Up
Next: Why Humans?

Disclaimer 1: These are my personal opinions. I emphatically do NOT speak for my employer (who probably wishes to remain anonymous--I haven't asked) or for anyone else, real or ficticious, living or long since dead; human, cetacean, or space alien; animal, vegetable, mineral or ethereal.

Disclaimer 2: I'm an aerospace stress analyst, not an economist or accountant. Take what I say about cost estimation with a grain of salt.


Why Are
LAUNCH COSTS
So High?

Peter A. Taylor
murmur@ghg.net
March 2000


Why do space launches cost so much? Specifically, why are the costs so far out of line with the cost of seemingly comparable airplane operations? Fuel is about 15% of the operations and maintenance (O&M) cost of a typical military airplane, and 38% for commercial aircraft, according to Aircraft Design: A Conceptual Approach, 1992, by Daniel P. Raymer. Space launches should be more energy intensive than airplane flights, so one would expect that propellants would be a larger fraction of the total operations cost for a launch industry that was as mature as the airline industry. Why are the non-fuel costs orders of magnitude higher for rockets?

If you don't think the question is important, please take a look at the companion essay, Where (in space) do we go from here? or Why do we have a manned space program? Specifically, I refer you to the discussion towards the bottom, of the economics of an Aerobraking Orbit Transfer Vehicle (AOTV, a "space tug"). I worked on this project, and thinking about it got me seriously upset about the economics of the manned space program in general, and launch vehicles in particular.

Why are space launches so expensive? I present here a list of answers I have heard. I will discuss each of these in turn. At the end I offer my own conclusions.



18 Candidate Explanations for High Launch Costs:

  1. # "It just does!"
  2. # Propellant costs.
  3. # "Energy costs."
  4. # Lack of reusability.
  5. # Lack of full reusability.
  6. # Conflicting vehicle requirements.
  7. # Learning curve; not enough generations of launch vehicles built.
  8. # Lack of adequate investment (solid boosters for Shuttle).
  9. # Politics I: Parasite control.
  10. # Politics II: Obstructionism.
  11. # Lack of technology I: Dry weight too high.
  12. # Lack of technology II: "Complexity" of multiple stages (need SSTO).
  13. # Living on the edge.
  14. # "Ground handling and launch operations" (need "ship and shoot" payloads).
  15. # Lack of intact abort capability.
  16. # Critical differences between rockets and airplanes (ie. thermal problems).
  17. # Lack of a clear sense of direction.
  18. # Economies of scale.


  1. "It just does!"
This is not an answer. It is the stuff of which light bulb jokes are made. It suggests that the launch industry is very mature and that no further significant improvements are possible, which is highly implausible based on the airplane analogy. Instead of dealing with the problem of high launch costs, we are urged to just throw money at it and go on to other things. Unfortunately, at current launch prices, it is hard to identify many "other things" that are worth what they cost. Under the circumstances, I find this sort of complacency very hard to understand.



  1. Propellant costs.
This is silly. If it takes 100 lbs of kerosene and LOX to put a pound of payload in orbit, with propellants on the order of $0.20/lb, this only accounts for $20/lb. of payload, out of something like $5,000-10,000/lb. That was the way I initially thought about the cost of the space program when I was 10 years old, so I tend to describe this as a ten-year old's view of launch costs.



  1. "Energy costs."
If "energy costs" refers to the cost of energy, $/lb., all of the energy in question comes from the chemical energy stored in the propellants, which we have just shown is not a significant cost factor. But if "energy costs" means energy requirements, in BTUs/lb., then how does this relate to dollars? Am I twice as dead if I fly into the side of a mountain at 2000 mph as I am at 1000 mph? Do I need twice as many pilots if my spacecraft is flying twice as fast? The "energy cost" doctrine depends for 99.6% of its plausibility on the belief that labor costs are an extremely strong function of energy requirements. But where is the basis for this belief? I have never heard of anyone even trying to make a quantitative estimate of how labor relates to energy. This "answer" is a cop-out. It sounds well grounded in physics, and conveniently deflects criticism of NASA. It presents the cause of high launch cost problem as being fundamentally unsolvable, and hence not worthy of further serious attention. This line of reasoning is often used as an argument for building a base on the Moon, where the "energy cost" in BTUs per lb. of a space launch is lower, but where the man-hours of skilled labor that account for the remaining 99.6% of Terrestrial launch costs will be astronomically more expensive for the forseeable future. Perhaps launching things from the Moon will be cheaper than from the Earth some day, but if so, it is likely to have more to do with the Moon not having an atmosphere than because of energy. In short, the "energy cost" explanation is an evasion. It doesn't tell me anything about money. Why does orbital speed cost so much? "Energy costs" doesn't answer the question unless we understand it in such a way that it is blatantly false. We are back to "It just does!"


energy and cost



  1. Lack of reusability.
This is much better than #2. Propellants may only cost a few cents per pound, but aircraft typically cost between $150 and $300 per pound (Raymer). If rockets are like very expensive airplanes, are expendable, and weigh empty several times what the payload weighs, it is easy to see how a launch could cost thousands of dollars per pound of payload. This was a plausible explanation for high launch costs until the Shuttle was built, and may still be plausible for many expendable vehicles.



  1. Lack of full reusability.
This is an attempt to modify #4 to make sense in the light of Shuttle experience. If the Shuttle External Tank (ET) costs $350 per pound (that figure was from the early '80s), and weighs more than the payload, that explains a significant fraction of the Shuttle's cost. But it is possible to build cheap cryogenic tankage, depending on how it fits in with the rest of the vehicle and how desperately critical the weight is. The liquid natural gas tanks used in LNG tankers cost something like $5 per pound (early '80s numbers), and the Atlas rocket propellant tanks cost $80 per pound (early '80s). There is surely some truth to explanations 4 and 5, but not enough to dismiss expendable concepts like the Big Dumb Booster and partially expendable "drop tank" concepts out of hand.



  1. Conflicting requirements: "Jack of all trades" philosophy.
Max Faget originally wanted a Shuttle with a 10,000 lb. payload capacity and minimal cross-range. The Air Force wanted 65,000 lbs. and large cross-range. It was supposed to carry stuff to a space station, then it was supposed to perform in lieu of a space station, and now it is carrying stuff to a space station. It was required to carry large, heavy cargo and a full load of people at the same time, safe enough for school teachers, and carry stuff both up and down. It was supposed to be inexpensive to develop and to operate, with the peak funding kept low. Now no one is happy.



  1. Learning curve: not enough generations of launch vehicles.
The idea here is that even if airplanes and rockets are similarly complicated, we are a whole lot better at building and flying airplanes because we have designed, built, flown, and crashed a whole lot more of them. This explains why the airplane fuel cost analogy doesn't apply to rockets. It also suggests a solution: build lots of cheap X-vehicles, build small launchers that fly often, and figure on spending lots of money on refining many little design details over many years, in small steps. Implicit in one version of this theory is a belief that rockets are complicated not because of any one reason, but because of a myriad of little details that have to be gotten right. Consequently, there is little hope for major cost breakthroughs. Another consequence is that the cost entries on a launch vehicle's balance sheet should be taken pretty much at face value; there is little point in trying to "read between the lines." A more optimistic version of this theory is that we need more diversity in our X-vehicles more than we need large numbers of them, and there is hope for a breakthrough.

Raymer reports (p. 502) that aircraft manufacturing man-hours typically follow learning curves between 75% and 80%. That is, doubling the number of a particular type of aircraft built reduces the time spent on the last one to 80% of the previous time. Although these curves are meant to apply to the construction of particular types of aircraft, one might hope that the state of the art of operating space launchers in general would improve with experience in an analogous way.

However, #John London (Reducing Space Mission Cost, Wertz and Larson, Figure 4-1, Historical Trends in Space Launch Costs) indicates that, while there was rapid progress in reducing launch costs in the late 50s and early 60s, since 1970 the costs have been roughly constant, between 40 and 100 man-years per ton of payload, depending largely on vehicle size.


80% learning curve



  1. Lack of adequate investment.
There may be a "chicken and egg" problem in the space launch business. It may not be possible to amortize the investment needed to build a dramatically cheaper launch vehicle without an enormous increase in the size of the launch services market. And a dramatic increase in the size of this market may not appear until decades after the prices have dropped by an order of magnitude. If this is true, a large investment may be required to lower costs dramatically, probably at public expense. However, this does not mean that you can solve an inefficiency problem by throwing money at it blindly--you might just make things worse. How would you spend the money? There is no shortage of launch vehicle technology "gimmicks" that would be expensive to develop (ie. #Single Stage To Orbit), but there seems to me to be a severe shortage of reasons for believing that most of these gimmicks would lower costs rather than raise costs, let alone that they would lower costs substantially.

The Shuttle was designed to use solid boosters and an expendable ET to minimize development cost, partly at the expense of operating cost. But current studies claim that switching to a liquid flyback booster can't be justified based on economics, but rather improved safety. If more money had been spent on developing the Shuttle, would that have made it an economic success, or just a more expensive flop? Until we have a convincing answer to the question, "How would you spend the money?" the lack of adequate investment is not a sufficient explanation.



  1. Politics I: Parasite control.
There is no question that the space program has what Larry Niven calls a "parasite control" problem, too many people whose services are not really required trying to use it as their meal ticket, and a political system that prizes their votes and responds to their lobbying. But we are not trying to explain why rocket costs are a little bit out of line with airplane costs, we are trying to explain why they are two orders of magnitude out of line. Does poor parasite control explain this big a discrepancy?

I don't think it does. Political scientist John Ambler has argued that popular democratic governments can do things efficiently if there is sufficient political will, and as an example invoked an oil company that was owned by the French government and operated very profitably under a Socialist administration, which viewed it as an important milk cow needed to support its social programs (a lecture I heard). The space program currently has no such political will. NASA hasn't had a clear sense of direction since 1969. Economist David Friedman suggests (Machinery of Freedom) that functions that are performed by government agencies typically cost a factor of two more than they cost in free markets. Friedman is a self-described libertarian anarchist, so this figure is surely not biased in favor of big government. Physicist Freeman Dyson (Infinite in All Directions) suggests that the politically driven waste in scientific programs depends in a strongly nonlinear way on the size of the program. But I am still thinking, "factor of 2," not "200."

Let's distinguish between development and operations costs (If we can! I have been told by an ex-boss whom I respect a great deal that the main reason the Shuttle is so expensive is that NASA refuses to "put the pencil down," and keeps adding or changing the requirements.). Our objective is to get the "marginal" cost of payload in LEO (the cost of the last little bit) down to where interesting things will happen. At that point, the development cost is a sunk cost, so even if the development cost was regrettably high, it is too late to back out, and so interesting things will still happen. If parasites double our operations costs, that means we don't achieve our objective, whereas if they double the development cost, we still get interesting things to happen; it just takes longer and costs more, as long as we keep working on it. So I want to focus on operations cost, and look at development parasites only if they threaten to create problems like conflicting requirements or inappropriate technology that will have serious operational repercussions. If the X-33 program is mostly waste, it is an infuriating setback, but it doesn't by itself explain why the Shuttle costs so much. On the other hand, the Space Shuttle Main Engine (SSME) is an example where Max Faget said we needed a "truck engine" but we got a "race car" engine instead that is very expensive to operate. This can reasonably be attributed to parasites.

Now let's look at competition. In a free market, if there are nine organizations with severe parasite problems and one without, I expect the parasite-infested organizations to soon change or go out of business. I imagine "soon" here as being on the order of a decade. Does this line of reasoning put an upper bound on the seriousness of the parasite problem? I think it does, at least for expendables. The launch industry is very fragmented and political, but I think there is enough competition, both domestic and international, that it would be clear if there were any organization that was head and shoulders above the others in efficiency. The Russians look good right now from a cost standpoint, but this is partly an artifact of the state of their economy. Ariane, Sea Launch, etc. are not that much cheaper than Atlas and Titan. Many of the organizations involved, Boeing, Lockheed, McDonnel Douglas, etc., are quite competant at building both economical airliners and a wide variety of sophisticated expendable military missiles. It is not plausible that all launch vehicle operators throughout the world should be roughly equally infested with an anomolously high level of parasites.

I note that none of the players in this market have particularly enviable records for reliability right now. Russia has dropped Protons on Kazakstan, Japan has given up on their H-2, Ariane is recovering from some recent failures, and it wasn't long ago that a $1B US spy satellite was lost on a Titan. It appears to me that the industry is at least somewhat competitive, and tries very hard to balance cost and reliability with expendables. If they have a vehicle with higher than 95% reliability, they cut their staff, and if reliability goes below 95%, they increase it. NASA is laughed at for having so many people supporting Shuttle launches, but they operate manned vehicles and need and get higher reliability, more like 99%. It appears that "the devil's in the details," and no one seems to have any spectacular advantage at being able to distinguish between "parasite control" (aka "better, faster, cheaper") and "cutting corners."

This argument isn't as strong for reusable launch vehicles (RLVs), because there is only one current example. But if a competant venture capitalist could find a competant aerospace contractor who thought they could build an RLV with operating costs that were in line with airline operations, what is to stop them from going into the launch business? At worst, they might have to go overseas, as Kistler (Australia), Beal (Sombrero Island), and Boeing (Sea Launch) are doing to escape some of the politics in the US. In order to explain high launch costs, the parasite control theory has to be combined with theories 4 (lack of reusability) and 8 (lack of adequate investment), and probably several others. It can at most be only one piece of the puzzle.

I should elaborate on the point about combining with theory 8. It may be that Boeing thinks they could develop an RLV with low operating costs, but that (A) it would cost a huge amount of money, (B) it could only capture a fraction of the total launch market, (C) the current market isn't all that huge, (D) it would take a decade or more for the market to expand to take advantage of low prices, (E) due to knowledge problems, the customers won't commit themselves to making the investments needed to take advantage of radically lower costs until after the prices have already fallen, (F) the results are uncertain, and (G) the innovations that would make their vehicle cheap would be easily observed and imitated. Factors E, F, and G suggest that there is enough of a "public good" problem in developing an RLV that we shouldn't be surprised if there is a "market failure." So this combination theory says that only RLVs can be cheap, the free market won't build them because of too large an element of public good, and governments don't build them because of parasites. We should also be alert to the possibilities that the customers are doing things that have a lot of public good in them, like quasi-religious public entertainment and medical research that can't easily be kept secret, or that the whole manned space program simply costs more than it's worth. But it seems to me that in any case, the parasite control theory by itself won't take us the whole way.



  1. Politics II: Obstructionism.
This theory is similar to the parasite control theory in that it doesn't do a credible job of explaining why all launch vehicles are expensive, but it is useful in explaining why there isn't more competition. My favorite version of this is that there is someone in the State Department who really doesn't like the idea of Bubba having his own ICBM, and is quietly but deliberately trying to discourage anyone from building any launch vehicles that aren't firmly controlled by the US government. There are certainly some turf battles between various Federal agencies, and a lot of the usual attempts by the relevant agencies to control everything they can without accepting any responsibility for anything. Another effect is paranoia or at least posturing about espionage or technology transfer. The resulting "red tape" certainly exists and certainly discourages groups like Sea Launch. Some NASA and Air Force officials undoubtedly regard private companies as unwanted and potentially very embarrassing competitors, and the large aerospace companies that are in the best position to develop launch vehicles privately are also dependent on govenment contracts for enough of their income to be vulnerable to blackmail, implicit or explicit.

Then there is the phenomenon of "displacement," familiar to conservative welfare reformers, wherein the existence of a government agency that is supposed to fix some problem tends to discourage volunteer efforts that would otherwise be working on the same problem. For a long time before the Clementine mission, Alan Binder was trying to raise money from private contributors and recruit workers for a volunteer version of Lunar Prospector. After Clementine, he was successful at getting government money. Before Clementine, when he approached potential contributors, he reported that they often responded that this was a great idea, but why isn't NASA doing it? I had heard libertarians claim that because of displacement, NASA was, on net, doing more harm than good, but I never took that claim seriously until I heard Binder talk of this experience.

Addendum: Andrew Beal stated in his 10-23-2000 press release regarding the Cease of Operations of Beal Aerospace, "There will never be a private launch industry as long as NASA and the U.S. Government choose and subsidize launch systems." However, I greet this announcement with a certain degree of skepticism. Beal invested how many years and how many millions of dollars in his launch company before discovering that the U.S. Government was funding competing schemes? A co-worker of mine suggests instead that he got egg on his face and is looking for a scapegoat.

Incidently, Max Faget complained about "pharisees" rigidly enforcing rules about things like the use of flammable materials in spacecraft without knowing why the rules were made, and when they can safely be relaxed. #Sellers and Milton argue on p. 70 of Reducing Space Mission Cost that too large a Quality Assurance (QA) bureaucracy can actually reduce quality, and it certainly raises costs. Political pressure after a high-profile failure such as the Challenger accident may increase the amount of QA paperwork even if it is already too high.

Again, like the parasite control theory, political obstruction may take us part of the way towards explaining why rockets are two orders of magnitude more expensive to operate than airplanes, but it doesn't take us the whole way.



  1. Lack of technology I: Dry weight too high. We should develop and use advanced technology and push it to the limit to reduce the dry weight.
Marshall Kaplan wrote a series of articles in Launchspace, starting in April 1999, that take what I describe as the "orthodox" view in the aerospace business that cost is largely a function of weight. Basically, the aerospace companies and think tanks have empirical Cost Estimating Relationships (CER) that relate cost to empty weight and various other parameters. Most of what I know I learned from reading a chapter in Aircraft Design: A Conceptual Approach, 1992, by Daniel P. Raymer. There is a different curve for each different class of aircraft (flying boats, fighters, military transports, etc.). Kaplan is mostly concerned with private companies wanting to amortize development costs, whereas I'm mostly interested in operating cost. But Raymer says on p. 512, regarding maintenence man-hours per flight hour, "MMH/FH is roughly proportional to weight because the parts count and systems complexity go up with weight."

On one level, these CERs are purely empirical, and one could argue that they do not depend on an underlying "story" in order to be valid. But I am reminded of Stephen Landsberg's criticism of econometrics, that if you don't understand the story behind the statistics, the fact that American football teams are statistically likely to punt the ball on fourth down would lead one to suggest that punting would be greatly reduced by changing the rules so that there were only three downs per possession. Also, even if these CERs give reasonable estimates for development costs for something like a Space Shuttle, they still don't say why the operations costs are so far out of line with airplanes.

Common sense tells us that cost is a function of weight, but it seems to me that common sense is talking out of both sides of its mouth here. Apart from having to buy more material, bigger cranes, etc. to build a bigger airplane, there is an underlying assumption which Raymer acknowledged that cost is driven largely by complexity, and that aircraft of the same class and weight will have the same complexity. Also, complexity goes up with weight--a machining operation to remove weight from a 2" thick piece of aluminum is more likely to be worth while than from a 1" thick piece.

But common sense also tells us that if you have a part with a certain weight that does a certain job, and you decide retroactively that it needs to weigh less, the cost will go up. Unless you did a bad job trading off weight and cost in the first place, the reason the part weighs as much as it does is probably because you thought it would be too expensive to take any more weight out. But an inexperienced cost estimator who doesn't have a firm grip on complexity may use the CERs to predict that a development program will be cheaper if you make it more technically challenging by lowering the allowable weight. Shortly after the Challenger accident, I was at a meeting to discuss crew escape systems where this seemed to happen, and I can only presume that the cost guy was trying to make a point that he didn't have enough information to answer the question he was being asked.

So complexity and cost increase with increasing weight except when they increase with decreasing weight. I want to use this sort of insight in order to compare the costs of reusables and expendables, and between such things as thick-wall monocoque (simple shell) propellant tanks for pressure-fed engines, thin-wall monocoque pressure stabilized tanks for pump-fed engines, and stiffened thin-wall designs like the Shuttle ET. Some people might suggest that dual-fuel engines would save weight and therefore save money. My reaction is that, at least as far as operating cost is concerned, I don't believe that an adequate data base exists to tell me what I want to know, I don't understand what the CERs purport to be telling me, and I don't think I would trust them even if I did understand them. I don't have a good enough handle on complexity, and I don't think anyone else does, either.



  1. Lack of technology II: Need Single Stage To Orbit (SSTO).
The idea here is that cost is a strong function of complexity, and the primary way to measure complexity is to count the number of stages. So we should develop and use advanced technology and push it to the limit in order to avoid staging rather than, for example, to reduce the dry weight of a staged vehicle.

My first problem with this theory is that, to me, rockets don't seem all that complicated compared to airplanes. In particular, staging a rocket seems rather simple compared with what an F-14 is able to do: catapult launch, variable geometry, all-weather, aerial refuelling, combat, separation of various external stores in non-predetermined order, and a tailhook landing at night on a moving ship at sea. In comparison, the complexity of staging a rocket seems more like that of changing planes in Dallas in order to fly from Houston to Toronto. It seems far from being absurdly complicated.

My second problem is that counting the number of stages seems an absurdly naive way of measuring complexity. How about the number of drawings needed to build it? How about the parts count? How about the number of dissimilar parts? How about the number of machining operations? How about dissimilar machining operations? How about the number of pages of stress analysis needed? How about the number of places where damage tolerance, corrosion resistance, and fatigue life are sacrificed in favor of materials that are just a tad lighter? What about the number of different ways that failures could result in death? We are told that one stage is simpler than two, but it seems very strange to me for someone to wring his hands about "complexity" without being worried that a rocket stage with performance requirements that are beyond anything that's ever been demonstrated might be more complicated than a rocket stage with only half the mission velocity requirement. It's like saying that one is less than two, therefore one Indy 500 race car is simpler and cheaper to operate than two touring motorcycles.

My third problem with this theory has to do with what Maxwell Hunter called "the standing army problem" in his 3-11-89 paper, "The SSX," later known as the DC-X. The overwhelming majority of the cost of a launch vehicle is ultimately attributable to labor, so any significant reduction in cost has to manifest itself as someone getting laid off. Whose salary could be eliminated if you could eliminate staging from a vehicle similar to the Space Shuttle? Unfortunately, the only information I have on this in anecdotal. The Orbiters typically spend on the order of 10 days in the Vehicle Assembly Building (VAB), where the External Tank (ET) and Solid Rocket Boosters (SRB) are mated, and 60 days in the Orbiter Processing Facility (OPF), where such things as corrosion and Thermal Protection System (TPS) tiles are repaired. The Orbital Maneuvering System (OMS) pods and their toxic propellants are handled elsewhere, in the Hypergolic Maintenance Facility (HMF). An SSTO vehicle will have more surface area of exotic TPS than a comparable two stage (TSTO) vehicle, and at least as much OMS work. SSTO would presumably spend less time in the VAB, but it would still need some crane work to install the payload. The empty SSTO would be heavier, but not enough to affect the transporter much. The launch pad might be simpler or more complicated, depending on whether the SSTO uses dual fuel. Launch operations at KSC might be simpler. Flight operations at JSC would be little affected. Neither SSTO nor TSTO would have SRBs or ET. I just don't see from this how SSTO is related to laying off a huge number of people.

Finally, I think I have a basis for skepticism in the way SSTO was sold. It was basically The Big Lie: the SSTO promoters equated SSTO with reusability. As Whitehead rather kindly puts it, "The concept of single stage to orbit is widely associated with the notion of reusable launch vehicles because the ultimate goal is to combine the two. In reality, designing for reusability initially requires robust hardware which is too heavy to reach orbit without staging.... Since reusability and SSTO are conflicting requirements, it is logical to first develop these capabilities independently." (AIAA 96-3108, Single Stage to Orbit Mass Budgets Derived from Propellant Density and Specific Impulse, John C. Whitehead, p.9). Raymer (Aircraft Design: A Conceptual Approach) mentions that maintainability requires access panels, which add weight, and that sometimes the ratio of access panel surface area to "total wetted area" is used as a figure of merit for maintainability. Some aircraft have figures of merit as high as 0.5. Raymer mentions the F-4 Phantom as a negative example, because in order to remove the radio, which was a common operation, one first had to remove the ejection seat, which was frequently damaged during removal.

Addendum: Oops. I forgot to mention the two most obvious weight impacts of reusability: the TPS and landing system (which for a VTOHL launch vehicle includes the wings). Roughly 15% of the Orbiter's empty weight is TPS. The wings, tail, and control surfaces are about 12% and the landing gear are about 5%.

Sometimes there are "dog and pony shows" where aerospace contractors show off their latest technology. At one of these, shortly before Lockheed was selected to build the X-33, I asked a Lockheed X-33 presenter how SSTO was going to save money. The conversation quickly moved to new TPS (heat shield) technology, built-in test equipment (BITE), payload isolation, and different management philosophy. Why can't a two-stage vehicle use the new TPS technology? Why can't a TSTO have BITE? Wouldn't a TSTO be less sensitive to the weight associated with payload isolation? Why can't a different management philosophy be applied to Shuttle? It seemed to me that he changed the subject a lot.



  1. Living on the edge.
Norm Augustine's Law XV states, "The last ten percent of performance generates one-third of the cost and two-thirds of the problems."

The basic idea here is that there are theoretical and practical limits to the performance of a rocket stage, and that the labor cost associated with developing and operating the stage depends strongly on how close to the limits it is required to operate. That last little bit of specific impulse may force me to increase the pressure in the combustion chamber dramatically. I may have to make the nozzle much more complicated. Removing that last little bit of weight may force me to use exotic materials that are hard to maintain and to tailor each inch of the structure for minimum weight rather than having a lot of common hardware. It's said that an airplane is a bunch of compromises flying in close formation. But if I am pushing the limits of performance, I can't afford to compromise much, and I will have to sacrifice robustness, maintainability, ruggedness, safety, and simplicity.

Common sense suggests that there is a certain amount of overhead for even a very low-performance rocket stage, but that it should not be costly to scale up a low-performance propulsion system slightly. On the other hand, costs may rise asymptotically near the limits of feasibility. As a thought experiment, consider a simple formula that has these characteristics:


cost = minimum_cost / ( 1 - ( delta_V / maximum_delta_V ) ^ 2 ) .


asymptotic curve


Suppose that maximum_delta_V was just a hair more than orbital speed, so that you could build an SSTO launch vehicle, but just barely. Would you want to?

This theory suggests developing "the big dumb booster," perhaps using pressure-fed rather than pump-fed engines. But if money is to be saved by designing launch vehicles with two or three low performance stages instead of one or two high performance stages, it raises the question of why current launch vehicles live so close to the edge. Is it due to some world-wide, quasi-religious, irrational worship of technology? Political pressure for glamor and prestige? Are rockets phallic symbols? This theory may be more helpful for explaining why the sole first generation reusable launch vehicle, the Space Shuttle, is expensive than it is for explaining why the relatively competitive expendables are.

Addendum: #John London (Reducing Space Mission Cost, Microcosm, p. 151) describes a large (1,112,000 Newton), pressure-fed engine built by TRW for under $100,000 in the late 1960s, inspired by the Lunar Module Descent Engine. Despite this, London reports (pp. 142-3) that Air Force studies in the 1980s continued to reject pressure-fed booster concepts on the grounds that cost was expected to be a strong function of weight. London argues (p. 144) that over-reliance on weight-based Cost Estimating Relationships (CER) is a serious problem in the aerospace industry in general. It also occurs to me that these CERs are vulnerable to deliberate misuse by interested parties whose positions would be jeopardized by any serious attempt at parasite control.

Addendum II: John Whitehead argues (AIAA 2000-3140, "Launch vehicle cost - a low tech analysis" and -3141, "Mass breakdown of the Saturn V") that an expendable "big dumb booster" using pressure-fed engines is incapable of significantly reducing launch costs because the performance (payload weight/vehicle empty weight) is just too sensitive to the propellant mass fraction of each stage. (Naive observers of the space launch industry tend to focus too much on specific impulse and not enough on propellant mass fraction as a measure of performance.) Figure 4 in his -3140 paper indicates that a modest amount of weight growth (reduction in propellant mass fraction from .9 to .77) in a 3-stage kerosene/oxygen rocket will cancel out all of the savings associated with an order of magnitude reduction in manufacturing cost (dollars per pound of empty vehicle weight). Launch vehicle designers are thus forced to sacrifice reliability, in the form of less redundancy and reduced factors of safety relative to airplanes, in order to achieve a reasonable payload mass. This argument reinforces the belief that expendable launch vehicles will never be cheap. Unfortunately, it also underscores the difficulty of building a reusable vehicle that is not prohibitively heavy.



  1. "Ground handling and launch operations" (need "ship and shoot" payloads).
I saw a memo from early 1986 that stated that propellant cost was insignificant, and that vehicle dry weight was secondary in importance to "ground handling and launch operations." Things that would help include built-in test equipment (BITE) and more automation in general during checkout, but the emphasis in this memo, at least through my biased eyes, seemed to be on reducing payload integration problems with the so-called "bolt-on" or "ship and shoot" payloads. Less sensitivity to weather conditions (high altitude winds that threaten to damage the wings during launch, and landing site visibility) would also help.

I am perhaps overly sensitive to payload integration problems, having been involved in too many "coupled loads analyses," a very non-airline like activity in which the dynamic (unsteady acceleration) loads that a payload designer needed before he made the design are calculated by combining a computer model of the Shuttle with one of the payload(s), after it has been built and tested. In other words, we take a "finite element" computer model (FEM) of the Orbiter, integrate FEM of all of the payloads that are manifested on a particular flight, and calculate all the structural loads specifically for that flight, perhaps several times if there are problems resulting in changes. You may find out at the last minute that your payload can't fly because it resonates with another payload to produce loads higher than what you designed for, or that the Orbiter is good for. There is at least one of these analyses per Shuttle flight, sometimes several. While the "coupled loads" costs may be hidden in "payload development" rather than seen as Shuttle operating costs, they are part of what makes it expensive and time consuming to put something in space. Coupled loads analyses for transport aircraft are normally limited to generic analyses when the aircraft are first developed, not something to do every flight. Part of the reason for this is that the vibration environment on rockets is dramatically worse that on airplanes. Part of it is "living on the edge" and not wanting to add weight. Part of it is also the way the Shuttle was designed, with the LOX tank on the "wrong" side of the payload bay from the main engines. More standardization among payloads would help.

Addendum: See my comments below regarding economies of scale (#18). It has also been noted that scrubbed launches are expensive, and suggested that money could be saved by only launching into orbits with very wide launch windows. An example would be due East launches into equatorial orbits, for which an equatorial launch site is always in the right plane. But Ariane launches from Kourou into geostationary orbit provide an example of this, and as noted above, they are only modestly less expensive than Titan or Altas launches from Florida.

Addendum II: Tom Sarafin (President, Instar Engineering and Consulting, Inc.) writes in Doing Things Right in Space Programs:
I am still looking for the launch-vehicle program that is seriously interested in breaking the mold and providing a better service to their customers with a standard mounting adapter that includes a tunable isolation system. The first company that succeeds in this endeavor will knock the socks off its competition as payload developers put more emphasis on the indirect costs of LV options.




  1. Lack of intact abort capability.
This theory is taken directly from Maxwell Hunter's 3-11-89 paper, "The SSX," (draft 2). Although Hunter also cites bureaucracy and various other problems, the point of this theory is that the size of the "standing army" is driven largely by the severity of the consequences of a failure, more than by the complexity of the systems involved. Essentially, the lack of "continuous intact abort capability" (safe abort during all phases of flight, including saving the payload) forces whomever is operating a launch vehicle to maintain an army of engineers to make sure that absolutely nothing can go wrong. Although "lack of intact abort capability" never appears as a line item on any budget, it indirectly drives the size of this "army," whose salaries in turn quite visibly drive launch costs.

While I don't buy a lot of other things Hunter says, I tend to buy the intact abort argument. This theory works for both expendables and reusables. Expendables inherently lack intact abort capability, and often carry frightfully expensive payloads, sometimes on the order of $1 billion, in addition to the launch cost in the tens or hundreds of millions. Similarly, the Space Shuttle Orbiters are essentially irreplacable nowadays, carry people on board, are politically sensitive, and are vulnerable to numerous potential catastrophic failures. If anything goes wrong with the SRBs during a Shuttle launch, the result is catastrophic. There is also a phase during flight when the only intact abort mode, Return To Launch Site (RTLS), requires a maneuver that has been said to require about eight miracles. By comparison, if an airliner has a serious failure during takeoff, the most likely result is that schedules will get messed up and the airline may have to put people up in a hotel overnight.

Based on this theory, it is quite conceivable that an expensive to replace reusable vehicle without good abort capabilities could be more expensive to operate than an expendable.

Addendum: Tom Sarafin points out in Doing Things Right in Space Programs that once a spacecraft has been launched, generally, if something goes wrong, we not only can't fix it, we can't even see what's wrong with it. This makes failures doubly (or triply) expensive.

Note also that the operations cost associated with providing good abort capabilities is a strong function of vehicle complexity (ie. living on the edge). Each flight controller or subsystem manager must not only know what can go wrong in his own system, and what can be done about it, he must also know how his system is affected by potential failures in other people's systems.



  1. Critical differences between rockets and airplanes (ie. thermal problems).
A lot of the labor involved in operating the Shuttle is repairing the Thermal Protection System (TPS) tiles. The thermal environment during Shuttle reentry really is unlike anything experienced by an airliner. There have been minor advances over the years, but major improvements seem to involve blunter vehicle shapes, lower weight/surface area, and redesign of the underlying vehicle structure.

Another unique problem for rockets is the oxidizers. There are no nice oxidizers. The most practical of them seems to be liquid oxygen. An air-breathing engine works with gasseous oxygen diluted with 80% nitrogen, at moderate pressures (ie. ten atmospheres). Liquid rocket engines use pure oxygen at pressures an order of magnitude higher, which eventually reaches very high temperatures (hopefully in places where it doesn't touch anything but fuel). Apart from the thermal stress and other problems of dealing with cryogens, it is difficult to keep oxygen from reacting with the walls of the plumbing it is flowing through. This makes it extremely difficult if not absolutely impossible to eliminate catastrophic failure modes, in which a failure in one engine destroys the entire propulsion system. In an airplane, if an engine shreds itself, you close the fuel valve. In a rocket, if an engine shreds itself, you notify the next of kin.

Addendum, 8-10-2001: To clarify, there are two problems here. One is to contain the blast and shrapnel from an explosion, and the other is to keep a fire from rapidly propagating up the oxygen line. The first problem is obviously solvable, but perhaps at a prohibitive price in terms of weight. It was not clear to me that the second is solvable, but I have since been informed that there are copper alloys such as monel whose fire resistance is good enough to keep a fire contained even in pure oxygen at moderate pressures. Again, the reason they are not used more extensively is their weight.

In comparison, several other differences I will mention are mere nuissances. One is the vibration environment. Good liquid rocket engines require good mixing in the combustion chamber, which requires extreme turbulence (or at least, that is my understanding of it). High pressure, turbulence, and combustion lead to the nasty random vibration loads that I complained about with regard to coupled loads analyses (along with large shocks like SRB ignition and nosegear "slapdown").

Another nuissance is acoustics. This seems an inevitable consequence of the shock waves that come from firing rockets with hypersonic exhaust velocities in the atmosphere, especially near the ground. It's just one more thing to make life difficult for spacecraft designers, especially for payloads with large surface areas and low mass.

Other nuissances include avionics that are more complicated than an airplane really needs, and the need for attitude control thrusters. (8-10-2001: In retrospect, this may be a very naive view of the complexity of launch vehicle avionics. This may be much more than a nuissance!)

Two more differences are more than nuissances, the impracticality of evolving orbital launchers from low speed cargo rockets, and the impracticality of incremental testing for many configurations. Airplane developers were able to learn by building economically useful low performance airplanes before moving on to faster and higher performance designs. But there is very little market for a reusable cargo rocket that doesn't make it into orbit. The process of testing a new vehicle is analogous. An airplane can do taxi tests, then low speed flights, and gradually "expand the envelope" to find out if there are problems before committing to performing at the limits it was designed for. The only rocket I know of that was able to do this was the DC-X. A VTOHL rocket that lifts off the pad is committed to making the transition to horizontal flight in order to land safely, and has no way to abort testing of the transition maneuver.



  1. Lack of a clear sense of direction.
#Wiley Larson writes (p. 18), "Managing the DoD's and NASA's large space programs is like trying to win a potato-sack race--with five people each having one leg in the sack while running in different directions, and 10 to 30 coaches and helpers are on the sidelines directing the people in the potato sack."

This theory sounds similar to #6, conflicting vehicle requirements, but they operate on different levels. #6 tries to explain "proximate" causes, why particular vehicles are expensive in terms of technical conflicts. #17 tries to explain ultimate or "root" causes on a human level, or why it is that the technical problems were not handled well by human decision makers. For example, if the Space Shuttle needed "truck engines," why were "race car engines" built for it? If the purpose of the Shuttle was to dramatically lower launch costs, why is there not a huge uproar within NASA that it failed to accomplish its mission? If the space station is being built to do research, why is its budget so out of proportion with the research budget? If the Shuttle is to be the primary logistics vehicle for the space station, why do they seem to compete with one another as research platforms more than they complement one another as related parts of a transportation infrastructure? Why does the space program seem to be fatally vulnerable to parasites when a French state-owned oil company is not? Why does NASA and much of Congress talk of private development of space launchers while other agencies seem to go out of their way to make things difficult for Sea Launch and Kistler? Why do volunteer organizations like the Mars Society criticize the Apollo program for "flags and footprints" missions while pushing for very similar Mars missions?

I am also struck by a sense that the space program is driven by cyclically repeating fads. There is a saying at NASA that you should never put a date on a briefing chart because it means having to do more work when you are asked to study the exact same thing and give the same briefing three years later. I feel like a fashion model watching hem lines rise and fall repeatedly over many years to no apparent purpose. Inflatables have been proposed before, but never pushed very hard. Why does a technically promising project like Transhab (today's inflatable) with obvious long term benefits flounder for want of a hard requirement for it in the short term? It mutates from purporting to support a Mars mission to wanting to be a space station module, or perhaps a hyperbaric chamber. It appears to be a solution looking for a problem, despite a very strong feeling among the people working on it that somebody, sooner or later, will want it, and probably sooner.

It seems to me that the lack of a clear sense of direction goes a long way towards explaining all of these things.



  1. Economies of scale.
The idea here is that if the number of space launches or, alternately, tons of cargo per month were as high as the number of airline flights or tons of cargo per month, the costs might be more directly comparable. This explanation overlaps to a considerable extent some of the others that we have already considered. Economies of scale result partly from increased specialization (fewer conflicting vehicle requirements, #6), the learning curve (#7) and the ability to amortize larger investments (#8). With a larger market, there might also be more political muscle available to overcome obstructionism (#10). Additional advantages can be wrung from specialization by trained personnel or specialized equipment that can be operated nearer to its full capacity. Also, in addition to the "learning curve" possibly representing how vehicle designers learn from previous designs, the learning curve definitely applies on a microscopic scale, as it was originally intended, reducing the number of man-hours needed to get repetitive work done on longer production runs.

There are certainly economies of scale in the sense that aerodynamic drag is relatively less important for larger launch vehicles. Furthermore, while complexity may increase with size, it doesn't increase as fast as payload capacity; if two vehicles fly equally often, the larger one should cost more per flight, but less per pound of payload. More demand for space launches would tend to make it easier for large launchers like Ariane 5 to find multiple customers who want their payloads delivered at similar times into similar orbits, and can thus share flights.

At one time it was often claimed that the Shuttle would reduce costs by flying frequently so as to take advantage of economies of scale. Nowadays it is generally thought the Shuttle can't fly more than about a dozen times per year without either compromising safety or greatly increasing staffing levels and perhaps capital investment. However, it would appear that economies of scale are part of the reason why Russian launch vehicles are relatively cheap--they've built a lot of them and lifted a lot of relatively heavy satellites with relatively short operational lifetimes. The Russian vehicles generally use larger numbers of engines than comparable American vehicles, but fewer different kinds. (See #London, and note the similarity to the Microcosm Scorpius design.) Maybe the relatively modular Russian expendable boosters really are dramatically cheaper than Western launchers in terms of man-years per ton of payload, and not just because of exchange rates.

But if the space launch industry were driven by economies of scale, that would tend to make it highly monopolistic, with the larger or busier firms much more efficient than the smaller or less busy ones. If this is so, why is there not a shake-out of smaller or less committed firms? There are several possible answers: Too many customers may have incompatible requirements for payload size, orbit, and schedules, creating a number of niche markets rather than one big monopoly. However, there seems to be competition within the niches. There is certainly political interference by many governments, such as subsidies for domestic launchers and import and "technology transfer" restrictions for foreign launchers, again resulting in a fragmented market. But again, there is still considerable competition within the niches. Various nations' military establishments all want to ensure that their access to space is not subject to foreign political risks, and the US military tries to avoid being too reliant on any one contractor. Cost is clearly a secondary issue to the military, but I note that the US military has gone for dissimilar designs (Atlas and Delta) rather than trying to take advantage of some economies of scale by having competing suppliers build and operate a common design under license.

If economies of scale were extremely pronounced, it would be cheaper to fly a single large vehicle than an assortment of large and small ones, even if the large one often flew half-empty. I also note that Boeing's Sea Launch program is using Ukrainian Zenit rockets while Boeing is simultaneously building more Delta vehicles.

One variation on the economies of scale argument is that my comment, "More standardization among payloads would help," under Launch Operations (#14), doesn't go nearly far enough. It is argued that very few launch missions are ever quite the same. For any two missions on the "same" launch vehicle, there is almost always a slightly different payload or set of payloads, different destination orbits, or differences in payload fairings or numbers and sizes of strap-on boosters. In this view, economies of scale could save a lot of money, but no one is currently taking advantage of them very effectively. The way to do this would be to design "cookie cutter" payloads and highly standardized launch mission profiles, using only a handful of initial "parking orbit" destinations. But like several of the other explanations we have considered, this one invites the question, "Then why doesn't someone do it that way?" The marginal cost of flying identical or near-identical payloads and mission profiles does not appear to drop spectacularly relative to dissimilar ones.

If the economies of scale in space launchers were overwhelming, there should be more consolidation in the launch business, and the Russians and French would have stronger advantages. The European Arianes aren't all that much cheaper than their American competitors. It's hard to say how much a Russian launch vehicle really costs, but if these vehicles are judged by their failure rates, they are again only somewhat better than their Western competitors.

There are certainly some economies of scale in the launch business, but we need to explain two orders of magnitude cost difference between airplanes and rockets. Economies of scale don't appear to be a sufficient explanation.


economics textbook sample curves

Are we near the left edge of this curve, or towards the right, near the optimum point where the two curves cross? (This curve is qualitatively similar to what you might find in an economics textbook, but the numbers are purely hypothetical.)



Now I will repeat my list of theories and try to sum up my conclusions. Remember that the original question was, "Why are the costs of space launches so far out of line with the costs of seemingly comparable airplane operations?"

  1. #^ "It just does!"
  2. #^ Propellant costs.
  3. #^ "Energy costs."
  4. #^ Lack of reusability.
  5. #^ Lack of full reusability.
  6. #^ Conflicting vehicle requirements.
  7. #^ Learning curve; not enough generations of launch vehicles built.
  8. #^ Lack of adequate investment (solid boosters for Shuttle).
  9. #^ Politics I: Parasite control.
  10. #^ Politics II: Obstructionism.
  11. #^ Lack of technology I: Dry weight too high.
  12. #^ Lack of technology II: "Complexity" of multiple stages (need SSTO).
  13. #^ Living on the edge.
  14. #^ "Ground handling and launch operations" (need "ship and shoot" payloads).
  15. #^ Lack of intact abort capability.
  16. #^ Critical differences between rockets and airplanes (ie. thermal problems).
  17. #^ Lack of a clear sense of direction.
  18. #^ Economies of scale.


I find it very difficult to take any of the first three answers seriously. "It just does" (1) clearly does not answer the question. "Propellant costs" (2) is an answer, but it is clearly wrong for any reasonable choice of propellants. "Energy costs" (3) either doesn't answer the question or is demonstrably wrong, depending on how it is interpreted.

The next two answers are a little better. "Lack of reusability" (4) may have a lot of truth in it in the sense that expendables may never be cheap, but if so, it still doesn't explain why a "big dumb booster" couldn't be made relatively cheap. It also fails to explain why the Shuttle is so expensive. "Lack of full reusability" (5) is much less plausible. While the Shuttle's expendable External Tank is very expensive, lots of other tanks, including "drop tanks" on airplanes, are not.

The next three answers, "Conflicting vehicle requirements" (6), not having gone far enough on the "Learning curve" (7), and "Lack of adequate investment" (8), are all somewhat helpful as partial explanations. They at least help explain why the Shuttle is so expensive, but none of them are very powerful explanations of high launch costs in general.

Similarly, the two classes of political answers, "Parasite control" (9) and "Obstructionism" (10), are helpful partial explanations, especially if they are combined with a belief that a considerable element of public good is involved in the conquest of space. Parasite control is mainly helpful in explaining the cost of the Shuttle and any proposed government-funded replacements for it. In my opinion, these explanations don't seem nearly powerful enough by themselves. However, they help, and most of the people I've talked to seem to believe that I seriously underestimate the importance of parasite control.

The next three answers are oriented towards technology: "Dry weight too high" (11), "Need Single Stage To Orbit (SSTO)" (12), and "Living on the edge" (13). Estimates of dry weight may be useful in predicting the development cost of certain kinds of vehicle, but it is a curve-fit story. This dry weight "story" doesn't explain why some points lie on a curve, it doesn't explain why space launch operations costs are so far off of the airplane curves, and it doesn't allow me to compare a "big dumb booster" against a lighter, more complicated design. The SSTO argument fails too many reality checks for me to take it seriously: complicated airplanes aren't that expensive, and staging isn't that complicated. "Living on the edge," like several of the previous answers, is helpful. It is especially helpful in explaining why certain vehicles such as the Shuttle are expensive. But in order to help much in explaining high costs throughout the entire worldwide space launch industry, it must be combined with an explanation for why such concepts as the "big dumb booster" are so generally out of favor, both politically and commercially. Again, this explanation is helpful, and perhaps I underestimate the mischief caused by the misuse of Cost Estimating Relationships (CERs).

Addendum: In retrospect, John Whitehead's "low tech analysis" answers several of the questions I raise in my discussions of theories 11 and 13. The big dumb booster (ie. an expendable vehicle with pressure-fed engines) is unlikely to save very much money because of the highly nonlinear relationship between the propellant mass fraction of a rocket stage and the resulting vehicle's payload capacity. In other words, we are living on the edge because we don't have a choice. This in turn answers the earlier question of why the Cost Estimating Relationships for rockets are so different from those of airplanes. It is economical for a rocket to live much closer to the edge than an airplane does.

The last five answers all seem to resist categorization.

"Ground handling and launch operations" (14), is a sort of grab-bag of minor problems, some of which can also be viewed in terms of "living on the edge" or "learning curve" problems. This explanation is only mildly helpful.

"Lack of continuous intact abort capability" (15), is my favorite. This to me is a plausible explanation of why neither expendables nor the Shuttle can be operated at anything like the costs of airliners. This may not be the full story, but it is a large piece of it.

"Critical differences between rockets and airplanes" (16) is helpful, especially in conjunction with intact abort (15). Although the Shuttle Orbiter's thermal protection (TPS) tiles are the first thing that comes to mind, a more serious problem may be keeping strong oxidizers at high pressure from reacting with metal parts in the main propulsion system. This makes engine failures much more dangerous in rockets than in airliners.

Painfully apparent to me is "Lack of a clear sense of direction" (17). Although this may not directly cause high costs, indirectly it goes a long way towards explaining why we haven't solved the problems that do directly cause high costs.

Finally, we have "Economies of scale" (18). This is only mildly helpful.



I consider the lack of continuous intact abort capability (15) to be the principal proximate cause of high launch costs. Several of the other answers are directly helpful, but none of them seem capable of explaining nearly as large a disparity in the costs of air travel and space travel. The learning curve (7) and lack of direction (17) are root causes for many of these problems. I also believe very strongly that lack of direction is largely responsible for our having made so little progress on the learning curve. These are the problems we should focus on in trying to reduce launch costs. Instead, much of the attention of space futurists has focussed on "advanced technology development." I regard this as a regrettable diversion.

Addendum (11-11-2001): In retrospect, the best root cause explanation of high launch costs may be the highly nonlinear relationship between propellant mass fraction and payload capacity. This makes it hard to build cheap expendables, or to build safe reusable vehicles (rugged ones with good abort capability and lots of redundancy) at any price. Consequently, I may have to retreat somewhat from my criticism of "advanced technology development," provided that this development is aimed at reducing empty weight, or to some extent for increasing specific impulse. My criticism of SSTO stands. I'm convinced that expendables are not the way to go, either. For reusables, I am still uncertain about the relative importance of weight vs. intact abort capabilities. I still want to see more X-vehicles aimed at reusable rocket operations research.




This essay has a companion: Where (in space) do we go from here? or Why do we have a manned space program?


If you found this article interesting, you may also be interested in the following:

  • Interviews with Max Faget and Guy Thibodaux
  • The Kistler K-1 two-stage reusable launch vehicle project
  • The Microcosm Scorpius modular expendable launch vehicle project (see their conference papers)
  • Reducing Space Mission Cost, by James R. Wertz and Wiley J. Larson, eds., Microcosm Press (Torrance, CA) and Kluwer Academic Publishers, 1996, ISBN 1-881883-05-1. Note ch. 4 especially, "Reducing Launch Cost," by John R. London III.


    Up
    Next: Why Humans?

    Comments?