[NotOscarWilde]

> Inertia, mostly.

I agree with inertia, but I want to add a few more details from experience that could shed a bit more light onto it:

* There are not that many open journals around (see e.g. [1], probably slightly outdated); plus the throughput of a journal is generally low (the linked journal ToC has published 20 articles in 2015).

* Graduate students often want to maximize the impact of the journal they're submitting it to simply because the work itself is say of medium importance to the field.

* The review process takes a lot of time, it can even take more than 100 days. This encourages the researcher into trying as few journals as possible, which benefits the more impacted, more numerous closed journals.

* At least at my university neither the administration nor the school library promote open journals in any significant way. Which is strange, seeing as they are the ones paying the journal subscription fees.

[1]: http://theoryofcomputing.org/toclinks.html