VC Matthew McCall has made some recent posts here and here on incubators that provide some clues.
The notion of co-locating start-ups together makes sense at a high level, but fails to work on the ground. Networking, mentoring, sharing, etc all work when the entities/companies involved are of similiar caliber, have similiar issues and are playing in the big leagues. However, often, you have a building full of inexperienced or small entrepreneurs attacking niche issues in a broad array of industries. There is not a lot that they can teach each other. If there is a breakout company, it is so focused on building its business, that it doesn't have the time or interest to pull up the other inhabitants with it. In fact, it will often move out for bigger space. It also experiences asymetrical benefits...it gives a lot of advice but gets limited value in return. Also, the companies in an incubator are often so diverse that they are each facing fairly unique issues.
Companies need to have global visibility in order to see what is going on in their industry...best practices, competitor challenges, etc. A local incubator can not give them this visibility or connectivity. By definition, the incubator is local and somewhat isolated.
Shared services and reduced rent are two potential benefits, but they don't make or break the success of a company. Many firms start up virtually or in low rent areas. They can provision internet, telephones, etc as needed.
Arun Natarajan is the Founder of Venture Intelligence, which tracks private equity and venture capital in India and Indian-founded companies worldwide. View sample issues of Venture Intelligence India newsletters and reports.