In auction theory, jump bidding is the practice of increasing the current price in an English auction, substantially more than the minimal allowed amount.
At first glance, jump bidding seems irrational. Apparently, in an English auction, it is a dominant strategy for each buyer whose price is above the displayed price, to always bid the minimal allowed increment (e.g. one cent) above the displayed price. By bidding higher, the bidder gives up the opportunity to win the item at a lower price.
However, in practice buyers increase the displayed price much more than the minimal allowed increment. Buyers may even sometimes offer an increase on their own high bid, seemingly irrationally.
Several explanations have been suggested to this behavior.
When bidding is costly, or when time is costly, jump-bidding allows the bidders to reduce their total costs and get to the outcome faster.[1]
Here jump-bidding comes into play. It works like a signaling game.[2] By jump-bidding, the jumper signals that he has a high value, and so the other bidder should quit immediately if his value is lower.
Two bidders, Xenia and Yakov, participate in an auction for a single item. This is a common value auction with the following parameters, where A B and C are independent uniform random variables on the interval (0,36):
X:=A+B
Y:=B+C
V:=(X+Y)/2=(A+2B+C)/2
The auction proceeds in two stages:
We show that there exists a symmetric perfect Bayesian equilibrium in which each bidder jumps if-and-only-if his value is above a certain threshold value, T. To show this, we proceed backwards.
In the second stage, there is a symmetric equilibrium in which each bidder exits at his observed value - Xenia exits at X and Yakov at Y.
In the first stage, we take Xenia's viewpoint. Assume that Yakov's strategy is to jump if-and-only-if his signal is at least T. We calculate Xenia's best response. There are four cases to consider, depending on whether Xenia/Yakov jumps/passes. The following table shows Xenia's expected net gain in each of these cases:
| Xenia jumps | Xenia passes | ||
|---|---|---|---|
| Yakov jumps | E[V-Y|T<Y<X] | 0 | |
| Yakov passes | E[V-K|Y<T] | E[V-Y|Y<X] |
At the threshold (X=T), Xenia should be indifferent between jumping and passing:
\begin{align} &&E[V-K|Y<T,X=T]&=E[V-Y|Y<X,X=T] \\ \implies&&K&=E[Y|Y<X=T] \\ \implies&&K&=2T/3&&(whenT\leq36) \implies&&T&=3K/2&&(whenK\leq24) \end{align}
So the symmetric PBE strategy (at least when
K\leq24
K
3K/2
The outcome of this PBE is substantially different than that of a standard Japanese auction (with no jump option). As an example, let the jump-level be
K=24
T=36
This outcome seems counter-intuitive from two reasons:
Y<X
2X/3
Jump-bidding is a very crude form of communication: it does not communicate my actual value, it only signals that my value is above a certain threshold. The careful selection of the threshold and the jump-height guarantee that this communication is a self-enforcing agreement: it is best for both bidders to communicate truthfully.
Since bidding proceeds in discrete steps, jump bidding can alter the outcome. For example, suppose the initial price is 0, the minimal increment is 2 and the values are 9 and 10. Then, without jumping, the 9-bidder will increase the price to 2, the 10-bidder will increase the price to 4, the 9-bidder to 6, the 10-bidder to 8, and the 9-bidder will have to quit, so the 10-bidder will win and pay 8. But, if the 9-bidder jumps from 0 to 8, the 10-bidder might quit and the 9-bidder will win and pay 8.[3]
Some authors claim that jump-bidding reduces the seller's revenue, since the signaling allows bidders to collude and reduce the final price.[2] Therefore, it may be more profitable for the seller to use an auction format that does not allow jump-bidding, such as a Japanese auction.
Other authors dispute this claim.[3]