Identifying Predictor Variables of Student Success in a Construction Management Program

Universities and colleges have continuously attempted to identify academic and non-academic variables that could best serve as indicators or predictors of student retention and success.  The main purpose of this investigation was to identify the variables that prove to be the best predictors of graduation in a construction management program.  Logistic regression was the statistical procedure used to make an association among the independent predictor variables (high school rank, high school GPA, high school class size, number of high school science courses, number of high school math courses, SAT composite score, matriculation age, gender, race, and residence) the dependent variable graduation.  A sample of approximately 400 students who originally enrolled in Purdue University’s Department of Building Construction Management program from the fall semester 1992 to the fall semester 1997 was used for this study.  This relationship was tested using logistic regression at the 0.05 level of confidence.  The results of the test show that two independent variables are significant; high school GPA and high school math semesters.

Key words:  Predictor variables, logistic regression, student retention, construction management students

Introduction 

The construction industry has always been a vast and complex industry.  Buildings are becoming more complicated to construct due to sophisticated technological demands.  For example, homes and offices are now being built with energy control management systems and state-of-the-art security, monitoring, fire systems, and time clocks that protect and track people as well as the property.  The construction industry requires an educated construction manager who has the skills and knowledge to manage the construction process.  The construction manager must have a thorough technical knowledge of the construction industry and the required managerial ability to ensure all phases of a construction job from the initial planning to final occupancy by the owner.  Each phase should be performed in such a manner as to result in a quality project (American Council for Construction Education [ACCE], 2000).   

Today there are approximately 100 universities and colleges throughout the United States that offer either four-year baccalaureate or two-year associate degree programs in construction, construction science, construction management, and construction technology in order to prepare qualified professional construction managers.  The American Council for Construction Education (ACCE) accredits 52 baccalaureate degree programs and six associate degree programs.  Additionally, there are eight baccalaureate degree and six associate degree candidate programs preparing for their initial ACCE accreditation process.  Construction education programs that are ACCE accredited are designed to produce a graduate who has the ability to manage and supervise the total construction process (ACCE, 2000). 

The members of ACCE, construction industry representatives, and construction educators establish and maintain standards and criteria for accreditation, provide guidance to those seeking to achieve accredited status, and carry out the accreditation and reaccredidation processes.  Approximately 50 of the schools that are accredited by ACCE also belong to Associated Schools of Construction (ASC), which has 96 members.  The professional association of ASC promotes the development and advancement of construction education, where the sharing of ideas and knowledge inspires, guides, and promotes excellence in curricula, teaching, research, and service.  Both of these organizations promote student retention so that a qualified construction manager can be produced.  Most construction related programs have seen a rise in their enrollments due to increased career opportunities with high salaries and 100 percent job placement after graduation.  However, student attrition seems to be on the rise in several of these programs due to student failure or students being hired by construction companies before they complete their degree program. 

Background of the Problem 

Since the 1950s, universities and colleges have been faced with the problem of an increase in student attrition.  Universities and colleges have continuously attempted to identify academic and non-academic variables that could best serve as indicators or predictors of student retention and success (Rogers, 1989).  According to Vincent Tinto (1993), a leading researcher on student attrition in college:

“More students leave their college or university prior to degree completion than stay.  Of the nearly 2.4 million students who in 1993 entered higher education for the first time, over 1.5 million will leave their first institution without receiving a degree.  Of those, approximately 1.1 million will leave higher education altogether, without ever completing either a two or four year degree program.  (p.1)” 

Today, the concern over why these students leave college before finishing their degree program has gained even greater importance among administrators of higher education (McLaughlin, Brozovsky, & McLaughlin, 1998).  Few topics in higher education have received as much attention as those of student persistence and departure (Mohr, Eiche, & Sedlacek, 1998).  Withdrawal from college is an important issue in higher education for several of reasons: (1) withdrawal is generally considered a painful process for many students because vocational and/or personal setbacks may result from impeded career development and the futile expenditure of time and effort; (2) withdrawal presents problems to school administrators because of inadvertent misallocation of limited educational resources (Peng & Fetters, 1978).  The idea of attempting to identify academic and nonacademic variables that could best serve as indicators or predictors of student retention and success in college dates back to as early as 1918 (Wilson, 1978).  Since then, a variety of investigations have studied particular variables that would best predict the academic success of college-bound students.  Predicting the success of students from academic and nonacademic variables in an educational program continues to be an inexact science, and student attrition continues to have an adverse impact on students, institutions, and society (Campbell & Dickson, 1996).  It appears that the implications of predicting if a student will graduate has gained even greater importance for universities, parents, and students in today’s higher education environment.  Parents and students are concerned about the likelihood of students completing their degree program of study, especially given the increasing costs of attending college.  Legislatures in some states have become involved in retention issues by collecting information on performance indicators while in other states they have used the information to make decisions about funding and policy.  University and college administrators cannot ignore the implications of such trends (McLaughlin, Brozovsky, & McLaughlin, 1998). 

For several years, universities and colleges have admitted students on the basis of their standardized test scores, high school grade point average, and/or class standing because these three variables seem to be the best indicators of predicting a students’ academic success.  Therefore, most universities have used the American College Test (ACT) or the Scholastic Aptitude Test (SAT) along with high school grade point average to predict if a student will graduate (Rogers, 1989; Merante, 1983; Kanoy, Wester, & Latta, 1989).  To assist in making more accurate predictions of a student’s academic success, universities are also using a student’s class rank (Merante, 1983). 

However, there are many hazards in making predictions about a student’s academic success.  Predicting a student’s academic success is similar to predicting how the stock market or economy is going to perform (Merante, 1983).  Nevertheless, someone is always trying to make predictions whether it is a presidential election or sporting contest.  Predicting a student’s academic success is important so that qualified candidates are admitted and unqualified ones are not admitted.  Predicting a student’s academic success will also help prevent the misallocation of funds for a student who should not have been admitted.

The previous studies have led to a greater understanding of student attrition and retention, and there are several studies which predict if a student will graduate from a specialized program, especially nursing.  However in my literature review, I have located no studies which predict student’s success in a building construction management program, so, this project will serve to initiate more studies.  Over half of the studies that predict academic success have focused either on freshman, the entire student body, or specific programs such as nursing, engineering, or business; thus, the extent of research that applies to building construction management students is uncertain.  Predictor variables need to be identified in order to have a good admission policy that will admit students who can be successful in completing a four-year baccalaureate construction management program. 

Purpose 

The purpose of this investigation is:  (1) to identify the variables that prove to be the best predictors of graduation in BCM, (2) to identify factors to use as a recruiting/marketing tool for the BCM program, (3) to provide a basis for educators to better understand the selection criteria for admission of students in the field of construction, construction science, construction management, and construction technology, and (4) to help other university construction management programs to identify predictor variables that can serve as valid predictors of college success for their own construction management programs. 

Study Design

Logistic regression was the statistical procedure used to make an association among the independent predictor variables (high school rank, high school GPA, high school class size, number of high school science courses, number of high school math courses, SAT composite score, matriculation age, gender, race, and residence) the dependent variable graduation.  Logistic regression is a method for investigating one or more independent predictor variables that determine an outcome that is measured with a dichotomous variable in which there are only two possible outcomes.  The objective of logistic regression is to find the best fitting model to describe the association between the dichotomous dependent variable (Pampel, 2000). 

Participants 

A sample of approximately 400 students who originally enrolled in Purdue University’s BCM program from the fall semester 1992 to the fall semester 1997 was used for this study.  The group of participants contained 385 males and 11 females with an overwhelming majority of the students being Caucasian American male.  The average age of the participants was 18.3 years.  The useable sample size was 343 due to missing data for some students.  All change of degree option (CODO) students were omitted.  A CODO student is someone who changed his or her originally declared major such as engineering, education, or history to building construction management.  The reason CODO students were omitted from the study is that funding to schools and departments is based on the initial freshman class as opposed to graduating seniors (personal interview, Ron Burkhardt February 19, 2002).  It is important for the department to target a student who will initially declare a BCM Degree and graduate so the BCM department will get the proper funding. 

Data Collection 

Student data used to develop this study was collected from University’s Office of the Registrar Research and Reporting Services.  Pre-enrollment data was obtained from University’s Oracle database by using the Decision Support System with queries being ran against the student information files.  Academic variables assembled are high school rank, high school GPA, number of science and math courses, high school class size, SAT composite score, and ACT composite score.  Personal variables are matriculation age, gender, race, and residency.  Both the academic and personal variables are classified as independent variables.  The dependent variable is graduate or non-graduate with a baccalaureate degree by the end of spring 2002 from the university’s construction management program. 

Predictor variables that are readily available were highly used because it did not require the students to take an additional test.  The high degree of reliance on academic variables could be questioned since some non-cognitive variables may be of equal value.  However, academic variables are still the strongest predictors presently available in the study of student persistence and attrition (Sanford, 1982).  The variables chosen for this study are justifiable both on the basis that they have been shown to be significant in previous research and that they are the usual indicators for admission decisions.  Although a number of other entering freshmen characteristics add significantly to the prediction of retention, student’s high school grades, admission test scores, race, and gender account for the bulk of the variance in retention that can be predicted from entering freshmen characteristics (Astin, 1997).  Universities can effectively assist those students who apply and thus, focus their recruitment efforts when precollege characteristics have been developed.  Additionally, this will assist universities in attracting those students who are predicted to perform best at their respective institution.  Obviously, this kind of focusing has tremendous benefits both for the university and for its prospective students (Merante, 1983). 

Statistical Analysis 

Data collected from the University’s Office of the Registrar Research and Reporting Services records were complied into a Microsoft Excel Spreadsheet for analysis.  Logistic regression was the statistical procedure used to show the relationship between the independent and dependent variables.  The Statistical Package for the Social Science (SPSS) version 11 was used to perform the mathematical calculations. 

Characteristics of the Sample 

The population being studied was 396 students who originally declared Building Construction Management as their major at Purdue University from fall 1992 through fall 1997.  However, 53 students were dropped from the study due to missing data so the final population was 343 students.  The sample population by gender and ethnicity is presented in Table 1.1. 

Table 1.1  Sample Population by Gender and Race

There were 334 (97.4%) males and 9 (2.6%) females in the sample.  The student sample by ethnicity was 95.3% Caucasian American (n = 327), 1.7% African American (n = 6), 0.9% Native American/Alaskan Native (n = 3), 0.6% Hispanic American (n = 2), 0.3% Asian American (n = 1), 0.6% other (n = 2), and 0.6% unreported (n = 2). 

The mean and standard deviation for each independent variable were developed.  The dependent variable was graduated or not graduated with a BCM Baccalaureate Degree.  Means and standard deviations are presented in Table 1.2. 

Table 1.2  Mean and Standard Deviation for the Independent Variables.

Matriculation age ranged from 17 to 26 years with a mean age of 18.3 years and a standard deviation of 0.708.  High school rank ranged from 0.02 to 0.95 with a mean of 0.6852 and a standard deviation of 0.2024.  Meaning, if a student had a high school senior class size of 100, the 0.95 was someone who ranked 5^th out of 100 or in the 95^th percentile.  The 0.02 means that a student was ranked 98th out of 100 or in the 2^nd percentile.  The number of high school math semesters completed ranged from 4 to 12 with a mean score of 7.68 with and a standard deviation of 1.436.  The number of high school science semesters completed ranged from two to ten with a mean score of 5.45 and a standard deviation of 1.653.  High school GPA ranged from 1.43 to 4.00 with a mean of 2.658 and a standard deviation of 0.4601.  SAT composite score ranged from 680 to 1380 with a mean of 999.450 and a standard deviation of 115.518.  The SAT mean 999.450 is in the 49^th percentile of all students who take the SAT based on scores from the 2002 College-Bound Seniors cohort.  Because some of the students had taken the ACT and not the SAT, each student’s ACT composite score was converted to an SAT composite score using concordant tables provided by the American College Testing Program (ACT).  The concordant tables allow for an ACT score to be converted to a corresponding SAT score. 

One of the purposes of this investigation was to identify the variables that prove to be the best predictors of graduation in BCM.  Therefore, logistic regression was performed using all independent variables (high school rank, high school GPA, high school science semesters, high school math semesters, high school class size, SAT composite, matriculation age, gender, ethnicity, and residency) to predict graduation (yes/no) with a BCM degree.  This relationship was tested at the 0.05 level of confidence.  When the full model was tested, there were two variables that had a significant positive relationship to graduation as shown in Table 1.3. 

Table 1.3 Results of Combined Model.

The results of the test show that two independent variables are significant in the full model: high school GPA and high school math semesters.  The Wald Chi-Square test for high school GPA = 3.888, df = 1, and p = 0.049.  The Wald Chi-Square test for high school math semesters = 9.452, df = 1, and p = 0.002. 

Recommendations

The following are several recommendations for future research.

1.        This study should be replicated in other construction management programs to investigate if the same variables have the same influence.

2.        In an effort to develop a more comprehensive profile of the successful college student, future studies should examine non-cognitive variables such as on/off campus living, involved/not involved with student activities, and their effect on graduation from a building construction management program.

3.        Because there was a small number of minorities and females, the hypothesis tests were inconclusive.  This variable needs to be further investigated.  It is recommended that other construction management programs be investigated to see if this variable is significant.

4.        Because the number of high school math semesters was significant, this variable needs to be further investigated to determine if the type and level of academic math course is significant.  It should be noted that only academic high school math courses such as algebra, geometry, calculus, statistics, or finite math are counted.  Non-academic math courses such as general math, business math, or computer math is not counted. 

Conclusion 

Based on the analysis, only high school GPA and high school math semesters were significant.  The practical implications of this study are for college admissions personnel and educators at the university’s construction management department who are interested in learning more about the characteristics of successful and unsuccessful construction management students.  By identifying predictor variables that can serve as valid predictors of college success, this study will provide university admission officers and educators in the construction with more valid, in‑depth, theoretically grounded, empirically obtained information on academic and personal characteristics of college students other than the descriptions presently available.  Other university construction management programs can use this information as well in helping them identify predictor variables that can serve as valid predictors of college success for their own construction management programs. 

First, the results of this study will provide a more descriptive profile of the successful and unsuccessful students in the field of construction, construction science, construction management, and construction technology than exists in the current literature.  Additionally, the study will provide a basis for educators to better understand the selection criteria for admission of students to the field of construction, construction science, construction management, and construction technology. 

Secondly, the results of this study will have implications for admission personnel and construction faculty by (1) establishing a body of information to serve as a foundation for future research and (2) providing a basis for decision‑making in the development of precollege enrichment programs. 

Finally, the results of this study will have implications on the student profile for university admission officers by: (1) providing a catalyst for more communication between high schools and colleges with regard to what students need to know in order to succeed in a construction, construction science, construction management, or construction technology program, (2) providing student counselors more information about successful student characteristics, thereby enhancing the counselors’ abilities to help prospective students achieve realistic perspectives about what they hope to accomplish at the university, (3) assisting administrators in formulating recruitment and admission policies and in allocating financial aid to enhance the probability that students will complete their studies, and (4) increasing the ability of university personnel to identify those students who have a high probability of being unsuccessful. Those students could be identified early and directed into appropriate intervention programs which could address their particular deficiencies, thereby improving their chances for success and retention. 

References 

American Council for Construction Education.  (2000, December 4).  Introduction to ACCE.  Retrieved December 5, 2000, from the World Wide Web: http://www.acce-hq.org/Intro/Intro.htm 

Astin, A. W.  (December 1997).  How good is your institution’s retention rate?  Research in Higher Education, 38, (6), 647-658. 

Campbell, A. R., & Dickson, C. J.(Jan-Feb 1996).  Predicting student success: A 10-year review using integrative review and meta-analysis.  Journal of Professional Nursing, 12, (1), 46-59. 

Kanoy, K., Wester, J., & Latta, M.  (Spring, 1989).  Predicting college success of freshman students using traditional, cognitive and psychological measure.  Journal of Research and Development in Education, 22, (3), 65-70. 

Lea, S.  (1997, March 11).  Logistic regression and discriminant analysis.  Retrieved October 4, 2002 from the World Wide Web: http://www.ex.ac.uk/~SEGLea/multvar2/disclogi.html 

McLaughlin, G. W., Brozovsky, P. V., & McLaughlin, J. S.  (Feb 1998).  Changing perspectives on student retention: A role for institutional research.  Research in Higher Education, 39, (1), 1-17. 

Merante, J. A.  (February 1983).  Predicting student success in college:  What does the research say?  NASSP Bulletin, 67, (460), 41-47. 

Mohr, J. J., Eiche, K. D., & Sedlacek, W. E.  (1998).  So close, yet so far:  Predictors of attrition in college seniors.  Journal of College Student Development, 39, (4), 343 – 354. 

Pampel, F. C.  (2000).  Logistic regression:  A primer.  (Sage university papers series on quantitative applications in the social sciences, series no. 07-132).  Thousand Oaks, California:  Sage Publications. 

Peng, S., & Fetters, W.  (1978).  Variables involved in withdrawal during the first two years of college:  Preliminary findings from the National Longitudinal Study of the High School Class of 1972.  American Educational Research Journal, 15, 361-372. 

Rogers, P. H.  (1989).  Predicting student success in college.  (Doctoral dissertation, Grambling State University, 1989).   

Sanford, T. R.  (Spring, 1982).  Predicting college graduation for black and white freshman applicants.  College and University, 57, (3), 265-278. 

Tinto, V.  (1993).  Leaving college: Rethinking the causes and cures of student attrition (2nd ed.).  Chicago:  University of Chicago Press. 

Wilson, H. E.  (1978).  An investigation of intellectual and non-intellectual variables as predictor of academic success of high risk college freshman at Southern Illinois University of Carbondale.  (Master’s Thesis, Southern Illinois University of Carbondale, 1978).