I hate to break up a meeting of the Douglas Elmendorf fan club, especially as the club counts many people I deeply respect among its members, but I think the Congressional Budget Office needs a change.
The issue isn’t so much Elmendorf himself. He’s a professional, and a level-headed liberal, which is a rare gift these days. I wouldn’t say the last few years have been a wonderful time for the CBO, though they haven’t been disastrous either. But I think the basic structure and character of the work of the agency (as well as of Congress’s other scorekeeper, the Joint Committee on Taxation, which assesses the revenue implications of legislation) needs to change pretty dramatically over the coming years.
It’s never easy to be the CBO — the agency gets thrown into the middle of complicated policy debates and is expected to offer relatively neutral assessments of highly contested questions in circumstances of great uncertainty. Congress pretty much requires the CBO to pretend it is more confident than its economists actually are, demanding point rather than spread projections and ignoring all caveats about uncertainty, yet members rarely have any compunction about criticizing projections they don’t approve of. My own interactions with CBO staffers in the last few years have always left me impressed with their professionalism and good intentions under these circumstances.
In a few instances in recent years, especially in its estimates of the effects of the proposed Senate immigration bill (or more precisely, in its unwillingness to be clear about its baseline estimates of immigration levels) and in its peculiar decision to effectively give up trying to score the deficit effects of Obamacare, the CBO has stood in the way of understanding more than it has had to. I certainly wouldn’t say these were reasons to fire Elmendorf, but they might weigh against renewing him in a Republican Congress.
More important, though, it is well past time for a more dramatic change in the way Congress’s scorekeepers do business. Simply put, the modeling and scoring done by the CBO and the JCT involves a “black box” process closed to outsiders. The agencies are both staffed by hard-working and highly professional economists who try to ensure their assumptions and methods keep up with the latest academic research, but their models are opaque and proprietary — which also makes them seem arbitrary and unpredictable. This, combined with the importance of the scores they provide, means the two agencies can throw key debates into heated (if often highly technical) confusion.
The “black box” nature of their work also means that the CBO and the JCT often act as major bottlenecks in the process of policy development, as members of Congress (except for a few committee chairmen) must generally wait many months to get a proposal scored, and can never know if the time and work they have invested in an idea will prove pointless when the agencies return with an unexpected score. This drives many members away from proposing ideas, and it has also launched a small industry of outside modelers at universities and think tanks who try not so much to model the world as to model the congressional models.
The JCT is a far worse offender on this front — often offering no hint about its methods even when it comes to very important projections and scores. But although the CBO’s opaqueness is less absolute, it is (because of the broader scope of its work) probably more consequential on the whole. MIT economist Jonathan Gruber, recently disgraced for (among other things) acknowledging the abject gaming of CBO’s model in the development of Obamacare, was paid $400,000 (in taxpayer dollars) by the Obama administration mostly for use of his academic model. It was worth so much to them, as the Washington Post recently noted, “in large part because it is similar to the model used by the Congressional Budget Office. That means administration policy-makers could predict with reasonable certainty how the CBO would score legislation.”
That’s ridiculous, but it is not uncommon. And it is not how 21st-century government modeling agencies should work. As Republicans take the reins of Congress, they should consider changing the role of these agencies: Rather than produce projections behind a heavy veil, the CBO and the JCT should ultimately act as developers and stewards of an open public model, available online to anyone with sufficient technical prowess to use it.
The two agencies, with outside help, would create and maintain the model, and would decide on (and make publicly known) a set of official economic assumptions and policy expectations to be used in modeling for formal budget-process purposes. Anyone could “score” any policy proposal using those official assumptions — and so could know the CBO score for that proposal immediately — and could also use the model with other assumptions to project a proposal’s consequences under alternative circumstances.
Users could also propose improvements to the model, to be considered by the two agencies. And outside developers could build applications that made use of the underlying model to make it easier for non-experts to use or to apply it to a variety of circumstances that the agencies might never think of. Such an open model would be a public resource, rather than a private secret. It would contribute to greater transparency — putting assumptions on the table to be debated and making the enormous uncertainty of projections better understood.
It would also help clear up the policy bottleneck the CBO and the JCT have created, unleashing some much-needed policy creativity on all sides of our politics. Congressional offices and committees, along with think tanks, universities, and companies could employ people with the knowledge to use the model, allowing ideas to be tested without requiring the two agencies to use their own manpower. And that kind of widespread experimentation would improve and strengthen the agencies’ ability to make projections for Congress.
I’ve been pushing this idea on an assortment of innocent victims for some time now (and several other people who actually ought to be listened to, like Pascal-Emmanuel Gobry and Avik Roy have recently argued for versions of it too), and I have encountered two broad objections above all. First, some worry it would undermine the authority of the CBO and the JCT, who can now give Congress hard-number conclusions rather than an open-ended mechanism for projection. But surely a more transparent model would increase the repute and authority of the two agencies, not undermine it. Their model would quickly become the foundation for all national policy debates, with candidates, officeholders, journalists, and experts relying on a common metric to test proposals and ideas.
A second objection, raised especially by current and former CBO officials, is that the two agencies don’t exactly have “a model” now but rather a series of complex spreadsheets and a careful professional process of applying them to proposed legislation. Replacing that system with a single universally useable model over time would involve a dramatic change in how the agencies do business based on an oversimplification of how they now work. But surely this too is actually an argument in favor of such a change. Today’s technology can enable the two agencies to standardize their work and develop a coherent yet constantly updated model that would serve them better than the sort of ad hoc procedure they have come to use over the decades. Social and physical scientists are increasingly making detailed models in many fields available to all as part of the standard transparency of peer-reviewed work. Doing the same would reinforce the authoritative stature of congress’s modelers, and would be good for them, as well as for lawmakers and the public.
There are some significant technical challenges to be thought through, of course. Keeping such a model updated would require a lot of work from the CBO and the JCT (as they have found over the years in keeping their existing models up to date), and some novel ideas just won’t easily fit into a system built to model today’s public-policy architecture and so will require constant improvements. Security will also be an issue, particularly if the model incorporates some confidential or proprietary datasets. But these are not insurmountable obstacles, they will just require a significant rethinking of the mission and structure of congress’s scorekeepers. There are also lessons to learn from work being done along these lines by others, especially by the Tax Policy Center (which has experimented with a publicly accessible face for at least one of its tax models) and the American Enterprise Institute, which is on the threshold of an important groundbreaking project in genuinely open-source modeling.
Getting there would take much work and trial and error, to be sure. It need not (and could not) be done in an instant. But Congress should instruct the CBO and the JCT to begin a gradual and transparent transition to a standard open model, calling on the aid of academic, think-tank, and commercial modelers as they do so. If the Budget Act of 1974 (which created CBO) were being written today, this is the kind of role it would give to the scorekeepers. The CBO and the JCT do immensely important work, and they generally do it well, but they have also contributed to some very great perversity in key national debates, and it is time for them to lead our economic debates into the 21st century.
Republicans could start this process on their own, since the agencies will now work for them, but there is no reason such a reform should not be thoroughly bipartisan. I do think, though, that it would make sense to launch it with a new director. Not because Douglas Elmendorf isn’t a professional and a man of integrity, but because a change of direction could be helped by a change of leadership.