Propellant, in the form of liquid oxygen and kerosene, is about as expensive as milk. For a vehicle with 2% mass fraction, and assuming that the entire vehicle is fuel in order to place an upper bound on fuel costs, that's 49 parts fuel to 1 part payload, by mass. If fuel and oxidizer are roughly the density of water, then each gallon weighs less than 8 lbs. (Closer to 5 or 6 for kerosene, and about 9 or 10 for oxygen.) This gives us about 6 gallons of propellant for each pound of mass, for a cost of about $22.
The fuel cost to launch something into orbit is about $22/lb. The rest of the $1e3 to $1e4 per pound is engineering and paperwork. A typical aerospace part that has, for example, $700-$800 worth of actual materials and labor will have a verification and paper trail that costs about $30e3 to produce, due to the mission-critical nature of all the highly stressed and low-factor-of-safety parts. (Car parts have an FS in excess of 2.5, for example, and aerospace parts are typically 1.1-1.2.)
Many people have suggested as much. The issue then is finding a large enough volume of customers / missions to make use of frequent flights. I've also heard (tho I don't have a link ready) that past 50 - 100 flights per year on a given vehicle, some more exotic methods of flight into orbit (that have been impractical to date) become cost effective compared to expendables. Not a few studies have suggested reusable chemical rockets are cheaper than expendables, if you can fit at least 50 flights a year on them. I once read that number as around 70 a year for the Shuttle (possibly from Antonio Elias, maybe someone else). If you can manage hundreds, some of the really nifty / whacky launch concepts might become viable, like cannon launch, laser launch etc. etc.
(I think Antonio Elias, on NSF made the argument that Kistler's now defunct K1 (COTS contract, along side SpaceX) rocket would have been viable at nearly the lowest possible launch rate for a reuseable, far sooner than the Shuttle.)
The biggest cost is the machine. High precision and exotic materials are usually involved, and they're usually hand-crafted in small numbers, and require gigantic tooling and facilities. And there's lots of testing and QA: since you don't get it back, you can't really test fly it.
Imagine if you were flying people to China via airliner. One way. The majority of your ticket cost would not be in fuel. (A 747 is about $350M; fuel across the Pacific is on the order of $100K.)
Note that reusing your vehicle solves a lot of those problems. But it's hard to do. Imagine that your 747 can just barely make it all the way to China, even if it's only carrying First Class, and there's no gas stations there.
Volume solves the rest of the problem--but that means we need to figure out a lot to do in space that's worth the cost.
There's development costs (which can be substantial). There's manufacturing costs. There's fixed recurring operational costs (which get averaged over the flight rate). There's incremental operational costs. And then way down in the noise is the fuel cost.
For most rockets manufacturing and fixed recurring operational costs are the biggies. There's only so small you can shrink a production line and an operations team, and there's only so high you can push your flight rate, the combination dictates your per flight costs to a substantial degree.
For a vehicle like the Shuttle which was quasi-reusable the fixed recurring operations costs were dominant. It cost several billion dollars a year just to maintain the standing army of engineers and technicians plus the facilities needed to keep the Shuttle operating. And because the Shuttle's couldn't fly more than once or twice a year per orbiter due to the long processing time between flights that resulted in a very high per flight cost even though manufacturing and fuel costs were low.
For Oribital's Pegasus, about 25% (solid fuel) engines, and 25% support labour [1].
http://www.mitre.org/sites/default/files/pdf/kane_mls.pdf Page 7 (Rest of that paper is also an awesome read about the problem space. Although it was from relatively early in Falcon 1 development... Pegasus has gone from alone in its class, to having competition, back to being alone).
Operations and infrastructure are not cheap either. If you require a massive pad, buildings, cranes, transporters, cryogenic infractructure, clean rooms for payloads, areas for hazardous fuel filling... And lots of people on the workforce with very specific skillsets.
And the kicker: launch only a handful of times per year so all that cost must be put into those launches.
Look at a modern airfield gate, all the activity when the plane rolls to it. See all those special vehicles, all the personnel. Some are putting the wheel chocks, some are waving guiding lights, someone connects the electricity line and then come the baggage handlers, the cleaners and the food people. The airplane requires more care on land than things like buses or trains for example. Or look at the people operating the security infrastructure. And the huge buildings out in the distance with the maintenance infrastructure and the workforce.
Think if there was just one flight per day per gate, what would the ticket cost be? Or one per month?
Actually flying used to be more infrastructure and labor intensive like that, and few people could afford it.
The amount of work per passenger mile has to be small for traveling to be possible for the masses. Otherwise we are in an old class society, since only a small elite can be served by a large amount people.
It was only with more efficient streamlined mass operations (as well as more efficient and more reliable aircraft) that flying became available to the common people, and finally even cheap. It's grown about 20 to 100 fold since 1950. Fuel cost is a significant portion of airline expenses by the way...
With rockets we are at an even worse starting point, since they cost roughly as much as an airliner (about a hundred million), yet are used only once.
That is the first point. But like demonstrated, solving the first problem is not enough. If the rocket must be "refurbished" after every flight and thus flies only a few times per year, nothing is gained, since all the other costs make the cost per flight so high that launches stay at their almost unusable price.
Hence real transformational reusable launch vehicles must also be cheap to operate per kilogram to orbit.
We don't yet really know how to do all that.
I am genuinely curious. I would've have thought that the amount of special-purpose propellant would be the biggest factor.