"Confronting the Numbers," a three-month Pittsburgh Tribune-Review investigation of the Pennsylvania lottery, began by confronting the kind of statistical conundrum that makes journalism such a fun job. The main point of my first computer-assisted reporting project was to see if the conventional wisdom - that the poor gamble on the lottery more than the rich - was true in our metro area. Not an easy question, it turned out. There are essentially two ways to use state data to determine who buys lottery tickets. One relies on sales information. The Pennsylvania Lottery will provide a spreadsheet of annual lottery sales by retailer. The spreadsheet includes the retailer's address, so a CAR reporter can sum the sales data by zip codes.
But numbers don't tell the whole story. Eventually, I had to put the mouse down and pick up the phone. I even had to (gasp!) leave the office.
Using demographic statistics, such as Claritas data, it's easy to figure per capita sales for a zip code, and compare that to median incomes in the same neighborhood.
The problem with this method is that it can't determine whether most, all, or none of a store's sales are to people who live in the neighborhood. Maybe the store is selling to commuters on their lunch hour or on the way home from work.
Show a lottery official your sales-based analysis, and he immediately will try to shoot holes in it for that reason.
Alternate method
The other method focuses on winners. Lotteries have to report anyone who wins more than $600 in a game to the IRS, and those records are public.
Because winners are a random sample of all players (if not, you've probably got a bigger story on your hands), a CAR reporter can extrapolate what share of sales comes from a given neighborhood based on what share of winners lives there.
The problem with this method is that $600 threshold. In Pennsylvania, the prize for hitting the three-digit Daily Number on a $1 straight bet is $500. Those winners aren't reported to the IRS. Numbers games are extremely popular in Pittsburgh, especially with its large elderly population. They are the game of choice among the heaviest players, the ones who collect lists of lucky numbers they notice on license plates or passing buses.
A Pennsylvania lottery official confirmed to me that if savvy numbers players want to bet $10 on 7-9-2 straight, they buy 10 $1 tickets. That way, if they hit they get 10 $500 payments instead of a single pot of $5,000 that will be reported to the IRS.
Given that choice, I opted to use the sales method. It is more straightforward and easier for readers to understand.
I eliminated from my analysis any ticket sales in the downtown area or in the airport, where sales were most likely to be to people who didn't live nearby.
Though Pennsylvania is notoriously secretive, they didn't balk at mailing the sales data on two disks of Excel spreadsheets. Because the data were statewide, I was able to give customized stats to sister papers in neighboring counties, allowing them to do similar stories.
The results
Surprise! Sales were highest in working class neighborhoods, and lowest in the wealthier suburbs.
Now comes the part most dangerous for CAR geeks - getting buried in spreadsheets. I made top ten lists of this, figured per capita of that, made pie charts and line graphs and all sorts of neat stuff. I became a data weasel. But numbers don't tell the whole story.
Eventually, I had to put the mouse down and pick up the phone. I even had to (gasp!) leave the office.
I visited the library and read what the experts and other pundits had to say about lotteries. I called the professors, the industry experts, and the state budget gurus. Several long interviews and hundreds of inches of notes yielded no quotes at all, but helped me understand the subject.
I found out about a 1987 legislative resolution to study how much the lottery causes problem gambling - which the state has since ignored.
Months of poking around caused some leaks to spring. We got a tip that 15 employees got custom Pennsylvania Lottery sportcoats, purchased from the lottery director's tailor, costing the state $500 each.
I visited a Gamblers Anonymous meeting in a church basement in the neighborhood where sales are highest. There I met several self-professed lottery addicts who were surprised and angry to hear the state director and revenue secretary both claimed such people did not exist.
At the region's top-selling store, our photographer met a poor family that spends $9 a day on tickets - two bucks for Mom and Dad, and one for each of their seven children. Their story sold the package and generated the most response from our readers.
I waited until late in the game to go to the capital and interview the bigwigs. By that time, I had a lot of pointed questions ready, and was able to anticipate likely evasions and press for further explanations.
The state still has a lottery; I don't think any revelations are likely to change that. But as a result of our story, the state legislature held hearings to review the lottery's financial health, which our package called into question. Soon after, the auditor general announced his office would launch a formal audit.
And the lottery director caught a tremendous amount of flak for those jackets. Three months after the series ran, he resigned, citing a need to spend more time with his family.
Mark Houser can be reached by e-mail at mhouser@tribweb.com