On the harsh reaction to Paul Graham’s post on inequality. Adapting Tyler Cowen’s laws to writing on the internet.

image credit
image credit

On January 2, technology investor Paul Graham published two posts. One on Refragmentation, about 20th century mass organizations breaking into pieces. Another on inequality, about why it’d be best to directly fix problems like poverty rather than fix inequality per se. What puzzled me was the visceral hate the inequality post produced, even by a number of tech insiders. Graham went so far as to write a second version using very simple words on Jan 4 (the implication of writing a dumbed down version wasn’t lost on anyone). Then on Jan 8 he wrote a response to Ezra Klein’s take. Now that this has mostly blown over, I remain puzzled. Why did this blow up? What does it mean about writing on the internet, if anything?

Paul Graham says the best way to disagree is to quote and restate the argument. Then refute the central point. Let me quote Graham:

  • Variation in productivity is far from the only source of economic inequality, but it is the irreducible core of it, in the sense that you’ll have that left when you eliminate all other sources.
  • economic inequality per se is not bad. It has multiple causes. Many are bad, but some are good. For example, high incarceration rates and tax loopholes are bad things that increase economic inequality. But startups also increase economic inequality…. And unlike high incarceration rates and tax loopholes, startups are on the whole good.
  • if we fix all the bad causes of economic inequality, we will still have increasing levels of it, due to the increasing power of technology.
  • Since economic inequality per se is not bad, we should not attack it. Instead we should attack the bad things that cause it. For example, instead of attacking economic inequality, we should attack poverty. Attacking economic inequality would be doubly mistaken. It would harm the good as well as the bad causes. But even worse, it would be an ineffective way to attack the bad causes. We will not do a good job of fixing the bad causes of economic inequality unless we attack them directly.

Now. Let me shorten Tyler Cowen’s laws as below, and adapt them into being a guide for how to write on the internet:

  1. There is a literature on everything.
  2. There are few decisive or knockdown articles or arguments, and furthermore until you have found the major flaws in an argument, you do not understand it.

I take Cowen’s laws as being applicable not just to economics, but any field where causal density is so high it’s hard to determine which factors are important. That is, most writing on the internet. So yes, there is in fact a large existing literature on inequality. And yes, until we place Graham’s arguments within that literature, and understand where his arguments are weak, we don’t really understand them.

Like much of economics, that literature can be tricky. Let me bullet point some basics at least as far as I understand how they apply to Graham:

  • A key economic lens for investigating inequality is partitioning out differences in income versus wealth. Also, other factors.
  • Thomas Piketty argues for increasing returns to wealth (r > g). Matthew Rognlie critiqued Piketty arguing for the importance of increasing land prices for housing. Disclosure: for what it’s worth, I’m team Rognlie. Details here.
  • Income inequality drivers are globalization, taxation, economic rents, discrimination. Additional causes are the ones Graham points to: differences in natural ability, differences in technology skills and the superstar hypothesis.
  • There is some data supporting productivity splits in firms, where technology frontier firms are continuing to improve productivity, with laggards flat.
  • Graham claims economic inequality will continue to get worse as superstar effects continue to become more important with ever better technology. This is an arguable point in the economics literature, with some arguing other reasons. For example the slipping of cultural norms for CEO pay. This can also be argued against by looking at previous economic technology cycles (see for example my post on Carlota Perez). I would treat this as a likely but unproven prediction about the future, not a fact.
  • Up until now we’ve only been discussing the economics of inequality. Graham also makes a policy claim. He claims that attacking inequality directly is an attack on the causes of the creation of wealth (e.g. startups, but the claim is broader to wealth creation in general). And directly attacking poverty, discrimination or education would be more effective.

I would boil Graham’s argument into econ jargon as:

  • Income inequality is driven by superstar effects in technology. There are other drivers, but superstar technology effects are innate to wealth creation itself, and so are more intractable.
  • As such it’s better to attack poverty, education, health directly rather than inequality, since that risks reducing economic growth.

Framed this way, it’s not clear why people are so angry. This is not an original argument on why inequality has grown. Nor an original argument on avoiding excessive taxes and regulation as killing the goose that lays the golden eggs. Graham’s partner Sam Altman made similar arguments in a post written two years ago. So Graham’s post is true as far as it goes. And it’s clear the point he wants to make is focusing on inequality rather than focusing on solving problems may not get important problems solved, which has some truth. But it’s hard to tell how much truth without details. His policy implications are too vague to refute, or even judge. Are higher taxes attacking startups? Or solving poverty? If attacking startups, at what threshold tax rate does this happen? And his inequality discussion is not placed in context either. Are superstar technology effects a larger driver of inequality than say, housing land costs? If bad zoning laws causing housing land costs in coastal cities to skyrocket is our biggest problem, then maybe focusing on this kind of inequality is a good idea after all.

So my working hypothesis is Paul Graham angered folks by not addressing Tyler Cowen’s laws. When I write, my shorthand check is to ask whether my post is “wikipedia worthy.” Would reading my post provide something beyond what wikipedia provides? Don’t laugh. This is a far harder challenge than one might suppose! Take a look at the wikipedia entries for Economic inequality and Income inequality in the United States. In Graham’s post, he didn’t know (or more likely decided not to discuss) potential weaknesses such as the cyclical nature of technology cycles causing superstar effects to diminish, the possibility of high tax rates not impacting economic growth as much as he implies, or the large role of housing land prices in driving inequality. As far as I know, the econ bloggers I read didn’t bother to comment on the whole thing. Which I suppose isn’t a surprise.

Additionally, there’s tone. Let me tell a story. A decade and a half ago when the startup I was at started to die, we got acqui-hired by another company. My boss (founder of the startup) became a VP at the new company. He had a brainstorming and collegial style, was great at a whiteboard, and so would often chat with people or sit in on technical design meetings. But in the new company we were at a larger scale (old company never got above 50 employees). And so at one point I pulled him aside and said if he idly commented on liking a green UI button better than a blue one, people would rush off and do it as an urgent order once he left the room. Career books tell you when you get promoted you have to adjust your style and do the new job. And stop doing the old one. Intellectually people know this. But it often creeps up on you unawares. After that talk, he changed his style a bit and all was fine. Why I’m reminded of this is I’ve read nearly all of Paul Graham’s essays, going back to 2001. They are really great. He has an original mind with an autodidact flair. He’s especially good on topics he has the deepest knowledge of, like startups. But in 15 years as Graham has risen in wealth and stature, their tone and style hasn’t changed much at all. I like Graham’s plainspoken tone, and didn’t mind his inequality post personally. But can see why it grated. In his takedown piece Ezra Klein introduced Graham as “Silicon Valley philosopher king Paul Graham.” Once you get a media title like that, you are a celebrity tech mogul whether you like it or not.

So I think this is the second part of what brought the crazy. When you write on the internet as an anonymous tech blogger, few will read you. And when they do, they are part of your tribe so cut you some slack. At least that’s my experience. As you reach the top of the food chain, 10000x more people pay attention. That’s great! But this changes the rules. It’s a change in scale so large it becomes a completely different job. You have to have thicker skin. You have to expect people will be cynical about your motives, and claim you only write things to hurt the poor. You have to assume people will not argue with what you wrote, but rather use your post to jump off and rant about their own views. And at that level it’s best to do your homework and craft your writing in accordance with Tyler Cowen’s laws, which I’d interpret as: 1) know and allude to the existing literature, 2) show your mastery by acknowledging the major flaws in your argument. Of course the major flaw in this recommendation is even that won’t always work. But it should help.

Finally, be gracious (or at least funny) once the inevitable troll attack begins.



Update: My blog is not widely read. So I didn’t expect Paul Graham to read this piece. But he did come across it and tweet about it:


Categorized as Economics

By Nathan Taylor

I blog at http://praxtime.com on tech trends and the near future. I'm on twitter as @ntaylor963.


  1. Thanks for pointing out the re-fragmentation piece. A pretty good read by someone who came of age about the same time as me, and finds himself straddling two economic models between what he grew up in and what we have now.

    I wonder if anyone is discussing it somewhere, because he tries too hard in the last paragraph to waive away the effects of policy. Companies were set free; while unions were crushed. Tax policy chose winners and losers and winning and losing uses of money. Capital became very mobile, while labor remains mainly not mobile. Physical fragmentation was encouraged and subsidized by policy.

    He talks about Microsoft as an example of a disruptor of the old order, but we all know its business model was every bit as oligarchic and monopolistic as the worst to them, more so in some ways, since its grip on its IP even more essential to the product than physical patents.

    And what’s more, although there are now many more choices of TV channels and products, there’s still a lot of crap and imitation. And in terms of products, it is still to a great extent all made by a single company (or set of carbon copy companies) in a single supply chain. And products still differentiate largely via adding “tail fins” and spending more on marketing than R&D.

    What’s really different is there is more money to make via arbitrage and financialization, than in making widgets, so the widgets have become a means to an end.

    1. I found the refragmentation post much better overall than the one on inequality. If i recall that was supposed to be his main post, and the inequality one was a side post that wound up getting all the attention.

      I’m not aware of it being discussed, though likely someone is. I think generally people agree that’s happening, but there’s some caveats.

      There are two frameworks I’m aware of. One is the technology cycles view, in which case we’re just in the “implementation” phase of the cycle, so this problem will work itself out. See my Carlota Perez post for that view.

      The second is the view that network effects and computers are fundamentally shifting the balance into smaller, winner take all firms. Which is more what he’s saying, but I think slightly different than his take. Graham sounds a bit old school to me, say Alvin Toffler 1970’s “Third Wave” demassification, e.g.

      The newer take is more about computers/network effects/disruption/winner take all. Which is similar but slightly different. And this version also ties more directly into inequality. Say

      Or from my post

      Neither of those last two all that good. But maybe now that you know keywords maybe you can find something better.

      1. Thanks for the reply. Lot of interesting stuff to digest.

        My immediate interest is how Graham tried to wave off the impact of policy (this may or may not have been his intent in the final nearly throwaway paragraph).

        It’s part of that chicken/egg question of whether our economy forms as a result of natural/social forces that sort of occur spontaneously (and then policies react or reflect those natural forces), or whether our economy takes shape as the result of deliberate policy choices made by a powerful government sector that picks winners and losers among several possible paths.

        He seems to agree on the power of policy in his opening about how the depression and WWII shaped the economy for a couple generations. But then at the end seems to be saying that forces beyond policy shaped the way it is now.

        The invention of the internet was a policy choice in fact. As was the decision to give it away to the private sector. As was the decision to provide strong long-lasting IP protections. And to lower tax rates on capital gains, and continue the interest deduction. Not to mention carried interest taxation (along with low discount rates), a policy choice that is likely responsible for contemporary M&As as much as anything.

  2. Well. Ok. If you put me on the spot, I’d say Graham’s Refragmentation post is overly pessimistic about the ability of policy, norms and institutions to mitigate the results of increasing productivity/technology. You can assist the poor directly with basic income or negative income tax. Or, what many policy planners like better, some kind of wage subsidy where if people earn below some threshold, then the government matches wages. This kind of policy avoids some of the downsides of government programs for the poor by encouraging work, but keeps their wages higher. So companies can afford to hire people and they get all the social benefits of contributing to society by holding down a job. That is to say, people and societies are very complicated and change more than we recall on generational time scales, so we should not give up on changing our laws and institutions to meet the challenges of new tech and increasing bifurcation of the market incomes technology is pushing towards.

    I have a post on technological unemployment, which is along these lines

    1. Thanks for your thoughts.

      Didn’t mean to put you on the spot necessarily. But as you alluded to above, such are the burdens of the internet blogger who leaves his comments feature open… 😉

Leave a comment

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s