Tip Sheets

Mining the Census for Health Reporting

Ronald Campbell, The Orange County Register
714/796-5030 | rcampbell@ocregister.com

See the accompanying presentation.

Since the metaphor of the day is mining, I will begin by referring you to some rich ores, or maybe tailings from past talks. Feel free to excavate them.

There are many buried treasures in IRE's wonderful Resource Center, including a couple that I left there:

In June 2011, Steve Doig somehow persuaded several reporters and Census people to visit Phoenix – yes, Phoenix, yes, in the summer – for a Census geek fest. You can find the whole thing online. I gave a semi-entertaining talk about FactFinder, complete with handouts, available at that website. A caution: FactFinder has changed a lot since I gave that talk. But the video will still give you a feel for navigating the site.

Among the thousands of topics that the Census covers, there are a few obvious gems for healthcare reporters and many that will require deeper digging. The obvious targets: health insurance and disabilities. Not only can you explore these topics geographically, but you can get insanely detailed information about each of them. How about the number of visually impaired people in your local workforce – employed vs. unemployed? Or look at part-timers and see how many buy health insurance on their own, through their employers or not at all.

Those, as I said, are just the obvious diamonds in the rough. Does your city inspect rental housing? How old is your housing stock, and how many people live in crowded housing? These are potential health problems, and the Census can tell you about them.

Is there a large immigrant community in your town? Do many of the immigrants speak a foreign language at home? If health institutions aren't equipped to talk with them in their native languages, your town is asking for health problems.

The Census Bureau conducts many surveys. But the one that is most valuable to healthcare journalists is the American Community Survey. It is conducted continuously nationwide and attempts to reach 3.5 million households each year. Because of its enormous sample size and rigorous quality control, the ACS can provide reliable statistics even for neighborhoods.

The key thing for journalists to remember is this: The smaller the area, the older the data.

Here's why: Sample sizes for a big state such as California are, well, big. Even obscure datasets tend to be robust with tiny margins of error. But by the time you get down to a census tract, or to hearing-impaired workers in Anaheim, the sample size is small and the margin of error is large. The best way to cope with small sample sizes – and the only way to cope with tracts – is to combine data for several years. That means a lot of your data will be old.

This raises another issue – the margin of error. Sometimes data that seems too good to be true really is too good to be true. The ACS helpfully lists the MOE at a 90 percent confidence interval for most statistics. So before you write a story about a number, look at the relevant MOE for that number. Is the number still newsworthy when you add or subtract the MOE to construct a range of values?

What if you want to compare the same statistics for different years? It gets a bit more complicated. You'll want to consult the ACS Accuracy of the Data documents, for example this one offering advice on comparing 2009-2011 3-year and 2007-2011 5-year estimates, to make sure you're doing it correctly.