Class Size Effects on Student Performance in a Hispanic-Serving Institution

Bellante (1972), Craig, et al. (1979), Saunders (1980), and Myatt and Waddell (1990) find student performance and class size to be statistically unrelated. Crowley and Wilton (1974), Raimondo, et al. (1990), and Stratton, et al. (1994) find a negative, albeit statistically small or weak, relationship. Differences in performance due to gender effects are examined, again with mixed results, in Lumsden and Scott (1987), Siegfried (1979), Williams, et al. (1992), Horvath et al. (1992), and Anderson et al. (1994).

Raimondo et al. (1990) count the number of students who are Asian American, African American or Hispanic when describing their data set, but do not include any race variables in their regression analysis. Borg and Shapiro (1996) include a dummy variable for African Amercian, Lopus (1997) includes a dummy variable for white students and McCoy et al. (1991) include a white dummy variable in their course sequencing model. None of these studies showed race to be statistically significant.

The Fall 2009 enrollment included 43 % Hispanic, 36 % white, with Native American, African American and Asian American students represent approximately four percent, three percent and one percent of the student population, respectively. Women and men represented 55 % and 45 % of enrollment respectively.

U.S. Department of Education, “Action Plan for Higher Education: Improving Accessibility, Affordability and Accountability,” September 26, 2006, http://​www.​ed.​gov/​news/​pressreleases/​2007/​09/​09262007a.​html. As recently as September 26, 2007, U.S. Secretary of Education Margaret Spellings stated in a press release to announce the awarding of $17.2 million in new grants to “colleges and other postsecondary schools that enroll a high percentage of Hispanic student,” that “Thanks to No Child Left Behind, the achievement gap is closing for Hispanic students and academic progress is on the rise…At the higher education level it’s a different story where Hispanic students still lag behind their peers in earning a post-secondary credential.”

This is course taken by non-business majors in fulfillment of general education requirements. It is divided into three parts, the first being general economic concepts (scarcity, opportunity cost, supply and demand), followed by discussion of microeconomic concepts (production costs, and profit maximization by competitive firms and the monopolist). The last unit is devoted to macroeconomics (aspects of the Aggregate Demand-Aggregate Supply model, and fiscal and monetary policy).

The semesters were Spring 1994, Fall 1995, Fall 1996, Spring 1997, Fall 1997, Spring 1998, Fall 1998, Spring 1999, Fall 1999, and Spring 2000.

The textbook used was Tucker, Survey of Economics (Southwestern Publishers). All students took short-essay-question tests.

Kennedy and Siegfried (1997) maintain that a study by Levin (1967) has results that are “clouded by the fact that in his study different instructors taught the large and small classes.” p. 387. The results obtained from a single instructor should alleviate statistical noise due to varying institutions and instructors.

To test the null hypotheses of no change, several Wald tests were conducted. The multiplicity of tests enabled a sensitivity of the Wald procedure to alternative specifications of the two required sub-samples. In all cases, the null hypothesis was rejected. For example, when the earlier period was defined to include Spring 1994 through Fall 1997, and the later period Spring 1998 through Spring 2000, the Wald test value was 1.20.

Students with less than 60 h of credit and a GPA of less than 2.0 may elect to have credit for courses with less than a grade of C deleted from the official GPA. The effect on the transcript is similar to transfer credit—the course and the grade remain on the transcript but not used in GPA calculation.

After 1997 students were also asked how many hours worked for pay during the semester. Students were also asked to provide ACT scores, but not all students are required to take this exam and many indicated they could not remember their scores or left the question blank.

The data are structured in varying levels of classes within years and also classes within different times of the year they were taught. A detailed analysis of possible nesting effects could yield more information about mechanisms driving the results. We note that Table 1 does not indicate extreme clustering of effects among any of the groups considered, with the possible exception of the postgraduate expectations of Hispanic males. Since the sample is relatively small, inclusion of more complex methods of controlling for nesting effects would result in a loss of significance. The straightforward approach taken in the analysis identifies a substantial result which we believe would not dissipate under a more refined analytical treatment. What may be generalized from these findings is that class size can wield a significant differential impact on minority students. Whether or not the mechanisms are fully understood, greater care should be exercised in considering the impact of class size on targeted student communities.

Withdrawals never exceed 10 students in any given semester and account for an average of 11 % of total enrollees.

The bivariate-probit sample selection model here applied is formally discussed in Vela (1998). For simplicity, use of subscripts has been kept to a minimum. The unit of analysis is the individual student.

Plots of course grades and class size (before and after student withdrawals from the course) suggested any relationship but linear.

This coding scheme is admittedly arbitrary. The College of Business has one classroom that seats 100 students. Two-thirds of the other classrooms seat 60–70 students and one-third seat no more than 30. Thus Class 1 is basically consists of those classes with 30 or less students, Class 2 consists of those classes with between 60 and 70 students and Class 3 is those classes with 100 students. The categorization of class size also tends to be somewhat arbitrary in the literature and more a function of the physical capacity unique to each institution. The divisions used here are comparable to those used in other studies. Kennedy and Siegfried (1997) cite several studies, such as Nelson (1959) where small and large classes are set within a range of 16–35 and 85–141 respectively. Bellante (1972) has small classes of 25–35 and large classes of 200–310. Raimondo et al. (1990) defines small classes as 25–35 and large classes as 200–350. Romer (1993) sets small classes as the bottom 33 % of all class sizes in his sample and large as the top 33 %.

See (Greene 2000; 322–325). A classic source on the use of spline regression is Piorier (1976). An application of spline regression to an analysis of class size is found in Akerhielm (1995). One of the merits of a spline specification over dummy variable scheme is that the well-known dummy variable trap can be avoided without the need to omit the reference group. Thus, parameter estimates of all class size-gender/ethnicity groups can be directly estimated.

See Anderson, et al. (1994) for a similar argument.

Additionally, for some students, their degree plan either required or highly recommended completion of this course. To incorporate this influence into their decision-making model, Z_a initially included a set of 13 indicators of degree plan. The inclusion of these indicators contributed very little to the explanatory power of the model. The estimates reported here were computed without these degree plan indicators.

Tabulations of degree plans by gender and ethnicity reveal, for example, that the nutrition, nursing, and education majors are heavily populated by females. The proportions of Hispanics in criminal justice/government and hospitality and tourism are also above average.

This would be the case if in the group of excluded students are above average achievers who are first time students, such as Freshmen and who would not have a GPA to report, as well as above average achieving upperclassmen who withdrew from this general education course because other courses required more attention and because they sought to protect their GPA.

Students with majors in Animal Science, Nutrition, Chemistry, Biology, and Computer Science, consistently, achieved grades of 85 or better.

For a very interesting analysis of the impact of learning style difference on course grades in economics see Borg and Shapiro (1996) and Borg and Stranahan (2002). Also see Charkins, et al. (1985).