Whenever legislation wends its way to the floor of Congress, there is a breathless period when everyone waits to see what score it will get from the Congressional Budget Office and the Joint Committee on Taxation. That score will indicate whether the proposed law will increase the deficit, or decrease it, and how it will affect the public and the economy. Those findings, in turn, may well determine whether the bill lives or dies.
This process can be quite complicated; often a bad score forces legislators back to the drawing board to come up with some more pleasing bill. The architects of health care reform spent a great deal of time tossing proposals at the CBO, seeing what the CBO model spat out and modifying things accordingly so that Democrats ultimately ended up with something that scored as both reducing the budget deficit and costing less than a trillion dollars over 10 years.
When Obamacare was finally voted on, critics (including me) complained that this iterative process of gaming the CBO scores was undermining the office's ability to discipline the policy process. Many of the things Democrats came up with to get that score were outlandish and unworkable — like strapping changes to an unrelated program (the student loan program) to Obamacare because it scored as reducing the deficit within the forecast window, and passing various taxes and add-on programs that ranged from politically fragile to completely disastrous, and seemed likely to expire in short order (as indeed, many of them did).
When the GOP passed its big tax bill, Democrats turned around and complained that Republicans were exploiting a hole in the CBO model that registered the repeal of Obamacare's individual mandate as saving an improbable amount of money (as indeed they were).
But both of these complaints are really a symptom of something larger: the fact that the CBO models are an inscrutable black box. We push bills into it; some numbers come out the other end. How were they generated? Further deponent sayeth naught.
In an essay at National Affairs, Matt Jensen, the director of the Open Source Policy Center at the American Enterprise Institute, argues that the inscrutability of these models ought to change. He makes a few basic points:
The CBO and the Joint Committee on Taxation have privileged access to government data that could be very valuable for the rest of us The CBO and the JCT are actually forbidden to follow the data-sharing and model openness that are common in much of social science, making it impossible to replicate what they do This creates a bottleneck for policy analysis, because their staff resources are limited It also makes it impossible for outsiders to check what they do in anything like a timely fashion. The people at the CBO are very bright, very dedicated and very knowledgeable; I am second to none in my admiration for them. But at the end of the day, they are one, relatively small group of people. They are going to make mistakes, and their modeling practices will keep those mistakes from being caught.
In other words, Jensen concludes, everyone would be better off if the CBO opened its models.
He spends much of the rest of the essay exploring how the CBO might do so within reasonable constraints involving things like data privacy. But I won't address those. Instead I'll ask whether he's right that the CBO should do this.
On its face, it seems like a crazy question. In the policy world, opposing transparency is a bit like opposing school lunches and mother-love. The one thing everyone can agree on is that more transparency is better.
But hold on: How many of us think that the world would be a better place if the CIA and the Defense Department were completely transparent about what they were going to do? (All right, Mr. Putin, put your hand down.) Most of the rest of us recognize that in fact, more transparency may make an organization less effective at its job.
