The Journal of Risk and Uncertainty, 30:1; 5–19, 2005 c  2005 Springer Science + Business Media, Inc. Manufactured in The Netherlands. It is Whether You Win or Lose: The Importance of the Overall Probabilities of Winning or Losing in Risky Choice JOHN W. PAYNE ∗ jpayne@mail.duke.edu Fuqua School of Business, Duke University, Durham, NC 27708, USA Abstract Imagine that you own a five-outcome gamble with the following payoffs and probabilities: ($100, .20; $50, .20; $0, .20; −$25, .20; −$50, .20). What happens when the opportunity to improve such a gamble is provided by a manipulation that adds value to one outcome versus another outcome, particularly when the opportunity to add value to one outcome versus another outcome changes the overall probability of a gain or the overall probability of a loss? Such a choice provides a simple test of the expected utility model (EU), original prospect theory (OPT), and cumulative prospect theory (CPT). A study of risky choices involving 375 respondents indicates that respondents were most sensitive to changes in outcome values that either increased the overall probability of a strict gain or decreased the overall probability of a strict loss. These results indicate more support for OPT rather than CPT and EU under various assumptions about the shape of the utility and value and weighting functions. Most importantly, the main difference between the various expectation models of risky choice occurs for outcomes near the reference value. A second study of risky choice involving 151 respondents again demonstrated the sensitivity of subjects to reducing the probability of a strict loss even at the cost of reduced expected value. Consequently, we argue that theories of how people choose among gambles that involve three or more consequences with both gains and losses need to include measures of the overall probabilities of a gain and of a loss. Keywords: decision, risk, preference JEL Classification: D81 The past fifty years have seen much theoretical and empirical work devoted to developing and testing models of risky choice behavior. However, developing a descriptive model of risky decision-making has proven much more difficult than originally assumed. In particular, “the issue of a suitable descriptive theory for gambles with three or more consequences is very much up in the air” (Luce, 2000, p. 286). This is unfortunate, since many important decisions involve risky options with multiple outcomes. This paper provides evidence that any descriptive theory of choice among gambles with multiple outcomes will need to include measures of the overall probability of a gain and the overall probability of a loss. Although the psychological meaningfulness of constructs reflecting the overall probabilities of winning or losing may seem obvious, such constructs are not part of the traditional expected utility model nor of more recent nonlinear expectation ∗ To whom correspondence should be addressed.