I was recently told about a manager of a ~500 person bank division decades ago. Seems this manager created an interesting info equilibrium. In his equilibrium, this manager could call anyone in the division at any time and get frank answers about their work. The manager would not tell anyone about the call, and he would viciously punish both anyone who ever lied to him in such a call, and anyone who punished anyone for giving frank call answers. So when you went to present something to this manager, you knew he might have called any of your subordinates for frank answers about your presentation.
It probably wouldn’t work to have an equilibrium where there are many people with this power to call anyone and get a frank answer. And it wouldn’t work if the organization was so big that the manager knew little about each person he might call. But at least within certain scale limitations, this is a way to cut through information barriers to get frank assessments on key issues into the hands of a pivotal decision-maker.
So would it work to nest this structure two levels deep? That is, could 200,000 people be organized into 400 divisions each like this, where the head of the whole thing could always call any division head and get frank answers?
Added 3Apr: My anonymous source elaborates:
People gave frank answers because they were seeking credit approval to credit submissions.
The submissions were the outcome of protracted and detailed documented process of sequential deliberation. They were also subject to annual audit and also specialist lending inspection.Verification was embedded into the process.
These were complex high value transactions that justified a significant investment of time.
The info king was the officer with the boards’ delegated discretion. The transactions were complex and/or high value and required elements of judgement.
The info king knew that the issue, given the high stakes for the transaction sponsors, was not fraud (because the verification processes protected against this), the issue was bias and nuance, and that required a subtler verification process.
That is interesting, and it suggests that there is a maximum size for optimum efficiency of around 400 employees.
What that implies is that companies larger than 400 employees are not more successful because of greater efficiency, but rather because of greater “power” due to size alone.
This suggests that limiting company size is a way to improve efficiency in the entire economy.
Competition could be limited to competition based on efficiency through regulation, but that doesn't work in real life because large companies have more political power and they use that political power to enforce regulations that favor the type of power they have, not efficiency.
Restricting company size would be simple to implement, simply increase taxes based on employee number above a certain level (the opposite of what happens now).
Sorry for the late comment on this thread, but I haven't visited for a while, but this is an interesting question to me. As noted by Gwern, definitely Coase's transactions theory is relevant here. I actually also have practical experience on this as well. A company I worked for actually was taken over by another large firm and the then CEO had exactly this philosophy. He believed that with business units greater in size than 400 people or so the leader of the unit lost touch with the business and so could not optimise things. So for a while we had a company consisting of around, I think, 40 business units. While it was probably true that the individual leaders had a very good hand on their individual business this meant that the top leadership basically had an impossible task in developing overall strategy and implementing overall corporate initiatives. There was also a lot of duplication of effort and local marketing or government relationships were not be handled very effectively. So after not very much time we abandoned the idea and went to much bigger units.