The most important effect is that the solutes are present (almost) only in the liquid phases. When that phase loses volume, either by freezing or boiling, the solute molecules are confined to less space. That means they have fewer states available to them- their entropy is decreased. Since the ultimate principle governing what's stable is the maximization of entropy, that favors the liquid phase over the others, lowering the freezing point and raising the boiling point.
Other effects can also be present, but so long as the solute is confined to the liquid and forms a stable solution, the net effect is always to lower the freezing point and raise the boiling point. We had a link to some lecture notes with the argument for that, but they seem to have been taken down. So here's a quick version of the argument.
Imagine that the water with the solute in it boiled at a lower temperature than pure water. That means that at that lower boiling temperature the liquid is in equilibrium with an atmospheric-pressure level of water vapor. Since the pure water isn't boiling, it's in equilibrium with a lower vapor pressure. That means that if there's some pure water near the solution, vapor from the solution goes into the pure water, making the volume of pure water grow. In fact, if you have an empty jar at the same temperature, water will leave the solution as vapor and condense in the empty jar. Now water can always leave the solution and find someplace to go- on top of the solution, for example. So the solution could, on its own, break up into a denser part and some pure water. The solution would not be stable, contradicting our assumption that it was stable.
Therefore we have proven that no stable solution in which the solute does not vaporize can have a lower boiling point than pure water.
(published on 04/10/08)