Two parallel groups and an eight-point gap
Two parallel L2 English groups in the same European School, with 24 and 22 students, follow an identical preliminary-mark scheme. In June the subject coordinator compares the two teachers' records: Group A's mean is 71 and Group B's is 63; six students in A are above 85 and none in B. Both groups sat the same Pre-Baccalaureate paper, each teacher marking their own class against the agreed grid. A blind exchange of four scripts shows the pattern: two B scripts re-marked by Teacher A gain 4 and 5 points; two A scripts re-marked by Teacher B lose 3 and 6. Teacher A's class spent three extra weeks on the examined text types; Teacher B's class had two changes of teacher in the autumn. Both teachers believe their own marks are right and that the other is either soft or harsh.
1. Decompose the eight-point gap. How much of it is attainment and how much is marking?
Read the worked answer
The gap is real in the sense that the numbers exist, but it is not one thing. Three sources need separating.
Genuine differences in opportunity to learn. Group A had three additional weeks on the examined text types, which is a curriculum advantage, not a marking issue. Group B's two changes of teacher cost continuity and probably some coverage. These plausibly account for a few points on the Pre-Baccalaureate paper, which carries half the mark, and they are legitimate: candidates in different classes are not guaranteed identical teaching.
Systematic marking difference. The exchange is the strongest evidence available. B scripts gain 4–5 points under Teacher A; A scripts lose 3–6 under Teacher B. That is a consistent directional effect of roughly 4–5 points in both directions, i.e. a severity/leniency gap of that order, not random noise. With only four scripts the estimate is imprecise, but the direction is consistent and the magnitude is large enough to explain most of the class-mean difference.
Distribution effects. Six students above 85 in A and none in B is a tail phenomenon that a mean difference alone does not explain. It suggests either genuinely stronger cohorts or a marking style that spreads marks at the top - both are worth checking against the exchange sample.
The honest conclusion: perhaps 2–4 points of the gap are attainment and 4–5 points are marking. The coordinator should say so rather than choosing one explanation.
2. Design the moderation procedure and state exactly what, if anything, you would adjust.
Read the worked answer
Before finalisation:
- Standardise on anchors. Both teachers independently mark the same three scripts (one strong, one middling, one weak) chosen by the coordinator, discuss criterion by criterion, and agree what a 10, a 13 and a 16 look like on this paper. Do this before any marks are compared.
- Blind second-mark a defined sample. All scripts within five points of a grade boundary, plus a stratified random sample of eight per group (four from the top half, four from the bottom). Remove names and the first marker's marks.
- Compute the mean difference per teacher, not per student, and check whether it is consistent across the mark range. A uniform shift suggests severity; a shift only at the top suggests a different interpretation of the top-band descriptors.
What to adjust. Adjust only what the evidence supports. Borderline and discrepant scripts are re-marked and the criterion-level evidence decides the individual mark. A group-level correction applied uniformly to all 46 candidates is defensible only if the sample shows a consistent bias across the range and the school's rules permit it; otherwise it is arithmetic dressed as moderation, and it will harm students whose marks were already accurate.
What not to do. Do not split the difference between the two teachers to keep the peace, do not adjust Group A downwards merely because its mean is high, and do not treat the 8-point gap as proof of anything until the exchange sample has been extended.
3. A student in Group B is at 59.5 and needs 60 to continue to the next level. What is the correct handling?
Read the worked answer
The 0.5 is not the issue; the evidence is. The script and the component marks should go into the blind re-marking sample described above, marked by Teacher A and by a third colleague who has not taught either group, with the 59.5 withheld.
- If two independent re-marks place the candidate clearly at or above 60 on the published criteria, the mark is corrected and the fact that it is a correction is recorded.
- If the re-marks confirm 58–59, the mark stands at 59. Rounding up because of the consequence is a decision to admit a student to a level for which the evidence says she is not ready, and it is unfair to the Group A candidates who were marked to a different standard.
- If the re-marks diverge by more than the tolerance, the procedure in force decides.
The severity finding about Group B is a reason to re-mark, not a reason to inflate. There is a legitimate separate conversation about progression - whether the level decision should rest on a single composite mark at all - but that conversation belongs to the school's progression policy, not to June's marking meeting. The coordinator should also check whether any other Group B candidate sits within a point of a boundary, so that the same treatment is applied to all of them; handling only the case that was noticed would be the real unfairness.