How student achievement is related to student behaviors and learning style preferences.

Terregrossa, Ralph A. ; Englander, Fred ; Wang, Zhaobo 等

INTRODUCTION

Previous research (e.g., Charkins, O'Toole and (1985), Hawk and Shaw (2007), Terregrossa, Englander and Wang, (2009)) has examined the relationship between students' academic achievement and their learning styles. Other research (e.g., Brokaw & Merz, 2000), Englander, Terregrossa and Wang (2009)) has examined the relationship between students' academic achievement, and their behavior, including alcohol consumption, and time spent on fitness activities, studying and non-academic employment. The present study combines these two perspectives and, furthermore, explores the relationships between student achievement and both student behavioral choices and student learning style preferences in a multivariate context.

A focal element in this research is the degree to which student behavioral choices (i.e., commitment of energies to alcohol consumption, Internet use, study hours, fitness and employment hours) can be explained by the learning style profiles of those students. In this study, the Dunn and Dunn learning styles model (DDLSM) (Dunn, 2000) is used to measure and explain learning styles. The DDLSM argues that one important characteristic that influences a student's success in processing information (learning) can be explained in terms of the degree to which a student falls along the analytic learner/global learner continuum. Factor analysis is used to reduce the twenty learning style preference variables of the DDLSM model to five instrumental factors which then are used to explain variations in the aforementioned student behavioral choice variables. The Ordinary Least Squares estimation approach is applied to a system of equations to examine the influence of behavioral choices, learning style preferences and academic ability on student performance in introductory microeconomics.

This study also utilizes the restricted least squares regression methodology to evaluate whether three separate categories of variables, student behavioral choices, learning style preferences and innate student ability, have a statistically significant influence on student achievement. Inferences are made regarding the relative contribution of these three categories to student achievement. From this analysis, it should be possible to evaluate the extent to which concerted efforts on the part of university instructors, university administrators, counseling staff and other student-support personnel to pursue strategies designed to enhance student academic achievement (and, therefore, progress toward graduation) are potentially worthwhile.

PUBLIC POLICY BACKGROUND OF THIS STUDY

In July 2009, President Obama announced the "American Graduation Initiative," a set of legislative proposals designed to address the low high school, community college and four year college graduation rates that he argues are impairing the economic growth and competitiveness of the American economy (Obama, 2009). Consistent with this theme, several influential writers have also recently written about the low graduation rates in the American educational system and the detrimental economic and social effects associated with such low rates of graduation. Herbert (2009) emphasizes that education is the key determinant of personal and societal success. He stresses that a two or four year college degree is a prerequisite to a middle class standard of living. He points to the high drop-out rate and reports that fifty percent of minority students drop out of high school. Subsequently, students who drop out of high school are less likely to attain a middle class standard of living and more likely to become a burden to society.

Krugman (2009) asserts that public higher education has been the main impetus to U.S. economic growth in the past two centuries. But because of the relative decrease in federal resources allocated to education over the past three decades and the greater recent fiscal restraint on the part of state governments, U. S. students are less likely to graduate from college compared to many other developed countries. He claims that it is more difficult for low income students to stay in school because of the need to work which reflects family financial pressures. Krugman laments, "One result, almost surely, will be lifetime damage to many students' prospects--and a large, gratuitous waste of human potential." (Krugman, 2009, p. A31) Krugman calls for a substantial increase in federal funding to reverse the neglect of the public education system.

Leonhardt (2009) states, "In terms of its core mission--turning teenagers into college graduates--much of the system is simply failing." (Leonhardt, 2009, p. B5) He reports that only half of the students enrolled in American colleges graduate. He directly links the low graduation rate in U.S. higher education to decreases in economic equality and economic productivity. Leonhardt (2009) explains that one cause of the low rate of graduation is under matching--the tendency of students to choose not to attend the best college for which they have been accepted, but instead to attend colleges that are in close proximity to their homes or are less expensive. Leonhard contends that part of the solution is greater financial support of students.

Brooks (2009) focuses on the role of community colleges. He reports that fifty percent of community college students do not graduate. Contrary to Krugman (2009) and Leonhardt (2009), Brooks (2009) contends, "Lack of student aid is not the major reason students drop out of college. They drop out because they are academically unprepared or emotionally disengaged or because they lack self discipline or because bad things are happening at home." (Brooks, 2009, p. A23) He observes, "Over the past 35 years, college completion rates are flat. Income growth has stagnated. America has squandered its human capital advantage." (Brooks, 2009, pp. A23). Brooks concludes,

"Real reform takes advantage of community colleges most elemental feature. These colleges educate students with wildly divergent goals and abilities. They host students with radically different learning styles ... You have to experiment with programs ... that are tailored to individual learning styles." (Brooks, 2009, A23)

This perspective articulated by Brooks (2009) aligns with the research results of Barr (2007) who studied the student characteristics and behaviors influencing retention behavior among over 19,000 freshmen at a large public community college in California. Barr (2007) found that the most important variable influencing whether students dropped out between their first and second academic years was course grade performance. This positive link between retention, or student progress toward graduation, and academic performance is reinforced by research results presented by Aitken (1982) and Murtaugh, Burns and Schuster (1999). The thrust of this argument, then, is that faculty and institutional efforts to better accommodate student learning styles improves learning and academic performance, and improving academic performance enhances student retention and progress toward graduation.

METHODOLOGY

Factor analysis is used to reduce the twenty learning style preference variables to five instrumental factors which then are used to explain variations in the aforementioned student behavioral choice variables. A multivariate estimation approach is applied to a system of equations to examine the influence of learning style preferences, student ability and student behavioral choices on student performance in introductory microeconomics.

Englander, et al. (2009) used a standard linear regression model to measure the contribution of various student behavioral variables (alcohol consumption, hours committed to fitness activities, study hours and student work hours) to student achievement in introductory economics. The model included qualitative variables to account for the level of difficulty of the different exams, student gender, student maturity, the student's overall academic ability and the relative strength of the student's math and communication abilities.

The methodology used by Englander, et al. (2009) improved upon previous research that focused on explaining variations in student academic achievement by virtue of the inclusion of student behavioral variables that account for actual time spent on physical fitness activities as well as individual choices regarding the consumption of alcohol. However, this study did not account for the potential influence of students' learning styles on student achievement.

Charkins, et al. (1985) used a simultaneous two-stage least squares regression model to examine the link between student achievement and learning style and the link between student attitude and learning style. In the first equation of the model, student achievement was regressed against student ability, student attitude, student effort, and the quality of instruction. In the model's second equation, student attitude was regressed against student ability, student achievement, student effort, quality of instruction and other socioeconomic variables. The quality of instruction was measured by the degree to which the instructors teaching style diverged from students' learning styles.

Charkins, et al. (1985) used the Grasha-Riechamann Learning Styles Questionnaire (GRLSQ) to differentiate students' learning styles. The GRLSQ identifies six learning styles "related to interaction between a learner and his or her peers and instructors: Avoidant, Collaborative, Competitive, Dependent, Independent and Participative." (Vaughn, Battle, Taylor, & Dearman, 2009, p.723) However, Charkins, et al. (1985) used the GRLSQ to identify three of the six types of learning styles.

Terregrossa, et al. (2008) used a standard linear regression model to measure the contribution of learning styles to student achievement. The multidimensional DDLSM was used to differentiate students learning styles. The central tenet of the DDLSM is that the best method of teaching is the method that most closely matches the way the student learns. The DDLSM is composed of a combination of environmental, emotional, sociological, physiological and psychological strands. Each strand is composed of several elements. The Productivity Environmental Preference Survey (PEPS) (Dunn, Dunn, & Price, 2006) was used to identify students' preferences for each of the elements in the environmental, emotional, sociological, physiological, and psychological strands.

The environmental strand includes preferences for sound, light, temperature and design, or formal versus informal seating. The emotional strand includes propensities for motivation, persistence, conformity (responsibility) and structure. The sociological strand reflects with whom a student prefers to learn and the preferred manner in which the material is learned (variety versus routines). The physiological strand includes preferences for perceptual modality (auditory, visual, tactile and kinesthetic), intake (snacks or drinks), varying energy levels throughout the day and mobility. The psychological strand refers to the way that a student absorbs and processes new and difficult information and whether the student is compulsive or reflective. Students' learning styles are hypothesized to be analytic, global or indifferent. The analytical and global styles correlate with preferences for sound, light, design, persistence and intake (identified as the five discriminating elements--i.e., the learning style preferences which are most relevant to categorizing a student as being an analytic or global learner).

Terregrossa, et al. (2008) also utilized the restricted least squares regression method to determine which of the five learning style strands have the greatest impact on student achievement and thus should receive the most consideration in crafting teaching styles. However, the study failed to account for the potential effects of differences in student

behavior and ability on student achievement.

Brokaw and Merz (2000) examined the effects that both learning styles and student behavior have on student achievement in principles of microeconomics courses. Their results indicated that the students whose learning styles matched their instructors' "chalk and talk" style had significantly higher grades than those students whose learning styles did not match.

In the present study, similar to Charkins, et al., (1985), a multivariate regression model is used to explain student achievement. In the first equation of the model, student achievement is regressed against the four behavioral variables suggested by Englander, et al. (2009), including the level of alcohol consumption, the number of hours committed to fitness activities, course related study hours and student work hours. In addition, students' time spent on the Internet is included as a behavioral variable. The PEPS instrument is used to identify the twenty learning style elements of the DDLSM, and factor analysis is used to decrease the twenty elements to five, independent learning style factors that reflect the alternative DDLSM learning styles. These five learning style factors are included as explanatory variables in the student achievement regression model. In this way, any inherent collinearity that exists among the learning style elements is eliminated. Dummy variables are included to account for differences in gender, the rigor of the three tests, and in the differences in the learning environment that may have existed among the alternative course sections. Total SAT scores are included as a measure of students' academic aptitude, and the ratio of the students' SAT math to verbal scores is included to examine whether a student would have an advantage in introductory microeconomics if that student has relatively strong math abilities. The number of credits completed is included to account for differences in students' maturity.

The remaining five equations of the multivariate model regress each of the behavioral variables against the same set of exogenous variables, including gender, academic ability, maturity and the five learning style factors. The learning style factors reflect the analytic, global and indifferent learning styles of the DDLSM. The DDLSM hypothesizes that indifferent learners may be either analytic or global in nature, or a combination of the two, depending on the material to be learned. The analytic learners learn best in a quiet, brightly lighted and formal learning environment. They prefer to study alone and to start and finish one project at a time. They do not snack or drink while learning. Global learners learn best with background noise, soft light in a relaxed learning environment. They simultaneously work on several projects, take frequent breaks, and enjoy snacks or drinks when learning. Global learners prefer to study with others and prefer that new and difficult information to be introduced anecdotally, especially in a way that humorously explains how the lesson relates to them.

Vaughn et al. (2009) examined the influence of learning styles on attachment styles and psychological symptoms in college women. Attachment styles, or ways in which people interact and relate to others, influence the behavior of individuals in personal and intimate relationships, in the workplace and in leadership roles. Psychological symptoms include, for example, anxiety, depression, obsessive and compulsive characteristics and hostility. The authors find a statistically significant correlation between learning styles and both attachment styles and psychological symptoms. The authors state, "Being aware of preferred learning styles may also increase the likelihood that students become more successful in making decisions inside and outside the academic environment ... " (Vaughn, et al., 2009, p. 733) The findings indicate that human behavior is significantly affected by learning styles.

In the present study, in accordance with Vaughn, et al. (2009), it is hypothesized that students' behavioral choices regarding alcohol consumption, Internet use, study hours, exercise and employment may be affected by their learning styles, gender, academic ability and maturity. Since analytic learners prefer not to snack or drink when learning when compared to global learners, who prefer to snack or drink while learning, analytic learners may consume fewer alcoholic drinks. Furthermore, the milieu of college parties, night clubs, bars and restaurants where alcohol is likely to be served, is typically noisy, dimly lit and crowded, and therefore is contrary to analytic learners environmental preferences but consistent with global learners' preferences. With regard to study hours, analytic learners are more persistent than global learners and, therefore, may be more likely to dedicate a greater, sustained period of time studying compared to global learners. Since analytic learners are more persistent than global learners, they may take fewer prolonged breaks when studying, for example, to exercise or participate in intramural sports. Since analytic learners prefer to work alone and global learners prefer to work with others, global learners may be more likely to participate in team sports and analytics may be more likely to participate in individual sports or exercise. For the same reasons, analytic learners may dedicate less time to employment when compared to global learners, ceteris paribus. Consistent with their environmental and sociological preferences, analytic learners may prefer to work, for example, in a quiet library or as an independent research assistant, whereas global learners may prefer to work in the din of restaurants, bars and dining halls or work with a group of students as a teacher's assistant. The Internet provides both analytic and global students with a milieu that is consistent with their learning style preferences. The Internet allows analytic learners to work alone, in a formal, quiet, brightly lit setting, without snacks or breaks. In contrast, the Internet allows global learners to take frequent breaks while working, to easily divert the focus of their Internet use, to conduct work on, chat on, or surf the Internet with others and to snack or drink while working on the Internet.

With regard to the relationships between student behavior, maturity and academic ability, it is hypothesized that a more mature student with a higher academic aptitude is more likely to exhibit a more salubrious life style. For example, a mature, intelligent individual may be more likely to exercise on a regular basis to stay physically fit, and drink alcohol in moderation. With regard to gender, it is possible that differences exist in the type and quantity of alcoholic drinks consumed, in the type, intensity or duration of physical exercise, and the types of employment between male and female students.

It therefore is quite possible that students' behavior may be related to their learning styles, gender, maturity and academic ability. Explicitly incorporating the relationships between student behavior and their learning styles, gender, maturity and ability into the regression model may substantially improve the specification of the student achievement model. However, it is necessary to separately account for the direct effects of learning styles, gender, maturity and academic ability on student achievement and the indirect effects of these variables on student achievement via their impact on student behavior. To accomplish this task, we utilize the residuals from the behavior regressions as explanatory variables in the student achievement model instead of the actual observations of the behavior variables. The residuals of the behavior regressions represent the behavior variables purged of the indirect effect of learning styles, gender, maturity and ability. The multivariate model, composed of six linear regression equations, is estimated in two steps. First, the five behavior regression equations are estimated using the Ordinary Least Squares (OLS) method. If the cross-equation residuals from the behavior regressions are correlated, then the Seemingly Unrelated Regression (SUR) method would produce efficient estimates. But, as Telser (1964) and Pindyck and Rubinfeld (1998) explain, "SUR estimation and ordinary least squares estimation are Identical ... when the explanatory variables ... are identical." (Pindyck & Rubinfeld, 1998, p. 360) Therefore, in this case, the OLS regression method produces results that are efficient and identical to the SUR method.

In the second step, the student achievement variable is regressed against the five purged behavior variables (i.e., the residuals from the behavior regressions), the learning style factors and the academic ability variables, gender and the control variables described above. Inclusion of the purged behavior variables and the learning style factors in the student achievement regression model improves upon previous research concerned with explaining student achievement.

The partial correlation coefficients of the student achievement regression model provide evidence regarding whether the individual explanatory variables have a statistically significant effect on student achievement. However, the partial correlation coefficients do not indicate the statistical significance of the alternative groups of behavioral, learning styles or academic ability variables. In the final stage of the analysis, the restricted least squares regression method is used to determine which of the three groups of explanatory variables has the greatest impact on student achievement. The hypothesis tested is whether or not a particular group, when considered as a subset of the total number of explanatory variables in the student achievement regression model, has a statistically significant impact on student achievement. The joint F-test for the restricted least square regression method is utilized to test the aforementioned hypothesis.

EMPIRICAL ANALYSIS

Data for 125 students were collected from eight sections of the same introductory microeconomics course over four semesters from spring of 2003 through fall of 2005 at a university in the northeast. All sections were taught by the same instructor. The introductory microeconomics course is the first semester of a two course economics sequence that is required of all business majors. The business program is accredited by AACSB.

The Productivity Environmental Preference Survey (PEPS) (Dunn, Dunn & Price, 2006) instrument, based upon the DDLSM, was used to identify students' preferences for the twenty elements comprising the five strands of the learning style model. The PEPS, designed to identify how college students and other adults learn and perform in their academic and occupational pursuits, is a self-report composed of 100 questions that can be completed in approximately twenty minutes.

Each question is designed to identify an individual's preferences for twenty separate elements which, in turn, are included among the environmental, emotional, sociological, physiological and psychological categories.

Factor analysis decomposed the information contained in student preferences for the twenty learning style elements into a smaller set of uncorrelated common factors that maintain and reflect the inherent characteristics of the original learning style elements. Each learning style element (X) is assumed to be a linear combination of a set of common factors (F) and a component (U) that is unique to the element as described in the following Equation 1.

[X.sub.i]=[b.sub.i1]+ [b.sub.i2] [F.sub.2]+ ... +[b.sub.ij] [F.sub.j]+ ... +[b.sub.im] [F.sub.m]+ [U.sub.i]

where [X.sub.i] (i = 1, 2,.. .20) is the ithlearning style element, [F.sub.j] (j = 1, 2, ... m) is the [j.sup.th] common

Notes: (1) P-values are reported in parentheses and a starburst (*) denotes statistical significance. (2) An (a) denotes an analytic learning style characteristic and a (g) denotes a global leaning style characteristic, based on the sign of the respec tive

factor, [b.sub.ij] is the factor loading, or standardized correlation coefficient between the learning style element i and the common factor j, and [U.sub.i] is the component unique to the leaning style element. In factor analysis, the number of factors is partially determined by the percentage of total variance (eigenvalue) that is explained by each factor. In this case, the five extracted factors cumulatively accounted for over forty-three percent of the variation in the twenty learning style elements. For each of the five extracted factors, the factor loadings and their p-values of the discriminating learning style elements, including the environmental preferences for noise, light, design, the emotional preference for persistence and the physiological preference for intake, are reported in Table 1.

Factor analysis has identified and differentiated all three learning styles hypothesized by the DDLSM: the global, analytic and indifferent learning styles. The first factor is strongly indicative of the global learning style. Four of the five factor loadings, three of which are statistically significant, have a negative sign which is consistent with the global leaning style. The second factor is weakly indicative of the global learning style. Four of the five factor loadings have a sign which is consistent with the global learning style, but only one, intake, is significant. The factor loading associated with light, an environmental element, is significant, but the positive sign is consistent with the analytic learning style. The third factor is indicative of the indifferent learning style. Three factor loadings, noise, design and intake, have expected signs that are consistent with the global learning style. The factor loadings for the intake and noise elements are significant. The factor loadings for the persistence and light elements have positive signs which are consistent with the analytic learning style and both are significant. Therefore, the evidence indicates a combination of the global and analytic learning styles. The fourth factor is weakly indicative of the global learning style. Three of the factor loadings have the expected sign consistent with the global learning style, but none are significant. The fifth factor is strongly indicative of the analytic learning style. Four of the factor loadings have signs consistent with the analytic learning style, all of which are significant. The alternative learning styles aligned with the extracted factors via factor analysis are summarized in Table 2. In the final stage of factor analysis, the factor loadings are used to generate factor scores for each student. In this way, factor analysis transformed each student's original, collinear preferences for the twenty DDLSM elements into five independently distributed factor scores that embody each student's inherent learning style.

In the second phase of the multivariate analysis the OLS method is used to estimate each of the five behavior models. The behavior model is described in Equation 2.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where B[V.sub.bi] denotes the ith observation of the bth behavior regression (b=1 to 5 and i=1 to 125); [c.sub.bk] denotes the partial correlation coefficients (b=1 to 5 and k=0 to 9); [F.sub.di] represents the dth learning style factors (d=1 to 5 and i=1 category; Credits denote the number of credits completed; SAT is the total SAT score; SATratio is the ratio of math to verbal SAT score and [V.sub.bi] is the error term. The results of the behavior regressions are reported in Table 3. With the exception of the hours spent on fitness, learning styles are significantly correlated with the behavioral variables. The strongly global learning style is positively correlated with alcohol consumption and significant at the .01 level. Gender and credits completed are both negatively correlated with alcohol consumption and significant at the .01 level. That is, during an average week, female students consume nearly six more alcoholic drinks than their male counterparts. With regard to time spent on the Internet, the coefficients of the strongly global and weakly global (F2) learning styles are positive and significant at the .01 level. The SAT total and SAT ratio variables have a positive coefficient and are significant at the .01 level. Both the strongly global and indifferent learning styles are negatively correlated with the number of hours spent studying economics and are significant at the .05 level. The weakly global learning style (F4) is positively correlated with the number of hours spent studying and significant at the .05 level. The SAT total coefficient is negative and significant at the .01 level. This result suggests that weaker students study substantially more hours than more able students in order to achieve a target grade, even though the marginal productivity of such efforts is relatively low. Gender, maturity, academic ability and learning styles all have a statistically significant impact on number of hours employed. The strongly analytic learning style is positively correlated and significant at the .01 level. The indifferent learning style is negatively correlated and significant at the .05 level. SAT total is positively correlated and significant at the .05 level. The number of credits completed is negatively correlated and gender is positively correlated with the number of hours spent working. Both are significant at the .01 level. Credits completed, negatively linked to time spent on physical fitness, is the only significant independent variable explaining fitness.

In the third phase of the analysis, the student achievement regression model is estimated using the OLS regression method. This regression model is described in Equation 3.

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where E[X.sub.i] is the ith score for the three examinations for each student (i=1 to 375), [c.sub.k] are the partial correlation coefficients (k=0 to 23); B[V.sub.bi] are the bth behavior residuals from Equation 2 (b=1 to 5), [F.sub.di] is the dth learning style factor (d=1 to 5); Dummy1 and Dummy2 are the examination dummy variables, the first exam is the base; Section1-[7.sub.i] are the dummy variables for the eight class sections, the first section is the base; and [V.sub.i] are the residuals.

The results of the student achievement regression model are reported in Table 4. Three of the behavioral variables are significantly correlated with student achievement. The number of hours spent studying economics is negatively correlated with student achievement and significant at the .10 level. The coefficient for the number of hours dedicated to fitness is negative and significant at the .01 level. There is a negative association between student achievement and employment that is significant at the .05 level. The global learning style is negatively correlated with student achievement and is significant at the .01 level. The indifferent learning style is positively correlated with student achievement and significant at the .05 level. The SAT score is positively correlated and significant at the .01 level. The dummy variables for the third exam and Section3 both are positively correlated with student achievement and significant at the .01 level. The coefficient of the dummy variable for Section5 is negative and significant at the .05 level.

In the last phase of the analysis, the alternative clusters of behavioral, learning styles, and academic ability variables are tested to determine if the cluster, when considered as a subset of the total explanatory variables, has a significant impact on student achievement. The results of the JointF tests for the restricted least squares regression models are reported in Table 5.

Each cluster of explanatory variables tested has a statistically significant effect on student achievement. The behavioral and the learning styles cluster are significant at the .05 level and the academic ability group is significant at the .01 level.

LIMITATIONS

The study is based on a relatively small sample size. The student subjects selected one instructor for one course at one private university. It is difficult to know the representativeness of those students. It is also likely that the academic qualifications of students at a private university are more homogeneous than those at public university. It is also possible that the results are influenced by specification bias resulting from the absence of such variables as student maturity, family relationships, and family support which are normally difficult to measure and, even if measured, difficult to include given reasonable precautions to protect student privacy.

CONCLUSIONS

Achen and Courant (2009) state, "Faculty give grades in part to provide awards, punishments, and signals of academic performance. Students know something about what they like and are good at--having learned in part from the grades they receive, relative to their peers, in the introductory courses." (Achen and Courant, 2009, p. 87) The results of this study suggest that, in addition to signaling the instructors' assessment and students' talents and preferences, grades also reflect students' life styles, learning styles and innate academic abilities. Student achievement is significantly correlated with student's behavioral choices and their learning styles. The findings of this paper, however, suggest that for the student sample examined here, student ability played the most significant role in explaining variations in student performance.

The authors believe that the results of this research reinforce the earlier work of Charkins et al. (1985), Browkaw and Merz (2000), Hawk and Shah (2007), Terregrossa et al. (2008) and Terregrossa, Englander and Wang (2009) who have stressed the importance of developing teaching approaches and tools to better accommodate the learning styles of students. The findings also support the analysis of Brooks (2009) who has argued that government efforts to address student dropout behavior and increase graduation rates would be advanced by offering greater support to and adopting the teaching strategies of community colleges because they have recognized the importance of student learning styles and have therefore made greater efforts to accommodate pedagogy to these various learning styles.

This research also highlights the importance of student behaviors and time management in accounting for differences in student performance, following the lead established by Kelley (1975) and others that were reviewed in Englander et al. (2009). The prescriptions that flow from the findings of this paper, however, suggest that despite the earlier findings in Englander et al. (2009) that alcohol consumption has a negative and significant impact on academic performance, student alcohol use was not found to have a significant influence on student performance. Student performance was found to be inversely related to student hours of Internet use, but this association was not statistically significant. That is, Internet use and number of drinks do not have a statistically significant link to achievement when those two behavioral measures are stripped of the mediating affect of learning styles.

Student grade performance, however, was found to be significantly and inversely related to hours of fitness activities, employment hours and study hours. The counter-intuitive finding that student performance is inversely related to study hours, a finding consistent with much of the earlier research on this question, as reviewed in Plant, Ericsson, Hill and Asperg, 2005), has been explained in prior research in terms of differences in the quality of study time (Plant, et. al, 2005), but, in this paper, by the possibility that weaker students may have to put in more study time to achieve a given target grade. It should be noted that student work hours have been linked to higher college dropout rates by Ehrenberg and Sherman (1987) and Horn and Malizio (1994). Further, commitment to athletics has been directly related to the athletes' college attrition by Maloney and McCormick (1993) and DeBrock, Hendricks and Koenker (1996). Although there are some studies offering contrary results, such studies are typically based on graduation rate data submitted to the NCAA. A recent article published by Steeg, Upton, Bohn and Berkowitz (2008) provides strong evidence that such data are misleading, if not fraudulent.

IMPLICATIONS

Colleges and universities may play some role in influencing student behavioral choices. However, it may be ironic that while a substantial majority of colleges and universities maintain counseling programs, residence hall staff and campus security efforts to discourage students from alcohol consumption, student performance in this study was found to be more threatened by excessive fitness activities and employment hours. The student counseling staff at the university from which the student subjects of this study were drawn reports that considerably more resources at that institution, and the vast majority of other higher education institutions, are committed to addressing the problem of student alcohol use than the problems introduced by poor student time management skills which may lead to poor student academic performance. In some instances, the role of the institution in sending the 'wrong' signals to students may go farther. Weisbrod, Ballou & Asch (2008) cite the contract between Auburn University and its football coach who, of course, has some responsibility for the athletic and academic performance of his players. This contract offers annual bonuses of up to $700,000 based on the team's on-field performance and up to $19,000 based on the team's overall GPA and graduation rate.

The authors, of course, do not advocate a thoroughly sedentary lifestyle for students nor do we deny that student employment may contribute to a student's maturity, sense of personal responsibility or greater understanding of the relationship between the cognitive concepts conveyed in the classroom and their application in the 'real world.' However, in the current environment, varsity athletes are routinely spending twenty hours a week on team activities. Moreover, students in this sample spend an average of 11.6 hours per week employed (one quarter of the sample spending at least 20 hours). The results presented here indicate that there may be many students who simply need to recalibrate their priorities, substantially reducing the time and energies spent on fitness and employment. College athletic departments and the NCAA may need to consider whether current schedules make a mockery of the term "student athlete." Just as student counseling services are routinely advising students against the harm that alcohol and drugs may be doing, these counseling professionals may need to put more emphasis on the temperance-related benefits of time management skills. Economists might preach adherence to the optimal allocation of time resources among competing goods; persons of the cloth may quote Ecclesiastes 3:1, "To every thing there is a season, and a time to every purpose under heaven."

REFERENCES

Achen, A. C. and Courant, P. N. (2009). What are grades made of? Journal of Economic Perspectives, 23, 7792.

Aitken, N. D. (1982). College student performance, satisfaction and retention: Specification and estimation of a structural model. The Journal of Higher Education, 53, 32-50.

Barr, J. (2007). Freshman dropouts. Journal of Applied Research in the Community College, 14, 105-113.

Brokaw, A. J., and Merz, T. E. (2000). The effects of student behavior and preferred learning style on performance. Journal of Business Education, 1, 44-53.

Brooks, D. (2009). No size fits all. Op-Ed, New York Times, July 17, A23.

Charkins, R.J., O'Toole, D.M., and Wetzel, J. N. (1985). Linking teacher and student learning styles with student achievement and attitudes. Journal of Economic Education, 6, 111-120.

DeBrock, L, Hendricks, W., & Koenker, R. (1996). The economics of persistence: Graduation rates of athletes as labor market choice. Journal of Human Resources, 31, 513-539.

Dunn, R. (2000). Capitalizing on college students' learning styles: theory, practice, and research. In R. Dunn, R. & S. Griggs (Eds), Practical approaches to using learning styles in higher education (pp.3-18). Westport, CT: Bergin & Garvey.

Dunn, R., Dunn, K., & Price, G.E. (1996). Productivity environmental proficiency survey. Lawrence, KS: Price Systems, Inc.

Ehrenberg, R. G., & Sherman, D. R. (1987). Academic achievement, and postcollege outcomes: A summary of results. The Journal of Human Resources, 22, 1-23

Englander, F, Terregrossa, R. A., & Wang, Z. (2009). Student behavioral choices influencing performance in introductory microeconomics. International Journal of Education Research, 4, 129-143.

Hawk, T. F., & Shah, A. J. (2007). Using learning style instruments to enhance student learning. Decision Sciences Journal of Innovative Education, 5, 1-19.

Herbert, B. (2009). Peering at the future. Op-Ed, New York Times, September 29, A39.

Horn, L. J., & Malizio, A. G. (1996). Undergraduates who work. Washington, D. C.: National Center for Education Statistics, U. S. Department of Education.

Kelley, A. C. (1975). The student as a utility maximizer. Journal of Economic Education, 6, 82-92.

Krugman, P. (2009). The uneducated American. Op-Ed, New York Times, October 9, A31.

Leonhardt, D. (2009). At colleges, too few diplomas. Business Day, New York Times, September 9: B1.

Maloney, M. T. , & McCormick, R. E. (1993). An examination of the role that intercollegiate athletic participation plays in academic achievement: Athletes' feats in the classroom. Journal of Human Resources, 28, 555-570.

Murtaugh, P. A., Burns, L. D., & Schuster, J. (1999). Predicting the retention of university students. Research in Higher Education, 40, 355-371.

Obama, B. (2009). "Remarks by the President on the 'American Graduation Initiative'," Washington, D.C. The White House, Office of the Press Secretary.

Pindyck, R. S., & Rubinfeld, D. L. (1998). Econometric models and economic forecasts. 4th Edition. New York City: McGraw-Hill.

Plant, E. A., Ericsson, K. A., Hill, L, & Asberg, K. (2005). Why study time does not predict grade point average across college students. Contemporary Educational Psychology, 30, 96-116.

Steeg, J. L., Upton, J., Bohn, P., & Berkowitz, S. (2008). College athletes studies guided toward 'major in eligibility.' USA Today, November 18. Retrieved November 15, 2009 from http://www.usatoday.com/sports/college/2008-11-18-majors-cover N.htm

Terregrossa, R. A., Englander, F., & Wang, Z. (2009). Why learning styles matter for student achievement in college economics. Journal for Economic Educators, Summer, 9, 16-31.

Telser, L. G. (1964). Iterative estimation of a set of linear regression equations, Journal of the American Statistical Association, 59, 845-862.

Vaughn, L M., Battle, J. B., Taylor, T., & Dearman, L. (2009). Learning styles and the relationship to attachment styles and psychological symptoms in college women, College Student Journal, 43, 723-735.

Weisbrod, B. A., Ballou, J. P., & Asch, E. D. (2008). Mission and money: Understanding the University. New York City: Cambridge University Press, 255.

About the Authors:

Ralph A. Terregrossa is an Associate Professor of Economics at St. John's University in New York City. He holds a Ph.D. in economics from the State University of New York at Binghamton. His current research focuses on pedagogical issues in college economics. He is the 2008 recipient of SJU's teaching excellence award.

Fred Englander is a Professor of Economics at Fairleigh Dickinson University in Madison, New Jersey. He received his Ph.D. in economics from Rutgers University. He has published articles in the Southern Economics Journal, the Business Ethics Quarterly, Science and Engineering Ethics and the Journal of Education for Business.

Zhaobo Wang is an Associate Professor of Production and Operations Management at Fairleigh Dickinson University, Madison, New Jersey. He received a Ph.D. in operations research from Rutgers University. He has published articles in the Journal of Educational and Behavioral Statistics and the Journal for Economic Educators.

Ralph A. Terregrossa

St. John's University

Fred Englander

Zhaobo Wang

Fairleigh Dickinson University

Table 1
Factor Loadings and P-values of the Extracted Factors
for the Five Discriminating Learning Style Elements

Element       Factor 1    Factor 2    Factor 3

Noise         -0.028a     0.075g      0.285g
              (0.756)     (0.408)     (0.001) *

Light         -0.337g     0.203a      0.189a
              (0.000) *   (0.023) *   (0.035) *

Design        -0.297g     -0.148g     -0.017g
              (0.001) *   (0.105)     (0.849)

Persistence   -0.615g     -0.033g     0.505a
              (0.000) *   (0.715)     (0.000) *

Intake        0.136g      0.575g      0.242g
              (0.131)     (0.000) *   (0.007) *

Element       Factor 4    Factor 5

Noise         0.103g      -0.248a
              (0.252)     (0.005) *

Light         -0.017g     0.544a
              (0.852)     (0.000) *

Design        -0.026g     0.553a
              (0.773)     (0.000) *

Persistence   0.028a      -0.055g
              (0.761)     (0.543)

Intake        -0.133a     -0.300a
              (0.140)     (0.001) *

Table 2
Alternative Learning Styles Identified by Factor Analysis

FACTOR      ASSOCIATED LEARNING STYLE

Factor 1    Strongly Global
Factor 2    Weakly Global
Factor 3    Indifferent
Factor 4    Weakly Global
Factor 5    Strongly Analytic

Table 3
OLS Regression Results for the Five Behavioral Models

                              Drinking            Internet

Variable                 Coeff     t-stat    Coeff     t-stat

Intercept                18.05    2.46 **    -9.46      -1.31

Gender
(Male=1)                 -5.67    -4.09 **    0.02       0.01

Credits-
Completed                -0.10    -3.01 **   -0.04      -1.10

SAT Total                -0.01      -1.20     0.01     2.54 *

SAT Ratio                 4.01       1.08    10.66    2.92 **

[F.sub.1]-Global
Learning Style            2.02    2.95 **     1.75    2.60 **

[F.sub.2]-Weak Global
Learning Style            0.44       0.61     2.59    3.63 **

[F.sub.3]-Indifferent
Learning Style           -0.47      -0.71    -0.58      -0.88

[F.sub.4]-Weak Global
Learning Style           -0.09      -0.11    -1.00      -1.37

[F.sub.5]-Analytic
Learning Style            0.37       0.46    -0.26      -0.32

[R.sup.2] and            0.105      0.083    0.111      0.089
Adj. [R.sup.2]

F- and p-value           4.753      0.000    5.076      0.000

                             Econ Study          Employment

Variable                 Coeff     t-stat    Coeff     t-stat

Intercept                 5.22    4.26 **    10.37       1.47

Gender
(Male=1)                 -0.21      -0.92     5.05    3.79 **

Credits-
Completed                0.004       0.68     0.09    2.63 **

SAT Total               -0.004    -4.00 **    0.01       1.00

SAT Ratio                 0.96       1.54    -8.57    -2.40 *

[F.sub.1]-Global
Learning Style           -0.25    -2.20 *    -0.58      -0.88

[F.sub.2]-Weak Global
Learning Style            0.02       0.21     0.46       0.66

[F.sub.3]-Indifferent
Learning Style           -0.24    -2.20 *     1.28     1.99 *

[F.sub.4]-Weak Global
Learning Style            0.26     2.11 *    -0.04      -0.06

[F.sub.5]-Analytic
Learning Style            0.19       1.39    -2.71    -3.50 **

[R.sup.2] and            0.100      0.077    0.116      0.094
Adj. [R.sup.2]

F- and p-value           4.486      0.000    5.298      0.000

                              Fitness

Variable                 Coeff     t-stat

Intercept                 7.30    2.82 **

Gender
(Male=1)                 -0.56      -1.15

Credits-
Completed                -0.03    -2.66 **

SAT Total                0.002      -0.89

SAT Ratio                 1.46       1.12

[F.sub.1]-Global
Learning Style           -0.26      -1.08

[F.sub.2]-Weak Global
Learning Style           -0.10      -0.41

[F.sub.3]-Indifferent
Learning Style           -0.20      -0.86

[F.sub.4]-Weak Global
Learning Style            0.18       0.69

[F.sub.5]-Analytic
Learning Style            0.03       0.11

[R.sup.2] and            0.032      0.008
Adj. [R.sup.2]

F- and p-value           1.352      0.208

Note: One asterisk denotes statistical significance at the .05 level;
and two asterisks denote statistical significance at the .01 level.

Table 4
Student Achievement OLS Regression Results

                                  Standard
Variable          Coefficient       Error    t-statistic   p-value

Intercept            2.0380        2.7057          0.75    0.4518
Drink                0.0030        0.0194          0.16    0.8766
Internet            -0.0196        0.0197         -0.99    0.3212
Econ Study          -0.2055        0.1117       -1.84 *    0.0667
Employment          -0.0375        0.0192       -1.95 *    0.0515
Fitness             -0.1746        0.0544     -3 21 ***    0.0015
F1.Global           -0.7802        0.2374     -3 29 ***    0.0011
F2.Weak Global      -0.1091        0.2507        - 0.44    0.6636
F3.Indifferent       0.4744        0.2309       2.05 **    0.0407
F4. Weak Global     -0.2813        0.2561         -1.10    0.2729
F5.Analytic         -0.3859        0.2779         -1.39    0.1658
SAT Total            0.0171        0.0018      9.39 ***    0.0001
SAT Ratio           -0.2579        1.2610         -0.20    0.8380
Credits             -0.0084        0.0122         -0.69    0.4896
Gender (male=1)     -0.2054        0.4945         -0.42    0.6782
Dummy 1              0.2400        0.5314          0.45    0.6518
Dummy 2              1.7680        0.5314      3.33 ***    0.0010
Section 1           -0.0304        0.8926         -0.03    0.9728
Section 2            0.3500        0.9649          0.36    0.7170
Section 3            2.6941        1.0016      2.69 ***    0.0075
Section 4           -0.7899        1.0570         -0.75    0.4554
Section 5           -2.3070        1.0037      -2.30 **    0.0221
Section 6           -0.5038        0.9101         -0.55    0.5802
Section 7           -1.3282        0.9429         -1.41    0.1598

   R-Square       Adj R-Square   F-statistic     p-value
    0.3437           0.3007       7 99 ***        .0001

Note: One, two and three asterisk(s) denote statistical significance at
the .10, .05, and .01 level, respectively.

Table 5
Joint F-Tests for Alternative Clusters of Explanatory Variables

Cluster of Variables           Behavior      Learning Styles

Degrees of Freedom: q/n-k      5/351         5/351

F -statistic                   4.36 *        3.63 *

F-critical values at 10%, 5%   3.115, 4.38   3.115, 4.38

Cluster of Variables           Academic Ability

Degrees of Freedom: q/n-k      2/351

F -statistic                   17.75 **

F-critical values at 10%, 5%   5.14, 8.535

Notes: (1) The Joint F-test degrees of freedom in the numerator and
denominator equal the number of restrictions (q), or the number of
variables within each group, and the number of observations (n=375)
less the number of parameters (k=24), respectively. (2) F-critical
is reported for the ten and five percent levels of significance,
respectively. (3) One and two asterisk(s) denotes statistical
significance at the .10 and .05 level, respectively.