The School-Based Assessment Moderation Dilemma
A school's Stage 2 Tourism class has 28 students. Two teachers each assess a different set of students against the same performance standards. Teacher A's marks for the 14 students in her group average 78%, with a range of 62–94%. Teacher B's marks for the other 14 average 64%, with a range of 55–81%. The school's moderation panel notes that the two groups had been randomly allocated, not streamed, and that the cohort's prior-year results in comparable subjects showed no meaningful difference between the groups. Teacher A's top student scored 94% on a task that Teacher B's students also completed, and Teacher B's highest mark on that task was 81%. The panel must decide whether to adjust marks before submitting final results to the external authority.
1. Identify the most likely problem and explain what evidence in the setup supports your diagnosis.
Read the worked answer
The most likely problem is inconsistent application of the performance standards between the two teachers, not genuine differences in student ability.
Evidence:
- Students were randomly allocated, so there is no reason to expect a 14-point gap in group averages.
- Prior-year results in comparable subjects showed no meaningful difference between the groups, which undercuts an ability explanation.
- The range compression in Teacher B's group (55–81%) versus Teacher A's (62–94%) suggests Teacher B may be marking more conservatively across the whole spread, not just at the top.
- The same task produced a 94% top mark in one group and an 81% top mark in the other, which points to differing interpretation of what a top-band response looks like.
The panel should treat this as a comparability issue requiring moderation, not as evidence that Teacher A's students are stronger. The diagnosis is about the reliability of judgements, not about the students.
2. Recommend a moderation action and justify why it is preferable to simply averaging the two teachers' marks.
Read the worked answer
A defensible action is to re-mark a sample of both teachers' work against the standards, then adjust marks so the two groups are judged on a common scale - for example, by having the panel independently re-assess the shared task across both groups, calibrating on annotated exemplars, and applying the resulting adjustment to Teacher B's marks (or to both sets if the panel finds Teacher A lenient).
Why not simply average the two teachers' marks?
- Averaging assumes both teachers are equally wrong in opposite directions, which the evidence does not show.
- It would penalise Teacher A's students for being marked against a standard the panel has not yet tested, and it would not fix the underlying inconsistency.
- Averaging is not anchored to the performance standards, so it does not produce defensible judgements.
Moderation should be evidence-based and standards-referenced: the panel compares actual student work to the descriptors, identifies where judgements diverge, and adjusts so that a given quality of work receives a comparable mark regardless of teacher. This protects both fairness to students and the credibility of the school's results.
3. Explain one risk if the panel does nothing, and one risk if it over-corrects.
Read the worked answer
Risk of doing nothing: Students in Teacher B's group may be systematically under-rewarded, which disadvantages them in any ranking or selection process that uses these results and undermines confidence in the school's assessment. It also leaves the inconsistency unaddressed for future cohorts.
Risk of over-correcting: If the panel inflates Teacher B's marks without evidence, it could overstate student achievement, misrepresent the cohort to the external authority, and create a new fairness problem for Teacher A's students. Over-correction also erodes trust between staff and the moderation process.
The judgement call is to adjust only as far as the evidence supports, documenting the reasoning so the decision is transparent and reviewable. This is the balance the exam rewards: recognising that moderation is about comparability of judgements, not about equalising averages for their own sake.