It’s trivial to cook dinner for yourself. It’s easy to do it for your family. A dinner party takes planning. But catering an event is absolute chaos.
This is an everyday example of an immutable law: As things get bigger, they get more complex. This is as true for organizations as it is for dinner parties: As an organization grows, it needs managers. More growth, and those managers need managers. At some point, organizing a system soon outweighs the surplus energy that the system can produce.
Joseph Tainter has spent his life thinking about why organizations—from corporations to entire societies—often collapse as they grow beyond a certain size. In Collapse of Complex Societies he says that societies collapse because of “diminishing returns on investments in social complexity.” In other words, the more complex something gets, the harder order is to maintain.
In Tainter’s work, organizations become more complex as they try to solve problems, starting to rely on analysts and coordinators not directly involved in production, which create more problems.
The same is true in societies. As every society confronts issues stemming from growth (such as theft or taxation or defense) it creates new layers to deal with it (such as a judicial system, or revenue assessors, or armies) which further increase its costs—a pattern Tainter has tracked across several civilizations.
Similar patterns exist in modern democracies. Since 1955, the population of the US has roughly doubled, growing from 161 million to 340 million residents. During that time, the number of local government employees in the US grew from 3.5 million in 1955 to 14.5 million in 2023—a factor of four.
The conclusion is bleak: The size (and cost) of the bureaucracy is growing exponentially compared to the size of the society it governs.
Information technology was supposed to save us. By using punched cards to conduct the census, government helped usher in the mainframe era. Governments are, at their core, information processing systems. While the edges of a government do physical work like building roads or monitoring fisheries, the chewy center of the public service does things like interpreting laws, granting permissions, collecting money, and allocating resources.
Since the 1960s, computing power has increased a trillionfold (if you need help visualizing that, check out this infographic from Visual Capitalist.) Storage and bandwidth have also dropped precipitously. And yet, public sector costs continue to climb. Tainter’s complexity beats Moore’s Law hands down.
There’s a solution, however.
I’ve referenced Herbert Simon many times over the years. He won the Nobel Prize; coined the term “the attention economy” back in 1970; and created the word “satisficing”, to describe how humans make decisions that are “good enough,” challenging the notion of economically rational buyers.
Simon tells the story of two watchmakers, Hora and Tempus. Both are skilled and in demand, and both made complex watches with a thousand separate pieces. Yet Hora prospered while Tempus eventually went bankrupt.
The two had very different ways of working. Tempus designed watches that, when he put down a partly assembled watch, it fell to pieces and had to be reassembled. By contrast, Hora designed his watches from sub-assemblies of about ten pieces that could be put down without falling apart. These were, in turn, combined into sub-assemblies, and these into larger sub-assemblies.
Herbert’s solution to Tainter’s doom and gloom is simply modularity.
The Boeing 787 has around 2.3 million parts, yet it’s a reliable aircraft that is, mile-f0r-mile, a safer form of travel than an automobile. That’s because Boeing doesn’t build the plane from scratch, the way Tempus did. Instead, the plane is made up of components: An engine from GE and Rolls-Royce; wings from Subaru; cargo doors from SAAB; wheel assemblies from Messier-Bugatti-Dowty, and so on.
By https://www.flickr.com/photos/markjhandel/ - https://www.flickr.com/photos/markjhandel/774759265/, CC BY 2.0, https://commons.wikimedia.org/w/index.php?curid=2838781
Complex systems scale when they’re made from subcomponents that are self-contained, tested, and improved independent of the overall system. Those subcomponents must be easily connected to one another through clear, well-documented standards. They must have known operating ranges, and be able to scale within those ranges reliably.
As a tech example, consider a cloud storage system, such as those from Amazon, Microsoft, or Google. The system performs a very simple function: Give it an object (such as an image) and it will give you a URL. Give it a URL, and it will return the image. It’s simple, bulletproof, and fast. Because of that, a significant amount of the modern Internet runs on these systems.
So why is so much of the public sector Tempus instead of Horus?
Jennifer Pahlka offers one reason in Recoding America: Going from legislation to implementation is a one-way street. Lawmakers don’t consider implementation when creating laws, which means implementers can’t push back. Laws get more complex, code gets more complex, and the end result is a Byzantine mess of custom tools because no off-the-shelf system can satisfy all the requirements. This in turn produces an almost guild-like team holding arcane knowledge, making recruiting hard.
Another challenge is the inherent fiefdom in government models (something we’ve written about in a post on Cheesegraters.) Structural incentives in government mean a not-invented-here attitude often prevails, rather than building for re-use across departments and jurisdictions.
But a set of standardized, well-maintained components—each owned and improved by a discrete group—can reverse Tainter’s curse. Re-usable infrastructure (such as broadband, computing, and storage) and re-usable components (such as forms, alerts, design frameworks, and messaging platforms) offer many advantages:
More wood behind fewer arrows: Improving one component benefits everyone.
Faster creation of new services: It’s cheaper and easier to build new services that work out of components that are already known to work.
Governance and compliance: When a component has been certified (for accessibility, language, security, and so on) then things built atop it inherit that certification.
Economies of scale and skill: Shared resources like bandwidth and cloud computing become more cost-effective and resilient the more users they have, because infrastructure is amortized across all users. Similarly, experts in security, reliability, and performance can fine-tune each module to get the most out of it.
Government is the most complex of systems, growing heavier and more sluggish each year as populations soar, despite the incredible power that modern IT offers. We can avoid this collapse through modular, re-usable components and shared infrastructure. We need to think like Horus.