Michigan GEAR UP:
What the Evaluation Measures and What It Has Found

Findings from the 2025 and 2026 evaluation
At a glance
  • Nearly all surveyed students plan to finish high school, and about four in five plan to continue their education after it. Of the nine confidence questions on the survey, students are least sure about paying for college. Section 3
  • Campus visits are the activity most consistently linked to students' confidence. The survey models, the staff survey, and the site interviews agree on this. Section 4
  • Relative to students at the same middle schools from the year before MI GEAR UP implementation, the initial MI GEAR UP student cohort had higher rates of on-time enrollment and attendance in grade 9, higher PSAT 9 and lower PSAT 10 scores, and no statistically significant differences in coursetaking. Section 5
  • Seniors at MI GEAR UP high schools completed the Free Application for Federal Student Aid (FAFSA) at higher rates than seniors at comparable, high-poverty high schools in Michigan, but below statewide and national averages, for each graduating class from 2018 to 2024. The lead over comparable schools was already there before the program reached these high schools. Section 6

1Where Michigan GEAR UP works and whom it serves

Michigan GEAR UP (Gaining Early Awareness and Readiness for Undergraduate Programs) is a federally funded program that partners universities with school districts across the state to prepare economically and academically disadvantaged students for postsecondary education.

The Michigan Department of Labor and Economic Opportunity coordinates it statewide, and each of Michigan's 15 public universities runs a site.

Sites work with schools where at least half of students qualify for free or reduced-price lunch, and they follow a cohort of students from 7th grade through high school graduation.

17,104
students enrolled
89,989
programming hours recorded
60%
of enrolled students with recorded participation
Program records, 2024–25 school year. Dots mark the partner universities' campuses, not the schools they serve; hovering a dot shows that site's full name and city.
Source
SCRIBE, the program's administrative data system; statewide totals for the 2024–25 school year. 60% participation = 10,262 of 17,104 enrolled students with recorded participation; programming averaged 8.8 recorded hours per participating student. Recording completeness varies across partners. In the first cohort, 23% of students had zero recorded hours, which can mean either non-participation or unrecorded participation.

The current grant cycle runs from 2020 to 2027. Its first cohort entered in 2020–21 as 7th graders or, for about two-thirds of the cohort, in 2021–22 as 8th graders, where COVID delayed the start. That cohort reached 12th grade in 2025–26 as the class of 2026, so the outcome findings on this page are preliminary. MI GEAR UP has operated in grant cycles since 1999; this page covers the current cycle. The numbers above describe the program in the 2024–25 school year.

2How the evaluation gathers information over the school year

Evaluators assess school programs on a yearly cycle, drawing on records, surveys, interviews, and site-visit observations, and return findings to the sites annually.

Fall

Sites occasionally hold fall events, such as the September 2025 Kick-Off to College, where students met with 21 colleges and other higher-education organizations. The evaluation surveys students before and after the events to see whether and how their answers change.

Winter

Site visits and a statewide staff survey take stock of how each site delivers the program. Each site's own student-survey results are reviewed with that site's staff.

Spring

The statewide student survey is offered to all participating students (in 2024–25, responses came from about one in six of them); sites also receive one-page summaries of their own service numbers from the program's records.

Summer

Evaluators present findings at the annual MI GEAR UP conference.

What the statewide student survey asks

Part of GEAR UP's charge is to raise students' expectations and knowledge of how to prepare for and pay for college, and over years of interviews, students said the program reinforced what they believed was possible for them, and for some it changed those beliefs. The confidence items follow a published definition of college-going self-efficacy (Gibbons & Borders, 2010).

1Plan to finish high schoolone question
"I plan to finish high school." Answered Yes / No / Not sure.
2Plans after graduationone question · eight options
"Which option best describes your plans for the fall after you finish high school?" Attend a 2-year or community college Attend a 4-year college or university Attend a certificate, technical, or trade program Attend a military academy Enlist in the military Work full- or part-time Volunteer work (AmeriCorps, PeaceCorps, etc) Other (please describe)
3College-going self-efficacy (CGSE)nine statements · rated 1–4
"How sure are you about being able to do the following:" Each item rated 1 (Not at all sure) to 4 (Very sure). I can find a way to pay for college I can get accepted to a college I can have family support for going to college I can choose a good college for myself I can get funding for college (example: scholarship, grant, financial aid) I can make my family proud with my choices after high school I can pay for college even if my family cannot help me I can choose the high school classes that will help prepare me for college I can go to college after high school
4Self-efficacy for regulated learning (SERL)eleven questions · rated 1–4
"How well can you…" Each item rated 1 (Not well at all) to 4 (Very well). …finish homework assignments by deadlines? …study when there are other interesting things to do? …concentrate on school subjects? …take notes during class? …use resources available to you (example: internet, books, library) to get information for class assignments? …organize your schoolwork? …plan your schoolwork? …remember information presented in class and textbooks? …arrange a place to study where you can focus? …motivate yourself to do schoolwork? …participate in class discussions?
Each spring's survey reaches whoever chooses to respond that year, so each year's respondents (each survey 'wave') are a different mix of students. Section 3 reports the answers.

3What students plan and how confident they feel about college

Students' answers from the 2024–25 statewide survey are reported below in three groups, in the order they were asked in the survey.

97.9% of surveyed students said they plan to finish high school.
(2024–25 statewide survey)
Students' plans for the fall after graduating high school, 2024–25 wave
Within each chart, items run from highest to lowest. About 1,700 students chose to take part, a minority of the program's participants. Section 6 turns to a related measure from school records: how many seniors complete the FAFSA, the federal financial-aid application.
Source
2024–25 statewide student survey; both scales from the same survey, wave, and respondents. About 1,700 respondents in the 2024–25 wave (item-level n: 1,635 on the finish-high-school question, 1,686 on plans, roughly 1,550–1,605 on the confidence items). The plans chart reports the survey's eight options as six categories: the two military options combined, work and volunteer combined. Respondents opted in; responses are about one in six of the 10,262 students with recorded participation in 2024–25 (section 1).

Students eligible for free or reduced-price lunch are less confident about paying

In the 2023–24 survey, students eligible for free or reduced-price lunch scored lower on affording college, a modest but statistically significant gap (−0.18 on the 1–4 confidence scale).

From statistical models of the statewide survey
Source
CGSE regression models on the 2023–24 survey wave, respondents matched to program records; Afford College scale, n = 1,505.Free or reduced-price lunch is the only student characteristic statistically significant on the Afford scale (−0.180).

The same item ranked last in two site surveys

At the two sites that surveyed students in winter–spring 2026, the lowest-rated item was the same item about paying for college that ranked last statewide ("I can pay for college even if my family cannot help me"), at around 2.0 on the 1–4 scale, with over 40% of respondents "not at all sure."

From two site surveys
Source
Surveys fielded at two sites, February–April 2026; 289 students combined, answering the same CGSE items as the statewide survey.

Confidence is linked to college plans

Students who scored higher on either confidence scale (confidence about attending college, or confidence about paying for it) were more likely to plan on pursuing a four-year degree. The link was stronger for confidence about attending college than for confidence about paying for it.

From statistical models of the statewide survey
Source
Regressions of four post-high-school plan categories on each confidence scale (2023–24 survey wave, n = 1,364), controlling for demographics and participation. Higher scores on either scale are linked to plans for a four-year college, and the estimates for confidence about attending run consistently larger: beliefs about being able to attend were more strongly associated with plans than beliefs about being able to afford.

"I think for paying for college, it's scary… but not as frequently when GEAR UP shows you different ways to pay for college, or little scholarships…"

Student, 2025 focus groups

4Which program activities are linked to students’ confidence

This section asks which program activities are linked to higher student confidence about attending and affording college. Three sources of information help answer this question: the survey of students, the staff survey, and the site-visit interviews. The cards below give each source's results.

Student survey models

Campus visits were the only activity associated with higher confidence in both attending (+0.127) and affording (+0.075) college. These are differences in points on the 1–4 confidence scale.

Source
CGSE regression models on the 2023–24 survey wave; n = 1,505 respondents matched to program records. On the Afford scale the campus-visit association is statistically significant, close to the .10 threshold (p = .094).The factor analysis behind the scales assigns seven items to the Attend scale and three to the Afford scale, with "I can get funding for college" loading on both.

Staff survey

18 of 18 staff who answered named campus visits among their site's highest-impact activities; financial aid advising (14) and postsecondary counseling (12) followed. When asked which outcomes each service affects, staff linked campus visits to more outcomes than any other service. When asked where the program has contributed most, staff rated postsecondary expectations and preparation highest (4.6 and 4.5 on a 1–5 scale) and school attendance lowest (3.2).

Source
Winter 2026 staff survey: 19 complete responses covering 13 of 15 sites (87%); respondents are mostly site leadership (directors, coordinators, PIs).The highest-impact item was answered by 18 of the 19; the contribution ratings by 17 of the 19 ("Based on your observations, to what extent has MI GEAR UP contributed to improvements in the following areas?", 1 = no improvement, 5 = major improvement).

Site visits

12 of 15 sites described campus visits as critical to building students' college-going confidence.

Source
Winter 2026 site visits: all 15 sites; interviews with site staff, analyzed thematically.Staff-survey respondents and site-visit interviewees overlap, so the two are read as converging lines of evidence, not independent confirmations.

"Just eating in the dining hall and walking around campus makes college feel real and achievable for our students."

Site staff, Winter 2026 site visits

The survey models behind the first card cover nine activities, not just campus visits. The chart below shows all nine on one scale, once for each of the two confidence scales: attending college and affording it.

Differences in confidence by activity, for each of the two confidence scales
How to read this chart: each dot represents the difference in average confidence ratings, on the 1–4 scale, between students who took part in an activity and students who did not. These results come from linear regressions on the 2023–24 statewide survey that include all nine activities together and adjust for student characteristics (gender, race and ethnicity, and free or reduced-price lunch status) (n = 1,505). On this page, a difference is called statistically significant when a difference at least this large would arise by chance less than one time in ten if there were no true difference (p < .10 in the legend). This is a looser threshold than the one-in-twenty (.05) often used. Filled dots are statistically significant; the line through each dot is its 90% confidence interval, the range of values consistent with the data.
Statistically significant (p < .10)Not statistically significant (p ≥ .10)90% confidence interval

Confidence about attending college. Students who went on campus visits or field trips reported more of it (+0.127 and +0.110 points on the 1–4 scale); students who received tutoring reported less (−0.105). The other six activities show no statistically significant difference.

Attend College scale (CGSE subscale)
Both charts show all nine activities in the models. Significance: * p < .10, ** p < .05, *** p < .01. All results are associations: respondents chose to take the survey, and students were not assigned to activities at random.
Source & method
The nine activity indicators (counseling, summer programs, workshops, mentoring, financial aid, field trips, tutoring, job shadowing, campus visits) mark any recorded participation (any hours) in the program's administrative records, linked to each survey respondent; the survey itself does not ask about activities. Sample: the 2023–24 survey wave matched to those records (n = 1,505). The models control for gender, race/ethnicity, and free or reduced-price lunch status.

What these links can and cannot tell us

  • Although the survey models show a statistically significant association, with students who went on campus visits reporting more confidence than students who did not, that association by itself cannot tell us whether the visits made them more confident. Statistical significance says only that a difference at least this large would be unlikely if there were no true difference; it does not say where the difference came from. Students were not assigned to visits at random, so the two groups the models compare may have differed before any visit: students already set on attending college, or those who were already more confident, may have engaged in activities more often than their peers. The confidence gap could reflect the visits, or who went on them, or both, and these data do not separate those possibilities.
  • By the same logic, the negative result for tutoring (−0.105, statistically significant) does not mean tutoring lowers confidence. Tutoring participants are not a random group, and tutoring is concentrated in a few schools. The data show that tutoring participants report lower confidence than other respondents on average; they do not show why.

5Analyzing academic outcomes for the initial MI GEAR UP cohort

Next, we move from what students say to what administrative records show, for the first cohort of GEAR UP students. The records answer two questions, each with its own comparison. Both sets of results are purely descriptive, indicating associations but not causal relationships. Each set is split into tabs by outcome.

Overall participation compares the first GEAR UP cohort with the class that went through the same schools just before the program arrived.

Activity by activity asks which of the program's activities are linked to better outcomes; each activity ran at only some schools, so each one is compared only within those schools, against the class that went through them just before.

Overall participation
Participants compared with the class before them: enrollment, attendance, PSAT scores, college readiness, coursework
Each dot is the adjusted difference between MI GEAR UP participants (the first cohort: 5,392 students who entered in grades 7–8 in 2020–21 and 2021–22) and a comparison group of students at the same schools in the year just before GEAR UP arrived. Outcomes were tracked through 2023–24.
Statistically significant (p < .10)Not statistically significant (p ≥ .10)90% confidence interval

Enrollment and attendance. Participants enrolled on time in grade 9 and attended school at slightly higher rates than the class before them: roughly 2 percentage points where the difference is statistically significant, nearly 3 on grade-9 attendance, and no statistically significant difference in grade 10 alone.

Enrollment and attendance (percentage points)
Significance: * p < .10, ** p < .05, *** p < .01. With this many outcomes tested, a few significant results can arise by chance. The differences are associational; attendance, PSAT, and coursework are observed for students who enrolled on time.
Source & method
Michigan K-12 administrative records linked to the program's participation records. Groups: the 5,392 participants include the 23% with zero recorded hours. The comparison group: 7,091 students at 59 middle schools, under the 14 partners with middle-school programming in those years. Of these students, 36% later joined GEAR UP in high school, which, if the program has any effect, pulls the estimated differences toward zero. Models: control for gender, race/ethnicity, economic disadvantage, special education, and English-learner status; standard errors clustered by school; binary outcomes reported as percentage-point differences (average marginal effects). Timing: the two groups reach grades 9–10 in different calendar years (participants 2022–24, comparison 2021–23, closer to the COVID-19 peak). Sensitivity: re-running everything excluding the 23% zero-hour students (4,173 participants) gives broadly similar results, with enrollment and attendance associations somewhat larger, at the cost of potential bias from dropping students who may differ systematically.
Activity by activity
The same comparison for six activities, one outcome at a time: enrollment, attendance, PSAT 9, PSAT 10, honors courses
Each dot is the adjusted difference between students who took part in one activity (2020–22) and a comparison group of students at the schools that ran that activity, in the year just before GEAR UP arrived. Outcomes were tracked through 2023–24. Because each estimate rests only on the schools that ran that activity (as few as 10), the confidence intervals are much wider than in the first set. The "Any participation" row repeats the overall estimate from the first set above (all 5,392 participants against the full comparison group), where the same outcome appears.
Statistically significant (p < .10)Not statistically significant (p ≥ .10)90% confidence interval

On-time grade-9 enrollment. Participants in every activity enrolled on time more often than the class before them, by 2 to 5 percentage points; campus visits and financial aid show the largest differences (+5.4 and +5.3).

On-time grade-9 enrollment (percentage points)
Six of the eleven activity types in the 2020–22 records are shown: the three most common (workshops, campus visits, financial aid) and the three most intensive (tutoring, summer programs, field trips). The other five reached too few students to be analyzed separately. Sites run more kinds of activities than the records capture. Activity categories follow the 2020–22 records scheme. That is a different measure, on different years and students, from the survey models in section 4. Comparisons across activities need care: participation is self-selected and baselines differ by activity (campus-visit participants, for example, were 16 percentage points more likely to be White and 7 points less likely to be economically disadvantaged); activities overlap within students; and some concentrate in one university partner (tutoring: 10 schools under one partner, 99 hours per participant). Financial aid here averaged about 1 hour per participant.
Source & method
Activity types follow the 2020–22 records scheme (eleven types); the five omitted are counseling, job shadowing, mentoring, academic support, and other. Participant counts range from 292 (summer programs) to 3,241 (workshops). Baselines for scale (comparison-group means across the six activities): on-time 9th-grade enrollment 94–95%, grade-9 attendance 79–83%, PSAT 9 totals 749–789.

6A school-level analysis of FAFSA completion

This section looks at FAFSA (Free Application for Federal Student Aid) completion: whether graduating seniors at GEAR UP high schools file the federal financial-aid application. FAFSA completion is a core outcome of this evaluation: it is part of the college readiness GEAR UP aims for, and an intended outcome of the program’s financial-aid activities. Here, “GEAR UP high schools” means the 70 high schools designated for the program, of which 50 to 62 appear in the federal files in any one class (see Source). Filing also matters to students directly: it is one requirement of the MI GEAR UP Scholarship, worth up to $5,000 per year (lifetime maximum $10,000) for the class of 2026. The first figure below takes a single class, the class of 2024, and builds the comparison step by step: GEAR UP high schools against the state, the nation, and comparable Michigan schools, then the differences between individual schools within each group. The second figure carries the same comparison across every class since 2018.

Source
MI GEAR UP Scholarship terms as published on the State of Michigan's student-aid pages (michigan.gov/mistudentaid, retrieved August 2026): filing a current-year FAFSA is among the eligibility requirements; for the 2026 cohort the maximum award is $5,000 per academic year, with a lifetime maximum of $10,000.

How to read this chart: keep scrolling to build the comparison one step at a time; the text beside the chart explains each step. The first two steps show each group's overall rate. From the third step on, each dot is one school, placed by its completion rate (in the national row, each square stands for about 50 schools), and the vertical line is the group's average. Point at any dot for its details.

Each rate is total completions divided by total seniors, computed the same way for every group (details under Source).
Across the seven classes, the completion rate at GEAR UP high schools ran 0.8 to 8.4 percentage points below the national rate, 1.1 to 6.7 below the statewide rate, and 0.4 to 3.5 above the rate at comparable Michigan schools. Within every group, in every class, rates at individual schools range from under 20% to over 75%; the middle half of GEAR UP high schools (from the 25th to the 75th percentile) spans 19 to 23 percentage points in every class.
Source
How each rate is computed: the number of a school's seniors who completed a FAFSA by June 30, from the U.S. Department of Education's school-by-school FAFSA files, divided by that school's 12th-grade enrollment in federal enrollment data (NCES Common Core of Data). All four groups are computed this way, from public schools only. Rates are shown to one decimal; differences between groups are computed before rounding, so they can differ by 0.1 from the difference of two rates as shown. Schools covered: the schools present in both federal files: 50–62 of the 70 GEAR UP high schools per year, 632–727 Michigan public high schools, and 18,006–19,420 high schools nationwide. The FAFSA files carry school ID codes only from the class of 2023 on; earlier classes are matched by school name, so earlier years cover somewhat fewer schools. Each of the fifteen university-anchored sites serves between 2 and 9 of the 70 designated high schools; a site or partner average would hide the differences between schools that the figures show. Comparable schools: Michigan public high schools outside GEAR UP where at least 50% of students were economically disadvantaged that school year, per state administrative records; 292–370 schools per year, with membership assigned year by year. GEAR UP schools average 27 percentage points more economically disadvantaged students than other Michigan public high schools. The groups are nested: statewide includes the GEAR UP and comparable schools, and the national group includes all of them. Class of 2024: GEAR UP 43.2% vs. comparable schools 40.7%. Site snapshots: at the two sites that fielded their own student surveys in winter–spring 2026, 49% and 74% of seniors said they had already completed the FAFSA. These are two small self-selected samples, each reflecting only its own site, consistent with the wide differences between schools. Programming reaching 12th graders (the 5.2% and 13.0% in the closing paragraph): from the program's SCRIBE participation records, as tabulated in the FAFSA report: among the schools where any 12th grader took part in any GEAR UP activity (25 schools in 2022–23, 35 in 2023–24), 5.2% and then 13.0% of 12th graders had a financial-aid activity on record; 12th graders with recorded hours averaged about 6 program hours in the year, and those with financial-aid hours about 2 to 3. Financial-aid participation sets (behind the closing paragraph of this section): classes of 2018–2024 trends compared across four nested sets of schools, from the program's SCRIBE participation records. This is a separate analysis from the federal method above, and unrelated to the four comparison groups in the chart. The widest set is all 70 designated high schools. The second is the 59 schools with at least one high-school participant on record in 2022–24. The third set is the 34 schools where at least one current student had ever taken part in financial-aid services, including in middle school. The narrowest set is the 12 schools where a 2022–24 senior personally received such services. Because the third set counts middle-school years, it is only approximately a subset of the second. If these rates differ from figures published elsewhere: there is more than one legitimate way to count a school's 12th-graders (on a single day or over the school year, from state or federal records, including or excluding private schools), and each yields a somewhat different completion rate. Both figures use one count, the federal enrollment data described above, for all four groups, because the groups can be compared only if their rates are computed the same way. A rate built on a different count can sit a few points higher or lower.

GEAR UP schools' lead over comparable schools could be read as a program effect. FAFSA completion is a step in paying for college, the area where surveyed students report the least confidence (section 3), and the lead holds in every class. The second figure argues against that reading: GEAR UP schools were already ahead in the classes that graduated before the program reached these high schools. Financial-aid programming has been limited so far. It reached 5.2% of 12th graders at the 25 schools where any senior took part in any GEAR UP activity in 2022–23, and 13.0% at the 35 such schools in 2023–24, a few hours per student, likely too little to move a whole school's completion rate. Nor does programming explain why schools differ so much: only 12 of the 70 high schools had seniors receive financial-aid services in 2022–24, and those 12 schools rose and fell with the rest, while running roughly 8 to 15 percentage points below the wider sets of GEAR UP schools in every year, including the years before the program’s high-school work began.

The frame the evaluation works from

Equity requires more than just program implementation. It requires evaluation systems that reveal where opportunity gaps exist and support collaborative improvement across diverse communities.