College tuition prices keep rising. State budgets are stagnant or shrinking. And policy makers, from President Obama on down, are increasingly calling for increases in the number of Americans who get some higher education or training.

Those factors have led more state legislators, trustees and others to argue that, to accomplish the latter goal given the former circumstances, colleges are going to have to lower what they spend to produce the average credential they award. But any discussion of lowering the "cost per degree" must start with a more fundamental question: What does a degree cost to produce now?

That question may be basic, but it is not simple, as a new report from the Delta Project on Postsecondary Education Costs, Productivity, and Accountability makes clear. The paper, prepared by Nate Johnson, associate director of institutional planning and research at the University of Florida, lays out a range of possible approaches to calculating the cost of a college degree and then calculates them using a rich set of data from the State University System of Florida, where Johnson formerly worked.

The paper shows that it is distinctly possible to come up with such a figure, but the wide variation in the numbers -- based on institution type, program, degree level, and other factors -- suggests that the answer will depend in large part on how the question is framed. And that decision is a surprisingly value-laden one, says Johnson. "You frame the question one way if you are only interested in students who graduate, and another way if you want to know the cost for people who go to college and don't complete," he says. "The point is, this is not just a data question. It's a question of what it is that we want from our colleges and universities."

The broad work of the Delta Project and its founder, Jane Wellman, is to analyze the "spending side" of the higher education cost and price picture; the group has released a series of reports that try to document the interplay of colleges' revenues and expenditures, and how those trends affect what they charge to students. The new study, which grew out of Johnson's work in Florida, he says, aims to develop a "common language," if not a common format, for focusing the discussion about how one might measure the cost of a degree in a particular institution, system or state. Toward that end, Johnson proposes several possible ways of calculating the average cost of a degree.

The analyses are based on data showing that the Florida university system incurred an average of $288 in direct and indirect instructional expenditures per credit hour, with wide variation by level ($188 for lower division undergraduate, $537 for master's, etc.), institution ($240 for an upper level undergraduate credit at the massive University of Central Florida, $677 for the same credit at the 700-student New College), and field of study ($159 in family/consumer sciences, $509 for natural resources/conservation). The analysis counts only those expenditures derived from state appropriations and student tuition, excluding endowment and other funds.

The first estimate, which Johnson calls the "catalog cost," calculates what a college would spend to educate a student who fulfills the "catalog requirements" of the average degree to the letter -- no more, no less. (The equation: cost per credit hour x instructional expenditures/credit hours.) The average cost is $26,485, with institutions within the Florida system ranging from $22,440 to nearly double that. Johnson also found significant variation by field because of vastly different requirements and program length, with mechanical engineering averaging $37,870 vs. $27,159 for elementary education.

The catalog method is easily understood, but it "does not reflect actual student behavior," Johnson notes. More accurate in gauging how students actually maneuver through institutions, he writes, is the "transcript method" of cost analysis, in which the total number of credit hours students take are multiplied by the cost per credit hour, and then divided by the number of degrees awarded. The average freshman who entered a Florida system university and graduated in 2003-4 "attempted" 131 credits, including failed or withdrawn courses and subtracting for any AP or dual enrollment courses that reduced their course requirements.

The average "transcript cost," then, was $31,763; converting to 2006 dollars, to make parallel to the figures from the "catalog cost" analysis, Johnson writes, the average figure is $33,672. (The 2003-4 figure for mechanical engineering was $47,257.)

Both the catalog and transcript cost methods factor into the calculation only those costs incurred by students who actually graduate. The third major analysis, "full cost attribution," examines the entire amount that an institution or system spent on instructional purposes to achieve an "aggregate level of degree completion." The equation looks like this: all credits taken at an institution over three years x the three-year average cost per credit hour/three years of degrees.

Not surprisingly, because all courses taken by all students would be allocated to the smaller proportion who actually earned degrees, this produces the highest cost per degree number; $37,757 in 2002-3 dollars, equivalent to $40,645 in 2005-6, Johnson writes. This analysis grows less predictable and valid the more narrowly it is drawn, he adds, because programs with high attrition, or into which many students transfer late in the game, can have their figures drastically altered. The overall high and low for the Florida university system, for example, were $170,831 for "multidisciplinary studies" and $21,473 for parks and recreation, and the variation by degree level was enormous: $33,425 for a law degree, $259,781 for an M.D., and $121,725 for a doctorate.

So which is the most accurate assessment of what a university spends to educate a graduate? The catalog cost of $26,485, the transcript cost of $33,672, or the "full cost" $40,645? The last is "probably closer to an answer" to the question that policy makers are increasingly asking now, about "what would we have to spend to get more graduates," though that assumes that colleges maintained their current enrollment and expenditure levels, he notes.

But the other key point, Johnson says, is that the choice of how you measure cost depends, to an extent, on how you perceive the role of colleges. Using the "full cost" measure, he asserts, more or less says that most of what a university does is designed to educate students, and that "all of those costs could be attributed to the cost of producing college graduates," as overhead, he says.

"If you highly value research or public service," though, "you could almost say that the graduates are free -- a byproduct" of what you spend on those other purposes.