Paper 4 — The Global Economy
Evaluations Guide
FORGE · WEC14 Evaluations Guide — compiled 2026-09-29. Every claim below traces to a named mark scheme or examiner report, either via research/veridian/WEC14-verified-facts.md or via a direct primary-source extraction performed for this document specifically (see §7 for exactly which files, how, and what was cross-checked across how many series). forge-level-threshold-guide.md and forge-exam-technique-guide.md were both read in full before writing this, so this document does not restate their worked level-boundary pairs (the Italy Q10 Evaluation ladder; the Argentina Q10 KAA ladder; the country gate) or their paper-wide technique findings (the KAA/Evaluation cutoff between the 6-mark and 8-mark Section B parts; "positive as KAA, negative as Evaluation"; two-stage chains capping Level 2 KAA on Section C). codex-complete-worked-paper.md (October 2021) was also read in full so this document's own worked example uses a different series and a different question. This document's job is the layer underneath all three: the actual mark-scheme text the paper uses to define an Evaluation mark, extracted directly from the primary PDFs rather than inferred from examiner-report paraphrase — which, per forge-level-threshold-guide.md §7's own honest gap, no source in this repo had done yet. Not affiliated with or endorsed by Pearson Edexcel.
WEC14 Evaluations Guide (Unit 4: Developments in the Global Economy)
0. What this document adds, and what it deliberately doesn't redo
Two documents already sit in this folder covering evaluation on this paper from two different angles. forge-level-threshold-guide.md works boundary-by-boundary, with worked just-misses-it/just-clears-it answer pairs, for the Section C Evaluation ladder (its §3) and the KAA ladder (its §2), built on Argentina's current-account deficit (Oct 2021 Q10) and Italy's public expenditure (June 2025 Q10). forge-exam-technique-guide.md works tariff-by-tariff across the whole paper, establishing where evaluation is credited at all and where it earns nothing (§2), and the paper-wide mechanics — the country gate, the flexible KAA/Evaluation direction, "informed judgement," two-stage chains — as examiner-reported findings (§4). Both are real, both check out, and neither needed to go one layer further down: into the actual printed words the mark scheme uses to define a Level, as opposed to what an examiner report says about a Level in prose.
That is this document's whole job. Every mark scheme in the archive prints two separate tables of generic Level descriptors for every levels-based question — one for Knowledge/Application/Analysis, one for Evaluation — before it ever gets to a question's own indicative content. These tables are the actual mechanism by which a Level gets assigned; the indicative content beneath them is only ever examples that "does not imply that any of these must be included" — the mark scheme's own words, repeated on every question. No document in this repo had transcribed or compared that generic table's wording directly until this pass — forge-level-threshold-guide.md §7 says so explicitly: "the generic, boilerplate wording of Pearson's own Level descriptor table... was not found verbatim in any source checked." This document closes that gap, and in doing so finds a real, precise, previously-unstated distinction between how the mark scheme itself defines the two ladders that carry Evaluation marks on this paper.
Method. The primary PDF archive for WEC14 (13 mark schemes, October 2020–October 2024, plus the June 2025 mark scheme already used elsewhere in this repo; 13 examiner reports, October 2020–January 2025) was re-fetched and re-run through pdftotext -layout for this pass. The generic Level descriptor tables were extracted and diffed by hand across three series spread across the archive — October 2021, January 2024, October 2024 — chosen to bracket the archive's early, middle and late thirds rather than clustering in one exam year. Every specific quote below was checked against at least one of these three extractions directly; where a claim is said to hold "across all three," it was checked in all three, not assumed from one.
1. Where AO4 lives on this paper — the map, briefly, before going one layer down
forge-exam-technique-guide.md §2 and forge-level-threshold-guide.md §1 both already establish this in full; it is repeated here only as the minimum scaffold the rest of this document needs.
| Component | Evaluation marks | Mechanic |
|---|---|---|
| Section A; Section B's 2/4/6-mark parts | 0 | No AO4 at all — confirmed directly in the mark scheme's own mark-breakdown line for each part ("Knowledge 2, Application 2" with no "Evaluation" listed) |
| Section B's 8-mark part | 2 | Flat, additive — not a level (§2 below) |
| Section B Q7(e) | 6 | Levels 1–3 (own descriptor table, §3 below) |
| Section C essay (×2) | 8 each | Levels 1–3 (own descriptor table, §3 below) |
Total Evaluation marks on the paper: 2 + 6 + 8 + 8 = 24/80, matching the spec's own 30.0% AO4 weighting for the unit — the same arithmetic forge-exam-technique-guide.md §2 already gives.
2. The Section B 8-mark part's Evaluation: "(2+0 or 1+1)," not a level at all
This is a genuinely different mechanic from the two ladders in §3, and worth stating in its own right because it is easy to assume "Evaluation" means the same marking shape everywhere it appears on this paper. It doesn't. The mark scheme's own instruction line for this part reads, in every series checked (October 2021, January 2022, January 2024, October 2024): "Up to 2 marks for evaluative comments." In two of those four (January 2022, and October 2024 with the order swapped) the line spells the split out explicitly:
"Up to 2 marks for evaluative comments (2+0 or 1+1)" — January 2022 "Up to 2 marks for evaluative comments (1+1) or (2+0)" — October 2024
October 2021 and January 2024 print the shorter header without that parenthetical, but their own worked indicative content shows the identical mechanic in practice (below) — so the split itself is not a one-series feature even where the header line doesn't spell it out; only the explicit parenthetical is inconsistently printed. Either way, the mechanic is the same: a candidate can earn the 2 marks either by developing one evaluative point far enough that it is worth 2 marks on its own (2+0), or by making two separate, shallower evaluative points, each worth 1 mark (1+1). There is no third mark, no level, and no requirement that the two marks come from a single connected argument — this is additive scoring exactly like the Knowledge/Application/Analysis marks either side of it on this same tariff, not a miniature version of the Level 1–3 ladder in §3.
A real worked instance. January 2024's Q7(c) asks candidates to "examine two measures" Greece's government could take to raise its international competitiveness, drawing on Extract A — 8 marks (K2, AN2, AP2, EV2). It is one of the two series whose header line doesn't print the parenthetical, which makes its own indicative content useful evidence that the split still applies. Its Evaluation bullets credit: "Increase in budget deficit/national debt" as one mark, paired with "resulting in crowding out" as a second, linked mark (a 1+1 pair built from one chain); and, in a genuinely different bullet, "Opportunity costs of government spending" marked explicitly "(1+1)" in the mark scheme's own notation on that single bullet — the mark scheme flagging its own example as a 1+1 pair even though the header above it didn't spell the rule out generally. A third bullet splits differently again: "Time lags/implementation lags may mean measures take a long time," one mark, with a second, genuinely separate consequence — that the measures might not be affordable for Greece long-term — as the second mark. Three bullets, three different internal splits, all valid, because the mark scheme is rewarding two marks' worth of evaluative content, however it is distributed, not a fixed shape.
What this means practically, and how it differs from Q7(e)/Section C. Because there is no level here, there is no "informed judgement" requirement (§3) and no penalty for stopping after one point instead of developing a chain across several — the ceiling is 2 marks either way. The self-mark checklist in forge-exam-technique-guide.md §3 already gets the practical instruction right ("write a genuine evaluation sentence, not a third analysis point relabelled"); what this section adds is why that instruction is correct: a second, shallower evaluative comment can score exactly as well as one fully chained argument on this specific tariff (1+1 = 2+0), which is not true anywhere else evaluation is credited on this paper.
3. The Evaluation ladder's own printed text — and the four words that separate Q7(e) from Section C
Both ladders (Q7(e)'s 6 marks; the Section C essay's 8 marks) are Level 1–3, no Level 4, exactly as forge-level-threshold-guide.md §3 and §5 already state. What hasn't been stated anywhere in this repo is that their printed descriptor text is not identical — and the one place it differs is precisely the "informed judgement" language forge-exam-technique-guide.md §4 already cites as an examiner-report finding.
Level 1 and Level 2, word for word identical between the two ladders, confirmed in all three series checked:
- Level 1: "Identification of generic evaluative comments," with "no evidence of a logical chain of reasoning" and no reference to context.
- Level 2: evidence of "evaluation of alternative approaches," "supported by a partially-developed chain of reasoning."
Level 3 — where they diverge. Q7(e)'s own Level 3 descriptor (5–6/6 marks, confirmed identically in October 2021, January 2024 and October 2024) opens with "evaluation recognises different viewpoints and/or is critical of the evidence," then separately requires appropriate reference to context and a logical chain of reasoning — three short clauses, no stated verdict among them. The Section C essay's Level 3 descriptor (7–8/8 marks, same three series) opens with the identical clause, but does not stop there: it continues, mid-sentence, "leading to an informed judgement." Four words, present in every Section C Evaluation Level 3 descriptor checked and absent from every Q7(e) Evaluation Level 3 descriptor checked — the same three series, the same comparison, run twice.
Why this matters, and what it resolves. forge-exam-technique-guide.md §4 quotes the January 2021 examiner report's own instruction that Level 3 evaluation "in Section C" specifically requires "an informed judgement." Read next to what this section just found, that instruction's own specificity — naming Section C, not the paper generally — is not a stylistic accident. It is the examiner naming, in prose, an actual textual difference printed in the mark scheme itself: the phrase the examiner is enforcing literally is not part of Q7(e)'s own Level 3 descriptor. This also directly answers the honest gap forge-level-threshold-guide.md §5 states and declines to resolve — whether the "informed judgement needed for Level 3" finding, confirmed there only for Section C, also governs Q7(e): checked directly against the primary text across three series, it does not — Q7(e)'s ceiling is a logical chain of reasoning that recognises or is critical of competing evidence; a stated verdict is not printed anywhere in its own descriptor. Practically, this means the highest Q7(e) can ask of an evaluative point is real, developed conditionality — not the closing "on balance, X, because Y" sentence a Section C essay's top Evaluation mark specifically requires.
4. The same split on the KAA side — Q7(e)'s ceiling borrows Section C's top rung, and skips its own middle one
This sits one level below Evaluation in the mark scheme's own AO structure (KAA, not AO4), but its descriptor table is printed directly alongside the Evaluation tables in §3, for the same questions, and the same comparison is worth stating once, briefly, because it resolves a second honest gap the sibling documents already flagged rather than guessed at.
Section C's KAA ladder (12 marks, Level 1–4) has a Level 2 (4–6/12) descriptor built specifically around "only two-stage chains of reasoning" — the exact phrase behind the "two-stage chains cap Level 2" finding both sibling documents already use, sourced there from examiner-report prose. Checked directly against Q7(e)'s own KAA descriptor table (8 marks, Level 1–3) across the same three series: that phrase, and the two-stage framing it describes, does not appear anywhere in Q7(e)'s own table. Q7(e)'s Level 2 (4–6/8) instead reads almost identically to Section C's Level 3 — "chains of reasoning... may not be developed fully or some stages are omitted," with no "two-stage" ceiling named — and Q7(e)'s own Level 3 (7–8/8, its top) reads almost identically to Section C's Level 4 (its top): "examples that are fully integrated to address the broad elements of the question," "logical and multi-stage chains of reasoning" (near-identical in January 2024 and October 2024; October 2021's earlier wording drops "fully" before "integrated" — a small, real drift across the archive worth flagging rather than smoothing over).
This directly resolves the specific gap forge-level-threshold-guide.md §5 states rather than assumes: "Whether the identical mechanism [the two-stage-chain ceiling] governs Q7(e)'s own Level 1/2 boundary was not independently checked... Treat that transfer as a reasonable inference, not a confirmed finding." Checked directly against the primary text: it is not a confirmed finding because the mechanism the inference was extending does not exist in Q7(e)'s own printed descriptor at all. Q7(e)'s three-level ladder is best read as Section C's Level 1, then Section C's Level 3, then Section C's Level 4, with Section C's own "two-stage chain" Level 2 rung simply not printed anywhere in Q7(e)'s three-level version — not a compressed copy of the same four rungs.
5. What "generic evaluative comments" (Level 1) is actually defined as — checked directly, not assumed from a different paper
The reference document for this file's own shape and rigour, content/veridian/technique-hub/wbs12/forge-evaluations-guide.md, exists because a raw draft for a different paper fabricated a "trigger phrase" rule — specific vocabulary supposedly required for a top Evaluation level — and that rule was checked directly against every WBS12 examiner report and found nowhere. No equivalent WEC14 draft claiming a trigger-phrase rule was found anywhere in ~/Downloads/FINSTO/ for this pass (the folder holds WBS12/WEC12/WEC11 "definitive guide to evaluations" drafts, but none named WEC14). Rather than skip the check because no specific claim needed debunking, this document ran the same check WBS12's own file ran, directly against WEC14's primary archive: grepped every mark scheme and examiner report text file converted for this pass for "best fit," "best-fit," "best match," "trigger phrase," and "trigger word." None of these appears anywhere in the WEC14 archive — neither the "best-fit, whole-response" marking language WBS12's own mark scheme uses, nor any trigger-phrase rule.
What the mark scheme actually prints instead, as the full and only definition of Level 1 Evaluation across both ladders (§3): "Identification of generic evaluative comments," with no supporting evidence or reference to context, and no logical chain of reasoning. That is a description of what an answer does — states an unsupported evaluative claim and stops — not a checklist of words it lacks. Nowhere in the descriptor, on either ladder, in any series checked, does a specific word, phrase, or sentence-opener appear as a requirement. An answer reaches Level 2 by supporting a claim with real, contextual evidence and a partially-developed chain, not by writing "provided that" or "only if" — the identical, evidence-based conclusion wbs12/forge-evaluations-guide.md §2 reaches for its own paper, reached here independently for WEC14's own primary text rather than assumed to transfer.
6. General marking guidance — how an evaluative answer is treated before any level is assigned
Printed once, near-identically, at the front of every mark scheme in the archive, before Section A's answers begin — genuinely new to this document, not cited in either sibling file:
"Mark schemes should be applied positively" and candidates "rewarded for what they have shown," not penalised for what is missing.
"There is no ceiling on achievement" — every mark on the scheme is meant to be used.
"All the marks on the mark scheme are designed to be awarded."
This is genuinely relevant to Evaluation specifically, not just a generic disclaimer, because of how the Section B 8-mark part's "(2+0 or 1+1)" mechanic (§2) and the three Level 1–3 ladders (§3) both work: an examiner is instructed to credit the evaluative content that is actually on the page — one strong point, or two thinner ones, or a partial chain that stalls at Level 2 — rather than mark down for the argument a stronger answer would have made instead. It is the same principle forge-exam-technique-guide.md §4's "no ceiling on achievement" instinct already assumes in its exam-day advice (§5 there: "an unfinished essay with a judgement present can still reach Level 3"), stated here in its own original, general, paper-wide form rather than only as a Section-C-specific inference.
7. Sources and honest gaps
Primary sources used, and how. research/veridian/WEC14-verified-facts.md for the paper structure and AO weighting in §1. For §§2–6, the primary mark-scheme PDFs in the local archive (~/Downloads/home 4/ubuntu/economics_ial_u4_complete/) for October 2021, January 2024 and October 2024 were re-fetched and re-run through pdftotext -layout specifically for this document, and the generic Level descriptor tables and general marking guidance page were transcribed and diffed by hand across all three. The January 2022 mark scheme was additionally checked for the Section B 8-mark part's "(2+0 or 1+1)" instruction line (§2), confirming it beyond the two series quoted directly. Every mark scheme and examiner report text file in that same local archive was grepped directly for the trigger-phrase/best-fit terms in §5. forge-level-threshold-guide.md and forge-exam-technique-guide.md were read in full before writing, specifically to confirm every claim above is additive to them rather than a restatement — the two honest gaps each of those documents states about Q7(e) (§5 of the threshold guide, twice) are the two gaps §§3–4 above close directly against the primary text.
Genuine gaps, stated rather than smoothed over:
- The three-series check (October 2021, January 2024, October 2024) is a real spread across the archive but is not all 13 series. Every claim above that says "confirmed across all three checked" should be read exactly that literally — not as "confirmed throughout the archive." A fourth or fifth series could in principle show a wording revision this pass didn't catch (the KAA Level 3 "integrated"/"fully integrated" drift between October 2021 and the two later series, noted in §4, is a real instance of exactly this kind of small change happening at some unidentified point between series).
- No mark scheme for June 2021 or January 2025 was available to check in this archive (June 2021 has only a provisional mark scheme; January 2025's mark scheme was not present in the local archive at all, only its question paper and examiner report) — neither is covered by any claim above.
- §2's "(2+0 or 1+1)" split is checked directly in four series (October 2021, January 2022, January 2024, October 2024), but the explicit parenthetical is only printed in two of those four (January 2022; October 2024, order swapped) — in the other two it is inferred from how the indicative content itself is marked, not from a stated rule, and that distinction is kept explicit in §2 rather than smoothed into "confirmed in every series checked." Not checked against every series where the 8-mark part appears under a different question letter.
- No likelihood percentage, grade boundary, or "every series without exception" claim is made anywhere in this document that was not checked series-by-series as stated.