A Good Start: CBO’s Return to GitHub
The scorekeepers showed their work. Some of it.
Part 6 in the Checking the Scorekeepers series
In our Checking the Scorekeepers series, we have argued that the Congressional Budget Office (CBO) and the Joint Committee on Taxation (JCT) operate with little transparency, limiting meaningful outside scrutiny.
In a WSJ op-ed last fall, we called on the agencies to post their code and data so that outside researchers could verify and stress-test the assumptions behind trillion-dollar scores. At the time, we noted that the CBO had posted just six models to its GitHub page in the five years since it started using the platform. When asked about the slow pace at a House Budget hearing last November, Director Swagel pointed to the demands of scoring the One Big, Beautiful Bill. He promised more models would be posted in 2026.
A few weeks ago, CBO delivered on that promise. The agency posted several new code repositories and outlined plans to publish more throughout the year. Its GitHub page now hosts 24 public repositories, which is up from six when we began writing this series last summer.
This is real, tangible progress and the CBO deserves credit for it. The new repositories span a wide range of the agency’s work: a conventional tariff analysis model, a permitting model, a state unemployment rate model, discount factor calculations for federal loan programs, Medicare Advantage data cleaning routines, a Social Security trust funds model, a Markov-switching macrosimulation model, and several others. The variety of models posted, covering defense, healthcare, macroeconomics, and tax policy, suggests a renewed institutional commitment to transparency rather than a token gesture.
Lawmakers Should Demand More
We don’t want to look a gift horse in the mouth, but we’d be remiss not to note that these models are components of a much larger scoring infrastructure. The question left unanswered is how these repositories connect to produce the scores that shape legislation. The conventional tariff model, for example, is valuable on its own. But understanding how its outputs feed into CBOs broader budget projections and how those projections interact with macroeconomic feedback effects is important in evaluating future scores. What we have now is transparency on parts of the pipeline, a code repository here, a working paper there, but not yet the ability to follow an estimate from these parts to the final score.
Additionally, JCT remains a black box. CBO’s progress makes the JCT’s silence all the more conspicuous. The JCT produces revenue estimates that CBO incorporates into its budget scores. These estimates can drive the reconciliation math that determines whether major legislation can pass. Until the JCT follows CBO’s lead, the scoring process will remain in the shadows.
In our reform proposals, we argued that the CBO Show Your Work Act, introduced regularly since 2017, should finally be passed and expanded to cover the JCT. Lawmakers should demand end-to-end transparency, and that is what the Act would require and what Congress should still demand.
We have been critical of CBO and JCT in this series because we believe they play an essential role in fiscal restraint, and that role requires broad public confidence. Last month’s release represented a meaningful step toward restoring that trust, but it is only a step. We hope the scorekeepers continue on this path, and that Congress keeps pressing them to show their work.