NEW HAVEN — Achievement First, the charter-school network that operates campuses in three states, on Friday released five years of longitudinal reading-recovery data that has, in the time since the data was first compiled, attracted attention well beyond the network's home districts.

The data, which the network has previously shared in academic settings but is now publishing in fuller form, tracks reading-proficiency trajectories for cohorts that experienced the most disrupted early-elementary years during the 2020-2022 period and that have since been the subject of the network's targeted reading-recovery programming.

What the data shows

The most consequential finding is that the proficiency gains the network's recovery programming produced in the initial year have, contrary to the early scepticism, held across the subsequent two academic cycles. The cohorts that experienced the recovery programming continue to outperform comparison cohorts on the network's internal assessments and on the relevant state assessments.

The persistence is, on the academic-research framing, the most important question. Many recovery programmes have produced initial gains that did not hold. The Achievement First data is one of the longer longitudinal records suggesting that some recovery interventions can produce durable outcomes.

What the programme does

The network's recovery programming combines structured small-group instruction, daily individual reading time, and a teacher-coaching model that has been the most labour-intensive element. The teacher-coaching component requires roughly forty hours of professional development per teacher in the first year and continues at lower intensity in subsequent years.

The labour intensity is the principal constraint on broader replication. Achievement First's network has maintained the resource commitment across all five years; whether other systems can sustain comparable commitments is the question on which broader replication turns.

The methodological caveats

The published data are accompanied by methodological caveats that the network has been careful to surface. The comparison cohorts are not perfectly matched on demographic and prior-attainment dimensions; the assessment instruments have changed in some respects across the five years; the participating schools' broader environments have evolved in ways that the analysis can only partially control for.

None of those caveats undermines the core finding, but they do limit the strength of the inference that can be drawn. The network has been explicit about this, in part because the alternative — over-claiming on the basis of imperfect data — would, in the network's view, ultimately reduce the credibility of the underlying programme.

What the broader research community is asking

The broader research community has, in the days since the data's release, focused on three questions. The first is whether the network can share more granular intervention-fidelity data that would allow the relative contribution of programme components to be assessed. The second is whether the cost-effectiveness data sufficient for cross-system comparison can be developed. The third is whether the persistence pattern continues into middle-school years.

The network has signalled it is open to engaging with each of those questions. The relevant data takes time to assemble in publishable form; the next round of public data is expected in roughly eighteen months.