pandas_datareader is maintained again (0.11.0/0.11.1, June 2026) — reconsider it as the builder-side entry point for dynamic data

[mmcky]

What changed

pandas-datareader had been effectively unmaintained since 0.10.0 in July 2021. That is no longer true — two releases landed in June 2026:

Version	Date	What
0.11.0	2026-06-23	pandas 3 compatibility; new pandas_datareader.macro API
0.11.1	2026-06-24	fix to OECDReader for large mostly-empty frames

Two parts of 0.11.0 matter here.

pandas 3 compatibility. It vendored deprecate_kwarg to drop the dependency on the private pandas.util._decorators module — the exact failure recorded in QuantEcon/lecture-python.myst#924, which blocked the anaconda 2026.06 upgrade. That incompatibility was the concrete reason the package was on its way out.

A unified macro API. 0.11.0 adds pandas_datareader.macro, described in the release notes as providing "unified access to FRED, Eurostat, and OECD macro datasets via read_macro, search_macro_datasets, and describe_*".

Why this is a question for this repo

The second point is a candidate answer to something PLAN Phase 5 and Phase 8 P4 already need: a single consistent fetch layer for dynamic macro series, living in a builder rather than in lecture notebooks.

Phase 8 P4 is UNRATE — "consumed today by 4 lectures across 3 repos via 2 access methods" — and it is the pilot that establishes the dynamic template plus the documented live-call ↔ snapshot switch. pandas_datareader.macro covers FRED, Eurostat and OECD behind one interface, which would make a builder's fetch stage uniform across sources instead of one client per source.

Note the direction this implies: pandas_datareader moves out of lecture runtime and into scripts/. As a builder dependency it runs on a cadence, its failures surface in the Phase 5 sources-alive canary instead of in seven lecture repos' CI — which is exactly the relocation Phase 5 sets out to achieve — and lectures read a stable URL.

Current usage, for the record

Ten published lectures across four repos reference it. Six genuinely use it; four carry an install line they never import — see the note below the table.

Repo	Lecture	Reference	Used?
lecture-python.myst	unemployment_linear	!pip install pandas_datareader; from pandas_datareader import data as web	yes
lecture-python.myst	unemployment_shocks	combined !pip install numpyro jax pandas_datareader arviz; same import	yes
lecture-python.myst	phillips_lost_conquest	!pip install pandas_datareader; same import	yes
lecture-python.myst	phillips_two_stories	!pip install pandas_datareader; same import	yes
lecture-python-intro	business_cycle	!pip install pandas-datareader; import pandas_datareader.data as web, alongside wbgapi	yes
lecture-python-intro	networks	!pip install pandas-datareader only	no
lecture-python-intro	input_output	!pip install pandas-datareader only	no
lecture-python-advanced.myst	doubts_or_variability	!pip install pandas-datareader only	no
lecture-datascience.myst	tools/classification	import pandas_datareader.data as web (pip dep in environment.yml)	yes
lecture-datascience.myst	tools/visualization_rules	from pandas_datareader import DataReader (same)	yes

Four of the ten are vestigial installs — networks, input_output and doubts_or_variability install the package and never import it. doubts_or_variability is the instructive one: a comment at line 2013 records that it deliberately avoids pandas-datareader by reading FRED's CSV endpoint directly, so the migration was done and only the install line was left behind. These four can be deleted with no code change, independent of whatever is decided below.

Of the six real usages, five are FRED readers. business_cycle is the one already half-migrated to wbgapi, and is also the lecture behind this repo's single existing dynamic snapshot (business_cycle_data.csv).

lecture-python.myst and lecture-python-intro are both on anaconda=2026.07 (pandas 3.x), and all four lecture-python.myst pages publish clean with no tracebacks — consistent with 0.11.0 working under pandas 3. lecture-datascience.myst is still on anaconda=2024.06 and carries pandas-datareader as an explicit pip dependency in environment.yml.

Questions

• Does pandas_datareader.macro become the standard fetch layer for dynamic snapshots, or do we keep per-source clients (wbgapi, fredapi) and treat this as one option among several?
• Should P4's UNRATE builder be written against it, making the pilot the test of this decision?
• Does the live-call ↔ snapshot switch keep a live pandas_datareader path available inside lectures at all, or is snapshot-only the target?

Related

• [deprecate] update package or remove the use of pandas_datareader across all QuantEcon projects meta#160 — the standing "deprecate or remove pandas_datareader" issue. Its premise, that the package "last received a release update in 2021 and the current maintainer is not inclined to maintain the readers", is now out of date; it should be re-read against this before anyone acts on it
• Accessing Data using Python meta#89 — closed as not planned; this carries forward the data-access question it raised
• pandas 3.0 (anaconda 2026.06) breaks hansen_singleton lectures via pandas-datareader lecture-python.myst#924 — the pandas 3 breakage that 0.11.0 fixes
• MAINT: fix FutureWarning and remove pandas_datareader lecture-python-programming#341 — removed pandas_datareader in favour of wbgapi (merged 2025-03-12), on the "not receiving updates" rationale

[mmcky]

Correction to the inventory above. The table originally said seven lectures across three repos. It is ten across four — the first pass searched for pandas_datareader (underscore) and missed the lectures that write it hyphenated in their !pip install line. The body has been updated; the three additions are networks and input_output in lecture-python-intro, and doubts_or_variability in lecture-python-advanced.myst, a repo previously reported as clean.

The correction sharpens rather than weakens the case. Four of the ten never import the package at all — they only install it. doubts_or_variability is the clearest case: a comment in that lecture records that it deliberately avoids pandas-datareader in favour of FRED's CSV endpoint, so the migration already happened there and only the install line survives. Those four lines can be removed with no code change and no dependency on the questions below.