Common mistake 1 with World map

Mapping a count rather than a rate. A map of total GDP is a map of country size with extra steps; per capita or per unit of area is almost always the question.

Common mistake 2 with World map

Ignoring the unmatched list. Aggregates like "World", "Euro area" and "OECD members" belong there and are harmless - a real country in that list means part of your sample is missing from the picture.

Common mistake 3 with World map

Choosing Mercator because it looks familiar. It inflates area with latitude, so it makes Scandinavia outweigh India. Robinson is the default for that reason.

Common mistake 4 with World map

Reading a difference between two adjacent classes as real. Two countries either side of a class edge may differ by almost nothing and be drawn in different colours.

EconStatsCraft › World map

World map

Shades every country by one of your columns. Country names, ISO codes and World Bank spellings all match, so a World Bank panel maps with nothing asked of you - and anything that does not match is named rather than quietly left off.

What it needs

Country column - Names, ISO codes (IND, IN, 356) or World Bank spellings.
What to shade by

The intuition

A choropleth colours each country by a value. It is the most persuasive chart in common use and one of the most misleading, for a reason that has nothing to do with the data: it colours AREA, and countries have wildly different areas and populations. Russia and Canada dominate any world map visually while holding a small share of the world's people, so the eye reads geography as magnitude.

When to use it

Cross-country data where where is part of the question: growth, income, inflation, tax take, life expectancy. It is a way of looking at data, not a test - use it before a cross-country regression to see whether the variation is regional, which is exactly the variation a country fixed effect would absorb.

Where the estimator comes from

Each region r is filled with f(v_r) for a value v_r and a colour scale f. The visual weight a region carries is proportional to its projected AREA, not to its population, its economy, or the reliability of its measurement. Projection compounds this: the Mercator projection inflates area by 1/\cos(\phi) at latitude \phi, so Greenland is drawn about fourteen times too large relative to the equator. An equal-area projection removes that specific distortion and leaves the deeper one - that area is the wrong variable - untouched.

What it assumes

That your place column identifies countries. Names, ISO codes (IND, IN, 356) and World Bank spellings all match; anything that does not is listed under the map rather than dropped silently.
One value per country. If your data is a panel, name the time column and say which period to show - the default is the latest, and the map says which one it drew.
That the classes are a choice. Quantiles put an equal count in each class; equal intervals cut the range evenly. The same data looks materially different under the two, which is why the rule and the class edges are printed.
The territories in your data match those on the boundary file in both name and existence. Countries split, merge and are renamed, and a historical series will not join cleanly to current boundaries.

Statistical properties

A choropleth must show a RATE, not a count. Population, GDP and case counts all map largest-country-wins; per capita or per unit area is what the colour should carry.
Use an equal-area projection. Mercator is a navigation projection and misrepresents relative size badly at high latitudes, which is exactly where the largest countries sit.
The classing method changes the map more than the data does. Equal intervals, quantiles and Jenks natural breaks on identical values produce visibly different maps, and the choice must be stated.
Small countries are invisible whatever their values. Singapore and Luxembourg cannot be seen on a world choropleth, so a table or a cartogram must accompany it if they matter.
Missing countries must be visibly distinct from low-valued ones. Grey for "no data" and the palest tint for "lowest value" are routinely confused by readers.
A cartogram resizes regions by the variable and fixes the area problem at the cost of recognisability. It is the honest alternative when the count, not the rate, is the point.

How to read the output

Read the legend first, then the map. The class edges are the only thing that tells you whether the darkest band is twice the lightest or a hundred times it.
Grey with hatching is no data, which is not the same as a low value. The count of matched places above the map tells you how much of your sample is actually shown.
Area is not importance. On any world map India and Nigeria look small relative to their populations and Russia and Canada look enormous relative to theirs - this is the fundamental limit of a choropleth, and it is why the largest and smallest units are also printed as a table underneath.
If the scale went diverging, the middle class straddles zero, so the hue tells you the sign and the depth tells you the size.

A worked reading

Map total COVID cases and you get a map of population: the United States, India and Brazil are dark because they are large, and the map has told you nothing you did not know. Map cases per 100,000 and the picture rearranges completely, with small European countries darkest. Same data, same chart type, opposite conclusions - and the only difference is dividing by a denominator. That division is the single most important decision in the whole chart, and it is made before the mapping software is opened.

Symptom → diagnosis → what to do

If you see this	It means	Do this
Large countries dominating	Area encodes nothing but area	Check you are mapping a rate, not a count
A map of totals	It will show population, whatever the variable	Divide by population
The pattern changing with the classing	Expected and under-reported	State the method; show a histogram of the values
Greenland enormous	Mercator	Use an equal-area projection
Missing countries looking like low values	The scale runs into the no-data colour	Make no-data visually distinct
Small states invisible	Inherent to the encoding	Add a ranked table

Common mistakes

Mapping a count rather than a rate. A map of total GDP is a map of country size with extra steps; per capita or per unit of area is almost always the question.
Ignoring the unmatched list. Aggregates like "World", "Euro area" and "OECD members" belong there and are harmless - a real country in that list means part of your sample is missing from the picture.
Choosing Mercator because it looks familiar. It inflates area with latitude, so it makes Scandinavia outweigh India. Robinson is the default for that reason.
Reading a difference between two adjacent classes as real. Two countries either side of a class edge may differ by almost nothing and be drawn in different colours.

Where to read this properly

Monmonier, How to Lie with Maps (2018), ch. 10.
Brewer (2016), Designing Better Maps, on classing and colour.
Gastner and Newman (2004), "Diffusion-Based Method for Producing Density-Equalizing Maps", PNAS 101(20) - cartograms.

All analyses

Data & variables

Each example is simulated from a known model, so you can check whether the estimator recovers the truth - and nothing here is anyone's private data.

Filter variables

Drag a variable into a slot on the right - or just click it.

Choose an example dataset above to begin.

Choose a transformation - it opens in the data editor, with a live preview.

Apply to

Multiply by

Columns ctrl-click for several

Group by ctrl-click for several

Within groups

Statistic

Bands

How many

Cut points

Keep rows where

Value

and

Fill with

Value

Also mark which rows were filled

Rules

Base period = 100

Which part

Percent at each tail

Threshold

Survey weight optional

Direction

Names column

Values column

Largest first

Delete them, not just count them

Read this column as

Tidy the text

New name optional

EconStatsCraft rigorous, plain-spoken econometrics

Econometrics that explains itself.

Regression, panel data, time series and the whole battery of tests. Every result is followed by a verdict you can defend.

Load a dataset to see it here.

World map

What it needs

The intuition

When to use it

Where the estimator comes from

What it assumes

Statistical properties

How to read the output

A worked reading

Symptom → diagnosis → what to do

Common mistakes

Where to read this properly

See also

Related Maps

All analyses

Charts

Maps

Explore

Regression

Time series - one series

Time series - several series

Panel data

Multivariate statistics

Marketing research

Causal

Finance & risk

Actuarial

Tests & breaks

Machine learning

Econometrics that explains itself.