Do Consumers' Reference Points Affect Their Buying Decisions?

ABSTRACT - This paper examines the applicability of prospect theory's reference point hypothesis to the consumer decision making process. We present a theoretical basis for reference point formation, and we report the results of a pilot study which tests this concept. In general, the results support the hypothesized relationship between the reference point and choice, and an individual difference factor, self esteem, was found to affect reference point formation.

Citation:

Debra Rowe and Christopher P. Puto (1987) ,"Do Consumers' Reference Points Affect Their Buying Decisions?", in NA - Advances in Consumer Research Volume 14, eds. Melanie Wallendorf and Paul Anderson, Provo, UT : Association for Consumer Research, Pages: 188-192.

Advances in Consumer Research Volume 14, 1987      Pages 188-192

DO CONSUMERS' REFERENCE POINTS AFFECT THEIR BUYING DECISIONS?

Debra Rowe, The University of Michigan

Christopher P. Puto, The University of Michigan

ABSTRACT -

This paper examines the applicability of prospect theory's reference point hypothesis to the consumer decision making process. We present a theoretical basis for reference point formation, and we report the results of a pilot study which tests this concept. In general, the results support the hypothesized relationship between the reference point and choice, and an individual difference factor, self esteem, was found to affect reference point formation.

INTRODUCTION

Researchers in a variety of academic and applied fields have long been interested in studying consumer choice, and the approaches used for studying choice phenomena reflect the varied backgrounds and interests the researchers bring to the task. Thus, for example, people's choices have been studied when they cast their vote for a political candidate, when they utilize a particular mode of transportation, and when they purchase a product or patronize a particular store. All of these choices have a common thread in that an individual decision maker considers two or more alternatives and their related outcomes, evaluates these alternatives along dimensions important to him or her, and selects the alternative which rates the highest on whatever criteria s/he wishes to use.

In this paper, we examine the applicability of prospect theory (Kahneman and Tversky 1979; 1982; 1984; Tversky and Kahneman 1981), a theory of individual choice under conditions of uncertainty, to consumer choice situations. Moreover, within the general scheme of prospect theory, we concentrate on exploring the foundations of the decision maker's reference point, which is hypothesized to determine the decision frame and, ultimately, the choice itself (Kahneman and Tversky 1979; Puto 1986). While we utilize only one form of consumer choice in this study--choosing between two retail sites--we propose that the framework for this investigation is readily adaptable to a variety of decision situations. In the remainder of this paper, we (1) briefly review prospect theory, especially with regard to our current knowledge of reference point formation, (2) present a series of research hypotheses derived from the literature, and (3) describe the results of a pilot study designed to explore the hypothesized origins of the reference point. The paper concludes with a summary of what we have learned and the resultant implications for future research.

THEORETICAL BACKGROUND

In prospect theory, Kahneman and Tversky (1979) propose that decision makers (1) evaluate decision alternatives by expressing their outcomes as either gains or losses compared with a specific reference point and (2) tend to be risk averse for choices involving gains and risk taking for choices involving losses. Prospect theory is a descriptive extension of expected utility theory, a theoretical framework long used by economists and decision scientists both to describe how individuals make decisions and to prescribe how they should make decisions. Prospect theory essentially addresses the descriptive shortcomings of expected utility theory, which are numerous and widely documented in the decision research literature (Schoemaker 1982). Issues relating to the normative (i.e., prescriptive) aspects of expected utility theory warrant separate treatment, and they are not discussed here (cf. Lopes 1983).

Central to prospect theory is the proposition that the choice process has two distinct stages - an editing stage and an evaluation stage. In the editing stage, the decision maker restructures, or frames, the decision problem into a more simplified form, such as a choice among simple prospects (i.e., gambles of the form: receive outcome "x" with probability "p"; else receive outcome "y" with probability "l-p"). A key aspect of this editing stage is that the prospects are derived on the basis of comparing each outcome with its deviation from some reference point, which exists in the mint of the decision maker. The reference point serves as the zero point on this scale of comparison, and each outcome then is seen as a gain or a loss compared to that reference point. The evaluation stage then posits that the decision maker assigns a value to each of the edited prospects and chooses the one with the highest value. This evaluation stage is governed by two functions: (1) a value function, which is hypothesized to be centered on the reference point, is concave for gains and convex for losses, and is steeper for losses than for gains; and (2) a probability weighting function, in which objective probabilities are modified by decision weights and thus do not necessarily adhere to the strict rules of mathematical probability theory, e.g., they do not have to sum to one (Kahneman and Tversky 1979). Moreover, small probabilities are overweighted, and large probabilities are underweighted. The hypothesized form of the value function suggests that individuals tend to make choices which are risk averse for gains and risk taking for losses. Gains and losses, however, are determined by the reference point, and currently very little is known about how reference points come into existence. If reference points play a similar role in framing consumers' judgments and decisions, then it is important for consumer researchers to know how they are developed. We turn now to a brief theoretical introduction to reference point formation.

Reference Points and Decision Frames

The original presentation of prospect theory is mute with respect to the formation of the reference point. This was not a problem during the early theoretical development because the original research in prospect theory centered on decision problems which contained explicit, unequivocal reference points. In later work, Tversky and Kahneman (1981) speculated that an individual's current wealth is a likely reference point in decisions involving monetary outcomes. Fischhoff's (1983) seminal paper on reference points raised several issues, but he was unsuccessful in relating experimental subjects' choices to their reference points. Recently, Puto (1986) proposed a conceptual framework for reference point formation, which he then demonstrated in the context of an industrial buying decision. The operationalizations Puto used in testing his framework are necessarily specific to industrial buyers; however, the conceptual framework is sufficiently general to extent to consumer buying situations. The following discussion is a brief restatement of Puto's framework. The reference point formation process is an iterative one in which the consumer first establishes an initial reference point (e.g., a target value on a key dimension, such as price) which is then continuously modified by environmental factors, such as subsequently acquired information about the choice, up to the time of the decision (Puto 1986). At the time of the decision, the decision maker evaluates the alternatives based on the current (i.e., the final) reference point and renders his/her choice. The reference point determines the decision frame which, in turn, governs the choice. Thus, a reference point which permits the outcomes to be framed as gains will tend to promote risk averse choice behavior, and a reference point which permits the outcomes to be framed as losses will tend to promote risk taking choice behavior. Decision frames which produce gains are termed positive frames, and decision frames which produce losses are termed negative frames. In the next section, we hypothesize several factors which affect the formation of the initial reference point and the final reference point.

Research Hypotheses

We begin by hypothesizing a relationship between the final reference point and choice matching that proposed in prospect theory. We then develop hypotheses regarding factors which affect the initial reference point, and we conclude with an exploratory issue regarding the effect of the stated probability on the reference point and choice. First, we turn to the prospect theory hypothesis. Prior research in prospect theory has reported that people tend to be risk averse for gains (i.e., they prefer a sure gain to a probabilistic outcome having equal or greater expected value) and risk taking for losses (i.e., they prefer a probabilistic outcome to a sure loss of equal expected value). If consumers frame buying decisions in a similar manner, then their choices should be similarly affected. Stated in research form, hypothesis one is as follows:

H1: Consumers whose reference points permit the choice to be framed as a gain will tend to exhibit risk averse choice behavior, and consumers whose reference points permit the choice to be framed as loss will tend to exhibit risk taking choice behavior.

As proposed by Puto (1986), the initial reference point derives from the individual characteristics of the decision maker (e.g., personality traits and the decision maker's perception of the choice environment), combined with whatever preliminary information that exists regarding the options to be evaluated. Consumer and marketing researchers have displayed an ongoing interest in relating personality traits to consumer behavior, but these efforts have at best met with mixed results (Kassarjian 1971). While several reasons have been advanced for this lack of correspondence between personality traits and brand choice, one plausible explanation is that the effects are indirect and thus may either be overwhelmed by stronger effects or measured at the wrong stage of the decision process.

If, for example, the reference point influences choice as hypothesized in prospect theory, then factors which influence the reference point are likely not to emerge as direct effects on choice (Puto 1986). One factor which may affect the reference point is the individual's self-esteem. The initial reference point is a target or preliminary goal, and Locke et al. (1981) propose that self-esteem affects goal setting in a predictable manner. Specifically, individuals with high self-esteem should be more likely to set challenging targets than those with low self-esteem. This suggests the following research hypothesis:

H2: People with high self-esteem will tend to form challenging initial reference points; people with low self-esteem will tend to form easily attainable initial reference points.

Another factor which may affect reference point formation is the person's attitude toward risk. Prospect theory essentially posits that, regardless of an individual's inherent risk taking propensities, the decision frame (as determined by the reference point) is the major determinant of risk taking behavior. Work by Huber and Puto (1985) suggests that the direction of causality is open to question. Huber and Puto created an informal risk attitude scale in which individuals gave their preference for several gambles alternatively framed by the experimenter as either gains or losses. Subjects who chose the "sure thing" alternative greater than 80% of the time were classified as "globally risk averse", and subjects who chose the risky alternative greater than 80% of the time were classified as "globally risk seeking". Subjects who were risk averse for gains and risk taking for losses were classified as conforming to prospect theory, and subjects who exhibited the opposite pattern, e.g., risk averse for losses and risk taking for gains, were classified as anomalous . The results showed that while the modal group chose as predicted by prospect theory, there were significant numbers of subjects whose choices did not vary with the decision frame. This leads to the speculation that inherent attitudes toward risk may lead people to form reference points consistent with their risk attitude. In research form, this hypothesis is as follows:

H3: Individuals who are globally risk averse will tend to form initial reference points which are less challenging than those formed by individuals who are globally risk taking.

One generally accepted facet of consumer behavior is that consumers operate in an informationally rich environment. They are exposed to large amounts of product related information, which they may or may not process (Bettman 1979), but they to not function in an informational vacuum. At some stage early in the decision making process, consumers begin to attend to preliminary information regarding the choice they may eventually make, and we propose that this preliminary information contributes to the formation of the initial reference point. Some evidence for this is provided in Puto (1986) in which industrial buyers were given information suggesting that price trends for a given commodity were either increasing or decreasing. Initial reference points for these experimental subjects coincided with the trend they were given. Thus, for example, prior published prices, such as the manufacturer's suggested retail price or advertised sale prices, will likely form the basis of the initial reference point for consumers (provided, of course, that price is a key dimension in the choice). This hypothesis is given in research form as follows:

H4: Preliminary information in the form of published prices will affect the formation of consumers' initial reference points such that relatively high published prices will produce correspondingly high initial reference points, and relatively low published prices will produce correspondingly low initial reference points.

We turn now to the last research hypothesis, which relates the initial reference point to the final reference point. Puto's (1986) conceptual framework posits that the initial reference point can be modified by environmental factors up to the time of the decision. We to not specifically investigate these environmental factors in this research, and in their absence, we hypothesize a relatively close mapping of the initial reference point onto the final reference point. Thus, we posit hypothesis five as follows:

H5: When there are no environmental factors to intervene, the initial reference point determines the final reference point.

Finally, most research on risky choice includes, by definition, an explicit statement of the probabilities associated with each outcome. There is, however, research which suggests that individuals to not treat probabilities with the save rigor as mathematicians and decision scientists (Bar-Hillel 1973), and it is reasonable to speculate that consumers may not be accustomed to dealing with or using explicit probabilities during normal buying situations. Thus, while uncertainties may exist in these situations, it is not clear how consumers use these uncertainties in evaluating alternative choice outcomes. We begin to explore this issue by examining the effects of two forms of phrasing the uncertainty -- as a 50-50 probability and as "a chance"--on consumers' reference points and choices. In the following sections, we report the results of a pilot study which examines the five research hypotheses and the exploratory probability issue.

A PILOT STUDY

The basic design is a 2 (typical price) by 2 (probability) between subjects factorial, with the individual characteristics of self-esteem and risk attitude treated as measured covariates. The four cells in the design matrix are: (1) high typical price, 50-50 probability; (2) high typical price, wa chance"; (3) low typical price, 50-50 probability; and (4) low typical price, "a chance". The self-esteem and risk attitude covariates were measured using a separate questionnaire administered approximately one week prior to the choice task.

Stimuli

The choice task was presented in the form of a hypothetical buying scenario in which subjects were asked to choose between two alternative retail sites when purchasing a video cassette recorder (VCR). The basic format for each scenario was as follows: Subjects were informed that they definitely had to purchase a VCR that day. They were told that the manufacturer's suggested retail price was $500 (which remained constant across all four scenarios), and they were told that they had remembered seeing the same model VCR priced at $400 ($350) but could not recall where. A chain store next to their workplace had the unit they wanted, and it was priced at $375. A shopping center directly on their route home from work had two competing chain stores that carried the same VCR; one was priced at $350 and the other at $400. The store with the low price had only one unit left in stock, and they could not hold it until the subject arrived to purchase it. There was a (50-50) chance the unit would be sold when they arrived. If that happened, they could not return to the first store and would thus have to buy the one at the higher priced store. Their choice, then, was always between taking the "sure" price of $375 at the store near work or going to the shopping center and taking a risk that the low priced store would be sold out when they arrived. The frame was manipulated by the typical price. The $350 typical price should make the $375 price at the store near work a sure loss of $25, and the $400 typical price should turn it into a sure gain of $25

The subjects were adult males enrolled in continuing education courses at a Michigan community college. Participation was voluntary, and subjects received no compensation for the study. The experimental materials consisted of two survey booklets, the first one containing the 20-item self-esteem scale and the 6-item risk attitude scale, the second one (administered one week later) containing the hypothetical purchase scenario and the dependent measures. Completion of each booklet was self-paced.

Measures

Measures were taken of (1) the initial reference point, (2) the final reference point, (3) the subject's assessment of the probability associated with the risky outcome, (4) the subject's self-esteem, (5) the subject's risk attitude, and (6) the subject's choice between the two possible retail sites.

The initial reference point was measured by asking subjects to indicate a target price they planned to pay for the VCR. The only information available was the manufacturer's suggested retail price and the price "they remembered seeing" during a previous search effort (either $350 or $400). If prior information has an effect on the initial reference point, it should be manifested in a downward adjustment of the manufacturer's suggested retail price, and larger adjustments should be observed for the $350 than for the $400 recently observed price.

The final reference point was measured following Puto (1986) in which subjects were asked to select from a list of possible reference points the one which came the closest to the basis of comparison they used in evaluating each alternative. Each scenario contained four possible final reference points as follows: (1) $350 (one of the risky outcomes); (2) $375 (the sure thing); (3) $400 (the other risky outcome); and $500 (the manufacturer's suggested retail price. Note that the price the subject "remembered seeing" was one of the experimental manipulations $350 if a negative (loss) frame was being induced and $410 if a positive (gain) frame was being induced -and thus coincided with either final reference point (1) or (3). The probability measure was an eleven-point scale (0-10), anchored by "extremely unlikely" and "extremely likely", on which subjects indicated the probability they actually used in determining whether or not the uncertain event would take place (Wyer 1974). Choice was measured categorically; either subjects chose the sure thing, or they chose the risky option.

Self-esteem was measured using Eagly's (1967) 20-item scale. Ten items were designed to reveal high self-esteem, and ten items were designed to reveal low self-esteem. Subjects responded by indicating their degree of agreement or disagreement with each item on a seven-point scale anchored by terms such as "always" and "never". The ten low self-esteem items were reverse scored, and the twenty items were summed to produce an overall measure of self-esteem. Risk attitudes were measured using a variant of the Huber and Puto (1985) risk attitude scale. Subjects were given 6 pairs of gambles (3 with positive frames and 3 with negative frames), each involving a choice between a sure outcome and a probabilistic outcome, and were asked to choose the one from each pair they would prefer to play. Subjects who chose the sure outcome five out of six times were designated globally risk averse; subjects who chose the probabilistic outcome five out of six times were designated globally risk seeking; subjects who chose the sure outcome for gains and the risky outcome for losses were designated as complying with prospect theory; and subjects who did not fall into these categories were designated anomalous.

RESULTS

We begin the results by examining the prospect theory hypothesis (El), which posits that high reference points will produce positive decision frames (and generate risk averse choices) and that low reference points will produce negative decision frames (ant generate risk taking choices). After examining the choice problem, subjects read a brief description of each reference point and indicated the one which best described the way they framed the decision. Table 1 shows the percentages of choices allocated between the sure thing and the risky alternative for each final reference point. The significance level for each term reported in Table 1 is based on the difference in chi-square between a full-model containing all of the reference point terms and a "reduced" model omitting the term being tested. This difference is distributed as chi-square with one degree of freedom (Harrell 1986). With the exception of FRP3 ($400), the results are in the hypothesized direction and at least marginally significant. For FRP3, the $400 final reference point should result in a positive frame and thus produce a higher proportion of risk averse choices, which did not occur. Within the limited range of the pilot study, we do find moderate support for H1.

TABLE 1

RELATIONSHIP BETWEEN FINAL REFERENCE POINT AND CHOICE

Research hypotheses H2, H3, and H4 all posit factors affecting the initial reference point, and they are examined using an analysis of covariance in which price (H4) and risk attitude (H3) are treated as experimental conditions and self-esteem (H2) is a covariate. (Note that risk attitude is technically a covariate, but since it was measured as a categorical variable, it was analyzed statistically as a treatment variable in an unbalanced design.) Detailed examinations of each hypothesis are given in order below.

Hypothesis two predicted that individuals high in self-esteem would tend to form more challenging initial reference points than those low in self-esteem. Eagly's (1967) 20-item self-esteem scale consists of ten items designed to reveal high self-esteem and ten items designed to reveal low self-esteem. The low selfCesteem items were reverse scored and then summed with the high self-esteem items to produce an overall measure of self-esteem (Cronbach's alpha = .895). The analysis of covariance revealed no empirical support for H2 (F1 31 1.37, P < .26). '

Hypothesis three predicted that globally risk averse consumers would have higher initial reference points (consistent with positive decision frames) and globally risk seeking consumers would have lower initial reference points (consistent with negative decision frames). Although there was a slight difference in the hypothesized direction for the initial reference point means ($395 for globally risk averse consumers versus $386 for globally risk seeking consumers), this difference tit not begin to approach significance (F . 1 = .32, P < .90). Thus, there is no empirical support for H3 .

Hypothesis four posited that high published prices (e.g., $400) would produce higher initial reference points than low published prices (e.g., $350). The results reveal absolutely no effect for the price manipulation (F1 31 .01, P < . 95), and in fact there is a slight ' (nonsignificant) difference in the opposite direction ($379 versus $382). This suggests that the subjects paid little or no attention to the attempted manipulation and results in no empirical support for H4.

Given the absence of hypothesized effects for H2, H3, and H4, an exploratory analysis was performed on the data to determine what, if anything, might be affecting the initial reference point. A scatterplot of the initial reference point data revealed that individuals with the highest levels of self-esteem formed initial reference points which were slightly less challenging than those formed by individuals with medium levels of self-esteem, suggesting that a quadratic term be added to the model. A model containing a quadratic term for self-esteem was estimated, both with and without the price and risk attitude treatment variables. There were no significant effects for the two treatment variables; so the model reported in Sable 2 contains the results of an ordinary least squares regression of the initial reference point on self-esteem and self-esteem squared. Although this is a post hoc analysis, it does indicate that self-esteem did play a role in influencing the initial reference points for these subjects. The low R (.22), however, suggests that this effect is modest in size.

TABLE 2

POST HOC ANALYSIS OF THE EFFECT OF SELF-ESTEEM ON THE INITIAL REFERENCE POINT

Hypothesis five posited that the initial reference point was a major factor in determining the final reference point. A multinomial logistic regression using the final reference point as a four-level categorical dependent variable and the initial reference point as a continuous predictor variable failed to produce a significant effect (chi-square - 1.55 with 1 d.f., P < .22). A frequency distribution of the initial reference point data revealed a series of clusters approximating discrete points along the continuum from $300 to $500. Based on this, the raw data were categorized into four levels of $350, $375, $400, and $500, converted into three effects coded (1,-1) dummy variables, and used as predictors in a multinomial logit regression. The results provide some measure of empirical support for H5 (model chi-square - 7.84 with 3 d.f., P < .05).

With regard to the exploratory manipulation of a 50-50 probability versus "a chance", the experiment produced no significant results. The probability manipulation occurred after the measurement of the initial reference point and before the measurement of the final reference point. Thus, it was possible to test for an effect of the probability format on the final reference point and on choice. Logistic regressions were computed in which the probability format was a dummy coded treatment variable, and either the final reference point or choice was the dependent variable, and no interpretable results were obtained. This, in turn, raised questions regarding the probability manipulation, which we investigate next.

Wyer's (1974) eleven-point probability measure was collected at the end of the experiment for use as a manipulation check. Subjects were asked to report on to 10 scale the number which best corresponded to the likelihood that the undesirable outcome in the risky option would occur (i.e., the likelihood that the low priced store would be sold out). This measure was then used as the dependent variable in an analysis of variance to determine whether the manner of presenting the probabilities had an effect on subjects' perceptions of the decision probabilities. No effects were observed. There does not appear to be any identifiable relationship between the probabilities reportedly used by the subjects and either the final reference point, choice, or the probabilities stated in the experiment. This finding prompted one final analysis, which is reported below.

Since the dollar amounts of the various outcomes were known and the subjects provided a measure of the probabilities they claimed to have used, it is possible to compute an expected value for each outcome. This expected value can then be compared with the final reference point to determine which, in this instance, is the better predictor of choice. A logistic regression was computed using both factors (expected value and final reference point) as predictors of choice. Then a second logistic regression was computed using only expected value as a predictor variable. The difference in chi-square between the two models represents the statistical significance of the final reference point term. Chi-square for the full-model is 6.92 (t.f. - 4), and chi-square for the expected value only model is 0.51 (d.f. 1). Therefore, the difference in chi-square is 6.41 (t.f. - 3, P < .08), and the final reference point is a better predictor of choice.

DISCUSSION

From the standpoint of using prospect theory to understand consumers' judgment and choice processes, these results are mixed. The empirical support for hypothesis one suggests that the final reference point is indeed a reasonable predictor of choice. Also, since the final reference point is essentially the result of a consumer judgment, it provides both a measure and another approach to exploring consumer judgment processes which, when combined with process tracing methods such as verbal protocols and information monitoring devices, may yield new insight into how such judgments are made. Moreover, the finding that the initial reference point, at least in the absence of effective manipulation, is affected by the individual's self-esteem represents an interesting link between individual differences and choice. It supports the idea that such factors may affect intermediate stages in the decision making process and thus only indirectly affect choice.

On the negative side, the manipulation we used was not successful in affecting either consumers' initial or final reference points, suggesting that this task is not as easy as it might appear. In this instance, it appears that the manufacturer's suggested retail price functions as an upper bound on the initial reference point for consumers, but factors affecting the lower limit remain to be determined. The results of the exploratory probability issue are also surprising. It seems that, in complex narratives, individuals assess their own probabilities, regardless of whether they are told the probability is "a chance" or "50-50". This finding suggests that Thaler's (1985) deterministic approach in applying prospect theory to consumer choice may be the appropriate route to follow. At this point we can only say that additional research into consumers' abilities to teal with probabilistic descriptions of uncertainty is warranted.

CONCLUSION

We began this paper by asking if consumers' reference points affected their choices, and the results suggest that they do. Furthermore, we report empirical support for an individual difference variable, self-esteem, indirectly affecting choice in a predictable manner. We were unsuccessful, however, in our efforts to experimentally manipulate consumers' reference points, and thus we did not learn as much about the decision framing process as we had hoped to learn. While single study, small sample results must be treated with caution, we are nonetheless optimistic that future research into the decision framing process is warranted and holds much potential for adding to our basic knowledge of consumer judgment and choice.

REFERENCES

Bar-Hillel, Maya (1973), On the Subjective Probability of Compound Events", Organizational Behavior and Human Performance, 9, 396-406.

Bettman, James R. (1979), An Information Processing Theory of Consumer Choice, Reading, Mass.: Addison-Wesley.

Eagly (1967), Revised Janis-Field Feelings of Inadequacy Scale, unpublished, original measure in Personality and Persuasion, C. Hovland and I. Janis (eds.), 1959, New Haven: Yale University Press.

Fischhoff, Baruch (1983), "Predicting Frames", Journal of Experimental Psychology: Learning, Memory, and Cognition, 9, 103-116.

Harrell, Frank (1986), "The Logist Procedure", in Robert P. Hastings (ed. ? The SAS Supplemental Library User's Guide, Version 5 Edition, Cary, N.C.: SAS Institute. Inc., 181-202.

Huber, Joel and Christopher P. Puto, (1985), working paper, The Graduate School of Business Administration, The University of Michigan, Ann Arbor, Michigan.

Kahneman, Daniel and Amos Tversky (1979), "Prospect Theory: An Analysis of Decision Under Risk , Econometrica, 47, 263-91.

Kahneman, Daniel and Amos Tversky (1982), "The Psychology of Preferences", Scientific American, January, 162-70.

Kahneman, Daniel and and Amos Tversky (1984), "Choices, Values, and Frames", American Psychologist, 39, 341-50.

Kassarjian, Harold (1971), "Personality and Consumer Behavior: A Review", Journal of Consumer Behavior, 8, 409-18.

Locke, Edwin, Lise Saari, Karyll Shaw, and Gary Latham, (1981), Goal Setting and Task Performance: 19691980", Psychological Bulletin, 90, 125-52.

Lopes, Lola (1983), "Some Thoughts on the Psychological Concept of Risk", Journal of Experimental Psychology: Human Perception and Performance,9, (February), 137-44.

Puto, Christopher P. (1986), "The Framing of Buying Decisions", working paper, The Graduate School of Business Administration, The University of Michigan, Ann Arbor, Michigan.

Schoemaker, Paul J.H. (1982), "The Expected Utility Motel: Its Variants, Purposes, Evidence and Limitations", Journal of Economic Literature, 20, 529-63.

Tversky, Amos and Daniel Kahneman (1981), "The Framing of Decisions and the Psychology of Choice", Science, 185, 1124-31.

Wyer. Robert S., Jr. (1974), Cognitive Organization and Change: An Information Processing Approach, Potomac, Maryland: Lawrence Erlbaum Associates.

----------------------------------------