# Summary Analysis of the Michigan Merit Scholarship Program

Donald E. Heller

Assistant Professor of Education

University of Michigan

In the spring of 1999, approximately 124,000 high school students throughout Michigan took the Michigan Educational Assessment Program (MEAP) High School Tests (HST). The tests were given in four subject areas: mathematics, reading, science, and writing. Students are not required to take the tests.

In June 1999, the Michigan Merit Award Scholarship Act was signed into law by Governor Engler. The law provides scholarships of up to $2,500 for students who score at Level 1 (exceeds Michigan standards) or Level 2 (meets Michigan standards) on all four subject area tests. All students in Michigan, regardless of family income or other characteristics, are eligible for the awards. In the fall of 1999, the Michigan Department of Education released a datafile with the spring HST results. The datafile contains information about each student, including racial/ethnic background, school attended, and test results. These data will be used by the Michigan Merit Award Board, the state body charged with responsibility for awarding the scholarships, to determine scholarship eligibility.

I used the HST datafile to analyze the characteristics of the students who are eligible for a scholarship based on the HST eligibility test. This report presents a brief summary of my findings.

The first step was to conduct bivariate analyses of the relationship between whether a student was eligible for a scholarship and various characteristics of the student and school he or she attended. Racial/ethnic background is strongly correlated with scholarship qualification. Table 1 shows for each group the total number of students in the datafile, the number that took all four tests (i.e., were eligible for a scholarship), and the number that passed all four tests at Level 1 or Level 2 (i.e., qualified for a scholarship).

Table 1 demonstrates that 21.9% of White students who took at least one of the subject area tests qualified for a scholarship by scoring at Level 1 or Level 2 on all four tests (column D). In contrast, 2.8% of African American and 11.3% of Hispanic students qualified for a scholarship. The White qualification rate was 7.8 times greater than the African American rate and 1.9 times greater than the Hispanic rate. If you examine the scholarship qualification rate of only those students who took all four tests and thus were eligible for a scholarship (column E), the White rate was still 4.8 times greater than the African American rate and 1.7 times the Hispanic rate. While not shown in Table 1, the bivariate analysis found that male students were less likely to qualify for a scholarship than were female students.

Table 2 presents similar information for students in schools with different free and reduced lunch rates. Free and reduced lunch percentages are often used as a proxy for the relative average income levels of the families with students attending the school. For the purposes of this analysis, schools in the database were divided into quartiles, with the schools with the highest percentages of students on free or reduced lunch in the first quartile and schools with the fewest students receiving free or reduced lunch in the fourth quartile.

Table 1: Scholarship Eligibility by Racial/Ethnic Group

Racial/Ethnic Group |
Total # of students |
# taking all 4 tests |
# qualifying for scholarship |
% of total qualifying for scholarship (C/A) |
% of eligible qualifying for scholarship (C/B) |

Native American | 942 | 537 | 104 | 11.0% | 19.4% |

Asian/PacificIslander | 1,719 | 1,068 | 424 | 24.7% | 39.7% |

African American | 15,916 | 6,388 | 448 | 2.8% | 7.0% |

Hispanic | 2,121 | 1,175 | 239 | 11.3% | 20.3% |

White | 58,081 | 37,458 | 12,710 | 21.9% | 33.9% |

Multiracial | 1,691 | 966 | 248 | 14.7% | 25.7% |

Other | 2,283 | 1,288 | 306 | 13.4% | 23.8% |

Missing | 41,597 | 17,539 | 5,659 | 13.6% | 32.3% |

Total |
124,350 | 66,419 | 20,138 | 16.2% | 30.3% |

Table 2 shows that the scholarship eligibility results are very strongly correlated with the percentage of free or reduced lunch students in the school. The fewer students on free or reduced lunch, the greater is the percentage of the students who qualified for the scholarships. Students in schools with the fewest free or reduced lunch recipients were more than three times more likely to qualify for a scholarship (column D) than were students in the schools with the highest number of free or reduced lunch recipients.

Table 2: Scholarship Eligibility by School Free or Reduced Lunch Percentage

School Free Lunch Quartile (range, median) | Total # of students | # taking all 4 tests | # qualifying for scholarship | % of total qualifying for scholarship (C/A) | % of eligible qualifying for scholarship (C/B) |

1st (26.6% - 100%, 41.8%) | 29,742 | 12,486 | 2,203 | 7.4% | 17.6% |

2nd (12.6% - 26.2%, 18.8%) | 29,813 | 18,186 | 4,740 | 15.9% | 26.1% |

3rd (5.9% - 12.6%, 9.1%) | 30,008 | 17,632 | 6,135 | 20.4% | 34.8% |

4th (0.3% - 5.9%, 3.0%) | 29,338 | 16,094 | 6,781 | 23.1% | 42.1% |

Missing | 5,449 | 2,021 | 279 | 5.1% | 13.8% |

Total |
124,350 | 66,419 | 20,138 | 16.2% | 30.3% |

Note: The size of each free lunch quartile is not exactly 25% of the total, due to the number of students right at each cutpoint.

The second step in the study was to conduct a multivariate analysis using logistic regression to examine the relationship between the predicted probability that a given student would qualify for a scholarship, and a combination of characteristics including the student's race and gender, and the percentage of free or reduced lunch students in the school. Multivariate analyses allow the researcher to examine the relationship between a particular outcome (in this study, the probability that a student would qualify for a Michigan Merit Scholarship) and another variable, controlling for the effect of other variables.

The results of the multivariate analysis indicate that even controlling for other factors, race and ethnicity are important predictors of the probability that a given student would qualify for a scholarship. For example, the predicted probability that an African American student would qualify for a scholarship was 16 percentage points lower than that of White students, controlling for the student's gender and the percentage of free or reduced lunch recipients in the school. Similarly, the predicted probability that a Hispanic student would qualify for a scholarship was five percentage points lower than White students. Asian American students had a predicted probability of qualifying for a scholarship that was 14 percentage points greater than Whites. Gender remained an important factor in the multivariate analysis, with male students five percentage points less likely to qualify for a scholarship than female students.

Looking at the free and reduced lunch proportions, every 10 point increase in the percentage of students receiving free or reduced lunch was related to a five point drop in the predicted probability that a given student would qualify for a scholarship, controlling for the other factors.

Certainly students' own academic abilities and experiences, and their ability to perform well on standardized tests, are important factors in performance on the MEAP tests. However, the results presented in this study demonstrate that a student's race and gender, and the percentage of students receiving free or reduced lunch in their school, are powerful predictors of whether they will qualify for a Michigan Merit Award Scholarship.

Notes

The law provides alternative methods for earning a scholarship, including scoring above certain levels on ACT or SAT standardized tests, but it is expected that most students will attempt to qualify for the scholarships by taking the MEAP tests.

The full $2,500 scholarship is awarded for students attending college or some other form of postsecondary training in Michigan. Students attending out-of-state institutions are eligible for a $1,000 award.

The differences in eligibility rates among the races are statistically significant at a level of p<.001. More information regarding the statistical analyses conducted is available from the author.

Data on the percentage of students in each school receiving free or reduced lunch were obtained from the Michigan School Report, compiled by the Michigan Department of Education.

The differences among the quartiles are statistically significant at a level of p<.001.

© 2000, Donald E. Heller

**Nearly 1000 four-year colleges and universities do not use the SAT or ACT **to admit substantial numbers of bachelor-degree applicants.