I read this in Pulleyblank's Outline of Classical Chinese Grammar, where he said it is a maybe.
In Old Chinese (OC), they have the following initial and rhyme groups with Baxter-Sagart reconstructions:
不 p + juw (*pə)
乎 h + u (*ɢˤa)
夫 b + ju (*ba)
My main concern is that 夫 fu2 has a voiced initial in OC [b] (note that it's not f that I am taking about, since f was a new development from the b group initials in late Middle Chinese), while 不 has a unvoiced one [p].
Is there any other theory, or other evidence so that this theory stands?
UPDATE
As @水巷孑蠻 mentioned, the well-recognized sound fusion includes 不 + 可 > 叵, which the initials differ only in aspiration. If difference in aspiration is not a problem, then I think difference in voicing is also not a problem.
But @水巷孑蠻 raised another problem, which is that if we rewrite 夫 to 不乎, then the sentence reads odd. For example, 吾歌,可夫? -> 吾歌,可不乎?
Yet I did find some cases in later documents the following examples:
不乎, 宰我以仁者必濟人於患難,故問有仁人墮井將自投下,從而出之不乎?
而父卒祖母後卒,當服三年不乎?