Coloring a map by raw count mostly draws a population map, not the pattern it claims to show

What it is

The choropleth's color channel has no normalization: filling with absolute counts (total population, total sales, case counts) makes the number large partly because "the place is big/populous," while the region's geographic area never enters the denominator. The result: large regions automatically look "serious"—they may simply be large or populous—while density and intensity, the genuinely comparable quantities, are masked. Mapping rates (population density, per-capita income, infection rate) strips area and base out of the comparison, so color reflects "intensity per unit."

Why it happens

The rate-first principle counters the area-total collinearity: absolute counts correlate strongly with region area/population (bigger regions tend to have more people and larger totals), so coloring by count largely repeats "how big the region is"—information the map already shows—rather than adding new information. Two layers of distortion follow: interpretation (readers turn "large total" into "high intensity"—a populous province reads as "severe problem" when it may just be populous) and comparison (between regions of different sizes the comparison loses a fair baseline—a region with 10x population naturally has more cases, regardless of containment quality). Rates (cases per square kilometer, income per capita) normalize the base, so color differences point to genuine structural differences. The canonical instance is US election mapping: filling counties by winning party is a binary rate (roughly fair), but filling by vote totals darkens populous counties into an intensity illusion; the same data rendered as "margin of victory" versus "population-weighted" paints entirely different pictures. When the business question genuinely needs absolute magnitude (total market size, total revenue), the more honest channel is a proportional-symbol map (symbol size = absolute value, position = geography)—absolutes use symbol area, and area bias no longer contaminates color.

Where it stops holding

Rate mapping has boundaries too: the small-denominator problem—rates in sparsely populated regions swing wildly (one case in a one-person community is a 100% infection rate)—so ratio noise amplifies on small bases, requiring significance filtering or region merging. The choice of rate itself carries a stance (per thousand vs per square kilometer answer different questions); decide "what is the unit" before mapping rather than casually computing one. When the absolute question is real ("where does the most revenue come from"), the right chart is not a choropleth but a symbol map or ranked bar chart—the chart type follows the question, not the question bends to the chart.

Applying it

• Default choropleths to rates or normalized indicators; present absolute counts as proportional-symbol maps instead.
• Check denominator stability: when small-base regions' rates are unstable, merge regions or flag low confidence.
• State the rate definition in the caption ("cases per 100,000 people"); maps with different rates are not comparable.
• Verification: place the count version and rate version side by side; if their "darkest regions" differ and the rate version matches business intuition better, the count version was misleading.

Related

• Same group: U12.02.1 Large-area regions gain disproportionate visual weight in choropleth maps · U12.02.3 Equal-interval, quantile, and natural-breaks classifications yield different map shapes · U12.02.4 The number of classes sets the grain of the visible spatial pattern · U12.02.5 Cartograms correct area bias at the cost of geographic recognizability
• Nearby: U12.02.1 Large-area regions gain disproportionate visual weight in choropleth maps · U8.06.1 Aggregation masks differences between subgroups
• Search terms: rate vs count · normalized choropleth · ecological correlation