Workshop Abstracts
1. Process and context in choice models
Workshop Chairs: Moshe Ben-Akiva, Andre de Palma and Dan McFaddenDiscrete choice models typically represent a rather pure situation in which a single decision-maker faces a set of alternatives and makes a choice. The objective of the workshop is to develop a framework to accommodate a broader view of choice models. We propose to explore the enrichment of choice models to explicitly represent the context and process leading to a choice.
In general, a choice is an ultimate outcome of a stream of evaluations of expectations and potential strategies, or more generally of a planning process. A shopping activity, for example, will usually be preceded by a planning stage that may include a review of needs and budgets and preparation of a tentative shopping list. Moreover, individual decisions are often not made in isolation. Choice of travel mode is dependent of the availability of a car, which may depend, for example, on car use by a spouse. Residential location choice often involves consideration of two jobs and multiple schools. The decision about who gets to use the car or where to locate are therefore likely to follow bargaining process among household members that could also be described as part of a planning process leading to a choice.
The workshop will consider other context and process aspects such as formation of intentions, response to uncertainty, sociality and networking; and attempt to develop insights into how enhancements of choice models with explicit representation of context and process address known violations of rationality.
2. Improving Medical Decision Making: Perspectives of Patients, Medical Care Providers & Policy Makers
Workshop Chair: Ziv CarmonMedical decision making can be extraordinarily difficult. Common decision hindrances such as overconfidence, status quo biases, sunk cost, and immune neglect, can be particularly significant. Patient decisions are further complicated by such factors as consequences being probabilistic, need for tradeoffs between dissimilar options & attributes, & strong emotions such as dread. Worse yet, medical decisions are often unique as patients have little or no experience with such situations (as well as poor insight into objective & subjective consequences of different health states). Healthcare providers & regulators can be of limited help as they have limited access to patients’ values, expectations, & beliefs; also, their interests can be misaligned with those of patients, ethical dilemmas can be substantial, & their personal values & beliefs color their impressions.
More generally, medicine is a fascinating, complex, & exceptionally important context for decision making research. Medical decisions provide abundant opportunities to explore how known phenomena interact with & are different under unique circumstances (e.g., how mixed emotions such as fear, stress, & hope, affect important decisions; how probabilistic information on important outcomes is processed). Medical decision research can also facilitate discovery of new insights that can inform other decision domains (e.g., knowing when & why it is detrimental to empower patients to choose can generally inform when & how to help people choose; studying people engaging in high risk behaviors such as unprotected sex can offer broader insights on such topics as risk seeking, self control, & intertemporal discounting). Finally, substantial private & public welfare implications of medical outcomes, provide ample opportunities for simple interventions that build directly on what we know about human decision-making to ‘nudge’ people (e.g., alter default options; encourage precommitment; modify food distribution in schools; cf. Thaler & Sunstein) to make better decisions.
There is often impressively little dialogue among medical decision researchers who work independently on related questions. Different research teams may originate from different scientific disciplines, & use different approaches & methodologies (which could be beneficial); but for a variety of reasons teams work independently, do not communicate with others during the research process, & even after research is complete communication is sparse as researchers may attend different conferences, publish in & read different journals, etc. In contrast, this workshop will bring together leading scholars from a wide spectrum of fields (medicine, healthcare management, psychology, economics, neuroscience, law, & business, among others). In the workshop, we plan to synthesize & integrate extant research in different disciplines relating to a few key questions. A preliminary set of potential questions assembled by a subset of participants include: detrimental effects of patient empowerment; harnessing placebo responses; communicating risk information; balancing ethics & effectiveness; unintended consequences of regulation & rationing; effects of monetary incentives on patients & providers; promoting healthy lifestyles; counteractive research priorities; how medical decisions research can inform decision research in other domains.
In conclusion, we believe that this workshop is worthy acceptance to the conference because medical decision making is a very important topic both theoretically & a practically, & due to the exceptionally high quality & breadth of its participants.
3. Understanding Choice Behavior when Agents Interact with Each Other
Workshop Chairs: Victor Aguirregabiria and Andrew ChingThis workshop will discuss ongoing research on different important issues and challenges in the structural empirical analysis of choice behavior under environments where agents interact with each other. Our definition of this type of environments is broad. It includes, naturally, static and dynamic games but also situations where agents are forward-looking and therefore, they interact with themselves in the future. For the last two decades, the structural modeling literature has combined game theory, dynamic programming, and econometrics to develop a new framework to empirically study choice behavior under this type of strategic environments, and to predict agents’ behavior in counterfactual scenarios. The literature has made significant progress in developing new techniques to implement this approach. Some recent examples include empirical bargaining models that explain delay in reaching an agreement (e.g., Diermeier, Eraslan and Merlo, 2003) or that endogenize prices (Merlo, Ortalo-Magne and Rust, 2006; Chen, Yang and Zhao, 2008), models that explain when to buy and stockpile (e.g., Hall and Rust, 2003 & 2007; Erdem, Imai and Keane, 2003; Hendel and Nevo, 2006), models that capture how consumers learn from others (e.g., Ching, 2009a), and how firms take advantage of this when setting prices (e.g., Ching, 2009b), models of dynamic competition between airline networks or between retail networks/chains (Aguirregabiria and Vicentini, 2007; Aguirregabiria and Ho, 2008), models of firms’ entry-exit decisions (e.g., Hong, Gallant and Khwaja, 2009), models of sales-force management (e.g., Chung, Steenburgh and Sudhir, 2009; Chan, Li and Pierce, 2009), models of intra-household behavior (e.g., Echevarria and Merlo, 1999; Yang, Zhao, Erdem and Zhao, 2009), and models of investment in individual’s health (e.g., Khwaja, 2009).
To reduce the computational burden of estimating this type of models, several methods have been proposed recently. Aguirregabiria and Mira (2002, 2007), Ching (2000, 2009a), Hu and Shum (2008, 2009) develop different non-parametric approaches to approximate agents’ policy functions and to avoid the repeated computation of equilibria in the model. Imai, Jain and Ching (2009) develop an algorithm that uses the past outcomes of the algorithm to approximate the expected future value to estimate models with forward-looking agents. These methods allow us to estimate fundamental structural parameters when we have data of a stable equilibrium environment. However, the ability to conduct counterfactual policy experiments has been one of the main motivations of estimating this type of models. When we change the environment, most of the literature so far has simply assumed that agents will eventually learn how to play equilibrium strategy and focus on studying the outcomes of the new equilibrium. Little is said about how long it might take to get to the new equilibrium. For models with multiple equilibria, we have known very little about which equilibrium will be selected after some policy parameters have been changed. In fact, little is known about how to use experimental or non-experimental data to investigate equilibrium selection mechanisms empirically, and to implement counterfactual experiments in models with multiple equilibria.
There are four main focuses in this workshop. Participants who will lead the discussion are listed in parentheses in alphabetical order.
(i) Discuss new estimation techniques, structural modeling or semi-structural modelling approaches that have become available in the past few years, and critically evaluate their pros and cons, as well as different alternative approaches to implement counterfactual experiments in models with multiple equilibria, and allow for serially correlated unobserved factors. (Aguirregabiria, Ching, Khwaja, Shum)
(ii) What have we learnt in the last few years from the applied works using these new estimation/modeling approaches? We survey their recent important applications, and study how these techniques have improved our understanding in areas such as marketing, health economics, industrial organization and political economy. We will also discuss potential applications of these methods and models in other areas. (Merlo, Nevo, Sudhir, Zhao)
(iii) In the context of empirical dynamic games, we discuss new methods to reduce the dimensionality of the state space such as oblivious equilibrium, continuous time, aggregation of state, limited information/bounded rationality of players, random grids, etc. (Aguirregabiria, Nevo, Rust)
(iv) Equilibrium selection or speed of convergence to new equilibrium strategies crucially depend on how agents adapt or learn from a new environment. How do we model choice behavior when agents’ beliefs are not in equilibrium? We have limited amount of real world data to study this type of important situations. How can we take advantage of experimental economics and use laboratory to generate data to enhance our understanding of how agents learn, adapt, and improve their strategies over time in disequilibrium? (Ching, Merlo, Rust, Shum)
We also plan to invite a few experimental researchers to join us (they are listed under “Other potential participants” in the cover page). With their knowledge and understanding of how agents behave under control environments, and their experiences in generating data in laboratory settings, we believe that they will have significant contributions to our agenda (iii) and (iv). In fact, three of our confirmed participants (Merlo, Rust, Shum) have been working with experimental researchers in this direction. We expect that they will bring many new insights and ideas about alternative modeling approaches that this research area could consider in the next few years.
4. Twenty Five Years of Preference Construction – Re-defining What it Means to Choose
Workshop Chairs: Ravi Dhar and Itamar SimonsonIt has been almost twenty-five years since behavioral decision researchers advanced the viewpoint that consumer preferences are not well defined but rather constructed in the process of making choices and judgments (Slovic 2006). Although the BDT literature has been critical of the normative framework to which it has most often compared its findings, both viewpoints assume that the person is consciously involved in evaluating options in the process of making a choice. For example, with few exceptions, most of the work in BDT can be characterized by four characteristics: 1. Focus on “isolated” or single-shot choices, 2. Design a test of construction vs. rationality principle based on some criterion of consistency or coherence, 3. Tests for mechanisms underlying the demonstration, based on cognitive processes operating under awareness of the decision maker. These characteristics built a strong foundation that led to rapid progress and creative demonstrations.
Like most foundations, the scaffolding also restricted the nature of phenomena and processes examined about consumer preferences. The advances in social psychology in the last twenty years or so have provided an incredible opportunity to re-define and re-examine the core knowledge around consumer choice and valuation and the purpose of future BDT research. For example, a question on how prior choices influence subsequent ones has generally been ignored because consistency principle offered no clear way to test for rationality when one moves beyond single shot choices (e.g., Dhar and Simonson 1999). In contrast, although research on conscious and non-conscious goals has important implications for understanding what impacts valuation and choice, it has been treated as mere “background noise” that is not worthy of much attention. Another example is the role of metacognitve experiences in affecting valuation that challenges the focus of prior work – to move beyond what people think about and incorporate how people feel about the information.
This session will bring together established and emerging leaders in the field of behavioral decision theory and social psychology to examine the fundamental questions around valuation and choice. The session will attempt at identifying some core psychological variables that are important to choice but yet not the principle focus of BDT researchers. The session will also propose a research program to frame the next generation of questions for BDT research.
5. Empirical Models of Decision-Making Under Uncertainty
Workshop Chairs: Bart Bronnenberg, Jean Pierre Dubé and Wes HartmannThe session will focus broadly on “the empirical role of beliefs in decision-making under uncertainty.” Beliefs are naturally unobserved in empirical analyses of revealed preferences. In practice, the researcher does not observe how an agent uses her current information to form a belief. In some instances, the information set itself may be partially or entirely unobserved to the researcher. The goal of this session is to highlight the advantages and disadvantages of how researchers have adapted to these problems in various empirical settings, while also broadly considering the non-parametric identification of preferences and beliefs in empirical settings. We will cover the types of data that would be amenable to estimating preferences and beliefs in practice. In addition, we will cover practical empirical problems such as consumer search, learning, and purchase timing in general. A large portion of the session will likely focus on the discrete choice modeling paradigm.
The econometrics of discrete choice models under uncertainty are subtle. In general, even when all the relevant state variables are observed, the distribution of an agent’s beliefs and her preferences (the deep structural parameters) are not jointly identified in a non-parametric sense with standard choice data. The standard approach consists of assuming that the agent forms rational (self-fulfilling) expectations. In this manner, beliefs can be estimated directly from the data. This assumption is typically made to circumvent the lack of available data on beliefs, yet has undesirable characteristics such as identical beliefs for all agents.
Identification becomes even trickier when a temporal element is added to the inference problem. In the context of the dynamic discrete choice model, beliefs, preferences and the discount factor are not jointly-identified in a non-parametric sense with standard choice data. Beliefs need to be expressed as a distribution, which is typically not known to the researcher. As above, researchers typically assume rational expectations so that this distribution can be inferred from the data. In addition, applied researchers need to make an assumption about the discount factor. A seemingly natural assumption is that agents discount future consumption according to the real rate of interest.
However, there is no a priori reason for an agent to discount future consumption according to the market interest rate. Similarly, agents may not all have the same degree of patience. The assumption also lies at odds with a large body of behavioral research with evidence supporting much more impatience than would be implied by the real rate of interest. The behavioral literature also routinely documents evidence of hyperbolic, as opposed to geometric, discounting. For instance, agents are often found in lab environments to exhibit an inherent taste for immediacy.
The session will tackle these issues from two angles. The first would consist of overviewing the econometric issues raised above more formally. This discussion would involve enumerating the various ways in which choice data could be supplemented to relax the assumption of rational expectations and, when applicable, the assumption of geometric discounting according to the real rate of interest. The second angle would consist of several practical applications in which the standard assumptions are imposed, critiqued and, if possible, relaxed.
6. Coming Full Circle: Studying Choice in the Field
Workshop Chairs: Gavan Fitzsimons and Duncan SimesterPrior to the mid-1950’s virtually all investigations of how people made choices occurred in the field. This was particularly true in consumer choice settings in which the vast majority of “research” performed prior to the mid-1950s consisted of case style studies of how consumers made choices. Unfortunately, as James Howell, the coauthor of the Ford Foundation’s 1959 report on the state of research and teaching in business schools stated, this led to a situation in which “Collegiate schools of business, with a few notable exceptions, were regarded as the slums of the educational community.” The Ford Foundation started pouring money into schools of business to transform how knowledge was generated, and how this knowledge would ultimately be disseminated. The results were largely positive, as highly trained researchers from economics, psychology and statistics began entering business schools and taking up the study of choice in earnest. One potentially negative byproduct of this transformation was a residual suspicion of studies of choice that generated their insights in the field, rather than in the laboratory, or through increasingly sophisticated mathematical models. This proposal seeks to address this shift and highlight some of the strengths of returning to the field to study choice.
After 50 years of incorporating insights, theories, and techniques from many base disciplines into the study of choice the current techniques used by choice researchers are often rewarded based on methodological sophistication. Increasingly, however, this technological advance comes at the cost of connection to the “real” world. One obvious fallout from this disconnect is that decision makers in companies view much academic research as “interesting, but not terribly germane” to their everyday business. An advantage of encouraging a return to the field will, in our opinion, increase the impact that choice researchers have on the world around us.
Interestingly, researchers in other fields have not abandoned field study to the same degree as those studying choices. Anthropologists, sociologists, and their counterparts within business schools (e.g., organizational behavior faculty studying relationships, leadership, network effects etc) have continued to maintain a strong presence in the domain of field research.
This is a particularly appropriate time to re-evaluate the role of field experiments in choice research. The introduction of new electronic media has already led to an initial resurgence of interest in this topic. As a result there is a new body of literature using field data collected from eBay, Amazon and other electronic marketplaces. We believe that these new seeds of interest provide a springboard that may spark a broader willingness to embrace field data.
7. Marketing and Politics: Models, Behavior, and Policy Implications
Workshop Chairs: Mitchell Lovett, Brett R. Gordon and Ron ShacharCandidates in the 2008 U.S. presidential election spent more than $711 million on media and advertising alone, nearly a three-fold increase over the previous presidential election. A key factor driving this growth is the increasing recognition that a candidate’s marketing campaign plays a critical role in the ultimate election outcome. Political campaigns employ every possible marketing tool, from direct mailings to television advertising to social media. Researchers are interested in measuring and understanding how marketing activities influence political participation and voting decisions, how candidates choose their policy positions and design their marketing strategies, and how regulators and policy makers may affect participation and oversee campaign spending.
Despite the inherent importance of marketing in political campaigns, a paucity of research exists in marketing on the topic. Research in political science itself is still equivocal on fundamental issues such as the general effects of advertising and the specific effects of negative advertising on voter behavior (Lau et al. 1999; Huber and Arceneaux 2007). Thus, both new and unresolved questions about political campaigns are fertile ground for choice research that exploits new methods and data sets.
First, recent methodological developments in structural modeling make it possible to frame politicians’ behavior as games of strategic interaction (Dubé et al 2005). This allows the researcher to conduct a rich set of counterfactual policy simulations, from understanding how advertising would be reallocated under alternative electoral college systems (Gordon and Hartmann 2009) to candidate's decisions to use positive and negative advertising (Lovett and Shachar 2009a). Modeling political campaigns presents several methodological challenges because candidates may possess asymmetric information about competitors’ strategies and face uncertainty about voters’ perceptions of the candidates' policy positions. Incorporating these factors into a structural model could help improve our understanding of political competition and might lead to new methods that are more broadly applicable.
Second, elections could serve as a test bed to help create new marketing theories and refine existing conceptual frameworks through application to the political science domain. Politics presents an important domain in which to study behavioral theories. For example, recent work shows that the entry of a third-party candidate creates an attraction effect that can permanently shift voting outcomes even after that candidate exits the election (Hedgecock, Rao, and Chen 2009). More behavioral research is needed to understand how various decision-making biases documented in a product choice context do or do not map onto the choice of a political candidate. Further, the political context may provide a better setting for identifying and testing new biases (e.g., Orhun and Urminsky 2009).
The importance of political campaigns and elections cannot be understated. Many academic disciplines study related topics, including political science, political economy, industrial organization, sociology, psychology, and marketing. Each discipline emphasizes a slightly different set of research questions and applies different techniques.
The Choice Symposium provides an opportunity to bring together researchers across these disciplines to share diverse perspectives and provide a common vocabulary to foster cross-disciplinary research. These interactions may spur new models that incorporate behavioral research and that enable marketing researchers to address questions that still vex political scientists. In terms of participants, as of this proposal, we have confirmed acceptances from top scholars in marketing, political science, economics, industrial organization, and psychology.
In summary, we believe our proposal focuses on an important and timely topic that spans multiple disciplines. We expect the output of the session to be beneficial to academic researchers in several disciplines, industry participants, and policy makers. For academic researchers, the session will provide a relevant and complete picture of the current state of research as well as new approaches to solve open problems in this area. The session will also serve as an introduction for academic marketers to the relevant body of knowledge. For industry, the session will provide a new perspective on the role of marketing in political campaigns that integrates and introduces new tools for analysis. Finally, policy makers may benefit from the session output in terms of the design and implementation of social programs to increase political participation and oversee campaign spending.
8. Empirical Examination of Behavioral Models of Managerial Decision-Making
Workshop Chairs: Avi Goldfarb and Teck-Hua HoHow do managers make choices? The dominant paradigm in empirical and theory work is to assume that manager choices are guided by fully rational models. These models often assume managers play a Bayesian Nash Equilibrium, solve a dynamic optimization problem, and seek to maximize the present value of current and future earnings. An increasing amount of laboratory research, however, has documented that these (and other) standard assumptions are often violated.
The proposed workshop will examine the extant empirical research on the rational decision-making in managerial decisions. The empirical research to be examined involves both laboratory experiments and field studies. We organize these ideas around two broad themes: non-standard solution methods and alternative utility functions. Non-standard solution methods involve boundedly rational approaches to manager choices including limited thinking steps in games and inability to make optimal decisions. Alternative utility functions include reference dependence, hyperbolic discounting, and social influences such as fairness and altruism.
The proposed workshop participants have studied both non-standard solution methods and alternative utility functions in both the laboratory and in the field. Based on this evidence, we will discuss when the fully rational model works well and when it needs to be supplemented with “behavioral” assumptions. For the situations where the fully rational model does not appear well-supported, we will examine which behavioral models have the most promise for increasing our understanding of managerial decision-making. In addition to discussing the small existing literature that applies behavioral models to manager decisions, we will identify the most promising areas for future research. Recent reviews of the literature include Ho, Lim, and Camerer (2006) and Della Vigna (2009). Our focus on bounded rationality and alternative utility functions in managerial decisions that sets our proposed workshop apart from the extant literature.
9. The Interactions of Perception, Learning, Thinking, and Feeling as Components of Adaptive Decision Making
Workshop Chair: Wes HutchinsonThe goal this workshop is to bring together people who work on focused issues in the areas of perception (e.g., visual attention, object and scene perception, aesthetics, perceptual fluency, spatial navigation & cognitive maps, etc.), learning (e.g., declarative and non-declarative memory, expertise, classical and instrumental conditioning, etc.), thinking (e.g., analogical, causal, and deductive reasoning, creativity, planning, problem-solving), and feeling (e.g., affect, emotion, motivation, hedonic experience, etc.) to exchange ideas about how these components interact in the service of goal-oriented, real-world decisions. Ideally, the result of the presentations and discussion will be an explicit "consensus theory of everything." We all have implicit notions of how our drilled down research fits into the big picture but this is seldom explicitly described, even in graduate level texts, and when researchers explicitly work on integrative models (e.g., cognitive architectures) they are typically very careful in claiming to have only integrated a few components of a much larger cognitive system (e.g., Anderson's How Can the Human Mind Occur in the Physical World).
This big picture goal is more modest than it sounds. We just want to make explicit what we already believe and see if there is, in fact, a high level of consensus -- and learn a lot about adjacent research domains in the process. Such consensus models can be very useful to researchers working in more molecular traditions such as sociologists and economists who would like a simple, but empirically accurate model of individual behavior to use as building blocks for more aggregate models. An initial integrative hypothesis (that may well be rejected early on) is that the four major components often interact in working memory as the result of the allocation of attention, but they also function independently in a variety of non-conscious and automatic processes.
In addition, our discussion will focus on the ways in which experimental tasks (which have been very carefully crafted to examine single sub-components) can be integrated to account for decision behavior in the real world. An interesting example is explaining the phenomenon recently described in the New York Times of how people who lived and worked in Manhattan in 2001 thought 9/11 would change their lives in many ways, but in fact, they were generally wrong in their predictions. Clearly, perception, learning, thinking and feeling all played a role in causing this discrepancy, but what was that role and will people now update their current expectations about the future (and this is happening again with the economic crisis). Of course, there are many more mundane real-world integrative phenomena such as car ownership (buying, driving, maintaining, etc.), professional decision making (business managers, public officials, etc.), and children in school (listening, working, taking tests, playing, etc.).
10. Constructing a Choice Architecture
Workshop Chair: Eric J. JohnsonThis group will structure a discussion focusing on certain topics with a goal of summarizing what we know about constructing an environment that aids in decision-making. For example, one topic might be identifying the right number of alternatives to present decision-makers. It would be useful to keep in mind certain possible applications or case studies.
One possible outcome of this would be a multi-authored or edited volume meant to be a guide or handbook on choice architecture. This is not certain and participating in the group does not obligate participants to such a project emerges. Each participant has a particular expertise in an area surrounding choice architecture and its application. Together, they represent an ideal team for such an exercise.
In addition, we may add one or more people with particular expertise. They might be interested in human-computer interface and another goal would be to involve real policy makers.
11. Dynamic Choice Models: Prospects and Problems
Workshop Chairs: Michael Keane and John GewekeThe term “dynamic choice model” actually encompasses a wide range of different types of models. In the earliest dynamic models, such as Guadagni and Little (1983), dynamics arise because lagged purchases affect current utility evaluations (“habit persistence”). An alternative meaning of the word “dynamic” refers to models where consumers make current choices not just to maximize current utility but also to maximize the expected present value of utility over a planning horizon. These could be called “forward-looking dynamic models.”
The first empirical model of forward-looking behaviour in the marketing literature was that by Erdem and Keane (1996). In that model, consumers have uncertainty about product attributes, and learn about attributes over time via use experience and ad signals. The model is “dynamic” in both senses noted above: Lagged signals about attributes affect current utility evaluations. And a consumer may buy a product that gives less than the highest expected utility on the current choice occasion, because he/she wants to learn about that product. This type of learning model has since been applied to many problems in both marketing and economics.
A second type of forward-looking dynamic model is the inventory model, which has been examined in recent work by Erdem, Imai and Keane (2003) and Hendel and Nevo (2006). A recent paper by Erdem, Keane and Sun (2009) found that inventory and learning models give rather similar predictions for patterns of dynamics in choice behaviour, so one key problem facing the area of dynamic models is how to distinguish between and/or integrate these two sources of dynamics.
It is important to note that the older Guadagni and Little (1983) type of habit persistence model could also be made forward-looking. That is, if consumers know that their purchases will shift their future preferences, it would be rational for them to take this into account when making current purchase decisions. Such models have been applied to study demand for addictive drugs in economics (although that work only considers Euler equations and not full structural choice model estimation). But it seems such models have rarely if ever been used in marketing. Such models do seem quite applicable to many types of goods, particularly luxury goods (e.g., I might maximize current utility by flying business class or staying at a 5-star hotel or buying a bottle of Chateau Margaux, but if I get used to that level of quality it might be bad for my future level of consumption enjoyment and/or my future bank account.)
Conversely, it is important to note that both learning and inventory models can be made “non-forward looking.” For instance, Roberts and Urban (1988) had a model where consumers are Bayesian learners but are not forward-looking. And recently Ching, Erdem and Keane (2009) have developed an inventory model where inventory affects the probability of considering a category, but where agents are not forward looking. In both cases, there is no clear evidence that the forward-looking model fits better than the non-forward looking model.
Thus, the literature appears to contain 6 main types of dynamic model, obtained by crossing the source of dynamics (habit persistence, learning, inventories) with whether or not consumers are forward looking. One important problem for the literature is to ask which of these models fits consumer behaviour better. But in order to do that, one must first ascertain how the predictions of the various models differ. For example, one model might predict that the increase in sales following a price promotion stems mostly from brand switching, while another implies it stems mostly from purchase acceleration. A key topic for the workshop to address is how we can best assess the different models’ ability to match “facts” about consumer behaviour.
A strong possibility is that no one model is best, and that some combination of the models would be superior. For instance, consumer behaviour in some categories may well exhibit aspects of learning and inventory behaviour, along with habit persistence. At the moment it seems computationally impractical to incorporate all of these aspects of behaviour into one model, but the workshop could discussion ways in which it might be done.
Another strong possibility is that the “best” model depends on the product category. That is, learning may be much more important in some categories than in others, and the same is true with inventories and habit persistence. The workshop could discuss the evidence on this issue, as well as methods that might be developed to make such judgements.
Of course, the assumptions that consumers are either (i) not forward looking at all (i.e., they are myopic), or that they (ii) have fully rational expectations about the future, are both rather extreme. An obvious topic for the workshop to discuss is the possibility of developing intermediate models where consumers take future payoffs into account when making purchase decisions but where they may not do so fully rationally. In addition, perhaps there are important sources of dynamics besides the three mentioned above to which the literature has paid insufficient attention.
Finally, the workshop will assess what we have learned from dynamic models to date. To give a key example, one strength of dynamic models is that they allow us to decompose the increase in sales from a price promotion into that due to (i) brand switching, (ii) purchase acceleration and (iii) category expansion. This is a key issue for managers. The workshop should assess whether a consensus has been reached on this issue, as well as whether the answer differs by product category.
12. Price Discrimination in Service Industries
Workshop Chairs: Anja Lambrecht, Katja Seim and Naufel VilcassimAcross industries, firms in recent years have introduced wide menus of offerings that grant access to virtually identical services under different pricing structures. This includes services as diverse as communications (landline or mobile telephony, internet access), media and entertainment (satellite or cable TV, digital music, DVD rental, amusement parks), transport (flights, car rental, public transport), or utilities (electricity, gas, water).
The prevalence of new and complex price discrimination strategies has inspired intensive research in marketing, economics, and operations research. The work to date has started to shed light on substantive and methodological challenges in the choice of price discrimination strategies, such as nonlinear pricing plans and bundling. Substantively, research aims to understand what drives customers’ discrete tariff choices and their continuous usage choices and, as a consequence, how firms should structure pricing plans. Such research relies on economic models of consumer decision making, but also allows for potential deviations from rational choice that may affect such decisions.
Methodologically, the nonlinear pricing structure introduces difficulties into the accurate modeling of the customer’s choice decisions between alternative offerings. Further complications arise from consumers’ uncertainty and learning that introduce dynamics into customer behavior over time. Models of firm decisions are equally challenging since profit maximizing pricing strategies are dependent on at most partially observed customer demand and need to pin down a larger number of pricing elements than under single-price strategies.
The objective of this session is to provide an interdisciplinary forum to discuss the current state of research on price discrimination in services and to develop directions for future research in this area. This will include a discussion of challenges in understanding customer choices, firm choices and how to appropriately model such decisions.
13. Towards a Theory of Scale
Workshop Chairs: Joffre Swait and Jordan Louviere“Scale” refers to the standard deviation of the unobserved or random component of utility in random utility based theory choice models. Historically, researchers in marketing and applied economics have viewed the unobserved component as unobserved effects that may be known to the chooser, but are unknown and/or unobserved by the researcher. Psychologists have long known that choices can vary from set to set or occasion to occasion within an individual, and thus there can be inherent randomness in humans that is unrelated to unobserved attributes or processes. While it was recognized by the 1980s (eg, Ben-Akiva and Lerman 1985) that parameter estimates in choice models inversely depended on scale, it was largely viewed as an unavoidable identification issue that had no real impact on understanding and predicting choices, with the possible exception of comparing data sources across space and time or ways of collecting data. That is, it was recognized that for whatever reason, some data sources exhibited more noise or stochastic variability than others. The latter view is evident in the Morikawa, Ben-Akiva and McFadden (2002) comparison of real and stated choices.
An evolution of Morikawa’s work was a stream of research comparing real and stated choices and/or comparing sources of preference data summarized in Louviere, Hensher and Swait (2000, Chapter 13). Generally, speaking it now is widely recognized that one needs to take scale differences into account when comparing sources of choice data, but this recognition has not extended very far beyond choices into the more general area of limited dependent variables. Hence, many researchers in many fields remain unaware of the issues associated with failing to take unobserved variability differences into account in data source comparisons (although, see Mood 2008).
Since 2000 a new view has emerged that sees scale as a much more fundamental and behavioral phenomenon. For example, Louviere (2001) proposed some tentative ideas about ways in which one should expect scale to differ behaviorally, the workshop on “dissecting the random component” (Louviere, et al 2002) considered ways in which scale would be expected to vary, and Swait and Adamowicz (2001a,b) demonstrated that scale varied with choice task complexity. Even more recently, Salisbury and Feinberg (2009) showed that failing to take scale differences into account within choice experiments can lead to incorrect and misleading conclusions about processes, such as misinterpreting higher choice variability associated with future compared with present choices as “inter-temporal discounting”. Similarly, Louviere and Eagle (2007) reviewed many results “within experiments” and showed that failure to take scale differences into account is associated with incorrect and misleading claims about process, such as claims about 1) the order in which individuals “see” choice scenarios being associated with “preference evolution” over choice occasions, when in fact what evolves is scale or error variability, or 2) preference heterogeneity, which is in many cases examined largely due to between-person error variability, or at least such variability is confounded with preference heterogeneity (see also, Louviere, Islam and Pihlens 2009).
More recently, researchers have developed new ways to conceptualize the role of scale, such as the scale-adjusted latent class model of Magdison and Vermunt (2007), models for single individuals (Louviere, et al 2008), the G-MNL model of Fiebig, et al (2009), and a new comprehensive and integrated theory of structural equation systems and choice models (Rungie, Coote and Louviere 2009). These new developments have significantly enhanced our ability to identify and estimate variance components in choice data, and also allow new and powerful tests to be made of differences in variability associated with and/or due to a variety of sources.
Thus, we think that it is timely and important to move towards a theory of scale in choice that can assist in anticipating and explaining how, when and under what circumstances one should expect scale to vary and the role that scale plays in choices observed under different ways of observing and collecting data, task complexity in experiments, differences in people, such as age-related choice consistency differences and/or differences associated with expertise or experience, and the like. The workshop will review what is known about scale, and will try to develop a conceptual framework(s) to understand and anticipate the behavior of scale in real and hypothetical markets, and propose a future research agenda.
14. Managing the Hedonic Consequences of Choice
Workshop Chairs: Karsten Hansen, Tom Meyvis and Leif D. NelsonConsumer choices have hedonic consequences, and for many choices, those consequences are the focal consideration. When choosing a movie at the theater, consumers will usually choose the movie they think they will enjoy the most. Other choices have a more utilitarian focus, but nevertheless have hedonic consequences. When choosing a physical therapist, people will probably be more concerned about qualifications and efficacy than about maximizing the pleasantness of the experience. Yet, even in that situation, consumers’ are aware that their decision will also have hedonic consequences—and this knowledge may influence their decision.
The focus of this proposed session is to explore consumers’ intuitions about the hedonic consequences of their choices—and discuss the extent to which these intuitions map on to actual consumer experiences. We are particularly interested in the hedonic consequences for the structuring of experiences over time. For instance, how is consumers’ enjoyment influenced by the order of experiences (Do you start with the most painful exercise or end with it?), the continuity of the experience (Do you interrupt the exercise or do you proceed without a break? How about interrupting a movie?), the frequency of the experience (How often should you go to the movie theater to experience maximum enjoyment?), and the variability of the experience (When is it more enjoyable to watch several episodes of the same TV show versus single episodes of multiple TV shows?). Whereas some of these issues have received attention in previous research—in particular research on preference for variety and improving sequences—several of these factors have been hardly addressed in the literature and almost never in combination with one another. Furthermore, even research on variety seeking and preference for improving sequences has mostly focused on consumers’ preferences and intuitions about the hedonic consequences of their choices, without systematically testing when these intuitions map on to actual experiences and when they do not. Part of the session time will be spent on discussing how hedonic consequences of choice might crystallize itself in real market consumer choices. Is the hedonic element in choice strong enough that we might find it in consumer panel choice data? How should researchers go about testing for the presence of hedonic aspects of choice outside the lab?
The end objective of this session would be to provide a general understanding of the way in which the structuring and sequencing of experiences influence consumers’ enjoyment of these experiences. By documenting consumers’ intuitions, we could provide recommendations about how to package experiences to maximize consumer appeal. By documenting the actual effect of structure on consumer enjoyment, we could provide recommendations about the ideal structuring of experiences to maximize consumer well-being. Finally, by comparing consumers’ intuitions to these actual effects, we could identify the situations in which consumers make suboptimal hedonic choices; and suggest interventions to facilitate learning or to “nudge” them towards more optimal decisions.
We have assembled a set of participants who bring a variety of specialized skills to a consideration of these problems. The participants have formal training in psychology, economics, and both behavioral and quantitative aspects of marketing. In combination, we think these researchers will be able to carry the problem from the basic theoretical concerns to some identification of market consequences.
15. Decision and Consumer Neuroscience
Workshop Chairs: Richard Gonzalez, Baba Shiv and Carolyn YoonConsumer and decision neuroscience has added a set of paradigms, techniques, and dependent variables to the standard study of consumer choice and decision making. There is not much consensus in the field as to how to use these new developments to address cutting-edge research questions about consumer choice and decision making. Ideally, we want these techniques to complement standard behavioral, theoretical, and modeling analyses that characterize the traditional study of consumer behavior. These new approaches should add value to standard debates and should provide direction for future debates and research questions. The workshop includes participants who are experts at many of these relevant components, including decision theory, analytic modeling, behavioral research, neuroscience, imaging techniques and various other techniques including study of lesion patients, genetics, endocrinology, and psychophysiological techniques. The workshop provides a unique opportunity to bring together top researchers in the field of decision neuroscience and consumer neuroscience for lively and meaningful exchange of ideas and agenda-setting. The timing of this workshop is very important. The general field of consumer choice and decision making is currently dabbling in various techniques of neuroscience. We need the top contributors in the general field of decision neuroscience to come together, exchange ideas, and help define the future role of neuroscience in consumer decision making.
Here is a subset of questions that will characterize the decision and consumer neuroscience workshop.
1. What is the status of the debate of the “common utility currency” in the brain? Behavioral researchers have resisted giving explanatory status to utility (i.e., the revealed preferences approach), yet new brain data appears to be challenging this strong position.
2. What are new directions for choice models as we add biological data into the decision and consumer domain (not only fMRI but EEG, MEG, TMS, psychophysiological measures, hormone assays, genetics, etc)?
3. What challenges do choice paradigms present in the fMRI framework? That is, can choice studies push the boundaries of what is possible with fMRI, which may lead to interesting synergistic collaborations with other fields such as biomedical engineering to find new approaches to imaging that are optimized for choice studies?
4. The terms “decision making” and “choice” mean different things in different fields. BDT researchers have had a limited view of those terms and now that neuroimagers examine choice the domain of those terms has expanded to animals, perceptual choices, hunger and food consumption, etc. Is it time for new conceptualizations of choice and decision making that are broader? Do we need to develop specialized choice models or is a model that highlights a single choice process sufficient at representing human decision behavior?
16. Theoretical modeling of choices in strategic situations
Workshop Chairs: Raphael Thomadsen and Robert ZeithammerThe goal of this session is to evaluate the role of analytical work in our understanding of consumer and firm choice. As consumer and business environments become more inter-connected – due to easier communication, lower search costs, and more efficient distribution networks – it is increasingly important to look at the world through a lens of strategic analysis that captures competition and/or coordination motives. Analytical work under the paradigm of game theory holds great promise in both generating internally consistent intuition and making empirically testable predictions for our new interconnected world. Despite this promise, the fields of Marketing and Economics continue to struggle in defining the role of analytics in relation to empirical observation and intuition. We intend to explore and clearly outline the role of analytics, hopefully charting the future directions of theory research on choice.
To investigate how theory can increase our understanding of strategic choice and its managerial implications (such as pricing, entry, advertising, etc.), we will consider specific research questions based on the interests of our session participants, all of whom are research-active experts in Marketing and Economics. We intend to focus the session on two broad types of questions.
The first set of questions will focus on the role of repeated consumption on pricing and price-discrimination in competitive contexts. When the repeated consumption occurs over time, such a situation involves competitive dynamics – an area of keen recent interest in both the empirical and theoretical literature. Some dynamics questions we would like to answer are: What is the role of finite vs. infinite horizon models in adequately capturing the real-world? How does word of mouth and/or behavioral-based pricing change our understanding of competitive dynamics? In markets where consumers exhibit a taste for a variety, do we adequately understand how variety-seeking alters our basic intuition about the role of competition and differentiation? The research of some of our participants demonstrates that the “conventional” answers to these questions are not always correct. For example, it turns out that in the presence of variety-seeking behavior, markets can be more competitive with moderate amounts of differentiation than with low amounts of differentiation.
The second set of research questions we want to focus on examines the structure of the market as a whole, i.e. the firms’ choices to enter, interact, advertise, form networks, etc. One such question is whether more products in a market necessarily reduce profits of an entrant (ceteris paribus, this is a key identifying assumption in much recent empirical work). A recent analysis of basic discrete-choice theory using commonly-used modes suggests that this is not actually an innocent assumption – profits of all firms can increase when one firm introduces an additional product. So what should be the prediction one takes to data about competitive entry? Another question asks how direct-to-consumer advertising strategies for prescription drugs are affected by the multi-agent interaction we see in the health care industry, which has a complicated web of moral hazards and informational asymmetries which can alter the effectiveness of different strategies. Finally, we want to discuss circumstances where competition can actually hurt consumers by reducing quality and variety offered in the market.
Ultimately, increased understanding of these key theoretical insights benefits all researchers studying questions of choices. By understanding how the institutional factors – including the extent of competition or the degree of repetition of choices – enveloping choice decisions affect market outcomes, choice researchers are more likely to offer better advice to firms and policy makers. Further, because our theoretical insights affect empirical modeling choices and identification strategies implemented today, our analysis will lead to improved empirical analysis of choice, as well.
