You can't contract out the compass

When government no longer understands its own systems, it no longer controls policymaking.

This is a lightly modified excerpt from The how we need now: A capacity agenda for 2025, by Jen Pahlka and Andrew Greenway.

Imagine American policymaking as a car. Government is ultimately responsible for getting the vehicle to its agreed-upon destination, but we’ve outsourced too much of the work of navigating us there. While government will always need to purchase the fuel that propels the vehicle and hire out for maintenance, the directions that will get us to the destination need to come from within.

If you’ve worked in or with government, you’ve probably heard the term “one throat to choke.” This refers to the need to have one prime contractor responsible for anything from the development of an IT system to the services a program is charged with delivering, and as the phrase suggests, to be able to blame them when things go wrong. The phrase is indicative of what has too often become a profound abdication of responsibility on the part of government. Well-intentioned attempts to ensure that government hires out for tasks that the private sector can and will do better have spiraled into a harmful ideology that assumes private contractors are better than government at pretty much everything and government should do as little as possible, of any kind of work, in-house. The reality is that if government is on the hook for the outcomes, which it should be if the work is funded by taxpayer dollars, public servants must make the critical calls, even if contractors may do some or even much of the work. And that requires that the competencies and capacity to make those calls — to navigate the vehicle to its destination — reside within government.

In 1966, the White House Office of Management and Budget released circular A-76, which built on previous policies stating that the federal government “will not start or carry on any commercial activity to provide a service or product for its own use if such service or product can be procured from private enterprise.” The specific methods A-76 prescribed are no longer relevant; use of A-76 was prohibited first by the Defense Department and later across government when the press and public officials blamed shocking conditions at Walter Reed Medical Center on the process by which the Pentagon had competed and awarded the contracts to run the center. But the memo is important because it formalized the distinction between functions that are “commercial” and those that are “inherently governmental.” That dividing line between work that must be done by government employees and work that should always be outsourced has been hotly contested before and since the memo, and it desperately needs revisiting today.

This debate is too often framed as a binary: Either you believe that government is bad at most functions and should outsource everything it can, or you believe outsourcing to the private sector (and the social sector, the abuses of which have gotten far more attention in recent years) is counter to public-sector values and want government staff to do it all. (Public sector unions of course have their own reasons for preferring insourcing, and it can be a surprise to leaders trying to, for instance, process a ten-fold increase in unemployment insurance applications that hired help is not able to touch the claims because legal provisions make those the sole domain of what’s called “merit staff” with many years of tenure.) But this is an unhelpfully reductive view of a complex issue. A more helpful debate would return to the core question of A-76 but return different answers, answers updated for the modern era.

For starters, the kinds of work that government contracts out for have changed. A-76 defined digital work broadly as commercial, and that thinking has pervaded government even as A-76 itself was rolled back. Digital is the domain probably most prone to the claim that government can never do it well. But an institution that doesn’t understand the digital domain is a very poor purchaser of digital products and services, as evidenced by the government’s repeated failures in technology, from the Pentagon’s struggles to even equip their staff with working computers to the Department of Education’s debacle with the FAFSA financial-aid form last year to the Commerce Department’s failed financial system that stopped tens of millions of dollars’ worth of invoices and reimbursements from being paid and shut down critical services in sub-agencies such as the National Weather Service.

These failures are the result of a system that tries to buy software design and development services as if they were pencils or cars. It’s not just that we evaluate bids on inappropriate criteria, it’s that we’re not soliciting for the right deliverables. We lack the internal core competence to outsource well. And we can never fully outsource the development of the bespoke software that manages government’s unique missions, because the software is so intricately tied to operations. If public servants aren’t fundamentally in charge of what’s being built, they won’t understand how their departments function.

In the cases above, internal core competence would be pretty cheap. Proper leadership can often be established with a relatively small internal team. Given that contracts for development of government software frequently cost hundreds of millions of dollars, failing to invest in a modest internal team with the appropriate skills and authorities is extraordinarily penny-wise and pound-foolish.

The folly of skimping on internal core competence is well established in many domains. One paper by Zachary Liscow of Yale finds a remarkable correlation between the staffing levels of state departments of transportation and the per-mile construction cost of highways: Increasing employment by one person per thousand in a population reduces costs by 26 percent. When there is insufficient staff to manage these projects, their work is outsourced to consultants who often lack the continuity, context, or incentives to perform the work properly. There is ample evidence across a wide variety of government functions that contractors cost more than government staff for the same work, and that outsourcing at those higher rates causes the very people the government has trained and invested in to leave for higher-paying jobs in the private sector. In 2011, former CIA Assistant Director Mark Lowenthal told Congress, “It’s the least experienced analytical staff since 1947, and this demographic trend will play out in years to come,” regarding the brain drain to the private sector.

Over-reliance on contracting exacerbates all the other dysfunctions of the modern bureaucracy. In an environment of procedural bloat, contracting adds even more procedures. It adds even more layers in the hierarchy, meaning the cascade of rigidity — in which broad mandates become ever more constricting as they move down the hierarchy — has even more steps through which the intent of policy can be perverted. It adds even more temporal, organizational, cultural, and structural distance between policy and implementation teams, making it far harder to engage in test-and-learn cycles. Software that’s contracted out often can’t be modified by government staff, making it hard to react when needs change. Top-performing teams at the Department of Veterans Affairs insist on having enough in-house staff on any given project that if a change needs to be made quickly, one of their engineers can do it. The norm, however, is to have to negotiate a change order or task order or, worse, begin a new procurement. This results in enormous cost and the loss of strategic flexibility and responsiveness. Government should not be beholden to suppliers, no matter how well known their brand or how global their reach, if it means that agencies lose the ability to configure and adapt systems for constantly changing circumstances. If a government is locked into a multi-year contract and a technology we can't change, then the whole idea of test-and-learn is a non-starter.

A system that can only see the need for leaders with policy-formulating expertise leaves enormous gaps in agencies’ capabilities, as we saw in the FAFSA debacle.

Having internal competencies is sometimes seen as risky, because it’s harder to argue you have that “one throat to choke.” Vendors know that absorbing blame for failures is part of what they are being hired to do, and they don’t mind because they know that they can (and will) continue to win contracts despite a history of failure, even very public failure, due to the nature of contracting rules. CGI Federal, famous for bungling the development of healthcare.gov, saw a miniscule drop in its valuation after it spent months in the headlines for the debacle. Since then, the stock price of its parent company CGI Inc. has nearly quadrupled, from roughly $30 a share to $114. Among the many government contracts the company has since won was a bid to work on the new FAFSA (albeit not as the prime contractor), a project that went as poorly as healthcare.gov.

The problems with FAFSA bring us back to our original question: What should be inherently governmental in today’s world? There is a lot wrong with how we contract with the private sector, to be sure, but changing those procurement rules will only help so much if the core competencies government hires for and values in leadership are limited to policy-formulating expertise. The team at Federal Student Aid who oversaw the FAFSA rollout were not necessarily bad leaders or derelict in their duties. They may simply have been political appointees in a system that can only see the need for leaders with their particular skill set, leaving enormous gaps in agencies’ capabilities. They came into FSA with a limited set of positions they could hire for, almost all defined for similar skill sets. The organizational structure they inherited, with all that entails (open billets, etc.) also assumed a policy-oriented leadership, with extremely limited affordances for operational leadership. Operational competencies were assumed to be contracted for, so the competency that should have been operational was instead represented by the procurement and contracting staff. Clearly, these procurement and contracting teams could have chosen vendors with more appropriate skill sets, as evidenced by the outcome, but contracting expertise is a poor substitute for in-house operational skills. We need a fundamental relook at the core competencies agencies hire for.

Rather than debating outsourcing v. insourcing on the basis of abstract theories about what is “inherently governmental,” we should ask: What in-house competencies and capacities allow government to be able to outsource well? This, in turn, is critical to a larger question: how might government develop a far healthier contracting ecosystem, one in which companies win contracts because they are good at what they do, rather than good at playing a highly specialized contracting game?

Read the full report.

---

Andrew Greenway is a co-founder of Public Digital and a former senior UK public servant who was one of the founders of Government Digital Service, a model of institutional change copied by many governments worldwide, including the U.S. He also co-wrote the influential Radical How report on mission-driven government, published in 2024.

Jennifer Pahlka is the author of Recoding America, a senior fellow at the Niskanen Center and the Federation of American Scientists, and a senior advisor to the Abundance Network. She served as deputy chief technology officer of the United States under President Obama and a member of the Defense Innovation Board under both Obama and Trump, and previously founded Code for America.

---