likely to choose the low default than its control. Although this

effect was unexpected, it is quite consistent with our theoretical explanation if we

concede that experts (in contrast to novices) might believe that marketers like to pass off

low-quality goods to the undiscriminating. The content of marketplace metacognition and

its heterogeneity across different classes of consumers would be fertile area for future

research.



39

The Social Benefit of Information-Based Effects. Information-based effects need

not necessarily be negative. For example, if a consumer has specific beliefs about a

particular category, invoking this prior knowledge could have a positive impact on the

consumer’s reasoning process. If a computer firm is known as an expert systems

integrator, or if an apparel firm is known as an arbiter of taste, firm-specific information

effects are likely to be positive, particularly for expert consumers who are aware of the

risks of mismatched computer hardware or the risks of wearing last year’s styles.

In the face of an overwhelming flood of evidence that consumers’ preferences are

perennially uncertain and must be constructed from the bits and pieces of information

floating around the brain, it seems critically important to appreciate how the consumer

herself or himself construes the situation the marketer has put him/her in. What is Firm X

up to? Why did they make the decision to do it that way? How should I as a consumer

respond to it? The default effects we have documented here are only one result of

consumers’ constant search for informative cues about their own preferences. Our studies

take a “dynamic” perspective on the study of consumer decision-making: one in which

the outcome is mutually determined by the marketer’s design and the consumer’s

interpretation of it. There has been a surge of recent interest in the research community in

such dynamic perspectives (Briley, Morris, and Simonson 2000; Campbell and Kirmani

2000; Friestad and Wright 1994; Iyengar and Lepper 1999, among others). Optimistically

this perspective no longer excludes the consumer’s viewpoint on marketing problems

(Bazerman 2001; Loewenstein 2001) for the eminently practical reason that we cannot

correct predict consumer behavior without it.



40

To our minds, information-based effects of marketplace metacognition are a

positive sign of consumers’ ability to cope with persuasive attempts rather than a sign

that they are subject to normatively inappropriate framing effects. Furthermore, our

results suggest that consumers do eventually become less vulnerable to defaults: high-

MM consumers clearly discriminated between types of defaults even when their expertise

in the category was low. Consumers’ knowledge of the marketplace may serve them well,

after all.



41

REFERENCES

Alba, Joseph W. and J. Wesley Hutchinson (2000), “Knowledge Calibration: WhatConsumer Know and What They Think They Know,” Journal of Consumer

Research, 27 (3 December), 388-396.

Bazerman, Max H. (2001), “Consumer Research for Consumers,” Journal of Consumer

Research, 27 (4 March), 499-504.

Bearden, William O., David M. Hardesty, and Randall L. Rose (2001), “Consumer Self-Confidence: Refinements in Conceptualization and Measurement,” Journal of

Consumer Research, 28 (1 June), 121-134.

Bettman, James R., John W. Payne, and Mary F. Luce (1998), “Constructive ConsumerChoice Processes,” Journal of Consumer Research, 25(3 December), 187-217.

Bettman, James R. and Mita Sujan (1987), "Effects of Framing on Evaluation ofComparable and Noncomparable Alternatives by Expert and Novice Consumers," Journal of Consumer Research, 14 (2 September), 141-154.

Briley, Donel A, Michael W. Morris, and Itamar Simonson (2000), “Reasons as Carriersof Culture: Dynamic versus Dispositional Models of Cultural Influence onDecision Making,” Journal of Consumer Research, 27 (2 September), 157-178.

Brown, Christina L. and Gregory S. Carpenter (2000), “Why Is The Trivial Important: AReasons-Based Account for the Affects of Trivial Attributes in Choice,” Journalof Consumer Research, 27 (2 September), 372-385.

Campbell, Margaret C. and Amna Kirmani (2000), “Consumers’ Use of PersuasionKnowledge: The Effects of Accessibility and Cognitive Capacity on Perceptionsof an Influence Agent,” Journal of Consumer Research, 27 (1 June), 69-83.

Chapman, Gretchen B. and Eric J. Johnson (1999), “Anchoring, Activation, and theConstruction of Values,” Organizational Behavior and Human Decision

Processes, 79 (2), 115-153.

Coupey, Eloise, Julie R. Irwin, and John W. Payne (1998), “Product Category Familiarityand Preference Construction,” Journal of Consumer Research, 24 (4 March), 459-468.

Dhar, Ravi and Itamar Simonson (1992), "The Effects of the Focus Comparison on.Consumer Preferences," Journal of Marketing Research, 29, 430-440.

Friestad, Marian and Peter Wright (1994), “The Persuasion Knowledge Model: HowPeople Cope with Persuasion Attempts,” Journal of Consumer Research, 21 (1June) 1-31.



42

Iyengar, Sheena S., and Mark R. Lepper (1999), “Rethinking the Value of Choice: ACultural Perspective On Intrinsic Motivation,” Journal Of Personality And Social Psychology, 76, 349-366

Johnson, Eric J., Steven Bellman, and Gerald L. Lohse (2002), “Defaults, Framing, and

Privacy: Why Opting In ≠ Opting Out,” Marketing Letters, 13 (1), 5-15.

Kahneman, Daniel, and Amos Tversky (1984), “Choices, Values, and Frames,” American Psychologist , 39 (4), 341-350.

Kahneman, Daniel, Jack Knetsch, and Richard Thaler (1991), The Endowment Effect,Loss Aversion, and Status Quo Bias,” Journal of Economic Perspectives, 5, 193-206.

Kunda, Ziva (1990), “The Case for Motivated Reasoning,” Psychological Bulletin, 108 (3 November), 480-498.

Loewenstein, George (2001), “The Creative Destruction of Decision Research,” Journalof Consumer Research, 28 (3 December), 499-505.

McFadden, Daniel S. (1974), “Conditional Logit Analysis of Qualitative ChoiceBehavior,” in P. Zarembka, ed., Frontiers in Econometrics. New York: AcademicPress.

Moreau, Page, Aradhna Krishna and Bari Harlam (2001), “The Manufacturer-Retailer-Consumer Triad: Differing Perceptions Regarding Price Promotions,” Journal of Retailing, 77 (4 December), 547-569.

Park, C. Whan, Sung Youl Jun, and Deborah J. Macinnis (2000), “Choosing What I WantVersus Rejecting What I Do Not Want: An Application of Decision Framing toProduct Option Choice Decisions,” Journal of Marketing Research, 37 (May),187-202.

Prelec, Drazen, Birger Wernerfelt, Florian Zettelmeyer (1997), “The Role of Inference inContext Effects: Inferring What You Want from What is Available,” Journal of

Consumer Research, 24 (1 June), 118-125.

Schweitzer, Maurice (1994), “Disentangling Status Quo and Omission Effects: AnExperimental Analysis,” Organizational Behavior and Human Decision

Processes, 58 (3), 457-476.

Sen, Sankar and Eric J. Johnson (1997), “Mere-Possession Effects without Possession inConsumer Choice,” Journal of Consumer Research, 24 (1 June), 105-117.

Simonson, Itamar (1989), “Choice Based on Reasons: The Case of Attraction andCompromise Effects,” Journal of Consumer Research, 16 (2 September), 158-174.



43

Stapel, Diederik A. and Willem Koomen (1998), “When Stereotype Activation Results in(Counter)Stereotypical Judgments: Priming Stereotype-Relevant Traits andExemplars, Journal of Experimental Social Psychology, 34, 136-163.

--------------- and --------------- (2000), “How Far Do We Go Beyond the InformationGiven? The Impact of Knowledge Activation on Interpretation and Inference,” Journal of Personality and Social Psychology, 78 (1), 19-37.

---------------, Leonard L. Martin, and Norbert Schwartz (1998), “The Smell of Bias: WhatInstigates Correction Process in Social Judgments?” Personality and Social

Psychology Bulletin, 24(8 August), 797-806.

Wegener, Duane T. and Richard E. Petty (1995), “Flexible Correction Processes inSocial Judgment: The Role of Naïve Theories in Corrections for Perceived Bias,” Journal of Personality and Social Psychology, 68(1), 36-51.

--------------- and --------------- (2001) “On the Use of Naïve Theories of Bias to Remove

or Avoid Bias: The Flexible Correction Model,” Advances in Consumer Research,28, 378-383.

Wernerfelt, Birger (1995), “A Rational Reconstruction of the Compromise Effect,” Journal of Consumer Research, 21 (3 March), 627-633.

Wright, Peter (2002), “Marketplace Metacognition and Social Intelligence,” Journal ofConsumer Research, 28 (3 March), 677-682.



44

TABLE 1

STUDY 1: ATTRIBUTE LEVELS

Attribute Low Middle High

Number of Musical Voices (Keyboard) 16 32 64

Number of Musical Styles (Keyboard) 10 20 40

Monitor Size (Computer) 17” 19” 21”

Hard Drive Capacity (Computer) 10.0 GB 13.6 GB 16.8 GB

Meals Included (Vacation Package) None Breakfast Breakfastand dinner

Transportation Included (Vacation Package) None Car Car anddriver



45

TABLE 2

STUDY 1: CHOICE PROBABILITIESa

Default

Percent Choosing: Control(no default)

Low Middle High

Low Option 30.43% 39.77% 21.59% 23.48%

Middle Option 49.28% 41.67% 58.71% 50.76%

High Option 20.29% 18.56% 19.70% 25.76%

aThe shaded cells suggest an effect of default designation on choice of the default option: subjects were

most likely to choose a particular option when it was the default, compared to the control column to theleft.



46

TABLE 3

STUDY 1: RESULTS

(Conditional fixed-effects logistic regression)

Model 1 Model 2Independent Variable Coef.

(Std. Err.) Z Coef.

(Std. Err.) Z

Middle Option .5600(.0721)

7.769 *** .6328(.0960)

6.589 ***

High Option -.3172(.0881)

-3.598 *** -.2120(.1122)

-1.890 ***

Default (any) .3731(.0751)

4.968 *** --- ---

Low Default --- --- .5752(.1542)

3.731 ***

Middle Default --- --- .3120(.1498) 2.083 **

High Default --- --- .2122(.1707)

.1243

Log likelihood -1091.15 -1089.89 a

* significant at p ≤ .05, ** significant at p ≤ .01, *** significant at p ≤ .001

a

We also tested this model against a model restricting the low and high defaultcoefficients to be equal (LL = -1091.43).



47

TABLE 4

STUDY 2: CHOICE PROBABILITIESa

Default Condition

InformationCondition

Choice Control (NoDefault)

LowDefault

HighDefault

Total

Limited Low Option 54.2% 56.2% 37.8% 49.5%

High Option 45.8% 43.8% 62.1% 50.5%

Enhanced Low Option 45.6% 59.4% 52.1% 52.5%

High Option 54.4% 40.6% 47.9% 47.5%

Total Low Option 50.0% 57.8% 45.0% 51.0%

High Option 50.0% 42.2% 55.0% 49.0%

aTo see the effects of default designation, compare the choice of an option when it is the

default to its control in the left-hand column. For example, in the 2 nd row illustrating thelimited information condition, we can see that 62.1% of subjects in the high defaultcondition chose the high option, whereas only 45.8% of subjects in the control conditiondid so.



48

TABLE 5

STUDY 2: RESULTS

Model

% Choosing HighOption)

Model 2

(%Counterarguing)

Model 3

(% Choosing HighOption)

IndependentVariable

Coef.(Std. Err.)

Z Coef.(Std. Err.)

Z Coef.(Std. Err.)

Z

--- --- --- --- ---

Low Default -.0842(.3130)

-0.27 -.2630(.3722)

-0.71 -.4661(.5172)

-0.90

High Default .6611(.3227)

2.05 * -.9076(.3854)

-2.36 * -.0337(.5577)

-0.06

Information .3453(.3100)

1.11 -.7273(.3728)

-1.91 * -.3388(.5367)

-0.63

Low Default*Info -.4735(.3834) -.12 .8550(.4681) 1.83 * .1787(.6441) 0.28

High Default*Info -.9227(.4156)

-2.22 * 1.309(.4770)

2.74 *** -.0507(.7256

-0.07

Counterarguments --- --- --- --- 4.069(.3178)

12.81 ***

Constant -.1671(.2566)

-0.65 1.984 0.70 -2.1467(.4632)

-4.63 ***

Wald χ2 (df) 16.62 (5) 10.74 (5) 179.62 (6)

* Significant at p ≤ .05, ** significant at p ≤ .01, *** significant at p ≤ .001



49

TABLE 6

STUDY 3 – CHOICE PROBABILITIESa

Default Condition(Enhanced Information)

Expertise MarketplaceMetacognition

Choice Control(No Default)

LowDefault

HighDefault

Tota

Novice Low MM Low Option 60.0 61.9 44.1 54.3

High Option 40.0 38.1 55.9 45.7

High MM Low Option 37.5 45.5 47.2 43.0

High Option 62.5 54.5 52.8 57.0

Total Novice Low Option 49.0 53.5 45.3 49.1

High Option 51.0 46.5 54.7 50.9

Expert Low MM Low Option 36.2 40.0 41.7 39.2

High Option 63.8 60.0 58.3 60.8

High MM Low Option 39.9 31.0 38.3 36.0

High Option 61.1 69.0 61.7 64.0

Total Expert Low Option 37.5 35.6 39.8 37.6

High Option 62.5 64.4 60.2 62.4

Total Low Option 42.9 43.3 42.4 42.9

High Option 57.1 56.7 57.6 57.1

aTo see the effects of default designation, compare the choice of an option when it is thedefault to its control in the left-hand column. For example, in the 1 st row, we can see that55.9% of low-MM novices in the high default condition chose the high option, whereasonly 40.0% of low-MM novices in the control condition did so.



50

FIGURE 1

STUDY 1: SAMPLE STIMULUS

High Default, Number of Voices

Number of Voices

The number of types of sounds the keyboard can mimic; for example, you can makeeach note sound like a violin, marimba, human voice or choir, electric guitar, etc.

Circle or fill in the box for the level of attribute you prefer. [If you do not circle one,the manufacturer will automatically provide the level indicated by a mark.]

16 voices

32 voices ● 64 voices

$90 different$40 different

$50 different



51

FIGURE 2

STUDY 1: PERCENT CHOOSING DEFAULT VERSUS CONTROLa

9.4

-7.6

-2.1

-8.8

9.4

-0.6

-6.9

1.5

5.3

-10

-8

-6

-4

-20

2

4

6

8

10

Low Default Middle

Default

High Default

Choosing MiddleChoosing Low

Choosing High

ai.e., the percentage of subjects in the low default condition choosing the low option was

9.4 higher than the percent choosing this option in the control (no-default) condition.



52

FIGURE 3


2.0

13.8

16.3

-6.5

-10

-5

0

5

10

15

20

Limited Enhanced

.

High

Low

ai.e., the percentage of subjects in the limited information condition choosing the highoption was 16.3 higher than the percent choosing this option in the control (no-default)condition.



53

FIGURE 4


1.9

8.0

3.8

-8.9

0.6

-5.3-9.7

15.9

-30

-20

-10

0

10

20

30

Novices/Low MM Novices/High MM Experts/Low MM Experts/High MM

Low Default High Default

ai.e., the percentage of low-MM novices in the high default condition choosing the high

option was 15.9 higher than the percent of low-MM novices choosing this option in thecontrol (no-default) condition.

Documents

Skeptical Shoppers