Expanding role of auctions
Auctions are used worldwide for selling all kinds of public and private resources. Their use has expanded in recent years, as organizations take advantage of their unique strengths, including their usefulness in accurately determining the fair market value of goods; their quick, easy and inexpensive implementation; and the chance for companies and governments to earn funds for public expenditures.
I believe auctions can be a good method of achieving social goals. Therefore, in my research my focus is not on private firms, but rather on governments, and how they can use auctions to allocate public resources. Auctions can be a complement to the price system, as well as play an important role in public finances — the money or revenue collected from auctions can be used to support public facilities or utilities.
The question researchers in this field always ask is, “What do we need to do to improve systems in real life?”
Designing a “good” auction
A well-designed auction should exhibit “pareto-efficiency,” that is, the items are allotted to bidders who value them most; social welfare is maximized, and it is impossible to reallocate items among bidders such that no bidder is better off, without reducing others’ well-being and the auctioneer’s revenue. This eliminates the risk of collusion between bidder and auctioneer, and reduces the possibility of a bidder “gaming” the system to his advantage.
It should show tractability, that is, it should be computationally efficient in the programming, and be reasonably simple to use for a non-professional.
It should be transparent, so that high selling prices and the price adjustment process are obvious to bidders. It should also reduce information leakage, so that the auction process reveals as little information about the bidders as possible.
In my research I try to strike a balance between practice and theory, so that a practical auction can also provide some good theoretical outcomes. In practice, of course, users don’t care about theory — they just care about their ability to implement their bid. So my key aim is to make the bidding process as simple and as usable as possible, while at the same time illuminating some theoretical properties.
We focus in particular on auctions with large money transfers, for example in allocating such things as housing, mobile phone spectrum, car licenses and the like. In those auctions the winning bids involve huge amounts of money.
For example, in 2003, the U.K. held an auction for five 3G spectrum licenses, for mobile phone carriers. As you can imagine, the winning bid was extremely large — the revenue totalled around 2.5% of the country’s GDP, and all of this money was spent on public utilities and services. So the auction resulted in a public good.
In another example, in 2013, the city of Shanghai held a car license auction, and there too the winning bids were very large. The average winning bid for a single car license was almost 1.5 times the average citizen’s annual income. The proceeds all went to the government.
The study of actions with large winning bids is a relatively new topic of research, because in traditional auction theory, the winning bid is relatively small, so that even if one wins the auction, it has no income effect, meaning the winning bidder does not need to cut his budget to be able to afford other expenditures.
But when the auction involves large amounts, the auction system is quite difficult. The current ones have lots of drawbacks. The basic question we ask is, with an auction with large winning bids, how can we design the auction? That is, how can we design an efficient and incentive-compatible auction?
A real world application
In my current research project, I am studying ways of assigning heterogeneous objects to agents and determining their payments. I present an alternate system for selling car licenses in Shanghai and Singapore.
I consider the situation where there are several items, and several bidders. Each bidder is interested in receiving at most one item. I use as my test case the auctions for car licenses in Singapore and Shanghai, since these cities impose restriction on car licenses to reduce pollution, and each bidder can get at most one license.
This is an auction that achieves a balance between theory and practice when dealing with several items; multiple bidders; each bidder interested in obtaining at most one item; and each bidder having general preferences.
This research could find wide application in commercial settings as well. In particular, companies like Google, Amazon or Facebook would benefit from this kind of approach for their ad slots sales and other auctions.
We know it’s hard to change research trends in auction theory, because people are familiar with the standard tools, and the standard technology. We do it in a standard framework, but we want to challenge that, because there’s still a big gap between theory and the real world applications.
So as this work progresses, I will continue to look for ways apply theoretical findings to the real world. I want to find gaps between theory and practice and bridge them.