Replaceability and the Economics of Disequilibrium

Sam is a 2014 blogging resident visiting us from his home blog at Moore’s Hand.

One characteristic of personal meaning is irreplaceability. If you’re the only one who can do what you’re doing, your actions suddenly seem a lot more important.

We’re familiar with this principle in our personal relationships — perhaps a friend needs a piece of advice, or a child needs parenting, that only we can give.

But larger systems are designed to make most people replaceable.

If you are replaceable, the system will equilibrate without you. If you aren’t replaceable, it won’t.

This is good for you but bad for the system. So your manager’s job is to make sure that the company will survive you being hit by a bus.

Replaceability is a spur to the ambitious. Every law school grad applying for a Supreme Court clerkship; every Ivy League grad interviewing at Goldman Sachs or McKinsey, knows that ten other people want the job, and can do it.

But replaceability is existentially demotivating. “Just a cog in the machine” has been an epithet for employment since Charlie Chaplin filmed Modern Times.

If a system will achieve roughly the same outcome no matter who’s inside it, these people are by definition replaceable.

And so in seeking irreplaceability, we must ask: what systems are capable of achieving genuinely different equilibrium outcomes?

Feedback loops

Differential equations are a branch of mathematics aimed to model systems where the system’s change over time is heavily influenced by the current state of the system.

A spiral wishing well is a convergent system. When you drop a coin in, its route will be different depending on where it starts, but all pennies eventually funnel down into the same final destination.

On the other hand, Glacier National Park in Montana is a divergent system. A cup of water poured in the west will flow down to the Pacific Ocean. Poured in the southeast, it flows to the Gulf of Mexico; in the northeast, to the Hudson Bay. The point at which a divergent system diverges is known as the inflection point, or colloquially, the saddle.

Would-be system nudgers should note: convergence/divergence are characterizations of a system at a point in time/space rather than overall system traits.

The spiral wishing well is a convergent system throughout the lifetime of the penny. But the watershed is only divergent in Glacier National Park; it is convergent elsewhere as the water flows out of the park and its final destination becomes inevitable.

Zooming one step out

In the market, the fiercer competition is, the more strongly the system equilibrates.

Sometimes this is obvious — for example, if we are on the job market. But if not, zooming out one level often helps us see equilibrating forces.

Let’s return to our Ivy League grad, who has decided to accept the offer at McKinsey. Frustrated by a sense of meaninglessness, he spends five years on East Asian projects, becoming the firm’s resident expert on, say, the South Korean export/import market.

Suddenly he’s needed for every project with an outsourcing theme. This figures prominently in a new project one of the firm’s partners is pitching. Due to the now-senior-associate’s expertise, the firm lands the contract.

That evening, our protagonist is walking home from the office, pleasantly musing on the day’s events, feeling a deep sense of satisfaction that his work is needed.

While thus engaged, he starts to explore the counterfactual if he wasn’t around.

We wouldn’t have been able to land the contract, he thinks. It would probably go to our competitors.

And they probably have someone just like me, who’s spent several years becoming an expert on the same topics I have, who would have been doing the exact same work I’ll be doing now.

Our associate may be irreplaceable in his firm, but if his firm’s work is commoditized, then he is not irreplaceable existentially.

‘Great men’, revisited

This line of thought can be a bit demoralizing. Alas, we have a bit more convergent-system bath water to throw out before we get to our divergent-system baby.

Perhaps many run-of-the-mill elites are replaceable. But typically, if we think about people who made some dent in the world, we think about more well-known historical figures; people who, popularly represented, stood at some fork in the road and pushed society one way.

How replaceable are these type of figures?

If not for Watson and Crick, DNA’s double-helix structure would probably have been found soon thereafter, eg by Rosalind Franklin.

But the search for a polio vaccine had been going on for seventeen years without success before Salk. And no one was looking for antibiotics when Fleming discovered penicillin.

Pizarro and Cortes’s conquests were stunning. But had they failed, the Aztecs and the Incas would have continued to be (a) lucrative targets and (b) vulnerable to smallpox.

Napoleon, on the other hand, marched over Europe in a manner difficult to imagine an alternate leader doing.

Inevitable battlegrounds; contingent winners

Our focus moves now to technology, the standard source of economic historians’ illustrations of divergent systems (they use the term ‘path-dependent’).

The textbook example is of the QWERTY keyboard layout in 1900. Widespread adoption of QWERTY created a virtuous cycle where it made sense for new typists to learn QWERTY rather than a different format.

Technological progress in general, and Moore’s Law in specific, tends to create an interesting interplay of convergent and divergent systems. It creates inevitable battlegrounds but contingent winners.

As processing power continued its relentless doubling, mainframes were replaced by minicomputers, minicomputers by desktop PCs, and desktops by laptops.

This progression of computing machines was more-or-less inevitable. But battlefield results — Microsoft and Apple won; Cray and Tandy lost; IBM pushed — were far more historically contingent.

As soon as the personal computers and the internet became widespread and fast enough to handle voice, something like Skype was going to become prevalent.

But once Skype became prevalent, it was, like most tech firms, an oligopoly.

Market structure is perhaps the largest factor in creating contingent results in tech — network effects and fast-moving markets typically result in one or two players dominating any tech-related market.

This market power enables flexibility in crafting product, which leads to idiosyncrasy.

Perhaps there could have been a sustainable mass-market social media platform that worked significantly different than Facebook. But we’re unlikely to find out now.

Convergent features; divergent backends

Another wrinkle in the tech story: among oligopolists, feature sets are more likely to converge than technical backends (just look at Twitter’s recent redesign).

Features are legible to customers and thus subject to the equilibrating effects of competition.

But on the backend, no one sees the programming language or technical architecture, there are usually multiple ways to achieve the same goal, and good gardening by software engineers tends to be a weaker force than than market pressures.

In college, I worked in the clothing department of the campus bookstore. All the sweatshirts in the retail area were neatly folded and the polos neatly hung. Our stockroom, frequented only by employees, was an ill-kept mess.

When an ill-kept mess becomes the initial condition for further technological development, things get interesting.

Don’t smash the butterfly

A good example here is the operating system market.

Back in 1997, Microsoft was king, Linux had not yet gained momentum, and Apple seemed on its death throes.

Asked what Steve Jobs should do, Michael Dell suggested the company close shop and return its cash to shareholders. Ultimately, Jobs convinced Bill Gates to invest $150 million and develop Office for Mac, and the company survived.

To the average consumer (though not to the Justice Department), this might not have seemed too consequential. So what if Macs disappear? There wasn’t a lot you could do on them that you couldn’t do on a PC.

But the technical gulf between these systems ended up playing a large roles in subsequent history.

Linux was open-source — thus both free to use, and easier than Windows to program on. Apple was vertically integrated, producing both its hardware and software.

In 1997, when software was “shipped” by mail and inserted into the consumer’s CD drive, software had to be written on the same OS it ran on. So ease of use for consumers, not programmers, was the relevant OS constraint.

But in 2014, when software runs on a remote server and is “shipped” via the Internet and a browser refresh, programmer ease of use suddenly matters.

Since 2000, the cost to start a tech company have fallen from around $5 million to around $5,000.  Linux-driven cloud computing played a large role in this trend, which has helped drive Silicon Valley’s current boom.

As for Apple — its hardware/software knowledge turned out to be pretty useful in dominating the smartphone market with the iPhone.

Google’s Android, in second place, is basically built on Linux. (and hardware integration has continued to be a pain point.)

Microsoft, of course, missed the boat.

Killing Linux/Macintosh in 1997 would have been, like Ray Bradbury’s butterfly-stomping time traveler, barely noticeable at the time but hugely consequential in the future. This is technical path dependency par excellence.

The scope tradeoff

Our journey began in a quest to make oneself irreplaceable. We’ve gone on what seems like quite a detour, looking for contingency and exploring path-dependent systems.

Let’s come full circle.

In deciding what to pour our energy into, we typically face a scope tradeoff:

Untitled Diagram (2)

If we work on large problems, perhaps in our professional lives, we usually see many other people doing the same thing. We wonder if we’re replaceable.

If we work on small problems, perhaps in our personal lives, we can likely see concrete impact. But we look around and see the vastness of the world and wonder if we’re actually doing anything consequential.

The standard coping mechanisms here are mostly internal, ie, identify with a larger narrative of which your actions are part.

But another tactic is to note that the above dichotomy is incomplete. Specific types of big systems are path-dependent and tend not to equilibrate.

And therein lies the opportunity.

Get Ribbonfarm in your inbox

Get new post updates by email

New post updates are sent out once a week

About Sam Bhagwat

Sam Bhagwat is an economist/data scientist by training. His ribbonfarm posts explore complex economic systems. Follow him on Twitter.

Comments

  1. “Killing Linux/Macintosh in 1997 would have been, like Ray Bradbury’s butterfly-stomping time traveler, barely noticeable at the time but hugely consequential in the future.”

    While I get the hypothetical point, I think this scenario is more of a Pizarro/Cortez story than a Microsoft-as-failed Napoleon story. While Apple could have died then, it’s entirely conceivable, and I would argue likely, that another competitor would have sprung up. And I don’t see how Linux was killable at all, short of (ab)using the patent system to monopolize the entire operating system space.

    • The scenario you laid out is certainly possible. Two possible complications:

      (1) would a different non-Apple competitor have integrated hardware/software?

      Typically you have tightly coupled, integrated components at the beginning of a market and then over time it moves to modularized components (ie the late 90s PC ecosystem). Yet it was exactly that integrated hardware/software — a legacy backend technical decision which Apple was rethinking in 1997 — that enabled Apple to pull off the iPod and iPhone.

      (2) It’s not that you would need to kill Linux per se, but could the Android of 2005/6, and the cloud computing of 2008, have been built on the Linux of 1997?

  2. And you are going to end the post on that note? Care to elaborate WHICH specific types of big systems tend not to equilibrate?

  3. Smart phones seem like what some people imagined PDAs would be 15-20 years ago “Personal Digital Assistants” — something we’d need to have at our fingertips all the time to function at our full capacity. The failure of PDAs left people pessimistic about realizing that sort of concept for several years, but it was actually there many years ago.

  4. Patrick Atwater says

    Fun little analysis that rings true — for me at least. Then again if I were to put on my skeptical cap, this whole argument seems like a justification for a “final vocabulary”1 for “change the world” types.

    1 To borrow Rorty’s excellent phrase for the foundational narrative beliefs we cling to regardless of what the world throws out.

  5. As a public high school teacher I can attest to schools being ‘big systems’ that often do not equillibrate well after losing an individual. Looking top down there is overwhelming evidence that replacing an experienced teacher with a rookie significantly decreases classroom learning potential. From the bottom up, many students can think of a time when an irreplaceable teacher left and the void was never filled. Quality teachers may be the most irreplaceable individuals in any system.

    Not quite sure how this translates to the business world. Maybe mentor-protege relationships within a firm mirror teacher-student relationships in schools?

    • Sam Bhagwat says

      I can certainly believe that.

      Schools have the resources to make teachers show up for their classes. They don’t have (or perhaps metaphysically can’t have) the resources to nudge teachers be *good* teachers, that business organizations have.

      – They can’t/don’t measure performance: who is a good teacher?
      – They don’t have the resources/flexibility to hire/fire/promote/give nonstandard raises to good tenured teachers.
      – Because of factors including tenure, experienced, good teachers are unlikely to be on the job market. Also, schools don’t get more funding if student welfare increases. So they’re likely to want to hire a rookie teacher, rather than an equally experienced teacher, in this situation to save money. That is not true in business — if the VP sales quits, the company finds someone of equal experience.

      Thus, motivation is largely internal to the teacher, which doesn’t equilibrate.

  6. Great example from Curtis about trying to replace good teachers (well actually the school doesn’t necessarily recognize good or try to replace… that’s the point)

    That example helped me reflect that the begged question in this fascinating ribbonfarm post is how we measure significance.

    A lot (for some of us, most?) of what we do has no publicly or objectively measurable value, yet we value it and orient ourselves using that value perception.

    Some of the most ‘valuable’ employees, teachers, parents, scientists, artists, humanitarians… are precisely those who ignore the question of whether their quality of production and influences are publicly quantifiable or big-picture-significant. They get on with it and focus on their sphere of quality. Sometimes the effect in history is big and visible, but it doesn’t have to be: by not caring about being recognized as making a difference, these are the real difference-makers.

  7. I used to suffer from a sense that it wasn’t worth doing anything because someone somewhere is doing better, or will do better some time in the future. I resolved this when thinking about what Elon Musk said about Tesla Motors: the goal is to accelerate the development of electric vehicles. If Tesla didn’t exist, someone else would do it eventually- but the idea is to put pressure on it, to live in the future and drag it to meet the present. That I can get behind; that strikes me as meaningful. To accelerate development in some way.

    • Elon Musk rationalizes his actions because doing R&D with a passion is a bipolar experience, something you go through phases of hope, joy and suffering. So when you are in a love story with your current project, enthusiasm, jealousy, fear that someone else may be the first and all kinds of other irrational desires are not unusual. In retrospect, when the story is over, no matter which outcome, one might grant oneself a lukewarm acknowledgment that one was part of a broader, ambient progress, that one helped to accelerate some general movement or something alike but this is never the whole story one tells to oneself.