The role of framing effects in performance on the Balloon Analogue Risk Task (BART)

https://doi.org/10.1016/j.paid.2006.11.026Get rights and content

Abstract

This study examined the impact of framing effects on performance in the Balloon Analogue Risk Task (BART). Seventy-two undergraduate participants were assigned to play the BART as either a gains task in which they had to pump up an animated balloon in order to win money (Group GBART) or as a losses task in which pumping up the balloon was necessary to avoid losing money (Group LBART). Each group went through the BART three times, playing for themselves, for their best friend, and for a charity president. Although there was no effect of the recipient on BART performance, individuals in the LBART group showed riskier performance (pumped up the balloon more) than GBART participants. Furthermore, GBART performance was significantly correlated with scores on the Zuckerman Sensation Seeking Scale while LBART scores were not. These findings support the idea that framing effects can affect performance on behavioral tasks aimed at measuring personality traits.

Introduction

Risk-taking is currently studied across a variety of areas in psychology, health care, and economics. In personality studies, risk-taking is often viewed as a stable trait assessed through self-report or behavioral measures and used to predict performance across situations and tasks (see Harrison et al., 2005, Lauriola et al., 2005, Reynolds et al., 2006). Risk-taking has also been a focus of interest in behavior decision theory. In that tradition, risk-taking is viewed as context-dependent and susceptible to “framing effects” in which the description of the situation can alter participant choices (see Harrison et al., 2005, Lauriola et al., 2005). Studies of decision making typically focus on responses to individual decision problems rather than on trait assessments of the participants. As Lauriola et al. (2005) discuss, there has been little research aimed at examining possible interactions between framing effects and personality trait measures.

One of the best known lines of research in decision making comes from “Prospect Theory” (Kahneman & Tversky, 1979) which states in part that an individual’s willingness to make a risky choice will depend on whether the decision outcomes are framed as gains or losses. Because the negative affective states associated with losses have more emotional force than the positive states associated with gains (the basis of so-called “loss aversion”), individuals will make riskier choices to avoid losses than they will to produce gains. Numerous studies have documented such framing effects (see Levin, Schneider, & Gaeth, 1998, for a review).

One factor playing a role in the expression of framing effects is whether participants make choices for themselves or for another. Krishnamurthy and Kumar (2002) showed that while people will show a gains/losses framing effect when making choices for themselves, this effect largely disappears when participants make the choice for another. Hsee and Weber (1997) suggest that this self-other difference occurs because individual choices are governed by emotional states which generate framing effects (the “risk as feeling” approach, see Loewenstein, Weber, Hsee, & Welch, 2001). When individuals make choices for others, their responses will depend on the degree to which they perceive those others as having emotional states similar to their own. Thus, choices made for abstract others should be less affected by framing effects than should choices made for the self or for close individuals who are perceived as similar to self.

There has been much attention recently to a behavioral measure of trait risk-taking called the Balloon Analogue Risk Task (BART) developed by Lejuez et al. (2002). The BART was developed in an attempt to avoid some of the limitations of self-report measures such as demand effects, failures of memory, and inaccurate introspections. In this computer-based task, participants are shown a series of animated balloons one at a time. For each balloon, participants repeatedly press a button to administer a puff of air that causes the balloon to increase in size (here called a “pump”) and a fixed sum of money to be added to a temporary account. On any given trial the balloon may pop, causing the participant to lose all of the money earned for that balloon. After each pump that does not result in the balloon breaking, participants can transfer money from the temporary account into a permanent one; doing so terminates trials on that balloon. An individual’s level of risk-taking is indexed by the average number of pumps they deliver per balloon.

BART scores have been compared both to self-report measures of sensation-seeking such as the Zuckerman Sensation Seeking Scale (SSS; Zuckerman, Eysenck, & Eysenck, 1978) and to reports of real-world risky behaviors. Some studies have found positive correlations between the BART and SSS scores (Lejuez et al., 2005, Lejuez et al., 2002), but others have failed to find such relationships (Aklin et al., 2005, Lejuez, Aklin, Jones et al., 2003, Lejuez, Aklin, Zvolensky, et al., 2003). Such results may indicate that the BART and self-report measures of personality capture different dimensions of the risk-taking construct (e.g. Reynolds et al., 2006), but they could also be the product of small sample sizes. BART scores are more consistently associated with self-reports of risk behaviors such as drug use, gambling, and risky sexual behavior (Aklin et al., 2005, Lejuez, Aklin, Zvolensky, et al., 2003, Lejuez et al., 2002, Lejuez et al., 2004). Furthermore, BART scores are higher in cigarette smokers than in nonsmokers (Lejuez et al., 2005, Lejuez, Aklin, Jones et al., 2003), higher in cocaine users than in heroin users (Bornovalova, Daughters, Hernandez, Richards, & Lejuez, 2005), and higher in adolescents with conduct disorder than in controls (Crowley, Raymond, Mikulich-Gilbertson, Thompson, & Lejuez, 2006).

Studies of personality measures of risk have typically not examined whether performance is consistent across different task framings. In this context, it should be noted that the BART is clearly a “gains” task; that is, participants pump up the balloon in order to accumulate earnings upwards from a starting point of zero. Furthermore, the BART is also clearly a “self” task; that is, participants themselves are the recipients of the money earned.

The present study was designed to examine whether the two framing effects discussed here (gains/losses and self/other) would impact BART performance. Half of the participants were exposed to the traditional BART in which players began with no money in a temporary bank at the start of each balloon and then accumulated a fixed sum with each pump (here called the Gains BART or GBART). These temporary winnings (or gains) disappeared if the balloon burst. The other half of participants received a revised BART in which they began each balloon with money already in a temporary account which they were told they would lose at the end of the balloon. Each pump on the balloon acted to offset that loss by a fixed amount; if the balloon burst, then all the money was lost (Losses BART or LBART). In this way, each pump led to exactly the same amount of money being won in both BART versions. However, in the GBART the money accumulated upwards from zero as a gain, whereas in the LBART the money accumulated upwards from a negative sum as a reduction of loss. Each subject then played either the LBART or the GBART under three conditions: for themselves, for their best friend, or for a local charity president.

We had three hypotheses. First, we predicted that participants would show riskier behavior (average more pumps per balloon) in the LBART than in the GBART. This prediction grows out of the prospect theory finding that subjects are more risk-seeking when avoiding losses than when accumulating gains. Second, we predicted that performance on the BART would vary as a function of the target recipient. On the LBART, participants should be most risky in the self condition and least risky in the charity president condition. In the GBART, we predicted the opposite pattern. In other words, we expected framing effects (risky on losses, conservative on gains) to be most pronounced when making decisions about the self and least pronounced when making decisions for a charity president. This hypothesis reflects the Krishnamurthy and Kumar (2002) finding that framing effects present in a “self” task may be diminished when choice are made for another (see also Hsee & Weber, 1997).

Our third hypothesis was more speculative and involved the relationship between the two BART versions and scores on the Zuckerman Sensation Seeking Scale. Past studies of the BART (here called the GBART) have provided some evidence for a positive relationship with SSS scores (Lejuez et al., 2002, Lejuez et al., 2004). Given the role of loss aversion in decision making, we wondered if scores on the LBART might be less well correlated with the SSS. That is, the general tendency to become risky in the face of losses might produce more homogenous scores on the LBART and therefore make such scores less well related to trait indices of risk-taking. Furthermore, if sensation-seeking as a construct is “gains-like” (presumably people are seeking gains rather than losses), then one might expect the SSS to better correlate with GBART scores on that basis as well. We do not know of existing data to support this idea, but thought it worth examining.

Section snippets

Participants

Participants were 72 undergraduate students (24 male and 48 female) from a local university who were recruited either by posters or by word-of-mouth referrals. All subjects were between the ages of 18 and 25 (M = 20.2, SD = 1.3) and all gave written consent to participate in the study.

Research design

The experiment was set up as a 2 × 3 mixed factorial design. We examined the role of gains/losses framing by assigning subjects through blocked randomization to either the LBART or GBART condition (between-subjects

Baseline scores on the Zuckerman Sensation Seeking Scale

Lejuez et al., 2002, Lejuez et al., 2004 found that BART scores were correlated with scores from the Zuckerman scale. Therefore, we first compared SSS-T scores between the LBART and GBART groups. The two groups produced nearly identical scores (GBART: M = 20.1, SD = 7.0; LBART: M = 20.6, SD = 6.7) which were not significantly different on a t-test (p > .05). This finding makes it less likely that group differences in BART score could be attributed to pre-existing differences in sensation-seeking.

Analysis of BART scores across conditions

Discussion

Two of the three study hypotheses were confirmed. First, participants faced with a losses-framed task (the LBART) were significantly riskier than those faced with a gains-framed task (the GBART). In both groups, single pumps on each balloon led to 5 cents being deposited in a temporary bank account with the potential to deposit up to $6.25 per balloon. Nevertheless, when this task was framed as avoidance of a loss rather than as production of a gain, participants made significantly more

Acknowledgements

This research was completed as part of the requirements for a B.A. in psychology by Alexander M. Benjamin. We would like to thank C.W. Lejuez for graciously making available to us the source code for the original BART and for taking the time to answer our many questions about the programming and scoring of the task.

References (17)

There are more references available in the full text version of this article.

Cited by (0)

View full text