Consumer Behavior II Case

Shaun Webbs

ConsumerBehaviorUnitIIArticle1AreConsumersRational.pdf

Home >Business & Finance homework help >Marketing homework help >Consumer Behavior II Case

Vol. 25, No. 1, January–February 2006, pp. 1–7 issn 0732-2399�eissn 1526-548X�06�2501�0001

informs ® doi 10.1287/mksc.1060.0196

Editorial

Are Consumers Rational? Experimental Evidence?

Steven M. Shugan∗ Warrington College of Business, University of Florida, 201 Bryan Hall, Box 117155, Gainesville, Florida 32611,

[email protected]

Despite some misconceptions, consumer rationality is a property of the researcher rather than the consumer.Consumers become more rational as we are better able to predict their behavior or other important out- comes influenced by their behavior. Perfect rationality results when we achieve accurate predictions. Conse- quently, at least for many Marketing Science articles, consumers are becoming more rational as we find better ways to predict. However, some experimental consumer behavior articles find the opposite. The difference between experimental and statistical controls explains the divergence in conclusions. Experimental controls test rationality based on whether previously absent variables exhibit significant explanatory power holding known explanatory variables constant. Statistical controls test rationality based on the incremental explanatory power of previously absent variables after accounting for known explanatory variables. Moreover, experimental tests tend to isolate consumer behavior predictions while statistical tests check for sufficient accuracy to choose among different firm strategies. Both perspectives are correct but ask very different questions.

Key words: bounded rationality; experiments; predicted choice; consumption; consumer behavior; econometrics

1. Rationality 1.1. The Importance of Consumer Behavior to

Marketing Most marketing activities seek to influence markets involving interactions among suppliers, competitors, regulators, the courts, government agencies, and cus- tomers. Some research topics, including contingent contracts (Biyalogorsky and Gerstner 2004), auctions (e.g., Shugan 2005), and exploiting historic data bases for marketing interventions (Rust and Verhoef 2005), examine marketing issues applicable in both con- sumer and business-to-business settings. However, most academic studies in marketing focus exclusively on consumer markets (Ankers and Brennan 2002), perhaps because we are all consumers. Consequently, the study of how marketing activities influence con- sumer buying behavior is a central part of the disci- pline of marketing. A clear and fundamental understanding of con-

sumer behavior should help us more accurately pre- dict consumer response to marketing interventions.

Editorial pages are not part of the regular Marketing Science page budget. We thank the INFORMS Society of Marketing Science for paying for all editorial pages. We also thank the Society for grant- ing every page supplement requested by the current editor.

We welcome and often post responses to editorials. Please see mktsci.pubs.informs.org. ∗Steven M. Shugan is the Russell Berrie Foundation Eminent Scholar in Marketing.

Those predictions would certainly be instrumental, if not invaluable, for designing more effective market- ing tactics and more profitable strategies. Understand- ing consumer behavior should allow both the iden- tification of the critical variables influencing behav- ior and the nature of that influence. It should also reveal which variables have relatively little impact on behavior and which marketing activities, conse- quently, might be ineffective.

1.2. The Debate About Consumer Rationality Given our great interest in consumer behavior, researchers in marketing sometimes find themselves entangled in debates about consumer behavior (e.g., see Firat et al. 1995, Howard and Sheth 1969). For example, one area of debate concerns consumer rationality. Sometimes, consumer rationality debates involve important implications for the effectiveness and implementation of numerous marketing activ- ities. Many marketing activities, such as signaling, require highly rational consumers (Kirmani and Rao 2000). Other marketing activities, such as the con- struction of loyalty programs, might require irrational consumers (e.g., Taylor et al. 2004). Unfortunately, these debates about consumer behavior, despite their fascinating aspects, are sometimes distracting, confus- ing, and divert research efforts into directions with no apparent direct impact on marketing activities. The manuscript review process reveals that some

researchers summarily dismiss marketing models that

assume extreme consumer rationality, i.e., extreme forms of optimal behavior where consumers effort- lessly ratiocinate through highly complex tasks with capacious memory. Other more moderate researchers suggest that marketing models should account for documented so-called departures from rationality found in experimental studies (e.g., Smith 2003). Some researchers, at the opposite extreme, summarily reject models that assume less than perfect rationality.

2. Some Definitions of Rationality Before discussing the debate over rationality, perhaps we should define the term “rationality.” As with other technical terms (e.g., utility, probability, product, opti- mization, equilibrium), the term “rationality” can con- vey different meanings in different disciplines. In fact, different meanings exist within the same discipline. Let us limit the following discussion to the meaning of rationality in the discipline of marketing and, pos- sibly, some sister disciplines. The everyday definition of “rationality” is “having

the ability to reason.” Technical definitions, in quest of precision, sometimes become far more complex and confusing. Confusion over the technical definitions of some technical terms often causes many unproductive debates about meaningless differences. Perhaps that confusion is, in part, deliberate. Researchers occasion- ally adopt less precise, simple everyday terms for their theories, both to better communicate the intended con- cept and to make assumptions appear more palatable. It seems far more reasonable to posit, for example, a normal distribution than to posit a Gaussian dis- tribution for the ubiquitous error term. Similarly, it seems more defensible to assume that consumers are rational, rather than to assume that consumers are adept optimizers, with perfect foresight and knowl- edge of the firm’s cost structure and never tire. In this sense, the usage of the term “rationality” is a market- ing strategy for selling (i.e., making more palatable) a set of technical mathematical assumptions that are sufficient for building a theory of consumer behavior. Like other assumptions, the attractiveness of “rational- ity” assumptions (as approximations to some complex real-world situations) will depend on whether the sub- sequent theory is able to explain (i.e., predict) impor- tant observables. In the economics literature, rationality is usually

associated with the sufficient conditions for the exis- tence of a consumer utility function (e.g., Malinvaud 1972). Traditional economic theory implicitly defines consumer rationality in terms of expected utility max- imization and a set of explicit axioms sufficient for utility functions to exist (Herstein and Milnor 1953). The econometrics literature defines rationality as util- ity maximization with an individual-specific addi- tive error term (Lewbel 2001). Game-theoretic applica- tions often define rationality as taking the best action,

given well-defined payoffs and rules of play (Bern- heim 1984). Hence, rational consumers do what is best for them in a context where all players (consumers, manufacturers, retailers, etc.) have different incentives (e.g., see Alba et al. 1997 for a discussion of conflict- ing incentives in interactive home shopping). Lipman (1991) defines rationality as choosing the best proce- dure for deciding. Of course, other disciplines have other definitions, including the idea that rationality is merely normal behavior.

3. The “Best-Action” Definition Most Marketing Science applications are consistent with the “taking-the-best-action” definition of ratio- nality. This definition implies that rationality is neces- sarily a function of the model (or theory) being pro- posed or tested because the best action depends on the postulated world of the model (e.g., parameters, decisions variables, relationships, measures). For example, when proposing a model of search

and consideration sets, Mehta et al. (2003) state that “consumer rationality implies that consumers will engage in price search to reduce [price] uncertainty.” Acquisti and Varian (2005) define consumer aware- ness of firm incentives to lower future prices as one property of rationality. Zwick et al. (2003) define opti- mal search behavior and the size of the consumer con- sideration set as properties of rationality. Akçura et al. (2004) define consumer learning as one property of rationality. Kalra et al. (1998) define consumer skepti- cism of manufacturer quality claims (i.e., without sup- porting evidence) as one property of rationality. Xie and Shugan (2001) argue consumer skepticism about service provider claims regarding future spot prices (i.e., that are not consistent with future spot profit maximization), ala Coase (1972), as one property of rationality. A variety of other factors might also pro- duce other definitions for rationality (e.g., dynamics, uncertainty, the preferences of others, cultural pres- sures, etc.). In sum, a rational consumer takes the best action

within the world of the model. Given that the dif- ferent models employ different decisions variables, different exogenous factors, different situations, and exhibit different properties, the precise meaning of the term “rationality” varies from model to model.

3.1. Why the Best-Action Assumption Is Really A Weak Assumption

The assumption that consumers will take the best action (within the world of the model) is often an extremely powerful assumption because it allows extraordinary consistency across and within myr- iad models that might appear completely unrelated. Hence, we can link diverse models related to advertis- ing budgets, promotions, advertising copy, shopping

behavior, and so on with this high-level assumption. We also get consistency between models of very dif- ferent phenomena (e.g., borrowing behavior and mar- riage). At first, this might seem like a strong assumption.

It is not. In virtually all situations, we could introduce ad hoc factors or arbitrarily modify the payoff func- tion to make any outcome appear best. We might, for example, allow consumers to consider the perceived fairness of the outcome, imagined legal constraints, perceived risks of litigation, social acceptability, pos- sible reputation effects, regret, intuition, and so on. A consumer might pay a higher price than necessary as a form of charity or a subsidy to help a valued firm stave off bankruptcy. A consumer might choose a lower-quality alternative as a means of experimen- tation (i.e., information gathering). A consumer might want to signal modesty in a social setting. Some consumers might deliberately try to make their own behavior unpredictable (as part of a more general strategy). Of course, some modifications might appear to resemble ad hoc ruses attempting to explain the irrational. This is not to say that all actions are reason-

able. Not all models are reasonable approximations of any conceivable real-world setting or real-world decision. This is only an argument that assuming that consumers take the best action is not as strong an assumption as it appears to be. The critical assump- tion, as argued later, is whether the model itself (i.e., the entire package of assumptions and condi- tions) provides a sufficient approximation of real- world settings. Moreover, outcomes might remain rational despite violations of the rationality assump- tions (e.g., see Mandler 2005).

3.2. Why Best Is Really Best Before arguing that model prediction is the key to testing rationality, we should concede that assum- ing that consumers do take the best action is still an assumption that warrants justification. Here are sev- eral justifications. 1. Most consumers would prefer to make the best

decision ceteris paribus. 2. The best action is often unambiguous (at least,

if the model is properly specified) and, hence, this assumption is directly testable—unlike assumptions that are less precise about which action will be taken. 3. Possible ambiguity related to the best action

alerts us of possible problems with the model’s spec- ification or formulation. 4. Given that firms seek to maximize expected prof-

its, assuming consumer maximization creates a sense of symmetry and consistency in the model formation. 5. Rather than requiring predictions for all con-

sumers, many marketing decisions need only consider

marginal consumers (i.e., only those few consumers who will change their purchase decisions—to buy or not—when we adopt a different marketing strategy). Hence, only marginal consumers need do what is best. 6. We are more interested in the eventual outcome

rather than in blips along the way (although, the blips are also interesting). Equilibria, for example, represent our targeted outcomes. 7. We would expect that learning and experience

would lead consumers toward the best actions. 8. When trying to persuade consumers, the conser-

vative assumption might be that we face the arduous task of persuading very astute consumers rather than the relatively easier task of fooling naïve ones.

3.3. A Practical Definition of Rationality Rather than quibbling with either the theoretical meaning of rationality or the particular rationality assumptions in any particular model, we should instead focus our concern on whether the rationality assumptions are sufficient to approximate the situa- tion being modeled. The key test is whether the model can accurately predict outcomes in that situation, at least, better than could be done without the model. Another way of looking at assumptions is that

the assumptions provide sufficient conditions when the model’s conclusions are justified. That viewpoint is true for every type of model (e.g., normative, descriptive, statistical, behavioral, aggregate, disag- gregate, etc.). The question is not whether the mod- eling assumptions are each good approximations for every situation or even most situations; the question is whether the model’s results are applicable in a suf- ficient number of situations so that the contribution justifies publication and application of the model. We hope that the conditions are sufficiently good approx- imations so that the model can accurately predict in a sufficient number of real-world situations.

4. Testing Whether Consumers are Rational

4.1. Rationality as a Model Property Inaccurate model predictions do not necessarily imply that reality is complex or unpredictable. High lev- els of uncertainty (in some situations) might only reflect an inadequate state of the art in modeling. As modeling technology improves, we expect that reality will appear simpler and more predictable. For exam- ple, navigation on the high seas was once onerous, but global positioning systems technology now allows accurate predictions and, consequently, easier naviga- tion. A similar argument is possible for consumer ratio-

nality. Consumers appear rational in situations in

which our models can predict their behavior. Con- sequently, consumers in well-studied choice situa- tions appear to exhibit high degrees of rationality because we have accurate models for these familiar situations. In other less-studied situations, consumers might appear irrational because our extant models are unable to accurately predict outcomes. In this sense, rationality is a property of our models and not a prop- erty of the consumer. The concept of a subjective probability is analo-

gous. The world is in some true state. For example, we might wonder whether the true box office of a movie is $1 million, $10 million, or $100 million. How- ever, there is some true box office. It is likely that time will reveal that true box office. In fact, we might know that true box office, but rather than using that information, we might predict it from other informa- tion to validate a model. A better model is better at predicting outcomes (i.e., explaining variance) than other models. However, the uncertainty in the out- comes (i.e., the variance) is a feature of the model and not reality. Reality consists of true states (which may or may not be known when predictions are made) while probabilities represent the researcher’s uncer- tainty about the true states. There are no correct prob- abilities, but there are correct predictions. Subjective probability reflects the researcher’s uncertainty. Simi- larly, irrationality reflects the researcher’s inability to predict behavior. Most marketing models (perhaps all) should be

tested on their predictions. Usually, predictions are made for qualitative or quantitative observations that are not used in the formulation, estimation, or calibra- tion of the model. Hence, a model should be capable of making predictions that we would be unable to make without the model.

4.2. What Is Being Predicted The prior argument suggests how we should test the rationality assumptions of a model. Given that con- sumer rationality assumptions are just a few of the many assumptions that comprise a model, it would be unproductive to test each assumption in isolation. Consider a road map that is a model of a geo-

graphic terrain. A particular map might show all the major highways but fail to show the location of hotels. The map model represents a simplification and approximation of the real geography. It can’t show every detail of reality, nor should it. It is difficult to evaluate, in isolation, whether ignoring lodging is a good or bad assumption. If the map is being used to navigate across the state, other assumptions in the map’s construction may trump the inclusion of lodging. If, in contrast, the user wishes to find lodging, ignoring hotels is a fatal flaw in the model. We are unable to evaluate the assumption in isola- tion. This argument also implies that the quality of

an assumption depends on the intent of the model, as well as on the other modeling assumptions. We are unable to conclude, in isolation, that some mod- els comprise more realistic behavioral assumptions than other models. A model for predicting industry sales, for example, might require different assump- tions about consumer behavior than a model attempt- ing to predict a particular consumer’s reaction to a direct-mail solicitation. Hence, the proper predictive test for rationality

assumptions need not focus on consumer behavior. Those assumptions only indirectly impact the valid- ity of the conclusions. For example, consider a model built to help select one of several new products for development. That decision might involve assump- tions related to consumer reactions, development fea- sibility, supply chain issues, costs, competitive reac- tions, inventory requirements, and so on. Whether a naïve consumer rationality assumption is an ade- quate approximation for expected consumer behavior depends on whether replacing that assumption with a more complex or realistic assumption would change the selection decision. In general, the adequacy of the rationality assumption depends on whether the assumptions lead to the adoption of the wrong mar- keting strategy, rather than on whether the assump- tions predict consumer behavior at some absolute level of accuracy. For example, the assumptions that consumers price shop at many or few outlets might each yield the same optimal marketing strategy when each assumption tends to yield the same prices across outlets. Of course, the rationality assumption might be

questionable if the model is unable to predict desired outcomes (e.g., profits, sales, market share) with suffi- cient accuracy to discriminate among strategies. Then, every assumption becomes suspect. Moreover, several assumptions could be flawed (i.e., bad approxima- tions).

4.3. A Brief Comment on Prediction Versus Explanation

Although the technical terms “prediction” and “explanation” certainly vary in meaning, this discus- sion treats the words as almost synonymous. Usually, after observing some qualitative or quantitative obser- vations, we propose a model or theory that explains those observations. We partially assess the validity of the theory or model by predicting different observa- tions (qualitative or quantitative). In some cases, the researcher arbitrarily defines explained observations (e.g., based on a point in time in the dataset, based on previous research at the time of submission, and so on). However, this distinction is less relevant here.

4.4. Irrationality Is the Default Assumption Authenticating irrationality is not necessarily our task. Our default assumption is that consumers are irrational, either because their behavior is inherently unpredictable or because we have not yet discovered how to predict it. The proof of rationality is straight- forward but, perhaps, daunting. We need only cre- ate a model that accurately predicts (i.e., explains the variance) in consumer behavior. If we are able to pre- dict consumer behavior as a function of the relevant variables in the situation of interest, we can conclude that consumers are rational (at least in that situation) and that our model accurately represents that ratio- nality.

5. Conflicting Findings on Rationality The prior reasoning suggests that consumers will appear to grow more rational over time as advances in model building technology ameliorate our ability to predict. For example, Wolfgang and Kannan (2005) discover how spatial multinomial models can bet- ter predict the spatial correlations among customer choices. Mittal et al. (2005) discover how customer satisfaction can better predict firm long-term finan- cial performance. Divakar et al. (2005) discover how to better predict microlevel consumer behavior. Nair et al. (2005) discover how aggregate data can better predict purchase incidence, brand choice, and pur- chase quantities.

5.1. Are Consumers Becoming More Rational? It seems clear that Marketing Science articles report increasing success at predicting consumer behavior— at least in purchasing situations. Moreover, many of these articles start with assumptions that are consis- tent with the strongest axiomatic representation of consumer preferences. Consequently, consumers are becoming more ratio-

nal because we are becoming better able to pre- dict their behavior. This greater ability to predict behavioral response to marketing interventions is also occurring at a more disaggregate level (e.g., Rust and Verhoef 2005). Although we have as yet not achieved perfect rationality, because consumer choice is not yet perfectly predictable, we are getting closer to achiev- ing that objective. It is also occurring with new forms of data including newsgroups (e.g., Godes and May- zlin 2004) and click stream data (Montgomery et al. 2004). However, some experimental consumer behav- ior articles appear to find the opposite. These arti- cles provide compelling demonstrations that influen- tial variables are absent from extant models.

5.2. Explaining Severely Conflicting Findings on Rationality

It might appear surprising that such a large num- ber of articles focusing on consumer behavior find

such a high level of irrationality (inconsistencies with typical extant rationality assumptions and the cor- responding models) among consumers. These arti- cles advocate inclusion of absent variables, including envy, relationships (Fournier 1998), framing, involve- ment (Zaichkowsky 1985), cognitive limitations, over- choice (Gourville and Soman 2005), social preferences, context effects, self-control, mental accounting, temp- tation, altruism, affective forecasting, bounded ratio- nality (Arthur 1994, Simon 1981), fairness, the diffi- culty of the decision (Shugan 1980), and so on. Moreover, many articles (e.g., Zeelenberg 1999)

claim that consumers are becoming more irrational, at least in the sense that these articles are find- ing more violations of the most common rational- ity assumptions. Loewenstein (1999), for example, states: “Despite the blossoming of the utility concept and expanding appreciation for the diverse determi- nants of utility, the list of human motives that have been codified in utility functions, and hence incor- porated into economic analyses, remains seriously incomplete.” Cohen and Dickens (2002) concede that behavioral studies have “been most successful in doc- umenting failures of the rational-actor model (e.g., failures of expected-utility theory, irrational coopera- tion, and time-inconsistent preferences).” The apparent conflicts in these findings and the

traditional assumptions in Marketing Science models sometimes cause debates between quantitative mod- elers who claim to have found high levels of con- sumer rationality (i.e., consistently with the model forecasts) and psychological researchers who tend to find high levels of irrationality (i.e., significant vari- ance explained by absent or overlooked variables). Despite appearances, there is little conflict between

these two different research streams. Differences in research objectives and differences in research meth- ods explain the differences in findings.

5.3. Experimental Controls Versus Statistical Controls

Most Marketing Science models focus on the amount of total variance explained by the model. Analytical models focus on whether qualitative outcomes are explained (i.e., occur when predicted). Statistical mod- els focus on whether quantitative outcome variance is explained (i.e., the difference between observed and predicted outcomes). Normative models, calibrated from past data, focus on whether the models produce the best strategies. In each case, the question asked is whether the model makes an adequately accurate prediction to discriminate among possible marketing strategies. For example, Hauser and Toubia (2005) find that

the errors and biases associated with adaptive metric utility balance (prior metric responses by consumers

are used to construct hypothetical choices for each consumer, keeping similar choice probabilities within each choice set) combined to less than the order of magnitude of typical response errors. Many consumer behavior models, however, use a

different criterion. These models start with theories or hypotheses that consider factors not commonly recog- nized by past research. These factors might be com- pletely absent from many extant analytical, statistical, and normative models. These consumer behavior articles provide unam-

biguous evidence that these new factors explain (or predict) a significant amount of consumer behavior. The obvious conclusion is that excluding these factors ignores important aspects of consumer behavior— hence, extant models of behavior are wrong. The key is that this experimental research asks a dif-

ferent question. This research employs experimental controls rather than statistical controls. Experimental controls test rationality based on whether previously absent variables exhibit significant explanatory power holding known explanatory variables constant. The question is whether there are still unexplored vari- ables that can alone significantly influence consumer behavior or enhance our understanding of consumer behavior (i.e., having the ability to predict behavior). Statistical controls test rationality from a different

perspective. These controls ask whether the incremen- tal explanatory power of previously absent variables is significant after accounting for known explana- tory variables. Consequently, if known explanatory variables are sufficient to produce predictions (either qualitative or quantitative) that are adequate for determining the best marketing strategy, we would be satisfied with known explanatory variables. More- over, in the quest for parsimony, stability, robustness, tractability, generality, and power, we would place greater value on models capable of isolating only the most critical variables that predict (i.e., explain the variance) in consumer behavior. With fewer variables, we are able to make more general predictions that are less dependent on factors that might be unknown in some situations. Both the experimental approach and the statis-

tical approach can yield remarkable insights. Both approaches can be extraordinarily useful, but they ask different questions, and each might be unable to answer the questions asked of the other.

5.4. A Brief Comment on Effect Size Note that this discussion regarding controls differs from arguments regarding the transparent report- ing of effect sizes in experimental inquiries (e.g., see Peterson et al. 1985). Effect sizes provide use- ful information about the absolute explanatory power of particular variables. However, as noted earlier,

effect sizes in experimental studies fail to consider the explanatory capabilities of variables held constant in the experiment. Although effect sizes do measure the total explanatory power of variables in experimental settings, while holding other variables constant, large effect sizes do not necessarily indicate large incremen- tal explanatory power after including known explana- tory variables. Finally, the strength of the manipulation often

determines the magnitude of the effect size. This could be problematic when the strength of the manip- ulation might not reflect the actual variance in real- world situations.

6. Conclusions Despite some misconceptions, consumer rationality is a property of the researcher’s model rather than the consumer. Consumer behavior appears more rational as researchers are better able to predict this behav- ior in more situations. Perfect rationality results when either consumer behavior is adequately predictable or when we can predict important outcomes influ- enced by consumer behavior. Consequently, at least for many Marketing Science articles, consumers are becoming more rational as new models more accu- rately predict consumer choice in more situations. However, some interesting experimental consumer behavior research finds the opposite. This research shows that extant models fail to consider critical variables that can explain significant variability in behavior. The illusion of conflict is resolved by understand-

ing the difference between experimental and statis- tical controls. This difference explains the apparent and dramatic divergence in the conclusions. Many articles in consumer behavior use experimental con- trols. Experimental controls test rationality based on whether previously absent variables exhibit signif- icant explanatory power holding known explana- tory variables constant. Hence, these articles ask whether previously unexplored variables have signif- icant explanatory power alone. Many Marketing Sci- ence articles use statistical controls. Statistical controls test rationality based on the incremental explanatory power of absent variables after accounting for known explanatory variables. Statistical controls ask whether absent variables have significant explanatory power beyond what is explained by known explanatory vari- ables. Both perspectives are correct, but they ask dif- ferent questions.

References Acquisti, Alessandro, Hal R. Varian. 2005. Conditioning prices on

purchase history. Marketing Sci. 24(3) 367–381.

Akçura, M. Tolga, Füsun F. Gönül, Elina Petrova. 2004. Consumer learning and brand valuation: An application on over-the counter drugs. Marketing Sci. 23(1) 156–169.

Alba, Joseph, John Lynch, Barton Weitz, Chris Janiszewski, Richard Lutz, Alan Sawyer, Stacy Wood. 1997. Interactive home shop- ping: Consumer, retailer, and manufacturer incentives to par- ticipate in electronic marketplaces. J. Marketing 61(3) 38–53.

Ankers, Paul, Ross Brennan. 2002. Managerial relevance in aca- demic research: An exploratory study. Marketing Intelligence & Planning 20(1) 15–21.

Arthur, W. Brian. 1994. Inductive reasoning and bounded rational- ity. Amer. Econom. Rev. 84(2) 406–411.

Bernheim, B. Douglas. 1984. Rationalizable strategic behavior. Econometrica 52(4) 1007–1028.

Biyalogorsky, Eyal, Eitan Gerstner. 2004. Contingent pricing to reduce price risks. Marketing Sci. 23(1) 146–155.

Brian, Arthur W. 1994. Inductive reasoning and bounded rational- ity. Amer. Econom. Rev. 84(2) 406–411.

Coase, Ronald H. 1972. Durability and monopoly. J. Law Econom. 15(1) 143–149.

Cohen, Jessica L., William T. Dickens. 2002. A foundation for behav- ioral economics. Amer. Econom. Rev. 92(2) 335–338.

Deighton, John A. 1997. Commentary on “exploring the implica- tions of the Internet for consumer marketing.” J. Academy Mar- keting Sci. 25(4) 347–351.

Divakar, Suresh, Brian T. Ratchford, Venkatesh Shankar. 2005. CHAN4CAST: A multichannel, multiregion sales forecasting model and decision support system for consumer packaged goods. Marketing Sci. 24(3) 334–350.

Firat, A. Fuat, Nikhilesh, Dholakia, Alladi, Venkatesh. 1995. Mar- keting in a postmodern world. Eur. J. Marketing 29(1) 40–56.

Fournier, Susan M. 1998. Consumers and their brands: Developing relationship theory in consumer research. J. Consumer Res. 24(4) 343–373.

Godes, David, Dina Mayzlin. 2004. Using online conversations to study word-of-mouth communication. Marketing Sci. 23(4) 545–560.

Gourville, John T., Dilip Soman. 2005. Overchoice and assortment type: When and why variety backfires. Marketing Sci. 24(3) 382–395.

Hauser, John R., Olivier Toubia. 2005. The impact of utility bal- ance and endogeneity in conjoint analysis. Marketing Sci. 24(3) 498–507.

Herstein, Israel N., John W. Milnor. 1953. An axiomatic approach to measurable utility. Econometrica 21(2) 291–297.

Howard, John A., Jagdish N. Sheth. 1969. The Theory of Buyer Behav- ior. John S. Wiley & Sons, New York.

Kalra, Ajay, Surendra Rajiv, Kannan Srinivasan. 1998. Response to competitive entry: A rational for delayed defensive reaction. Marketing Sci. 17(4) 380–405.

Kirmani, Amna, Akshay R Rao. 2000. No pain, no gain: A criti- cal review of the literature on signaling unobservable product quality. J. Marketing 64(2) 66–79.

Lewbel, Arthur. 2001. Demand systems with and without errors. Amer. Econom. Rev. 91(3) 611–618.

Lipman, Barton L. 1991. How to decide how to decide how to …. Modeling limited rationality. Econometrica 59(4) 1105–1125.

Loewenstein, George. 1999. Because it is there: The challenge of mountaineering … for utility theory. Kyklos 52(3) 315–344.

Malinvaud, Edmond. 1972. Lectures on Microeconomic Theory. North- Holland, Amsterdam, The Netherlands.

Mandler, Michael. 2005. Incomplete preferences and rational intran- sitivity of choice. Games Econom. Behavior 50(2) 255–277.

Mehta, Nitin, Surendra Rajiv, Kannan Srinivasan. 2003. Price uncer- tainty and consumer search: A structural model of considera- tion set formation. Marketing Sci. 22(1) 58–84.

Mittal, Vikas, Eugene W. Anderson, Akin Sayrak, Pandu Tadika- malla. 2005. Dual emphasis and the long-term financial impact of customer satisfaction. Marketing Sci. 24(4) 544–555.

Montgomery, Alan L., Shibo Li, Kannan Srinivasan, John C. Liechty. 2004. Modeling online browsing and path analysis using click- stream data. Marketing Sci. 23(4) 579–595.

Nair, Harikesh, Jean-Pierre Dubé, Pradeep Chintagunta. 2003. Accounting for primary and secondary demand effects with aggregate data. Marketing Sci. 22(3) 444–460.

Peterson, Robert A., Gerald Albaum, and Richard F. Beltramini. 1985. A meta-analysis of effect sizes in consumer behavior experiments. J. Consumer Res. 12(1) 97–103.

Rust, Roland T., Peter C. Verhoef. 2005. Optimizing the marketing interventions mix in intermediate-term CRM. Marketing Sci. 24(3) 477–489.

Shugan, Steven M. 1980. The cost of thinking. J. Consumer Res. 7(2) 99–112.

Shugan, Steven M. 2005. Editorial: Marketing and designing trans- action games. Marketing Sci. 24(4) 525–530.

Simon, Herbert Alexander. 1981. The Sciences of the Artificial, 2nd ed. MIT Press, Cambridge, MA.

Smith, Vernon L. 2003. Constructivist and ecological rationality in economics. Amer. Econom. Rev. 93(3) 465–508.

Taylor, Steven A., Kevin Celuch, Stephen Goodwin. 2004. The importance of brand equity to customer loyalty. J. Product Brand Management 13(4/5) 217–227.

Wolfgang Jank, P. K. Kannan. 2005. Understanding geographical markets of online firms using spatial models of customer choice. Marketing Sci. 24(4) 623–634.

Xie, Jinhong, Steven M. Shugan. 2001. Electronic tickets, smart cards and online prepayments: When and how to advance sell. Mar- keting Sci. 20(3) 219–243.

Zaichkowsky, Judith Lynne. 1985. Measuring the involvement con- struct. J. Consumer Res. 12(3) 341–352.

Zeelenberg, Marcel. 1999. The use of crying over spilled milk: A note on the rationality and functionality of regret. Philos. Psych. 12(3) 325–340.

Zwick, Rami, Amnon Rapoport, Alison King Chung Lo, A. V. Muthukrishnan. 2003. Consumer sequential search: Not enough or too much? Marketing Sci. 22(4) 503–519.