Driving CAR on deadline is not hard. It's kind of like the old saw about the tourist in New York who asks an old man on the street: "How do you get to Carnegie Hall?" The answer: "Practice, practice, practice." Except, in the case of computer-assisted reporting, it's more like: "Prepare, prepare, prepare." Now, this may seem like a contradiction in terms – How can we prepare for the unexpected? But it's really no different than any other deadline reporting we do.
We all carry around a mental list of sources to use on breaking stories, whom to call when there's a major crime, an unexpected political battle or a big education story that demands coverage for tomorrow's paper or tonight's broadcast. It's no different with CAR. In CAR's case, though, it's a matter of knowing what data are available, knowing how to analyze it, whom to call to verify or interpret it.
It's a matter of what can be called "thinking in rows and columns" – thinking always about how data can supplement a story. Now, we're not talking about big projects here. We're talking about daily or two-day stories; sometimes we're not talking about stories at all but "computer-assisted paragraphs," those little bits of context that make a good story even better. Here are some examples from recent history at the Broward County edition of The Miami Herald: Tobacco and kids.
A couple of months ago, we sent a FOIA to the FDA requesting the database they used to build their "Children & Tobacco Compliance Checker" on their Web site. It contained information on how states and various retailers were doing in the effort to enforce laws barring sale of tobacco products to young people. The response to our letter was typical: "We don't have any way of doing that, but we've asked our outside contractor for a program and they're working on it." A few weeks after that, they called: "We've got the data and we've put it up on the Web site.
You can download it." Great. The FDA in its wisdom had decided to make the data "easy" to get by releasing it to the world instead of to the handful of reporters who'd asked for it. Now what had been a leisurely, spare-time kind of story became competitive. (The Herald's Broward edition competes fiercely with the Sun-Sentinel, the Chicago Tribune daily located across town.) We downloaded the data, saw it was in a beautifully simple, comma-delimited format and imported it into Access.
Browsing through it, there were a few undefined codes, so we put a call in to the FDA for interpretation. We also saw that while cities were in the data, counties weren't. That meant quickly creating a look-up table to group South Florida cities into the counties in which The Herald was interested. Finally, the analysis: How was Florida doing compared to the rest of the country? How was South Florida doing?
How were our local retailers doing? The result: We were doing better on all counts. Florida's push – and it's a big one – to get kids off tobacco was bearing fruit. Though this was not scientifically gathered data – it was not random but showed only what happened at the retailers FDA checked, a fact made clear in the story. So what we had was a nice, incremental story, a pat on the back for Florida, maybe not destined for 1A but for sure good on 1B.
Total time: Two days. And it would have been one had the FDA not taken its sweet time calling back. Elections. Local elections are an example of how CAR can add depth and context to stories we already do. There are basically two questions we ask after an election: Who won? How did they do it? That often means staring at a table: Voting districts or precincts down the side, candidates across the top.
Reporters often eyeball the results in certain "key" precincts, call a couple of pundits and throw a story in the paper or on the air. We do it a little differently. We've arranged with our county election supervisor to download precinct-by-precinct results at the end of election evening or, at worst, early the next morning. We import that into an Access database in which we've already prepared several lookup tables showing voter registration, precinct or district demographics, etc.
With that, it's easy to pull out areas or demographics in which we're interested to show, for example, how predominantly black districts voted or to look at the "condo canyons" full of retired, New Deal Democrats. Add a mapping program such as ArcView, and you can draw a precinct-by-precinct picture of exactly how the vote went. And that lets us ask the candidates and the pundits exactly why they think various demographics voted the way they did instead of asking who made the difference.
Another example: Every week, we survey 34 gas stations around the county and print the price of regular gasoline at each. Since the beginning of the year, we've also keyed the results into an Access database. Because we'd taken the time to prepare, we were in an ideal position to add local context – county and neighborhood averages, week-by-week price fluctuations and the like – to our stories when gas prices jumped dramatically a few months ago.
Conclusion. What these examples have in common is this: We think in rows and columns, that is, we look for stories in which data might be used. We take our opportunities when we see them. We anticipate stories we know we're going to need to do and prepare data appropriately. And we do a couple of other things: We prowl constantly for useful data and maintain a data warehouse, available to all reporters at their desks.
We use these databases to background people, find neighbors, check salaries, etc.