How to Scale an Educational Intervention Effectively

How to cite

Holm, Jakob Majlund; Andersen, Simon Calmar; Ahrensberg, Nanna Vestergaard; Hansen, Jesper Asring Jessen; Hjortskov, Morten. 2023. "How to Scale an Educational Intervention Effectively." AEA RCT Registry. AEARCTR-0011098. https://doi.org/10.1257/rct.11098-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Jakob Majlund Holm* , Simon Calmar Andersen , Nanna Vestergaard Ahrensberg , Jesper Asring Jessen Hansen , Morten Hjortskov

* Primary investigator

Registry ID
AEARCTR-0011098
DOI
https://doi.org/10.1257/rct.11098-1.0
First registered
Last updated
Status
In development
Trial dates
2023-03-24 – 2024-10-01
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/11098

Abstract

Many interventions aimed at improving students' reading skills have proven effective in small-scale evaluations. However, when educational interventions include more than a thousand students, effect sizes tend to approach zero. This problem of scalability is recognized in the literature, yet there exists very little systematic, research-based knowledge on how to solve it.

If behavioral interventions cannot scale effectively, the societal value of the rigorous experimental studies is limited. Therefore, it is of utmost importance—and the purpose of this study—to understand how interventions can be made effective at scale.

To study this question, we use an experimental approach where we randomly assign approximately 630 schools including approximately 6,000 students to either (i) a Control condition, (ii) a Treatment-as-Usual condition, or (iii) an Implementation Support condition.

Intervention

The intervention consist of physical and electronic children's books combined with advice to parents on how to do shared-book reading. Additionally, in one of the treatment groups, the intervention contains support to schools and families in order to facilitate implementation of the program.

Experimental design

The Control condition will receive no invitation to participate in the program.

The Treatment-as-Usual condition will receive an invitation to participate but with only the support that have negligible marginal costs (such as e-mail reminders) and limited amount of human capital intensive implementation support.

The Implementation Support condition will receive additional personal, human capital intensive support throughout the intervention period.

Design details

Randomization method
Randomization is done in office by a computer
Randomization unit
We randomize at the school level
Clusters
Approximately 630 schools
Observations
Approximately 60,000 students
Arms
Approximately 420 schools invited, 210 schools control
Primary outcomes
The primary outcome is children reading test score measured by the national reading test around September 2024.

Cited by

No citation data available for this registration yet.

Keywords