Reference-Dependent Consumption PlansI. The ModelII. Information PreferencesIII. Monetary PreferencesIV. Wealth and Consumption in Intertemporal ChoiceV. ConclusionAppendix: ProofsREFERENCES909American Economic Review 2009, 99:3, 909–936http://www.aeaweb.org/articles.php?doi=10.1257/aer.99.3.909In this paper, we develop a general dynamic model of reference-dependent utility. Building on ideas and models in Astrid Matthey (2005), Christopher K. Hsee and Clair I. Tsai (2008), Miles Kimball and Robert Willis (2006), and Ko˝szegi and Rabin (2006, 2007), we assume that utility depends on recent changes in rational beliefs about present and future consumption, and bad news is more painful than good news is pleasant.1 We derive implications of our model for preferences over receiving information about an exogenous future event, monetary risk prefer-ences, and intertemporal consumption decisions. If news about more imminent consumption is felt more heavily than news about distant consumption, a person prefers to receive the same information sooner rather than later, increases consumption immediately in response to good surprises regarding wealth but delays cuts following bad surprises, and—since surprising herself with extra immediate consumption is pleasant—may overconsume early in life relative to the optimal committed plan. To reduce the impact of losses she may suffer, the decision maker pre-fers to receive bits of information together rather than apart, and—to lower the marginal utility associated with any surprise losses—prepares for future uncertainty by increasing savings.We present the basic framework in Section I. In each period t ∈ {1, … , T}, the decision maker consumes a K-dimensional consumption bundle, ct. Overall instantaneous utility in period t is the sum of reference-independent “consumption utility” that derives purely from ct, and gain-loss utility that derives from recent changes in beliefs about consumption in each dimension in each 1 The model in this paper is closely related to contemporaneous work by Matthey (2006). Building on her earlier intuitions in Matthey (2005), she develops a formal model of preferences that is similar to our formulation. But her solution concepts embed different assumptions from ours about how people form their beliefs, and she develops a dif-ferent set of results.Reference-Dependent Consumption PlansBy Botond KO˝szegi and Matthew Rabin*We develop a rational dynamic model in which people are loss averse over changes in beliefs about present and future consumption. Because changes in wealth are news about future consumption, preferences over money are reference-dependent. If news resonates more when about imminent consump-tion than when about future consumption, a decision maker might (to generate pleasant surprises) overconsume early relative to the optimal committed plan, increase immediate consumption following surprise wealth increases, and delay decreasing consumption following surprise losses. Since higher wealth mitigates the effect of bad news, people exhibit an unambiguous first-order precautionary-savings motive. (JEL D14, D81, D83, D91)* Ko˝szegi: Department of Economics, University of California, Berkeley, 508-1 Evans Hall #3880, Berkeley, CA 94720 (e-mail: [email protected]); Rabin: Department of Economics, University of California, Berkeley, 508-1 Evans Hall #3880, Berkeley, CA 94720 (e-mail: [email protected]). This paper updates and replaces earlier ver-sions titled “Dynamic Reference-Dependent Preferences,” first presented and posted in June 2006. We are grateful to Avinash Dixit and Neel Mukherjee, and seminar audiences at the Conference in Honor of Eytan Sheshinski, University of California, Berkeley, CERGE-EI, University of Chicago, Cornell University, University of Edinburgh, Harvard University, London School of Economics, University of Michigan, Princeton University, University of California, San Diego, University of Southern California, and University of Texas, Austin, and especially three anonymous referees for useful comments. We thank the National Science Foundation (grant SES-0648659) for financial support. Kristóf Madarász provided research assistance.JunE 2009910THE AMERICAn ECOnOMIC REVIEWperiod starting with t. The person experiences “contemporaneous” gain-loss utility from any contrast between current consumption and her prior expectations of current consumption, and “prospective” gain-loss utility from changes in her beliefs about future consumption. In all these comparisons, she is loss averse: bad news about consumption is more unpleasant than good news is pleasant. Normalizing the weight γt,t on contemporaneous gain-loss utility to one, in period t the decision maker puts weight γt,τ ≤ 1 on prospective gain-loss utility regarding outcomes in period τ > t. Her goal in period t is to maximize the sum of instantaneous utilities starting in period t.Because preferences depend on the sequence of expectations, our model is complete only when combined with a theory of how expectations are formed. We follow much previous work on beliefs-based preferences and assume that beliefs must be rationally based on credible plans for state-contingent behavior.2 In particular, an implemented plan must be a “preferred personal equilibrium”: in each period it maximizes expected reference-dependent utility given the expec-tations generated by the plan, with the constraint that continuation plans must be consistent with a similar maximization in the future. In the Web Appendix (available at http://www.aeaweb.org/articles.php?doi=10.1257/aer.99.3.353) we propose a framework for thinking about the dynamic formation of rational beliefs and plans without imposing that initial beliefs about the future are necessarily correct, and explore some circumstances when this more basic rationality assump-tion justifies the use of preferred personal equilibrium.In Section II, we explore some implications that our assumption of loss aversion over changes in beliefs has for informational preferences. On the one hand, when prospective gain-loss utility does not loom as large as contemporaneous gain-loss utility—so that possible bad news is less painful when outcomes are not imminent—the decision maker likes receiving the same information sooner rather than later. On the other hand, consistent