Goal-Induced Risk Taking in Strategy Choice

We test whether specific, challenging goals increase risk taking. We propose that goals serve as reference points, creating a region of perceived losses for outcomes below a goal (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992). According to the Prospect Theory value function, decision makers become more risk seeking in the domain of losses. In three experiments we compared a “do your best” condition with a “specific, challenging goal” condition. The goal condition increased risky behavior in a skill task, monetary gambles, and bargaining. The discussion considers additional implications of the reference point perspective as well as the relationship between goal-induced risk taking and innovation. Goal-Induced Risk Taking 3 Goal-Induced Risk Taking in Strategy Choice One of the most widely-documented findings in psychology is that goals improve task performance (Locke & Latham, 1990). In general, having a specific, challenging goal increases effort and persistence compared to a vague intention, such as “doing one’s best.” Although the effectiveness of goals is undisputed, some researchers have been concerned with identifying boundary conditions and unintended consequences of goal setting. For example, researchers have found evidence that difficult goals can harm performance on complex tasks (see Wood, Mento, & Locke, 1987, for a review) and can detract from performance on other tasks or task dimensions for which goals are not set (Bavelas & Lee, 1978; Polzer & Neale, 1995; Rothkopf & Billington, 1979). In the current research, we draw from research on individual decision making to propose that specific, challenging goals have another critical consequence: They make people more willing to take risks. Although previous researchers have speculated on the possibility that difficult goals lead to “excessive” risk taking (Locke & Latham, 1984, p. 160), there has been no systematic analysis or empirical demonstration of this claim. Recently, however, we proposed (Heath, Larrick, & Wu, 1999) a theoretical explanation of goal-setting effects that predicts changes in risk preference. Specifically, we proposed that goals change the value of outcomes according to the principles identified in the Prospect Theory value function (Kahneman & Tversky, 1979; Tversky & Kahneman, 1992). One of the main implications of the value function is that risk preferences change depending on whether decisions involve gains or losses: In choices between a sure gain and a risky gain, most people take the sure gain, but in choices between a sure loss and a risky loss, most people prefer to gamble. We proposed that goals serve as reference points that make people feel as if they are in the domain of losses (e.g., “10 units behind a goal”) rather than the domain of gains (e.g., “20 units ahead of where I started”). By changing the frame of reference, goals change risk preference (March & Shapira, 1992; Payne, Laughhunn, & Crum, 1981). Although the prediction of risk-seeking is a central prediction of the value function Goal-Induced Risk Taking 4 approach, other theories of goals have not systematically considered the mechanisms by which goals affect risk preference. If goals do increase risk seeking, it may have useful theoretical and practical implications. On a theoretical level, it may shed light on some puzzling results in the traditional goal-setting literature. A small body of evidence has shown that goals lead people to make more sizeable and frequent changes in their strategies. Below, we argue that this is a form of riskseeking and that this kind of search for new strategies can be predicted by the same theoretical mechanism we use to predict risk. From the standpoint of organizational practice, risk-taking may be associated with positive outcomes like innovation and creativity (Longswirth, 1991), as well as negative outcomes like reckless behavior (Maremont, 1995). If goals lead people to become more willing to take risks, organizations must manage the process so that they realize the advantages of innovation and creativity while avoiding the potential damage of reckless actions. In the next section, we provide a brief review of our prediction that goals increase risk taking. We then consider findings in the goal-setting literature on strategy selection that are consistent with goal-induced risk taking. We finish by describing three specific experiments that provide a direct test of goal-induced risk taking. Goals, Reference Points, and Risk Preference A common theme runs through literatures concerned with goals: Goals motivate because they provide a comparison for evaluating performance (Lewin, Dembo, Festinger, & Sears, 1944; Locke & Latham, 1990). Specifically, goals transform a somewhat ambiguous stimulus— a performance level such as “27 sales so far this month”—into an outcome that has a clear valence—“better” or “worse” than the goal. Fundamentally, the effect of goals on behavior depends on a cognitive process of comparative evaluation (Locke & Latham, 1990, p. 78). Because comparative judgment is central to goal-setting effects, research on the cognitive psychology of comparative evaluation can enrich our understanding of goal-related behavior. Goal-Induced Risk Taking 5 Recently, we have proposed that a well-known theory of comparative evaluation, Prospect Theory, provides an important link between goals and motivation (Heath et al., 1999). Specifically, we predict that goals systematically transform the valuation of outcomes consistent with Prospect Theory’s S-shaped value function shown in Figure 1. The value function embodies three principles that determine how tangible outcomes x are translated into psychological experience. First, the value function assumes that people judge outcomes relative to some neutral point of comparison, or reference point, and thus encode them as gains or losses. Second, it assumes that people exhibit loss aversion; they find losses to be more painful than comparable gains are attractive. Thus, the value function is steeper below the reference point in the domain of losses than above it in the domain of gains. Finally, the value function assumes that people experience diminishing sensitivity to outcomes—they are less and less sensitive to changes as they move away from the reference point (i.e., the marginal value is less and less). Because of diminishing sensitivity, the value function is concave in the domain of gains and convex in the domain of losses. The principle of diminishing sensitivity is the most important of the three for our analysis because it determines risk preference. Diminishing sensitivity implies that decision makers will be risk averse for choices involving gains (where the value function is concave) but risk seeking for choices involving losses (where the value function is convex). The tendency for people to shift from risk aversion in gains to risk seeking in losses has been called the “reflection” effect (Kahneman & Tversky, 1979) and it has been widely documented (Lattimore, Baker, & Witte, 1992; Payne, Laughhunn, & Crum, 1980, 1981; Tversky & Kahneman, 1992). Our argument is that goals serve as reference points, so when people evaluate performance relative to a goal, the value function predicts how they perceive their performance. Specifically, goals shift decision makers from the domain of gains to the domain of losses and change risk preference accordingly. Consider someone who is below their goal by 10 units, and is considering a risky tactic that has a 50/50 chance of advancing them 5 units towards their goal Goal-Induced Risk Taking 6 or setting them back by 5 units. This person will tend to prefer taking the risk because an advance of 5 units brings substantial satisfaction, while a set-back of 5 units is less painful because of diminishing sensitivity. Overall, the value function makes a strong prediction that if people treat goals as reference points, they will typically behave in a risk seeking manner when they are below their goal. Strategy Selection and the Value Function If our argument about the value function is true, then it may help to explain some interesting results in the traditional goal-setting literature. Research has suggested that goals may affect the way that people choose and develop strategies. However, aspects of these results on strategy development are hard to explain using traditional mediating mechanisms. In their reviews of the literature, Locke and Latham (1990, 1991) listed a number of mediating mechanisms to explain why goals increase performance; the two most central are that goals increase effort and persistence. Another potential mediating mechanism listed by Locke and Latham is that goals may alter the way people develop their strategies. The evidence that goals change strategy development is somewhat difficult to interpret. Some of the evidence can be interpreted as increased effort or persistence. For example, when people are given a goal, they tend to pursue their strategies more carefully and consistently (Earley & Perry, 1987), particularly when they are provided with a specific strategy that will reliably attain the goal (Earley, Connolly, & Lee, 1989). If strategy depends primarily on effort and persistence, then it can be subsumed under the first two mediators of Locke and Latham. However, other results on strategy are difficult to explain in terms of effort and persistence. For example, people who have high goals make changes to their strategies that are more drastic and more frequent. Evidence that goals produce larger changes in strategies is provided by studies of a complex computer simulation (Bandura & Wood, 1989; Wood & Bandura, 1989; Wood, Bandura, & Bailey, 1990). In these studies, participants with high goals were more likely to change multiple factors of the simulation at once (i.e., to make larger Goal-Induced Risk Taking 7 changes), which made it more difficult for them to interpret feedback and led them to perform worse. Evidence that goals produce more frequent changes is provided by studies of a challenging multi-cue probability learning task; when participants had high goals, they changed their strategies more often and their strategies were less consistent as measured by the fit of a linear regression to their choices (Earley, Connolly, & Ekegren, 1989; see also Earley, Connolly, & Lee, 1989). The results of this study are particularly difficult to explain with the standard mediating mechanisms of effort or persistence. Indeed, if anything, in these studies a goal made participants less persistent. When people with high goals make larger and more frequent changes, they are accepting an element of uncertainty that has not been emphasized by previous researchers: Large, frequent changes increase both upside and downside opportunities, thereby increasing the variance in final performance. Why then, do people with high goals accept this unpredictability? We suggest that the value function provides a plausible answer. If goals serve as reference points, then people who are below their goal will see themselves as in the domain of losses. The value function in this area is convex, so people experience a lower opportunity cost if they remain far from their goal and they receive a higher upside return if they move toward it. This analysis, which depends on the property of diminishing sensitivity, is conceptually identical to the analysis of risk-seeking that we offered above. Thus, if goals lead people to become more risk-seeking, then the value function promises to explain some heretofore puzzling results in the goal setting literature on strategy development. In order to show that goals affect strategy development, it is important that we separate the effects that goals have on strategy choice from the effects of goals on strategy performance (e.g., effort and persistence). In the studies in this paper, we do this by (1) separating strategy choice from strategy performance (Studies 1 and 4) or (2) by selecting tasks that make performance depend primarily on strategy choice, not effort and persistence (Studies 2 and 3). Goal-Induced Risk Taking 8