WORK IN PROGRESS — NOT FOR GENERAL CIRCULATION Animal Spirits: Affective and Deliberative Processes in Economic Behavior George Loewenstein Ted O’Donoghue Department of Social and Decision Sciences Department of Economics Carnegie Mellon University Cornell University Draft: 3/30/2004 Abstract The economic conception of human behavior assumes that a person has a single set of well-defined goals, and that the person’s behavior is chosen to best achieve those goals. We develop a model in which a person’s behavior is the outcome of an interaction between two systems: a deliberative system that assesses options with a broader, goal-based perspective, and an affective system that encompasses emotions and motivational drives. Our model provides a microfoundation for many departures from full rationality discussed in the behavioral-economics literature, and captures the familiar feeling of being "of two minds.” By focusing on factors that moderate the relative influence of the two systems, our model generates a variety of novel testable predictions. Acknowledgements: We thank David Laibson and participants at the 2004 ASSA meetings in San Diego for useful comments, and we thank Christoph Vanberg for valuable research assistance. For financial support, Loewenstein thanks the Integrated Study of the Human Dimensions of Global Change at Carnegie Mellon University (NSF grant # SBR-9521914), and O’Donoghue thanks the National Science Foundation (grant SES-0214043).1At the center of the brain lies a cluster of strange-shaped modules that together are known as the limbic system. This is the powerhouse of the brain – generator of the appetites, urges, emotions and moods that drive our behavior. Our conscious thoughts are mere moderators of the biologically necessary forces that emerge from this unconscious underworld; where thought conflicts with emotion, the latter is designed by the neural circuitry in our brains to win. Rita Carter Mapping the Mind. (1999:54) I. Introduction The standard economic conception of human behavior assumes that a person has a single set of well-defined goals, and that the person’s behavior is chosen to best achieve those goals (or at least that it is “as if” the person’s behavior is determined in this way). Such an approach is intuitive, tractable, and has shed light on a wide range of economic behaviors, from mundane activities such as consumer decision making to more exotic behaviors such as those associated with drug addiction. While this approach seems to suffice in many situations, we show that a more nuanced, “two-system” perspective that takes account of interacting processes within the human brain permits us to better understand a wide range of economic phenomena that are difficult to reconcile with the standard unitary preference approach. Specifically, we develop a two-system model in which a person’s behavior is the outcome of an interaction between a deliberative system that assesses options with a broader, goal-based perspective (roughly along the lines of the standard economic conception), and an affective system that encompasses emotions, such as anger and fear, and motivational drives such as those involving hunger and sex. Our model provides a microfoundation for many of the departures from full rationality discussed in the recent behavioral-economics literature. At the same time, it captures the familiar feeling of “being of two minds” — of simultaneously wishing one were behaving one way while actually behaving in a different way. By focusing on factors that influence the relative influence of the two systems, our model not only sheds light on well-known phenomena such as addiction, but also generates a number of novel, testable, predictions. In Section II, we motivate our particular two-system account of behavior, drawing on research findings and insights from diverse fields such as philosophy and psychology. We also review some of the large body of evidence from neuroscience — on both humans and animals — which suggests that behavior is the outcome of multiple neural processes operating in parallel.2Our two-system approach is motivated by, and roughly similar to, many of distinctions made in these literatures. We then discuss some of the insights that emerge from these literatures concerning interactions between the two systems. We describe how the two systems often respond differently to the same stimuli or cues, leading to conflicts between them. We also describe how many of the most important divergences between affect and deliberation stem from the differential impact on the two systems of what we term cue-proximity. Specifically, the affective system is much more responsive to cues that are proximate, on many dimensions, than is the deliberative system. Finally, we discuss evidence on the concept of willpower, which is key to our model. We view willpower as a resource expended by the deliberative system to exert influence over the affective system. Specifically, there are dynamic effects in the sense that recent use of willpower depletes current willpower and hence increases the cost of exercising willpower, and there are more static effects such as the impact of stress or 'cognitive load'. In Section III, we develop our model of interactions between the affective and deliberative systems and discuss its connections to and differences from some recent, closely related approaches in the economics literature — notably, the planner-doer approach of Shefrin and Thaler (198x), the hot-mode/cold-mode approach of Bernheim and Rangel (2002, 2003), and the effects of cues and visceral states of Loewenstein (1996) and Laibson (2000). In Sections IV, V, and VI, we apply our model to time preferences, risk preferences, and social preferences. In Section IV, we show how our framework predicts discounting without assuming that the deliberative system has any time preference, and also how it provides a nice reinterpretation of the differences between two recent models of self-control — between hyperbolic discounting as in Laibson (1997) and O’Donoghue and Rabin (1999) and temptation utility as in Gul and Pesendorfer (2000). Finally, we describe how our model makes predictions in terms of when people should show more discounting, “steeper” hyperbolic discounting, or stronger temptation (dis)utility. II. Motivation for a Two-System Approach