Statistics can be useful, not useful, or misleading. Nothing about them is inherently meaningful or valuable--it depends entirely on the question and process generating the data and statistic. GIGO applies.
Despite the meme that an anecdote isn't data, it actually is no different from data from the Bayesian point of view. It is an n=1 posterior with a prior of one's past life experiences. And many anecdotes together can be thought of as a multi-level model (if you don't believe this, just see what the methodology of many behavioral/observational studies looks like--it's collating anecdotes into "data"--including the crime statistics that you are after).
> the anecdote is a powerful emotional narrative that feels true
And researchers can absolutely be emotional! A study biased by the emotions and beliefs of a researcher will produce biased results. Statistics isn't an escape hatch from human bias; it actually compounds whatever bias exists in the first place.
