Methodology · USA Data Map

Q: How are color scales chosen?

Sequential maps use 7-stop quantize bins so data ranges divide evenly across the color ramp. Diverging maps anchor at zero (or the U.S. average) so values above and below are visually distinct. Where the underlying distribution has a long tail or single outlier (Hawaii's RPP, for instance), the bins are recomputed to avoid letting one extreme distort the rest of the map.

Q: How are margins of error handled?

ACS estimates carry margins of error (MOEs) that grow rapidly for low-population counties. We suppress counties whose coefficient of variation exceeds 25% on any value-driven map and disclose the suppression count on the article page. We never silently smooth or impute MOE-flagged values.

Q: What renders the maps?

A custom Node CLI built on D3 and resvg-js reads a per-topic configuration, joins normalized data to the base geometry, and outputs an SVG plus PNG variants at three sizes. Every article links to its source SVG and JSON data on the page.

Where the data comes from

Every map is sourced from a publicly accessible federal or state dataset. The most common are the U.S. Census Bureau's American Community Survey (ACS), Population Estimates Program (PEP), and Business Trends and Outlook Survey (BTOS); the Bureau of Economic Analysis (BEA) Regional Price Parities; the IRS Statistics of Income county-to-county migration files; the BLS Quarterly Census of Employment and Wages (QCEW); the EIA Form 861 utility-level electricity series; and the USGS Water Use of the United States. Specific table IDs and data vintages are listed at the bottom of each article.

What projection the maps use

Choropleths use the U.S. Census Bureau's TIGER/Line shapefiles via the us-atlas TopoJSON pre-quantization (10m resolution). All maps use the Albers USA composite projection, which insets Alaska and Hawaii at a comparable visual scale rather than projecting them at their true geographic size.

How color and scale are chosen

Sequential maps use 7-stop quantize bins so that the data ranges divide evenly across the color ramp. Diverging maps anchor at zero (or the U.S. average) so values above and below are visually distinct. Where the underlying distribution has a long tail or single outlier (Hawaii's RPP, for instance), the bins are recomputed to avoid letting one extreme distort the rest of the map.

How margins of error are handled

ACS estimates are accompanied by margins of error (MOEs) that grow rapidly for low-population counties. We suppress counties whose coefficient of variation exceeds 25% on any value-driven map and disclose the suppression count on the article page. We never silently “smooth” or impute MOE-flagged values.

How the maps are rendered

The map rendering layer is a custom Node CLI (D3 + resvg-js) that reads a per-topic configuration, joins the normalized data to the base geometry, and outputs an SVG plus PNG variants at three sizes. The pipeline is open for inspection. Every map links to its source SVG and JSON data on the article page.

Corrections

Email [email protected] with corrections. Significant corrections are noted directly on the article with a dated update line.

How the maps are built.