Romney's hypothesis - that all reliably Democratic voters are dependent on money from the government and don't want to work - was testable and failed to meet the basic demands of evidence.
First, such a broad definition of government-sector dependency and "people who don't work" will include people (a) too young to work (otherwise known as "children"), (b) people who used to work but have since retired (otherwise known as "the elderly"), and (c) people who have careers that are paid for by local, state, and the Federal government (otherwise known as "police officers, fire fighters, public school teachers, and other civil servants" and "military personnel"). Even a cursory examination of these categories makes their designation as moochers a bit of a stretch, to say the least. Second, one can actually verify, through polling, how these groups of people actually vote. Children, of course, cannot vote. Those who fall into category (c) are, for the most part, evenly distributed in their preferences between Republicans and Democrats. The elderly tend to vote Republican.
Clinton's hypothesis that half of Trump supporters are either racist, sexist, homophobic, xenophobic, islamophobic, etc, or all of the above, is also one testable through polling. There's the
Reuters/Ipsos poll from late June that, for instance, showed that half of all Trump supporters rated blacks as more criminal and more violent than whites. And even though I think its fairly safe to assume a significant degree of overlap between the various categories within the basket of deplorables, that's just
one of the criteria Clinton included within the basket's panoply. The polling on Trump supporter's attitudes towards Muslims all throughout the primary campaign, for instance, leaves one with the distinct impression that "half" is, if anything, a wholly conservative estimate.