Why do people who get paid the most do the least?
Third derivative work
In No One is Really Working, I provide possible justifications for the high salaries of early-career professionals.
This post presents a theory of “derivative levels” as a way to understand why certain roles and types of work are highly compensated.
Both CEOs and professors are highly compensated, with different combinations of financial and social capital, yet neither appears to do much on any given day.
Consider the average day of a CEO:
Wake up
Go to the gym
Go to the office
Get briefed by your assistant
Respond to some emails
Go to some meetings
Lunch
Sit through a strategic initiatives meeting
Send some emails
Go home
And now consider the average day of a professor:
Wake up
Drink coffee
Give the same lecture you’ve done 1000 times with nobody listening
Go to a research meeting
Lunch with other faculty you don’t really like
Talk with graduate students about research
Write a grant you probably won’t get
Go home
Everybody who isn’t a CEO or professor looks at these schedules and thinks to themselves, “These people aren’t doing anything”, followed by “I can do that.” On most days, this is probably correct. The trajectory of Chipotle would not change if I was CEO for a day. College students around the world would still get their protein slop bowls that day, and life would go on.
Some people consider this oppression: “Why do CEOs get paid more than I do, when they’re writing leisurely emails and I’m digging ditches in the hot sun?” Although this might happen in a few cases like nepotism, in a competitive labor market one should not expect CEOs to get paid out of proportion to the value they add. In No One is Really Working, I offer seven explanations as to why professionals get paid high salaries to do seemingly nothing. One rationale goes as follows:
2. A single breakthrough covers everything.
A worker comes up with the idea of a widget that increases internal productivity 1000-fold or creates a new product that everyone wants. The firm asymmetrically benefits from capturing the economic value of this breakthrough and does not compensate the employee proportionally to the value they’ve created.
You don’t know who will do this ex-ante (and neither does the employee) so you have to pay everyone an inflated salary to attract the innovator.
Compensation impact: High in select industries, low otherwise
This explanation provoked the most emails and comments. One commenter wrote:
At first glance, this feels intuitively true. Adam has a skill that not many people have (programming), Brenda has a skill that more people have (copywriting), and Carl has a skill that everyone has (talking). Adam and Brenda have work outputs that clearly translate to the bottom line of the company while Carl does not.
Contrary to popular belief, Carl actually deserves the highest compensation. This is because his work has the highest potential impact on increasing the company’s bottom line.
In this post, I describe derivative levels as a model for understanding a worker’s leverage in changing a company’s output.
Just as a derivative in calculus measures how quickly one variable changes with respect to another, a worker’s derivative level reflects how a worker’s actions affect company output.
First Derivative Work (Brenda, Writer)
Every day, Brenda gets to the office at 9 and sits down to write. She is able to pump out 500 words, which are then sent to the editor for review. She doesn’t have super strict deadlines, but aims to publish one piece a week to the company’s website. She gets her work done and heads home at 5.
To do her job, Brenda uses her computer to do work. The hardware, software, and internet infrastructure required were made by other people.
Brenda’s work is the modern day version of working on a factory floor. There are tools available to her that she uses for her job and allows her to do her job efficiently. This is what is so great about the modern structures of the economy: somebody is able to make a widget that allows everybody else to be more efficient.
While Brenda might not be getting paid per hour, her output and compensation scales linearly with how much she works.
Second Derivative Work (Adam, SWE)
Adam works at Roblox, a game platform that enables game creators to create new games and experiences for users. Adam works on the game tools team, building the infrastructure to allow creators to make new games.
Adam’s work helps creators make more engaging games for the users. Adam, Roblox, and game creators are incentive-aligned as improved tooling increases Adam’s chances of increased compensation, Roblox’s increased revenue, and more money to game creators.
Adam engineers and implements these new solutions. Every day he writes a couple hundred lines of code. These days, he’s using AI tools so it’s less and less of him writing the code, with Claude writing most of the code and him reviewing it.
Even though Adam only works for two real work hours a day, his work impacts many hundreds of thousands of game creators on Roblox.
What features and projects that Adam works on come from his manager or his skip via Jira tickets. Endless meetings occur where context and knowledge transfer about goals go from the managers to the engineers, like Adam. Adam asks questions and provides some technical insight before ultimately starting to implement the solution. As the project or feature is being developed, Adam may have technical feedback about constraints or offer some ideas for marginal improvements on these features.
Unlike first derivative workers, the output of second derivative workers like Adam scales superlinearly with how much he works, but is ultimately constrained by the number of hours in the day that he is able to implement and engineer these new solutions.
Third Derivative Work (Carl, Consultant)
Carl works as a Strategy Consultant at a Big 4. He works with Fortune 500 companies on their AI strategy and on developing systems to increase aggregate worker productivity.
The productive Carls of the world actually increase aggregate worker productivity. They use a combination of data science, intuition, and tacit knowledge to identify what are the important problems to solve and design solutions for it. Examples of this include Deloitte integrating a predictive scheduling engine for Starbucks based on demand forecasts and customer-flow analytics to increase labor efficiency and improve customer satisfaction.
Even something as trivial as outsourcing basic calendar functions to machines is so insanely high leverage that entire businesses can be built around it.
The nature of third derivative work includes a lot of talking and thinking without concrete work product. As in the CEO and professor day-in-the-life anecdotes, this can make it seem like they are replaceable and not really contributing anything. But the few things that they identify in a year, maybe just one in their entire career, pays for everything else.
The Carl who identifies and manages the predictive scheduling engine project will generate tens of millions of dollars of business value. Carl pays for the modal management consultant who is absolutely worthless, imposing a negative impact on the world by wasting people’s time in meetings and allocating resources to dead-end projects.
Strategy and consulting gets shit on so hard because the modal person who says they want to work in strategy is an unskilled MBA student whose only talent is resource extraction. They see a high status, well-compensated job with lax day-to-day lives and think “I want that.” Many of them do end up getting it by leveraging their credentials, and extract value from the system for decades.
Third derivative workers are the hardest to interview and select for. Oftentimes, the candidate themself has no idea if they are a productive third derivative worker or an imposter resource depleter. Carls implicitly negotiate and operate as a quasi-union to normalize their salary as the mean of the group, which, given the power law nature of their work, explains their high wages.
The output of third derivative workers is a multiplicative factor on the superlinear function of second derivative workers based on the number of second derivative workers they can coordinate.
Coda
So what does this all mean for the ambitious young person?
Institutions and credentialing can help parlay yourself into new derivative levels but if you want to move up within your derivative level or up to the next derivative, you will have to apply novel ideas to the right contexts. Most of the ideas you will have will be very bad or already implemented. Fortunately for you, agentic reader, the bar is very low: the modal third derivative worker is worthless because they can’t even pattern-match correctly.
Context comes from difficulty, exposure, and stakes. Hate to break it to you, but there’s no shortcuts.
Having one semi-original idea in your entire life requires lots of preparation and lots of luck. So ask yourself: What would I do if I was the luckiest person in the world?
Bonus: Derivatives, Continued
We can extrapolate the derivative analogy further into a fourth and fifth derivative:
Compensation and influence increase as you work your way up the derivative ladder because abstraction, fueled by financial and social capital, enables compound leverage. This model explains a possible not-too-distant future where the people who control the models are more powerful than the elected President.
Popularity prediction hash: a68bf3b67ae721265e49cd5d1e96f3f1432fc977851fda7fe6a3d96d74d40c76
Thanks to Vaniver Gray for discussion and comments.
Professors are highly compensated? That’s news to me.
I thought that the academy was notorious for appalling pay, driving many PhDs into consulting, banking, corporate strategy, etc.