Andy the pugh is correct on this point. If the nut material has a yield strength equal to or higher than the bolt, a very large portion of the load will be concentrated on the first thread.
The nut material should have a lower yield strength than the bolt. You should generally match the nominal property class of the nut with the nominal property class of the bolt (or use a weaker nut, though this often won't give you enough tensile capacity). E.g., the material of a property class 8.8 nut, even though you may not know, actually has a lower material yield strength than the material of a property class 8.8 bolt, in order to allow the first few threads of the nut to slightly yield prior to bolt thread yielding, thereby redistributing the load more evenly over a larger number of threads.
In other words, an 8.8 nut goes with an 8.8 bolt, or a nominal "1040 MPa" nut goes with a nominal "1040 MPa" bolt, as a matched set. Stated differently, you shouldn't use a nut having a higher property class with a bolt having a lower property class.
So andy the pugh is correct. You want the nut material to have a lower yield strength than the bolt material. But James Jennings is correct in that you *do* want sufficient thread engagement length so that the nut (or tapped hole) will have a higher overall tensile capacity than the bolt, thereby allowing a bolt to break prior to nut stripping (in spite of the lower yield strength of the nut material).
I agree with James Jennings in that metric standard fastener threads are inherently better and more fatigue resistant than standard imperial (UNC, UNF) threads due to their improved thread root radius design. However, I disagree that Shigley "is not a good source, while Bickford is a very good source." Actually, the reverse is true. Shigley is an excellent source, whereas Bickford is an overpriced anthology of simplistic, redundant, fairly low-quality information. I do agree that VDI 2230 is a good source.