The 31 people on the cover of Parade magazine’s “What People Earn” issue make an average income of 5.8 Million a year. They over sample the elities and under sample the poor.
Meanwhile I want to rant a bit on Dunbar’s number. Dunbar’s number is nicely explained here. The theory goes that you weight the monkey’s noodle and then you plot that versus group sizes you find a correlation. Humans are on the perimeter of this plot (aren’t we?). If you want to play along then our group size is then around a 150. I guess that might work if we were monkeys.
But we aren’t. Language changes the game. It’s a lot easier to solve coordination problems with a bit of language. Technology changes the game. Rolodex, Outlook, oh and the printing press.
Specialization changes the game. In both the business models that give rise to income and the social models that give rise to group structure humans are pretty clever at finding frameworks that enable them to work around limits that might otherwise put a limit on the upper bounds of these things. The distribution of group sizes, the distribution of social contacts, the distribution of wealth do not appear to be capped by physical limitations. My first problem with Dunbar’s number is that it posits the existence of an upper limit when the data doesn’t seem to show any such limit. For example we don’t see any sign that village/city/nation sizes are capped by some limit.
My second problem with Dunbar’s number is that everything seen of it suggests that its fans are uninformed by the power-law distribution in the underlying data. Which means that they aren’t aware of the extremely skew’d distribution of affiliations in real populations.
Dunbar’s number is a measure of social capital. Social capital, like economic capital is not a zero sum game nor is it uniformly distributed. Reasoning by analogy or on the basis of small samples sets is almost always extremely misleading in the presence of highly skew’d distributions.
The bottom 20% of households in the United states captured 3.4% of the total income. Think about that! Should the same distribution holds for affiliations, and I see no reason to think otherwise, that means that one out of five people participate less than 4% of the the benefits generated by the social fabric.
Parade? It misleads you about the elite portions of the social network and the long tail. So does Dunbar’s number. Parade magazine isn’t pretending to be professional, it’s just trying to sell you mattresses, celebrities, and herbal remedies. Dunbar’s number on the other hand is just, well, wrong.
I think the point of my article on the Dunbar Number is that there is a limit to the number of people who can share unstructred trust. Unstructured trust is required of various kinds of groups, in particular survival groups, but also applies to a variety of social groups. There is ample evidence of some sort of limit to unstructured trust as shown by research on primitive tribes, terrorist organizations, mafias, and military organizations — you don’t need brain size to show that.
Communication and culture *can* provide a context that can allow for more structure and thus larger groups, but the nature of that stucture imposes some costs. There is a price for coordination, whether it be time, or through architectural problems like the stovepiping problem in bureaucracy.
In addition, unstructured trust is more emotional and thus seems deeper, and to many people more satisfying.
I like the term “unstructured trust” Care to flesh out a definition?
Regarding the evidence I don’t trust the experimental designs. Consider a seemingly straight forward example. What’s the typical group size in open source? If you go draw off answer from source forge. The only thing that’s really compelling in that data is how skewed the distribution is. So skewed that the term average and median becomes effectively meaningless. To get a group size number out people ignore first discard the tiny groups. They then ignore the level of involvement. They then ignore the high probability that the large groups aren’t in source forge at all. What’s the data worth at that point? Given that is one of the very best pools of data on group size available in the history of mankind it all gets pretty ungrounded.