Yes, starting a new scientific journal is relatively easy, but getting a prestige is a chicken-and-egg problem.
Scientists get points for publishing in prestigious journals with high impact factor; that means in journals where articles are often cited. If you start a new journal, there is no reward for a scientist that would give you their article. A scientist can publish an article only once, and getting points for publishing articles is very important for them (it influences if they get grants or tenure). In other words, the system strongly penalizes scientists for publishing in new journals. This is what reduces competition.
So when you start a new journal, you usually pay scientists for publishing in your journal. And you have to send a lot of "spam" to make yourself known, which makes you rather unpopular (you advertize that you have zero prestige). It takes a time to slowly increase your prestige.
When you are at the other end of scale, the situation is opposite. You are famous, and your articles are good, therefore scientists get a lot of points for publishing in your journal. You can refuse less-than-perfect articles, which increases your score (impact factor = number of citations per article). Everyone works for free, and sometimes scientists even pay you for publishing their excellent articles.
The system by its definition supports monopolies, because players are rewarded for cooperation with established winners. Not only implicitly (the winners probably have better quality selection, larger audience, etc), but they get some explicit points for the sole fact that they cooperated with established winners. For example, some grants may be awarded based on your articles in "current contents" journals. As a scientist you often do not know which grants you will need in future, so you try to maximize all criteria by ignoring the non-winners, even if they would treat you more nicely.
Thanks for clarifying how this works!
link
Razib Khan found this paragraph rather striking (who is reminded of this episode of South Park) and I would tend to agree that its a rather convincing argument.
Are publishers really so successful as rent seekers or is there something the original article is missing here? Also what useful strategies would LWers recommend to help minimize costs for someone trying to practice the virtue of scholarship? The obvious suggestions (implied in the article) seem to be emailing authors (and perhaps those suscribed) asking for the papers and acquiring and paying for membership in some libraries.
Another obvious option is using ... liberated databases of such academic papers.
Edit: Just wondering, has this been discussed before on Lesswrong?