As best I can tell from word searches on the Washington Post and New York Times websites, my favorite part of the President's budget did not make the mainstream media's near-exhaustive coverage at all.

The overlooked gem is the discussion of how the budget is seeking both to increase funding for new program efforts where there is actual good evidence, based on rigorous research, that they work -- and also to provide more funding for efforts to use social science research to find out whether programs work. Together, this approach goes under the moniker "evidence-based government." This section of the budget reflects what grew out of a call in instructions to agencies for this year's budget to include evidence-based initiatives in their submissions to OMB.

So the budget proposes funds for experiments to test the effectiveness of various potentially promising interventions to keep people with disabilities in the labor force or to improve the quality of college education while reducing costs. For some programs where there already exists evidence of success, the budget recommends funding increases. The budget also recommends increased funding for so-called "pay for success" programs in various policy areas, including interventions designed to reduce costs. (In the contracting area, this has been called "share-in-savings" contracting, to which the budget has therefore given a boost.)

However, one thing is sadly missing from the discussion: examples of programs being cut back because of evidence they don't work.

There is a certain bipartisan wonk coalition that gets excited about this. I have blogged in support of this approach. My Republican friend Robert Shea, who was in charge of performance measurement in the Bush administration, enthused on his Facebook page that this was "what playing Moneyball in Government looks like."

But there is also a bipartisan coalition that does not like this. Some Democrats are worried that if you establish the outrageous criterion that we should be hesitant about funding programs that don't work, government will get smaller. Some Republicans are worried about handing too much influence over policy formation to pointy-headed professors who do research. There is a huge know-nothing constituency in Washington that wrongly believes (based on many of the egregious products that partisan think tanks or paid-for consultants put out) there is no such thing as genuine scholarly research following accepted methodological standards. And evidence-based government does not get the ideological or combative juices of either side flowing.

Evidence can't answer all questions, such as how to deal with a program that does achieve benefits, but costs a lot per amount of benefit delivered, or how to deal with programs that benefit different groups differentially. In politics, there will always be a role for values as well as evidence. But surely an additional dose of evidence in political debates would be a good thing for our decision-making processes.

So I would say that if you are inclined to see problems with evidence-based government, I ask you to consider the alternatives.

(By Steve Kelman)