Include business_cycle: what it reads, what data-lectures now serves, and the sequence

[mmcky]

business_cycle is one of the eight lectures this repo excludes ("WASM-unsupported", README; commented out of lectures/myst.yml). It was excluded because every one of its data reads is a live API call — wbgapi against the World Bank and pandas_datareader against FRED — and Pyodide cannot make them (no CORS on the API hosts, and wbgapi needs requests patched). That reason no longer has to hold: QuantEcon/data-lectures now serves cached snapshots over raw.githubusercontent.com, which is CORS-clean and already works in this build (long_run_growth, inflation_history, french_rev, heavy_tails, inequality all read it in-browser today), and it has a weekly refresh-as-PR workflow so a snapshot of a moving series stays current. This issue is the review of what including the lecture would take, and the proposal for doing it.

What the lecture reads today

Measured against lectures/business_cycle.md on 2026-09-01. Nine live reads across two sources; the ones a snapshot already covers are marked.

Line	Source	Series	Selection / window	Snapshot today
66, 83	World Bank	wb.series.info(q='GDP growth'), wb.series.metadata.get(...)	teaching the API, not data	n/a — see below
72, 95	World Bank	NY.GDP.MKTP.KD.ZG	USA ARG GBR GRC JPN	yes — business_cycle_data.csv (five economies, 1960→2025, refreshed annually)
485	World Bank	NY.GDP.MKTP.KD.ZG	CHN USA DEU BRA ARG GBR JPN MEX	partly — the snapshot has 5 of the 8
549	World Bank	SL.UEM.TOTL.NE.ZS	USA FRA GBR JPN	no
728	World Bank	FS.AST.PRVT.GD.ZS	GBR	no
313	FRED	M0892AUSM156SNBR	1929-01 → 1942-06	no
321	FRED	UNRATE	1948-01 → 2022-12	no
344, 607, 684	FRED	USREC	1929 / 1978 / 1919 → 2022-12	no
608, 626	FRED	UMCSENT, CPILFESL	1978-01 → 2022-12	no
686	FRED	INDPRO	1919-01 → 2022-12	no

JupyterLite's "Run all" stops at the first failing cell, so the lecture is runnable only when every row above is served from a snapshot. A partial repoint buys nothing for readers.

What data-lectures would add

Per the design agreed on the data-lectures side (one builder per source, composite files, a shared fetch library — QuantEcon/data-lectures#26):

• World Bank: business_cycle_data.csv extended from five economies to the eight-economy union (it has no consumers yet, so that is free today and a breaking change once this repo reads it), plus the unemployment and private-credit tables from the same builder.
• FRED: one composite monthly file for the lecture — UNRATE, USREC, UMCSENT, CPILFESL, INDPRO and the 1929–42 historical unemployment series on one date index, 1919 → latest — built by a builders/_fred.py helper against fredgraph.csv (no pandas_datareader in the builder, per the decision on data-lectures#26). FRED's UNRATE carries one structural null, 2025-10 (the shutdown month BLS never published), which the manifest will declare.
• Both class: dynamic-snapshot, refreshed by the weekly workflow; this repo would be listed in their manifests as a consumer with on_refresh: rebuild, so a merged refresh triggers a build here.

What this repo would change

• lectures/business_cycle.md: every wb.data.DataFrame(...) and web.DataReader(...) becomes a pd.read_csv('https://raw.githubusercontent.com/QuantEcon/data-lectures/main/lectures/<file>') with the same selection applied in pandas (.loc[[...]], a date slice). The %pip install wbgapi … pandas-datareader cell and both imports go. The API-teaching cells (wb.series.info, wb.series.metadata.get) cannot run here and are not the point of this edition — shown as non-executing code blocks with a one-line note that the intro edition runs them live. lecture-python-intro keeps every live call: the API is the lesson there, and that is a deliberate split, not drift.
• lectures/myst.yml: uncomment business_cycle; README: remove it from the unsupported list.
• One thing to settle while here: the README says "do not edit lectures directly in this repository — sync from intro's wasm branch", but that branch is 125 commits behind intro main and carries none of the data-lectures repoints, while the last ten PRs here (long_run_growth: read mpd2020.xlsx from data-lectures, not from intro's repo #52–short_path: write the graph with %%file, and move the cell so it can run #63) all edited lectures/ directly. The rule is dead and should say so, or the sync should be revived; either way this PR follows the practice.

Sequence

1. data-lectures: extend the World Bank snapshot and land the FRED composite (two PRs, builders + manifests, consumers: []).
2. This repo: the lecture edit, toc and README — a preview build on Netlify proves the in-browser run.
3. data-lectures: the flip — consumers recorded with on_refresh: rebuild, migration.yml → repointed, which is what re-arms the audit's drift check for these files.

Repoint rules apply as usual (data-lectures PLAN.md): this repo publishes on push to main, so no repoint/publish/delete split is needed, and nothing is deleted anyway — there is no committed copy to remove.

Part of the P4 dynamic-snapshot pilot (QuantEcon/meta#338; data-lectures PLAN.md Phase 8). See QuantEcon/data-lectures#109, #110 and #112 for the snapshot, the workflow and its first refresh.

[mmcky]

Stage 1 is in review: QuantEcon/data-lectures#114 lands every file the table above needs — business_cycle_data.csv grown to the nine economies the lecture plots, unemployment_rate_annual.csv, private_credit_to_gdp.csv, and us_business_cycle_monthly.csv (UNRATE, USREC, UMCSENT, CPILFESL, INDPRO and the 1929–42 historical unemployment series on one monthly grid from 1919-01). All four are dynamic-snapshot, refreshed by the weekly workflow, and read from raw.githubusercontent.com/QuantEcon/data-lectures/main/lectures/<file> — the CORS-clean form this build already uses. Filenames of the three new files are provisional pending the naming discussion at QuantEcon/data-lectures#113; they stay free to change until this repo reads them, which is why stage 2 should follow that decision or accept a possible rename in the flip.

One thing worth knowing for the lecture edit: the composite's UNRATE and CPILFESL are empty for 2025-10 (the shutdown month BLS never published), and the historical series is named by its FRED id M0892AUSM156SNBR as a column — the lecture's existing rename to UNRATE still applies.

[mmcky]

Stage 1 merged: QuantEcon/data-lectures#114 is on main and all four files serve from raw.githubusercontent.com/QuantEcon/data-lectures/main/lectures/ with the hashes their manifests record (checked 2026-09-01). Stage 2 — the lecture edit here — can start whenever the naming question at QuantEcon/data-lectures#113 is settled enough to commit to the three new filenames, or sooner if a possible rename in the flip is acceptable.

[mmcky]

The naming question is settled and the filenames are final (QuantEcon/data-lectures#113, landed in QuantEcon/data-lectures#124 on 2026-09-07). One of the four changed: business_cycle_data.csv is now gdp_growth_annual.csv. The other three stay as recorded above. So stage 2 — the lecture edit here — can start against these four URLs with no rename risk left:

what the lecture reads today	file to read instead
wb.data.DataFrame('NY.GDP.MKTP.KD.ZG', …) (GDP growth, both selections)	https://raw.githubusercontent.com/QuantEcon/data-lectures/main/lectures/gdp_growth_annual.csv
wb.data.DataFrame('SL.UEM.TOTL.NE.ZS', …) (unemployment, four economies)	…/lectures/unemployment_rate_annual.csv
wb.data.DataFrame('FS.AST.PRVT.GD.ZS', 'GBR') (private credit to GDP)	…/lectures/private_credit_to_gdp.csv
the six web.DataReader(…, 'fred') calls	…/lectures/us_business_cycle_monthly.csv (one composite; M0892AUSM156SNBR keeps the lecture's rename to UNRATE)

All four are raw.githubusercontent.com, the CORS-clean host this repo's data-url-guard check requires. Work plan for the edit: QuantEcon/data-lectures#118, block B.