Good morning. Even by the standards of post-lockdown labour market strength, Friday’s jobs report was a doozy. The US economy added over 350,000 jobs in January, and the details of the report looked healthy, too. The canned explanation is companies “labour hoarding”, which feels unsatisfying to us, a bit like saying a car moved because the wheels spun. Have a better explanation? Email us: firstname.lastname@example.org and email@example.com.
I used to think competition was a stronger economic force than it really is. I believed that all companies with very high returns would be forced back to the mean eventually. So I kept getting surprised by the way that dominant companies, such as Apple, only seemed to get bigger and more profitable with each passing year. Surely what happened to the handset businesses of Motorola, Nokia and BlackBerry would happen to Apple’s sooner or later. “Trees don’t grow to the sky,” I would think. Then I read W. Brian Arthur’s paper “Increasing Returns and the New World of Business”.
Arthur makes a distinction between classic production industries, for steel and the like, and knowledge-based industries. Production companies’ returns diminish as competition pushes returns back to the cost of capital. But some knowledge companies demonstrate increasing returns. Arthur cites three reasons for this: these special companies have some combination of high development costs and low unit costs; significant switching costs for customers; and network effects. Network effects get the most attention these days, but the other two are different and important.
The strategist and professor Michael Mauboussin, of Morgan Stanley Investment Management, has a new paper tracing increasing returns to five separate causes: economies of scale; international trade; learning by doing; network effects; and recombination of ideas. The first three sources are not the exclusive realm of knowledge industries such as tech or pharma. Economies of scale happen because unit costs fall when higher sales volume allows ever more refined divisions of labour; international trade allows highly specialised producers to find large markets; “learning by doing” refers to the fact that as production volume increases, producers gain experience and discover efficiencies.
The last two sources are more native to high tech, and are sometimes mistakenly treated as if they are one and the same. Network effects are familiar by now: information processing products and marketplace services become more valuable the more users they have. “Recombination of ideas” is more complex. If a company’s products are ideas (algorithms, formulas, designs), the development process may be expensive and difficult. But ideas, once developed, tend to combine and propagate. If the offspring of the ideas can be made the company’s exclusive property (with a patent, say), the result is a valuable product that has a zero or near-zero incremental cost.
Here is Mauboussin’s table:
Mauboussin does not trace out the investment implications of his ideas, but they might be used to make powerful distinctions between companies. Think of the Magnificent 7, for example. All enjoy increasing returns. But from where? Google and Meta lean very heavily on network effects. What distinguishes them commercially is not the development and recombination of ideas. Other companies have offered very similar products, but couldn’t reproduce their networks.
Microsoft also enjoys network effects, but its intellectual property is important — perhaps more important. It sells patented ideas for others to use. You can see the difference this way. A company that had legal access to all the technology behind Facebook’s social network or Google’s search engine still couldn’t compete with them. They’d need the users, too. If a company could legally sell perfect copies of Microsoft’s software, it could immediately compete with Microsoft on price.
Now turn to Amazon’s retail business. It enjoys network effects, but also depends very heavily for its increasing returns on traditional economies of scale and learning by doing. Hence the company’s huge investments in physical infrastructure, and its relentless experimentation. It is also a huge beneficiary of increasing returns from international trade, having created a tech and logistics layer for getting cheap manufactured goods reliably from Asian factories to a huge end market in the West. Interestingly, though, competitors like Temu and Shein are challenging Amazon’s dominance on this front.
Apple’s astonishing returns also flow from a blend of all five causes. It does not have particularly low incremental unit costs: the iPhone 15 Pro Max reportedly costs between $500 and $600 to produce. But the company can protect returns because it never seems to lose pricing power. That’s likely down to a combo of network effects and the attendant high switching costs (consider iMessage lock-in in this context), recombination of ideas (those constant software updates), learning by doing and international trade (has anyone ever produced a high tech product at a scale matching Apple and its stable of global sub-contractors?).
With Nvidia, a key question is whether its current lead in AI chips (recombination of ideas) will allow its technology to become an industry standard, akin to Microsoft’s PC operating systems, or whether it can be displaced by superior or less expensive rivals.
Cloud computing businesses are also an interesting case. Do they have true network effects, or just high switching costs? Do incremental advantages of scale in cloud computing diminish at some point?
Tech investors have to ask whether a given company’s increasing returns are sustainable or not. They need to be able to tell a BlackBerry from an Apple. So it is worth thinking about whether increasing returns stemming from different causes have different half lives.
Larry David, grump.