A non-marketable output is something that has relatively little value outside of one firm. One company’s tax return won’t help any other firm file its return. A half-finished Chevy is not of much use to Toyota.
1. I claim that a typical firm buys marketable outputs, produces non-marketable outputs, and turns at least some of these non-marketable outputs into marketable outputs.
2. I can imagine a firm that produces only non-marketable outputs because it works only for a single buyer. However, there is a sense in which the firm and its buyer can be treated as a single entity.
3. Non-marketable outputs are what determine the configuration of firms. Suppose that there are ten stages of production. If at each stage the output is marketable, then there might exist firms at each stage of production. On the other hand, if only the final stage is marketable, then there will be just one firm.
4. The value of a non-marketable output is indeterminate. It has to be worth at least as much as the cost of the inputs required to produce it, and it cannot be worth more than the entire marketable output of the firm. But that leaves a wide range. Consequently, workers engaged in the production of non-marketable output do not have a well-defined marginal revenue product.
5. I conjecture that the larger the firm, the higher the proportion of non-marketable output relative to total output. If all you are doing is buying marketable outputs and selling marketable outputs, then you can be a tiny firm, like somebody who sells on e-Bay. On the other hand, if you manufacture airplanes, then most of your effort goes into producing unfinished airplanes, so you need a large firm.
Compare the old-fashioned general store to Wal-Mart. Wal-Mart has important non-marketable output in its supply chain, consisting of logistical systems and contracts with sellers. That supply chain might be worth something to an old-fashioned general store, and you can imagine a different Wal-Mart acting solely as a wholesaler/distributor. However, the supply chain is worth even more when it is integrated with large, strategically-located retail outlets, namely Wal-Mart stores.
6. I conjecture that non-marketable output tends to become increasingly important as the economy becomes more complex. That in turn would suggest that the trend would be for firms to get larger.
7. Some important creative destruction takes place in the arena of non-marketable outputs. Uber and taxi companies both offer on-demand rides. But Uber replaces the taxi company’s non-marketable output, its dispatching system, with something different. Amazon and traditional sellers both offer books. But Amazon replaces the non-marketable outputs of the traditional sellers (inventory management, product display, and customer fulfillment) with something else.
Note that Amazon found that a lot of its infrastructure proved to be marketable. Other companies rent server space from Amazon or sell products using Amazon.
A very Austrian idea.
Counterexample: Apple?
The firms they outsource to couldn’t sell the iphone components they make if apple didn’t want them anymore, i think.
That’s directly analogous to the car part companies. Nissan may not want a GM assembly, but the part company can retool easily to produce the same thing to Nissan’s specifications. So we have a category of non-marketable output that isn’t tied very tightly to the needs of one company, rather than an entire industry. Those types seem to be sensible to outsource.
“I conjecture that non-marketable output tends to become increasingly important as the economy becomes more complex.”
There’s a countertrend that people get cleverer about figuring out how to slice and standardize subsystem functionality so that the subsystems become marketable. A similar trend and countertrend can be seen in computer software: as software systems become more complicated they tend to acquire more subsystems that are too specialized to be useful independently, but as time goes on people sometimes learn how to tease out such functionality into useful standalone products (typically libraries, but also things like software frameworks and specialized pieces of hardware with lots of firmware). E.g., the role of serving web applications is no longer so dominated by big monolithic applications like Apache as it was a decade or more ago, people have gotten very clever about teasing the problem apart and producing standalone software that solves part of the problem and gains flexibility from being able to be plugged into other things. (It’s hard to describe without huge digression, but e.g. Plack is one of several libraries that do this kind of thing, and I personally have found it useful.) For that matter, even the web itself is in part a vast effort in teasing software (and other information-heavy stuff like publication and distribution) apart into individually useful bits. Or as another example, as operating systems have become more complicated, things like windowing systems and filesystems have sometimes been teased out into separate components. They aren’t useful in isolation any more than e.g. automobile batteries or spark plugs or tires are, but even if they are only useful as part of a larger piece of software (like a spark plug is useful in an engine in an automobile or lawnmower) they have been turned into components that can be sold independently of complete OSes/automobiles/whatever.
I am skeptical of #6 “… that non-marketable output tends to become increasingly important as the economy becomes more complex.” It smells of the old socialist argument that central planning is needed complex industrial society while property rights evolved for the simpler needs of medeival artisan-burghers.
It might be that as economies and systems get more complex, the giant integrated firms turn into dinosaurs. Outside those firms, standardized products come into being which are marketable, and the old way is replaced by a market of buyers and sellers of those standardized products.
I had the computers and the internet in mind when I wrote the above paragraph. Think mainframes diminishing in importance relative to Unix & Windows servers. Or prorietry networks replaced by applications running over TCP/IP. But I can imagine the same process works in a humbler way in all fields.