Census data is a treasure trove of information about American demographics. Journalists often mine its data to write about emerging social trends. But the Census is technical, and sifting through hundreds of tables with . Where do you even start?
We spoke to Joe Germuska, chief nerd at Knight Lab and the project lead for Census Reporter, a user-friendly website that lets people search Census data. It was built with journalists in mind.
Hopefully, some of the resources and concepts Germuska discusses will be useful as you do your reporting about population trends.
The types of Census data
The Census has dozens of projects, Germuska says, but the ones journalists use are the decennial census (the population count that comes out every ten years), the American Community Survey and the Current Population Survey. Each of these data sets has pros and cons.
The decennial census provides “fine-grained accuracy” in that it actually attempts to count every person in the U.S. But the limitation is that the questions asked are pretty basic, such as about age, sex, race, family relationships and housing characteristics.
The American Community Survey is what Germuska describes as the journalistic “sweet spot” because ACS data is released annually, and the survey includes information that is journalistically interesting, such as economic status and educational attainment. Plus the data drills down to the local level so journalists can observe local trends.
The Current Population Survey comes out every month to measure employment information, but it doesn’t give you much information about local trends.
Given the daunting task of sifting through all this information, Germuska and his team developed the Census Reporter. It is a website that presents the most recent ACS data in a user-friendly and searchable way.
Census Reporter lets you search for information based on a geographical subdivision (such as city, state or census tract), and lays out the broad demographic information about that place with some context.
Census Reporter also lets you compare specific datasets (or tables as they are known in the Census) between multiple geographic areas.
The Census Reporter is really good at providing an entry into the ACS, and comparing multiple areas at the same point in time.
But to track change over time, you’ll have to delve into the Census Bureau’s ACS search tool, American Factfinder.
Things to keep in mind when working with Census data
I asked Germuska about common errors he’s seen and confusion I’ve had and seen other reporters have. Here are some topics we discussed:
1. Census tract vs. neighborhood
Census tracts are the subdivided units into which cities and counties are broken down. A tract encompasses a population of 1,200 – 8,000 people. Census tracts are not the same as neighborhoods. It is often difficult to map a neighborhood onto a corresponding group of Census tracts, since Census tracts get redrawn from time to time, and neighborhoods aren’t official designations whose boundaries are fluid.
Another complication arises when you may have to split Census tracts in order to fit them into the boundary of a neighborhood. This raises another set of questions about how to most accurately split the tract.
Germuska adds some cities — such as Seattle and Chicago — have made a list of census tracts that correspond with particular neighborhoods, but this is still rare.
Don’t try to force the data to be more exact that it is.
“One of the things we really advocate that journalists get comfortable with in writing from Census data, is to not write as though there’s more precision than is appropriate,” Germuska said.
The ACS is a survey, which means there is uncertainty built into the data — it’s an estimate. You want your writing about this information to reflect that, Germuska says. So instead of using exact numbers like 33,617, use more general terms to describe the spirit of the data, such as “more than half” or “about a third.”
The Census is highly technical, and you should familiarize yourself with the terms used, the questions that people are asked and the options people have to give as answers. This information is called the documentation, and it changes.
Knowing the documentation is important for two reasons: you avoid making presumptions and inferences about the data before you actually know what it’s measuring, and if you’re tracking change over time, it’s important to note whether — and how — the questions and answer options have changed, too.
The Census Bureau has a useful handbook for journalists that explains these concepts.
4. Treat Census data like you’d treat any source
Germuska’s rule of thumb: “The more fine-grained the question, the higher the likelihood of significant statistical error.”
That means if you have a specific question, looking at a small group of people, in a small geographical unit, the survey result could be wrong. Germuska gave the example of looking for people who identify as multiracial in Ohio. He said one tiny segment in Toledo had the highest number of multiracial people one year, and the following year, it had zero multiracial people. Germuska explains with such a small population (people who identify as multiracial) in such a small geographical unit, the estimates can vary widely, making them almost nonsensical.
“You need to bring your local knowledge,” Germuska said. “You don’t ever trust a single source. Data is a single source, and it can tell you some things, and you have to check it out by checking human sources or possibly other data. You have to run it by your sniff test wherever possible.”
This article originally appeared on Navigator, GroundTruth’s newsletter for early-career journalists. You can subscribe to Navigator here.