<...>
There is a largish number, but not huge (and certainly not "most"). Just over a year ago, around 15K account had negative trust (from a DT perspective). I haven’t got a more recent set of data, but another indicator I’ve got is that nearly 4.800 merited profiles (out of over 40k profiles) bear negative trust (from a DT perspective). They may nevertheless have positive ratings too.
Note that I’ve mentioned from a DT perspective a couple of times. As things stand, the Trust score counters (positive, neutral and negative) display the values provided by the ratings in accordance to your trust view (be it your custom trust list, or the Default Trust view (DT) which most view). You cannot counter the negative values with positive values, as they are independent counters, and only the person who gave you a negative rating can change that.
It is though pretty usual to have negative ratings that do not count towards the Trust counter values (these will be the untrusted feedback). For example, I’ve got a couple of these negative ratings, originated as a worthless retaliation for calling out a couple of scammers. This is quite a usual modus operandi, but rather futile as an attempt to damage one’s account, lest the retaliation come from the DT network (top levels).
In theory, the Trust system should be commerce related. That applies to commerce transactions you could have made with a given person (where one would give feedback in accordance to that experience) and/or the ability of that person to scam you or others (desirably, providing a reference to some substantial evidence). Nevertheless, if you read through feedback on people’s profiles you’ll see all sorts of feedback that differs from this commerce nature, thus undermining in some cases the nature of the Trust System’s principle.
That’s why, besides looking at the trust ratings (which are a numerical summary so as to say) you need to read a profile’s trust feedback (both trusted and untrusted) in order to get a general composition of the person’s trustworthiness.