When we last left Rocketman, he was accusing Gregg Maryniak of comparing launch vehicles to submersibles, an accusation that, knowing Gregg, I found quite unlikely.

He has since had an email discussion with Gregg, and clarified the issue. I found this little bit of Gregg's response interesting, because it wasn't something to which I'd previously (or at least recently) given much explicit thought.

Bottom line is that space stuff costs perhaps 500 times as much to develop historically as (very challenging) undersea stuff. Why? It may be largely because of the expectation that it should (based on the history of governments racing each other without regard to normal engineering cost contraints.) When I speak to big audiences of traditional government space program engineers and program managers I usually pray that one of them will ask me: "You mean to say that mere expectations can be cost drivers--ridiculous"...to which I say, I have two words for them...."stock market."

Is Gregg right? Is space hardware and operations expensive because we expect it to be?

There's actually quite a bit of evidence that it is.

Most space contracts, particularly government contracts are cost plus fixed fee. This means that the contractors are reimbursed for the actual costs of executing the contract, as reported by them, plus some amount for profit (typically a few percent of the contract value). This is because high-technology research and development is recognized to be high risk--that the schedule might slip, or the costs be greater than originally estimated, and few if any private companies
are willing to absorb those costs, and NASA knows that none would bid on any other basis.

The problem with this, of course, is that it skews incentives in a way that are bad for the taxpayer (though not necessarily for the true constituencies of the space program--the contractors themselves, and the congresspeople in whose districts the contractors employ people). Perversely, the more they spend, the more they earn. There are occasionally attempts to mitigate this by putting in bonuses for hitting cost targets, and penalties for missing schedules or overrunning the budget, but they're largely ineffective, at least judging by the space station program.

But there's a more pernicious result of this, that's less often considered. In order for NASA to project the cost of the contract, they have to have a way of estimating the costs, even if it's something that may have never been done before. The way they (and the contractors) typically do this is called parametric cost analysis. They have cost models that are built up by examining many past programs, and incorporating the cost and schedule data from those programs. The models might use factors such as complexity (which is hard to measure), weight, technology level, and so on. The hope is that a good cost estimator can come up with a valid estimate for the program cost and schedule, based on similar efforts that have been performed in the past.

One problem with this is that it's more art than science, and heavily dependent on the assumptions that the modeler uses. Another problem, of course, is that reinforces notions of how expensive things will be, because by definition, it's based on how expensive things were in the past. It doesn't provide any way to model true innovation. In addition, because almost all of the experience comes from government programs, the data base for private space activities is very sparse, so they don't have any way of modeling that with any degree of credibility. And it turns even government programs, given the right team and incentives, can beat the estimates. As an example, consider the DC-X program:

Prior to letting the DC-X contract our program office conducted a cost estimating study. We used three models, one developed internally, one used by the US Air Force and one from NASA. The results were that our cost estimate based on the rapid program assumptions I described earlier and projected a cost between $60 and $70 million, the Air Force model using standard aerospace procurement practices produced an estimate of $365 million, the NASA model based on highly technology development based shuttle program experience projected the program would cost over $600 million. The actual DC-X program cost through the first test series came in around $65 million.

In other words, they beat the conventional Air Force costing model by a factor of more than five, and its NASA equivalent by almost an order of magnitude, or factor of ten.

Sadly, here's the process (slightly oversimplified). NASA comes up with a program idea. They come up with a cost estimate for it. They request a budget. If Congress authorizes it, they put out a procurement for that budget target. The contractors write their proposals, and then come up with their own cost estimates that magically, and almost invariably turn out to be close to what NASA has money for. And thus the expensive game is perpetuated.

But as one more example of how such estimates and quotes can be voodoo, let me relate a story (possibly apocryphal, but it's certainly believable to anyone with experience in the business) that was told to me by a program manager from the seventies. In the process of submitting a proposal, a small, almost insignificant typo found its way into the final version as delivered to the customer. It was a decimal point, misplaced one place to the right, resulting in a bid for that part of the program ten times too high, relative to the contractor's internal estimate.

The contractor was downselected for a Best And Final Offer, which is an opportunity to negotiate a little bit. The contractor fully expected to be raked over the coals for their outrageously high bid (I think that it was something like ten million dollars, when it should have been one), and they weren't disappointed. The NASA contracting officer excoriated them, calling them crooks and cheats, and other names not mentionable in a family web site, and finally finished up his lecture with the words, "...and we're not going to give you a dime over nine million!"

And of course, the outraged response from the contractors' representatives (as they sighed with relief) was, "But we can't do it for that!"

As someone who personally witnessed a Lockmart rep give the details on their X-33 proposal, where "Subcontractors in 38 states and 120-something congressional districts" was a bulleted item under the heading of "Technical Specifications," this story is completely believable to me.

(This was at the Space Access conference, shortly before the contract was awarded. I still remember how the room groaned when they saw that line.)