What is the motivation to start a new journal in a crowded field?

Being able to set editorial policies can have a significant impact on how science is done in the field. Some examples of such policies include: encouraging the use of preprint services, mandatory data/code sharing, transparent peer review process, encouraging replication studies, etc. For the big names who are not satisfied with simply setting the direction of their own labs, setting up a new journal can be another ideal career ambition.

Great question. I've asked various people about this and thought about it myself (I used to work in publishing). My sense of this is the correct question is less "why?" and more "why not?". Here're some factors in the macro-publishing view:

• Global publishing is increasing and has been increasing for a while. This report for example finds that "The number of articles published each year and the number of journals have both grown steadily for over two centuries, by about 3% and 3.5% per year respectively"; furthermore, "The reason is the equally persistent growth in the number of researchers, which has also grown at about 3% per year and now stands at between 7 and 9 million."
• It's common, especially for researchers in first-world countries, to think that good papers will be published in good journals and everything else isn't worth reading (you didn't write exactly this, but by writing "highly regarded" journals, you imply something similar). Lucky them - many authors from developing countries will never publish in one of these top journals. For example in many developing countries a performance indicator is number of papers published in SCI-indexed journals (example from a brief Google search). If that sounds like nonsense to you because all your papers are already published in SCI-indexed journals, you're one of these privileged authors.
• Why can't these authors publish in top journals? From the journal's point of view, because the papers aren't very interesting/novel/high-impact. From the authors' point of view, it could be because they aren't as capable as those people working in first-world countries; equally possible is that they just don't have as many resources. For example without access to LHC data, one is not likely to make breakthroughs in particle physics. So one ends up with boring, low-novelty papers that will never get published in top journals.
• These papers still have to go somewhere, however. Where there's demand, someone will come up with a way.

A few more factors from the publisher's point of view:

• You may not be aware that many of these not-so-good journals are actually solidly unprofitable. It's absolutely not a money grab. The flip side is, they also don't lose that much money. Sure their revenue is low, but their expenses are also low (usually because they publish relatively few papers).
• Publishers are well set-up to cope with changes in the number of papers received (they have to - the number of accepted papers is not by any measure a constant). That's why many desk editors handle both journals and books. When there are fewer papers than average they spend more time on books; when there are more papers they outsource to freelancers. The result is that estimating the cost of a new journal is a difficult process. There's some estimate of how much each paper is going to cost, but the last journal I helped set up literally cost nothing in extra personnel. It was simply assigned to one of the desk editors and he incorporated it into his duties.
• The cost can become substantial if the journal starts publishing a lot of papers; however if the journal actually manages to do that then chances are it will become profitable. (High volume is a key driver in getting indexed by databases such as Web of Science, which in turn is a key driver in getting more submissions & subscriptions, etc.)
• Based on that then the cost of the publisher setting up a new journal is pretty low. Therefore if a group of academics come along and suggest starting a new journal, the publisher is likely to be interested. Why not? You risk a relatively small amount in the hope that the journal works out, when it does become lucrative (since it's a reliable and recurring source of revenue).
• Besides, even if the journal fails, you've hopefully built contacts with a few more people who might collaborate with you in the future, e.g. by writing books.

And then there's the academic end.

• Sometimes the would-be editors think they've genuinely identified a niche in the journal market.
• Sometimes they're switching publishers. For example the Journal of Topology was started by the disaffected editorial board of Topology when they switched publishers.
• Other times they're doing it for prestige. For example, another journal I helped set up involved the publisher identifying what they thought was a niche and writing to an expert in that niche. The professor was agreeable but wanted to name two of his postdocs as associate editors. Talking to the two postdocs, it sounded like they were doing it in part because it's prestigious, and in part because it's a novel challenge they've never had to deal with before.
• As for the professor himself, he was paid a fairly substantial honorarium to handle the journal: several thousand USD a year. I don't think he agreed to the journal because of this money, he probably also felt it was a reasonable idea and he could contribute to it, but the money probably influenced his decision.

Edit: One more thing - being able to point to a successful journal later in one's career and be able to truthfully say "I had a hand in starting that journal" is hugely satisfying.