You Are Not So Smart posted an article on the anchoring effect. The anchoring effect happens when you need to estimate a number of some kind — like a statistic or a price point — and you use some other number for reference, even if there is no real reason to think that other number tells you anything.
This is why you’re more likely to buy a leather jacket for $400 if the price tag originally said $1000. And it apparently also works when you know the anchor number has nothing to do with your estimate. The article recounts a 2006 MIT study where subjects wrote down digits from their social security numbers and then bid on items in an auction. Subjects who wrote down higher (essentially random) numbers bid higher in the auction.