Reaching out first costs more per student and less per degree

In randomized trials, the programs that put a person on each student, an adviser, a coach, or a tax professional, kept more students enrolled and nearly doubled graduation. The largest spent 63% more per student and cut the cost of each degree earned by 11.4%; its Ohio replication produced each additional graduate for about $51,000, versus $99,000 under usual services. The same outreach run as broadcast messaging across 800,000 students produced no effect on aid receipt or enrollment for any subgroup.

Published August 2026. Every figure carries a named source or its arithmetic.

In short

  1. CUNY's Accelerated Study in Associate Programs (ASAP) nearly doubled three-year graduation in a randomized trial: 40.1% versus 21.8% (MDRC, 2015).
  2. ASAP spent $16,284 more per student, 63% more. The cost per degree earned fell 11.4%, from $118,248 to $104,825 (MDRC, 2015).
  3. Ohio's replication produced each additional graduate for a published $51,000. A graduate under usual services cost about $99,000 (MDRC, 2020).
  4. InsideTrack coaching at roughly $500 a semester raised 12-month enrollment by 5.3 percentage points across 13,555 students (Bettinger and Baker, 2014).
  5. Campaigns messaging more than 800,000 students about the federal student aid application, the FAFSA, produced no effect on aid receipt or enrollment for any subgroup (Bird et al., 2021).
  6. Five years of texting and online-coaching experiments across nearly 25,000 students moved no academic outcome (Oreopoulos and Petronijevic, 2019).

In two randomized trials, graduation nearly doubled

Every figure here comes from a randomized controlled trial: a study that admits students to a program by lottery, so the two groups start alike and any gap that opens later belongs to the program. Students left in usual services are the control group.

The largest gap on record belongs to CUNY's Accelerated Study in Associate Programs, known as ASAP. The evaluation firm MDRC ran the lottery in 2010 with 896 low-income students at three community colleges, all placed into developmental coursework, the catch-up classes below college level. Three years later, 40.1% of the program group held a degree. In the control group, 21.8% did. The next-largest three-year graduation gain in MDRC's community college evaluations was 4 percentage points.

Figure 1. Graduation nearly doubled in both trials.
Three-year graduation for the program group and the control group in the CUNY and Ohio ASAP randomized trialsshare earning a degree within three years0%10%20%30%40%50%21.8%40.1%+18.3 ppCUNY ASAPn=896, randomized 201019.2%34.8%+15.6 ppOhio ASAPn=1,501, randomized 2015-2016
program groupcontrol group

Three-year graduation, program versus control, in the two ASAP randomized trials (MDRC 2015; MDRC 2020).

Source: MDRC, 2015; MDRC, 2020.

The program wrapped tuition assistance, required full-time enrollment, and gave every student an adviser carrying 60 to 80 students, with required meetings twice a month. An adviser in that program opens the fall with a roster of about seventy names. The twice-a-month rule turns the roster into a calendar: a first-semester student choosing between two math sections, a father of two whose shift changed and whose classes have to move with it, a student who missed a meeting and gets a phone call the same week. The evaluation charged her salary to the program's ledger and counted her students' degrees three years later.

The model traveled. Three Ohio community colleges, Cincinnati State, Cuyahoga, and Lorain County, replicated ASAP with 1,501 students starting in 2015, with MDRC again running the lottery. Degree receipt in three years reached 34.8% for the program group and 19.2% for the control group, and MDRC counts 126 additional graduates among the program's 806 students.

Cost per student and cost per degree give opposite verdicts

ASAP cost real money. Over three years, CUNY spent $42,065 per program student and $25,781 per control student, a net additional investment of $16,284, about 63% more than usual services (MDRC, 2015).

MDRC published the second denominator. Within the evaluation sample over the three-year window, a degree earned cost $118,248 under usual services. ASAP brought it to $104,825. The added spending lowered each degree's cost by $13,423, an 11.4% saving, because it was spread across nearly twice as many graduates.

Figure 2. The same program, judged per student and per degree.

cost per student, three years

Cost per student over three years, control group against ASAP+$16,284 (+63%)$25,781$42,065ControlASAP

computed from the same cost and outcome data

cost per degree earned, three years

Cost per degree earned over three years, usual services against ASAP-$13,423 (-11.4%)$118,248$104,825Usual servicesASAP

ASAPusual services (control group), the same colors in both panels

Within-study comparison, evaluation sample only, degrees earned within three years (MDRC 2015). Each panel carries its own dollar scale, and both scales start at zero.

Source: MDRC, 2015.

The arithmetic at the margin: $16,284 of added cost divided by 0.183 added graduates per student comes to about $89,000 for each graduate who exists because of the program (computed from MDRC's published inputs; see methodology).

Ohio ran the model at community college prices: $5,521 in direct cost per student over three years, and a net additional investment of $8,030 per student, a 42% increase. The cost per degree earned came to $77,783 for the program group and $99,162 for the control group, 22% lower. MDRC published the marginal price directly: about $51,000 per additional degree, versus a cost per graduate of about $99,000 in the control group.

Figure 3. Ohio published the cost of an additional graduate.
The published cost of an additional graduate in Ohio ASAP, with the published cost per degree earned beneath itper additional graduate, ASAP Ohio (published)$51,000per graduate under usual services (published)$99,000COST PER DEGREE EARNEDASAP program group (-22%)$77,783Control group$99,162$0$50,000$100,000

Ohio ASAP replication, 1,501 students at three community colleges, randomized 2015-2016. Bars start at zero on one shared dollar scale.

Source: MDRC 2020, Table 6.3 and executive summary.

A coach at $500 a semester moved persistence for two years

Persistence is the share of students still enrolled at a later check-in. The largest randomized coaching trial measured it four times. Bettinger and Baker studied 13,555 students, many older and enrolled part time, at eight institutions that hired InsideTrack, a phone-based coaching firm, between 2003 and 2008. Coached students were 5.2 percentage points more likely to be enrolled at six months (63.2% versus 58.0%), 5.3 points at twelve months, 4.3 at eighteen, and 3.4 at twenty-four, all significant at the 99% level. The gap held for at least a year after the coaching ended.

Figure 4. Coaching's persistence effect over two years.
The intention-to-treat enrollment effect of InsideTrack coaching at 6, 12, 18 and 24 months+2+4+6+5.2 pp6 months+5.3 pp12 months+4.3 pp18 months+3.4 pp24 months63.2% enrolled vs 58.0%0 (no effect)

All four estimates significant at the 99% level; effects persisted at least a year after coaching ended.

n=13,555.

Source: Bettinger and Baker 2014; InsideTrack cost roughly $500 per semester.

InsideTrack charged roughly $500 per semester. Two semesters is about $1,000, which works out to roughly $18,900 per additional student still enrolled at twelve months (computed; see methodology). The federal What Works Clearinghouse reviewed the trial, found it meets design standards without reservations, and rated its persistence effects potentially positive, with no discernible effects on degree attainment.

Fort Worth's Stay the Course trial tested a heavier version: a social worker as each community college student's navigator, plus up to $500 a semester in emergency assistance, at $1,880 per year. Among all women offered the program, associate-degree receipt within three years rose 7.4 percentage points; among the women who enrolled, 22% of those offered, the estimated gain was 31.5 points.

Ten minutes of hands-on help moved enrollment and completion

In 2008, H&R Block tax professionals offered low- and moderate-income families help completing the FAFSA, the Free Application for Federal Student Aid, using data already in the family's tax return, plus a personalized aid estimate. The assist took less than ten minutes, and families received $20 for participating. Among families with a high-school senior, FAFSA submissions rose 15.7 percentage points, a 40% increase. Fall college enrollment reached 34.5%, versus 26.8% in the control group. Three years on, 36% of the treated students had completed two years of college, compared with 28%.

A second arm received the aid estimate with no hands-on help; it produced no significant change in FAFSA submission and no enrollment effect. The verified cost of the version that worked: under ten minutes of a professional's time, with the form's data already on the screen. The enrollment gain those minutes bought was 7.7 percentage points.

The broadcast version reached 800,000 students and produced zero

Bird and coauthors tested outreach as pure messaging. Two campaigns, one statewide in Texas and one national through the Common Application platform, sent FAFSA completion texts and emails to more than 800,000 students, with many message designs. The published finding, in the authors' words: "no impacts on financial aid receipt or college enrollment overall or for any student subgroups." No variation in framing, delivery, timing, or access to one-on-one advising changed the result.

Oreopoulos and Petronijevic spent five years running the same test with more variations. Across nearly 25,000 students on three campuses, text and online coaching improved mental health and study time, and none of the interventions significantly moved any academic outcome, including for students at higher risk of dropping out. Students studied five to eight fewer hours a week than they planned and answered reminders by lowering their grade expectations. In the same team's earlier trial, students matched with a personal coach, an upper-year student who proactively started regular conversations, earned higher average grades, a gain of 0.3 standard deviations; the trial's text campaign and one-time online exercise produced no effect on any academic outcome.

In Florida, texts simplifying re-enrollment for 27,028 former community college students in good academic standing produced a small and statistically insignificant effect on their own; the arm that added a one-course tuition waiver, worth $303 to $354, lifted re-enrollment by 1.5 percentage points, a 21% relative gain. In Fort Worth, offering the emergency money without the navigator produced no difference from the control group. Across eleven public universities, a 10,037-student proactive-advising trial called MAAPS found no significant average effect on graduation or persistence at six years.

Figure 5. Five trials produced no effect.
Effect estimates from five null trials on a bold zero line, with two contrast rows beneathBird et al., 2021800,000+ studentsFAFSA messaging at scale: zero for every subgroupOreopoulos and Petronijevic, 2019~25,000 studentstexting and online coaching: no significant academic effectOrtagus et al., 20219,009 studentsinformation-only re-enrollment texts: small and insignificantMAAPS, six-year average10,037 studentseleven-university proactive advising: no significant effectEvans et al., 2020299 assignedemergency-aid-only arm: no difference from controlFOR CONTRASTOrtagus et al., 2021, waiver armsame Florida trialtexts plus a tuition waiver: re-enrollment +1.5 pp+1.5 pp-10+1+2 ppOreopoulos and Petronijevic, 2018grade effecta personal coach raised average grades +0.3 SD+0.3 SDown scale: standard deviations

A solid dot is a precisely estimated zero or a significant estimate; an open dot is a statistically insignificant estimate shown at zero. The bottom row is plotted on its own scale because its units are standard deviations, not percentage points.

"No impact" for Bird et al. is a precisely estimated zero at very large sample size; the other nulls are statistically insignificant estimates.

Source: Bird et al., 2021; Oreopoulos and Petronijevic, 2019 and 2018; Ortagus et al., 2021; Ithaka S+R, 2023; Evans et al., 2020.

Inside the MAAPS average, sites varied: at Georgia State, the one university that sustained the program past three years, program-group graduation ran 7 percentage points higher; at UC Riverside, program-group persistence ran 5 points lower.

Every effect in the record has a person or money attached

Per student, the two successful programs cost 63% and 42% more than usual services. Per degree earned, they cost 11.4% and 22% less.

Figure 6. What one additional outcome cost in each trial.
The cost of one additional outcome in each trial, published or computedInsideTrackcomputedper additional student enrolled at 12 months~$18,900Stay the Coursecomputed, treatment-on-the-treatedper additional associate degree among participating women~$17,900ASAP Ohiopublishedper additional degree~$51,000CUNY ASAPcomputedper additional graduate~$89,000FAFSA messaging at scalezero effect: no additional enrollees means no price to compute$0$25,000$50,000$75,000

Rows buy different outcomes, and a 12-month persister and a completed degree are different purchases. Each row carries its own outcome label and is read on its own.

Source: ASAP Ohio figure published (MDRC, 2020); the other rows computed from each trial's published inputs; see methodology.

Ohio's published figures put each additional graduate at about $51,000, in a system already paying about $99,000 per graduate.

Methodology

Every effect estimate in this piece comes from a randomized controlled trial; the What Works Clearinghouse entry is a federal evidence review of one of those trials, and the MDRC cost analyses are cost studies run inside the trials. All figures were verified against the listed primary-source URLs on August 21, 2026.

These computed figures are shown in full and labeled "computed" in the text:

  • InsideTrack cost per additional persister: the paper reports coaching "charged roughly $500 per semester," so a two-semester engagement is about $1,000 per treated student. $1,000 / 0.053 (the 12-month enrollment effect) = ~$18,900 per additional student still enrolled at 12 months. This is our arithmetic on an intention-to-treat basis; the authors publish no such figure.
  • CUNY ASAP cost per additional graduate: $16,284 (published net additional cost per student) / 0.183 (published graduation effect) = ~$89,000. Our arithmetic from MDRC's published inputs. The comparison figure, $118,248 per degree earned under usual services, is published.
  • Stay the Course cost per additional degree: $5,640 (published three-year cost) / 0.315 (published treatment-on-the-treated effect for women) = ~$17,900 per additional associate degree among participating women. Costs on an intention-to-treat basis would differ because take-up was 22%.
  • The Ohio figures are published, with no computation: ~$51,000 per additional degree and ~$99,000 per control-group graduate appear in MDRC's 2020 report (Table 6.3 and executive summary), as do $77,783 and $99,162 cost per degree earned.

Denominator discipline: every cost-per-degree figure is a within-study cost-effectiveness comparison, covering degrees earned by the evaluation sample within a three-year window. None of these figures describes an institution's overall budget.

Exclusions: the widely circulated ~$88 per-participant cost for the H&R Block experiment could not be verified at a primary source and is excluded; the piece uses only the verified cost facts (an assist of under ten minutes plus a $20 participation payment). Vendor-reported outcome claims (coaching, early-alert, and re-enrollment vendors) are excluded throughout; the only InsideTrack figures used are the independent trial and its federal review. Intention-to-treat estimates are the default; the single treatment-on-the-treated figure is labeled at every use.

Questions this answers

Did ASAP's effect fade after three years?
The six-year follow-up, published in the American Economic Journal: Applied Economics in 2019, estimates that ASAP raised six-year graduation rates by 10 percentage points, an effect the authors call exceptional among randomized experiments in higher education.
Do text messages ever change student outcomes?
They moved outcomes when they carried value. Florida's re-enrollment texts on their own produced a small, statistically insignificant effect, and the arm that added a one-course tuition waiver lifted re-enrollment by 1.5 percentage points, a 21% relative gain (Ortagus et al., 2021). In the H&R Block trial, the hands-on assist raised enrollment 7.7 points and the information-only arm showed no significant effect. The largest pure-messaging test, across more than 800,000 students, found no impact for any subgroup (Bird et al., 2021).
Are the cost-per-degree figures published research or advocacy?
MDRC's evaluation reports publish them directly as within-study cost-effectiveness analyses: $118,248 versus $104,825 for CUNY (2015) and $99,162 versus $77,783 for Ohio (2020). An independent estimate (Levin and Garcia, 2012) put the CUNY saving at 14.7%. The $89,000 and $18,900 figures in this piece are computed from published inputs and labeled as such.
What do "intention to treat" and "treatment on the treated" mean in these results?
Intention to treat measures the effect across everyone offered a program, including students who never showed up, which makes it the conservative default and the basis for every headline figure here. Treatment on the treated measures the effect among actual participants; the 31.5-point Fort Worth figure is treatment-on-the-treated behind 22% take-up, and it is labeled wherever it appears.

Sources

Cite this as

Risely AI Research, "Reaching out first costs more per student and less per degree," August 2026.

Last updated August 2026. Published by Risely AI.