The new Department of Government Efficiency (DOGE) is a promising adjunct to the second Trump administration that proposes to identify and root out billions or even trillions of dollars of federal waste, fraud, and abuse. To properly manage the problem of government inefficiency, it is best to measure it with big data on federal spending. But as those of us who have been studying wasteful federal programs know, getting the necessary data has become more and more difficult.
The primary tool available to outside analysts is the Freedom of Information Act (FOIA), a 1966 law that allows members of the public to obtain government documents. Sadly, the heydays of executive branch compliance with FOIA are long behind us. Today, departments and agencies rarely respond to complex requests within the twenty working days specified by the law and often have to be taken to court to disclose anything at all. FOIA includes a variety of exceptions that bureaucrats interpret liberally, claiming the right to heavily redact documents or withhold them.
While FOIA abuses related to federal mismanagement of the COVID-19 pandemic are now well known, there are many low-profile cases of unjustified secrecy. For example, any government that receives more than $750,000 of federal funds in a given year must provide the federal government with audited financial statements. In 2016, the Obama administration made all of these audits publicly available, but Native American tribes were exempted. As a result, outside researchers cannot assess wasteful spending directed to tribal governments.
Medicaid and Medicare are two very large programs that generate reams of data unavailable to the public. The Center for Health and Human Services has granular data detailing medical visits by public healthcare program beneficiaries, including amounts paid. These data can be linked to patient health outcomes (on an anonymized basis to avoid disclosing personal medical information), enabling researchers to assess the cost-effectiveness of various treatments and medical providers. However, this information is not accessible by FOIA and is instead made available only to a handful of academic researchers and corporate users who pay hefty subscription fees for the data.
Transit Capital Investment Grants are a major category of inefficient federal spending. For example, the federal government spent $1.55 billion to partially fund the construction of an elevated train line in the Honolulu suburbs. It opened several years late and only carries about 3,000 passengers per day. These riders could just as easily reach their destinations via express bus. Active federal transit grant programs are subject to a monthly review by an independent Project Management Oversight Contractor (PMOC).
However, the Federal Transit Administration (FTA) rarely publishes these PMOC reports. At the Cato Institute, we FOIAed these reports three months ago and are still waiting for them. If researchers could obtain and systematically analyze a corpus of PMOC report data, they may better understand the early warning signs that transit projects will run far above budget and behind schedule.
Executive branch officials have a not unreasonable explanation for the slowdown in FOIA compliance: understaffed FOIA departments deal with a deluge of often complex and sometimes poorly formed requests. But modern technology provides a couple of solutions to this problem.
First, agencies should be able to use artificial intelligence to supplement or even replace FOIA liaison staff. If AI has access to a department’s full corpus of potentially responsive documents and can interpret FOIA requests, it should be able to fulfill these requests with limited or no human intervention.
A second, much simpler technology solution obviates FOIA entirely. Departments, especially those that do not deal with large volumes of documents with national security implications or contain personally identifiable information (PII), should just automatically publish all the documents they produce or receive to the public cloud. There, non-government organizations can apply crawlers and large language models to make the document stash digestible to a wider audience. Even documents with PII can be published by default if there is satisfactory technology to redact them automatically.
Because DOGE’s success will rely on access to an enormous volume of federal information, the creation of this department presents an incredible opportunity to open up government processes to external oversight. To be fair, a heightened level of transparency will not only expose government waste but may also reveal government agencies that function relatively well.
But let the chips fall where they may: DOGE should act as a battering ram against the walls of federal secrecy.
