By age, size, income and race.
For reasons of both practicality and privacy, the Census Bureau releases no dataset that contains information on every individual household in America. The demographic data we more often see from sources like the American Community Survey is based on a sample of the U.S. population, and it extrapolates from there to tell us about communities down to the county, census tract or block group level (this smallest geographic unit usually contains several hundred to a few thousand people).
That data can tell us, for instance, that 100 of the households in a given block group make less than $20,000 a year. It cannot tell us that, of those 100 households, 87 of them contain three people, or 34 have black heads of household, or six are headed by people older than 65. Want that dataset?
"It doesn't exist. You can't get it," says Bill Wheaton, the director of the Geospatial Science and Technology Program at the research institute RTI International. "But we can try to represent, or synthesize that list."
Which is what RTI has done here with its newly launched Synthetic Population Viewer.
"We want to actually represent every household as a dot on a map, and as a record in a database," Wheaton says. "Instead of having a record for the block group that says how many different kinds of households there are in there, we have a record for every household."
RTI's mapping tool, with an attached public dataset, is a "synthetic" representation of 112,596,000 individual households in the U.S. that was developed to match the household characteristics at the census block group level from the 2005-2009 American Community Survey. The distribution of the dots draws on density data from the LandScan database (so you don't see people living, for instance, in the middle of Central Park). The model is also populated with information from the Census Bureau's Public Use Microdata Sample.
In effect, each dot isn't an actual household, but it represents one as closely as possible. The project is conceptually similar to this Racial Dot Map (which plotted equally distributed individuals as single dots within Census block groups). The RTI project, though, also pulls in information on household size, householder age, and income. And you can view all four maps side-by-side in the tool, as shown up top.
That added data illustrates, for example, the concentration of young, solo and low-income people (portrayed below in dark blue for 15-34 year-old heads of household) around the University of Pittsburgh and Carnegie Mellon in Pittsburgh:
Or here you can see older populations (hot pink) spread out along the golf course retirement communities of Central Florida:
The project was funded as part of a Models of Infectious Disease Agent Study grant from the National Institute of General Medical Sciences, with the goal that this type of synthetic population data could support modeling of the spread of diseases. But it could also be used for complex simulations studying all kinds of elements of society, from housing to transportation to land use.
Or, of course, you could just gawk at it for kicks. The full tool is embedded below, or open the map on a full screen from here.