How well does NASA spend its money?

When NASA says a space program will cost $X, how much does that program end up costing on average? The programs that get air time are the problematic ones that are way over budget or behind schedule. I’m thinking about the Space Launch System (SLS) and Orion, Mars Sample Return, etc.

NASA receives ~0.5% of US federal spending every year. Every dollar is carefully budgeted, but do those budgets mean anything? I’ve always assumed the answer is no, and the actual amount spent is almost unrelated from the initial estimate.

After hearing so much about the runaway costs of mega-programs, I was curious to look at the data.

Cost charts

I looked at 26 programs, mostly science missions, all launched after 2000. I added the other big, well-known missions for comparison. (For notes on the accounting assumptions, see the bottom.)

Here is the % over budget when you compare to the initial program cost estimate (NASA’s estimate of the total lifecycle costs when the program starts).

Positive values (red) mean the program exceeded the budget. Negative values (green) means the program came in under budget.

In the top right, we see the Mars Sample Return mission. By status quo, I mean the price tag of the latest architecture proposed by NASA. There are other proposals, and this program may be descoped or cut.

Notice anything missing?

Here is the same chart with 2 additional huge programs: the James Webb Space Telescope (JWST), and SLS and Orion, the key elements of NASA’s Artemis program. I am grouping SLS and Orion together since they’re both part of the larger Artemis program, which is NASA’s current moon exploration program.

These programs are so much more expensive than everything else that they might as well be paid for with monopoly money. This plot uses conservative estimates of cost. Some estimates put total costs for SLS / Orion higher, around $60B (as opposed to $40-50B).

While there are a few projects that are utterly out of control, most programs actually come close to their budgets. A few bad apples are spoiling the crop.

It really surprised me how many small and mid-sized programs came in under the initial lifecycle cost estimates, or were close to the estimates. But all this careful budgeting doesn’t matter much when you have 3 mega-programs decimate everything in their path and consume nearly unbounded amounts of money and time: JWST, Mars Sample Return, and SLS / Orion.

Of these 3 programs, only JWST has launched and is performing its intended purpose. JWST is a giant space telescope to succeed the Hubble Space Telescope. It images distant objects in space in the infrared spectrum. Much of its cost was due to engineering complexity. For example, one component needed to be folded 12 times to fit inside the payload fairing of the launch vehicle, and then be carefully unfolded in space.

The other 2 programs have not accomplished their mission yet, and likely never will in their current formulation. The goal of Mars Sample Return is to retrieve rock samples from the Martian surface and fly them home to Earth. These samples stashed by earlier rovers. SLS and Orion are notionally a large rocket and spacecraft, respectively, to go back to the moon. Over time, they have become everything to everybody.

Outside of these mega-programs, other projects look boring and routine. Programs that involve international coordination, or multiple NASA centers, or are designated as “flagship” missions, are more likely to be slightly over budget. No surprise there.

Sticking to budget might be less important than budget-setting in the first place.

It’s great that the Dawn program came in under the $472M estimated at the outset of the program. But was $472M the right number?

It’s been hard to cut the costs of science missions because their requirements tend to be one-off and custom. This begets expensive and exquisite designs. In other areas of space, SpaceX and Rocket Lab have aggressively driven down costs.

NASA could bring down the cost of science missions by leaning into a commercial mindset like they did with launch and LEO satellites. One promising example of how this model could be brought to deep space science missions is Rocket Lab’s recent proposal to take over Mars Sample Return end-to-end for $4B. The company has offered a milestone-based, firm-fixed-price structure.

Taxpayers should be jumping at this offer vs. the status quo, i.e., a program that has ballooned ~3x to $11B, with return-to-earth timelines slipping past 2040. Hopefully there will be more of this kind of proposal in the future! We could do a lot more science programs if each program cost 70% less!

What is NASA’s mission?

NASA’s stated mission is to “explore the unknown in air and space, innovate for the benefit of humanity, and inspire the world through discovery.” But it turns out that a big reason we are funding the biggest programs is to create jobs at the agency. SLS and Orion have held onto support throughout delays and budget overruns by using political engineering. They have splintered work as widely as possible across states and NASA centers.

Spending up to $4B per expendable SLS launch is a boondoggle when SpaceX’s Starship will carry 15,000 times the payload for 1/500th the cost (h/t Casey Handmer). Another commercial rocket–New Glenn–is coming online as well.

SLS and Orion are rightfully in the crosshairs of the new administration. Cancelling them would certainly free up resources to do a lot of science! There are also rumors that NASA will see significant staffing cuts next week. I hope we can spend more resources on space in the future, but focus them on the right things.

Appendix

Thanks to The Planetary Society and Casey Dreier for putting together the Science Budget Dataset I started from! It’s a great resource.

Here is the underlying data for the charts if you’re curious.

A lot of the work was just escalating and de-escalating annual costs to account for inflation. I used the NASA New Start Inflation Index (FY24).

It turns out accounting choices have a significant effect on whether programs look like they’re within budgets or not. This is mostly due to inflation, which matters a lot for 10 or 20-year programs! I tried to do like-for-like comparisons in original dollars (as of the program formulation year).

Moving cost baselines are another accounting issue. Many programs get “rebaselined” as time goes on, meaning the budget gets reset at a higher number. I used the original / earlier baseline in most cases.

There were also choices to make about which costs to include. Should unexpected ops costs be counted if a mission lasts longer than expected? I tried to include all costs.

For international missions, I tried to include US contributions only.