This research employs factor and multiple regression analyses to develop the Personal Automobile Policy Shopping Intentions Model (PAPSI). Extant research is combined with the results of nine nominal grouping sessions to produce 82 independent variables for this investigation of their relationships with the near-term shopping intentions of standard personal automobile (PAP) insureds. It is concluded that a some of PAP insureds' variance in shopping intentions can be explained by the variables that entered the PAPSI equation.

Private passenger automobile insurance is the single largest form of property and liability insurance. In 1991, the property and liability insurance industry generated $79 billion of personal automobile insurance premiums, which represented over 35 percent of the total property and liability premium volume in this country [Insurance Information Institute, 1993]. The Personal Automobile Policy (PAP) is "... the coverage form now most commonly in use" [Wood, Lilly, Malecki, Graves & Rosenbloom, 1989, p. 65] to cover personal automobile loss exposures.

Surprisingly, a thirty-year review of the literature shows that little is known empirically about the variables involved in insureds' decision processes with respect to their behavioral intention to shop or not shop for a PAP. In fact, only one published study [AIRAC, 1985] can be found that investigates PAP shopping behavior. No known property and liability studies provide an in-depth investigation of consumer attitudes toward agent service, insurer service, and PAP shopping behavior.

Clearly, PAP insurers would benefit from any information that could improve the effectiveness of their total marketing effort. The model developed in this research identifies a set of attitudes, characteristics, and shopping behaviors of insureds with respect to their behavioral intention to shop for a PAP in the near future. This information, in addition to traditional underwriting criteria, can be used by insurers to segment, target, and promote to their markets with greater precision, resulting in both successful product positioning and reduced resource waste.

As Ajzen and Fishbein [1980, p. 223] suggest, "before we can provide guidelines for the formulation of persuasive communications that will be effective in changing behavior, we must have an understanding of the factors that determine behavior." Thus, the present research examines 82 attitudinal, behavioral, and demographic variables that may influence or be associated with PAP insureds' shopping intentions.

Specifically, the purpose of this research was the development of an empirically-based model that identifies significant variables which explain some of the variance in PAP insureds' shopping intentions. Thus, a multivariate investigation of insureds' temporal behavioral intention to shop for a PAP within the next twelve months was conducted. The formal objectives were:

1. To develop conceptually, and validate empirically, a multivariate model of the PAP shopping process, utilizing demographic, normative, cognitive, behavioral, and attitudinal variables and constructs. These descriptors and constructs are posited to predict and explain the criterion variable -- the behavioral intention to shop for a PAP within 12 months.

2. To evaluate the model in terms of the statistical strength of the relationships and the reliability of the constructs in the model.

In the following section the theoretical basis for the study is introduced. This is followed by a discussion of the questionnaire development process and data gathering methods. Data analysis and conclusions are then presented, followed by suggestions for future research.

Fishbein's theory of reasoned action was the theoretical basis for this research. The Theory of Reasoned Action [Ajzen & Fishbein, 1980] recommends the determination of a target population's primary beliefs that are functionally related to the behavior in question. The theory defines primary beliefs as an individual's beliefs about salient outcomes or consequences, and about referent others. The salient outcomes or consequences are determined by measuring the individual's belief that performing the behavior will result in a certain consequence and the individual's evaluation of that consequence. The subjective norm (referent others) is determined by measuring the normative belief that a reference group or person thinks that the individual should or should not perform the behavior and the individual's motivation to comply with the influence of the reference group or person. Behavioral intention (reasoned action) is determined by adding the attitude toward performing the behavior and the subjective norm. These primary beliefs are then used to construct a model of behavioral intentions that can be empirically tested.

All behavioral intentions model components were measured in strict conformance with Fishbein's conceptualization. First, the behavioral intention must be under the respondent's volitional control and must be time- and situation-specific [Ajzen & Fishbein, 1980; Fishbein & Ajzen, 1975; Ryan & Bonfield, 1975]. Since the decision to shop/not shop for a PAP is made by someone within the family, behavioral intention is certainly within each insured's control. In addition, the PAP-shopping process is time- and situation-specific. If the insured overtly shops at all, s/he begins the process by developing an intention to shop.

Fishbein and Ajzen [1975] posits that only five to nine outcomes will be salient, and that they should be elicited in a free association situation. They suggest that saliency refers to the notion that when an individual is conscious of an outcome, salient outcomes have a high likelihood of being developed in such a session. In the present research, the development of lists of relevant consequences and referent others was accomplished in a similar manner -- through nominal grouping sessions (NGS).

A combination of extant research and the results of the nine nominal grouping sessions provided enough information to determine a priori research hypotheses -- one for each determinant variable and the PAPSI model. The general expression of each null and alternate hypothesis is:

Ho: There is no statistically significant relationship between insureds' (each predictor variable) and their time-specific behavioral intention to shop among insurers for a PAP within 12 months.

Ha: There is a statistically significant relationship between insureds' (each predictor variable) and their time-specific behavioral intention to shop among insurers for a PAP within 12 months.

Ajzen and Fishbein's [1980] composite variables (A-act, SN, and multiplied pairs) proved to be inappropriate for this research due to low reliabilities of scales (Cronbach's alpha statistics ranged from .59 to .74). Thus, factor analysis of all scaled variables was used to produce new constructs with higher reliabilities for use in multiple regression analysis.

After testing for statistically significant relationships with the criterion variable, the data were then used to test for the predictive capability of the developed model. The following research hypothesis addresses the concurrent validity of the Personal Automobile Policy Shopping Intentions (PAPSI) model:

Ho: There is no statistically significant predictive relationship between the model's variables/ constructs (the predictor variables) and insureds' behavioral intention to shop among insurers for a PAP within 12 months.

Ha: There is a statistically significant predictive relationship between the model's variables/ constructs and insureds' behavioral intention to shop among insurers for a PAP within 12 months.

Nine nominal grouping sessions (NGS) were conducted to ascertain the primary (salient) beliefs of the target population [Kraft, Ross & Paolillo, 1979], which is standard PAP insureds in the United States. Two insurers participated in the study. Both insurers are a major, national, multi-line property and liability insurers; one using the independent agency system channel of distribution (agency-writer) and the other using the direct writer channel (direct-writer). These large PAP insurers write over 1,300,000 standard PAPs in 47 states. The agency-system insurer drew eight independent systematic random samples by policy numbers for six nominal grouping sessions, the questionnaire testing session, and the national mail survey. The direct-writer drew five independent simple random samples by policy numbers for three nominal grouping sessions, the questionnaire testing session, and the national mail survey.

The primary objective of the nine sessions was to determine the considerations that would be important to insureds in their decision to shop/not shop for a PAP within 12 months. The results of these sessions were merged with extant research to create a developmental questionnaire, which then was tested by an independent group of randomly selected insureds of each insurer.

After the first pretest by a sample of the agency-writer's insureds, the questionnaire was redesigned and submitted to scrutiny by a group of ten university business professors to refine the questionnaire. The resultant questionnaire was used in a second pretest by a sample of the direct-writer's insureds. The final questionnaire was used in a national mail survey of the two companies' insureds to obtain the data for this study.

The nine nominal grouping sessions were composed of randomly selected PAP insureds of the sponsoring insurers, and were conducted from December of 1991 through February of 1992 to ascertain insureds' primary beliefs about PAP insurers, agents, policies, and the consequences of shopping for a PAP. It is believed that the size and diversity of the insurers, the use of nine grouping sessions in eight different states, and the use of simple/systematic random samples increased the likelihood that session members possessed a broad spectrum of demographics, beliefs, and behaviors. It also is believed that this design enhanced the discovery process as well as the studies's validity.

The nominal grouping sessions for each insurer produced almost identical results across sessions. Participants in each session listed their consensus top four positive and negative consequences of PAP shopping and significant others. The respondents in the three direct-writer sessions did not produce a consequence or significant other that was not listed in the six agency-writer sessions.

Two randomly selected nominal groups, one from each insurer, pretested the developmental questionnaire. This pretest aided in the elimination of improper or confusing wording and enhanced the reliability of the scales and the study. The present study scaled all attitudinal items on 7-point, bipolar scales which ranged from -3 for strongly disagree to +3 for strongly agree [Ajzen & Fishbein, 1980; Ryan & Bonfield, 1975]. To capture the subjective likelihood of salient outcomes (consequences of PAP shopping), the belief statements were worded in the form If I shop for automobile insurance among several insurers during the next 12 months, I may.... The criterion variable -- the behavioral intention to shop among insurers for a PAP in the next 12 months -- was scaled in the same way to keep all scaled data on the same measurement level. Also, all of the belief statements concerning agent service, insurer service, and policy knowledge were scaled in the same manner as the Fishbein-based attitudinal measures. Even though the questionnaire included 83 items, brevity was maximized in questions, statements, and instructions because parsimony of the instrument and the time required to complete the questionnaire was a concern since the study utilized a mailed, self-administered questionnaire.

To accommodate multivariate analysis, every effort was made to enhance the measurement level of the raw data. All attitudinal statements were developed with the objective of achieving interval level measurements. Although researchers disagree on whether or not Likert scales generate interval level data [Crask & Fox, 1987; Dowling & Midgley, 1991; Traylor, 1983], the present research assumes the described method of preparing statements and scaling responses to those statements produced interval level, metric data for all scaled statements, including the criterion variable. As Tull and Hawkins [Tull & Hawkins, p. 308] suggest, "most researchers treat the data from such [Likert] scales as if they were equal interval in nature since the results of most standard statistical techniques are not affected greatly by small deviations from the interval requirement."

The mail survey closely followed the procedures recommended by Dillman [1978]. The final questionnaire, a cover letter, and a postage paid business-reply envelope were mailed to 2038 and 2050 standard PAP insureds of the agency-writer and direct-writer, respectively. The subjects were provided a small incentive for returning usable surveys. Table 1 presents the response rates for each survey wave by insurer, by survey wave, and for the total sample. The usable response rate was 23.0 percent for the total sample. This response rate was judged adequate since Kress [1988, p. 84] notes that "it is not unusual for a national survey mailed to a general group of households to receive only a 5 to 10 percent response," and since it was observed that this rate is higher than most of the published national surveys researched for the present study.

variables were Chi-square tested for independence as was the categorical criterion variable. The responses from the two waves may be statistically related to the wave that produced them on only 2 and 3 of 77 categorical variables for the agency-writer and direct-writer, respectively.

These results suggest the absence of nonresponse bias with respect to the 77 categorical variables, especially since the tests did not signal a problem with an important demographic variable. Of the 82 predictor variables and the single criterion variable tested for nonresponse bias, only 1 predictor variable exhibited a statistically significant difference -- Spouse's Education Level (SEDUCA). However, this finding causes little concern about possible nonresponse bias for two reasons. First, the possibility exists that chance observations could account for the existence of significant differences in a small number of variables. Second, this variable should be given a low level of importance in comparison with most of the other demographic variables investigated, especially those of the respondent. This second assertion is based on the fact that the responses investigated in the present study are those of the respondent, rather than his or her spouse. It was concluded that this analysis does not suggest that the two waves were substantially different, which suggests that the presence of nonresponse bias in this survey is unlikely. Thus, the two survey waves for each insurer were pooled for further analysis.

In contrast, responses by the insureds of the two insurers may be different. However, no statistical differences were found. Thus, the two sets of data were pooled into one national sample for further testing.

The null hypotheses that each predictor variable does not significantly relate to the criterion variable, the behavioral intention to shop among insurers for a PAP within 12 months (SHOPONEY), were then tested. Table 2 presents all hypothesis test results produced by correlational and ANOVA procedures. Correlational analysis attempts to identify the strength of relationships between each metric predictor variable and the criterion variable. Correlational analysis was conducted to facilitate testing the null hypotheses that each metric predictor variable does not relate to the criterion variable. A one-way analysis of variance was used to assess the significance of relationships between each nonmetric measure and the criterion measure.

Variables Tested Coef.b Prob.

First time PAP buyer last year .1548 .694d

Switched PAP agent or insurer last year 10.1228 .002dc 

Agent initiated PAP quotation last year 16.5137 .001dc 

Respondent solicited agent for PAP quotation last year 42.1919 .001dc

Respondent sought family/friend PAP advice last year 28.4502 .001dc

Respondent compared PAP with advertisements last year 10.1336 .002dc

Number of years since last shopped for a PAP -.2140 .001c 

Major reason why not shopped among insurers more recently .1904 .001c

How chose to buy current PAP -.0024 .941

Last shopped among insurers through one or more than one agent .0200 .791d

Most important PAP information source -.0590 .404

Major reason began the last shopping process -.1080 .005c 

Perceived maximum price difference between PAPs .0683 .042c

Automobile insurance decision maker .0084 .797

Respondent's age -.1069 .001c

Respondent's sex 3.4938 .062d

Respondent's race 1.6603 .198d

Respondent's marital status .3100 .578d

Respondent's personal annual income -.0289 .392

Respondent's occupation .0002 .994

Respondent's education level -.0906 .006c

Spouse's age -.0839 .025c

Spouse's occupation -.0344 .364

Spouse's education level -.1074 .004c

Household annual income -.0459 .180

Number of male licensed drivers in household .0659 .047c 

Number of female licensed drivers in household -.0471 .158

Age of youngest licensed driver in household -.0646 .056 

Sex of youngest licensed driver in household 1.2657 .261d 

Number of insured vehicles in household .0007 .983

Number of uninsured vehicles in household .0326 .860d 

Attitudes toward most agents' willingness to provide quotations -.0876 .007c

Attitudes toward most agents' willingness to answer questions -.0832 .011c

Attitudes toward most agents' coverage recommendations -.0804 .014c

Attitudes toward most agents' willingness to help settle claim disputes -.0614 .060

Attitudes toward most agents' aggressiveness in soliciting other

lines of insurance -.0484 .139

Attitudes toward most agents' service deliveries -.0544 .095

Attitudes toward current agent's PAP shopping activities -.0761 .020c

Attitudes toward current agent's willingness to answer questions -.2279 .001c

Attitudes toward current agent's coverage recommendations -.2442 .001c

Attitudes toward current agent's willingness to help settle claim disputes -.2345 .001c

Attitudes toward current agent's aggressiveness in soliciting

other lines of insurance -.1043 .001c

Attitudes toward current agent's service deliveries -.2772 .001c

Attitudes toward the financial soundness of most insurers -.0859 .008c

Attitudes toward most insurers' fairness in handling claims -.0048 .884

Attitudes toward most insurers' willingness to answer questions -.0295 .367

Attitudes toward most insurers' PAP pricing -.1371 .001c 

Attitudes toward most insurers' service deliveries -.0276 .398

Attitudes toward current insurer's financial soundness -.1154 .001c

Attitudes toward current insurer's fairness in handling claims -.2065 .001c

Attitudes toward current insurer's willingness to answer questions -.2394 .001c

Attitudes toward current insurer's PAP pricing -.2087 .001c

Attitudes toward current insurer's service deliveries -.2413 .001c

Attitudes toward sameness in PAP coverages across insurers -.0541 .097

Attitudes toward state regulation of PAP pricing -.1330 .001c

Attitudes toward state regulation of PAP coverages -.0884 .007c

Attitudes toward the complexity level of the PAP -.0101 .757

Attitudes toward the quality of available PAP information .0679 .038c

Attitudes toward the value of using personal time PAP shopping -.3165 .001c

Importance of using personal time for PAP shopping -.2699 .001c

Attitudes toward finding a better PAP agent by shopping .3570 .001c

Importance of finding a better PAP agent .4052 .001c

Attitudes toward acquiring an inferior PAP insurer by shopping -.1310 .001c

Importance of acquiring an inferior insurer -.0454 .164 

Attitudes toward learning about the PAP through shopping .2554 .001c

Importance of learning about the PAP .3152 .001c

Attitudes toward the potential unpleasantness of the

PAP shopping experience -.0731 .025c

Importance of the potential unpleasantness of the

PAP shopping experience -.2032 .001c

Attitudes toward finding a better PAP value than currently have .3271 .001c

Importance of finding a better PAP value than currently have .3919 .001c

Attitudes toward the potential for after-shopping solicitations

by agents over time -.1281 .001c

Importance of the after-shopping solicitations by agents -.1290 .001c

Belief in the ability of current agent to influence PAP decisions -.1994 .001c

Motivations to comply with the advice of current agent -.2456 .001c

Belief in the ability of family members to influence PAP decisions .0212 .519

Motivation to comply with the advice of family members .0132 .687

Belief in the ability of current insurer to influence PAP decisions -.0819 .012c

Motivation to comply with the advice of current insurer -.1006 .002c

Belief in the ability of interested organizations to influence PAP decisions .1094 .001c

Motivation to comply with the advice of interested organizations .1103 .001c

Belief in the ability of friends and coworkers to influence PAP decisions .1640 .001c

Motivation to comply with the advice of friends and coworkers .1064 .001c

There is no predictive relationship between the predictor variables/constructs of the PAPSI model and insureds' behavioral intention to shop among insurers within the next 12 months (SHOPONEY). F=27.19975, p=.001c

a The criterion variable is SHOPONEY - Behavioral intention to shop among insurers for a PAP within 12 months.

b A Pearson correlation coefficient for metric variables unless designated an F-ratio for nonmetric variables.

c Significant at alpha =.05; Reject hypothesis of no relation with SHOPONEY.

d Coefficient is an F-ratio from a t-test of the nonmetric variable.

Two principal-components factor analyses were performed on the scaled predictor measures to create reliable measurement scales. These analyses facilitated substantial data reduction toward improved interpretation of the regression analysis which follows this section.

Varimax rotation produced the most interpretable factors and was used in both solutions. The resulting factors were assessed for reliability using Cronbach's alpha. Reliabilities of .70 are considered reliable since they indicate internal consistency, while alphas greater than .80 indicate very good reliability [Nunnally, 1978].

The first solution was developed from the 27 predictor variables conceptualized to measure agent service, insurer service, and policy knowledge. These variables were not included in a single factor analysis with the Fishbein-based attitude and social norm variables because, although scaled in the same manner, the two sets of scaled variables were constructed and worded in a different way. The most notable difference is that the Fishbein-based variables used two measures for each attitude and social norm measurement, while the service items consisted of a single measure of attitude. In addition, the Fishbein-based attitude items measured attitudes toward the consequences of PAP shopping, while the non-Fishbein variables measured attitudes toward the services provided by agents and insurers.

The factor analysis of the non-Fishbein scaled measures produced the seven-factor solution shown in Table 3. No variables were excluded from this solution for any reason. The first three factors exhibit very good reliabilities and were used subsequently in multiple regression analysis.

The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy for this first factor solution is .88171, which suggests a high degree of homogeneity of variables. Norusis [Norusis, 1988, p. B44] notes that Kaiser "...characterizes measures in the 0.90's as marvelous, in the 0.80's as meritorious, in the 0.70's as middling, in the 0.60's as mediocre, in the 0.50's as miserable, and below 0.50 as unacceptable." This meritorious result suggests the scale is appropriate for factor analysis. Further, the Bartlett's test of sphericity statistic is large (10934.320) and significant (.00001), thereby indicating linearity of correlations between sets of data. This result also indicates the appropriateness of factor analysis for this scale.

The high coefficient alphas of the first three factors shown in Table 3 demonstrate good internal reliability of the scales. Tukey's test for additivity tests the assumption that there is no multiplicative interaction among items in the scale. The low Tukey statistics indicate that transformation of the data is not needed to improve reliabilities which suggests that a linear scale is appropriate [Norusis, 1988]. The first three scales were judged to be of very good quality.

Factor 1 clearly refers to insureds' attitudes toward the services provided by their current agent. This construct was named Current Agent Service scale. Factor 2 refers to insureds' attitudes toward the services provided by agents and insurers other than their current agent and insurer. This construct was named Agent/Insurer Population Service scale. Factor 3 refers to insureds' attitudes toward their current insurer. This construct was named Current Insurer Service scale. Factors 4, 5, 6, and 7 were dropped from further analysis due to low reliabilities.

dimension of each factor were captured by each of the four scales. Finally, since the reliabilities of factors 8, 9, 10, and 11 were acceptable, these scales were judged to be of reasonable quality and were used in regression analysis. Factors 12, 13, and 14 were dropped from further analysis due to poor reliabilities.

Factor 8 refers to insureds' attitudes about acquiring value in PAP purchases. Presumably, their desire for learning about the PAP through shopping relates to a feeling that additional knowledge will facilitate the attainment of additional value. Further, insureds apparently include an agent's services as some portion of that value. This construct was named Value Consciousness scale. Factor 9 refers to insureds' propensity to follow the advice of insurance professionals. This construct was named Professional Influence scale. Factor 10 refers to insureds' apprehensions concerning the PAP shopping experience. This construct was named Shopping Anxiety scale. Factor 11 refers to insureds' propensity to follow the advice of interested organizations, friends, and coworkers. This construct was named Third-Party Influence scale.

The factor scores of the seven factors indicated in Tables 3 and 4 and the 24 significant predictor variables (including 14 Likert-scaled variables) from the correlation and analysis of variance tests that were not in those factors were then subjected to multiple regression analysis to develop the PAPSI model of prediction. The stepwise method was used to maintain parsimony of the model. A tolerance of .30 was employed to prevent variables that were multicollinear with the variables in the model from entering the equation. The PAPSI model was developed in 10 steps as 10 variables entered and none were removed until the 0.05 probability of F-to-enter triggered completion of the regression. The .30 tolerance did not come into play. Table 5 shows the results of multiple regression analysis.

Since all scaled predictor variables/factors and the criterion variable were measured using the same Likert scale, the absolute value of the coefficients of the scaled predictor variables/factors are comparable with respect to the order of their importance in the equation. Based on the size of the coefficients, the predictor variables/factors included in the PAPSI model are shown in Table 5 in order of their importance in predicting/explaining shopping intentions. The last four variables in the equation were not Likert scaled and are listed in descending order of their standardized regression coefficients (Beta). Table 2 shows the results of the hypothesis test of the PAPSI equation. The criterion variable is SHOPONEY -- Behavioral Intention To Shop Among Insurers For A PAP Within 12 Months.

Predictor Variables SHOPONEY 

Coefficient t-Statistic 

Constant .70478 1.597

Value Consciousness .72929 11.165

Shopping Anxiety -.30160 -4.486

Third-Party Influence .23431 3.722

Professional Influence -.19850 -2.991

Current Insurer Service -.13731 -2.057

Attitude Toward Agents' Post-Shopping Solicitations -.09355 -2.062

Respondent's Level Of Education -.07275 -2.898

Major Reason Why Not Shopped Around More Recently .15111 2.228

Number Of Years Since Last Shopped For A PAP -.05152 -2.101

Respondent Solicited Agent For PAP Quote Last Year -.24666 -1.789

Adjusted R-Square .34428

Durbin-Watson 2.01884

F Statistic 27.19975

The Shopping Anxiety construct exerts a negative influence on PAP shopping intentions. Thus, it may be concluded that the more insureds fear the shopping experience, the less likely they intend to shop in the near future.

The Third-Party Influence construct exerts a positive influence on PAP shopping intentions. This result suggests that insureds who tend to follow the PAP advice of friends, coworkers, and interested organizations are more likely to intend to shop in the near future.

The Professional Influence construct exerts a negative influence on PAP shopping intentions. Thus, it may be concluded that insureds who tend to follow the PAP advice of agents and insurers are less likely to intend to shop in the near future. The Current Insurer Service construct exerts a negative influence on PAP shopping intentions. This result suggests that insureds who think highly of their current insurer's service performances are less likely to intend to shop. Since it is unlikely that insureds accurately distinguish between agents' and insurers' services, it is possible that this reference to insurer service is actually a reference to perceptions of services provided by the agent-insurer team.

The Attitude Toward Agents' Post-Shopping Solicitations variable exerts a negative influence on PAP shopping intentions. This result suggests that the more insureds fear that shopping among insurers will generate subsequent sales calls from the agents involved in the shopping process, the less likely those insureds intend to shop in the near future.

Respondents' Level Of Education exerts a negative influence on PAP shopping intentions. This result may imply that insureds with more formal education are more confident in their previous PAP choice, which results in reduced intentions to shop.

The negative relationship of the Number Of Years Since Last Shopped with PAP shopping intentions suggests that the greater the time since insureds shopped around, the less likely they intend to shop in the near future. The positive relationship of the Major Reason Why Not Shopped Around More Recently with shopping intentions suggests that the more that insureds feel satisfied with their current agent or insurer (a low score), the less likely they intend to shop within the next 12 months (a low score on the criterion variable). Finally, the negative relationship of Respondent Solicited Agent For A PAP Quotation Last Year with PAP shopping intentions suggests that insureds who recently asked at least one agent for a PAP quotation (a low score) are more likely to shop again in the near future.

The regression results are intuitively pleasing and relate well to insurance sales experience. Policy holders are prone to shop when they want to find a better value, listen to the advice of friends, and engaged in previous shopping within the last year. The desire for better value is the greatest positive influence on shopping behavior. Combining these variables suggests that a search for better value supported by information from friends and the personal experience of obtaining recent price quotes motivates individuals to shop.

Policy holders are less inclined to shop when they fear the shopping experience, believe shopping will generate future sales calls, think highly of the present service, have high levels of satisfaction with present service, follow the advice of agents, have not shopped recently, and have a higher level of education. As with most services, satisfaction with the present service level and faith in the service provider lead to less interest in searching for a new insurance provider. Interestingly, fear of the results of the shopping experience influences the shopping decision. Those who perceive the shopping process to have high psychic costs (e.g. time, poor results, future sales calls) will probably not shop. As more time lapses between shopping activities, the less likely an individual will shop. This possibly relates to the previously mentioned issues of satisfaction and negative shopping outcomes, i.e. there continues to be no reason to shop.

The only surprising finding is the relationship between education and shopping. Consumer behaviorists would suggest that better educated individuals should be more efficient and effective shoppers and would tend to shop. However, the previous comment on policy holders' confidence in his/her past choice decision is an acceptable alternative explanation.

The general conclusion regarding the Personal Automobile Insurance Shopping Intention Model (PAPSI) is that it accurately represents the relationship between the predictor constructs/ variables that entered the equation and the criterion variable SHOPONEY. The obtained results support the formulation of the model, and the model demonstrates concurrent validity.

The adjusted r-square statistic (34.4%), however, suggests that most of the variance in the criterion variable was not explained by the predictor variables in the model. There are at least three possible explanations for this result. First, there may exist some important variables which were not investigated. Second, insureds' shopping intentions may include substantial randomness. If the PAP shopping decision process is some combination of these two possibilities, the only practical problem created by the low r-square statistics is a reduced ability to predict shopping intentions. However, all of these results can be used to influence insureds' attitudes concerning the PAP and services provided by the insurer and its agents. It just can not be said that the model can be used to accurately predict how much such efforts will alter shopping intentions. A third possibility must be considered. The PAP shopping decision may be one of low-involvement, while the Fishbein model and attitude investigations in general research high-involvement decisions. Involvement refers to a strong motivation which is reflected in a high perceived personal relevance of the product in a particular context. The degree of search employed by insureds may depend upon the level of consumer involvement with the PAP product and/or the decision process [Beatty & Smith, 1987]. The present research provides evidence that PAP shopping intentions may entail low-involvement decision making for some shoppers. 

For instance, insureds seem to accept moderately higher prices while avoiding shopping. This fits Assael's [Assael, 1987, p. 4] suggestion that "uninvolved consumers seek acceptable, not optimal products. They seek to minimize problems, not to maximize benefits." However, Assael also suggests that "low-involvement products are frequently purchased on the basis of price alone, since brand comparisons are unimportant and there are few differences between brands," and that most involved consumers consider more than price. The present research suggests that insureds do consider more than price in their PAP decisions and discovered some of those determinants. Considering the large annual cost of the PAP for most households, it is surprising that insureds do not consider those determinants more often. Perhaps their passivity results from their general perception of product sameness among insurers. If most insureds are uninvolved with the PAP, an insurer must create marketing programs to increase prospects' involvement in order to break the renewal habit and increase shopping activity.

As mentioned in the conclusions, future investigation should focus on the high and low involvement decision making process of insurance purchasers. Identifying the characteristics of consumers who are classified into each category and the variables important in their decision would be of value to researchers and practitioners.

The psychic costs of shopping for insurance needs additional investigation. The specifics of psychic costs and how to lower these costs to encourage insurance shopping would be of value to agents and insurers. How and why consumer perceptions of psychic costs are formed also need to be understood. The dimensions of customer satisfaction with an insurance provider must be further examined. While some may consider satisfaction dimensions to be transparent, additional research may uncover satisfaction aspects beyond the obvious.

AIRAC [All Industry Research Advisory Council] (1985), Patterns of Shopping Behavior in Auto Insurance, Oak Brook, IL.

Ajzen, I., & Fishbein, M. (1980), Understanding Attitudes and Predicting Social Behavior, Englewood Cliffs, NJ: Prentice Hall, Inc.

Armstrong, J. S., and Overton, T. S. (1977), "Estimating Nonresponse Bias in Mail Surveys," Journal of Marketing Research, Vol. 14, pp. 396-402.

Assael, H. (1987), Consumer Behavior and Marketing Action, 3rd ed., Boston, Mass: Kent Publishing Company.

Babbie, E. (1990), Survey Research Methods, 2nd ed., Belmont, California: Wadsworth Publishing Company.

Beatty, S. E. and Smith, S. M. (1987), "External Search Effort: An Investigation Across Several Product Categories," Journal of Consumer Research, Vol. 14 (June). pp. 83-95. 

Brightman, H. J. and Schneider, H. C.(1992), Statistics for Business Problem Solving, Cincinnati: South-Western Publishing Company.

Crask, M. R. and Fox, R. J. (1987), "An Exploration of the Internal Properties of Three Commonly Used Research Scales," Journal of the Market Research Society, (October), pp. 317-39.

Dillman, D. A. (1978), Mail and Telephone Surveys: The Total Design Method, New York: John Wiley.

Dowling, G. R. and Midgley, D. F.(1991), "Using Rank Values as an Interval Scale," Psychology and Marketing, (Spring), pp. 37-41.

Fishbein, M. and Ajzen, I. (1975), Beliefs, Attitude, Intention, and Behavior: An Introduction to Theory and Research, Reading, Mass.: Addison-Wesley.

Insurance Information Institute (1993), 1993 Property/Casualty Insurance Facts, New York: Insurance Information Institute, pp. 1-17.

Kraft, F. B., Ross, R. H., and Paolillo, J. G. (1979), "The Nominal Group Technique: An Alternative to the Focus Group Interview." In R. Franz, et al., (Eds.), Proceedings of The Southern Marketing Association, Lafayette, LA: Southern Marketing Assn., pp. 199-202.

Kress, G. (1988), Marketing Research, 3rd ed., Englewood Cliffs, N. J.: Prentice-Hall, p. 84.

Norusis, M. J. (1988), SPSS/PC+ Advanced Statistics V2.0, Chicago: SPSS, Inc.

Nunnally, J. (1978), Psychometric Theory, New York: McGraw-Hill Book Company.

Ryan, M. J., & Bonfield, E. H. (1975), "The Fishbein Extended Model and Consumer Behavior," Journal of Consumer Research, Vol. 2, pp. 118-36.

Traylor, M. (1983), "Ordinal and Interval Scaling," Journal of the Market Research Society (October), pp. 297-303.

Tull, D. S. and Hawkins, D. I. (1993), Marketing Research: Measurement and Method, 2nd ed., New York: Macmillan Publishing Co., Chapters 9 and 16.

Wood, G. L., Lilly III, C. C., Malecki, D. S., Graves E. E., and Rosenbloom, J. S. (1989), Personal Risk Management and Insurance: Volume I, 4th ed., Malvern, Pennsylvania: American Institute For Property And Liability Underwriters, Chapter 2.