Paul Graham has recently penned an essay on income inequality. Holly Wood wrote a pretty good critique of this piece, but it is addressing a huge amount of pre-existing context, as well as ignoring large chunks of the essay that nominally agree.
Eevee has already addressed how the “simplified version” didn’t substantively change anything that the longer one was saying, so I’m not going to touch on that much. However, it’s worth noting that the reason Paul Graham says he wrote that is that he thinks that “adventurous interpretations” are to blame for criticism of his “controversial” writing; in other words, that people are misinterpreting his argument because the conclusion is politically unacceptable.
Personally, I am deeply ambivalent about the political implications of his writing. I believe, strongly, in the power of markets to solve social problems that planning cannot. I don’t think “capitalist” is a slur. But, neither do I believe that markets are inherently good; capitalist economic theory assumes an environment of equal initial opportunity which demonstrably does not exist. I am, personally, very open to ideas like the counter-intuitive suggestion that economic inequality might not be such a bad thing, if the case were well-made. I say this because I want to be clear that what bothers me about Paul Graham’s writing is not its “controversial” content.
What bothers me about Paul Graham’s writing is that the reasoning is desperately sloppy. I sometimes mentor students on their writing, and if this mess were handed to me by one of my mentees, I would tell them to rewrite it from scratch. Although the “thanks” section at the end of each post on his blog implies that he gets editing feedback, it must be so uncritical of his assumptions as to be useless.
Initially, my entirely subjective impression is that Paul Graham is not a credible authority on the topic of income inequality. He doesn’t demonstrate any grasp of its causes, or indeed the substance of any proposed remedy. I would say that he is attacking a straw-man, but he doesn’t even bother to assemble the straw-man first, saying only:
... the thing that strikes me most about the conversations I overhear ...
What are these “conversations” he “overhears”? What remedies are they proposing which would target income inequality by eliminating any possible reward for entrepreneurship? Nothing he’s arguing against sounds like anything I’ve ever heard someone propose, and I spend a lot of time in the sort of conversation that he imagines overhearing.
His claim to credentials in this area doesn’t logically follow, either:
I’ve become an expert on how to increase economic inequality, and I’ve spent the past decade working hard to do it. ... In the real world you can create wealth as well as taking it from others.
This whole passage is intended to read logically as: “I increase economic inequality, which you might assume is bad, but it’s not so bad, because it creates wealth in the process!”.
Hopefully what PG has actually been trying to become an expert in is creating wealth (a net positive), not in increasing economic inequality (a negative, or, at best, neutral by-product of that process). If he is focused on creating wealth, as so much of the essay purports he is, then it does not necessarily follow that the startup founders will be getting richer than their customers.
Of course, many goods and services provide purely subjective utility to their consumers. But in a properly functioning market, the whole point of of engaging in transactions is to improve efficiency.
To borrow from PG’s woodworker metaphor:
A woodworker creates wealth. He makes a chair, and you willingly give him money in return for it.
I might be buying that chair to simply appreciate its chair-ness and bask in the sublime beauty of its potential for being sat-in. But equally likely, I’m buying that chair for my office, where I will sit in it, and produce some value of my own while thusly seated. If the woodworker hadn’t created that chair for me, I’d have to do it myself, and it (presumably) would have been more expensive in terms of time and resources. Therefore, by producing the chair more efficiently, the woodworker would have increased my wealth as well as his own, by increasing the delta between my expenses (which include the chair) and my revenue (generated by tripping the light pythonic or whatever).
Note that “more efficient” doesn’t necessarily mean “lower cost”. Sitting in a chair is a substantial portion of my professional activity. A higher-quality chair that costs the same amount might improve the quality of my sitting experience, which might improve my own productivity at writing code, allowing me to make more income myself.
Even if the value of the chair is purely subjective, it is still an expense, and making it more efficient to make chairs would still increase my net worth.
Therefore, if startups really generated wealth so reliably, rather than simply providing a vehicle for transferring it, we would expect to see decreases in economic inequality, as everyone was able to make the most efficient use of their own time and resources, and was able to make commensurately more money.
... variation in productivity is accelerating ...
Counterpoint: no it isn’t. It’s not even clear that it’s increasing, let alone that its derivative is increasing. This doesn’t appear to be something that much data is collected on, and in the absence of any citation, I have to assume that it is a restatement of the not only false, but harmful, frequently debunked 10x programmer myth.
Most people who get rich tend to be fairly driven.
This sounds obvious: of course, if you “get” rich, you have to be doing something to “get” that way. First of all, this ignores many people who simply are rich, who get their wealth from inheritance or rent-seeking, which I think is discounting a pretty substantial number of rich people.
But it is implicitly making a bolder claim: that people who get rich are more driven than other people; i.e. those who don’t get rich.
In my personal experience, the opposite is true. People who get rich do work hard, and are determined, but really poor people work a lot harder and are a lot more determined. A startup founder who is eating rice and beans to try to keep their burn rate low and their runway long may indeed be making sacrifices and working hard. They may be experiencing emotional turmoil. But implicitly, such a person always has the safety net of high-value skills they can use to go find another job if their attempt doesn’t work out.
But don’t take my word for it; think about it for yourself. Consider a single mother working three minimum-wage jobs and eating rice and beans because that’s the only way she can feed her children. Would you imagine she is less determined and will work less hard to keep her children alive than our earlier hypothetical startup founder would work to keep their valuation high?
One of the most important principles in Silicon Valley is that “you make what you measure.” It means that if you pick some number to focus on, it will tend to improve, but that you have to choose the right number, because only the one you choose will improve
A closely-related principle from outside of Silicon Valley is Goodhart’s Law. It states, “When a measure becomes a target, it ceases to be a good measure”. If you pick some number to focus on, the number as measured will improve, but since it’s often cheaper to subvert the mechanisms for measuring than to actually make progress, the improvement will often be meaningless. It is a dire mistake to assume that as long as you select the right metric in a political process that you can really improve it.
The Silicon Valley version - assuming the number will genuinely increase, and all you have to do is choose the right one - really only works when the things producing the numbers are computers, and the people collecting them have clearly circumscribed reasons not to want to cheat. This is why people tend to select numbers like income inequality to optimize: it gives people a reason to want to avoid cheating.
It’s still possible to get rich by buying politicians (though even that is harder than it was in 1880)
The sunlight foundation published a report in 2014, indicating that the return on investment of political spending is approximately 76,000%. While the sunlight foundation didn't exist in 1880, a similar report in 2009 suggested this number was 22,000% a few years ago, suggesting this number is going up, not down; i.e. over time, it is getting easier, not harder, to get rich by buying politicians.
Meanwhile, the ROI of venture capital, while highly variable, is, on average, at least two orders of magnitude lower than that. While outright “buying” a politican is a silly straw-man, manipulating goverment remains a far more reliable and lucrative source of income than doing anything productive, with technology or otherwise.
The rate at which individuals can create wealth depends on the technology available to them, and that grows exponentially.
In what sense does technology grow “exponentially”? Let’s look at a concrete example of increasing economic output that’s easy to quantify: wheat yield per acre. What does the report have to say about it?
Winter wheat yields have trended higher since 1960. We find that a linear trend is the best fit to actual average yields over that period and that yields have increased at a rate of 0.4 bushel per acre per year...
In other words, when I go looking for actual, quantifiable evidence of the benefit of improving technology, it is solidly linear, not exponential.
What have we learned?
Paul Graham frequently writes essays in which he makes quantifiable, falsifiable claims (technology growth is “exponential”, an “an exponential curve that has been operating for thousands of years”, “there are also a significant number who get rich by creating wealth”) but rarely, if ever, provides any data to back up those claims. When I look for specific examples to test his claims, as with the crop yield examples above, it often seems to me that his claims are exaggerated, entirely imagined, or, worse yet, completely backwards from the truth of the matter.
Graham frequently uses the language of rationality, data, science, empiricism, and mathematics. This is a bad habit shared by many others immersed in Silicon Valley culture. However, simply adopting an unemotional tone and co-opting words like “exponential” and “factor”, or almost-quantifiable weasel words like “most” and “significant”, is no substitute for actually doing the research, assembling the numbers, fitting the curves, and trying to understand if the claims are valid.
This continues to strike me as a real shame, because PG’s CV clearly shows he is an intelligent and determined fellow, and he certainly has a fair amount of money, status, and power at this point. More importantly, his other writings clearly indicate he cares a lot about things like “niceness” and fairness. If he took the trouble to more humbly approach socioeconomic problems like income inequality and poverty, really familiarize himself with existing work in the field, he could put his mind to a solution. He might be able to make some real change. Instead, he continues to use misleading language and rhetorical flourishes to justify decisions he’s already made. In doing so, he remains, regrettably, a blowhard.