The recent news that a U. of Utah college student managed to sabotage an entire Bureau of Land Management (“NAMBLMA”) auction (1, 2) presents an opportunity to write about game theory for no particular reason.
Leaving aside for the moment the issue of whether or not it was a good idea, what Tim DeChristopher did is pretty illegal. By winning several auctions, he entered into contracts that he doesn’t intend to honor (presumably he can’t afford to pay the $1.7 million that would be required). But that doesn’t seem to be what people are upset about. Many bidders, as well as the BLM, don’t seem as concerned with that as they are that somebody might have “overpaid” for a parcel of land.
The bidders have every right to be annoyed. In the absence of interference, they probably would have won their parcels for less money. And even though bidder interference should net them more money, the BLM probably should be annoyed, too. First, because they’ll need to help make things right with their irritated bidders, which may be costly. I read somewhere that they were offering bidders the opportunity to retract bids that they felt were placed in error, although I can’t seem to find the article that said that anymore. Second, and more to the point of this post, DeChristopher’s stunt embarrasses them; it demonstrates that their auction system is critically flawed. How did they just let any chump walk in and bid with money he didn’t have? From the BusinessWeek article:
Shea said the BLM didn’t require bidders last week to show proof of a bond or their ability to pay for leases.
That was a practice Shea said he followed as head of the bureau for two years ending in 1999.
“Somehow, the regulations changed, an indication of their rush” to sell oil and gas parcels before President George W. Bush leaves office next month, said Shea. “It was rush before the door slams behind them: ‘Let’s get as many leases out as possible.’”
From the sound of it, the auction was the kind everyone is used to: an open auction (in particular, an English auction). All the bidders are in a room, and when the auctioneer calls for bids, a bidder can place a bid by raising a paddle. The auctioneer continues asking for higher bids until nobody is willing to bid, and then the last bidder wins the item at that price of his or her bid.
English auctions are useful because they’re pretty transparent; all bids are in the open, so everybody can verify that the highest bid did, in fact, win the item (as opposed to a secret auction, where the party who announces the result must be trusted by all other parties to be honest). It is also a form of second-price auction: the winning bidder only has to pay (a little more than) the highest amount that the runner-up was willing to bid, no matter how high the winner would have been willing to go.
What, then, is bad about English auctions? For one, they’re vulnerable to collusion; an unscrupulous auctioneer can have someone drive up the price with fake bids, or a ring of colluders could bid in such a way as to discourage others from entering. Another issue is that bidders have no privacy; all bids are open to inspection, so participating may divulge a bidder’s strategy. Finally, DeChristopher’s attack is more of a social attack than a game-theoretical one: it’s absurdly easy to disrupt an English auction just by placing ridiculous bids. This is pretty common on high-profile eBay auctions, such as when Stephen Colbert auctioned his signature-saturated wrist cast. Pre-approving bidders is one solution, but can be time-consuming, and the extra paperwork potentially discourages some potential participants, thus reducing the efficiency of the auction.
Naturally, there are other solutions. One is to use a secret auction. Anybody who wants to can submit a bid in a sealed envelope. The auctioneer privately opens all the bids and announces the winner, along with the winning price. (In a second-price auction, the winning price is the second-highest bid.) If some jackass shows up an submits a bogus bid, no problem—because that action didn’t affect anyone else’s strategy, it’s just a matter of throwing out the bogus bid and announcing the result as if that bid didn’t exist.
This works great, so long as everyone trusts that the auctioneer, or an appointed proxy, is honest. A clever auctioneer can work around even this hurdle by [WARNING: NERD TIME] publishing a zero-knowledge proof that the computation was carried out correctly.
Interference hasn’t been particularly common in the past, but if the BLM wants to prevent further displays like this one, it’s probably time to switch to a secret auction. It’s not without precedent: the Federal Reserve uses sealed bids in the Term Auction Facility and in Treasury Bond auctions (although those also require pre-approval; the main benefit there is actually bidder privacy, and that not all parties have to be in the same place at the same time).
As it turns out, I don’t have any real opinion on whether DeChristopher should or should not have done what he did. And I’ve even managed to bury any opinion at all deep into TL;DR territory.
Glancing at the name of the blog, I think I’m doing it wrong.
Actually, it would have been simpler to prevent this just by bonding all participants. They waived those requirements because they were in a rush to get the sale in at least 30 days before the end of the administration.
But you’re missing something key, here. Dechristopher didn’t just disrupt the bidding on the 13 parcels he ‘purchased,’ but on the rest of them, as well. On dozens of parcels, he drove the price substantially higher than it otherwise would have been. And that highlights another inefficiency of ascending auctions – they only work well when there are enough bidders, and the bidders don’t collude. The fact that most participants were willing to pay much, much more than they would have had DeChristopher not pushed the bidding northward suggests that these auctions haven’t been working efficiently, anyway. They’re securing good deals for the bidders, but not discovering the highest potential price for the seller – or anything close to it. I can’t say whether it’s collusion or illiquidity. But either way, it highlights the point that the land-auction system is a lousy deal for the public.
Actually, it would have been simpler to prevent this just by bonding all participants. They waived those requirements because they were in a rush to get the sale in at least 30 days before the end of the administration.
I did mention that (…. more or less). I also mentioned offhand that from a theory standpoint, it introduces a barrier to participation, and thus isn’t efficient. But of course it’s the simplest solution, especially if you’re not an administration completely devoid of foresight.
As for the inefficiency of the English auction process, you’ve a very good point. And I was planning to get there (hence the seemingly irrelevant title of the post) before I drifted into TL;DR territory and decided that having more wine was more immediately important than fixing it.
A second-price auction is suboptimal from the seller’s standpoint, in that the seller doesn’t get as much money. But, as noble a goal as it is for the taxpayers, I don’t think it’s practical to actually sell the item at the maximum price a bidder is willing to pay; that’s just not an equilibrium of the auction. That’s why I emphasize the second-price nature of the English auction, as well as the easily-implemented second-price secret auction.
We may have to get into nerdy math time once we get LaTeX installed here. (VCG FTW?) It’ll be just like those times arguing with Silvio Micali, except with better handwriting.