Our method is straightforward to implement and applicable in any setting with subjective assessments and a way to identify marginal assessees!

Take a look at the paper for more info.
econpapers.repec.org/paper/uclcep...

7/7

We can also study heterogeneity by subject and grade boundary:

We find more bias in subjects where teachers have more discretion (prior attainment is less informative)

There is heterogeneity by grade boundary but no clear patterns

6/7

Our findings are robust to a load of different specifications and samples -- our main analysis uses ALL marginal students (ranked adjacent to a boundary) across all subjects and boundaries

5/7

We find teachers favour female students over male and high-SES over lower SES -- inline with previous work

Crucially we find no effect at "placebo" grade boundaries between adjacent students within an assessed grade

4/7

We check for gender/SES/ethnicity balance across grade boundaries, using only the most marginal students (as ranked by their teachers) -- like an RDD balance test but with a twist, as our running variable is ordinal

3/7

We estimate bias directly, without comparing blind and non-blind assessments, so without the usual assumptions.

Our unique data comes from when high-stakes A-level exams were cancelled in 2020, and replaced by teacher-assessed grades and ranks

2/7

We provide new evidence -- via a brand new method !! -- on the issue of bias in teacher assessments. Very relevant right now, with many US colleges going "test-optional" and implicitly relying on (teacher assessed) GPA

1/7