The effect of regulatory focus on the shape of probability-weighting function: Evidence from a cross-modality matching method

Prospect theory (Kahneman & Tversky, 1979, 1984) suggests that when people are presented with objective probabilities they (a) underweight high probabilities (e.g., behave as if 99% likelihood of an event is lower than 99%), (b) overweight low probabilities, and (c) are relatively insensitive to differences among moderate probabilities. We hypothesized that these biases will be found under prevention focus (Higgins, 1997), which can be triggered by security needs, and monetary considerations; but reversed under promotion focus (Higgins, 1997), which can be triggered by self-actualization needs. To test the hypothesis, we developed a crossmodality matching task that allows tapping probability transformations independently from the value of an event. In two studies, participants (N 1⁄4 116 and N 1⁄4 156) drew portions of circles that represented their transformations of 13 different stated probabilities regarding three scenarios (either promotion or prevention). Results in the prevention condition were consistent with prospect theory—providing validity for the cross-modality matching method. Results in the promotion condition indicated both a general elevation (overweighting), which was most evident for moderate and moderate-high probabilities, and minor underweighting for probabilities larger than .80. In the second study, we also assessed chronic-regulatory focus which yielded effects similar to the manipulated-regulatory focus. In both studies, some individuals in the promotion focus groups yielded probability weighting functions with a curvature opposite the predictions of prospect theory; and within each experimental condition there were additional significant differences in the transformation yielded by the putatively similar three scenarios. The results indicate that our crossmodality matching method is very sensitive to context effects and hint at the possibility of applying similar cross-modality matching methods to explore other decision-making processes such as value functions (Kahneman & Tversky, 1979). 2004 Elsevier Inc. All rights reserved. The authors thank Amnon Rapoport for encouragement to start this project and Amnon Rapoport and Yuval Rottenstreich for comments on an earlier draft of this paper. This research was supported by a grant from the Reccanti Fund at the School of Business Administration and an ARI contract # DASW01-04-K-0001 to the first author. The view, opinions, and/or findings contained in this paper are those of the authors and should not be construed as an official Department of the Army position, policy, or decision. An earlier version of this paper was presented at the Israeli Conference on Cognitive Psychology and on Judgment and Decision Making at the Hebrew University, Jerusalem, July 2002. * Corresponding author. Fax: +972-2-5881341. E-mail address: [email protected] (A.N. Kluger). 0749-5978/$ see front matter 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.obhdp.2004.05.003 Introduction Prospect theory (Kahneman & Tversky, 1979, 1984) suggests that people psychologically transform stated (objective) probabilities ðpÞ into weighted probabilities ðwðpÞÞ in a non-linear fashion. Specifically, people overweight very low probabilities ðwðpÞ > pÞ and underweight very high probabilities (p > wðpÞ). The transformed probabilities are not observed directly, but can be inferred from choice decisions, and hence are called decision weights. For example, most respondents (72%) prefer a gamble of receiving 5000 pounds with a probability of .001 rather than receiving 5 pounds for sure (Kahneman & Tversky, 1979; Problem 14). If 2 A.N. Kluger et al. / Organizational Behavior and Human Decision Processes xxx (2004) xxx–xxx ARTICLE IN PRESS people behaved according to normative prescriptions they would evaluate each prospect by its value and expectancy (by multiplying the value with its probability). If people would follow this prescription they should show indifference between the prospects of 5 1.00 and 5000 .001 because they carry the same expected value (5). But the modal choice suggests that people have a clear preference for the gamble. If people treated monetary gains in a linear fashion, one could infer that wð:001Þ > :001 because 5/50001⁄4 .001. However, it is accepted that psychological value (v) of money is also psychologically transformed with a concave function such that 1000 vð5Þ > vð5000Þ. Thus, the more concave is the value function, the stronger is the bias that can be inferred in wð:001Þ. Similar phenomena can be demonstrated for large values of p. To explain this effect, two constructs are evoked: reference points and diminishing sensitivity (e.g., Tversky & Fox, 1995). The reference points on the weighingprobability function are certainty and impossibility. As one moves away from the reference points one loses sensitivity to change in probability. Thus, people are most sensitive to changes in probabilities nearest to the endpoints. The predicted bias in wðpÞ relative to p received substantial support (Camerer & Ho, 1994; Gonzalez & Wu, 1999; Tversky & Fox, 1995; Tversky & Kahneman, 1992; Wu & Gonzalez, 1996). However, the magnitude of the effect was predicted to vary slightly for probabilities framed as successes versus failures (Tversky & Fox, 1995). Moreover, the transformation bias was shown to be affected by the degree of the context emotionality (Rottenstreich & Hsee, 2001). We further suggest that context with high emotionality (both regarding success and failure) will influence wðpÞ differently as a function of regulatory focus (Higgins, 1997). The typical research of wðpÞ involves probabilities associated with monetary gains and losses. Contexts that involve money are likely to evoke either power or security needs (Ronen, 1994, Chapter 5). Security needs, and to some degree power needs, are likely to evoke prevention focus where the dominant goal is to avoid pain (Higgins, 1997, 1998). In contrast, needs such as self-actualization are likely to evoke promotion focus where the dominant goal is to approach pleasure (Higgins, 1997, 1998). Given that self-actualization needs are typically absent from decision-making research, the key goal of this paper is to develop and test hypotheses regarding the varying effects of underlying motivations on probability transformation. Testing the effect of needs on the probability transformations with the standard method is problematic because the standard method relies on inferring wðpÞ from choices regarding quantifiable prospects. To overcome the difficulty in assessing wðpÞ for non-monetary prospects, the second goal of this paper is to develop a cross-modality matching method to estimate wðpÞ directly. To derive the hypotheses, we first review the pertinent motivational literature. Higgins’s self-regulation theory Higgins (1997, 1998) proposed that people have two basic self-regulation systems. One system regulates the achievement of rewards and focuses people on a promotion goal. In contrast, the other system regulates the avoidance of punishment and focuses people on a prevention goal. The focus of self-regulation—prevention or promotion—is determined by at least three antecedents (Higgins, 1998): prevention focus is activated by security needs, strong obligations, and the framing of the situations in ‘‘loss versus non-loss’’ terms; promotion focus can be triggered by growth and development needs (Brockner & Higgins, 2001; labeled nurturance needs in prior works), strong ideals and the framing of the situations in ‘‘gain versus non-gain’’ terms. Each focus has different consequences for perception, for decision-making, and for emotions (Higgins, 1997, 1998). Under prevention focus people are more likely to be sensitive to the presence or absence of punishment, use avoidance strategies, monitor errors of commission, and experience emotions ranging from agitation to quiescence. In contrast, under promotion focus, people are more likely to be sensitive to the presence or absence of rewards, use approach strategies, monitor errors of omission, and experience emotions ranging from elation to dejection. Two feature of Higgins s theory are relevant here. First, one important feature is the prediction that security needs evoke prevention focus and hence sensitivity to punishments, and that growth and development needs evoke promotion focus and hence sensitivity to rewards. This motivational dichotomy is well recognized in the decision-making literature in the form of security versus aspiration or potential-mindedness (Lopes, 1987, 1995). Second, the regulatory foci are thought of as rich syndromes that differ from each other on multiple variables. The activation of a focus entails changes in perceptual process such as eagerness versus vigilance in signal detection, physiological processes such as the flexion versus tension of arm pressure (Forster, Higgins, & Idson, 1998), affective processes (Brockner & Higgins, 2001) and cognitive processes. Thus, rather than searching for the core variable that either induce or is influenced by these foci, we assume that multiple variables operate in differentiating these foci and that by considering the nature of the needs inducing these foci, we would better understand the total effect of the sets of variables induced by each focus. Accordingly, we next review the relevant need literature. Fig. 1. A schematic map of the dimensions that underlie Ronen s (1994) and Schwartz s (1992) empirical findings regarding the structure of needs and values. A.N. Kluger et al. / Organizational Behavior and Human Decision Processes xxx (2004) xxx–xxx 3 ARTICLE IN PRESS Promotion and prevention in needs and values Higgins proposed that the prevention focus system relates to duties and obligations (oughts) and satisfies security needs, whereas the promotion focus system relates to accomplishments and aspirations (ideals) and satisfies nurturance needs (Higgins, 1997) and growth needs (Brockner & Higgins, 2001). Higgins specifically refers to Maslow s need hierarchy theory (Maslow, 1965) and argues that Maslow s security needs are linked with prevention focus and that Maslow s self-actualization needs are linked with growth needs and with promotion focus. Maslow s (1965) theory has two elements: taxonomy of needs and an argument about a hierarchy among the needs, where only the taxonomy is relevant to our investigation. The putative basic needs are physiological, safety (security), belongingness (love), esteem, and selfactualization needs. Early reviews rejected the five-need taxonomy, but recognized a possibility of two classes of needs: (Wahba & Bridwell, 1976): deficiency needs (safety, love, and respect from others) versus growth needs (self-respect, achievement and self-actualization). An alternative two-level classification and hierarchy— Wahba and Bridwell s (1976)—suggests maintenance needs (physiological and safety) versus growth needs (belongingness, esteem, and self-actualization). It is also interesting to note that a two-class dichotomy similar to the prevention-promotion dichotomy also appears in Maslow s early writing (Wahba & Bridwell, 1976). More recent lines of research on motivation and values are more supportive of Maslow s taxonomy. Researchers who applied smallest space analysis (a type of non-metric multidimensional scaling) to 14 work-related needs found four groups of needs: Physical and Security, Self Actualization, Social and Esteem (Ronen, 1979; Ronen & Kraut, 1980; Ronen, Kraut, Lingoes, & Aranya, 1979; Ronen & Shenkar, 1986; Shenkar & Ronen, 1987). In review, Ronen (1994) indicated that work needs in 15 different cultures yield similar twodimensional maps of needs containing four distinct regions conforming to the a priori classification. In these analyses, security items were clustered in the same region and were most separated from self-actualization items that were clustered in a different region. The extremes of the smallest space analysis maps found by Ronen are shown schematically in Fig. 1. Fig. 1 also shows other constructs that are likely to correspond to Maslow s constructs, consistent with Ronen (1994) who recognized the similarity between various need theories. A structure similar to Fig. 1 is proposed in value theory (Schwartz, 1992) that was developed and tested independently from need research. Values are considered to be desirable, trans-situational goals reflecting both biological needs and conformity to social pressures that serve as guiding principles in one s life (Schwartz, Lehmann, & Roccas, 1999). The theory (Schwartz, 1992) postulates ten values organized as a circle: Power, Security, Conformity, Tradition, Benevolence, Universalism, Self-Direction, Stimulation, Hedonism, and Achievement. Adjacent values tend to be compatible whereas values in opposing positions tend to be in conflict. For example, ‘‘the pursuit of novelty and change (stimulation) is likely to undermine preservation of time-honored customs (tradition)’’ (Schwartz et al., 1999, p. 110). The value circle is organized by two sets of higher-order values that are in conflict: change (stimulation and self-direction) versus conservation (security, conformity, and tradition) and self-enhancement (power, achievement, and hedonism) versus self-transcendence (universalism and benevolence). Empirical maps of 57 values obtained with smallest space analysis across different national cultures (for a recent report that contains already 57 cultures see Sagiv & Schwartz, 2000) uncovered a circumplex structure. In this structure, security value conflicts with the value of self-direction, defined as ‘‘independent thought and action-choosing, creating, exploring’’ (Schwartz et al., 1999). This structure is very similar to a circumplex found in the analysis of needs (Ronen, 1994) where security needs are most different from self-actualization need (Ronen, 1994). In summary, two different approaches to the content of human motivations—an empirical tests of Maslow s theory (Ronen, 1994) and values theory (Schwartz, 1992)—have yielded a similar two-dimensional space representation. One dimension or axis ranges from security needs, and security values, to self-actualization, and self-direction. We suggest that this axis in the structure of needs and values reflects Higgins s differentiation between security and development needs, which are considered in his theory as precursors of prevention and promotion focus. Thus, activating 4 A.N. Kluger et al. / Organizational Behavior and Human Decision Processes xxx (2004) xxx–xxx ARTICLE IN PRESS security needs should yield behaviors that characterize the prevention system, whereas activating self-actualization needs should yield behaviors that characterize the promotion system. In contrast, activating needs that can be projected to the middle of this axis—belongingness (affiliation) and esteem (power) needs—can activate either promotion or prevention focus. Hence, needs and values that lie in the middle of this axis are likely to yield mixed effects. Importantly, if need for money (salary) is projected on the two-dimensional space into an imaginary axis ranging from security to self-actualization it will be found on a point in between the extremes of security on the one hand and self-actualization on the other hand (see Ronen, 1979, 1994; Ronen & Kraut, 1980; Ronen et al., 1979). Similarly, the value of power (containing items such as wealth) is also located on a point between values representing change such as self-direction and values representing preservation such as security with a greater proximity to security (Schwartz, 1992). Yet, on the basis of empirical data (N > 10; 000), power and security are generally more correlated than either of them is with self-direction; and self-direction is less positively correlated with security than with power (Schwartz & Boehnke, 2004). Thus, choice problems involving money are likely to induce a mix focus, where prevention focus is more likely to dominate monetary decisions for most people under most situations. This suggests that issues involving money cannot create a pure promotion focus. Indeed, several researchers in the domain of risk-taking proposed that decision makers oscillate between fear (prevention) and hope (promotion) and thus their decisions should be modeled as a mixture of these motives (Lopes, 1987). Interestingly, prospect theory was demonstrated ‘‘mainly with monetary outcomes’’ but was argued to be ‘‘readily applicable to choices involving other attributes, e.g., quality of life’’ (Kahneman & Tversky, 1979, p. 288). Yet, Tversky and Kahneman (1991) noted that ‘‘loss aversion appears to be more pronounced for safety than for money . . . and more pronounced for income than for leisure’’ (p. 1054). Similarly, scenarios regarding security (life and death) produce stronger framing effects than scenarios regarding money and property (Wang, 1996). This pattern seems like traversing between security values to hedonism and self direction values on Schwartz s map which might reflect decisions largely driven by prevention focus to decisions largely driven by promotion focus. This pattern suggests that pure promotion focus may lead to biases not typically explored in the decision-making literature. Before we develop our hypotheses, it is important to note that context effects are well recognized in the literature (for a comprehensive review see Rettinger & Hastie, 2001). Context may differ not only in their motivational properties, but also in the ease of constructing a narrative account of scenarios. Moreover, some motivational biases are predictably orthogonal to those considered here. For example, choices involving moral dilemmas are likely to invoke unique decision processes (Rettinger & Hastie, 2001). Yet, moral issues are likely to be orthogonal to regulatory focus because some moral issues are more about prevention (the obligation to help a needy family member) where others are more about promotion of ideals (teaching writing and reading in other nations or helping ‘‘doctors without borders’’). Therefore, our hypotheses can be useful for organizing one set of contextual features that is likely to have an important contribution to the decision-making process, but cannot account for all the known contextual effects on decision making.