[apodohmv] Quantifying "dressing like a slut"

Consider the problem of quantifying to what extent "dressing like a slut" (said the Toronto police officer) currently increases sexual assault.

Of course, we hope / wish it doesn't affect it, and we should work toward that being the case if it isn't already, but here we are considering the "descriptive" not "normative" question:  what is the current state of affairs?

This is the most rigorous quantification I could think of, using a probabilistic framework.

First we quantify "dressing like a slut" versus not.  We consider the choice between two outfits, one which 60% of the population (not more, not less) considers "dressing like a slut", and another which 5%.  (It's a "trick" to rely on polling to quantify this variable, rather than specific characteristics of the outfit itself.  The percentages were picked kind of arbitrarily.)

What is the probability distribution of total number of lifetime sexual assaults if you wear outfit A today?  What is the distribution if B?  We care about the probability distribution of the delta change in number of lifetime sexual assaults, that is, the change that can be attributed to the difference in outfit.  As a Bayesian, we care about the one distribution for your particular situation, but empirically we measure among "people like you", perhaps same geographical area and social group.

Do anyone even remotely know what this difference distribution looks like?  I don't know if the mean (expected value) is near 0.5 or near 0.000001.  It would be surprising but not unthinkable that the mean is negative (dressing like a slut currently decreases your number of expected sexual assaults due to a side effect of "empowerment").

Ideally ("normatively"), this difference distribution should look like a Dirac delta function with the spike at zero.

This is a serious problem.  Without some guess as to what this distribution looks like, you cannot make good decisions of what to wear.  Omitted from this analysis were any other costs and benefits of each outfit.  There are probably externalities.

Without knowing what the distribution looks like, how can we quantitatively tell if we are making progress toward decreasing sexual assault attributable to "dressing like a slut"?  For example, how effective was a public awareness event like Slutwalk?