ContentsTwo measures on one map
Two measures on one map
Of 2,192 tracts in New York City with both measures, 466 are in the lowest third on both income and education and 482 in the highest third on both, while the two mixed corners hold 69 (high income, low education) and 55 (low income, high education). Each axis is cut into thirds (ACS five-year estimates, 2019 to 2023), and the nine-colour key beside the map is the only way to read it.
Income against Education, by census tract, New York
2,192 of 2,324 tracts painted
51 missing Income only · 81 missing both
- Income
- $68k · $97k
- Education
- 26.5% · 43.0%
Every cell, and how full it is
| Education ↓ / Income → | Low | Middle | High |
|---|---|---|---|
| High | |||
| Middle | |||
| Low |
Tertiles guarantee equal thirds on each axis and nothing about the cells: two strongly correlated measures empty the off-diagonal corners, and a grid with two near-empty cells is a univariate map wearing a bivariate legend. Select a cell to find its areas on the map; select it again to clear.
Cuts fitted to
Cuts are fitted to each axis's own values, including areas the other axis is missing, so they describe a slightly larger set than the map paints. Ties stop the thirds being equal: 731 / 730 / 731 horizontally and 716 / 736 / 740 vertically.
ACS 2023 5-year · boundary vintage 2020 · tertile cuts recomputed per city, level and measure pair, so the colours are not comparable between two of these maps
Change city, level, axes or palette
City
Level
Horizontal axis
Vertical axis
Palette
Every other map on this site can be half-read. A reader who ignores the legend on a sequential ramp still knows that darker means more, and will usually be right. Here that is not available: nothing about a muddy blue-green announces which of the two axes is high, so the map cannot be read until the key has been read.
That is the reason it is worth building. This site’s standard is that a form has to be able to say what it hides; a bivariate choropleth cannot work at all unless it says what it encodes, so the key carries the map. It is the plain version of a problem every map here has.
What it hides is both marginals. A cell says an area is in the top third on each axis and nothing about where inside that third it sits, so two areas sharing a colour can be far apart on both measures. The third palette is deliberately bad, two adjacent greens that are barely separable, because that is what most published bivariate maps look like.
The key is a control as well as a picture. Selecting a cell, in the rotated diamond or in the table under the map, outlines its areas, so the map answers “where are the high-high places” as well as “what is this polygon”. Clicking a polygon selects its cell, so the two directions are the same selection.