GiveBetter x Chicago

Saga Education

High-dosage, in-school math tutoring for ninth and tenth graders

Research time: 9 min on Claude Opus 5.5
  • Research — reviewed programs, finances and impact evidence.
  • Modeling — estimated costs, QALYs and uncertainty.
  • Source audit — checked claims, assumptions and calculations.

Updated: 2026-10-18

Donate

Summary

What do they do? Saga Education (legally Saga Innovations) places trained tutors in public high schools to teach ninth and tenth graders math in pairs, as a daily class during the school day. Its Direct program served 3,083 students in 29 schools in Chicago, New York City and Baltimore in 2024–25; a separate Widespread program trains district tutors elsewhere. This in-depth review prices the organization’s whole annual cost per Direct student against adult earnings inferred from math gains in two Chicago randomized trials.

Why we’re interested in this organization:

  • Two randomized trials with about 5,300 Chicago students found math gains of 0.16 and 0.37 standard deviations, math grades up by about 0.4 to 0.6 points and math course failures roughly halved.

  • Gains persisted: in 11th grade, a year or two after tutoring, math scores were still 0.22 standard deviations higher and math grades 0.18 points higher.

  • Later randomized trials of a lower-cost, technology-assisted version, one of them in Chicago and New York City schools, found math gains of about 0.19 to 0.23 standard deviations.

Our main reservations:

  • Our working estimate is about $102K per DALY; all of it is adult income inferred from test scores, which is the weakest link in the chain.

  • The trials did not measure adult earnings, and the graduation effect (+1.9 points) is too imprecise to rule out a decline.

  • Current Chicago students average about 70 tutoring hours, roughly 60% of the daily trial dose, and Saga ran deficits of $11.5M and $6.7M in the last two years as stimulus and AmeriCorps funding ended.

What do you get for your dollar? $102K per DALY in Chicago. DALY is our one-year comparison unit. QALY-based health gains and income-welfare equivalents retain their original assumptions; they are not measured clinical DALYs. Whole-organization annual cost per Direct student against adult earnings inferred from randomized math gains; income-only welfare comparison, not a verified marginal gift offer.

Math test scores, math grades and course failures in grades 9–11; imprecise graduation; no adult outcomes.

1. What do they do?

Saga Education, founded in 2014 as SAGA Innovations (EIN 47-2168942), runs high-dosage math tutoring inside public high schools. Tutors work with two students at a time for a full class period every school day, using a curriculum matched to each student’s level. Its Direct program served 3,083 students in 29 schools in Baltimore, Chicago and New York City in 2024–25. A Widespread program trains and supports tutors employed by districts and states, reaching over 15,000 students. School districts pay part of the cost through contracts; philanthropy pays the rest. A donation supports the whole organization.

2. Monitoring and information sharing

Saga’s Chicago model was tested in two randomized trials run by the University of Chicago Crime and Education Labs (Guryan, Ludwig and coauthors; NBER working paper 28531, IPR WP-21-12). Study 1 randomized 2,633 9th and 10th graders in 2013–14; study 2 randomized 2,710 in 2014–15. Outcomes came from district records. Saga publishes annual reports with student counts and cites later evaluations. We did not find current Chicago-only student counts or a cost per student; we have not contacted the organization.

3. Qualitative assessment

The evidence on learning is strong. Among students offered a slot, about half participated; for participants, study 1 found a 0.16 SD math gain after one year, a 0.56-point rise in math grades on a 4-point scale and 49% fewer math course failures. Study 2 found 0.37 SD, a 0.42-point grade gain and 44% fewer failures. Pooled, the 11th-grade math gain was still 0.22 SD. The authors find no evidence that the effects come from mentoring or social-emotional change rather than instruction. Arnold Ventures rates the program Top Tier. The graduation effect is +1.9 points, with a confidence interval from -4.8 to +8.6, and no adult outcomes have been measured. Benefits beyond schooling therefore rest on how much test-score gains raise adult earnings, which we treat as uncertain.

4. What do you get for your dollar?

Our central estimate is $102K per DALY, using one modeled equivalent year. It is an income-only comparison: we credit no health gain. This revision replaces the first in-depth estimate of $121K per DALY.

Cost. FY2025 total expense was $25.4M, the same year as the 3,083 Direct students. Saga also supported over 15,000 Widespread students through district-employed tutors; we count each as a tenth of a Direct student, giving $5,548 per Direct student. Charging everything to Direct students gives $8,247. For comparison, the trials measured about $3,800 per participant in 2013–15 (about $5,206 today), and a co-founder put the cost at $3,100 in 2020.

Benefit. The persistent 0.22 SD gain is about 8 percentile points. Chetty et al. (2011) associate each 8th-grade percentile with $150 a year of earnings at age 27, so the trial authors’ own mapping implies about $1,819 a year in 2025 dollars. We treat half of that association as causal: Chetty, Friedman and Rockoff’s quasi-experimental teacher study implies about 11% higher earnings per standard deviation of scores, roughly half of this mapping, and a meta-analysis of 29 programs with adult follow-up (Bailey, Watts and Hart) finds the link between short-run scores and adult outcomes is real but not robust. We keep 0.6 for transfer to today’s program, because Chicago Direct students now average about 70 hours of tutoring against roughly 120 in the daily trial model, 0.9 for displacement and 0.8 for the share of adult life spent in the Chicago metro, and count 60% as net household welfare after taxes and lost benefits. Gains run from age 25 to 59, discounted at 3% from age 15, on a $35,000 adult-income base.

Range. At the trial cost the estimate is $95K per DALY. If the full association is causal, as the authors assume, it is $51K per DALY; if only a quarter is, $203K per DALY. Using only the smaller study-1 effect gives $139K per DALY. Charging the whole budget to Direct students gives $151K per DALY; assuming the trials’ full dose transfers gives $76K per DALY. A graduation-only channel would be tiny (+1.9 points times a diploma premium under 5%), so the estimate relies on broader skill effects. If test-score gains never reach adult earnings, there is no credited benefit.

Model, assumptions and sensitivity

FY2025 whole-organization expense divided by 2024–25 Direct students plus Widespread students at one-tenth weight.

Chicago Direct students; 0.8 of adult earning years assumed lived in the Chicago metro.

Price = cost per student ÷ [0.5 × discounted years(ages 25–59, from 15) × ln(1 + net gain ÷ base) × causal × transfer × non-displacement × Chicago share]. Gross gain = 0.22 SD × 37.5 percentiles/SD × $150 × CPI. Health = 0.

FY2025 total expense
25426248 USD (observed). Form 990, fiscal year ending June 2025. [propublica]
Direct students 2024–25
3083 students (observed). Annual report; same fiscal year as the expense. [saga-annual] [saga-annual-2025]
Widespread students 2024–25
15000 students (observed). Over 15,000 students reached through district-employed tutors. [saga-annual-2025]
Widespread cost weight
0.1 Direct-student equivalents (judgment). Training and support of district tutors costs far less per student than Saga-employed daily tutoring.
Published cost per student (2020)
3100 USD (observed). Co-founder statement. [hechinger-2020]
Trial cost per participant
3800 USD (observed). Authors’ best estimate for 2013–15. [saga-wp]
Persistent math effect
0.22 SD (observed). Pooled treatment-on-treated effect in 11th grade. [saga-wp]
Percentiles per SD
37.5 percentiles per SD (observed). Study 1: 0.16 SD equals 6 percentile points. [saga-wp]
Earnings per percentile
150 USD per year at age 27 (observed). Chetty et al. (2011) association as cited by the trial authors. [saga-wp] [chetty-2011]
CPI adjustments
[1.47,1.37] ratio (judgment). Approximate CPI-U changes.
Causal share of score-earnings link
0.5 share (judgment). Quasi-experimental teacher value-added implies about half of this mapping; fadeout meta-analyses find weak, non-robust links. [saga-wp] [cfr-review] [fadeout]
Transfer to current program
0.6 share (judgment). Chicago Direct students now average about 70 hours vs roughly 120 in the daily trial model; later lower-cost trials found 0.19–0.23 SD. [saga-tech-rct] [saga-tech] [saga-annual-2025] [saga-wp]
Non-displacement
0.9 share (judgment). Skill gains are mostly not zero-sum.
Chicago share of adult years
0.8 share (judgment). Allows for moving away.
Net welfare share
0.6 share (judgment). After income and payroll taxes and lost means-tested benefits.
Adult base income
35000 USD per year (judgment). Reference personal income for adults from these schools.

Allocated FY2025 cost per Direct student, causal half of the score-earnings link: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 5.46. FY2025 total expense over 2024-25 Direct students plus Widespread students at 0.1 weight; persistent 0.22 SD gain; half the Chetty association treated as causal; 0.6 transfer for lower current dose; 0.9 non-displacement; 0.8 Chicago residence; 60% net welfare.

Whole FY2025 expense charged to Direct students: Cost: $825K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 5.46. All Widespread and technology work treated as overhead of the Direct program.

Published per-student cost: Cost: $388K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 5.46. Co-founder’s $3,100 per student (2020), inflated to 2025.

Current program as effective as the trials: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 7.28. Prior review’s transfer factor; ignores the lower current dose.

Program cost as measured in the trials: Cost: $521K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 5.46. Central benefits at the authors’ $3,800 per-student cost, inflated to 2025.

First in-depth review, 18 October 2026 (superseded): Cost: $882K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 7.28. FY2023 expense over 2024-25 Direct students with 0.8 transfer; retained for comparison.

Full score-earnings association is causal: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 10.92. As the trial authors assume.

Only a quarter of the association is causal: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 2.73. Stronger fade-out or weaker causal link from test scores to earnings.

Smaller study-1 immediate effect only: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 3.99. Uses 0.16 SD instead of the pooled persistent 0.22 SD.

Current program much less effective than the trials: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 2.73. Scale-up or implementation drift halves the transferred effect.

All adult years in the Chicago metro: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 6.83. No out-migration discount.

5% discount rate: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 3.50. As in the authors’ benefit-cost calculation.

Test-score gains do not change adult earnings: Cost: $555K; Chicago health QALYs: 0.00; all-population health QALYs: 0.00. Income-equivalent healthy years: 0.00. Full fade-out of any earnings effect.

Counterfactual: Randomized control students on the regular CPS schedule, often double-dose Algebra.

Attribution: A gift is credited with the transferred average effect of an added Direct student at whole-organization average cost; whether a gift adds a Chicago student is unverified.

Dominated by the causal link from test scores to adult earnings, then by current cost per student, transfer to today’s program and out-migration.

Sensitivity

  • Whole FY2025 expense charged to Direct students: $151K per DALY. All Widespread and technology work treated as overhead of the Direct program.
  • Published per-student cost: $71K per DALY. Co-founder’s $3,100 per student (2020), inflated to 2025.
  • Current program as effective as the trials: $76K per DALY. Prior review’s transfer factor; ignores the lower current dose.
  • Program cost as measured in the trials: $95K per DALY. Central benefits at the authors’ $3,800 per-student cost, inflated to 2025.
  • First in-depth review, 18 October 2026 (superseded): $121K per DALY. FY2023 expense over 2024-25 Direct students with 0.8 transfer; retained for comparison.
  • Full score-earnings association is causal: $51K per DALY. As the trial authors assume.
  • Only a quarter of the association is causal: $203K per DALY. Stronger fade-out or weaker causal link from test scores to earnings.
  • Smaller study-1 immediate effect only: $139K per DALY. Uses 0.16 SD instead of the pooled persistent 0.22 SD.
  • Current program much less effective than the trials: $203K per DALY. Scale-up or implementation drift halves the transferred effect.
  • All adult years in the Chicago metro: $81K per DALY. No out-migration discount.
  • 5% discount rate: $158K per DALY. As in the authors’ benefit-cost calculation.
  • Test-score gains do not change adult earnings: no finite positive estimate. Full fade-out of any earnings effect.

Unresolved inputs

  • Chicago-only Direct students and cost per added student.
  • Any measured adult outcome (college, earnings) for trial participants.
  • Dose-response evidence linking tutoring hours to persistent gains.

Health and income breakdown

For the modeled $555K: 0.00 health QALYs + 5.46 income-equivalent healthy years in Chicago: $102K per DALY. Income equivalents are welfare comparisons, not observed QALYs or DALYs.

Method and caveats.

5. Funding and previous grants

Saga Innovations’ expenses peaked at $29.4M in FY2024 and fell to $25.4M in FY2025, after $27.2M in FY2023. Revenue fell from $28.8M in FY2023 to $17.9M and then $18.7M, leaving deficits of $11.5M and $6.7M and net assets of $12.6M. District contract revenue fell from $9.9M to $5.1M. The 2024–25 annual report says stimulus (ESSER) funding waned and Saga’s AmeriCorps grants ended; it lists $5.6M from philanthropic partners including Blue Meridian Partners, the Dell Foundation, Overdeck Family Foundation, Arnold Ventures and the IMC Charitable Foundation, which funded tutoring in 21 Chicago Public Schools. CPS itself faces a deficit of over $500.0M. A gift may therefore sustain existing Chicago capacity rather than add new students; we have not seen a costed plan.

Annual expenses

Organization-level spending, including programs, administration and fundraising. The research list averages three comparable, consecutive full fiscal years when available.

  • FY 2023: $27.2M; Saga Innovations, EIN 47-2168942, 12-month period, IRS Form 990 total functional expenses, fiscal year ending June. Source
  • FY 2024: $29.4M; Saga Innovations, EIN 47-2168942, 12-month period, IRS Form 990 total functional expenses, fiscal year ending June. Source
  • FY 2025: $25.4M; Saga Innovations, EIN 47-2168942, 12-month period, IRS Form 990 total functional expenses, fiscal year ending June. Source

6. Sources

  1. Not Too Late: Improving Academic Outcomes Among Adolescents (IPR WP-21-12). Northwestern Institute for Policy Research. Published: 2021-03-08; retrieved: 2026-10-18.
  2. Not Too Late: Improving Academic Outcomes Among Adolescents (NBER w28531). National Bureau of Economic Research. Published: not stated; retrieved: 2026-10-18.
  3. Not too late: study summary. Stanford National Student Support Accelerator. Published: not stated; retrieved: 2026-10-18.
  4. An RCT of Saga Technology: high-dosage tutoring and an education technology platform for ninth graders. Arnold Ventures. Published: not stated; retrieved: 2026-10-18.
  5. How Does Your Kindergarten Classroom Affect Your Earnings? (as cited in the trial paper). Chetty et al., Quarterly Journal of Economics. Published: not stated; retrieved: 2026-10-18.
  6. Saga Education annual report 2024–25. Saga Education. Published: not stated; retrieved: 2026-10-18.
  7. Saga Innovations, Form 990 filings (EIN 47-2168942). IRS, hosted by ProPublica. Published: not stated; retrieved: 2026-10-18.
  8. Can Technology Facilitate Scale? Evidence from a Randomized Evaluation of High-Dosage Tutoring (NBER w32510). University of Chicago Education Lab (Bhatt, Guryan et al.). Published: 2024-06-01; retrieved: 2026-10-18.
  9. Saga Education annual report 2024–25. Saga Education. Published: not stated; retrieved: 2026-10-18.
  10. This math tutoring program gets blockbuster results in high-poverty schools. The Hechinger Report. Published: 2020-01-29; retrieved: 2026-10-18.
  11. Review of Chetty, Friedman and Rockoff teacher value-added study. National Education Policy Center. Published: not stated; retrieved: 2026-10-18.
  12. Why Do Most Education Interventions Fade Out Over Time?. Education Next (Bailey, Watts and Hart). Published: 2026-02-12; retrieved: 2026-10-18.
  13. The Signaling Value of a High School Diploma. RAND (Martorell and Clark). Published: not stated; retrieved: 2026-10-18.
  14. Cost-effectiveness: health and log-income comparison. Coefficient Giving. Published: not stated; retrieved: 2026-10-18.