Comment by yummyfajitas | Hacker News Reader

Comment by yummyfajitas | Hacker News Reader

Parent Full thread

If you like academic papers, there is a great one by Claudia Goldin. Her topic of interest is gender pay gaps (boring), but it goes into detail on pay nonlinearities which is quite interesting. Basically, she goes into a lot of literature and does some work herself showing that for many jobs, work output is a nonlinear (e.g. y=x^2) function of hours worked.

A nonlinear pay function is generally related directly to nonlinear work output. Her working example is pharmacists - you can easily swap two half-time pharmacists for 1 full time one or vice versa.

On the other hand, consider your top trader - his job is keeping the entire strategy in his head (roughly a 30 hour/week job) and being around when the market is open (another 35 hours/week). You can't replace him with 2 30 hour/week traders, each of whom keeps half the strategy in their head and are available half the time.

http://www.aeaweb.org/aea/2014conference/program/retrieve.ph...

Flexible hours will, of course, be more prevalent in jobs and industries where Output(hours) is a linear function.

(Gender pay gaps are not intrinsically boring, but the inherent "turn of your mind because politics" makes it far less interesting.)

Ties in well with Paul Graham's essay "Mind the Gap":

It may seem unlikely in principle that one individual could really generate so much more wealth than another. The key to this mystery is to revisit that question, are they really worth 100 of us? Would a basketball team trade one of their players for 100 random people? What would Apple's next product look like if you replaced Steve Jobs with a committee of 100 random people? [6] These things don't scale linearly. Perhaps the CEO or the professional athlete has only ten times (whatever that means) the skill and determination of an ordinary person. But it makes all the difference that it's concentrated in one individual.

http://paulgraham.com/gap.html

This Paul Graham's essay is hands down the best argument for libertarianism I have ever read. Yet, even if there are 3 orders of magnitude differences in productivity among different people (and I doubt that), this doesn't entirely account for some 7 or so magnitudes of wealth and income disparities we see in the real world.

The problem is, in modern society, almost nobody deserves to be paid for the work they do (I don't know about you, but my life would be exactly the same without Steve Jobs and random American athletes). We have automated almost everything, and the biggest chunk of productivity is due to capital (and innovations of our ancestors). So the question, how is this capital controlled and distributed, becomes central. (That's why many people here so oppose intellectual property - essentially the privatization of public information, which is analogous to land enclosures during 18th century.)

So I think, if someone genuinely thinks that people should be paid based on value which they personally bring to society, he shoots himself to the foot. This value is probably too small for a living wage. That doesn't mean the improvements are not useful; it's just all the low hanging fruits have already been harvested.

even if there are 3 orders of magnitude differences in productivity among different people (and I doubt that)

That's not very hard to imagine, at all. Take a guy who flunked out of 8th grade due to his inability to read. Point him at a computer. Ask him to fix a problem in a Ruby on Rails application which you've benchmarked an intermediate engineer at requiring 1 day. If the problem is solved within 1,000 days, he's within 3 orders magnitude of that engineer's productivity.

I would assume "nonlinear" output means that their productivity increases the more hours they spend. Any job requires ramp-up time, some more than others, but it does not mean the output in non-linear. All your example illustrates is some jobs (like software development) can't be brute-forced by hiring more people irrespective of skill - two mediocre programmers are not a replacement for one good one.

It means that the person's output is not simply a multiple of the number of hours that they work, but that for each unit of time they work their output increases compared to the previous unit.

For example:

    Linear
    output = time * k

    Non-Linear
    output = time ^ k

In practice it means that for jobs that are linear, it doesn't matter how many hours any individual works as long as the total hours worked remains the same - assuming that they're all of the same skill level. Doctors, for example, are mostly shift workers so the output of one doctor working for 12hrs is pretty much the same as two doctors working six hours each.

For a job that has non-linear output, two people of the same skill who each work six hours, will be less productive than one person of the that skill working twelve hours. Easy enough to see:

    12 ^ 1.5 = 41.6
    2 * 6 ^ 1.5 = 29.4

In that case, splitting the job between two people results in an output that's 70% of the output of that job being done by one person.

Basically, it's saying that many workers are not fungible assets, no matter that they share the same level of skill.

Another thing to consider (for software engineers at least) is hours per day, vs days per week.

It can take a few hours to ramp up productivity and get in the zone. On the other hand, in a well organized team if a few people take a day or two off, not much would be lost.

Yeah -- "nonlinear" isn't wrong here, but a more precise term would be "superlinear" (or "disproportionate" or "super-proportionate"). IOW, what drives the result is that one worker with N hours does more than m workers each with N/m hours.

If you (bizarrely) had sub-linear ("under-proportionate") returns to hours worked -- like logarithmic or square root function -- then it would make sense to have as many workers as possible and work as little as possible.

For a lot of knowledge work I think it's closer to a bell curve, in that if you work too few hours you'll be swamped by overhead, but if you work too many hours, your per-hour productivity drops due to lack of focus caused by insufficient rest or dropping motivation. I thought it was common knowledge that overworked developers create more bugs, so I wonder how a trader is able to work 65 hours a week without performance suffering in the long run. Or could you rotate traders in a one week on / one week off cycle?

It is 38 pages long so I have to ask. Would you mind telling me what her inferences were as to this non-linearity and gender ?

Women work less than men and nonlinearity exacerbates the effect this has on pay. The more jobs become commoditized, the more gender equality and flex time we will have.

Reply on news.ycombinator.com