Quantify the causal impact of teacher quality on student outcomes, diagnose why teacher development systems in most countries are ineffective, and assess the evidence base for interventions that reliably improve classroom instruction at scale.

Meta-analysis of teacher effectiveness studies using value-added modeling across 25 OECD countries and 15 LMICs. World Bank TEACH classroom observation data analyzed across 40 countries to measure instructional quality distribution. J-PAL evidence synthesis on teacher incentive and development interventions. Cross-country analysis of pre-service and in-service teacher development systems.

Teacher quality is the single most powerful in-school determinant of student learning, explaining 30% of variance in outcomes — more than class size, school resources, or curriculum combined (Hanushek & Rivkin). Yet most education systems have no reliable mechanism for measuring, developing, or differentiating teachers based on instructional quality.

World Bank TEACH data across 40 countries reveals that in the median LMIC classroom, students are academically engaged less than 50% of instructional time. The top quartile of teachers produces learning gains equivalent to 1.5 additional years of schooling compared to the bottom quartile — within the same school.

The pre-service training failure: 68% of teacher training programs in LMICs are primarily theoretical with less than 10% of time on supervised classroom practice (OECD TALIS). Teachers graduate without demonstrated ability to manage a classroom or deliver structured instruction.

The feedback desert: 75% of teachers globally receive no formal classroom observation or instructional feedback in a given year. In the US, the average teacher receives 38 minutes of observation annually. Performance appraisal systems that exist are almost universally disconnected from development support.

What works: Structured pedagogy programs (scripted or semi-scripted lesson guides with embedded formative assessment) show consistent 0.2-0.5 SD learning gains across LMICs in RCTs. Coaching models with regular classroom observation and feedback show 0.15-0.3 SD gains. Performance-linked pay in isolation shows weak and inconsistent effects — only effective when paired with development support.