Programmers and “productivity”

Your test doesn't work.  The problem is what one makes as an independent is a function of both one's skill at programming and one's skill at marketing.  If you are 10x as good at coding but stink at marketing you won't produce 10x the income.

I think it's a measurement problem. People who say some programmers are 10x as productive as others are stating the matter sloppily. What "productive" means in this context is generally not producing 10 times as many lines of code in a given amount of time, but in producing code that's 10 times as valuable in a given amount of time. This could be a function of being better at code re-use and so producing a 100-line program that achieves the same outcome as another programmer's 1000-line reinvent-the-wheel program. But it could also mean writing code with fewer bugs (meaning it saves the company 10x the programmer's salary in QA/customer support expenses) or it could mean making better architectural decisions that lead to more scalable, maintainable code down the line.

The problem is that only an above-average programmer can identify other above-average programmers ex ante--and even that's going to be iffy. In some cases, a non-programmer might be able to identify the best programmers ex post--say by counting the number of problems the QA team finds--but even that will be an imperfect measure since a dozen bugs involving minor corner cases is less serious than a deep show-stopper bug that's detected just before the project ships. And of course in any nontrivial code base, it's going to be hard to be sure which programmers are responsible for any given bug, to say nothing of more subtle issues with good architectural decisions.

This is a persistent problem with skilled professions. When I take my car to the shop, I don't have a clue if the mechanic did a good job or a lousy job of fixing it--even if it breaks later that doesn't necessarily mean the previous fix was incompetent. When I hire a lawyer, I don't know if I've made a good case until I win or lose my case--and even then I can't be sure if it was an easy case or a hopeless one. Likewise, if I hire a surgeon I don't know if he's good until I wake up after the surgery.

The fundamental problem is that what you're buying in all these cases is knowledge and judgement, and you can't evaluate those unless you already have them yourself. So the vast majority of the market under-values the best talent because it doesn't know how to identify it. And the small minority of companies that do know how to identify the best talent--like Google--pay a little more than average and make out like bandits.

If the only way to be a(n economically) productive programmer is to start a successful startup, then programming itself has no productivity. You're saying the only way to be a productive programmer is to develop skills in business. By your argument, productive programmers do not exist. What if you're right? What if productive programmers don't exist? What if Bill Gates is not an epic productive programmer, but an epic businessman who happens to be literate in programming? What if the same is true for all these guys?

I think you're right, and productive programmers don't exist, and further that programmers don't exist. I think programming is just a type of literacy. Most "programmers" are beauraucrats or assembly-line workers. Many are designers. A few are entrepreneurs. Only the entrepreneurs make real money. There's no 10x difference there. There's a 10,000x difference. You could take programming out of the equation and make the same statement about any area of the economy.

Thank you for writing so lucidly about this. As and economist, the original question bothered me a lot, but the words didn't come. I agree wholeheartedly with this assessment.