I think they're actually quite bad for morale in the long run. Why? People who have good rapport with their managers and are implicitly trusted will graduate to self-directed work within less than a year and be full-fledged "intrapreneurs" (i.e. have most of the autonomy that a startup founder would have, if not the same upside) within 2 or 3. You don't need a 20%-time policy to back that up. It naturally happens. So 20%-time policies exist to set a minimum on the allowance for self-directed work, but here's the problem: people who don't win the implicit trust of their managers get pissed off and disgusted-- regardless of whether they're working 32 or 40 or 50 hours per week-- and no side project can fix that: the only solution is for the employee to transfer before he burns out and gets fired.
A 20%-time policy dangles a carrot. It reminds the people who don't have autonomy and implicit trust of what their work lives would be like if they had better managers.
20%-time often becomes a negative space and a justification for reducing employee autonomy during the other 80%. You want to mentor interns? Maintain a legacy module you depend on? Work on a cross-hierarchical project (from a cynical manager's PoV, this looks like a careerist attempt to become a "socialite" within the company rather than remaining under one thumb)? That's what 20% time is for.
Also, 20% time might have been a real benefit in 1975 when average people didn't have access to a computer, but in 2012, most people would be happier just taking a day off to work on a side project. If you get fired or quit, you still have your side projects, but your 20%-time project ends up in /dev/null.
Finally, here's a dirty secret about what has happened in the professions. Until about 1980, it was typically understood that metered work (i.e. what your boss is directly asking you do to) expectations in the professions would be about 20 to 25 hours per week. A boss who asked for more than that, except with an aggressive once-in-a-career type deadline, was setting an unreasonable expectation. The other 20-25 hours were to be spent on off-meter work, such as networking, continued education, and research-oriented side projects that might not work out but would teach something in the process, all of which were intended to keep the professional current and improve his or her long-term career. After the slow-motion terrorist attack that was the Reagan era and aggressive corporate cost-cutting, that devolved. The new expectation, after that, became 40+ hours of metered work, and any off-meter work was to happen on your own time. (The consequence of this is that most professionals don't learn much and are replaced by younger versions of themselves when they burn out.) So "20% time" is just a ghost to remind us of what things were like back when the professions really were professions.