I agree with the author's thesis but I'm bothered by the way the numbers are presented. Other articles in favor of making immigration for skilled workers easier usually are guilty of the same number fudging.
In the article: "My research team documented that one quarter (http://ssrn.com/abstract=990152) of all technology and engineering startups nationwide from 1995 to 2005 were started by immigrants."
The linked source says it more precisely: "We found there was at least one immigrant key founder in 25.3% of all engineering and technology companies established in the U.S. between 1995 and 2005 inclusive."
The key words in the second quote are "at least one immigrant key founder"; there are typically 2-4 founders in a startup (YC's avg is 2.5). If we just calculate what fraction of founders are immigrants across all startups (=total_immigrant_founders/total_founders), the number should only be ~ 11% [1], which sounds less dramatic but is more accurate in terms of how much credit immigrant founders deserve for the startups.
Similarly "more than 25% of U.S. global patents had authors who were born abroad" should really mention what fraction of inventors listed on patents are foreign-born, since patents typically have many coinventors.
[1] - There's probably actual data on this somewhere, but if we assume the probability p of a founder being immigrant is independent of other founders, and the number of founders per startup = 2.5:
In the article: "My research team documented that one quarter (http://ssrn.com/abstract=990152) of all technology and engineering startups nationwide from 1995 to 2005 were started by immigrants."
The linked source says it more precisely: "We found there was at least one immigrant key founder in 25.3% of all engineering and technology companies established in the U.S. between 1995 and 2005 inclusive."
The key words in the second quote are "at least one immigrant key founder"; there are typically 2-4 founders in a startup (YC's avg is 2.5). If we just calculate what fraction of founders are immigrants across all startups (=total_immigrant_founders/total_founders), the number should only be ~ 11% [1], which sounds less dramatic but is more accurate in terms of how much credit immigrant founders deserve for the startups.
Similarly "more than 25% of U.S. global patents had authors who were born abroad" should really mention what fraction of inventors listed on patents are foreign-born, since patents typically have many coinventors.
[1] - There's probably actual data on this somewhere, but if we assume the probability p of a founder being immigrant is independent of other founders, and the number of founders per startup = 2.5:
=> 1 - (1-p)^founders_per_startup = probability_that_startup_has_at_least_1_immigrant_founder
=> 1 - (1-p)^2.5 = .25
=> p = .1087