Thanks to a heads-up from Cardboard Connection, I was turned on to a study done which looked at racial affects of sales on eBay. The authors of the study chose to sell baseball cards. In each auction, a single hand was shown holding the card. The study then evaluated the original purchase price versus what the card sold for in the study's auction based on the race of the card holder's hand: that is - do African-American sellers get lower bids than White sellers, based on only the skin color of the seller?
In the study, the authors do a fairly good job of eliminating most of the variables related to eBay auctions: sellers have non-identifying IDs, paypal accounts, etc; different sellers offer up a variety of player cards; no interaction between sellers and potential buyers was conducted during the auctions, etc.
The authors prided themselves on NOT selling the same cards among the participants. That is, no two sellers sold the same card. I see this as a serious design flaw in the study. Why? Because in card collecting, WHO is pictured represents a very significant piece of the puzzle. Even if superstars are spread among the sellers (which they were, as well as memorabilia/auto cards), one must still take into account the actual player(s) being depicted. Granted, the authors were not trying to see if the cards sold for more or equal value to what they were originally purchased (though that is a part of the data). They wanted to see the overall difference: did white sellers have higher sales than non-whites. To me, they were not comparing items on a level playing field. This was akin to comparing sales of Granny Smith apples to Red Delicious. I do understand that they did not want to appear to be flooding the market with any particular card, but to me anyway, the study should be re-done using cards of the same players, and preferably of the same cards (or at least have some control cards involved: each seller has, say, 8-10 of the exact same card as the other sellers for sale). For clarity here: I do not mean each seller has 8-10 Rollie Fingers autograph cards. I mean each seller has a Rollie Fingers auto card, each one has a game-used McGwire card, etc and preferably it is the SAME card. So, potential buyers would see FOUR Fingers cards, FOUR McGwire cards, etc - the ONLY difference being the hand which holds the card.
The second flaw in the design: Players/teams being offered. The study was done around May 2006 (judging by the posted photos of completed auctions). April/May is actually not a bad time of year to do such a study: Baseball season is gearing up/just starting and potential collectors tend to buy more/bid higher. I admit, though, that the players generally being offered are of a caliber that they should garner decent sales results regardless of the time of year. However, collectors are a finicky lot. Some player cards will warrant higher selling prices for all kinds of reasons: does the collector mainly collect Hank Aaron, does the collector mainly collect Astros, how common is the card being offered up, etc? I think they did a nice job of mixing graded and non-graded, spreading out the production years, etc, but still: WHO is on the card? As mentioned above, a more consistent offering of the players depicted would have evened the playing field.
The study reports on the race(s) of the player(s) shown on the card. As a collector, I say they wasted a lot of time on that one. Generally, collectors do not collect based on the player's race (though, admittedly, I am sure there are collectors who do that very thing) - they collect based on WHO is shown on the card (whether that be a specific player, team, etc). While the data showing seller race compared to depicted card race(s) was a fun aside, that is all it was: a distracting aside that bears no meaning on the true focus of the study.
Perhaps a "better" control study would have been to offer one copy of each card without any hands being depicted at all. Would the lack of any kind of identifying race have changed the outcome of the study? Something to think about.
Of course, the number one issue at hand has nothing to with race. The number one issue is the use of baseball cards as the study's vehicle. Baseball card prices are some of the most liquid figures in the known world. Even when offered for a week at a time, any given card might garner higher sales or lower sales based on absolutely nothing whatsoever. In fact, perhaps a follow-up study should be done with the same cards in order to see if the data returns the same. Naturally, that wouldn't really be conclusive either based on the third sentence in this very paragraph: card prices are liquid.
Here's another thought: What if this study were done with, say, football cards instead? Or how about hockey cards? Would we see the same results? Would it matter, given the flaws inherent in such an experiment? What if the experiment were done with jewelry, videogames, or Fruit Loops?
I appreciate what the authors were trying to achieve: a determination of whether race plays a significant role in eBay auctions. But, one set of sales based one type of product is hardly conclusive. They even say so themselves, though not in so many words.
The entire study can be found here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1934432