The Framing of the Insurance Purchase Decision

ABSTRACT - Previous research (Hershey and Schoemaker 1980; Tversky and Kahneman 1981) has shown that choice is dependent upon the manner in which the decision is framed. If framed in the asset domain (as is assumed by Utility Theory), the purchase of insurance would seem to be a realistic alternative. On the other hand, if the insurance purchase is framed in the loss domain as suggested by Prospect Theory (Kahneman and Tversky 1979), then it would appear that the purchase would be much less likely. An extensive pilot study of the decision to purchase flood insurance finds some tentative support for the greater likelihood to purchase insurance when the respondent is using an asset decision frame.


Joshua L. Wiener, James W. Gentry, and Ronald K. Miller (1986) ,"The Framing of the Insurance Purchase Decision", in NA - Advances in Consumer Research Volume 13, eds. Richard J. Lutz, Provo, UT : Association for Consumer Research, Pages: 251-256.

Advances in Consumer Research Volume 13, 1986      Pages 251-256


Joshua L. Wiener, Oklahoma State University

James W. Gentry, Oklahoma State University

Ronald K. Miller, Oklahoma State University


Previous research (Hershey and Schoemaker 1980; Tversky and Kahneman 1981) has shown that choice is dependent upon the manner in which the decision is framed. If framed in the asset domain (as is assumed by Utility Theory), the purchase of insurance would seem to be a realistic alternative. On the other hand, if the insurance purchase is framed in the loss domain as suggested by Prospect Theory (Kahneman and Tversky 1979), then it would appear that the purchase would be much less likely. An extensive pilot study of the decision to purchase flood insurance finds some tentative support for the greater likelihood to purchase insurance when the respondent is using an asset decision frame.


An individual's behavior is influenced by the task encountered, the individual's characteristics, and their interaction (Punj and Stewart 1983). The influence of the task environment is so strong that Dawes (1975) has argued that many behavioral theories test well because they are actually models of the tasks which the subjects are performing.

One of the most powerful task environment elements is how the task is presented to the subject. The manner in which the task is presented can directly influence how an individual forms his or her decision frame (Tversky and Kahneman 1981). A decision frame is the individual's perception of the decision problem being faced. It includes the perceived alternatives, consequences derived from selecting an alternative, and likelihood of particular consequences occurring. The developers of the decision frame concept, Tversky and Kahneman (1981), emphasize that a single decision may be framed in multiple ways. In other words, how a communicator presents the objective information to an individual (i.e., frames the problem) can influence how the receiver perceives the problem (i.e., forms the decision frame).

Dawes' (1975) suggestion that behavioral models are task environment specific provides a method for the a priori identification of how decision formulation manipulations will influence behavior. If different behavioral models predict different responses to the same objective situation, then behavior can be influenced by encouraging the use of a particular model. The use of a particular model can be encouraged by presenting the task so that it better fits the model's decision frame.

To illustrate how the presentation format's influence upon behavior can be predicted by use of the decision frame construct, a study of decision making in an insurance purchase situation was conducted. Both economists and decision theorists have investigated the insurance purchase decision. Although there is considerable variation within each research framework, there are two fundamental differences which delineate the frameworks. Both expected utility theory and the decision theory frameworks, such as prospect theory (Kahneman & Tversky 1979), explicitly postulate different framings of the uncertainty decision, and make different predictions. In particular, they make opposite predictions about an individual's willingness to purchase flood insurance (Kunreuther 1978).

Decision theory frameworks postulate that an individual frames a decision in terms of the immediate gains or losses he or she may incur. For example, if one consequence of a bet is losing $17.00, an individual will frame this consequence as a $17.00 loss. On the other hand, economic theories generally postulate that an individual frames a decision in terms of changes in his or her assets. In other words, it is assumed that the immediate loss or gain is combined with the current level of assets to produce a net asset level. For example, if an individual's assets are worth $3,000, then the consequence of losing a $17.00 bet is framed as a reduction in wealth from $3,000 to $2.983.

Decision theory and economic frameworks make opposite predictions about the choice an individual will make when faced with the following decision. The decision is to choose between a situation with a small certain loss and a situation in which the two alternatives are a large loss or no gain. Economic theory predicts that most individuals will select the situation with the small certain loss. The decision theory framework predicts that most individuals will select the uncertain prospects over the certain loss. This prediction is made if the individual's aspiration level of wealth is his or her current level, and the loss is not ruinous (Payne, Laughhunn, and Crum 1980).

Prospect theory predicts that the uncertain prospect will be selected over the sure loss, unless the probability associated with the sure loss is very small. Prospect theory postulates that individuals make risky decisions by acting as if they used a two stage process. In the first stage they edit the problem. An outcome is viewed as either a gain or loss. A value is assigned to the outcome. The function that assigns values to outcomes is convex for losses and concave for gains. The function is steeper for losses. A second important feature of the editing process is that probabilities are translated into decision weights. Most decision weights are lower than their corresponding probabilities. The exception is that very low probabilities are associated with relatively higher weights (Kahneman & Tversky 1979). The majority of the experimental studies discussed in Hershey, Kunreuther, and Schoemaker (1982) support this conclusion.

The insurance purchase situation, stripped of its nonmonetary aspects, fits the structure of the problem discussed above. If insurance is purchased, t small certain loss (the premium) is incurred. If insurance is not purchased, then a situation exists which includes the likelihood of a large loss (the flood, fire, accident, or robbery) and the likelihood of no gain.

The analogy to the insurance decision provides a framework for evaluating the respective approaches. A number of recent studies of decision making under uncertainty have used a flood insurance framework (Slovic and Kunreuther 1974, Slovic et al. 1977, Kunreuther 1978). Flood insurance provides a particularly useful framework because some of the nonmonetary dimensions, such as moral hazard and enforced legal requirements to buy, are minimal. Moreover, the propensity of individuals to purchase flood insurance is a matter of major public concern (Taft 1972, Kunreuther 1978).

Flood Insurance

The most salient fact about flood insurance is that the vast majority of flood-plain homeowners in the United States do not own any flood insurance. Only one in four homes in 100 year flood plains are covered by flood insurance (Barton 1985). [Homes in a 100 year flood plain have a .01 likelihood of being flooded within a given year.] Although virtually no homes in 500 year flood plains have flood insurance, 40% of all flood damage (measured in dollars) occurs in these areas (Barton, 1985). The number of homeowners residing in flood-plains who are not insured for the peril of flood is startling. As White and Haas (1975, p. 255) pointed out, "Nearly every community in the nation has some kind of flood problem, chiefly resulting from inadequate drainage systems for runoff water produced by heavy rainfall from storms. For example, 97.5% (2483 out of 2547) of the communities in Pennsylvania have been identified as being flood prone (Luloff and Wilkinson 1979). The Federal Insurance Administration estimates that one of every ten Americans resides in locations where flooding is likely to occur (Kunruether 1978).

Survey research by Kunreuther et al. (1978) has identified three key reasons why individuals fail to purchase flood insurance. They may be unaware of it; they may miscomprehend it; or they may not want it. A critical finding of this study was that when the survey respondents' subjective assessments of the likelihood of a natural disaster, the monetary damage caused by such disasters, and premium rates were incorporated into expected utility analysis, between 30 and 40 percent of their insurance decisions violated the predictions of expected utility theory.

The finding that insurance purchase decisions violate the predictions of expected utility theory has been replicated in experimental situations (Slovic et al. 1974, 1977).


Expected utility theory, as originated by Bernoulli (1954), and axiomatized by Von Neumann and Morganstern (1947), predicts that most individuals who fully comprehend the flood insurance purchase decision will buy flood insurance. This prediction is due to the assumption that individuals are risk averse with respect to changes in their wealth. In contrast, the results of a series of experiments reviewed by Hershey et al. (1982) generally support the postulate that individuals are risk takers in the domain of losses. Slovic et al. (1977) argue that this postulate implies that individuals who fully comprehend the flood insurance purchase decision will not buy flood insurance unless the nonmonetary dimensions compensate for the monetary dimensions.

Numerous experiments investigating the validity of expected utility theory have been conducted since 1948. A common element of their design is that information is presented co subjects in the form of a win or lose gamble. A common finding is that individuals are risk preferring in the domain of large losses. The finding is integral to prospect theory (Kahneman and Tversky 1979) and supported by the findings of Slovic et al. (1977), Payne, Laughhun and Crum (1980), Hershey and Schoemaker (1980), and Hershey et al. (1982).

A key difference between the two approaches is that the direct consequences of the action are not integrated with the individual's existing wealth in the decision theory framework. Although the two equations are isomorphic (both can be derived from the same rationality axioms, Borch (1972)), they cay represent very different mental framings of the insurance decision. In the former case purchasing insurance is associated with a sure positive consequence, in the latter purchasing insurance is associated with a pure negative consequence. This sign difference is potentially significant since individuals appear to be risk averse in the positive domain of gains (Tversky and Kahneman 1979) but risk takers in the negative domain.

Both economists and decision theorists explicitly recognize that insurance provides more than simply monetary protection. there are both normative reasons for buying insurance, e.g., it is the socially right thing to do, and self interest reasons, e.g., it permits one to take less care. Studies by Hershey and Schoemaker (1980) and Hershey et al. (1982) support these arguments. They find that individuals are more likely to select the small certain loss alternative when it is labeled "buying insurance."


The alternative of purchasing flood insurance was presented to subjects in five forms. Two decision frames were used (asset and loss) and each frame was presented in two ways (tabular format and lottery format). Also, one form of the advertisement was used as a control group and did not present the decision in either a loss or asset frame. Thus the experiment involved a two x two with control group design. The subjects were college students from a variety of courses.

Advertisements. The students were asked to role play the following situation:

The year is 1990 and you have just purchased a $100,000 home (with an 80: mortgage) in a large midwestern city. You have received the following information in the mail.

All subjects were shown the advertising content shown in Figure 1. Those subjects in the control group were shown only the information in Figure 1.



The loss and asset frames were manipulated as shown in Figures 2 and 3. The loss frames contrasted the possible loss of a large amount of money with the sure loss of a smaller amount of money (the premium). The asset frames discussed a possible large reduction in assets as opposed to a sure small reduction in assets. The tabular format was developed because it is the most comprehensive means of showing what happens in all circumstances. The lottery format was used also because most previous work with decision frames has used gambling tasks expressed as lotteries (for example, Hershey and Schoemaker 1980; Tversky and Kahneman 1981).



Subjects. The use of student subjects presents serious problems, since few of them have faced the issue of whether to insure their home from flood damage. Consequently, the role playing scenario was required. Factors which made the scenario somewhat realistic were (1) a large number of the students planned to obtain employment after graduation in the city mentioned in the scenario, and (2) this city had suffered serious flood damage within the past year.



While college students are homogeneous in many ways (age, for example), we attempted to investigate diverse groups of students. The insurance classes were included, as it was expected that they would be most familiar (among student groups) with the content areas of the study. Similarly, undergraduate decision theory classes and an MBA class were included with the expectation that these students would have more expertise with the cognitive aspects involved in the decision making. At the other end of the spectrum, sophomore level business law students were included with the expectation that they had little understanding of the insurance decision in general and little training in systematic decision making processes. Finally, several senior-level marketing courses were surveyed, with the expectation that these students would fall somewhere in between the sophomores and the other students.

The students provided information on their (and their family's) experience with floods and with insurance purchases (car, life, health, and property). Also, personal information (age, family income, and sex) was obtained.

Manipulation Checks. Previous decision frame research manipulated the loss and asset perspectives, but did not use any manipulation checks. Given the different risk propensities found in the gambling scenarios, it can be inferred that the manipulations were successful. The translation of the positive and negative quadrant frames from gambling to applied decision-making situations is not straightforward. While the manipulations shown in Figures 2 and 3 may well be criticized, they are enormous improvements over the first advertising appeals we developed. Given the applied nature of the decision setting, we used manipulation checks (shown in Figure 4) to investigate the success of our attempts. Further, their use should provide insight into the perspective that subjects bring into the insurance purchase, as the control group received neither the loss nor asset manipulations.

Dependent Variables. The subjects responded to three items about their intentions to purchase flood insurance and to two items dealing with their beliefs about flood insurance. One intention measure was the subject's assessment of how likely (7 point scale) s/he would be to buy insurance for $388. The other four measures are also shown in Figure 4. The eight Likert-type items shown in that figure were ordered randomly in the survey completed by the subjects.




The profile of the 541 respondents is shown in Table 1. Only 8% of the sample had ever owned a home, indicating that the role-playing was somewhat irrelevant to most of the subjects. The vast majority of respondents were covered by some form of insurance, although many of them had it purchased for them by their parents. About 9% had experienced flood damage themselves and 17% had family members who had experienced flood damage.

Manipulation Checks. Unlike previous studies investigating the impact of decision frames, this study used manipulation checks to see if the treatments would affect one's perception of the decision frame. The results presented in Table 2 show that the asset manipulations were successful for one of the two checks. For that check, those who received the asset manipulation were more likely to agree that buying insurance either allows you to keep your current wealth or reduce your wealth slightly than were those in the control group. The control group was more likely to agree than those who received the loss manipulations. The loss treatments did not elicit more agreement with statements viewing insurance in the loss domain than did the asset treatments or the absence of a treatment.



Impact of Decision Frame on Intention and Attitudes. Table 2 also presents the mean responses for the intention to purchase and the attitude measures. One point to note is the highly positive mean responses to all five items (between 5.0 and 6.0 on a seven-point scale). Given the low acceptance of flood insurance among the general population as a whole (as discussed earlier), these results either indicate the wonders of advertising or that demand characteristics are present. The second explanation will be investigated in more detail later.

The decision frames showed no differential impact on intentions or attitudes. The presence of advertising information (whether framed in the asset or in the loss domain) resulted in higher intentions to purchase than did the control. [This was not found for the third intention measure (wanting to seek more information about flood insurance). However, this is to be expected since the control group's treatment presented the smallest amount of information.] It had been hypothesized that the asset decision frame would result in higher intentions to buy. There is a slight trend in this direction, but the differences are not significant at all (p > .25).

It may be that people are predisposed to view the insurance purchase situation in either a loss or an asset decisions frame. Further, it appears that our treatments did not manipulate that decision frame significantly (especially for losses). Consequently, we attempted to categorized subjects as to having a loss decision frame, an asset decision frame, or neither. In order to operationalize these constructs, we assumed that our manipulation checks are capturing the concepts adequately. Those agreeing with the loss checks more than the asset checks were classified as having a loss frame, and vice versa.



When the intention and attitude measures for these groups are investigated (as shown in Table 3), we find that the direction of the mean responses is consistent with our theory for all five measures. Further, the means are significantly different for two of the measures, marginally significant for two others, and non-significant for only the "investigate further" intention (which received the uniformly favorable response, possibly tue co demand characteristics).

Impact of Context on Intentions and Attitudes. Previous studies of decision frames have presented the alternatives in the form of lotteries. Since lotteries would seem to be more appropriate for gambling situations than for the purchase of insurance, it was believed that the use of a two x two table would present the possible consequences of the insurance purchase in a more systematic manner. While we can see no reason to expect differential effectiveness for the two modes of presentation; past research (Dawes 1975; Hershey and Schoemaker 1980) would suggest that context effects might occur. However, the mean responses to the tabular presentations were not different from the mean responses to the lottery presentations for any of the dependent variables.



Evaluation of the Nature of the Subjects. As this study represents a first attempt to investigate decision framing in the context of a specific purchase situation (as well as the first attempt to measure the decision frame through the manipulation checks), student subjects were used as a pilot study. As acknowledged earlier, students are far from ideal in this scenario (92% in this study have not experienced home ownership). In our defense, it can be noted that previous research on decision frames has used student subjects almost exclusively. This section will attempt to investigate the nature of this limitation.

Even though we must acknowledge that our student sample is largely homogeneous, we did attempt to vary the familiarity with insurance and with decision making processes by using subjects from a variety of classes. No differences in the intention and attitude responses were noted for the five class groupings, but differences were noted (p < .05) for two of the decision frame measures ("insurance is like thinking about losing money" and "insurance makes you think about how much you are worth'). Students in the insurance classes were the most likely to disagree with the former item and to agree with the latter item, while the MBAs were the most likely to have the opposite views.

The intention and attitude measures were related directly to the demographic, insurance-experience, and flood-experience variables using Chi square analysis. For ease of interpretation and because most previous decision framing studies (Hershey and Schoemaker 1980) had used a dichotomous dependent variable, we dichotomized the dependent variables by coding the three positive scale responses (for example, "extremely likely," "very likely," and "likely") as "yes" and the other responses as "no."

The majority of cross tabulations were not significantly related, but the pattern of significant findings is insightful. Previous family experience with flood damage was directly related to the intention to buy flood insurance (for the first two intention measures). Ownership of property and health insurance was directly related to the willingness to investigate flood insurance further.

The rest of the results point out the weaknesses of using a student sample. Females, those who do not buy their own car insurance, and younger respondents were more likely to intend to purchase and to hold positive attitudes toward flood insurance. Further, those who do not own their own homes were more willing to seek further information about flood insurance. These results indicate that it is likely that a survey of homeowners would result in much lower intentions to purchase. Whether the different decision frames have a differential impact on those intentions remains to be investigated.


This study represents an attempt to investigate the impact of decision framing on one's intention to purchase flood insurance. A case is made for the expectations that people viewing the purchase from an asset perspective will be more likely to buy than those viewing it from a loss perspective. Subjects were asked to role play a situation in which flood insurance might be a relevant alternative. They were presented with information, including a manipulation of the decision frame.

Contrary to the expectations, subjects receiving the asset frame did not express greater intention to purchase. However, further analysis did find that those viewing insurance in the asset domain expressed greater intentions than those viewing insurance as being in the loss domain.

The study suffered from several limitations. One, the use of student subjects, received much discussion. It seems clear that students are prone to the demand characteristic to say that they will purchase flood insurance. A study of homeowners will undoubtedly find a lower level of intentions being expressed. Whether the decision frames will have greater impact remains to be seen.

Recently, Peterson, Albaum, and Beltramini (1985) reported that the use of both student subjects (as opposed to nonstudents) and intentional (as opposed to behavioral) response variables decrease the size of treatment effects by 42% and 85% respectively (X2 being the effect size measure). By analogy, this study's use of students and intention measures may have contributed to the failure of the manipulations to have any significant effect.

Another limitation lies in the asset and loss manipulations. The failure to find significant manipulation checks indicates the need to develop better treatments, especially for the loss decision frame. There appears to be little difference in the effectiveness of a lottery or tabular presentation of the information.

Part of the problem with the insignificant manipulation checks may be due to the construction of the manipulation checks themselves. As noted earlier, previous studies have not attempted to measure subjects' decision frames. One reason may be that this is less than a straightforward process. Clearly improvement is needed in the manipulation checks used.

Thus our conclusion is that there is sufficient evidence to support a recommendation that more research in this area is needed.


Barton, Bill (1985), Manager of Region VI of the National Flood Insurance Program, personal communication with the authors, February 7, 1985.

Bernoulli, Daniel (1954), "Expositions of a New Theory on the Measurement of Risk, Econometrica, 22 (January), 23-26 (translated by Louise Sommer from the original work written in 1738).

Borch, Karl (1972), The Economics of Uncertainty, Princeton, NJ: Princeton University Press.

Dawes, Robyn M. (1975), "The Mind, and the Task," in Cognitive Theory, Vol. 1, Frank Restle et al., eds., Hillsdale. NJ: Lawrence Erlbaum, 119-139.

Hershey, John C., Howard C. Kunreuther, and Paul J. H. Schoemaker (1982), "Sources of Bias in Assessment Procedures for Utility Functions, Management Science, 28 (No. 8, August), 936-954.

Hershey, John C. and Paul J. H. Schoemaker (1980), "Risk Taking and Problem Context in the Domain of Losses: An Expected Utility Analysis," Journal of Risk and Insurance, 47 (No. 1, March), 395-418.

Kahneman, Daniel and Amos Tversky (1979), "Prospect Theory: An Analysis of Decision Under Risk," Econometrica, 47 (No. 2, March), 263-291.

Kunreuther, Howard (1978), Disaster Insurance Protection: Public Policy Lessons, New York: John Wiley & Sons.

Luloff, A. E. and Kenneth P. Wilkinson (1979), "Participation in the National Flood Insurance Program: A Study of Community Activeness," Rural Sociology, 44 (No. 1), 137-152.

Payne, John W., Dan J. Laughhunn, and Roy Crum (1980), "Translation of Gambles and Aspiration Level Effects in Risky Choice Behavior," Management Science, 26 (No. 10. October). 1039-1060.

Peterson, Robert, Gerald Albaum, and Richard Beltramini (1985), "A Meta-Analysis of Effect Sizes in Consumer Behavior Experiments," Journal of Consumer Research, 12 (June 1985), 97-103.

Punj, Grish N., and Davit W. Stewart (1983),-"An Interaction Framework of Consumer Decision Making," Journal of Consumer Research, 10 (No. 2, September), 181-196.

Slovic, Paul, Baruch Fischhoff, Sarah Lichtenstein, Bernard Corrigan and Barbara Combs (1977), "Preference for Insuring Against Probable Small Losses: Insurance Implications," Journal of Risk and Insurance, 44 (June, No. 2), 237-258.

Slovic, Paul, Howard Kunreuther and G. F. White (1974), "Decision Process Rationality and Adjustment to Natural Hazards," in Natural Hazards: Local, National and Global, G. F. White, ed., New York: Oxford University Press.

Tversky, Amos, and Daniel Kahneman (1981), "The Framing of Decisions and the Psychology of Choice," Science, 211, 453-458.

Von Neumann, John and Oskar Morgenstern (1947), Theory of Games and Economic Behavior, Princeton, NU: Princeton University Press.

White, G. and J. E. Haas (1975), Assessment of Research on Natural Hazards Cambridge: MIT Press.



Joshua L. Wiener, Oklahoma State University
James W. Gentry, Oklahoma State University
Ronald K. Miller, Oklahoma State University


NA - Advances in Consumer Research Volume 13 | 1986

Share Proceeding

Featured papers

See More


Tuition Myopia: Temporal Discounting Induces a Myopic Focus on the Costs of Higher Education

Haewon Yoon, Indiana University, USA
Yang Yang, University of Florida, USA
Carey K. Morewedge, Boston University, USA

Read More


Identity Threats, Compensatory Consumption and Working Memory Capacity: When and Why Feeling Threatened Leads to Heightened Evaluations of Identity-Relevant Products

Read More


Worse is Bad: Asymmetric Inferences on Items and Assortments From Logically Equivalent Comparisons

Yoel Inbar, University of Toronto, Canada
Ellen Evers, University of California Berkeley, USA

Read More

Engage with Us

Becoming an Association for Consumer Research member is simple. Membership in ACR is relatively inexpensive, but brings significant benefits to its members.