I have encountered, I believe mostly in works from Generativist phonological traditions along the lines of Chomsky and Halle's The Sound Pattern of English, the idea that words like potency, latency, decency, presidency end at some level of morpho-phonology with the sound /j/ rather than some nuclear vowel sound like /i/.
I've realized I don't quite understand the point of this analysis; could anyone explain it to me and tell me if it is still considered to be correct by anyone, or by most people, or by only a few people?
Stress & vowel length
The main motivation that I think I understand is stress and vowel length: if we say that "potency" has only two syllables at the level of phonology where stress and vowel length rules are applied, this explains why it doesn't show "trisyllabic laxing"/"Luick's law" and has a long rather than a short vowel in the initial syllable. Likewise, the stress of "presidency" can be analyzed as antepenult rather than pre-antepenult. However, this seems kind of a lame motivating factor in itself considering that there are many other extra-metrical suffixes in English that aren't, as far as I know, analyzed as lacking syllabic nuclei, such as -ness, -less. Also, I know of at least one exception to the apparent extrametricality of -y in -ency/-ancy nouns: the word discrepancy, which is typically pronounced with antepenult stress on the e, which is "short". It seem to me that it would be bizarre to explain this by saying that "potency" ends in /j/ but "discrepancy" ends in /i/.
Apparent consonant changes
I suppose another reason might be the apparent palatalization of /t/ to /s/ in these words (latent > latency). While palatalization is undoubtedly the diachronic source of /s/ in words ending in -cy, it is not obvious to me that we can explain the /s/ as the result of a synchronically active process of palatalization.
Palatalization before written "i"/phonetic /i~j/ in word-medial contexts usually yields /ʃ/, not /s/ (or at least can): consider potential, potentiality. Unlike these words, potency cannot be pronounced with /ʃ/. This suggests to me that "potency" has to be analyzed somewhat differently.
Also, we see /s/ in -ance/-ence words, where there is no (at least, no overt) following segment corresponding to /j/.
There are also odd formations like normalcy < normal, where there is no /t/ in the source word; baronetcy, with /t/ followed by /s/; and, in spelling if not in pronunciation, also bankruptcy, vicountcy, paramountcy. These examples suggest that in at least some cases, /si/ is apprehended as a suffix, in which case "ancy" and "ency", rather being analyzed as -ant + -y and -ent + -y, can or should be analyzed as -ant + -cy and -ent + -cy, with no need to consider the /s/ as derived synchronically from palatalization of /t/.