In theory do we expect lenses of identical specs and quality to be cheaper when designed for smaller sensors? My expectation is that the answer is yes, because all of the things that lenses struggle to control — chromatic aberration, vignetting, distortion, etc. — are exposed over a smaller central region. Or, put another way, if you take a full-frame image and center crop it you always end up with less of these bad things. I'm also thinking of the exit image quality problem as sort of an inverse of the entry image problem, and we know that larger apertures are optically more difficult.
Now what about in practice: Is the answer different? For example, it could be "yes" in theory and "no" in practice because everyone designs and builds to full-frame specs anyway, and economies of scale cancel out whatever theoretical savings there might be for lenses designed for smaller sensors.
Update: Some of the answers so far assume that the complexity of the lens must be constant, and thus only address the marginal cost of reducing the size. Obviously size/material savings are not going to be significant when talking about tradeoffs like this in quality glass.
I was thinking more along the lines of being able to reduce the quality and maybe even number of the lens components. Again, because I have a hard time visualizing the optics of the exit cone I hope the engineering is roughly analogous to the input side of the lens: We know that a high-quality large-aperture lens faces the biggest challenges shooting wide open because it has to take incident light and bend it through more extreme angles without introducing chromatic aberration, distortion, etc. If you're willing to shoot a prime at, say, f/10+ then a lot of those challenges go away. So I'm assuming that on the exit side it's somewhat similar: the larger the circle the more you have to work the optics to keep quality constant edge-to-edge. Just as closing the aperture relieves you of challenges in maintaining quality, doesn't reducing the angle of the exit cone we care about by reducing the sensor size make preserving quality across the sensor easier?
Another Update: Let us assume that we're talking about sensors with the same pixel pitch and technology. I.e., we aren't asking the lenses for smaller sensors to compensate for any shortcomings that often accompany reduced sensor size. Though, again, I assume that in practice "good" glass is generally running close to the upper (diffraction) limit of resolution.