Before choosing a form, name what is actually spatial about the data. Not all data that has a place-name dimension is genuinely spatial. Employees by office location is a ranked comparison that happens to have city names. A bar chart answers that question better than a map, because comparison is the relationship , not spatial proximity, not geographic clustering, not the question "where?" Geographic visualization earns its complexity when the location itself is the answer.

Choropleth. Geographic regions shaded by a value. The mark is the region polygon; the channel is color luminance. The choropleth is the default spatial form in data journalism, public health, and policy work. It works when the data is a rate (a ratio per population or per unit), the regions are roughly comparable in area, and the reader's question is geographic pattern rather than precise value. Dot density map. Dots placed within regions, one dot per N units.

The choropleth is the most common spatial form and the most commonly misused. Its failure mode is structural, not just a matter of poor implementation. The problem: the reader's eye processes the area of each region before it processes the color. Large regions attract attention. Small regions recede. This happens regardless of what color is encoded. A choropleth of "quality of life" shading Luxembourg alongside France will always make France visually dominant, even if Luxembourg has the highest quality of life by a factor of two.

The ratio-vs-absolute rule is Cairo's "compared with what?" check applied geographically. A choropleth of absolute counts answers a different question than the designer usually intends. "Number of COVID cases by state" mostly shows "where do people live?" , because the number of COVID cases in a large state is mostly a function of its population, not of anything specific about how the state managed the pandemic. The chart's signal is dominated by the confound.

John Snow's 1854 cholera map of the Soho neighborhood in London is the canonical example of geographic visualization enabling causal reasoning. It is also the best argument for the dot density form over the choropleth. Snow's question was: where in Soho are the cholera deaths concentrated? He had a list of addresses , every household that reported a death during the 1854 outbreak. He placed a dot at each address on a detailed street map.

Every geographic visualization involves a projection , a mathematical transformation from the sphere of the Earth to a flat surface. All projections distort something: area, shape, distance, or angle. There is no perfect projection. The choice of projection matters for geographic charts because different distortions interact differently with different encodings.

The bubble map beats the choropleth when the data is an absolute value and the visual priority is magnitude rather than geographic pattern. A choropleth of absolute refugee counts shades large countries darkly because they contain more people , Syria, Afghanistan, and South Sudan generate large refugee flows. But the visual result favors the areas (Russia, Canada, Australia) that have few refugees but large land masses. The map does not show what the data says.

Geographic charts have form-specific specifications that belong in every prompt. For choropleths: the projection (d3.geoEqualEarth() or d3.geoAlbersUsa()); the encoding explicitly stated as rate or ratio ("encoding is uninsured rate, not absolute uninsured count"); the color scale type (sequential for unipolar data; diverging for data with a meaningful zero); the bin count (5, 7) and bin method (quantile or equal-interval); the legend.

The test for this chapter: given any geographic dataset, answer three questions before touching Claude Code. First: is geography the right family, or is a ranked comparison chart the honest answer? If the question is "which state has the highest uninsured rate?", a bar chart answers it faster and more accurately than a choropleth. If the question is "what is the geographic pattern of uninsured rates?", the choropleth answers it. The word "pattern" is the signal.

Exercise 14.1 , Form selection. For each of the following, name the right geographic form (choropleth, dot density, bubble map, or connection/flow map) and justify the choice in one sentence: Per-capita income by US county (rate, roughly comparable-area units). Total cancer cases by US state (absolute count, large region area variation). Locations of every reported opioid overdose in a single city over one year. Volume of container shipping between the top 20 port pairs globally.

Choropleth. Region-shaded map; color luminance encodes value. Invented by Dupin (1826) for French illiteracy by department. Requires rate encoding and comparable-area regions. Area-size distortion. Choropleths are perceptually dominated by the size of geographic regions regardless of the data value encoded. Structural to the form; mitigated by equal-area projections and comparable-area regions; eliminated by switching to bubble maps or dot density maps.

Choropleth. Region-shaded map; color luminance encodes value. Invented by Dupin (1826) for French illiteracy by department. Requires rate encoding and comparable-area regions. Area-size distortion. Choropleths are perceptually dominated by the size of geographic regions regardless of the data value encoded. Structural to the form; mitigated by equal-area projections and comparable-area regions; eliminated by switching to bubble maps or dot density maps.