What the study found
Computer-based formative assessment (CBFA) data could predict later standardized assessment performance, but not as well as past standardized assessment measures. The strongest model used mean abilities across different competence domains and explained a considerable proportion of variance, about 30–48%.
Why the authors say this matters
The authors say the findings provide insights into how learning progress connects to future achievement. They conclude this may help teachers adapt instruction earlier and may inform policies that reduce reliance on high-stakes testing.
What the researchers tested
The researchers estimated student abilities in a large sample of children assessed at different time points during compulsory schooling. They then compared regression models trained to predict standardized assessment abilities using different subsets of features derived from formative assessment abilities and auxiliary variables.
What worked and what didn't
A model including mean abilities in different competence domains performed best for predicting standardized assessment abilities. The most predictive formative assessment features generally came from the same or a similar competence domain as the standardized assessment ability being predicted. Even so, the predictive power was still below that of past standardized assessment measures, and the authors report systematic model biases that would matter for decision-making.
What to keep in mind
The abstract does not give detailed limitations beyond noting systematic model biases. The results are based on one large sample of children and on the specific features and models described in the study.
Key points
- Computer-based formative assessment data predicted later standardized assessment performance.
- The best model used mean abilities across competence domains and explained about 30–48% of variance.
- Predictive formative assessment features usually matched the same or a similar competence domain as the outcome being predicted.
- Past standardized assessment measures still predicted better than the formative assessment-based models.
- The authors report systematic model biases that would need attention in decision-making.
Disclosure
- Research title:
- Formative assessment data predicted standardized assessment performance
- Authors:
- Benjamín Garzón, Stéphanie Berger, Charles Driver, Martin J. Tomasik
- Institutions:
- Kantonsschule Enge
- Publication date:
- 2026-02-23
- OpenAlex record:
- View
- Image credit:
- Photo by hamiltonpaviana on Pixabay · Pixabay License
Get the weekly research newsletter
Stay current with peer-reviewed research without reading academic papers — one filtered digest, every Friday.