In addition to KarlG’s answer, there’s a paper that argues that speakers have actually reanalyzed the reduced have as “of” acting as a complementizer, i.e. conversically. . working similarly to the ‘to’ in English infinitives, giving the bracketing I would (of worked). On the other hand that the argument is that have cannot further reduce from to while of can always reduce from to … In the construct under discussion the proposed reduced version of have can in fact decrease from to , and the construct under discussion in general is not necessarily justified. Hence, according to the paper’s argument,is not actually have, but of.