Wednesday, September 24, 2003

 

More on Zipf volitility


Ben Hyde (bhyde at pobox dot com) writes:
The is a little data about zipf volatility for incomes, but it is almost useless because it deals with quintiles so there isn't enough resolution to really think clearly about it.

The most interesting data about zipf volatility is from firm size data. Firms are power-law distributed along most metrics [1] (income, size, employees, etc). There is also data on year to year variation in their rankings.

So the data shows that grow is largely independent of size; i.e. there doesn't appear to be tendency for firms of size A to do better or worse than firms of size B.

But, there is a significant variance of growth rates. Small firms have extremely variable growth rates, and huge firms have little variation.
I.e. large firms tend to grow with the economy and there is little variation from that, while small firms suffer huge variations from one time frame to the the the next. The distribution of the variance is quite sharp, a double exponential.

I think this is another example of the scale free nature of these things - i.e. there isn't a typical firm, blog, website. You can't sample your neighbors to get an accurate model of the behavior of the entire set.

[1] http://linkage.rockefeller.edu/wli/zipf/axtell01.pdf



Comments: Post a Comment

This page is powered by Blogger. Isn't yours?