In 1910, 49 percent of New Yorkers lived in Manhattan; by 2020, 19 percent did.
The Census Bureau has counted New York City tract by tract since 1910: 4.8 million people that year, 8.8 million in 2020. This map spreads each tract’s count over its land, smooths it and raises the ground where more people live, so the higher and brighter the ground, the more people per square mile. In 1910 the most crowded square mile, on the Lower East Side, held about 380,000 people. In 2020 the most crowded, on the Upper East Side, held about 140,000.
Data: IPUMS NHGIS, University of Minnesota · How it was made
1910
Between census years the map blends the counts on either side.
People per square mile, averaged over about an eighth of a mile around each point. Thin dark lines mark 25,000, 50,000, 100,000 and 200,000.
Click or tap the map to follow any spot through all 12 censuses.
Each census’s tract counts come from IPUMS NHGIS at the University of Minnesota, with tract boundaries drawn for that census. Each tract’s people are spread evenly over the land inside it, on a grid of squares a twentieth of a mile on a side, using today’s shoreline from the Department of City Planning. The grid is then smoothed over about an eighth of a mile. In every census the tract counts add up to within 0.6 percent of the official city total. New York kept its 1930 tracts in 1940, so the 1940 counts are drawn on 1930 boundaries. The “most crowded square mile” is the square-mile window on the grid holding the most people before smoothing; since people are spread evenly within each tract, it is an estimate. Full methodology, sources and checks.
The idea comes from an animated density map of France, “Là où vit la France,” shared by Yan Holtz of the R Graph Gallery.
And please read this AI caution. The idea and the choice of sources are the author’s (a human). The code and the data processing were done by Anthropic’s Claude Code under close direction. Quality control on this project: the tract counts for every census were summed and checked against the official city total; borough totals for 1970 through 2010 were checked against the Department of City Planning’s community-district counts and land within 0.8 percent; the 1940 tract codes were matched to 1930 boundaries and each tract’s 1940 count was compared with its 1930 count to catch mismatches. The risks of using AI remain, and red flags are welcome.