| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
1 Department of Psychology, University of Toronto, Canada.
2 The Rotman Research Institute of Baycrest Centre, Toronto, Canada.
3 Department of Psychology, Michigan State University, East Lansing.
Address correspondence either to Sunghan Kim or Lynn Hasher, both of whom are at the Department of Psychology, University of Toronto, 100 St. George Street, Toronto, Ontario M5S 3G3, Canada. E-mail: shkim{at}psych.utoronto.ca or hasher{at}psych.utoronto.ca
 |
Abstract |
|---|
 TOP Abstract MethodResultsDiscussionReferences |
|---|
THE term framing effect refers to a phenomenon whereby the choices people make are systematically altered by the language used in the formulation of options. For example, in the well-known Asian disease task (Tversky & Kahneman, 1981
), participants in the so-called positive frame condition make a treatment choice based on a description of lives saved (e.g., 200 out of a total of 600 people will be saved); by contrast, in the so-called negative frame condition, participants make a treatment choice based on a description of lives lost (e.g., 400 out of a total of 600 people will die). Despite the identical underlying basis of the two versions of the problem, people make different choices: They are more likely to be risk averse (i.e., to move away from a risky option) when questions are framed as gains (i.e., positively) and more risk seeking (i.e., to move toward a risky option) when questions are framed as losses (i.e., negatively).
The literature on these and other framing-effect problems is extensive (for reviews, see Levin, Schneider, & Gaeth, 1998; also see Kühberger, 1998). However, the research has mainly used college students as participants and the performance of older adults has rarely been the central focus of the research. Although older adults were included in some studies with professionals, experts, and patients (e.g., Loke & Tan, 1992; McNeil, Pauker, Sox, & Tversky, 1982; Roszkowski & Snelbecker, 1990), the age of the samples in these studies was also confounded with other characteristics such as expertise and some medical conditions.
Previously, we (Kim & Hasher, 2005) demonstrated superior performance by older adults in decision making using attraction-effect tasks, suggesting that older adults can, under some circumstances, show performance better than that of younger adults, perhaps because they have greater life experience in decision making (e.g., Hertwig, Barron, Weber, & Erev, 2004). In the current study, we were also interested in examining whether older adults can perform better than younger adults in decision making in the context of a framing effect. As in our previous study, it is conceivable that older adults are less likely to show framing effects than younger adults because they have greater life experience in decision making. However, an alternative hypothesis is also possible.
In the framing-effect literature, there is some evidence that even people with experience or expertise are not immune to framing effects (e.g., Loke & Tan, 1992; Roszkowski & Snelbecker, 1990). Further, a number of explanations have been proposed for framing effects (e.g., see Kühberger, 1997, for a review); there has been some agreement that, in many instances, "framing" is the product of heuristic information processing (e.g., McElroy & Seta, 2003). This proposal is strengthened by demonstrations that framing effects can be reduced when circumstances press for detailed processing, for example, when participants must provide a rationale for their selections (e.g., Sieck & Yates, 1997; Takemura, 1993). Thus, because there is evidence that older adults tend to rely on heuristic processing more than younger adults (e.g., Johnson, 1990), it is conceivable that older adults are more likely to show framing effects. Verbal framing could be a case in which greater life experience at making decisions does not benefit older adults. If it is such a case, then our question becomes whether asking for a rationale for a decision can increase reliance on detailed processing and thus make it possible for older adults to make decisions that are similar to those of younger adults.
|
METHOD |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
), than younger adults (M = 19.39, SD = 7.04), t(368) = 11.75. Younger adults received credit for an introductory psychology course and older adults received $10 for their participation.
Materials
We used two problems, each with a positive and negative frame. We adopted one problem, the "fatal disease" problem, from Wang, Simons, and Brédart (2001). We did not use the well-known Asian disease problem in this study because of the demographic characteristics in the Greater Toronto area, in which a considerable portion of the population is Asian. We adopted the other problem, the "cancer treatment" problem, from McNeil and colleagues (1982; see Appendix). Note that, for both problems, the option labeled as B is the one that previous work shows is chosen reliably more in the negative frame than in the positive frame.
|
). Participants who were in the without-justification condition were not asked to provide a rational for their choice.
Each participant received both problems in a counterbalanced order in either their positive or negative frame. Because research has shown that a reliance on analytic strategies is more likely at optimal times of day whereas a reliance on heuristic strategies is more likely at nonoptimal times (e.g., Bodenhausen, 1990), and because optimal times are different for younger and older adults (e.g., Yoon, May, & Hasher, 1999), we tested at normatively optimal times for each group, that is, in the morning for older adults and in the afternoon for younger adults. Our goal was to increase the likelihood that analytic strategies would be used by both groups in the standard condition.
|
RESULTS |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
2(1, N = 106) = 0.99, and they showed a marginal effect in the cancer treatment problem, 2(1, N = 106) = 3.13, p =.08. (The effect size of the original Asian disease problem is actually exceptionally large; its effect size was categorized as an outlier in the meta-analysis of Kühberger, 1998. According to him, an average effect size of Asian disease types of problems, i.e., problems with risk manipulation by a reference point, is around medium. To give room for an effect to go up or down with younger and older participants, we selected problems with effect sizes slightly lower than medium in the current study. Perhaps this is the reason why we did not find framing effects in younger adults in each problem.) Older adults, by contrast, showed a significant framing effect for both the fatal disease problem, 2(1, N = 106) = 9.95, p <.01, and the cancer treatment problem, 2(1, N = 106) = 16.06, p <.01.
|
2(1, N = 80) = 0.25, or of cancer treatment, 2(1, N = 80) = 1.26. Similarly, older adults did not show a framing effect for either the fatal disease problem, 2(1, N = 80) = 1.32, or the cancer treatment problem, 2(1, N = 80) = 1.35.
Across Problems: Pooled Data Analysis
We combined the choice data from the two problems to enable a powerful test of our hypotheses. (In order to examine the independence of participants' responses across two problems, we used McNemar's test. McNemar's test is a test used to examine whether repeated categorical responses in two different conditions are independent of each other. The test results showed that responses or choices are independent, i.e., significantly different, from each other across the two problems for younger and older adults, respectively, in each justification condition. Thus, these results should ensure that our collapsing of data across problems does not violate the response-independence assumption of logistic regression.) We analyzed these pooled data by using logistic regression to examine main effects and interactions among three independent variables: frame (positive and negative), age (young and old) and justification (with and without justification). First, we carried out model selection by using stepwise and backward selection methods to find a logistic regression model that best fits the data (e.g., SAS Institute, 1989).
The final model we selected (Hosmer–Lemeshow goodness-of-fit 2 = 1.58, p =.95, df = 6) excluded the three-way interaction (Age x Justification x Frame) and the one two-way interaction (Age x Justification), and it included only three main effects (frame, age, and justification) and two two-way interactions (Age x Frame and Justification x Frame). Of these five parameters in the final model, three (one main effect and two interactions) were statistically significant. Specifically, the main effect of frame was significant, 2(1, N = 744) = 6.84, p <.01, indicating an overall framing effect. This effect, however, was qualified by two hypothesized interactions—Age x Frame, 2(1, N = 744) = 4.66, p <.05, and Justification x Frame, 2(1, N = 744) = 4.53, p <.05.
The Age x Frame interaction indicates that participants' age had an effect on the magnitude of the framing effect. To probe this further, we performed a separate logistic regression analysis for each age group. The main effect of frame was significant only for older adults, 2(1, N = 372) = 22.29, p <.01, but not for younger adults, 2(1, N = 372) = 3.13, p =.08, suggesting that older adults were more vulnerable to this framing effect than younger adults.
The Justification x Frame interaction indicates that the justification manipulation had an effect on the magnitude of the framing effect. Separate analyses that we performed for each justification condition revealed a significant main effect of frame only for the without-justification condition, 2(1, N = 424) = 23.14, p <.01, but not for the with-justification condition, 2(1, N = 320) = 2.12. These results suggest that the justification manipulation significantly reduced the framing effect for both younger and older adults.
|
DISCUSSION |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
Furthermore, as we expected on the basis of earlier studies asking younger participants to provide a rationale for their decisions (e.g., Sieck & Yates, 1997; Takemura, 1993), the justification manipulation significantly reduced the present framing effect for both younger and older adults. Clearly, older adults can use analytic processes that enable them to base their decisions on a problem's deeper structure, rather than on its superficial language.
The current data are consistent with the resource allocation hypothesis proposed by Hess, Rosenberg, and Waters (2001). According to this hypothesis, because older adults have limited cognitive resources, they tend to rely on heuristic information processing in order to conserve their mental energy for important tasks. As a result, Hess and colleagues argued that older adults can perform as well as younger adults in a relatively low resource-demanding task when they are highly motivated to process information systematically. Our data fit with this explanation, if this decision situation can be thought of as one making a low demand, in that older adults were more likely to show framing effects than younger adults because they tended to rely on heuristic information processing more than younger adults, but their performance improved up to the level of younger adults when they were encouraged to process information systematically.
Our current findings have important practical implications for older adults' decision making, especially in medical decision domains. This is a domain full of risky decisions that have to be made. Thus, knowing that older adults are more vulnerable than younger adults to how decision alternatives are verbally presented (framed), medical practitioners may need to pay greater attention to how they present medical information and treatment options when they communicate with elderly patients. However, our current finding that older adults' heightened susceptibility to verbal framing can be ameliorated by a simple justification manipulation suggests that a relatively easy, everyday solution may be available to older adults—along with others—to reduce some biases in choice.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication September 3, 2004. Accepted for publication February 16, 2005.
|
References |
|---|
|
TOPAbstractMethodResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  K. Brummel-Smith and J. Spike Refusal of Care by Patients JAMA, December 27, 2006; 296(24): 2922 - 2923. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
