Thunder & Lightning News Service - Core Skills Seminars

TEN COMMANDMENTS OF CAR

People are the point of every CAR story, just as they are the point of every other news story. Numbers – in and of themselves – are not news. Nor are computer processes, or intricate tales of how the data was obtained, the point of CAR.

The dataset is the beginning of the reporting process, not the end product of it.

Data is not inherently honest. It is no more reliable than any other news source. Subject your data to the same scrutiny you would any other news source: Know where it is coming from, what its motivations, and where possible, corroborate it against the known record.

Don’t bite off more bytes than you can chew. Computer equipment and software is expensive. Statistical problems can be complex and confounding. And gremlins have an odd way of creeping into the best-laid plans. Don’t commit to any project you’re not sure you or your equipment can handle. If you have even the slightest doubt, agree to try it... but don’t commit to delivering.

Designing data queries is similar to formulating interview questions. The best ones are open-ended and presented in such a way that any answer is a potentially interesting and newsworthy one. (It is much stronger, for instance, to structure the query to find out what county has the highest proportion of people on public assistance rather than structuring it to find out if XYZ county has a higher proportion than another county.)

Double-check the data and debug the program after every operation, rather than waiting until you get the end result. If there’s no published record to check your data against, spot-check your calculations with a desktop or hand-held adding machine.

Whenever possible, avoid CAR on deadline. Just as you wouldn’t promise an interview with a prominent newsmaker before finding out if the newsmaker was available, don’t promise CAR results you’re not sure you can deliver. It is far safer – and, in the long run, more efficient for your news operation – to derive your data and process it before committing to the story.

Write twice. It’s the only way to insure that you bring context to the numbers. Leave time to write your story. Then put it aside and write it from the top again. Don’t think a quick pencil-edit will suffice. The only way to bring insight and understanding is to work hard on the words.

Make friends with the data-processing people from whom you are obtaining the data. Many times they are going to be your best source of help on sorting out problems or explaining apparent inconsistencies in results. They work with the data all the time, and often have knowledge to share about it.

When in doubt, get help. There are many sources of it all around you. For computer help: the station propeller-head (someone who works with computers for a living, or plays with them for fun) may know programs; the local college has staff and faculty who may be familiar with hardware and software; a local computer store (many have on-site trainers who know the machinery); NICAR (an invaluable resource). Statistical help: station accounting staff (people with business degrees usually have been through at least one stat class); ratings researcher (heavy statistical backgrounds here, and virtually every station has one on staff or in the station’s consulting firm); local college; NICAR, NICAR-L or CARR-L (on the Internet). Subject help: government agencies and legislative committee staff; Internet lists; academics and associations.