Measures of Effective Teaching: The Rise and Retreat of the Gates MET Study

Majid Khan
37 Min Read

Key Takeaway

>

  • The measures of effective teaching study, funded by the Bill & Melinda Gates Foundation at roughly $45–52 million, ran from 2009 to 2013 across six urban districts with about 3,000 volunteer teachers.
  • Its January 2013 culminating reports concluded that no single measure identifies effective teaching on its own, and that a composite of value-added scores, classroom observation and student surveys performed better than any single component.
  • Its most durable finding is a reliability figure: a single observation score reached only 0.14–0.37 reliability, and four observations scored by four raters were required to reach 0.65.
  • As of November 2026 there is no successor study, no MET 2.0 and no updated dataset; the surviving longitudinal database is available only through restricted-access application.
  • At least 30 states rolled back high-stakes evaluation policies after ESSA removed the federal condition, and a 2024 national analysis by Bleiberg, Brunner, Harbatkin, Kraft and Springer found no meaningful effect on student outcomes.

The measures of effective teaching project closed in January 2013 and has never been reopened. As of November 2026 there is no successor study, no second data collection and no announced MET 2.0 from the Gates Foundation or any other funder, yet the 2013 reports remain the baseline citation for classroom observation protocols and student-survey design across U.S. teacher preparation and district coaching programmes. The research outlived the policy it was used to justify because the measurement findings were descriptive — they described how noisy a single observation score is — while the high-stakes evaluation systems built on them collapsed under rater and staffing demands districts never funded. At least 30 states have since retrenched.

About this analysis. This article follows the money and the measurable outcome: stated amounts, dated commitments, named researchers and the published findings that ended the policy. Institutional positions are attributed to the Bill & Melinda Gates Foundation; where Bill Gates has spoken personally, that is identified as his own statement rather than Foundation policy.

Transparency note. Every figure below is drawn from the cited sources listed at the end of this article. Where a value has not been published — including the Foundation’s internal rationale for specific grant amounts — this article says so rather than estimating.

Where the Measures of Effective Teaching findings stand today

The measures of effective teaching project is, as of November 2026, a closed 2013 research programme with no successor, whose observation and student-survey findings remain in active formative use while its high-stakes evaluation applications have been abandoned. The MET project Gates Foundation funded between 2009 and 2013 produced three sets of reports, a longitudinal database and a set of reliability estimates that are still reproduced in current literature.

Who still cites it is a narrow and identifiable group. Teacher preparation programmes use MET’s video methodology to assess candidates. Video coaching platforms built their review cycles on it. District survey offices run student-experience instruments descended from the Tripod 7Cs. And the published research literature continues to reproduce MET’s reliability figures, including the 2025 handbook chapter by Eric S. Taylor of Harvard University, which restates the single-observation reliability range directly.

What has been discarded is the summative application: value-added-weighted composite ratings attached to tenure, dismissal and performance pay. The split is not accidental. The Measures of Effective Teaching findings survived because they told coaches something operationally useful about measurement error, which means they hold whether or not a score carries consequences. The policy failed because acting on those same findings required four trained raters per teacher — a capacity cost that districts were never funded to carry, resulting in observation systems that recorded almost no variation at all.

2009 — Race to the Top makes teacher evaluation a funding condition

In 2009 the federal Race to the Top competition made teacher and principal evaluation systems incorporating student achievement growth a scoring condition for state applications, and in the same year the Bill & Melinda Gates Foundation committed roughly $45–52 million to the Measures of Effective Teaching project to build the measurement evidence behind it. The sequence matters: the federal condition arrived first, which created immediate state demand for a defensible instrument that did not yet exist.

Race to the Top teacher evaluation requirements pushed dozens of states to rewrite statute at speed. States competing for federal awards needed to show reviewers that their evaluation systems incorporated student achievement data, and most had neither validated rubrics nor any evidence on how many observations a reliable rating required. That evidentiary gap is precisely what the MET project set out to fill.

The Foundation’s stated rationale, as an institutional position, was that better measurement of teaching would improve the identification and development of effective teachers. The project ran across six urban districts with the participation of research partners including Harvard, where the Tripod survey instrument had been developed. Consequently a philanthropic research study acquired policy weight it was never designed to carry: a validity study funded by a private foundation became the technical justification for personnel law in states that had not commissioned it and were not party to its design. That mismatch between the research question and the policy use is the root of nearly everything that followed.

2009 to 2012 — Three years of fieldwork: video, peer observers, value-added and Tripod 7Cs

Between 2009 and 2012 the project collected classroom data at a scale no district had attempted. Roughly 3,000 volunteer teachers participated, thousands of hours of lessons were captured on multi-camera rigs designed to show both teacher and students, and trained raters outside the participating schools scored the recordings remotely rather than visiting classrooms.

Four instruments were tested rather than one, and the reason is statistical. Each measure carries a different error structure: value-added estimates are distorted by small class sizes and test noise, observation scores by which lesson an observer happens to see, student surveys by class composition. The design tested whether combining measures with unrelated errors would cancel some of that noise, which is the logic the later composite conclusion rested on.

The four measures tested in the MET fieldwork

  • Value-added measures — statistical estimates of a teacher’s contribution to student test-score growth, used as the achievement anchor against which the other measures were validated
  • Classroom observation with structured rubrics — lessons scored against instruments including the Danielson Framework for Teaching and CLASS, applied to video rather than live visits
  • Trained peer observers — teachers and raters from outside the school scoring recorded lessons, which removed the single-principal dependency of conventional evaluation
  • Tripod 7Cs student survey — a student perception instrument covering seven constructs including care, challenge, control and clarify, developed by Ronald Ferguson at Harvard

The use of video was the methodological innovation with the longest afterlife. Because recorded lessons could be scored by raters who had never met the teacher, the design separated the measurement of instruction from the politics of the building — and that separation is what later made video a practical coaching medium rather than an audit tool.

Thomas J. Kane and the research team behind the design

The project’s principal investigator was Thomas J. Kane, an economist at the Harvard Graduate School of Education who had previously directed the Gates Foundation’s data and research work in education. The Thomas Kane MET study design therefore reflected an economist’s priorities: measurement validity first, operational feasibility second.

A university research team rather than a school district led the work because the questions required capacity no district held. Establishing whether a measure predicts future achievement gains demanded randomised reassignment of teachers to classes, common assessments across sites, and statistical modelling across six districts simultaneously. No individual district research office could run that, which is why the design sat with a university economist and a national research consortium.

The consequence is the core of the teacher effectiveness research criticism that followed. Because the design was optimised for statistical validity, it was not optimised for the staffing realities of the schools expected to implement it. The study established what a defensible score requires; it did not establish that a typical school could produce one. Later analyses, including the post-mortem published by the American Enterprise Institute, identified exactly that gap — administrator capacity — as a principal mechanism of implementation failure.

January 2013 — The culminating reports and the composite argument

In January 2013 the MET project released its culminating reports, and the headline conclusion was that no single measure identifies effective teaching on its own. The reports argued that the most defensible approach combined measures of effective teaching into a weighted composite, because each individual instrument carried a substantial and distinct error component that the others partly offset.

What the January 2013 MET reports concluded

  1. No single measure of teaching is sufficient on its own; each carries a distinct and substantial error component
  2. A weighted composite of value-added scores, observation ratings and student survey results predicted future student achievement gains better than any component alone
  3. Observation ratings required multiple lessons and multiple raters before they became stable enough to compare teachers
  4. Student perception surveys were the most internally consistent of the three measures and predicted achievement independently
  5. Teaching effects were identifiable and non-random — the finding states seized on to justify test-linked personnel decisions

The objection arrived almost immediately. A review published through the National Education Policy Center at the University of Colorado Boulder, with critiques led by economists including Jesse Rothstein and NEPC’s William Mathis, argued that the reports overstated what the composite could support and that the three measures agreed with one another too weakly to carry high-stakes consequences. That objection mattered because it attacked the exact inference states were about to write into statute: not whether teaching quality varies, but whether a composite score was stable enough to end a career. The statutes were drafted anyway.

The reliability finding that outlasted everything else

The single most consequential number MET produced was a reliability coefficient: the reliability of a single observation score fell between 0.14 and 0.37, and reaching 0.65 — the threshold MET treated as acceptable for comparing teachers — required four observations scored by four different raters. Those figures, reproduced in Eric S. Taylor’s 2025 AEFP handbook chapter, have outlived almost every other MET conclusion.

For a school leader the arithmetic translates plainly. A reliability coefficient of 0.2 means roughly four-fifths of the variation in a score reflects which lesson was watched and who watched it, not how the teacher teaches. Which means a single principal visit cannot support a personnel decision — not because the principal is unskilled, but because one lesson is too small a sample of a teacher’s year.

Observation setup Reported reliability Raters required Lessons scored
Single observation, single rater (MET range) 0.14–0.37 1 1
MET’s acceptable comparison threshold 0.65 4 4
Typical district practice before MET Not stated 1 (principal) 1–2
Formative coaching use in 2026 No summative score produced 1 coach or peer Repeated cycle

The implementation consequence was decisive. Four raters across four lessons for every teacher in a building is a staffing cost most districts never funded, and therefore schools did one of two things: they cut the number of observations, or they rated nearly everyone proficient. Both responses destroyed the very classroom observation reliability the reforms claimed to deliver. A 2026 applied study from Mississippi State University documented the same squeeze directly, describing a capacity-demands imbalance in which the hours required for rigorous multi-measure evaluation exceeded what high school administrators had available.

The student-voice result: surveys as a predictor

MET validated student perception surveys — specifically the Tripod 7Cs student survey — as a predictor of student achievement gains and engagement that performed more consistently than single classroom observations. Of the three measures tested, student surveys showed the highest internal consistency.

The reason is sample size. Every student in a class responds, which means a teacher’s survey result rests on 20 to 30 independent judgements accumulated across an entire term, while an observation rating rests on the two or three lessons a rater happened to see. The effect of that difference is a measure that moves less from one administration to the next.

After 2013, districts kept the instrument and changed its purpose. The 7Cs constructs — including care, challenge, control, clarify, captivate, confer and consolidate — are now used to measure school climate and student experience rather than to grade individual teachers, as Education Elements’ overview of the Tripod 7Cs Assessment describes. The driver of that repositioning is straightforward: once survey results carried consequences for a teacher’s pay, every adult in the building had an incentive to shape the responses, and consequently formative use preserved the data quality that summative use degraded. In 2026 the instrument and its descendants sit in district student-experience and climate reporting, feeding coaching conversations rather than ratings.

After 2013 — States write MET’s logic into law and the backlash builds

Alongside the research, the Gates Foundation committed roughly $500 million to the Intensive Partnerships for Effective Teaching, a companion initiative that funded school districts and charter management organisations to build and operate multi-measure evaluation systems at full scale. MET supplied the measurement argument; Intensive Partnerships supplied the implementation money. Dozens of states supplied the statute.

State legislation after 2013 tied tenure decisions, dismissal procedures and performance pay to composite ratings containing test-linked growth components. The specific weightings varied by state and are not uniformly documented, but the structural change was consistent: a quantitative student-growth component became a mandatory input into employment decisions for teachers in tested subjects.

The backlash followed from how the systems were used. Because composite ratings were attached to consequences rather than to coaching, teachers and principals read an observation visit as a compliance event, which destroyed practitioner buy-in and with it the developmental purpose the Foundation had articulated. Union opposition, test-boycott campaigns and complaints about administrator time burden all trace to that same cause. The Mississippi State analysis quantified the administrative side of it, and the AEI post-mortem names political opposition and administrator capacity as the two mechanisms that recur across states. Gates Foundation teacher evaluation funding had bought a system that the people operating it had no incentive to operate well.

2018 to 2024 — The national retreat and the evidence behind it

The national push to reform teacher evaluation produced no meaningful improvement in student outcomes, according to the 2024 national analysis by Joshua Bleiberg, Eric Brunner, Erica Harbatkin, Matthew A. Kraft and Matthew Springer. That conclusion, reached across a decade of state reform data, is the closing entry in an evidence trail that began six years earlier.

In 2018 the RAND Corporation and the American Institutes for Research published their evaluation of the Intensive Partnerships initiative and found no significant or sustained gains in student achievement, nor measurable improvement in low-income and minority students’ access to effective teaching. Because the Foundation had commissioned that evaluation itself, the finding removed the empirical case for the model from inside the programme rather than from its critics.

The policy ground had already shifted. The Every Student Succeeds Act, signed in 2015, removed the federal condition that had driven adoption in the first place, which means states no longer gained any funding advantage from maintaining a system their teachers, principals and unions opposed. The political cost stayed; the federal benefit disappeared. At least 30 states retrenched as a result, weakening test-score weighting, restoring multi-year tenure protections or dropping growth measures from evaluation entirely.

The evidence trail behind the retreat

  1. 2015 — ESSA replaces Race to the Top conditions and ends the federal requirement to tie evaluation to student growth
  2. 2018 — RAND and AIR report no significant achievement gains in the Intensive Partnerships districts
  3. 2018 to 2023 — At least 30 states weaken or repeal high-stakes evaluation provisions
  4. 2024 — Bleiberg, Brunner, Harbatkin, Kraft and Springer find no meaningful national effect on student outcomes
  5. 2025 to 2026 — Taylor’s AEFP handbook chapter and the Mississippi capacity study reframe the era as an implementation failure rather than a measurement failure

The two named mechanisms — administrator capacity and political opposition — point to the same conclusion about teacher evaluation reform results: the limit was the delivery system, not the measurement science.

The Gates Foundation’s own pivot away from evaluation

The Gates Foundation stopped centring teacher evaluation because its own commissioned evaluations found no achievement gains, and its U.S. Program now funds educator preparation, foundational learning, curriculum quality and support for the state and intermediary agencies that implement instruction. The pivot is institutional: it is visible in what the Foundation funds and in the strategy language it publishes, and it should be read as a Foundation position rather than as a personal statement by Bill Gates, who has at various points reflected publicly on the limits of the Foundation’s education work in his own voice.

Funding area 2009–2013 emphasis 2026 emphasis
Teacher evaluation and measurement Central; MET plus $500M Intensive Partnerships Not a stated priority
Educator preparation Secondary Central to U.S. Program strategy
Curriculum and instructional materials Limited High-quality instructional materials funded
Student assessment and value-added data Core measurement anchor Not a stated priority
State and intermediary implementation support Primarily district-level grants “Middle-tier” implementation support funded

The Foundation’s educator preparation strategy page sets out the current K-12 position, which emphasises preparation programmes that build a diverse teaching workforce equipped to serve Black, Latino and low-income students. The measurement attention has moved with the money: the National Academy of Education’s 2025 report on evaluating and improving teacher preparation programmes now occupies much of the analytic space MET-era in-service evaluation once held.

For readers seeking funding, the practical point is that the Foundation’s K-12 work is overwhelmingly invitation-led rather than open-call, which means the actionable route is alignment with a named programme strategy and partnership with an existing grantee or intermediary rather than an unsolicited application. Anyone tracking Gates Foundation funding priorities should read the current strategy language first and treat the published priority areas as the eligibility signal.

November 2026 — Who still cites MET

MET’s conclusions are now cited to justify formative coaching, not summative rating — and that distinction is what makes the survival of a closed 2013 study legible. The 2013 reports appear today in teacher preparation programme design, in video coaching platform documentation, in district student-experience survey instruments and in the observation rubrics MET validated at scale.

Who still cites MET in 2026

  • Teacher preparation programmes using video evidence for candidate assessment and licensure portfolios
  • Video coaching teacher observation platforms including GoReact and Edthena, whose review cycles trace to MET’s video observation methodology
  • Districts running student-experience and school-climate surveys built on the Tripod 7Cs constructs
  • Observation rubrics MET validated at scale, principally the Danielson Framework for Teaching and CLASS
  • Research literature on observation reliability, including Eric S. Taylor’s 2025 AEFP handbook chapter
  • Vendor and practitioner roundups of observation and feedback tools, such as Perspective’s 2026 guide for teacher educators

The formative use stuck for a specific reason. MET demonstrated that recorded lessons could be scored consistently by raters who were not in the room, which made video a workable coaching medium, resulting in a tool market that cites MET’s video methodology while leaving its composite scoring argument alone. Edthena’s AI Coach feature, which automates parts of the review cycle for teacher self-reflection, is a current example of a product whose lineage runs back to MET’s video observation work rather than to its rating formula.

What is not coming: no updated dataset, no MET 2.0

As of November 2026 there is no MET 2.0, no updated MET dataset and no announced successor study from the Gates Foundation or any other funder. All three of the questions readers ask — is there a version two, is new data being collected, has a successor been commissioned — have the same answer, and nothing in the Foundation’s current published K-12 strategy indicates otherwise.

What does exist is the original data. The MET longitudinal database remains available to qualified researchers under restricted-access terms through the archive housed at the Institute for Social Research, and access is controlled rather than public because the files contain student-level records and classroom video of identifiable children and teachers.

Requesting access to the MET longitudinal database

  1. Confirm institutional affiliation and secure IRB or equivalent ethics approval for the proposed analysis
  2. Submit a research plan specifying which MET components are required, including whether classroom video is needed
  3. Execute the restricted-data use agreement covering storage, access control and publication review
  4. Complete the archive’s secure-access onboarding before the files are released
  5. Cite the database and the original MET reports as the data use agreement requires

What comes next is a different kind of evidence. Research and philanthropic attention has moved to curriculum quality, educator preparation and formative coaching, which means the next generation of findings on teaching quality will come from preparation programme evaluations and instructional materials studies rather than from a second national measurement project.

Frequently asked questions

What was the Gates Foundation’s Measures of Effective Teaching study?

The measures of effective teaching study was a Bill & Melinda Gates Foundation research project running from 2009 to 2013 that tested whether classroom observation, student surveys and value-added test-score measures could together identify effective teaching. It operated across six urban districts with roughly 3,000 volunteer teachers and thousands of hours of classroom video, culminating in reports released in January 2013.

How much did the Gates Foundation spend on the MET project?

Roughly $45 million to $52 million went to the MET research itself. That figure is separate from the companion commitment of approximately $500 million to the Intensive Partnerships for Effective Teaching, which funded districts and charter networks to implement multi-measure evaluation systems rather than to conduct research.

What did the MET study actually conclude in January 2013?

No single measure identifies effective teaching on its own. A weighted composite of value-added scores, observation ratings and student surveys predicted future achievement gains better than any component alone. The National Education Policy Center review objected that the three measures agreed with one another too weakly to carry high-stakes personnel consequences.

How many classroom observations does a reliable teacher score require?

Four observations scored by four different raters, to reach a reliability of 0.65. A single observation by a single rater reached only 0.14 to 0.37. Because that rater load exceeded what most schools staffed, scores produced in practice carried far more measurement error than the policies using them assumed.

Did teacher evaluation reform improve student achievement?

No. The 2024 national analysis by Bleiberg, Brunner, Harbatkin, Kraft and Springer found no meaningful effect on student outcomes across a decade of state reforms, and the 2018 RAND and AIR evaluation of the Intensive Partnerships districts found no significant or sustained achievement gains and no measurable improvement in access to effective teaching.

Why did the Gates Foundation move away from teacher evaluation?

Because its own commissioned evaluations found the systems did not raise student achievement. Administrator capacity shortfalls and sustained political opposition compounded the problem. The Foundation’s U.S. Program now emphasises educator preparation, foundational learning and curriculum quality — an institutional repositioning rather than a personal statement by Bill Gates.

How many states rolled back their teacher evaluation policies?

At least 30 states weakened or repealed high-stakes evaluation provisions. The trigger was the Every Student Succeeds Act, which in 2015 removed the federal requirement to tie evaluation to student growth, which means states gained no funding advantage from keeping a system their educators opposed.

Is the Tripod 7Cs student survey still used in 2026?

Yes, but for a different purpose. Districts use the Tripod 7Cs student survey and comparable instruments to measure school climate and student experience rather than to rate individual teachers, because attaching pay and employment consequences to student responses degraded the quality of the data the instrument was built to collect.

Is there an updated MET study or a MET 2.0?

No. As of November 2026 there is no MET 2.0, no updated dataset and no announced successor study. Research and philanthropic funding has moved to curriculum quality, educator preparation and formative coaching, and no funder has commissioned a measurement project at MET’s scale.

Can researchers still access the MET project data?

Yes, under restricted access. The MET longitudinal database is held in an archive at the Institute for Social Research and requires institutional affiliation, an approved research plan and an executed data use agreement, because the files include student-level records and classroom video of identifiable participants.

Limits of this evidence and the alternatives now being tested

The measures of effective teaching evidence base is narrower than its policy influence suggested, and several limits are worth stating plainly.

  • Participants were volunteers in six urban districts, which means the reliability estimates do not transfer automatically to rural schools, to very small classes, or to subjects without standardised assessments.
  • Value-added measures were never available for most teachers, because the majority of teaching assignments have no annual standardised test. Therefore the composite MET validated could only ever cover a minority of a school’s staff, and states applied composite logic to teachers for whom the validating evidence did not exist.
  • The 2024 Bleiberg et al. finding is an analysis of policy effects, not of measurement validity. It demonstrates that the reforms did not move student outcomes, which is a different claim from saying the measures themselves were invalid.
  • Critics including Jesse Rothstein argued through the NEPC review that the measures agreed with one another too weakly to carry personnel consequences, and later research has not resolved that objection.
  • Alternatives now in district practice include trust-based observation models, sustained instructional coaching cycles, high-quality curriculum adoption with implementation support, and student-experience surveys used formatively. None of these carries national effect evidence at the scale a MET successor would have required, which is itself a gap in the current record.

Where the timeline ends and what to watch

As of November 2026 the measures of effective teaching research is closed, effectively uncited in high-stakes policy and widely cited in formative practice — a reversal of the position it held a decade ago. The durable contribution turned out to be the reliability arithmetic rather than the composite formula, because that arithmetic told schools what a defensible observation actually costs in observer hours, and that finding holds regardless of whether a score carries consequences.

Three things are worth watching from here. First, the Gates Foundation’s educator preparation and foundational learning grants, which now absorb the attention teacher evaluation once held. Second, the emerging evidence base on curriculum implementation, which is where the next credible claims about instructional quality will originate. Third, whether any funder commissions a measurement study at MET’s scale — and on the published record, none has been announced.

For ongoing coverage of where the money is moving, see billgates.pk coverage of grant eligibility and funding timelines.

References

  • National Education Policy Center, University of Colorado Boulder — review and critique of the MET culminating findings: https://nepc.colorado.edu/blog/billions-dollars
  • Eric S. Taylor (Harvard University), 2025 AEFP handbook chapter on teacher evaluation, reproducing MET’s observation reliability figures: https://erictaylor.scholars.harvard.edu/sites/g/files/omnuum3716/files/2025-03/Taylor%20AEFPHbk.pdf
  • W.T. Grant Foundation — the MET longitudinal database at the Institute for Social Research: https://wtgrantfoundation.org/met-database-at-the-institute-for-social-research
  • American Enterprise Institute — post-mortem on teacher evaluation policy, covering the Bleiberg, Brunner, Harbatkin, Kraft and Springer 2024 analysis: https://www.aei.org/commentary/teacher-evaluation-policies-have-flopped-where-did-they-go-wrong/
  • Bill & Melinda Gates Foundation — U.S. Program, K-12 Education, Educator Preparation strategy: https://www.gatesfoundation.org/our-work/programs/us-program/k-12-education/educator-preparation
  • Mississippi State University — 2026 applied dissertation on capacity-demands imbalance in high school teacher evaluation: https://scholarsjunction.msstate.edu/cgi/viewcontent.cgi?article=7896&context=td
  • National Academy of Education — Evaluating and Improving Teacher Preparation Programs (2025): https://naeducation.org/wp-content/uploads/2025/01/Evaluating_and_Improving_Teacher_Preparation_Programs.pdf
  • Education Elements — Tripod 7Cs Assessment overview and current district use: https://www.edelements.com/tripod-7cs-assessment
  • Perspective — observation and feedback tools for teacher educators, 2026: https://getperspective.ai/blog/observation-and-feedback-tools-for-teacher-educators-2026
Share This Article
Leave a Comment

Leave a Reply

Your email address will not be published. Required fields are marked *