Yeah, electrolytics are healthiest near ratings, say 60-90% of ratings. Less and you're simply wasting money and space; more and you're risking damage*.
*However, it's quite possible to reform a capacitor beyond its rated voltage. I've heard of 200V parts used at 250, 300VDC. The problem is, as the oxide layer grows (you're anodizing the aluminum as you increase the voltage slowly), the metal goes away, so the capacitance value and ESR change. Too much and the foil itself will be oxidized through, significantly reducing capacitance and increasing ESR. I don't know at what point this normally occurs, if it's 50 or 100% or what beyond rated voltage.
Ceramic capacitors depend on type: C0G (historically, NP0, or erroneously, NPO) and similar type I dielectrics are stable and work up to basically any voltage. I've operated 50V discs at 500V AC before; they don't even heat up (oh, I was also running an ampere through them at that frequency). Not that this is recommended operation for reliability, but needless to say they can handle it.
Type II dielectrics have larger values and poor voltage dependency. Typically an X7R type will fail its tolerance (10 or 20%) within just 30% of rated voltage. Thus, you typically specify a 35 or 50V part for 12V operation, and etc. They also age (logarithmically over time), so they aren't suitable for timing or precision applications.
Film capacitors are almost as stable and low loss as C0G ceramics. They're available in much larger values, especially at higher voltages. They should be chosen for equal or better voltage ratings, as with electrolytics. Typically, some nominal overvoltage is acceptable, but an excess will cause incremental damage (self healing) or failure.
Tim