Examiner Intelligence

SIGNAL · WEC14 Examiner Report Intelligence — compiled 2026-09-29, drawing on research/veridian/WEC14-verified-facts.md (14 real exam series, October 2020–January 2025, 13 with a matching examiner report; June 2021 has no examiner report in the archive; a narrower Summer 2025 pass added later checked one question only and is not counted in the "13" below). Every quote is a short fragment (under ~15 words) with its own series/question citation; anything longer than that is paraphrased. Not affiliated with or endorsed by Pearson Edexcel.

#WEC14 Examiner Report Intelligence (Unit 4: Developments in the Global Economy)

#What this is, and how it differs from what's already built

This paper's technique hub already has a content-error taxonomy (forge-error-pattern-tracker.md, ten codes, each tied to what candidates get wrong), a term-pair and command-word compression (forge-vocabulary-precision-list.md), and a topic-frequency table (signal-frequency-and-boundaries.md, which sub-sections get tested and how often). None of those asks a different question this document exists to answer: not what the errors are, but how the reports talk about them — which sentences get reused, almost word for word, series after series, and which are freshly written each time; who a given sentence is actually addressed to; and what the vocabulary an examiner reaches for implies about how a script is actually being read. Several quotes below overlap with evidence already used for content purposes elsewhere in this hub — that's unavoidable, since both documents draw on the same 13 reports — but every pattern here is about the language, not the underlying mistake, and is flagged as new where it is.

A pattern below is only called a pattern where at least two independently named series show the same wording or structure; a single instance is labelled a single finding instead, per this course's own standing rule.

---

#1. Some descriptive sentences are boilerplate; most are bespoke — and you can tell which from the wording alone

Most performance sentences in these reports are written fresh for that series: "Many candidates discussed education and wages" (January 2024, Q8) or "several candidates copied paragraphs from the extract" (October 2021, Q7(d)) name a specific stem and a specific answer pattern that will not reappear in that exact form next series.

One sentence breaks that rule. The single weakest Section A question is repeatedly reported in language that is not just similar but textually recycled:

• June 2022, Q6: "the least well answered question" (terms of trade)
• October 2022, Q6: "the least well answered question" — identical wording, a different series
• January 2023, Q6: the same clause again — the verified-facts research names this instance "verbatim identical" to October 2022's
• June 2023, Q5: the same clause, third consecutive year
• June 2024, Q4: the same clause, a fourth time

Five series, one recycled sentence, naming terms of trade every time it appears. It is not an empty ritual reused regardless of the facts: October 2021 is the honest counter-example — that series names regressive tax and tariffs as its weakest MCQs and reports terms of trade among the best-answered instead, and the boilerplate sentence does not appear that year. The sentence shows up only when it is true, which is exactly what makes its recurrence worth reading as a signal rather than noise: when this specific wording appears in a report, it is naming a standing weakness of the paper, not a one-year fluke, and it will very likely name the same topic again next series. A bespoke sentence about a specific stem tells you about that stem only; this one sentence, wherever it lands, tells you about the paper as a whole.

#2. "A common response was to explain it in terms of…" reports the modal answer, not the mark

Section B's 4-mark "explain what is meant by [term]" parts get a distinct reporting formula: examiners state what most of the cohort actually wrote, in a fixed phrase, and then separately state whether that was worth anything. The two clauses are logically independent, and the formula is used identically whichever way the second clause goes:

Four series use the exact stem "a common response was to explain it in terms of…"; October 2020 uses the adjacent variant "a common response was to define…". Either way, the phrase names the statistically dominant answer across the whole cohort, not the correct one — the report's opening clause is a description of what most people wrote, drawn from reading a large number of scripts, and the credit judgement is a second, separate sentence you have to read for independently. Do not treat "a common response was X" as Pearson telling you X is right; read past it to whatever the report says next.

#3. Two audiences share the same reports, in two different sentence shapes

Most of a report's sentences narrate that series' candidates: "many/some/few candidates were/were not able to…" — descriptive, past tense, about a specific cohort. A smaller, separate set of sentences is not about the candidates at all; it is an imperative addressed to whoever prepares them:

• "Ensure that all parts of the specification are taught and internally assessed" — this clause, or a close paraphrase of it, appears in at least 10 of the 13 reports read for the facts bank, named directly in October 2020, January 2021, October 2021 and January 2022.
• "It is important to use the data carefully for calculation-based questions" — recurs across Q7(a) commentary in at least October 2020, January 2023 and October 2023, and per the facts bank's own reading, in nearly every other series' Q7(a) section too.

Neither sentence describes what that year's candidates actually did; both are standing advice to centres, reissued close to verbatim regardless of how that specific paper went. Read alongside pattern 1, this gives three sentence types inside the same document, not two: bespoke narration (most of the report, tells you about that stem only), the one recycled descriptive sentence in pattern 1 (tells you about a structural weak point, not that year), and recycled imperative advice like the two above (tells you nothing about any specific series at all — it is the same reminder whether or not it was heeded). Weighting revision by report language without separating these three is a way to mistake "Pearson says this every year regardless of outcome" for "this year's cohort specifically struggled here."

#4. What the credit-outcome vocabulary implies about how a script gets read

Set side by side, the closing clause of a trap sentence splits cleanly into two vocabularies, and the split tracks this paper's own KAA/Evaluation structure rather than being random word choice.

Paired-term and definitional confusions close in total-loss language: "unable to access any marks" (October 2022, Q7(d), fiscal/monetary policy; October 2023, Q7(d), IMF/World Bank), "did not get any marks" (June 2024, balance-of-trade deficit/surplus), "did not attain any marks" (June 2022, Q10, causes substituted for policies). Each of these is a discrete, binary KAA-marking point on the mark scheme — the term is either the one asked for or it isn't, so the report's language records a pass/fail outcome because the mark scheme itself has no partial setting for it.

Evaluation shortfalls close in graduated, ladder language instead: "did not achieve more than Level 1" and "only carried a two-stage chain of reasoning" (January 2024, Q8), "struggled to evaluate effectively" with points scoring no better than Level 1 (October 2021, Q9). Nothing here is described as scoring zero — because Evaluation is banded (Level 1–3 or 1–4, not a fixed list of discrete points), so a weak answer still lands somewhere on the ladder, and the report's vocabulary reflects that graduated structure rather than a binary one.

That the reports' own word choice mirrors the mark scheme's KAA-versus-Evaluation architecture this precisely is a real, checkable claim about how these scripts are read: a definition is checked for presence or absence of the required term, credited or not; an evaluative argument is read for how far up a developing chain of reasoning it gets, and marked on where it lands. It is also the practical payoff of this whole document — AO4 (evaluation) is 30.0% of Unit 4 alone, the largest single assessment objective in this unit (spec p. 51 AO table, already verified in the facts bank) — so the graduated-ladder half of this vocabulary split describes where most of the unit's marks, and its risk, actually sit.

---

#Sources, coverage and honest gaps

Every quote above is drawn from research/veridian/WEC14-verified-facts.md, itself checked against the primary Pearson PDFs for 14 series (October 2020–January 2025); this document did not independently re-open those PDFs for this pass, and inherits that source's own stated limits — no June 2021 examiner report exists in the archive; June 2022 and June 2023 reports were text-extracted from image-heavy PDFs; June 2024's report was read by rendering pages as images, not pdftotext, because of an unmapped embedded font. "Repeated in at least 10 of the 13" and "nearly every series" in section 3 are the facts bank's own stated counts, not a fresh tally against the raw archive for this document.

This is a language-and-register analysis, not a fifth content-error list — where a quote above also appears in forge-error-pattern-tracker.md or forge-vocabulary-precision-list.md, that is the content-level treatment of the same underlying mistake; this document's only claim is about the recurring wording and what it implies, not a restatement of the mechanism. No pattern above generalises past its named series, and none converts a frequency into a probability — "seen in N of M series checked" is as far as any claim goes.

---

Provenance: this document sits alongside forge-error-pattern-tracker.md (content-level error codes), forge-vocabulary-precision-list.md (term-pairs and command words) and signal-frequency-and-boundaries.md (topic frequency) in WEC14's technique-hub build; compiled 2026-09-29.