Numbers Game

In re my post on the end of the CBO process, Kevin Drum asks "What's wrong with hitting budget targets?"

Well, to start with, a bunch of the mechanisms used to hit those targets were pretty sketchy--as Kevin himself noted a few days ago, the revenue raising starts now, but the spending starts in 2014; this has the effect of both increasing the deficit reduction, and shrinking the price tag of the bill.  If liberals can't pass a $2 trillion ten-year bill with voters, then they shouldn't shift the time window so they can artificially call it a $1 trillion bill.  The Bush sunsets may have been a gimmick, but at least the sunset will actually occur.  Their estimates of the cost of the Iraq war may have been wrong, but at least they were genuinely, idiotically wrong, not the artifact of a budget metric.

But to answer the more general question, I will . . . pose another question:  what was wrong with the ratings on mortgaged-backed securities?

One of the reasons that we got so deep into the crisis is that people were substituting ratings for common sense.  Not all people, but enough.  If a security was rated AAA, well then, it must be an excellent security.  People confused the rating with an actual prediction of the future.

Indeed, the rating was supposed to bear some (not perfect) resemblance to the future.  But the more important the rating got, the more money there was to be made by gaming it.

The raters used metrics--perhaps more than they should have.  So if you knew the metrics, you could tweak the security to get a good rating on your toxic waste.  In the past, a well designed security would (arguably) have naturally passed the metrics.  But now the point was not to build a well-designed security; it was to design a security that would just barely pass the standards used by Moody's or S&P.  Then people would buy your crap, and . . .  oops! we crashed the financial system.

This also appears to have been the genesis of substantial fraud in the mortgage market.  Computers had arbitrary cutoffs, so many future metric-gamers got their start tweaking inputs by small amounts to see if they could hit the numbers--raising the annual salary by $500, or the FICO score by a couple of points.  Individually, this actually makes reasonable sense--if you make $75,000 a year, $40 a month is probably not going to break your budget.  Collectively, and taken further, they were disastrous.

People are treating the CBO estimates as if they are predictions.  They aren't.  The CBO does not make accurate predictions of the future, because it can't.  Moreover, the CBO very explicitly does not forecast the most likely scenario; it forecasts what is written into current law, no matter how unlikely it is that this current law will be implemented.  Thus the ritual annual "temporary" fixes in things like the AMT and the SGR, performed to keep the full cost of the measure from showing up anywhere, even though we know we are never going to let them go into effect.  The CBO's goal is not accuracy, but consistency:  a method that is not influenced by political predictions, and generates similar results for similar bills.

Accurate predictions would require estimating, for example, whether the Medicare cuts will prove unworkable for either political or practical reasons, and be repealed.  But that is not the CBO's job.  And it shouldn't be; the agency could not function if it were injected into politics.

Doug Elmendorf, the head of the CBO, has made it as clear as he is allowed to that he is very skeptical that these cuts will take effect as written.  They are not scored because he thinks this bill will reduce the deficit.  They are scored because it is his mandate to score them.  But the scores are being treated by the media as if they were the CBO's best estimate of the actual future, rather than the product of a highly artificial model.

Producing bills that will--just barely!--fit the CBO criteria is not the same as producing good bills, or bills that are fiscally sound.  As Michael Cannon has pointed out, the Democrats seem to have been very carefully dancing around criteria that would force the CBO to estimate, and publish, the cost to individuals of the mandate.  One can argue about whether or not these costs should be included.  But they're not being left out because it's a good idea, but because Democrats know exactly what they have to do to avoid getting it put in.  We're not getting a bill that starts in 2014 because we need lead time, but because Democrats want to be able to use a fairly deceptive estimate of its costs.

It was totally legitimate for securities issuers to go to Moodys and say, "What kind of rating does this tranche get?"  But at some point, the relationship got too close.  Many allege that this is because of conflicts of interest, or financial malfeasance; I withhold judgment on these accusations.  Whatever the cause, it has become clear that if raters and the rated work too closely together, the ratings begin to bear less and less resemblance to actual reality.

This is not a partisan complaint.  I do not think that the Democrats are sneakier, only that they thought of it first--and happened to have a former CBO director who worked closely with Ron Wyden at the OMB.  I am sure that Republicans will use the same tactics, at which time I am sure that most of the liberals who currently admire them as political masterstrokes will find reasons to be outraged.  But my complaint is more general.  CBO scores, like securities ratings, used to have some rough relationship to the future.  Now they have almost none--they are mostly arbitrary constructions aimed simply at gaming the rating system.  We are losing the best tool we have for evaluating major legislation.  This is not, as Kevin has interpreted it to be, an objection to passing health care, nor is it some kind of complaint about the bill's legitimacy. It's just a pressing worry about the future, and a plea not to let this sort of thing become common.

But I suspect it's too late.  Even if Democrats relinquished their new toy, the Republicans would insist on their right to use it when it is once again their turn in power.  Political innovations are not so often undone.