HOW POSITIVE AND NEGATIVE FRAMES INFLENCE THE DECISIONS OF PERSONS IN DIFFERENT CULTURES

Abstract

Levin et al. (1998) identified three distinct types of framing manipulations that may affect consumer judgments and decisions. The present study tested the reliability of these effects across samples of subjects in the United States and Australia.

Introduction

Researchers studying consumer decision making processes have frequently shown that objectively equivalent information is responded to differently depending on the manner in which the information is labeled or "framed." For example, Levin and Gaeth (1988) showed that ground beef was evaluated more favorably when it was labeled as 75% lean rather than 25% fat. The lesson here appears to be "accentuate the positive." On the other hand, Meyerowitz and Chaiken (1987) showed that women were more apt to engage in breast self-examination (BSE) when presented with information stressing the negative consequences of not engaging in BSE than when presented with information stressing the positive consequences of engaging in BSE. Here the lesson appears to be "accentuate the negative."

More generally, recent reviews of framing studies by Rothman and Salovey (1997), Kuhberger (1998) and Levin, Schneider, and Gaeth (1988) revealed that the magnitude and direction of framing effects varied considerably across studies, as a function of task characteristics such as the presence or absence of risk and subject characteristics such as level of involvement and prior experience. Noticeably absent however, are studies that test the reliability of these effects across cultures. The present study adds an analysis of cross-cultural difference in framing effects and asks whether people in different cultures are equally susceptible to framing effects. If differences are found, their contribution to advances in theory and research will depend on the ability to identify fundamental processes underlying differences in overt behavior.

Our ability to identify fundamental processes underlying framing effects comes from Levin et al.'s (1998) typology and theoretical analysis. According to this typology, three different kinds of framing effects can be distinguished on the basis of operational definitions, typical results, and likely theoretical mechanisms. Attribute framing effects occur when evaluations of an object or product are more favorable when a key attribute is framed in positive rather than negative terms. According to Levin and Gaeth (1988), this is because positive labels tend to evoke positive associations while negative labels tend to evoke negative associations, which are then mapped onto the response scale. Goal framing effects occur when a persuasive message has different appeal depending on whether it stresses the positive consequences of performing an act to achieve a particular goal or the negative consequences of not performing the act. One observation is that the negative message is more persuasive because most people are more motivated to avoid a loss than to achieve a gain. This is called "loss aversion." However, not all studies of goal framing have reported such a result. Risky choice framing effects occur when willingness to take a risk (e.g., elect a risky medical procedure) depends on whether potential outcomes are positively framed (e.g., in terms of success rate) or negatively framed (e.g., in terms of failure rate). Typically, people are more willing to take risks with negatively framed outcomes than with positively framed outcomes. That is, they are more apt to take a risk to avoid a loss than to achieve a gain. This is explained by prospect theory's value function which is concave in the domain of gains and convex in the domain of losses (Kahneman & Tversky, 1979).

In the current study, subjects are presented with an integrated series of three decision making tasks, one involving attribute framing, one involving goal framing, and one involving risky choice framing. To date, subject populations are college students from the United States and Australia. Additional data collection is anticipated in Italy and the Netherlands. Comparisons across these populations will allow us to address the following questions: (1) How reliable is each type of framing effect across cultural differences? (2) How do samples from each country differ in terms of responsiveness to positive frames and in terms of responsiveness to negative frames? (3) How do the samples compare in terms of risk taking tendencies?

Method

One hundred and thirty college students from the University of Iowa, USA, and 97 students from the University of Western Sydney, Australia, were given a three-part questionnaire. The first part asked their reactions to an attribute framing task.

Participants were told to assume that they were inviting a special friend to dinner and that they were making their favorite lasagna dish with ground beef ("minced beef" for the Australians). Half the participants within each population were told that the beef was labeled "80% lean" (the positive condition) and half were told that the beef was labeled "20% fat" (the negative condition). They were then asked to evaluate the beef on the following 7-point bipolar scales: fat-lean, greasy-greaseless, low quality-high quality, and bad tasting-good tasting.

The second part of the questionnaire asked their reactions to a goal framing task. Participants were told to imagine that they were considering eliminating or reducing the amount of red meat in their diet. They were then shown an excerpt from an article describing the effects of eating red meat. Half the participants (the positive condition) were told: "If you discontinue eating red meat you will be able to reduce the level of cholesterol in your blood. Thus, you will significantly decrease the likelihood of the early onset of heart disease." The other half (the negative condition) were told: "If you continue eating red meat you will not be able to reduce the level of cholesterol in your blood. Thus, you will fail to significantly decrease the likelihood of the early onset of heart disease." Participants in each condition were then asked to write a number between 0 and 100 to indicate how likely they are to eliminate red meat from their diet, and to write a number between 0 and 100 to indicate how likely they are to reduce by at least 1/3 the amount of red meat in their diet.

Finally, participants were asked their reactions to a risky choice framing task. They were to imagine that one of their parents was diagnosed as having dangerously high levels of cholesterol. They were told that two programs have been developed for treating high levels of cholesterol. Participants assigned to the positive condition were given the following descriptions: "If Program A is adopted, 1/3 of the persons treated will succeed in reducing their cholesterol. If Program B is adopted, there is a 1/3 chance that all of the persons treated will succeed in reducing their cholesterol and a 2/3 chance that none of the persons treated will succeed in reducing their cholesterol." Participants assigned to the negative condition were given the following descriptions: "If Program A is adopted, 2/3 of the persons treated will fail to reduce their cholesterol. If Program B is adopted, there is a 1/3 chance that none of the persons treated will fail to reduce their cholesterol and a 2/3 chance that all of the persons will fail to reduce their cholesterol." Participants in each condition were then asked which of the two programs, A or B, they would recommend for their parent. Choice of Program A represents a risk averse choice and choice of Program B represents a risky choice.

Results

Results for each country in each task are summarized in Table 1.

Attribute framing

On each of the four rating scales and for each country, responses were more favorable in the positive (80% lean) condition than in

the negative (20% fat) condition. Statistically significant differences (p < .05) were found for three of the four scales in the U.S. sample and for all four scales in the Australian sample. This replicates Levin and Gaeth (1988) and other studies showing that evaluations are more favorable when a key attribute of an object or product is framed positively rather than negatively.

Goal framing

For American subjects, both the rated likelihood of eliminating red meat from the diet and the rated likelihood of reducing by at least 1/3 the consumption of red meat were significantly higher in the negative condition which stressed the disadvantages of failing to reduce or eliminate red meat than in the positive condition which stressed the advantages of reducing or eliminating red meat. This replicates Meyerowitz and Chaiken (1987) and others who found that health-related messages were more effective when they stressed the negative consequences of failing to perform a desired act than when they stressed the positive consequences of performing the act. However, neither result approached significance for the Australian subjects.

Risky choice framing

For each country, participants in the negative condition which compared a risky and a riskless option in terms of failure rate were significantly more apt to choose the risky option than were participants in the positive condition which compared a risky and a riskless option in terms of success rate. The majority of participants in the positive condition chose Program A while the majority of participants in the negative condition chose Program B. (Note the similarity of percentages in each country.) This "preference reversal" is consistent with results for Tversky and Kahneman's (1981) Asian disease task and other similar tasks where subjects chose between a risky and a riskless option of equal expected value.

Discussion

The reliability of framing effects across American and Australian samples was established for two categories of framing effects: attribute framing and risky choice framing. Based on Levin et al.'s (1998) analysis of these categories, it appears that associative processes to positive and negative labels in attribute framing and attitudes toward risk for gains and for losses in risky choice framing are similar for the two samples. The two samples, however, differed in reaction to our goal framing task. The negatively framed message was more persuasive than the positively framed message for the U.S. sample but not for the Australians. Two possible reasons come to mind. First, goal framing seems to be the least reliable of the three categories of framing effects, as seen in both the Levin et al. (1998) review and in the present study where even for the U.S. sample, the level of significance of observed differences was less extreme for this category than for the others. Lack of significance in this category for the Australians may simply be a reflection of this lower reliability.

A more interesting possibility is that the process captured by the goal framing manipulation, loss aversion, differs across cultures.

Loss aversion is a process that potentially affects risk taking in a variety of situations in which a potential loss of a given magnitude is felt more than a possible gain of the same magnitude and thus causes people to avoid situations with potential losses. If persons in a particular culture are less subject to this tendency, then it should affect a variety of decisions involving possible loss. The present study should thus be followed up by comparing Australians and Americans on a variety of tasks involving the operation of loss aversion.

Of course, for the purpose of examining cultural differences in decision making, cultures more diverse than the United States and Australia must be compared. This is a relatively new domain for decision making researchers but one receiving increased attention. (See review by Weber and Hsee, 1999.) The study of framing effects is an important part of "the psychology of decision making" because such effects clearly show the operation of subjective processes in evaluating objectively-equivalent choice options and have important marketing implications for the effective presentation of product and service information. Present plans are for collecting data on framing effects in a number of different countries with possibly different attitudes toward risk and loss.

TABLE 1

Summary of Results

	U.S. Sample						Australian Sample
Attribute Framing
	80% lean		20% fat			80% lean			20% fat
	n	mean	n	mean	t	n		mean	n	mean	t

fat-lean	65	4.29	65	3.51	2.70**	47		4.49	50	3.50	2.99**
greasy-greaseless	64	3.61	64	3.22	1.58	47		4.55	50	3.48	3.37**
low quality-high quality	65	4.49	65	3.68	3.10**	47		4.98	50	3.68	3.91**
bad tasting-good tasting	65	5.08	65	4.45	2.50**	47		5.17	50	4.18	2.93**

Goal Framing
	positive message		negative message			positive message			negative message
	n	mean	n	mean	t	n		mean	n	mean	t

likelihood eliminate red meat	65	36.11	62	49.36	2.26*	47		42.72	50	47.34	0.70
likelihood reduce by at least 1/3	65	56.08	62	67.50	1.93*	47		51.26	50	53.22	0.30

Risky Choice Framing
	success rate		failure rate		c²	success rate			failure rate		c²
	freq	%	freq	%		freq		%	freq	%

choose a (riskless)	40	71	19	37		30		73	16	36
choose b (risky)	16	29	35	63	14.51**	11		27	28	64	11.58**

* p < .05

** p < .01