It's fairly well known among people who have been involved in the domain aftermarket . . cough . . for a few decades . . groan (getting old here) that many - probably most - domain sales "go unreported" . . soooo . . I take all generalizations regarding domain sales (like "people prefer..." or "most . . whatever") with a grain of salt. That is, with a bit of a "ho-hum . . when will they learn" attitude.
There was a time when folks went after "keyword domains" because there was evidence that keywords in a domain name / website address might afford that website-domain a bit of a lift in search results for queries targeting those keywords. Eventually, Google downplayed keywords in domains.
Today, I think of domains / website addresses "in all the media" -> not just SERPs (Search Engine Results Pages) -> which includes social media, radio, print and display ads, etc. I also think about "what's the cost of branding", i.e., creating a service/product -> website name association for a non-descriptive domain. In that context it's not a stretch to consider deploying a descriptive domain, one that clearly and concisely delivers a memorable message about a website's service or product(s) - becauuuuse - I don't expect folks in the social media space to spend much time or effort "explaining things" - so let the website name help get the job done.
Historically, based on my own aftermarket sales during the last 20 years, I've got the evidence that there's a decent market for descriptive (you say "keyword") domains in the .Com space. That may be partially based upon the domains I've registered or acquired, which invariably are generic, descriptive, industry defining domain names. In other words, I'm not a big fan of "brandable domains", though I'll admit there's a few opportunities I passed along the way (such as CVCV domains) and a few opportunities I grabbed (such as 0101.com, which I sold a few years back.)
I'm a small fish in a very big pond of domains, so my observations aren't conclusive of anything.