The study’s hard data will not be revealed until the paper is published this month, but the broad conclusion was unmistakable: people who offered or asked for precise amounts — including such arbitrary-seeming figures as $5,015 for a piece of jewelry as opposed to $5,000 — almost always wrung more concessions from the person on the other side of the table than those who opened with neater numbers. The reason, Mason believes, is that a precise number — rightly or wrongly — implies you’ve done your homework and know the actual value of the thing.

This is also true when asking for a raise:

http://pandawhale.com/post/20787/asking-for-a-raise-avoid-round-numbers-wsjcom

Other pricing experiments you can learn from:

http://pandawhale.com/post/18375/pricing-experiments-you-might-not-know-but-can-learn-from

A long time ago the Santa Fe Institute Complexity Institute had a contest to explore the effects of automated pricing and purchasing in commodity markets.  This was long before day trading, online trading, automated trading, and high frequency trading.   I spent a lot of time trying to implement an algorithm and ran out of time.  The program ended up asking arbitrary-seeming figures and then fluctuated wildly which threw the opposing algorithms into making sub-optimal sells or buys.  The secret?  Memory errors in C.   I ended up taking home a big portion of the $10,000 prize money!

Ha! Greg, that's a great story.

I've used this tactic (using non-rounded/arbitrary figs) in price proposals in the past. It does work. 

Good to know. Avoid round numbers!