As always, I think it is worth be a little cautious about what “everyone knows”. My understanding is a few colleges over time have openly suggested they only had a legacy preference in ED, and some people have then generalized that to all colleges doing that, but that is not necessarily a valid assumption.
In terms of data, as another poster pointed out, colleges do not willingly give out really detailed admissions data, and legacy data in particular is not necessarily something colleges really love to discuss. However, Harvard was forced to as part of the litigation it was involved in.
This is complicated by the fact Harvard only introduced REA partway through the litigation (which helps explain why there were more RD legacy applicants than you might expect). But as I understand it, the numbers (admitted/applied) were:
Total: 1560/4644 (33.6%)
RD (including when there was no REA): 709/3011 (23.5%)
REA (once introduced): 851/1633 (52.1%)
Aha, some people said, Harvard gives more of a legacy boost in REA.
Well, maybe, but maybe also stronger, better-fitting legacies were more likely to apply REA to Harvard once that became an option.
Still, this is confirmation that the legacy admit rate is likely higher in ED/REA/SCEA at most colleges. But whether there would actually be a higher legacy “boost” at every college once you controlled for other factors? Not so clear.
And for what it is worth, Yale has specifically said no, meaning it tells alum parents early legacy applicants must meet exactly the same criteria for admission as regular legacy applicants, and that they only accept legacies early when they are confident they would accept them regular as well.