In California, justice isn't color- blind. Constitutional guarantees of equality notwithstanding, that's what the San Jose Mercury News found recently in a computer-as- sisted study of nearly 700,000 crimi- nal cases over a nine-year period. As more and more cases are decided by plea bargain, whites as a group get significantly better deals than Hispanics or blacks who are accused of similar crimes and who have similar criminal backgrounds.

Instead of judges or juries handing down verdicts, the fates of hundreds of thousands of defen- dants a year often are negotiated over the telephone, in hurried hall- way or parking lot conversations or in closed-door meetings. What happens? At virtually every stage of pretrial negotiation, whites are more successful than non-whites. They do better at getting charges dropped. They're better able to get charges reduced to lesser offenses.

Our story began on the cam- paign trail in 1990. In get-tough-on- crime rhetoric, the two major candi- dates for governor including soon- to-be Republican Gov. Pete Wilson - were sounding a lot like outgoing Gov. George Deukamejian, who himself was elected mostly on a law-and-order pitch. So we decided to look at what nearly a decade of enacting new crime laws, appointing hundreds of "common sense" Deukamejian judges and building cavernous new prisons had done to cut crime and make neighborhoods safer.

The answer was: Not much. One reason, we were told, was California's "determinate sentencing" law, which removes discretion from judges' hands by mandating specified prison terms for specified crimes. The idea behind this law was to remove racial disparity from sentencing. But, one of our sources suggested, what if dis- crimination had simply shifted elsewhere in the proc- ess? And thus our story was born.

As we moved into it, it became clear the story was tailor-made for computer-assisted reporting. Simply doing some limited sampling of cases wouldn't work. Once you start looking at specific crimes by criminal background and by specific racial/ethnic group, the numbers can get pretty thin. And when that happens, it doesn't take much to distort results. Our main data sources were two state files: O. detailing adult felony arrest dispositions, the other pro- viding information on those sent to state prisons.

We're pleased to say we handled the entire project on a PC, without having to resort to mainframe or systems people help at all. Our largest of several files was about 100 megabytes, and although we bumped into a limit on the number of fields allowed in a record, Paradox was a speedy database manager for combing through all the information. Getting the data was a best-of-times, worst-of- times affair.

The felony arrest data came from a state agency still willing to make information available quickly and (in this case) at no charge. Getting the prison data, however, was an ordeal. In an experience very much like that recounted in Uplink recently by the Orange County Register, it took months of negotiation and haggling over price before we got what we wanted. Standing up for criminals' rights isn't politically popular these days, but we thought the story important enough to devote about seven months to it.

But as time as that might seem, it still pales next to how lon, would have taken without our trusty nine-track tape reader and a 33-megahertz 386 clone.