|
By
Hal R. Varian
New York Times
Online
auctions offer a wonderful laboratory for experimental economists:
The participants are intelligent adults, spending real money,
who hope to purchase goods in which they have an intense
interest. This is a far cry from the reluctant sophomores
whom experimentalists have had to rely on in the past to
test economic theories.
Analysis of online auction data has yielded a wealth of
insights, and a few puzzles. A particularly intriguing puzzle
has been the tendency for "late bids." In a representative
sample of eBay auctions, researchers found that 37 percent
of them exhibited bids in the last minute and 12 percent
had bids in the last 10 seconds. These data actually understate
the actual number of bids submitted in the closing seconds
of the auction, since bids that arrived at eBay after the
auction had closed were not counted.
The late-bid puzzle is particularly interesting, since eBay
offers an automated "bidding agent" that is intended to
eliminate any incentive for late bidding. I only need to
tell my bidding agent the most I am willing to pay for an
item, along with my initial bid.
If someone bids more, my agent will automatically increase
my bid by the minimal bid increment, as long as this doesn't
raise my bid over my maximum.
In theory, each bidder should report his true maximum willingness
to pay and let the agent do the work of bidding. In the
end, the person with the highest willingness to pay for
the item will win the auction, paying a price equal to the
second-highest willingness to pay plus the bid increment.
So why doesn't this happen? One theory, advanced by Patrick
Bajari and Ali Hortacsu, two auction researchers at Stanford,
is that experts don't want to reveal their interest in the
auction too early.
Suppose you are a recognized experts on coins. If you bid
early on a particular coin, you may reveal to others that
this coin is highly desirable. Perhaps it is better to delay
bidding in the hopes that others won't recognize the coin's
value, and then sneak in at the last minute with a winning
bid.
This explanation makes a lot of sense for auctions where
expert opinions matter. But late bidding doesn't just occur
in markets for coins, stamps, and collectibles; it also
happens in auctions for computer parts, where one player's
bid isn't likely to influence other bidders' valuations.
How can late bidding be explained in these cases?
Recently Al Roth and Axel Ockenfels, two Harvard researchers,
have offered a different, but complementary, explanation.
Their hypothesis is that late bidding is a way for participants
to avoid bidding wars, allowing them to get the item at
a lower price than they would by bidding early.
To give the flavor of the analysis, put yourself in the
place of a bidder for a Pez dispenser - just the one you
need to complete your collection. It's worth $10 to you,
and you know there is only one other possible bidder, who
you believe is also willing to pay $10. The seller has set
a reservation price (the minimum price at which he will
sell) of $2.
If each of you tells your bidding agents your true willingnesses
to pay for the Pez dispenser early on, the agents will get
in a bidding war, raising the final selling price to $10.
Even if the tie is resolved in your favor, this is no bargain:
You have to pay $10 for a Pez dispenser that is worth just
$10 to you.
What happens if you each bid $10 in the last few seconds
of the auction? In this case, there's a good chance that
one of the bids will be dropped by the server, leaving only
a single bidder, who wins by default and must pay only the
seller's reservation price of $2.
You don't have to be an economist to recognize that paying
$2 is a lot better than paying $10. True, when you bid late
you run the risk of your bid being dropped, but even if
you get the Pez dispenser for $2 only half the time, you're
still better off than you would be by bidding early.
Simple as this example is, it exhibits the essential intuition
of the Roth-Ockenfels analysis: Bidding high at the last
minute and letting chance determine the outcome is better
for both players than bidding high early and precipitating
a bidding war.
This is an appealing explanation for late bidding, but is
it right? What one would like to do is to vary the auction
design, to see how players behave when "letting chance decide"
isn't an option.
Luckily, there are other auctions on the Internet that use
different rules. Amazon.com, for example, has a "Going,
Going, Gone" rule. If there is a bid in the last 10 minutes
of the auction, it is automatically extended for another
10 minutes. The auction terminates only when 10 minutes
have gone by with no bids.
If the Roth-Ockenfels theory is right, we would expect to
see much less last-minute bidding in Amazon auctions. Indeed,
that's what the researchers found. While in the sample of
eBay auctions 37 percent had bids in the last minute, on
Amazon fewer than 1 percent of the auctions exhibited this
sort of behavior. Furthermore, on eBay experienced bidders
tend to submit late bids, while on Amazon experienced bidders
submit early bids.
So perhaps the late-bid phenomenon makes sense after all.
The online auction bidders, either through cunning or through
trial and error, have managed to figure out a winning strategy.
|
