What's the Deal with CBO's Scoring?

There's been a lot of wonky discussion about the CBO's scoring methods lately. The other day, Ezra Klein defended the CBO against conservative critics, insisting that the agency's costs estimates represent the "best guess of the town's most rigorous guesser."

Ezra writes:

[B]e very careful with any criticism of CBO that seems to be merited by a particular score rather than a particular methodological difficulty. To put that slightly differently, does anyone think that conservatives would be squawking if CBO had disappointed Democrats by saying the bill would save less money than either the House or Senate incarnations? If not, then keep in mind that this is a political, not technical, dispute. To establish my own credentials on this, here's the post I wrote defending the CBO when liberals were arguing that it was underestimating health-care reform's savings.

I think Ezra is a bit disingenuous when he pretends that he's never engaged in partisan attacks of specific CBO cost estimates. (See here, for example.) But there is a much bigger criticism of Ezra's point, which Greg Mankiw articulates very well here.

By convention, the CBO uses something called "static" budget scoring to determine a bill's impact on the federal budget. This kind of scoring essentially disregards the macroeconomic implications of federal actions.

Here is how former CBO director Douglas Holtz-Eakin explained it:

For every piece of legislation . . . the budgetary impacts are estimated using the same, unchanging baseline projections of overall gross domestic product (GDP) and its aggregate income components. Specifically, the estimates do not include the effects of legislation on the supply of labor or on saving (and hence on overall economic growth); they do not include effects on income that might result from the influence outlays and taxes, say, may have on technological progress; they do not include the increases or decreases in output that are caused by the way subsidies or taxes reallocate resources among various activities; they do not include the effects on national saving and other incentive effects that result from the government’s financing of the budget change; and they do not include the income and employment effects that arise from the impact of fiscal policy on aggregate spending in the economy in a recession.

To steal a premise from Paul Ryan, this means that the CBO's current scoring methods would assume no macroeconomic impact if the federal government increased spending by, say, 50 trillion dollars . . . just as long as that new spending was offset by 50 trillion dollars in tax increases.

The alternative to static scoring is some kind of "dynamic" budget analysis, which would account for macroeconomic feedback effects rather than simply holding baselines GDP estimates constant. This would give us a much more accurate cost estimate for large policy changes.

Under the current assumptions, we're virtually guaranteed to get an faulty score.