---
title: Teaching at the Right Level
summary: Grouping primary school children by what they can actually read and calculate, rather than by grade, and teaching each group the basic skills it is missing.
depth: 2
evidence: strong
mechanism: A simple assessment sorts children by learning level. For a set time each day, or in short intensive camps, they work in groups on the reading and math skills they lack, and teachers use data on what each child can do to guide the groups.
causes: [education]
timeToEffect: months
durability: years
feedback: strong
agency: top-down
cost:
  outcome: 0.1 standard deviation gain in one student’s broad foundational math score (Zambia’s government-run Catch Up program)
  low: 6.38
  high: 6.38
  asOf: 2022–2024 trial
  source:
    title: Targeting Foundational Skills to Improve Learning at Scale in Zambia
    publisher: J-PAL
    url: https://www.povertyactionlab.org/evaluation/targeting-foundational-skills-improve-learning-scale-zambia
  note: A point estimate from one randomized trial, not an uncertainty interval. The program cost $9.63 per student over two years. Counting only the procedural math skill it targeted gave a different estimate, $2.43 per 0.1 standard deviation, so the choice of test changed apparent cost-effectiveness by a factor of 2.6. The calculator uses the broader measure.
conditions: [measurement-gap]
sources:
  - title: Teaching at the Right Level to improve learning (evidence-to-policy case study)
    publisher: J-PAL
    year: 2019
    url: https://www.povertyactionlab.org/case-study/teaching-right-level-improve-learning
    note: Last updated 2022.
  - title: 'Mainstreaming an Effective Intervention: Evidence from Randomized Evaluations of “Teaching at the Right Level” in India'
    publisher: Banerjee, Banerji, Berry, Duflo, Kannan, Mukerji, Shotland and Walton, NBER Working Paper 22746
    year: 2016
    url: https://www.nber.org/papers/w22746
  - title: 'From Proof of Concept to Scalable Policies: Challenges and Solutions, with an Application'
    publisher: Banerjee, Banerji, Berry, Duflo, Kannan, Mukerji, Shotland and Walton, Journal of Economic Perspectives 31(4)
    year: 2017
    url: https://www.aeaweb.org/articles?id=10.1257/jep.31.4.73
  - title: Targeting Foundational Skills to Improve Learning at Scale in Zambia
    publisher: J-PAL
    url: https://www.povertyactionlab.org/evaluation/targeting-foundational-skills-improve-learning-scale-zambia
  - title: New education “Smart Buys” report outlines how cost-effectively supporting teachers and parents can lead to significant learning improvements
    publisher: World Bank
    year: 2023
    url: https://www.worldbank.org/en/news/press-release/2023/05/09/education-smart-buys-cost-effectively-supporting-teachers-and-parents-can-lead-to-significant-learning-improvements
---

## What the evidence says

Teaching at the Right Level was developed by the Indian nonprofit Pratham and has been tested in randomized evaluations since 2001. J-PAL counts six randomized evaluations in seven Indian states and says the approach has produced some of the largest effects rigorously measured in education. In Uttar Pradesh, short learning camps doubled the number of children who could read a paragraph or story.

The evaluations also show where it fails. In Bihar and Uttarakhand, government teachers did not adopt the method, even after well-received training. It worked when it had protected time and support. In Haryana, teachers gave it a dedicated hour a day with mentoring from trained government staff, and language scores rose by 0.15 standard deviations. In Uttar Pradesh, intensive 40-day camps run by Pratham volunteers during school hours, plus summer camps, raised language scores by 0.70 standard deviations. J-PAL notes that training alone or materials alone did not improve learning.

At scale in Zambia the gains are smaller. A randomized trial of the government-run version, Catch Up, in 273 schools found gains of 0.10 standard deviations in literacy and 0.15 in numeracy after two years. Only about a third of students attended a remedial class on a given day, largely because many were absent, had dropped out or had moved schools. Students fully exposed to the program later scored 0.14 standard deviations higher in language and 0.11 in math on grade 7 exams.

In 2023 the Global Education Evidence Advisory Panel, convened by the World Bank and partners, listed teaching by learning level among its “Great Buys,” its top tier for cost and impact. The grade is strong because many rigorous trials agree that the approach raises basic skills when it is well implemented. How large the effect is depends heavily on delivery.

## Where it sits on the depth scale

Charitology places it at **prevention** (depth 2). It gives children who have fallen behind dedicated time at their own level, so that they don’t leave primary school without basic reading and math. When a ministry runs it through its own teachers, as Zambia does, it moves toward structure. Even then it works alongside the grade-based syllabus rather than replacing it.

## Open questions

- Why are effects so much smaller in Zambia’s government program than in the Indian camps? See [what randomized trials can tell us](/record/disagreements/what-randomized-trials-can-tell-us/).
- How can governments keep the protected time and mentoring that made the Indian versions work? See [the government question](/open-problems/the-government-question/).
- Would changing curricula and exams prevent the gap more cheaply than remedial instruction?