Attracting Potential Switchers to Mass Transit: Mode Choice As a Multi-Attribute Decision Model

ABSTRACT - Policy makers and transportation planners frequently have to make decisions regarding the characteristics to incorporate into new or revised mass transit systems. To ensure consumer patronage, it is necessary to identify the trade-offs which are most likely to attract potential consumers. This paper uses conjoint analysis to identify such trade-offs and presents results and policy implications of using these procedures in one medium-sized American city.


Linda L. Golden, John F. Betak, and Mark I. Alpert (1979) ,"Attracting Potential Switchers to Mass Transit: Mode Choice As a Multi-Attribute Decision Model", in NA - Advances in Consumer Research Volume 06, eds. William L. Wilkie, Ann Abor, MI : Association for Consumer Research, Pages: 519-525.

Advances in Consumer Research Volume 6, 1979      Pages 519-525


Linda L. Golden, The University of Texas at Austin

John F. Betak, The University of Texas at Austin

Mark I. Alpert, The University of Texas at Austin

[This study was financed by the U.S. Department of Transportation under contract DOT-OS-30093 to the Council for Advanced Transportation Studies, University of Texas.]


Policy makers and transportation planners frequently have to make decisions regarding the characteristics to incorporate into new or revised mass transit systems. To ensure consumer patronage, it is necessary to identify the trade-offs which are most likely to attract potential consumers. This paper uses conjoint analysis to identify such trade-offs and presents results and policy implications of using these procedures in one medium-sized American city.


A frequent objective of transportation planners is to attract persons from low density modes (private automobile) to high density modes (mass transit). Given that the planner has some flexibility in determining the system attributes or characteristics, and given a desire to attract patrons of the system beyond those which might be considered a captive market, it becomes important to look to the potential consumer.

Due to limited resources, the planner and policy maker may not be able to incorporate all the characteristics into the system that the consumer finds necessary. However, just as the planner and policy maker must accept trade-offs, the consumer is also willing to accept trade-offs. And, with acceptable levels of certain attributes, although the consumer may not consider the system optimum, the consumer may patronize the system when he or she would not have otherwise. In a less than perfect world, it then becomes important to identify the trade-offs among system characteristics that are most likely to attract the consumer. With knowledge of these trade-offs, and knowledge of how the public views current characteristics of the automobile and mass transit, policy and planning decisions may be facilitated.

The focus of this paper is upon identifying trade-offs among levels of system attributes for potential users of mass transit and the resulting policy implications. The problems are couched in terms of a mode choice situation whereby the consumer has mode choice options and makes mode choice decisions on the basis of characteristics of alternative modes of transportation. In addition, this paper uses a methodology, conjoint analysis, which may be of considerable use to transportation planners. As Green and Srinivasan (1978) state,. "Although conjoint analysis has been extensively applied mainly in the private sector, it has a large potential for public sector applications as well". While it is not new to stress the need to identify system attributes relevant to consumer mode choice decisions in an attempt to model and promote public transportation usage (Blattberg and Strivers 1970; Golob, Dobson and Sheth 1973; Hill and Von Cube 1963; Hill and Martin 1967; Mundy, Cravens and Woodruff 1974; Sheth 1975; Tybout, Hauser and Koppleman 1978), this new application of conjoint analysis may facilitate a further understanding of consumers' transportation decisions:, hence, more efficient system design and potentially broadened consumer acceptance.

Mode Choice Problem

Mode choice is manifest by the behavior of individuals and/or groups of individuals. This is predominately a purposive, adaptive behavior. As such, mode choice behavior represents the outcome of a complex decision process which encompasses: (1) trade-offs between system characteristics and non-system characteristics, including user requirements and attributes; (2) past decisions with respect to mode choice, origins and destinations, life style, etc.; and (3) future decisions with respect to origins and destinations, life style, etc., i.e., goal-directed decisions.

It is the view of this study that mode choice decisions are one part of a large decision-making system in which each part of the system affects, and is affected by, the other decision components. A decision-making system may be conceptualized as having three main parts: "(1) a set of decision variables, (2) interrelationships among the decision variables, and (3) criterion functions depending on values of the decision variables" (Whinston, 1966). Given this orientation, this paper characterizes mode choice as a multiple criteria decision task. And, the problem of determining utilities in mode choice is considered to be one of determining part-worths of the mode's attributes at different levels.

Conceptualizing mode choice as a multiple attribute decision problem in which the individual chooses among alternatives which are described in terms of their attributes, it is possible to formalize the mode choice task as a multi-attribute utility model with the following essential features:

u(s) = c1u1(x1) + c2u2(x2) + ... + cnun(xn)


x = (x1, x2, ..., xn).

A vector that specifies a consequence with n attributes, and xi specifies the ith attribute.

(x) = Set of possible consequences. The probability distribution of x depends on the decision maker's action choice.

u(.) = Utility function defined on (x).

Ui(.) = Utility function defined on (xi).

ci = Scaling or weighting factor.

This is the so-called additive utility model which is discussed in some detail by Willie and Pessemier (1973). Using this model as normative framework for mode choice, it is argued that the individual will choose the transportation alternative that maximizes expected utility.

Given this specification of a model of mode choice decision-making, it now remains to determine a framework for obtaining utility values for a set of mode choice attributes. Conjoint analysis is a technique which decomposes the respondents' overall evaluations into separate, and compatible utility scales by which original global judgments can be reconstructed. This contrasts with the compositional approach of expectancy-value models. Conjoint procedures require only rank-ordered output, yet yield interval scaled output. In the case of finite data, the scale is technically an ordered metric, but as the number of input values increases a unique representation at the interval scale level is approached (Green and Rao 1971).

Thus, conjoint measurement is concerned with the joint effect of two or more independent variables on the ordering of a dependent variable. For example, one's preference for various modes of transportation may depend upon the joint influence of such variables as cost, travel time, convenience, dependability, privacy and so on. Algorithms and applications of conjoint analysis have been discussed in the literature (Carroll 1969; Coombes, Dawes and Tversky 1970; Green and Carmone and Wind 1972; Green and Rao 1971: Johnson 1974; Krantz 1964; Kruskal 1965; Luce and Tukey 1964; Srinivasan and Shocker 1973; Tversky 1967; Westwood, Lunn and Beazley 1974; Young 1969), and the purpose of this paper is not to discuss the mathematics of the methodology, but to investigate its usefulness in a modal choice situation. Green and Wind (1975) provide a discussion of the technique, for those unfamiliar with it, and Green and Srinivasan (1978) provide an excellent paper on issues and outlook for consumer behavior.


In order to identify those persons who were most likely to switch from the automobile to mass transit and the attributes to be used in the trade-offs analysis, a pilot study was conducted. The pilot study consisted of 252 usable personal interviews obtained from a stratified by census-tract (quotas proportional to population) area random sample of households in a medium-sized southwestern city. The purpose of this study was to identify determinant attributes (Myers and Alpert, 1968) for mass transit and demographic characteristics of potential switchers to mass transit.

Selection of Attributes

The results of the pilot study indicated that 11 attributes of the 27 tested were determinant (both perceived as important and varying across modes, in this case the bus and car). These were: dependability, low energy use per passenger, economy, low pollution per passenger, convenience, flexibility, freedom from repairs, freedom from accidents, no parking problems, brief travel time, and safe from dangerous people.

On the basis of the above results and extensive literature reviews (Betak and Betak 1969; Davies and Alpert 1975; Alpert, Golden, Betak and Story 1977), nine attributes of transportation modes were selected for trade-off evaluation: cost per mile, fuel use per passenger, transportation available--hours per day, travel time is--minutes, possibility of encountering dangerous people, level of comfort, opportunity to socialize, and transportation available--days per week. With the exception of level of comfort and opportunity to socialize, the attributes used for the trade-off analysis were determinant. These non-determinant attributes were reviewed in the literature and may be operationalized by policy makers to improve the system.

To operationalize some of the determinant attributes it was useful to redefine them in terms which could be related to observable phenomena. For example, the attributes of dependability, flexibility, and convenience were operationally defined in this study to mean "transportation available ________ hours per day" and "transportation available ______ days per week." To operationally define the attributes of economy and energy," cost per mile" and "fuel use per passenger'' were utilized. To operationalize the attribute of brief travel time, "total travel time is minutes" was used. No attempt was made to provide operational definitions of comfort, safety from dangerous people or socializing.

Each attribute was conceptualized as a three-level variable. In the case of "cost per mile," the levels were defined as being present cost, 154 less than present cost, and 154 more than present cost. To assist the respondents in calculating their present cost, estimates of typical current operating costs of an automobile or a bus ride were provided in the introduction of the questionnaire. The attribute of "level of pollution per passenger" was defined as low, medium and high. The levels of "transportation available ________ days per week" were defined as Monday through Friday or five, Monday through Saturday or six, and Monday through Sunday or seven. The levels of "transportation available hours per day" were defined as twelve, eighteen, and twenty-four. "Total travel time is _______ minutes" was defined as fifteen, thirty and sixty minutes (Harman 1974; Redding 1970). "The possibility of encountering dangerous people" was defined as never, sometimes, and often. The attribute "level of comfort" was defined as having three levels of low, medium, and high. The attribute of "opportunity to socialize" was defined as having three levels of never, sometimes, and often. "Fuel use per passenger" had three levels of low, medium, and high.

Sample of Potential Switchers

The pilot study identified potential switchers as those who do not now ride the bus, but said they would definitely ride if it were improved. Ten percent of the random sample answered definitely yes to this question, and thus, were identified as potential switchers. These people tend to be relatively younger, have smaller households, are most likely to be full or part-time students, are more likely to work or shop downtown, and tend to be higher educated then non-potential switchers. Potential switchers did not differ from the aggregate sample by income, occupation, number of cars or race.

On the basis of this information, eleven areas of the city were identified from census tract data as having a high proportion of potential switchers who live within one-quarter mile of mass transit routes. An enumeration of households in these areas was obtained from Cole's Directory. Computer-generated random numbers were used to identify every nth person to be included in the sample frame. Three hundred and seventy-five potential respondents were mailed personal letters asking for their participation in the study which was followed by a phone call from the interviewer establishing a date, time and place for the interview. Of the 201 people contacted by phone, 60 agreed to participate, resulting in 48 usable personal interviews.

Final Instrumentation

The nine attributes and their levels were developed as pairwise matrices. The thirty-six matrices, each with three levels by three levels trade-offs for each pair of attributes, requires a respondent to make 324 rankings (36 times 9 rankings per matrix). To facilitate this task, illustrative graphics were utilized throughout the matrices on the final questionnaire to characterize the attributes and their levels. In a pilot study, it had been shown that these graphics increased respondents' speed of completing the rankings without altering the relative utilities obtained for levels of the attributes rated (Alpert, Golden. Betak, Story and Davies 1977). Three separate pre-tests were used to refine the instrument and test the usefulness of graphic and verbal presentation of attribute levels.

The final instrument provided a set of instructions, "warm-up" matrices, trade-off matrices, attribute level perceptions for bus and private car, demographics and questions concerning current ridership patterns. Respondents were asked to respond in the frame of reference of trips to work or school. (Trade-off matrices elicited information about the importance of attribute levels for a transportation made in general, not for a specific mode). Interview time averaged forty-five minutes.

A parallel study (using the same sampling procedures) was undertaken, formatting the trade-offs into an alternative method for collecting data, in which unique combinations of attributes and levels are placed on separate cards, each representing a possible combination that might be available. The potential advantage of the "card-sort" approach is that it illustrates the "Gestalt" and has respondents make choices involving wholes rather than pairs of attributes and levels (Green and Wind 1975). Once the data are obtained, it is possible to analyze card-sort data in the same manner as matrix data, since an orthogonal array is used to generate equivalent combinations of levels of attributes to be trade-offs against other combinations, yielding similar choice problems for levels of attributes (Johnson 1973 and 1974). Using an orthogonal array substantially reduces the number of combinations that are needed, in this case, from 39, or 19,683 combinations to 27 combinations that are minimally necessary to secure trade-offs equivalent to those obtained in the matrix approach, also assuming no interactions among attribute levels.

However, even with this limited amount of cards, the 50 respondents involved in the "card sort" procedure were faced with an information-overload situation. Measures of the ability of the algorithm to capture the utilities sufficient to reproduce accurately the rankings given showed the card-sort procedure to be substantially less reliable than the matrix one, FOR THIS NUMBER OF ATTRIBUTES AND LEVELS (Alpert, Golden, Betak, Story and Davies 1977). Accordingly, we present results from the 60 matrix method respondents whose demographics correspond closely to those of previously identified potential switchers.


Data analysis evolved through several stages. The first stage investigated the validity of the matrix data. The second stage involved determination of utility curves and linear equations for the trade-off data, and the third state of analysis provided information from which policy statements could be made concerning attribute trade-offs for the car and the bus.

Validity Measures

Two types of analysis were used to evaluate the validity of the utility model mode choice proposed. The first analysis used M (theta) to evaluate the goodness of fit of the data by examining the relationship between the input rank order of the data and the obtained rank order as derived from the trade-off algorithm. The algorithm used is the non-metric regression analysis developed by Johnson (1973). The lack of fit measure is q, where:


For pairwise, two attribute trade-offs, 8 will be zero if the dij have the desired rank order, and unity if their order is perfectly reversed (Johnson 1973).

Table 1 presents 8 values for selected control groups. Eight categories of controls are used. In the first category 8's from all respondents were analyzed. In the second category respondents were grouped into three classes on the basis of a post-interview evaluation of their seriousness and level of effort in completing the instrument. The post-interview evaluation was done by a non-interview team using the remarks of the interviewers written on each instrument. In this second category, the first three quality levels of the respondents were grouped together and their data submitted for analysis. The remaining controls were for sex, age, and satisfaction levels, all using the first three quality levels of respondents. The satisfaction category is limited to those respondents who are very satisfied with their present mode of transportation.

As can be seen from Table 1, the 8 values ranged from .327 to .142. The O's for the respondents are relatively low, indicating that the derived weights for the attributes are reasonably consistent with the input rank order data. In short, it is possible to interpret the rank ordering of the attributes of the respondents with some degree of surety that these weights are a meaningful representation of the part-worths of the attributes investigated.



To consider further the issue of the validity of the results obtained, the consistency of the rank order of the attributes obtained was compared with the results of previous research. This issue was considered by comparing the range of weights obtained for each attribute, and the rank order of the attributes for all of the respondents. The difference between the weights (utilities) for the high and low levels of an attribute indicate how sensitive that attribute is to level changes, hence, how much influence it may have in determining choice (see Table 2). A large range indicates that variation in the amount of the attribute available in a mode will significantly affect the utility of that mode in a choice situation. The attribute ranges were: Possibility of Encountering Dangerous People, 1.365; Energy Use, 1.234; Pollution, 1.228; Total Travel Time, 1.210; Cost, 1.066; Available Days/Week, .945; Available Hours/Days, .848; Level of Comfort, .670; Opportunity to Socialize, .506.

The data indicate that the rank order of the attributes derived from the trade-off questionnaire were generally consistent with other research (Betak 1969; Davies and Alpert 1975; Alpert, Golden, Betak, Story, Davies 1977). Some differences did occur and were expected given the nature of the task confronting the respondents, the sample and the particular set of variables used.

It is clear from the thetas and consistency with other research that meaningful interpretations of the trade-offs may be made. Subsequent discussion focuses on these analyses, and are restricted to the 48 respondents contained in the group of quality 1, 2, 3 interviews.

Attribute Level Utilities (Part-Worths)

The first form of analysis was a determination of the sample's utilities for each attribute in each pairwise trade-off. Due to the volume of data, Figure 1 illustrates a matrix presented to respondents and this form of analysis. In this figure, the sample's utilities for the attribute levels are indicated by the decimal values at the right and bottom of each matrix. The algorithm, as previously described, computes the joint additive utility for each attribute and level pair, and is indicated in the top part of the respective cells in the matrix. The utility of each combination is the sum of the two level's utilities in each matrix.



Table 2 provides a summary of the average utility or part-worth, given by the sample for each specified attribute level. The average was calculated as the arithmetic mean at all derived weights (as in Figure 1) for each level of an attribute, as determined through all possible trade-offs with all other attributes and levels.

As noted above, the rank order of the attribute determinances show an essentially similar pattern to those studies that have included similar variables. Time, convenience, and psychological safety variables tend to dominate over level of comfort and time of day availability, given current levels of these attributes on public transportation systems (vs. cars) as represented in the trade-off study. Caution should be taken to realize that these utility ranges are sensitive to the selected levels shown in Table 2; variations in ranges chosen might produce different rankings of relative determinance of attributes. Care was taken to select levels which typified the relevant range for current transportation modes, but the utility ranges are nevertheless unique to the trade-offs here.



Another general observation from Table 2 is that with one exception the part-worths are all relatively linear representations of increasing or decreasing utility, in intuitively appealing directions. For example, the utility of cost per mile decreased as cost increased, and the same was found for energy use per passenger, travel time in minutes, possibility of encountering dangerous people, and so forth. The reverse results were shown for service availability, in that utility increased as did availability. The linear equations which might prove useful for interpolating within these ranges, and making modal combinations with equivalent utilities yet differing costs to deliver, are described elsewhere (Alpert, Golden, Betak, Story, Davies, 1977).

The utilities for opportunity to socialize showed a relatively more interesting non-linearity, in the sense that utility peaked at the middle level, implying a "golden mean" for socializing (perhaps too much involves perceptions of crowding or obtrusive conversations). This non-linear relationship with modal satisfaction would be considerably more difficult to discern with the more conventional multiple regression approaches to utility estimation (Banks, 1950; Hughes, 1966).

Private Automobile versus Public Transportation

The final stage in the analyses concerned using the calculated utilities to assess how respondents viewed private automobile and public transportation at the time of the interview. Viewed in total, Table 3 has a great deal of information that can aid in policy decisions. Columns 1 and 2 provide perceived utilities for the respondents' perceptions of the images of public and private transportation features. Assuming rationality, there is no reason to expect public transportation to be chosen, since the private auto clearly has the highest overall utility, given the present level of system attributes. Thus, the remainder of Table 2 provides guides to policy makers with respect to focal points for making changes in the modes of transportation to obtain increased patronage for one mode or another (typically trying to increase public transportation patronage).

The third column of Table 3 shows the relative advantage or disadvantage, per attribute, of public transportation's image. Five attributes of the nine show negative gaps for public transportation. Column 4 shows that both modes are quite distant from the maximum utility levels obtainable, across all attributes, given the perceived (and actual) characteristics.

One way for public transit to gain on private autos would be to improve toward the maximum utility level for each attribute. Column 5 indicates the potential gains that could be obtained, utilizing a positive or "carrot" (vs. "stick") strategy of improving the perceived image of public transportation to the maximum utility level. This column should be balanced against the relative economic costs of moving perceptions from their current point to or toward the maximum utility level. Rather than lowering revenues by instituting lower fares, for example, it might be less costly, and more effective to decrease the perceived probability of encountering dangerous people on public transportation.



As this variable is in part a proxy for social class perceptions of riders (mainly by non-riders), part of the non-rider's fears can be countered by scheduling routes through "safe" neighborhoods. Additional strategies might involve providing between lighting at bus-stops, instituting dial-a-ride (essentially door-to-door) service, placing security personnel on buses, featuring persons in bus advertisements who are similar to the target market of potential switchers (and somewhat counter to their expectations), and so forth. Several of these means are relatively inexpensive, yet may succeed in raising the perceived utility of this feature for enough of the .781 potential gain that the current overall disadvantage is largely offset. Although dial-a-ride would be considerably more expensive than lighting and promotional activities, this would also serve to increase perceived comfort utility, as well as that for total travel time, both of which Table 2 shows have high potential gain. Further research is needed to measure the dollar costs of moving perceptions along these attribute levels, and the incremental ridership levels of closing the gap (or surpassing) private transportation's utility for commuting.

Columns 6 and 7 suggest alternative policy and forecasting scenarios for public transportation's relative utility versus private autos. If, for example, a tax were levied on auto use sufficient to increase the price per mile to the lowest level of utility (column 6), the relative utility of buses would increase by .540, which is the difference between auto's current perceived utility (0.17) and the minimum utility in the scale range (-.523). If achieving positive gains may be viewed as the "carrot" approach, then imposing penalties on auto usage may be viewed as the "stick". Many planners have suggested more emphasis on the latter. While planners may wish to incorporate mostly positive improvements in the public transportation modes, to the extent that increased costs are enacted or may be forecasted (due to the increasing fuel costs of private cars, for example), the conjoint utility data presented here may prove useful in evaluating the quantitative impact of projected changes in either mode's features. Increased traffic congestion in inner cities, coupled with bus lanes and express service, may be forecasted as leading to net utility gains for public transportation's travel time dimension, where it currently has its greatest perceived disadvantage. However, persons attempting to manipulate such relative travel time by bus lanes should be aware of the potential factors that partially led to the abortive bus lane preferential treatment experience in Los Angeles during 1976 (Tischer and Shea 1978). Taking into account the political and environmental factors, column 7 can be useful in indicating alternative, albeit negative, approaches to influencing (or forecasting) improvements in the relative utility of public vs. private transportation.

Clearly, there are several combinations of policies or environmental changes that may occur to change modal perceptions and/or perceived utility values for particular levels of attributes. Given relatively stable utilities for various modal attribute levels, Table 3 can prove useful in guiding policies for altering the relative advantage in favor of public transportation. In addition, it is recommended that repeated studies be done to check on changes in the perceived utilities of levels over time and changing conditions and social values. If we can assume that respondents would generally behave rationally and choose the most preferred alternative, and that public transportation were to gain in perceived utility for sizable numbers of respondents, significant shifts might be possible. If the appropriate promotional techniques were utilized to apprise the respondents of the alteration in the transportation system, and accounting for a certain amount of lag or inertia in choice behavior, an increasing utilization of public transportation would be expected for respondents having characteristics in common with the sample studied here.


Viewing modal choice decisions as part of a large decision-making system in which each part of the system affects and is affected by the other decision components a multi-attribute utility model can be used to describe the consumer's mode choice decision. Applying conjoint analysis to this framework, the researcher can identify utilities and trade-offs among system attributes for both mass transit and private transportation. This yields information which can assist the marketer and planner of public transportation in determining changes in the system most likely to increase ridership.

This study indicated that policies directed toward improving total travel time, service availability in hours per day and days per week, safety from dangerous people, and comfort would be most likely to improve the overall utility of public transportation. Further, ridership would be likely to increase if scheduling and headways were arranged to provide total travel time of 15 minutes, 24 hours per day, 6 days per week and at a cost of 15 cents more per mile than current cost. The market segment is willing to pay more for an improved combination of attributes designed to meet their needs.

If policies were directed to improving the utility for these attributes, or others with higher utility gains relative to costs of changes, and assuming that the respondents behaved rationally and chose the most preferred alternative, and that the characteristics of the automobile were not altered, then public transportation would have a perceived total utility higher than a private automobile. Alternatively, we have shown how changes imposed on auto characteristics by policy and/ or environmental conditions may similarly affect this relative utility. Assuming effective promotional techniques, a campaign to inform individuals like those in the sample about the improvements in the public transportation system should yield an increasing utilization of public transportation.

Limitations on the type of analysis performed in this study are in assessing whether the policy options having the greatest potential for altering choice behavior were feasible politically or economically. Furthermore, these types of analyses cannot indicate which combinations of the changes in transportation attributes would yield the most cost-effective option. Also, no information was obtained for attributes outside the range presented in this study. Thus, analyses beyond the scope of this research must be undertaken to fully utilize the results reported herein.

From the preceding results, several suggestions for future research appear germane. First, incremental changes in the attributes having the greatest potential for altering utilities should be implemented and monitored. Second, analytical models for evaluating the political and economic viability of alternative attribute combinations for transportation systems need to be developed, possibly linear programming models. Finally, in conjunction with the recommended study of incremental changes, further development should be undertaken of more parsimonious instrumentation for eliciting trade-off data from potential users of transportation services, minimizing the respondents' time investment, and reducing the computational costs of analyzing trade--off data.

In conclusion, the trade-off analyses developed in this study provide indications of the areas where policy may be most effective in increasing the relative utility of public transportation services. These findings provide at least a first handle on some of the policy levers that may be available to decision makers confronted with choosing alternative strategies for the provision of public transportation in their communities.


Alpert, M. I., Golden. L. L., Betak, J. F., Story, J., and Davies, C. S. (1977), An Evaluation of Promotional Tactics and Utility Measurement Methods for Public Transportation Systems, Research Report 39, Council for Advanced Transportation Studies, The University of Texas at Austin.

Banks, S. (1950), "The Relationship Between Preference and Purchase of Brands," Journal of Marketing, 15, 145-157.

Betak, J., and Betak, C. (1969), Urban Modal Choice: A Critical Review of the Role of Behavioral Decision Variables, Research Report, The Transportation Center, Northwestern University.

Blattberg, R. C., and Stivers, R. (1970), "A Statistical Evaluation of Transit Promotion," Journal of Marketing Research, 7, 293-299.

Carroll, J. D. (1969), "Categorical Conjoint Measurement," Ann Arbor, MI: Meeting of Mathematical Psychology.

Coombs, C. H., Dawes, R. M., Tversky, A. (1970), Mathematical Psychology, An Elementary Introduction. Englewood Cliffs, New Jersey: Prentice-Hall, Inc.

Davies, S., and Alpert, M. I., (1975), Modal Choice and The Value of Passenger Travel Time Literature: A Selective Bibliography, Research Report 22, Council for Advanced Transportation Studies, The University of Texas at Austin.

Golob, T., Dobson, R., and Sheth, J. N. (1973), "Perceived Attribute Importance in Public and Private Transportation,'' American Institute for Decision Sciences Proceedings, 5, 7-10.

Green, D. E., Carmone, F. J., and Wind, Y. (1972), "Subjective Evaluation Models and Conjoint Measurement," Behavioral Science, 17, 288-299.

Green, D. E., and Rao, V. R. (1971), "Conjoint Measurement for Quantifying Judgmental Data," Journal of Marketing Research, 8, 355-363.

Green, D. E., and Srinivasan, V. (1978), "Conjoint Analysis in Consumer Research: Issues and Outlook," Unpublished research paper.

Green, D. E., and Wind, Y. (1975), "New Way to Measure Consumers' Judgements" Harvard Business Review, 53, 107-117.

Harman, E. J. (1975), "A Behavioral Analysis of the Concepts Used in Housing Choice" Ph.D. Thesis, Department of Geography, McMaster University.

Hill, D. M., and Von Cube, H. G. (1963), "Development of a Model for Forecasting Travel Mode Choice in Urban Areas," Highway Research Record, 38, 78-96.

Hille, S. J., and Martin, T. K. (1967), "Consumer Preferences in Transportation," Highway Research Board Record, 42, 6-43.

Hughes, G. D. (1966), "Developing Marketing Strategy Through Multiple Regression," Journal of Marketing Research, 3, 412-415.

Johnson, R. M. (1973), "Pairwise Nonmetric Multidimensional Scaling," Psychometrika, 38, 11-18.

Johnson, R. M. (1974), "Trade-off Analysis of Consumer Values," Journals of Marketing Research, 11, 121-127.

Krantz, D. H. (1964), "Conjoint Measurement: The Luce-Tukey Axiomatization and Some Extensions," Journal of Mathematical Psychology, 78, 248-277.

Kruskal, J. B. (1965), "Analysis of Factorial Experiments by Estimating Monotone Transformations of the Data," Journals of the Royal Statistical Society, Series B, 27, 251-263.

Luce, R. D. and Tukey, J. W. (1964), "Simultaneous Conjoint Measurement: A New Type of Fundamental Measurement," Journal of Mathematical Psychology, 78, 1-27.

Mundy, R. A., Cravens, D. W., and Woodruff, R. B. (1974), "Potential for Marketing Management Applications in Public Transportation Planning," American Marketing Association Proceedings, 274-279.

Myers, J. H., and Alpert, M. I. (1968), "Determinant Buying Attitudes: Meaning and Measurement," Journal of Marketing, 32, 13-20.

Redding, M. J. (1970), "The Quality of Residential Environments: Preferences for Accessibility to Residential Opportunities," Unpublished Ph.D. dissertation, Department of Civil Engineering, Northwestern University, Evanston, Illinois.

Sheth, J. N. (1975), "A Psychological Model of Travel Mode Selection," Association for Consumer Research Proceedings, 2, 425-430.

Srinivasan, V., and Shocker, A. D. (1973), "Estimating the Weights for Multiple Attributes in a Composite Criterion Using Pairwise Judgments," Psychometrika, 38, 473-493.

Tischer, M .L., and Shea, C. G. (1978), "Attitude-Behavior Changes in a Before-After Mode Choice Situation," Association for Consumer Research Proceedings, 5, 456-464.

Tversky, A.(1967), "A General Theory of Polynomial Conjoint Measurement," Journal of Mathematical Psychology, 81, 1-20.

Tybout, A.M., Hauser, J. R., and Koppelman, F. S. (1978), "Consumer Oriented Transportation Planning: An Integrated Methodology for Modeling Consumer Perceptions, Preference and Behavior," Association for Consumer Research Proceedings, 5, 426-434.

Westwood, R., Lunn, T., and Beazley, D. (1974), "The Trade-off Model and Its Extensions," Journal of the Market Research Society, 16, 227-241.

Whinston, A. (1966), "Theoretical and Computational Problems in Organizational Decision-Making," in J. R. Lawrence (ed.) Operational Research and Social Sciences. London, New York: Tavistock Publishers, 191-207.

Wilkie, W. L., and Pessemier, E. A. (1973), "Issues in Marketing's Use of Multi-Attribute Attitude Models," Journal of Marketing Research, 10, 428-441.

Young, F. W. (1969), "Polynomial Conjoint Analysis of Similarities: Definitions for a Special Algorithm," Research Paper No. 76, Chapel Hill, N.C.: University of North Carolina, Psychometric Laboratory.



Linda L. Golden, The University of Texas at Austin
John F. Betak, The University of Texas at Austin
Mark I. Alpert, The University of Texas at Austin


NA - Advances in Consumer Research Volume 06 | 1979

Share Proceeding

Featured papers

See More


What Converts Webpage Visits into Crowdfunding Contributions: Assessing the Role of Circumstantial Information

Lucia Salmonson Guimarães Barros, Universidade Federal de Sao Paulo
César Zucco Jr, Brazilian School of Public and Business Administration, Brazil
Eduardo B. Andrade, FGV / EBAPE
Marcelo Salhab Brogliato, Brazilian School of Public and Business Administration, Brazil

Read More


How Do Consumers React to Anthropomorphized Brand Alliance? Applying Interpersonal Expectations to Business-to-Business Relationships

DONGJIN HE, Hong Kong Polytechic University
Fangyuan Chen, Hong Kong Polytechic University
Yuwei Jiang, Hong Kong Polytechic University

Read More


E9. “Power Distance, Social Aspiration, and Fair Trade Products” – the Interaction Effect of Power Distance Belief and Status Motivation on Fair Trade Product Consumption

Sunghee Jun, Seoul National University
Libby Youngjin Chun, Seoul National University
Kiwan Park, Seoul National University, 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.