On the practice of benefit-cost analysis

04 Oct 2026

I’m taking today a bit slow, and thought I’d reflect a bit on some of my experiences in the Federal government. I’ve been meaning to write these up for a while, but a combination of wanting to give it time and being busy meant I’ve put it off. Today’s post is about my experiences with the practice of benefit-cost analysis (BCA).¹

---

Some context. From late 2023 to early 2025, I served as an economist in the Office of Technology, Policy, and Strategy (OTPS) at NASA. OTPS was an incarnation of the HQ strategy shop, a spiritual successor to offices like PA&E. Similar functions exist elsewhere in the government, e.g., the Office of Net Assessment and CAPE. One of the functions of the NASA strategy shop was to support the Administrator with independent assessments of things that different parts of the agency disagree about. You see, like any large bureaucracy, NASA is a collection of distinct entities facing different incentives and constraints, and Administrators past had found it useful to have a shop that could (say) measure the size and composition of the market for private space station services without having to go through the office responsible for putting the private space stations program together.

One of the things I worked on at OTPS was benefit-cost analysis of space sustainability investments, specifically relating to risks from orbital debris in low-Earth orbit. When I arrived, OTPS had published one report on the topic and was working on another; I was to lead the third phase.² Given that NASA’s budget is limited while its mission grows seemingly without bound, there was interest in finding out how to reduce the most risk at the least cost. Basically, you can reduce risk by removing objects from orbit (“remediation”), by better tracking objects so satellites can dodge collisions (“tracking”), or by shielding satellites so they’re better protected (“mitigation”). There are multiple ways to do each type of action, and trade-offs among the different approaches. The phase one report focused on remediation, and the phase two report on remediation, mitigation, and tracking as separate investments. In phase three, we were looking at how to design a portfolio of remediation, mitigation, and tracking investments to efficiently reduce risk in orbit.³

---

Textbooks teach that one of the advantages of BCA is that you can bring multiple quantities of interest together into a single number, and that dollars are a useful conversion factor for this purpose. This is pretty true in my experience. Before the first phase of BCA that OTPS did, discussion of space debris risks was complicated by all the different types of orbital debris. Small, medium, large; tracked vs not; this orbit vs that orbit; batteries passivated vs not; orbital debris is quite heterogeneous. That’s all well and good for research, but when you need to start allocating limited funds to competing uses, it’s helpful to have a common measuring stick and language so you can at least understand the trade-offs you’re dealing with. Dollar-valued risks serve that purpose, and the phase one report made that case well.

A popular conception of BCA is that in doing it you find the action with the biggest net benefits, and that’s what must be chosen.⁴ This picture breaks down, at least at NASA.⁵ Remember how I said that NASA is a big bureaucracy with many entities facing different incentives and constraints? Well, these entities had different perspectives on what mattered, how to count it, and what that meant for budget allocations. For example, an office that operated spacecraft may have been wary of proposals that would make their missions more expensive; an office that was already investing in a particular approach to risk reduction may have been wary of proposals that would downrank the investments they were making; and an office that focused on purchasing capabilities from industry may have been wary of proposals that would limit their options. Even as everybody cared about space sustainability, their specific situations meant they had different perspectives on what “efficient investments in space sustainability” looked like.

One of the useful functions of BCA, then, was to create a kind of forum for people across the agency to discuss different conceptions of value and what ought to be quantified. For example, some in the agency may have disagreed with how we modeled atmospheric drag or the effects of debris smaller than 10 cm. Some might have wanted to consider how well the systems might inspire people, or enable other missions to inspire people. Others, perhaps in industry, may have found our cost or performance assumptions too pessimistic. Part of the work in doing the BCA was to engage with these people, learn about their considerations, and find ways to reflect them in the analysis or else to clearly indicate why it wasn’t reflected and how to think about that omission. There was never a sense that the numbers we produced were the final word, that they didn’t involve value judgements, or that we quantified everything. Our job was to develop the most credible numbers we could, and a large part of credibility is socially constructed. Credibility therefore couldn’t emerge as a function of the numbers and estimation procedures alone; the whole thing had to be tested and challenged by others who were themselves seen as credible and still remain standing. Put differently, the value of the BCA wasn’t so much the numbers (though they were obviously valuable) as much as it was the consultation, debate, and consensus-building that went into producing the numbers.

---

To be clear, there was a lot of technical work involved. We developed some pretty cool simulation and estimation methods to model hypothetical systems like ground-based lasers to zap debris as well as existing capabilities like better spacecraft shielding. We reviewed a lot of technical literature and spoke to experts to get reasonable parameter estimates, including discount rates. We even did some theorem-proving to establish that our conclusions weren’t driven by some modeling choices. But, with the benefit of hindsight, I think we could have done some of those things differently and gotten to a similar desired end state. The work of bringing people together and facilitating conversations around shared values in a common language, on the other hand, was non-negotiable. The quantification had to be solid not because the numbers would be used mechanically to make a decision, but because others with different perspectives would push hard against our numbers and we had to convince them the numbers were reasonable. There were technical folks in just about every party involved, and sloppy work wouldn’t have passed muster. But in practice the point of the exercise was less to get a single number or set of numbers that dictated action than to develop a shared set of assumptions, understandings, and estimates that produced better discussions and decisions about what to do with limited resources. To quote Arrow:

Although formal benefit-cost analysis should not be viewed as either necessary or sufficient for designing sensible public policy, it can provide an exceptionally useful framework for consistently organizing disparate information, and in this way, it can greatly improve the process and, hence, the outcome of policy analysis.

Before my stint as a BCA practitioner, I don’t think I really appreciated what BCA does to the process of decision-making in large bureaucracies as much as the outputs it produces. Now, it seems perhaps the most important part of the whole deal. I am, somehow, an even bigger fan of BCA now than I was then.