I am doing research about collaboration in projects.
If a company has an innovative idea that they want to put on the market, but they don't have the knowledge to develop the idea, then they are dependent on another company's knowledge. Most of the time the company(s) with the innovative idea won't be willing to collaborate with other companies that have the knowledge they need.
I think your question falls under the more general umbrella of collective intelligence - the underlying forces involved are similar. This is a very interesting area of development - how do we extend the Wikipedia model of collaboration and Creative Commons to the business world in industrial and enterprise solutions? Can we do this with patents? Taking this question further, why are are individuals more likely to collaborate than business entities (assuming the premise is true)? Looking for these distinctions may help you find the answers.
What is the reason they won't collaborate? Some reasons that come to mind are:
-IP barriers and theft risk (company A contributes an idea based on patented technology that company B does not have a license to)
Scott Kurnit (founder of About.com back in the day, well respected industry vet) was just interviewed in Business Insider on the subject of why Corporate Joint Ventures are generally doomed to fail, specifically talking about the ongoing questions about the future of Hulu:
SCOTT KURNIT: JVs Like Hulu Are Slow-Motion Train Wrecks The point being - joint ventures are designed to play to the strengths of the participants (as you describe above), but are generally hamstrung by the priorities of each organization backing the venture.
First, you might want to top up your view of the subject with the related area of "open innovation." You'll find there are specific companies, industries and technology areas where collaborating on innovations is either locally present or is the norm.
Somewhat related is the idea of a cluster - where a particular geography becomes home to a variety of companies in the same broad market space. There's plenty of writing in this area - not least because it's sometimes very hard to pin down what's going on and therefore much scope for competitive theories to be developed, tested and docuemnted. Formal collaborations may exist in clusters, highly specialised companies may emerge that depend on the total ecosystem, or porosity may see projects moving forward through people moving between companies - which is a kind of systemic collaboration (the cluster generates benefits that are distributed and sustained over time, but individual companies may never collaborate).
My own experience of formal joint ventures between large organisations has been mixed - and mainly negative. Almost by definition, where competences are different (hopefully 'complementary' - but different nonetheless) it's quite likely that management styles, attitudes to risk, adaptability and a host of other factors will also differ. That is likely to produce tensions between the JV and its parents, and within the JV itself.
More speculatively, it seems to me that business schools primarily position collaboration as an answer to a problem for one partner. I have everything but geographic reach - so I collaborate. I have customers and delivery but a weak product pipeline - so I collaborate. And, as soon as I can, I extract myself from or else buy out the partner. If the thinking is asymmetric, many potential collaborations may be prevented because each party wants to collaborate only from that 'upper partner' position.
I take a different perspective on this. Joint Venture and partnerships occur and collobaration occurs often in certain industries. Some fail but not all.
It's like why do most inventions result in no inovation and then saying inventions don't create inovation. Or most startups fail and then saying all startups fail.