Cornell University has a research paper that starts with the assumption that parents naming a new baby are rational creatures working on reasonable assumptions, and pick a name for its uniqueness. They also work with the assumption that parents are myopic, meaning they can't see the possible consequences of their actions. The paper is authored by Katy Blumer, Kate Donahue, Katie Fritz, Kate Ivanovich, Katherine Lee, Katie Luo, Cathy Meng, and Katie Van Koevering. So you can see where this is going.
The entire paper is available as a PDF. This is real research, which they admit is incomplete in several places. The theory is that parents find a name they like that has the low popularity they desire, seen as the frequency of it being used, which can be ascertained by Social Security records. However, they should (but don't) also look for changes in that frequency year over year. This omission results in a shock when the child they so carefully named ends up at school with an extra initial attached to their name because there are so many others with the same name. Those who know say the study has some really good math jokes, but don't let the math put you off; you can skip over the numbers and just read the text to understand just how tragically real yet funny this project is. -via Metafilter
PS: Current naming strategies have left us with several songs about the phenomena, like Multitude of Amys, Daves I Know, and 27 Jennifers.
My niece got a very obscure family name that had been used only for men. It was unique until someone famous gave it to their son ten years later.
Then page 2, "This set-up gives us convenient parameters for the model and just enough Greek letters to sound smart enough for publication." And how myopic parents with "perfectaccess to baby name data" is "a realistic assumption".
Oh, and citation 6 ("They would be wrong.") is a SIGBOVIK (a satirical computer science conference organized by Carnegie Mellon students) presentation by the authors themselves!
Thanks for the pointer!