The statistics of eBay

I'm eBaying for a new laptop this morning, as my Compaq Presario 732 (about four years' old) has begun having serious issues. The power button has taken on the personality of a devil child and responds only about 5% of the time. I take this to be the Grim Reaper calling and am cherishing every moment we have together.

I used the "Buy It Now" option on an IBM T40 Thinkpad, and started having buyer's remorse as soon as the initial rush wore off. At any given time there are about two or three dozen T40s for sale in various configurations. One auction (for a T40 with better specs, wouldn't you know?) is finishing in a few minutes, and I'm wondering how much the price is going to shoot up. Has anyone ever done a study on that? "If only I had a statistical model...." How many times have I said that? Don't even get me started!

A quick Googling of the issue shows only a couple academic papers using eBay data. This one from Stanford seems most promising. Let's see... "Multi-variable regression results"... perfect! Reference to a Table III... ugh, that's a lot of variables. It's not exactly what I was looking for, but maybe the problem's more complicated than I thought.

Well, the auction's over now. It looks like the final price represents only about a 7% increase over the t-2 minutes' price. But who needs 1+ GB of RAM anyway? (sigh)