The revenue or cost of each additional unit is called the margin. Economists assume that corporations seek to maximize profit, which occurs at the unit for which marginal revenue equals marginal cost. They also assume that people experience diminishing marginal utility as they consume more of a given good, and that corporations experience diminishing marginal returns to production. Diminishing marginal returns to production means that as more of an input is applied to production, more is produced, but each additional unit provides less additional output than the previous unit. This means that marginal costs increase with production. In a market with many competing producers, a situation which economists call perfect competition, the supply curve is equal to the marginal cost curve…
This academic theory is very much divorced from reality, particularly as it applies to the market for manufactured goods. There is a phenomenon in manufacturing called the learning curve. Learning in this context means that as more units are produced, the cost of producing additional units, the marginal cost, decreases. It was first discovered in the 1930s in the manufacturing of aircraft and has been documented for a variety of products such as missiles, tanks, spacecraft, calculators, and refrigerators. The general rule is that cost decreases by a constant percentage every time the quantity doubles. This percentage decrease can be as much as 30%.
That was the wise Christian Smart in his post, “Neoclassical Economics: The Cult of Marginality.” It seems he started a new blog in Feb 2020 with several interesting posts. He has a follow-on post about marginal cost here.
As Christian notes, decreasing marginal costs isn’t just for aircraft and missile production. It is the new trend as we move away from the industrial era and into the information era. The creation of intangible assets (like software, databases, product design) all have high upfront costs and low-or-zero cost of reproduction. Other intangible assets (business processes like lean manufacturing, supply chains, context specific training) also have high upfront costs, but continual improvement leads to lower and lower marginal costs.
One important thing about intangible assets is that the dollars put into them usually doesn’t track well to specific outputs:
(1) When a firm invests in an employee training program or enterprise software, that costs real money but the productivity improvements accrue across the entire firm and all its product lines. These indirect costs are only growing as the assembly line gives way to knowledge work. Most costs of tech firms cannot be traced to a widget.
(2) When working on new product designs or software, adding more labor, materials, and capital won’t necessarily increase output. Intangible capital not only requires skilled/creative labor, it requires a lot of context and on-boarding so that the new worker becomes useful. While assembly lines have linear handoffs between people, intangible assets require most participants to communicate with each other. This means communication becomes exponentially harder with more people. Hence Bezos’ two-pizza team rule.
(3) When a firm designs new software, that also costs real money but it isn’t clear if and when the software will generate revenue. Because the software can be replicated on everyone’s computers at no marginal cost, the “cost” of each unit depends on how many get sold in the past and in the future. With future sales unknown, the cost of sales is also unknown. This is a consequence of scalability.
(4) Intangible assets have significant spillovers, both within a firm and between firms. When a company designs a more efficient battery, for example, elements of that design can also improve other product lines at no cost. Other firms may also piggy-back on the design.
And so the government’s obsession with tracking money costs to specific end items — which made perfect sense in the industrial assembly line era — is no longer workable. It incentivizes firms to operate in industrial era methods of tracking all expenditures, and that means giving up investments in intangible capital. So defense firms have stovepipes around each government program, and builds the entire thing full-stack.
But as Marc Andreessen argues, software is eating the world. The corollary is that software-native firms will get better at doing hardware than hardware-native firms can do software. SpaceX is beating out ULA in part because it doesn’t do government cost accounting — only accepting fixed-price contracts. That way it can invest in enterprise tools, software, new designs, and so forth, rather than focusing on getting paid for turning wrenches or writing lines of code.
Leave a Reply