I think scale might matter too. Companies are partly public goods problems. Each individual's hard work benefits the whole company. As a result they will undersupply hard work. If the company is size N, they only share 1/N of the total benefit they provide. So, the public goods problem bites harder when companies get big - even holding the level of competition in the market constant.
The graph of the distribution for value per person must be wierd, and definitely not flat. Maybe based on a power law, but I am unsure how to model the people that provide negative value, for example bad managers.
Well, the question is how much of the output they get :-)
Though actually note that the classic Mancur Olson result is "the exploitation of the strong by the weak". That is, the person who has the largest share of output contributes disproportionately much to the public good.
Simple proof: write individual i's marginal benefit from the company's total effort X as
s_i f(X) - x_i
where s_i is i's share of the output f(X), and x_i is i's own effort.
Taking everyone else's effort as given, i will equate marginal benefit of own effort to marginal cost, so that total effort solves
s_i f'(X) = 1
equivalently
f'(X) = 1/s_i
Now suppose the person with the largest share satisfies the above. Then everyone with a smaller share must have
f'(X) < 1/s_i
and therefore, if they are contributing more than the minimum possible, they have an incentive to reduce their contribution. So, this extreme version has that only the largest "shareholder" does any work. (A more realistic model has a convex cost of effort, so smaller shareholders do something, but they still contribute less than their output share.)
Your assumption is that the organisation is a cooperative - the employees get a share of the output of the organisation.
Intel is not a cooperative.
Fungible employees (the majority?) are paid a market clearing price (salary) for their effort. The company can derive much more value from an employee and the employee earns very little of the excess (perhaps bonuses to align working incentives).
Supposedly, in knowledge work like programming, the distribution follows a pareto distribution: for n people, sqrt(n) produce half the productive output.
