Let's leave aside the definitional reading -- the dominance of unmarked terms is a consequence of what "unmarked term" means. If some phenomenon is dominant in a language (which is an empirical matter), then it's unmarked in the transcription of that language (which is a theoretical matter). And not the other way around; data comes before theory.
The other sense of the question seems to ask why those phenomena -- whatever they are -- that are dominant in a language (and therefore unmarked in its transcription) became dominant. The question is thus not about markedness at all; it's about individual language history, typology, and the comparative effectiveness of structures in a given language.
There is, let us say, a lot of variation in these matters from one speech group to another (never mind from one language to another), and while there is plenty of theorizing about source and causes, there is no agreement in practice or in principle.
Markedness is not an empirical but a typographical convention -- like the phonemic principle, it mandates using only a minimum of necessary symbols in language transcription. If something is predictable, it needs no marking; economy of marking requires that only unpredictable features be actually marked in transcription.
Markedness thus has little if any connection to origins. If it did, linguists wouldn't be able to talk about it usefully, because the origins are in dispute, and we'd just be arguing about theory, instead of processing data.