We present a systematic study of the role of elemental identity in determining electronic, energetic, and geometric properties of representative A 28 B 28 , A 30 B 30 , and A 36 B 36 III-V (A=B, Al, Ga, or In and B=N, P, or As) and II-VI (A=Zn or Cd and B=S or Se) fullerene allotropes. A simple descriptor comprised of electronegativity differences and covalent radii captures the relative fullerene stability with respect to a nanoparticle reference, and we demonstrate transferability to group IV A 72 (A=C, Si, or Ge) fullerenes. We identify the source of relative stability of the four-and six-membered-ring-containing A 36 B 36 and A 28 B 28 fullerene allotropes to the less stable, five-membered-ring containing A 30 B 30 allotrope. Relative energies of hydrogenpassivated single ring models explain why the even-membered ring structures are typically more stable than the A 30 B 30 fullerene, despite analogies to the well-known C 60 allotrope. The ring strain penalty in the four-membered ring is comparable to or smaller than the nonpolar bond penalty in five-membered rings for some materials, and, more importantly, five-membered rings are more numerous in A 30 B 30 than four-membered rings in A 36 B 36 or A 28 B 28 allotropes. Overall, we demonstrate a path forward for predicting the relative stability of fullerene allotropes and isomers of arbitrary shape, size, and elemental composition.