Measuring MESH: How to Measure Social Emotional Skills

By the Transforming Education editorial team. Reviewed for accuracy against the sources listed below.

Schools teach social and emotional skills whether or not they measure them. The question this guide answers is whether measuring those skills helps, which methods are trustworthy enough to act on, and where the limits sit. It draws on the largest real-world test of SEL measurement in US public schools, California's CORE Districts, and on the measurement scientists who have been most candid about what surveys can and cannot do.

What MESH meant, and how it maps to CASEL

MESH stands for Mindsets, Essential Skills and Habits. The original Transforming Education toolkit introduced the term as shorthand for a deliberately narrow set of constructs: growth mindset, self-efficacy, self-management and social awareness. The former Boston nonprofit behind that work was founded in 2013 and partnered with the CORE Districts to choose which skills to measure. This site is an independent publication and is not that organization; we use the term because many educators still search for it.

The selection logic matters more than the acronym. According to Martin West's Brookings analysis, the four skills were chosen on three criteria: they could be measured, they predicted important outcomes, and schools could plausibly change them. Here is how they line up with CASEL's five competencies.

Two CASEL competencies, relationship skills and responsible decision-making, were not part of the MESH set. That was a measurement decision, not a judgment about their importance. Reliable, brief survey items for the four chosen constructs existed in the research literature; items for the other two were less mature in 2014. If your district's SEL standards cover all five competencies, you will need instruments beyond a MESH-style survey. Our SEL assessment guide walks through those options.

Why measure social emotional skills at all

Three practical reasons justify the effort, and one common reason does not.

First, these skills matter for outcomes schools already care about. CASEL's research summary reports that students in SEL programs addressing the five competencies scored 11 percentile points higher academically than non-participants, and that years later, participants' academic performance averaged 13 percentile points higher. If a district invests in SEL, it is reasonable to check whether the intended skills are actually growing.

Second, the data reveal patterns adults cannot see from the hallway. The PACE analysis of the CORE survey, Trends in Student Social-Emotional Learning, found that self-efficacy and social awareness declined markedly between grade 4 and grade 12, with the steepest drops in middle school, while growth mindset rose steadily through grade 10. Girls' self-efficacy fell behind boys' by grade 7 and the gap exceeded a quarter of a standard deviation by grade 11. No teacher's gut sense would produce those curves.

Third, measurement forces definition. Districts that commit to measuring self-management have to agree on what it means, which grade-level behaviors show it, and which practices build it. That shared language is often the most valuable byproduct.

The reason that does not hold up is accountability. We cover why below, and the ESSA and SEL resource explains how federal policy handled the question.

What the CORE Districts learned

The CORE Districts are a network of large urban California districts. According to the PACE brief, six districts implemented the federal waiver that made the SEL survey part of school accountability: Fresno, Long Beach, Los Angeles, Oakland, San Francisco and Santa Ana unified school districts, with Garden Grove and Sacramento City also part of the network. The network received its waiver from the US Department of Education in 2013 to build an alternative to the No Child Left Behind accountability system.

The timeline, from West's Education Next and Brookings accounts and Education Week's December 2015 report:

• 2013-14: pilot with about 10,000 students, including experiments comparing alternative survey items. Survey burden was capped at under 20 minutes per student each spring.
• 2014-15: field test with more than 450,000 students in grades 3 through 12.
• 2015-16: the four measures were publicly reported and counted for eight percent of a school's score on CORE's School Quality Improvement Index.
• December 2015: the Every Student Succeeds Act voided the waiver's accountability requirement. The districts kept collecting the data for school improvement.

The instrument itself was short. The PACE brief describes 25 questions on a five-point scale, answered by students in grades 4 through 12, with the same items each year. Roughly 390,000 students, about 70 percent of those enrolled in the surveyed grades, completed it in each of 2014-15 and 2015-16.

What did the field test show about quality? West reported that the scales showed strong reliability, with growth mindset the exception at about 0.7 against a conventional 0.8 benchmark, and that correlations between the SEL measures and academic indicators ranged from 0.33 to 0.69. He also ran a check for reference bias: if students in high-expectation schools systematically rated themselves lower, within-school correlations with achievement would be stronger than the overall correlation. They were weaker, which he read as limited evidence that reference bias was not a major problem across the network as a whole, while noting it could still distort results for individual schools with distinctive climates.

What did districts do with it? The 74 reported in 2019 that CORE schools see results on a school-level dashboard and can compare themselves with similar schools. Long Beach Unified set no mandated response. Some schools adopted an SEL curriculum, others ran growth-mindset nights for parents and teachers, and one high school built a two-period math course for struggling students that alternates algebra with SEL lessons. The same article quotes Education Analytics' Andrew Rice cautioning that the measurement science is still early and that the data should not drive high-stakes decisions.

Research: The PACE authors stress that CORE's measures are self-reports. Differences across students or over time may reflect differing normative standards rather than real skill differences. They urge caution in interpreting subgroup gaps and trends, even while noting that the measures relate to academic and behavioral indicators in expected ways.

Four ways to measure, and what each is good for

Angela Duckworth and David Yeager's 2015 paper Measurement Matters compared the main approaches. The table below summarizes their analysis and adds administrative data, which districts already hold.

Duckworth and Yeager's advice is to use multiple methods where possible and to choose the most valid measure for the intended purpose. A measure developed for basic research is not automatically valid for a district decision.

Try this: Before adopting any instrument, write one sentence that begins "We will use these results to decide..." If you cannot finish the sentence, you are not ready to survey.

Reliability, validity and reference bias, honestly

Reliability asks whether a measure gives consistent results. Validity asks whether it measures what you think it measures, for the use you have in mind. SEL surveys can be reliable and still be invalid for a particular use.

Reference bias is the central threat. Duckworth and Yeager describe it as respondents judging themselves against systematically different implicit standards. Their examples are uncomfortable: within countries, self-reported conscientiousness tracks performance, but between countries higher conscientiousness ratings went with lower math and reading scores. Students at middle schools with higher admissions standards rated themselves lower in self-control. Seniors rated themselves grittier than juniors while performance tasks showed the opposite. Students at KIPP charter schools reported more homework and scored higher on tests but did not report higher self-control. What one student calls "rarely," another calls "often."

A 2022 study in Scientific Reports led by Benjamin Lira with Yeager and Duckworth among the authors extended this with large adolescent samples. The pattern held: students surrounded by higher-achieving peers rated their own self-regulation lower, and task-based measures did not show the same distortion.

The consequence Duckworth and Yeager draw is blunt. For accountability, the answer is "not yet," and they do not believe any available measure is suitable for between-school accountability judgments. They also warn, citing Campbell, that indicators become corrupted once they drive decisions: students can be coached to give desired answers and schools can alter testing conditions. And a handful of questionnaire items cannot diagnose an individual child.

Education Week's March 2016 coverage made a further point: even if a measure is valid and reliable, high-stakes use raises separate questions about how low-scoring schools will respond and whether those responses help students. West's own summary was that CORE's results said nothing about how self-reports would perform in a high-stakes setting.

Anchoring vignettes, short descriptions of hypothetical students that respondents rate before rating themselves, have been proposed to reduce reference bias. Duckworth and Yeager note the approach is unverified in this context.

What to do with the data, and what not to do

Do:

• Report at the school and grade level, and look at change over time within the same school. Within-school trends are less exposed to reference bias than cross-school rankings.
• Disaggregate by gender, income, English learner status, disability and race, and treat gaps as questions to investigate rather than verdicts. The CORE pattern of girls' falling self-efficacy in middle school is a prompt for teacher conversation, not a label on girls.
• Pair survey results with implementation data. If self-management scores are flat, first ask whether the self-management lessons were actually taught.
• Debrief results with students. Washoe County School District, which built its own free competency assessment, describes its primary use as low-stakes decision-making, discussion and trend monitoring, including debriefing data with students, families and staff.
• Tell families in plain language what you measure, why, and how the results will and will not be used.

Do not:

• Rank schools or teachers on SEL survey scores.
• Place individual students in interventions on the basis of a survey score alone.
• Put a social emotional skill score on a report card.
• Compare your district's averages to another district's as if the difference were a skill gap. Different norms, different items and different survey conditions make the comparison unreliable.
• Survey without a plan for who will look at the results and when.

The data-informed SEL resource turns these principles into a meeting protocol.

A Monday-morning sequence for a school team

1. Name the constructs. Pick two to four skills tied to your school's SEL standards. Fewer constructs, measured well, beat a long battery nobody reads.
2. Define them in grade-band language. For elementary self-management, that may be "returns to work after a frustration within a few minutes." For high school, "plans the steps of a long assignment and checks progress."
3. Choose instruments by purpose. Student self-report for trends and reflection; teacher report for within-school conversation; existing attendance and discipline data for triangulation. See the instrument comparison.
4. Set the calendar. One fall and one spring administration is enough for most schools. Keep each under 20 minutes, following CORE's cap.
5. Handle consent and privacy before day one. Notify families, offer an opt-out, and check whether any item touches a protected area under federal survey law.
6. Teach the skills you measure. Use the strategy banks in the growth mindset, self-efficacy and self-management toolkits.
7. Review results in a team meeting with a protocol: describe what you see, then interpret, then decide one action.
8. Reflect at year's end. Did the data change a decision? If not, change the measure, the meeting or the question.

Common pitfalls

• Measuring before teaching. A spring survey after a year with no explicit SEL instruction measures the weather, not the program.
• Treating a decline as failure. PACE's trend data show that self-efficacy and social awareness typically fall through middle school. A flat line may be a win.
• Over-reading small differences. A 0.1 shift on a five-point scale, from a sample of 60 students, is noise. Look for consistent direction across grades and years.
• Letting the dashboard replace the conversation. Washoe's approach of data debriefs with students and staff is the model; a PDF emailed to principals is not.
• Changing items every year. CORE kept the same 25 questions so trends meant something. Resist the urge to tinker.

Where this leaves measuring social emotional skills

The honest position is the one the CORE Districts arrived at: measure, publish to educators, and act on the results, but keep the stakes low. The surveys taught California schools where confidence collapses and which practices seemed to help. They did not, and could not, tell anyone which school was better. Build your measurement plan around the first set of questions and you will get most of the value with little of the harm.