Reference points, anchors, norms,...
ORGANIZATIONAL BEHAVIOR AND HUMAN DECISION PROCESSES 51, 296-312 (1992) Reference Points, Anchors, Norms, and Mixed Feelings DANIELKAHNEMAN University of California, Berkeley The article reviews theoretical analyses and some recent evidence on the role of reference points and anchors in individual choice, and considers pos- sible implications of these analyses for a treatment of negotiation. Reference points are characterized by the abrupt changes in the valuation of gains and losses and of acceptable or reprehensible behavior. Norm theory is applied to the treatment of mixed feelings about outcomes to which multiple refer- ence points are relevant. Anchoring effects in judgment are analyzed in terms of a memory process that assigns normality to possible values. 8 1992 Academic Press. Inc. The notion of reference points is frequently mentioned in the negotia- tion literature, and it plays an important role in a recent text on that subject (Neale & Bazerman, 1991). Reference points are important be- cause other outcomes are compared to them, and are coded and evaluated in terms of this comparison. In the simple case of a continuous outcome variable with monotonically increasing value (e.g., salary), the reference point separates the domain into regions of desirable outcomes (gains) and undesirable ones (losses). However, most decisions and almost all nego- tiations involve multiple reference levels for any valued dimension. For example, Neale and Bazerman (1991) mention five reference points that might affect how the union team will evaluate a wage offer: last year���s wage management���s initial offer the union���s estimate of management���s reservation point the union���s reservation point the union���s publicly an- nounced bargaining position. The valuation of a modest offer by manage- ment could range from a substantial gain to a substantial loss depending on which of these reference values is adopted, and the response to the offer might be expected to vary accordingly. It is a reasonable surmise that some of the messages that negotiators exchange are attempts by each side to communicate its reference point and to affect the reference point of the other side. One aim of this article is to review some experimental and theoretical work that may shed light on the determinants of reference points and on their functions in negotiations. This article draws extensively on collaborative work with Amos Tversky, Jack Knetsch, Richard Thaler, Dale Miller, and George Loewenstein. I am grateful for comments from Maggie Neale. Address correspondence and reprint requests to Daniel Kahneman, Depart- ment of Psychology, University of California, Berkeley, Berkeley, CA 94720. 296 0749-5978192 $3.00 Copyright 0 1992 by Academic Press, Inc. AI1 ri&s of reproduction in any form reserved.
REFERENCE POINTS 297 The concept of anchoring is also frequently mentioned in discussions of negotiation: negotiators are said to be anchored on a position (conveying stubborn adherence to it), unacceptable offers are introduced as anchors, and so on. Another aim of this article is to elucidate the function of anchors and to distinguish them from reference points. The treatment of anchoring is embedded in a cognitive analysis of the norms that represent the internal variability of categories, the uncertainty of beliefs, and the ambivalence of affective responses (Kahneman & Miller, 1986). The first section discusses the role of reference points in the evaluation of outcomes, focusing on the asymmetric treatment of gains and losses and on the aversion to concessions. The effect of reference transactions on judgment of fairness is discussed next. The third section focuses on the ambivalent valuation of the same outcome from several reference points the concept of norm is introduced as a representation of the relative frequencies of the alternative valuations. The final section elaborates on the function of norms and discusses the process of anchoring. THE FUNCTIONS OF REFERENCE POINTS The location of the reference point affects the coding of outcomes as gains and losses. This coding, in turn, affects preferences because of characteristic differences in the evaluation of positive and negative out- comes. Two differences between the psychophysics of gains and losses were emphasized in prospect theory (Kahneman & Tversky, 1979). (i) The value function is S-shaped, concave in the domain of gains and con- vex in the domain of losses this shape favors risk aversion for gains and risk-seeking for losses. (ii) The value function is loss averse, steeper in the domain of losses than in the domain of gains. As we see next, the two characteristics of the value function suggest different psychological pro- cesses that could cause negotiations to fail. In both their experimental work and their excellent text, Neale and Bazerman (1991) have emphasized the importance of the willingness to accept the risk of impasse. The argument is simply that the probability of negotiation failure increases when the negotiators are both risk-prone. This prediction has been tested by manipulating reference points to in- duce different groups of subjects to frame the same negotiation problem in contrasting ways: some subjects view it as a problem of allocating gains, others as an allocation of losses. The negative framing is expected to induce greater risk-seeking. In accord with this hypothesis, negotiators bargaining over losses tend to make fewer concessions, find fewer inte- grative solutions, and more often fail to reach agreement than negotiators bargaining over gains (Bazerman, Magliozzi, & Neale, 1985 Neale 8z Bazerman, 1985 Neale & Northcraft, 1986 Neale, Huber, & Northcraft,
298 DANIEL KAHNEMAN 1987). The interpretation of this result in terms of risk attitudes is plau- sible, but direct evidence of the contribution of risk attitudes is lacking. For example, would the effect of negative framing vanish, as the risk hypothesis entails, if negotiators were faced with a final offer that they could only accept or reject? As will be seen, there is another reason to expect agreement to be particularly difftcult to reach when the issue is the allocation of losses: concessions that increase one���s losses are much more painful than con- cessions that forego gains. Of course, the two mechanisms are not incom- patible, and both could be at work. Loss Aversion Loss aversion refers to the discrepancy between the valuations of gains and losses, which is captured by a value function that is steeper in the negative than in the positive domain. In the context of risky decisions, loss aversion entails a strong reluctance to accept gambles on the toss of a coin, unless the payoffs are extremely favorable: most subjects will only accept such a gamble if the gain/loss ratio exceeds 2:l. In the context of riskless choice the concept of loss aversion has been formalized in terms of the effects of a shifting reference point on indifference curves (Tversky & Kahneman, 1991 Kahneman, Knetsch, & Thaler, 1991). The general principle is straightforward: when an option is compared to the reference point, the comparison is coded in terms of the advantages and disadvan- tages of that option. A particularly important case arises when the refer- ence point is the status quo, and when the retention of the status quo is an option. Because the disadvantages of any alternative to the status quo are weighted more heavily than its advantages, a powerful bias in favor of the status quo exists (Samuelson & Zeckhauser, 1988). The location of the reference point also affects the evaluation of differ- ences between other pairs of options. Differences between disadvantages will generally have greater weight than corresponding differences be- tween advantages, because disadvantages are evaluated on a steeper limb of the value function. For example, the difference between salary offers of $40,000 and $45,000 will be viewed as a difference between two gains by someone whose current income is now $35,000, and as a difference between losses if current income is $50,000. The psychological difference between the alternatives is likely to be greater in the latter case, reflecting the steep slope of the value function in the domain of losses. Acceptance of the lower salary will be experienced as an increased loss if the refer- ence point is high, as a foregone gain if it is low it will be more painful in the former case. The following classroom experiment has been interpreted as an illus- tration of loss aversion in a riskless context (Kahneman, Knetsch, &