The Evidence on Teaching

Changing How Literacy Is Taught: Evidence on Synthetic Phonics

Machin, S., McNally, S., & Viarengo, M. · 2018

grade Bnatural-experimentindependentreplicatednumbers spot-checked
Sample
Whole-population England pupil census (millions of pupils, cohorts 2005-2010)
Population
England state-school pupils tracked ages 5-11; subgroups FSM (low-income) and EAL (non-native).
Design
Difference-in-differences on the staggered rollout of England's CLLD synthetic-phonics teacher-training programme (early- vs later-adopting local authorities). A low-cost intervention retraining existing teachers. Cleanest at-scale identification in the phonics-policy literature.
Key findings
The corroborating natural experiment. Real early reading gains (0.22-0.30 SD at age 5) fade to ~0 on population average by age 11, persisting only for disadvantaged/EAL pupils (0.05-0.18 SD). Reproduces the RCT signature at national scale: modest, decoding-specific, fadeout-prone, durable mainly for the at-risk tail.
Genetic confound
Quasi-experimental staggered adoption; treatment assigned at LA/school level, not pupil self-selection - addresses passive rGE.
Replication notes
Pattern (early gain + mean fadeout + at-risk persistence) converges with Reading First RD (Gamse 2008) and Suggate 2016.
DOI / URL
10.1257/pol.20160514

Effects

OutcomeMetricValueMeasureTimingVsHorizonClass
Reading, age 5 (end of treatment)SD+0.22 to +0.30mixedend-of-treatmentbusiness-as-usualend-of-treatmentdomain-skill
Reading, age 7SD~+0.08 (attenuating)standardized2yr follow-upbusiness-as-usual1-2yrdomain-skill
Reading, age 11 (whole population)SD~0.00 (mean fadeout)standardized6yr follow-upbusiness-as-usualover-2yrdomain-skill
Reading age 11, disadvantaged FSM / EAL subgroupsSD+0.05 to +0.18 (persists)standardized6yr follow-upbusiness-as-usualover-2yrdomain-skill

Cited by