Donna Arduin has a quaint idea about government finance: She thinks every state program should have specific goals laying out what it is trying to accomplish and clear measures to help the public determine whether those goals have been achieved.

Arduin, Gov. Arnold Schwarzenegger's finance director, is spending much of her time these days helping the boss with his most immediate fiscal problem -- trying to close a $15 billion gap between California's revenues and its projected spending. But Arduin, who came to California after budget stints in Michigan, New York and Florida, also has her eye on the long term.

She wants California's budget writers to change their focus from how much is going in to what is coming out. Are we, in other words, getting our money's worth?

But her first step in that direction was actually a step back: a clear assessment of how much the state is spending, and how much that spending has grown, for every major program in the budget. She coupled that with a look at each program's growth relative to the state population and the caseload it is serving, and a quick review of the factors driving those costs.

Her findings, part of the audit Schwarzenegger promised during the campaign, have not been collected into a formal report but are sprinkled throughout the governor's budget summary.

On Page 98, for instance, is a review of the Medi-Cal program, which provides health care to the poor. Here we find that the program grew by $3.1 billion, or 41 percent, between 1998-99 and 2003. But health care costs are soaring everywhere, right? And California's cost-per-recipient is the second-lowest among the 10 largest states.

But Arduin's findings show that more than a third of the recent growth was due to policy decisions by the Legislature and the governor to expand eligibility for the program, adding 1.5 million people, or 23 percent of the total number now served. The changes were phased in over time, and the first-year costs totaled a relatively modest $158 million. But in 2004-05 the total will come to more than $1.3 billion.

In CalWorks, the state's biggest welfare program, Arduin notes that spending has remained relatively flat at about $2 billion a year even as the caseloads have declined by nearly 20 percent over the past five years. But California's caseload hasn't been shrinking as quickly as many other states' caseloads, and she projects that costs will soon grow again if current rules aren't changed.

And in the prisons, Arduin found, costs have increased by $1.6 billion, or 41 percent, in five years, even as the number of inmates behind bars has declined slightly. The reasons: higher inmate health costs, medical and psychiatric supplies, workers' compensation for prison employees, utilities, and higher pay and retirement benefits for prison guards.

Gathering such numbers was the easy part, however. They are all readily accessible to the governor's budget staff. The next step will be more difficult: trying to get the managers of every program and agency to define their objectives and develop ways to measure their progress.

If Medi-Cal is measured by how many Californians it serves, for example, or its cost per client, the program would have to be considered a smashing success. But are there other ways of assessing its value? Is it possible, or even desirable, to hold the program accountable for the health of the people it is supposed to serve?

And if the point of CalWorks is to move people off public dependency, it appears to be succeeding. But is it moving welfare recipients into the work force, or just getting them off the rolls? And how does its performance compare with what other states are doing?

Arduin wants each program to measure two different kinds of results: what she calls outputs and outcomes. Outputs, roughly defined, are a measure of how many people are served or how much of a thing -- say, new highway miles -- are produced.

Outcomes are a higher level of result: What effect, if any, are all those new highway miles having on rush-hour gridlock?

"What's the reason we are doing all these things?" she asks.

In criminal justice, for example, "We are trying to reduce crime. What's happening to the crime rate? It's going up. Why? It might be because these programs are not effective. But there are other circumstances we need to take into account. It's hard to get people to say, 'Wait a minute, all these programs we are putting in place, trying to reduce crime but the crime rate is going up, maybe we are not doing something right.' "

This is tough stuff. It seems simple at first but can lead to a bureaucratic morass, and to people gaming the numbers. The state tried something similar in the 1990s under Gov. Pete Wilson and eventually gave up in frustration. But other state and local governments, including Florida, Arduin's last stop, have implemented performance-based budgeting and are happy with the results.

Maybe it is time for California to give it another try.