I usually avoid politics, but since we now seem to have an administration that's open to suggestions, I'm going to risk making one. The single biggest thing the government could do to increase the number of startups in this country is a policy that would cost nothing: establish a new class of visa for startup founders.
The biggest constraint on the number of new startups that get created in the US is not tax policy or employment law or even Sarbanes-Oxley. It's that we won't let the people who want to start them into the country.
Letting just 10,000 startup founders into the country each year could have a visible effect on the economy. If we assume 4 people per startup, which is probably an overestimate, that's 2500 new companies. Each year. They wouldn't all grow as big as Google, but out of 2500 some would come close.
By definition these 10,000 founders wouldn't be taking jobs from Americans: it could be part of the terms of the visa that they couldn't work for existing companies, only new ones they'd founded. In fact they'd cause there to be
more jobs for Americans, because the companies they started would hire more employees as they grew.
The tricky part might seem to be how one defined a startup. But that could be solved quite easily: let the market decide. Startup investors work hard to find the best startups. The government could not do better than to piggyback on their expertise, and use investment by recognized startup investors as the test of whether a company was a real startup.
How would the government decide who's a startup investor? The same way they decide what counts as a university for student visas. We'll establish our own accreditation procedure. We know who one another are.
10,000 people is a drop in the bucket by immigration standards, but would represent a huge increase in the pool of startup founders. I think this would have such a visible effect on the economy that it would make the legislator who introduced the bill famous. The only way to know for sure would be to try it, and that would cost practically nothing.
Thanks to Trevor Blackwell, Paul Buchheit, Jeff Clavier, David Hornik, Jessica Livingston, Greg Mcadoo, Aydin Senkut, and Fred Wilson for reading drafts of this.