economiststyle

An Economist-inspired Matplotlib style: a style sheet, the colour palette behind it, the chart furniture rcParams cannot express, and the notebook that derived the whole thing.

Install

uv add economiststyle          # or: pip install economiststyle

From a checkout:

uv sync

Use

Importing the package registers its style sheets and named colours with Matplotlib, so the style resolves by bare name from any working directory:

import economiststyle
import matplotlib.pyplot as plt

plt.style.use("theeconomist")        # or: economiststyle.use()

Add the title block — the red corner tag, bold headline and source line that a style sheet has no vocabulary for:

from economiststyle import econ_title

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(x, y)

econ_title(
    fig,
    "Corruption and human development",
    source="Sources: Transparency International; UN Human Development Report",
)

The grayscale variant

For print and photocopy, layered over the base style so the two can't drift:

economiststyle.use("gray")

The palette

economiststyle.CYCLE      # the ten cycle colours, in order
economiststyle.RAMPS      # tonal ramps by hue: RAMPS["blue"][0]
economiststyle.BRAND      # brand colours by name: BRAND["economist"]

Every ramp and brand colour is registered as a Matplotlib named colour, so it works anywhere a colour string does:

ax.plot(x, y, color="blue:0")
ax.axhline(0, color="economist:red")

The demo dataset

The corruption-vs-development data the style was built against travels with the package:

df = economiststyle.load_cpi_hdi()

The Altair theme

An optional Vega-Lite theme, kept out of the base install and out of economiststyle's own import-time registration since Altair is a separate dependency:

uv add "economiststyle[altair]"    # or: pip install "economiststyle[altair]"

import economiststyle.altair as economist_altair
import altair as alt

economist_altair.enable()          # or: economist_altair.enable("gray")

alt.Chart(df).mark_line().encode(x="date:T", y="value:Q", color="country:N")

It reproduces the colour cycle, the y-only grid, the spineless axes, and the bold left-aligned title with a grey subtitle — everything a Vega-Lite config maps onto cleanly. It does not attempt the red corner tag: Vega-Lite has no drawing surface outside a view's own scales, so there is no equivalent of fig.add_artist to place it with.

A note on the font

The Economist sets its charts in a face that isn't freely distributable, and the nearest common substitute — Gill Sans MT — ships with Microsoft Office rather than with any OS. The style sheet therefore names a stack:

font.family     : sans-serif
font.sans-serif : Gill Sans MT, Gill Sans, Lato, Fira Sans, ..., DejaVu Sans

Gill Sans MT is used when present and the stack degrades gracefully when it isn't. This matters more than it looks: setting font.family to a face name directly, as earlier versions of this style did, makes Matplotlib skip the fallback list entirely, so a machine without that font silently renders in DejaVu Sans. For pixel-identical output everywhere, install one of the free faces in the stack.

Development

uv sync
uv run pytest                            # 41 tests
uv run python examples/make_gallery.py   # regenerate gallery/

CI runs the tests on Linux, macOS and Windows across Python 3.9–3.13, and renders the gallery on every push. The Linux runners have no Gill Sans MT, which is deliberate: it means the font fallback is exercised on every run rather than only on a machine that happens to have Office installed.

The notebook

MatplotlibLearning.ipynb is the original derivation, kept as a narrative: defining the palette, registering it with Matplotlib, building the rcParams up piece by piece, and finally loading the packaged sheet. It now imports from economiststyle rather than duplicating its logic, and CI executes it end to end so it cannot rot.

Licence

MIT — see LICENSE.

This is an independent homage, not affiliated with or endorsed by The Economist. The palette was eyeballed from published charts for the purpose of learning Matplotlib; no proprietary asset is redistributed here.