Interesting thread. I have a few thoughts on it:
a) Surveys asking people if they have experienced a particular act of violence depend only on the respondent's interpretation about what took place, and there is no external way of checking if the response is true or not.
In larger samples most of this doesn't matter, because the individual variations are washed out in the averaging process.
But it does matter if the people participating in the survey consist of different cultural communities with somewhat different language and definitions of words and different views on what constitutes violence. It might even matter when we interpret answers by men and women, at least in some more conservative communities.
Police reports don't suffer from that problem, but they obviously suffer from other problems, so a combination of data sources would ideally be needed. I'd also like to see more in-depth open-ended investigations to tease out what all kinds of people mean with such words as sexual harassment or verbal violence etc.
b) The Williams Institute link gives a summary of this study (both are linked to separately in this thread). The abstract of the study itself does not include some of the things the Williams Institute chose to report, such as this:
Transgender women and men had higher rates of violent victimization (86.1 and 107.5 per 1,000 people, respectively) than cisgender women and men (23.7 and 19.8 per 1,000 people, respectively).
That's because this particular finding is more likely to be produced by chance than the level the authors set as the cutting point in their study. So the Williams Institute shouldn't have reported that bit without also reporting about its likely statistical non-significance.
c) The statistics on violence against transgender people are different for the most extreme kind of violence than other kinds of violence: Homicide rates in countries outside South America.
A few sources:
This study (which I believe has one transgender researcher) shows that the risk of becoming a homicide victim was lower for transgender people than the rest in US between 2010 and 2014. This site reports on some UK data, and this one gives some rough calculations for the US.
So why are the reported overall violence figures against transgender people so high, yet the homicide rates are not, outside South America? This clearly needs good research.
All violence is unacceptable, and transgender people should be free of assaults and violence, just as everyone else should be free of those. But considering the possible solutions to the problem requires much better and more comparable data collection and analysis, and the likely solution really is not to demand only one other demographic to change its behaviour (women) when the violence is overwhelmingly carried out by another demographic (some men).
d) My last comment on this, I swear:
It's always good to remember that ideally we wish to compare like with like in these kinds of studies, i.e., we should compare a transgender person to someone who is otherwise the same except for gender identity, lives the same kind of life, takes the same types of risk and so on. Only then could we be absolutely certain that the cause of the differences we find can be attributed to the difference in gender identity (or rather the performance of this vs. one's sex)
One reason why men are more likely to be the victims of violent crime than women is that women take more mitigating actions (not going out alone late at night, avoiding certain places and forms of behaviour).
It's possible that some of the findings in this specific field are also partly driven by differences in other kinds of variables which correlate with being transgender. Asking more questions about where the reported violence happened (work, home, leisure activities), and who the alleged perpetrators were (strangers, sexual partners, spouses etc.) would cast further light on this.