We can do that only if population is not very well mixed and Y haplogroups are recently coming from well known and defined ethnicities, and his autosomal DNA still could be mostly from that group. It is so rare that I can't give you a real life example. It is theoretical possibility only. The main issue is that ethnicity is a cultural phenomenon, not a genetic one. For example we can have a person living in Turkey having half of European genome, and this person is a patriotic Turk and Sunni Muslim. On top of it his language is actually from Central Asia, not from Near East, never mind Europe. This person also has not more than 10% original Central Asian Turk DNA. See, it is complicated even on autosomal level. His Y haplogroup could have come from Turkic, Near Eastern or European great, great... grand father. Complicated.
Not sure I am convinced by your answer LeBrok. I am not talking here about cultural ethnicity but genetic ethnicity like the ones based on autosomal dna in geno 2.0 ?
you said we can do that only if population is not well mixed. could you please be more explicit ?