The flood of Census data is over. For now.

U.S. Census Bureau servers, like journalists across the U.S., worked overtime for two months as the Summary File 3 became available, state by state, in August and September.

IRE's Census-L listserv buzzed with questions and stories. Journalists communicated by code, it seemed. Sample question on the Census-L: "Are there numbers for median household income by race for 1990? I'm looking in STF3 and not seeing such a table. Anyone?" Answer: "Table P82."

Incredibly, most Census-L subscribers understood that answer. It refers to a specific table released after the 1990 Census.

The SF3 data came through as a series of tables, derived from the Census Bureau's April 2000 long-form survey, sent to about one of six households. With that, the agency took the responses and crafted summaries, released in 76 data files for each state, plus a file containing geographic information.

And while the data can - and will - keep reporters busy for months to come, there are stories aplenty to examine and learn from.

The gender gap

The Detroit News, for example, looked at the slow improvement in the "gender gap" - the difference between what men make and what women make in the same jobs.

Brad Heath conducted data analysis for the newspaper. A big challenge, he said, was not to go too far.

"The biggest problem was that the data stopped just shy of giving me the smoking gun I was looking for," he said. "The data the Census Bureau provided show inequality to be sure, but they can't get you far enough to show actual bias."

Still, The Detroit News could report that while every state has made progress in closing the gender gap from 1989 to 1999, Michigan lagged. In that state, women who worked full time made 67 cents for every dollar earned by men, a figure that budged less than 5 cents in the 1990s. Nationwide, women who worked full time in 1999 earned 73 cents for each dollar men earned.

Like almost any database, though, the Census data didn't answer the question: Why?

"The census made clear that women and men work at different jobs and have different educations, both of which clearly play a role in the earnings disparity," Heath said. "But there wasn't enough there to come up with a solid answer to why Michigan ranks so low in pay for women. That was frustrating. And because the Census Bureau changed some of its job categories in the past decade, it wasn't really possible to show how well women are making inroads into higher-paying fields here."

Complaints about comparing 1990 and 2000 data items were common headaches expressed on Census-L traffic for more than a year.

Rising black income

Dwight Worley of The Journal News of White Plains, N.Y., encountered challenges in reporting on the rise of the black middle class population in New York in general and The Journal News' circulation area in particular.

One worry: error margins. What might look like a big change or trend could turn out to be statistically insignificant, depending the error margins or how much of the data is imputed - that is, estimated - by the Census Bureau.

"So you have to be careful - especially when you're comparing data over several decades," Worley said. "That's why we stuck with using state and county-level data, which have lower error margins than town or place level data."

To make those several-decades comparisons, Worley looked at income, education, employment, poverty and home ownership factors, from the SF3 in 2000 and from either Summary Files 3 or 4 from earlier years. He perused PDF files on the bureau's Web site for 1990 SF4 information. Going back to 1980 and 1970 was even more difficult. He finally went to government document libraries in New York City and Westchester County to go through books and microfiche images and just plugged in numbers by hand.

Then there's the definition of a household vs. a family. Here's how the Census Bureau divides the two groups: A household includes all of the people who occupy a housing unit. A family includes a householder and others, living in the same household, who are related to the householder by birth, marriage or adoption. Households, then, may include more members than do families. And not all households contain families, since a household may be composed of a group of unrelated people or of one person living alone.

There are tables that break down the number of workers by family type and the aggregate income of each type, but not by race. Household income is broken down by race, though, and counts everyone - not just those who constitute a family.

But, Worley said, even that can leave questions. "If there are, on average, four workers in a black household now when there used to be only two in 1970 then that could seriously alter the analysis of rising incomes. In that instance, household income is rising, but it could be because there are more workers in the house, not necessarily because individual blacks are earning more money."

Race, ethnicity, money

The Miami Herald found a rise in black income in Florida. That newspaper used median income by race and ethnicity to make a comparison down to the level of places that have a population of at least 10,000.

The Census Bureau provided a 1990 table of those places and median income.

Said The Miami Herald's Tim Henderson: "That was fine as far as it went, and showed us that there was a large increase in black income in Florida, but it wasn't reflected in our urban county (Miami-Dade) - only in our suburban county, Broward. There, black income soared and reached 81 percent of white non-Hispanic income, better than anywhere else in the state."

One of the problems of using only places and counties, he said, is that it could mask changes in areas that weren't marked by municipal boundaries.

"We could see this clearly when we mapped tracts by black income, screening for tracts with a significant black population [at least 10 percent]," he said. "A new regional area of black affluence stood out clearly at the western border of the two counties, and the ethnic makeup of those tracts made it clear that Jamaicans were playing a role in that new affluence. African-Americans also dominated in the adjoining Dade County tracts."

Those points made a story about the region's ethnic chemistry a positive note in light of the recurrent criticism that the region has no black middle class to provide hope and support for the poverty-stricken inner cities, Henderson said.

Commuting with CAR

Any movement from the middle of a city means an increase in the number of commuters and the time they spend on the roads. Matt Waite followed commuters in Pasco County, Fla., for the St. Petersburg Times.

Economic demographics play a large part in the commuter game.

"Pasco County is made up of three distinct areas: west, central and least," he said. "The east is the same as it was 40 years ago - rural. Central is the rapidly developing area of $150,000-plus homes. West is where all the retirees moved in the '60s and '70s, and they are all now dying off, leaving their $40,000 to $60,000 homes behind. The people moving in are 20- or 30-something, with kids."

Those moves, of course, mean an interesting mix of commuters.

"You have a population of people buying low-cost homes who have to commute long distances, which is counterintuitive," Waite said. "You expect more well-off people driving a ways to their downtown offices. But you don't expect to find an area with a $23,000 median household income driving an average of 30 minutes to work every day."

And, like the gender gap or income stories, the data can't tell the whole story. In Waite's case, the big question is where the commuter is going. The data can only tell where the commuter lives.

Now journalists are awaiting the arrival of still another data release - called Public Use Microdata Sample, or PUMS. Beginning with a release tentatively set for December, PUMS data is a sample of the actual responses. With that, journalists can go beyond the Census Bureau's cross-tabs and build their own summary tables.

Contact Jeff Porter by e-mail at jeff@nicar.org.