College Benefit Segmentation Analysis: Approach and Results

Ronald Hoverstad, Texas Christian University
Charles W. Lamb, Jr., Texas Christian University
Patrick Miller, Texas Christian University
ABSTRACT - Market segmentation is an important topic to higher education researchers and administrators. The authors explored the relative importance that prospective students attach to various benefits offered by a medium-sized private university with moderately selective admissions criteria. The purposes of the study were to test, and provide data to illustrate, the efficacy of one approach to college benefit segmentation analysis.
[ to cite ]:
Ronald Hoverstad, Charles W. Lamb, Jr., and Patrick Miller (1989) ,"College Benefit Segmentation Analysis: Approach and Results", in NA - Advances in Consumer Research Volume 16, eds. Thomas K. Srull, Provo, UT : Association for Consumer Research, Pages: 332-338.

Advances in Consumer Research Volume 16, 1989      Pages 332-338


Ronald Hoverstad, Texas Christian University

Charles W. Lamb, Jr., Texas Christian University

Patrick Miller, Texas Christian University


Market segmentation is an important topic to higher education researchers and administrators. The authors explored the relative importance that prospective students attach to various benefits offered by a medium-sized private university with moderately selective admissions criteria. The purposes of the study were to test, and provide data to illustrate, the efficacy of one approach to college benefit segmentation analysis.


In a paper presented at a recent ACR conference, Chapman (1986) noted that college selection may be viewed as a process consisting of a sequence of interrelated stages. The five stages in this process are as follows: (1) pre-search behavior; (2) search behavior; (3) application decision; (4) choice decision; and (5) matriculation decision.

Pre-search behavior begins when a prospective college student first begins to consider the possibility of someday going to college. Pre-search may begin when a child is in elementary school and may last for several years. Previous research on pre-search has focused primarily on the "go - no go" decision.

The search behavior stage is characterized by active information gathering about alternative colleges and/or universities. Search activities include writing for college catalogs, examining reference books, talking with friends, counselors. parents, and other potential information sources, visiting campuses, and/or engaging in a variety of other search activities. Since college selection clearly entails high involvement decision making, extensive search behavior is normally expected.

The search behavior stage ends and the application decision stage begins when the prospective student decides to apply to a specific set of colleges for admission. This group of institutions then becomes the prospective student's evoked set.

The choice decision stage is the time between when he or she begins to receive admission notification and when the final selection or matriculation decision is made. In the choice decision stage, the prospective college student is presumed to possess relatively complete information on all relevant college attributes regarding all colleges in the choice set. Some further information search may take place regarding determinant attributes, but this is

expected to be limited.

The matriculation decision is the final step in the process. Sometimes it must be made with incomplete information about variables such as financial aid, housing options, or even admission decisions from some institutions in the evoked set. Since the decision is revokable with little or no penalty, the amount of anxiety involved in the decision may not be as great as is normally expected when consumers make high involvement decisions.

Purposes of the Study

The study reported here provides an empirical link between the choice decision and matriculation decision stages of Chapman's (1985) college search and choice behavior model. The purposes of the study were to test, and provide data to illustrate the efficacy of one approach to college benefit segmentation analysis.


Benefit Segmentation

In 1968, Haley stated that "The idea that all markets can be profitably segmented has now received almost as widespread acceptance as the marketing concept itself' (p. 30). Since that time, the concept and practice of market segmentation have slowly diffused into a number of application areas including higher education (Kotler and Andreasen 1987). For example, Goodnow (1980) found that adults attending the College of DuPage in Illinois fell into the following five benefit segments: (1) social improvement learners; (2) career learners; (3) leisure learners, (4) submissive learners; and (5) ambivalent learners. Based upon these findings, she recommended a separate marketing strategy for each benefit segment.

According to Haley (1968, p. 31), ". . . benefits sought by consumers determine their behavior much more accurately than do demographic characteristics or volume of consumption." He further notes that though most people would like as many benefits as possible, the relative importance they attach to individual benefits can differ importantly, and can be used as an effective lever in segmenting markets.

Ideally, a benefit segmentation study would entail empirically identifying determinant, rather than simply salient, product-market attributes. As Engel, Blackwell and Miniard (1986,p. 96) note, "Sometimes a salient evaluative criterion does not influence the evaluation process. This occurs when the alternative; under consideration perform equally well (or poorly) on this criterion." Alpert (1971) discussed the value of alternative methods of identifying determinant variables and their implications for solving various marketing research problems.

College Choice

Numerous studies have been conducted to identify the factors that influence college selection decisions (e.g. Chapman 1979; Chapman and Staelin 1982; Hossler and Gallagher 1987; Houston 1979; Kellaris and Kellaris 1988; Krone, Gilly, Zeithaml and Lamb 1981; Litten 1982; Manski and Wise 1983; Rodner and Miller 1975; Sternberg and Davis 1978; Stordahl 1970; and Tierney 1980). Most, though not all, of these studies employed cross-sectional and correlational analyses to examine various aspects of the school choice process including variables such as what attributes of a school influence the choice decision and what demographic and socioeconomic characteristics are linked to school choice behavior. A few researchers (e.g. Chapman 1986; Punj and Staelin 1978), have focused on how school choice can be explained as a process and whether models can be used to predict school choice behavior.

Chapman (1986) reported that

'The disportionate emphasis that researchers have placed on studying the college choice phase (of the college search and choice behavior process) is no doubt due to the relatively low costs with which such research efforts may be conducted. A one-shot post-choice retrospective survey is typically employed n college choice studies" (p. 249).

His assessment is that search behavior cannot be accurately studied using retrospective methodologies. 'The study of search behavior requires intervention during the search process" (Chapman 1986).


Data were obtained from a random sample of applicants for the freshman class entering a medium-sized private university with moderately selective admissions criteria. The population consisted of all admitted applicants to the freshman class of Fall 1986 (N = 2322). A random sample of 1600 accepted applicants was selected for potential respondents.

Mailing of the survey instrument began four months prior to matriculation. Thus, potential respondents received the questionnaire right after the traditional notification period for admission and financial aid. Most students were therefore still in the process of finalizing decisions; the choice decision stage in Chapman's model.

Each sample member received a mail questionnaire with enclosed postage paid return envelope, followed two days later by a reminder postcard. Questionnaires were numbered to allow a second follow-up mailing two weeks after the initial mailing for non-respondents. None of the mailed materials, including the return address, were traceable to any specific institution. Pocus group interviews after matriculation confirmed that students did not link the questionnaire with a specific school.

Of the 1600 admitted applicants who were mailed surveys, forty-seven had bad addresses, leaving 1553 in the sample. From the 1553 potential respondents, 733 returned usable questionnaires, yielding a response rate of forty-seven percent (47%).

Subjects were instructed to list their first three college choices in blank spaces provided and then to rate each school on the following 43 items which were derived from previous studies and focus group interviews held prior to the survey.

1. Size of school

2. Quality of college faculty

3. Attractiveness of campus


4. Distance from home

5. Student/faculty ratio

6. Location of campus


7. Campus visit

8. A specific department or major

9. Religious opportunities on campus


10. Quality of computer facilities

11. Friendliness

12. Admissions brochures, pamphlets, catalogues


13. Research orientation of faculty

14. Quality of student body

15. Accelerated programs (Advanced Placement. Honors)


16. General reputation

17. Quality of laboratories

18. Academically challenging


19. Fraternities and sororities

20. Quality of classroom facilities

21. Opportunity for personal contact with faculty


22. Variety of majors

23. Quality of library

24. Reputation of alumni


25. Strong graduate program

26. Availability of financial aid

27. Opportunity to attend cultural events


28. Cost of tuition, books, and fees

29. Living expenses

30. Athletic facilities (student recreation)


31. Good residence halls

32. Athletic events (inter-collegiate)

33. Information received at dinners or receptions


34. Religious affiliation of school

35. Where my friends are going

36. Private school preference


37. Parents' preference

38. Contacts with admission personnel

39. Undergraduate teaching emphasis of faculty


40. Social activities

41. Contacts with students

42. College information books (such as Peterson's)


43. Career opportunities after graduation


Subjects were instructed to rate each college on each item on a scale ranging from very negative influence (1), to neutral (3), to very positive influence (5).

At the end of the survey, subjects were asked to identify which college was their first choice. They were also asked to list in rank order the three most important items in their decision.


Since the focus of the study was those students who were considering enrolling in the subject institution, the first step was to eliminate all questionnaires that did not list the subject school as a first, second, or third choice. One hundred fifty five surveys were excluded from further analysis.

The study was designed to examine the relative (rather than absolute) importance that respondents attach to individual college benefits. Respondents' answers to each of the 43 questions regarding the subject institution were standardized to create measures of relative importance for each attribute. In other words, relative importance was measured by adjusting for within-subject variance.

Each respondent rated 43 attributes on a scale of 1 to 5. The total number of points a person could allocate ranged from 43 (assigning an importance rating of 1 to each attribute) to 215 (assigning an importance rating of 5 to each attribute). The mean attribute rating could therefore range from 1 to 5.

The relative importance of attribute i for an individual was computed by (1) calculating his or her mean attribute rating on all 43 attributes; (2) subtracting the mean attribute rating from the rating of attribute i; and (3) dividing that number by the standard deviation of the person's responses to the 43 attributes. Expressed mathematically, the relative importance of attribute i is

(Ii-X) / (SQRT [(1/43) (SUM (Ii - X)2)])


Ii is the individual's rating of the importance of the attribute i:

X is the mean importance rating of the 43 attributes.

This approach is similar to the method recently used by Muller (1985) to determine relative importance.

The standardized importance ratings of the 578 respondents were factor analyzed to avoid any redundancies in the data and to prepare it for subsequent cluster analysis (Morrison 1967). Scores were computed for each factor by averaging the standardized ratings of the key attributes related to each factor (those that loaded + .4 or more). Ratings for items that loaded negatively were reverse coded before averaging. The factor analysis results are reported in Table 1.

Because of the difficulties involved in choosing an optimal cluster solution, the following procedure was used. First, the sample was split randomly in half and separate K-means cluster analyses were performed on each half with the five factor index scores as input. Because K-means cluster analysis is quite sensitive to outliers, an iterative process was used in the cluster analysis. The final cluster centers were saved from the first cluster analysis and used as the starting points for a second cluster analysis. The final cluster centers from the second analysis were saved as input for the third analysis, and so on. The cluster memberships from each stage of the above analysis were cross-tabulated with the cluster memberships from the previous stage until less than one percent of the cases changed from one cluster to another. This iterative process was used on each split-half of the data for the 2, 3, 4, 5, and 6 cluster solutions.

Cluster solutions 2 through 6 were examined to identify solutions that were similar for each half. Discriminant analysis also was performed on each cluster solution for each half. The resultant discriminant functions were applied to the other half of the data to determine which discriminant function was most accurate in duplicating cluster membership.

The split-half analysis showed the five cluster solution to be 96.07 percent accurate and the two cluster solution to be 97.86-percent accurate. The three, four and six cluster solutions were less accurate than the two and five cluster solutions. The split-half solutions for the two cluster solution were markedly different, but the clusters for the two split-halves of the five cluster solution were quite similar. Therefore, the five cluster solution was chosen for further analysis.

A "hit rate" was then computed for each cluster. That is, each cluster contains some respondents who subsequently enrolled in the subject institution and some who did not. The cases in each cluster were compared to enrollment data to identify which respondents enrolled at the subject institution. The number of enrolled students in each cluster was divided by the total number of respondents in that cluster to determine the hit rate. These results offer important insight regarding past and future student recruitment strategies.


After completing the cluster analyses, the halves of the five cluster solution were combined and another five solution cluster analysis was performed. Table 2 shows how the clusters compare in terms of the five factors and the relative importance attached to each factor.

Describing the Clusters

Cluster one contained 113 respondents (19.5%), who are primarily attracted to the university by specific programs. The religious affiliation of the university is also important. Academic excellence, financial considerations and a small school atmosphere are also considerations. In short, cluster one describes a segment of the college applicant market that considers everything except social activities when selecting a university.

Cluster two was the largest cluster with 141 respondents (24.3%). This cluster is attracted by academic excellence. While some other considerations may have a minor impact on the college choice decision, this segment looks almost exclusively for an academically challenging university.







Cluster three contained 107 respondents (18.5%). This cluster focuses on specific academic programs and academic excellence. Social activities were the third most important factor for this group. Cluster four contained 132 respondents (22.8%) who focus primarily on religious affiliation. They may look for a university affiliated with their own church, or may look to avoid universities affiliated with certain other religious groups. This cluster also looks to a lesser extent at a small school and financial considerations.

The final cluster contained 85 respondents (14.7%) and rates financial considerations as most important. This is the smallest cluster of the five, and may consist of people who will attend the school which they can afford.

Cluster "Hit Rates"

For each cluster, the sample respondents were compared to enrollment data at the subject university. The hit rate in the cluster reflects the number of applicants in each cluster who subsequently enrolled at the subject university. Table 3 shows the hit rates in each cluster.

Overall, 232 (40%) of the 578 sample respondents chose to attend the subject university. The hit rates in each of the five clusters were all very close to 40%, indicating that the likelihood of attending the subject university did not vary among the clusters. This is not surprising since the subject university's current admissions effort includes information about all aspects of the university and is directed equally at all acceptable applicants. With a concentrated effort by the university toward a particular cluster, the hit rate would be expected to change.

It is important to note that, in the current study, it is possible to calculate the hit rates. In most segmentation studies conducted in marketing, segments are defined without any follow-up to determine the effectiveness of the marketing effort in each segment. The nature of this study provides a unique opportunity to monitor not only the hit rate in each segment, but also the academic progress of students in each segment as their academic careers progress.

The calculation of a hit rate is only possible for the subject university. While information regarding the relative importance of various attributes was collected for several universities (each subject rated three universities), access to the enrollment records of other schools was not possible. Therefore, all analyses in this study concern only the subject university.


The current study has important implications for university enrollment management. First, the methodology employed demonstrates a way of identifying benefit segments among prospective students. University admissions personnel probably have an intuitive understanding that prospective students differ widely in the specific benefits they wish to receive from a college education. This methodology clearly identifies what those segments are and how large they are.

Once a prospective applicant has been identified as a member of a particular segment (cluster), the admissions effort can be directed at the specific benefits sought by that segment. A more concentrated effort in a particular segment would yield a higher hit rate due to the more efficient admissions effort. Universities using this procedure can do a better job of attracting the type of students they wish to have enroll.

A second implication of this study concerns admissions policies. Now that different benefit segments of applicants can be identified, university administrators can decide whether some segments are more attractive than others. For example, some universities emphasize overall academic excellence, and others stress a specialization in certain programs like business or engineering. Still other universities may wish to be all things to all people. Universities wishing to promote academic excellence would concentrate their efforts on attracting segment two. Some larger universities may wish to appeal to segment number one. Many universities with a religious orientation may wish to attract segment number four.

The advantage to this procedure is that university officials can establish priorities among different benefit segments, now that they have the means to pursue those segments more efficiently than before. Universities can revise their recruiting strategies based on administrative priorities for certain segments. The effectiveness of this recruiting effort for each segment can be evaluated.


There are several directions for future research. First, the academic progress of those who enrolled at the subject university within each segment can be monitored. Such things as retention, grade point average at various stages of their academic career, their majors, and whether they graduate can be used to ascertain which segments are the most attractive from the university's perspective. This information, used along with test scores and other traditional sources of admission information, is valuable in the long run in establishing recruiting priorities among the segments.

A second direction for future research is to compare analysis results from year to year to determine the stability of the segments. While some shifting among the segments can be expected over time, confidence in this procedure increases if the results of the analysis are replicated in the future. This gives university officials more confidence in their strategies to pursue certain segments.

A third possibility is to evaluate the effectiveness of recruiting strategies by monitoring the hit rates in the various segments in future years. As previously mentioned, recruiting strategies could be designed to attract applicants within particular segments. If this is done, it is necessary to follow through to see if these strategies actually do attract more of the desired applicants to campus.

A fourth possibility is to follow up with the applicants who do not enroll at the subject university. There may be several reasons why this occurred. For example, they could have selected a different university, they could have chosen not to attend any university, or they could have experienced financial problems which forced them to go elsewhere. Knowing why the subject university is losing accepted applicants is important in the enrollment management process.


Alpert, Mark I. (1971), "Identification of Determinant Attributes: Comparison of Methods," Journal of Marketing Research, 8 (May), 184-191.

Chapman, Randall G. (1979), "Pricing Policy and the College Choice Process," Research in Higher Education, 10, (1), 37-57.

Chapman, Randall G. (1986), "Toward a Theory of College Selection: A Model of College Search and Choice Behavior," in Richard J. Lutz, ed., Advances in Consumer Research, vol. 13, Provo, Utah, Association for Consumer Research, 246-250.

Chapman, Randall G. and Richard Staelin (1982), "Exploiting Rank Ordered Choice Set Data Within the Stochastic Utility Model," Journal of Marketing Research, 19 (August), 288-301.

Engel, James F., Roger D. Blackwell, and Paul W. Miniard (1986), Consumer Behavior, New York: The Dryden Press.

Goodnow, Wilma E. (1980), "Benefit Segmentation: A Technique for Developing Program and Promotional Strategies for Adults in a Community College," unpublished Ph.D. dissertation, Northern Illinois University, DeKalb, Illinois.

Haley, Russell I. (1968), "Benefit Segmentation: A Decision Oriented Research Tool," Journal of Marketing 32 (July), 30-35.

Hossler, D., and Gallagher, K. S. (1987), "Studying Student College Choice: A Three-Phase Model and Implications for Policymakers, College and University (Spring), 207-221.

Houston, Michael J. (1979), "Cognitive Structure and Information Search Patterns of Prospective Graduate Business Students," in Jerry Olsen, ed., Advances in Consumer Research, vol. 7, Provo, Utah, Association for Consumer Research.

Kotler, Philip and Alan R. Andreasen (1987), Strategic Marketing for Nonprofit Organizations, 3rd edition, Englewood Cliffs, New Jersey: Prentice Hall, Inc.

Kellaris, James J. and William K. Kellaris, Jr. (1988), "An Exploration of the Factors Influencing Students' College Choice Decision at a Small Private College," College and University (Winter), 187-197.

Krone, Flavia, Mary Gilly, Valarie Zeithaml, and Charles W. Lamb, Jr. (1981), "Factors Influencing the Graduate Business School Choice Decision," in Kenneth Bernhardt et al. eds., The Changing Marketing Environment: New Theories and Applications, Proceedings of the 1981 American Marketing Association, 453-456.

Litten, Larry H. (1982), "Different Strokes in the Applicant Pool: Some Refinements in a Model of Student College Choice," Journal of Higher Education, 53-54, 383402.

Manski, Charles F. and David A. Wise (1983), College Choice in America, Cambridge, Massachusetts: Harvard University Press.

Morrison, Donald G. (1967), "Measurement Problems in Cluster Analysis," Management Science, 13 (August), 755-780.

Muller, Thomas E. (1985), "Structural Information Factors Which Stimulate the Use of Nutritional Information: A Field Experiment," Journal of Marketing Research, 22 (May), 143-57.

Punj, Girish N. and Richard Staelin (1978), 'The Choice Process for Graduate Business Schools," Journal of Marketing Research, 15 (November), 588-598.

Radner, Roy and Leonard S. Miller (1975), Demand and Supply in US Higher Education, New York: McGraw-Hill.

Sternberg, Robert J. and Jeanne C. Davis (1978), "Student Perceptions of Yale and its Competitors," College and University (Spring), 262-279.

Stordahl, Kalmer J. (1970), "Student Perceptions of Influences on College Choice," Journal of Educational Research, Vol. 63, No. 5 (January), 209-212.

Tierney, Michael L. (1980), 'The Impact of Financial Aid on Student Demand for Public/Private Education," Journal of Higher Education, 51 (September/October), 527-545.