Faculty Working Paper 91-0102 The Library of the FEB 1 * |99l University of Ui4fioi» irtana-Champalgr Gain/Loss Asymmetry in Risky Intertemporal Choice Marjorie K. Shelley Department ofAccountancy t Bureau of Economic and Business Research College ofCommerce and Business Administration University of Illinois at L'rbana-Champaign BEBR > FACULTY WORKING PAPER NO. 91-0102 College of Commerce and Business Administration University of Illinois at CIrbana-Champaign January 1991 Gain/Loss Asymmetry in Risky Intertemporal Choice Marjorie K. Shelley Department of Accountancy This paper is taken from my dissertation completed at the University of Texas at Austin. I would like to thank my dissertation committee members, Urton L. Anderson, James S. Dyer, David V. Hinkley. Garry A. Marchant, and Lawrence A. Tomassini for their support and for their helpful comments throughout the project. would also like to thank Don N. Kleinmuntz, George F. Loewenstein and Ira I Solomon for helpful suggestions on earlier drafts of this paper. Finally, would like to thank the > I colloquia participants at The University of Texas at Austin, the University of Illinois at Urbana- Champaign, The Ohio State University, the University of North Carolina, Duke University and the University of Kansas for insightful comments during the early stages ofthe work. The financial support of the Ernst & Whinney (Ernst & Young) Foundation is gratefully acknowledged. » I Gain/LossAsymmetry in Risky Intertemporal Choice 1 Abstract This paper investigates the claim that distant payoffs generate more risky choices than immediate payoffs. Decision makers will make more risky choices if loss discount rates are higher than gain discount rates or if the implicit risk of an option appears greater for loss than gain outcomes. These hypotheses were tested by comparing the loss and gain discount and implicit risk rates inferred from ratings of lotteries formed from a 4x4x4x2 (gain, loss, time, probability) experimental design. The ratings were used to estimate four alternative models of the lottery evaluation process. Implied loss discount rates were higher than gain rates, but the difference diminished substantially once the implicit risk rate had been extracted. Gain and loss implicit risk rates were also different and in a direction that implies greater risk tolerance for decisions with delayed consequences. Conventional discounting models (e.g., the net present value model) imply constant discount rates across gain and loss, as well as over time. A constant rate is enforced when the conventional model is applied explicitly, but decision makers discount for delay and for implicit risk whether or not a formal model is invoked. The results of this study imply that decision makers' natural discounting processes do not conform to conventional theory. The departures from conventional discounting detected in this study can affect any intertemporal decision to which a formal model is not applied, the options selected to submit for formal evaluation, and the estimates made of potential outcomes and likelihoods for use in formal evaluations. (I Digitized by the Internet Archive 2011 with funding from in Urbana-Champaign University of Illinois t http://www.archive.org/details/gainlossasymm102shel Gain/Loss Asymmetry in Risky Intertemporal Choice 2 1. Introduction "The riskiness of racetrack wagers declines as post time approaches. . . . Anxious public speakers are easier to sign up several months in advance than when the date of the speech is close at hand" (Jones & Johnson, 1973, pp. 613-614). Women who want to experience natural childbirth demand in advance that their doctors withhold pain medication to preserve the memory of the birth, but at the onset of labor, they renege and demand relief (Christensen-Szalanski, 1984; Schelling, 1984). Homeowners readily assume home equity loans with immutable five-year "balloon" payments, but regret their decision as the payment date draws near because, even though they expected the payment and knew its magnitude, they apparently failed to appreciate its eventual impact when they weighed against the anticipated benefit. it These are anecdotal illustrations of what appears to be a discrepancy between the way decision makers weight gains relative to losses for immediate versus future outcomes. One explanation for the alleged asymmetry is a shift in the decision maker's reference point that occurs as the outcome approaches. For example, before the onset of labor, a woman's choice is between pain in the future, a negative but short-term outcome, and the experience and memory of natural childbirth, presumably a positive and long-term outcome. Once she is in labor, however, pain is the status quo - the reference point - and she must choose between relief, an immediate short-term positive outcome, and the experience of natural childbirth - two gains. On the other hand, when risk is involved, the intertemporal difference in the cost/benefit balance could be due to decision makers' perceptions of risk; decision makers may exhibit an optimistic tendency to believe they can control the odds or the magnitude of potential future loss (March & Shapira, 1987). This perception would create a discrepancy between subjective discount rates for future negative and positive outcomes. Losses, pain, payments, etc., may lose their disutility faster into the future than gains of comparable magnitude lose their utility. A gain/loss discounting asymmetry has been noted in previous intertemporal choice research, but previous studies have not Gain/Loss Asymmetry in Risky Intertemporal Choice 3 usually addressed risk (Thaler, 1981; Benzion, Rapoport & Yagil, 1989; Loewenstein & Prelec, 1989b). The objective of this study is to estimate a descriptive model of an intertemporal lottery evaluation process and use the resulting parameter estimates to determine whether time preferences are influenced by asymmetric discounting when subject reference points are constant. In what follows, Section 2 gives some background information concerning discounting theory and previous investigations of gain/loss asymmetry and implicit risk. Section 3 introduces four simple descriptive models of the subjective evaluation process for two-outcome risky decision options. The models' components are explained and specific hypotheses relating to predicted values of the estimated model parameters are discussed. Section 4 describes the experimental design and method of data analysis, Section 5 presents the experimental results, and Section 6 discusses the results, their implications, and directions for future research. Background 2. 2.1 Discounting In general, an individual's "subjective rates of time preference are derived from his consumption utility function They are independent of the market rates of interest and . . . his borrowing and lending opportunities" (Henderson & Quandt, 1980, p.327). Since Fisher's (1930) formalization of discounted utility (DU), discount functions have usually been assumed to depend only on time distance, given a particular discount rate. But several departures from Fisher's model, including certain gain/loss asymmetries, appear consistently in recent literature (Thaler, 1981; Benzion et al., 1989; Loewenstein & Thaler, 1989; Loewenstein & Prelec, 1989a and 1989b; Shelley, 1990). Some of these departures were identified by early theorists as causes of inconsistent intertemporal choices and suboptimal planning (e.g., Strotz, 1955 and Thaler, 1981). The departures from Fisher's DU theory noted in the literature tend to fall into the four categories identified by Loewenstein & Prelec (1989a & 1989b) as: (1) A common