NEWLY PUBLISHED: Understanding the Response to Financial and Non-Financial Incentives in Education: Field Experimental Evidence Using High-Stakes Assessments.
Can incentives for high school students work?
Yes
Want more detail? Read on …
(1/10)
Can incentives for high school students work?
Yes
Want more detail? Read on …
(1/10)
The paper is out now:
authors.elsevier.com/c/1e2rP~1QBz22E
Joint with Sally Sadoff and Rob Metcalfe.
The original large-scale RCT generously funded by @EducEndowFoundn
(2/10)
authors.elsevier.com/c/1e2rP~1QBz22E
Joint with Sally Sadoff and Rob Metcalfe.
The original large-scale RCT generously funded by @EducEndowFoundn
(2/10)
Key design feature: We incentivize pupil inputs not outputs, that is, behaviours not grades.
Who will this affect?
Students already working hard – likely not much. But for other students, this might encourage greater effort.
(3/10)
Who will this affect?
Students already working hard – likely not much. But for other students, this might encourage greater effort.
(3/10)
Our field experiment implemented in high schools in England, in the final year of compulsory schooling.
The experiment included over 10,000 students in 63 high schools across England.
The outcome measures were (high-stakes) tests in English, Maths and Science (GCSEs).
(4/10)
The experiment included over 10,000 students in 63 high schools across England.
The outcome measures were (high-stakes) tests in English, Maths and Science (GCSEs).
(4/10)
Average impact is positive, but small and insignificant.
We study the distribution of treatment effects: half of the students have economically meaningful positiv effects.
For these, GCSE scores improve by 10% - 20% SD; the chance of hitting 5A*C increases by 8 ppt.
(5/10)
We study the distribution of treatment effects: half of the students have economically meaningful positiv effects.
For these, GCSE scores improve by 10% - 20% SD; the chance of hitting 5A*C increases by 8 ppt.
(5/10)
How do we identify these highly affected groups?
Simple subgroup analysis.
Also machine learning techniques to guard against researcher discretion: well-established approaches and the more recent Random Causal Forest methods. They all point the same way.
(6/10)
Simple subgroup analysis.
Also machine learning techniques to guard against researcher discretion: well-established approaches and the more recent Random Causal Forest methods. They all point the same way.
(6/10)
The “right tail” of highly-responsive students is well proxied by characteristic: disadvantaged students who are native English speakers.
Students who have English as an additional language score highly at GCSE and are largely unaffected by the incentives.
(7/10)
Students who have English as an additional language score highly at GCSE and are largely unaffected by the incentives.
(7/10)
The students predicted to be most responsive to both financial and non-financial incentives in maths are those with lower performance at baseline.
Our results suggest that incentives could close achievement gaps among these students by about half.
(8/10)
Our results suggest that incentives could close achievement gaps among these students by about half.
(8/10)
We discuss feasible and acceptable implementation of these policies in the paper.
While it is not nudge-like cheap, the fact that it was particularly effective with low-performing, hard-to-reach groups of students makes it worth considering.
(9/10)
While it is not nudge-like cheap, the fact that it was particularly effective with low-performing, hard-to-reach groups of students makes it worth considering.
(9/10)
May be of interest: @hhsievertsen @AnnaVignoles @rabe_b @Sandra_McNally @mike_t_sanders @lindsey_mac @jdportes @timleunig @Woessmann @hooster1
@EducEndowFoundn
authors.elsevier.com/c/1e2rP~1QBz22E
(10/10)
@EducEndowFoundn
authors.elsevier.com/c/1e2rP~1QBz22E
(10/10)
• • •
Missing some Tweet in this thread? You can try to
force a refresh
