I would guess that hard evidence is going to be difficult to come by -- doing anything resembling a controlled experiment would be enormously expensive, and cyclists are too disorganized (by this I mean lacking unified organizational structure) to do some sort of sampling-based measurement or epidemiological study.
I would just observe that, in my opinion, bike thieves come in several varieties, with differing motivations, and the likelihood that a particular bike will be selected for theft depends a lot on the type of thief:
- Glitter-attracted impulse thief. Someone sees a fancy/shiny bike (by their standards) that catches their attention, has low self-control, and so just rides off with it, on a lark.
- Necessity/opportunity thief. Someone is in an immediate situation where a bike would be handy (eg, it's raining, or they're in a hurry), sees one, and rides off with it.
- Theft for personal acquisition. Someone sees a bike they like, feel that it would be a good bike for them, and takes it.
- Theft for resale. Someone steals bikes to sell for money.
In the last two categories there's also the division of theft for the whole bike and theft for parts (or removing parts from the sitting bike).
Of these categories, the first will probably be put off by "uglification", but it will have relatively little effect on the second type of thief. (In fact, the second thief may be apt to pick a less-expensive looking of several equally available bikes, to lessen the seriousness of their crime -- at least in their own minds.)
The third thief will have specific attributes in mind for a bike and may or may not be put off by the obfuscated appearance, depending in part on how perceptive he is. (But keep in mind that of the four he's the most likely to spend time studying the fine points of the bike.)
The fourth thief will likely know how to recognize a valuable bike, even obfuscated, but may choose to pass over one that has been permanently defaced or which would take too much effort to clean up.
You can also evaluate these categories based on how effective locks will be. The first thief is apt to be deterred by a halfway secure lock but may (or may not) have at hand the wherewithal to defeat a lightweight one. The second thief will almost certainly pass over a bike that is at all encumbered with a lock, even skipping a bike locked only by the QR-attached front wheel. The third thief may be willing to spend some time defeating a lock on a bike he desires but may not be especially skillful, while the fourth thief will have the skills and tools to defeat many locks quickly.
The character of the neighborhood, time of day, etc, will affect the mix of these different thief types.