I recently looked again at Ausubel’s multi-unit auction paper, and really liked the crystallization of the first paragraph in it:
The auctions literature has provided us with two fundamental prescriptions guiding effective auction design. First, an auction should be structured so that the price paid by a player—conditional on winning—is as independent as possible of her own bids (William Vickrey, 1961). Ideally, the winner’s price should depend solely on opposing participants’ bids—as in the sealed-bid, second-price auction—so that each participant has full incentive to reveal truthfully her value for the good. Second, an auction should be structured in an open fashion that maximizes the information made available to each participant at the time she places her bids (Paul R. Milgrom and Robert J. Weber, 1982a).
I would say that this is useful practical insight gained from studying theory.
Turing's Invisible Hand 
The first part does not seem to be true. In GSP, the price totally depend upon opposing bids, still it is not fully truthful.
Some people actually call GSP auction as freshman’s VCG. If theory is misunderstood then it gives rise to some interesting things, such as GSP.
Regarding the second part, Yahoo provided the information about competing bids, and Google did not. See who changed its practice. Yahoo had a website, where you could put a keyword and see all the bids. No username/password was required.
Yahoo provided more information, and Google less. Now Yahoo changed its practice and not Google.
Actually, it seems to me that Ausubel’s wording is just non-specific enough as not to rule GSP out despite it being non-IC.
Your point about the 2nd part, where giving information did not seem to work well, is really interesting, and I don’t really have a good explanation — I still can’t understand why Yahoo’s open way was not better.
In the second pararaph, “sometime” is missing. The paragraph should read.
Some people actually call GSP auction as freshman’s VCG. If theory is misunderstood then *sometime* it gives rise to some interesting things, such as GSP.
Naom, the answer to the 2nd point lies in an effect, I pointed out a few years ago, and that is monopoly.
A search engine buys and sells intent. And this process is short duration, and this gives a monopolistic power to the search engine for that short duration. I called it Atomic Monopoly. One direct proof of this theory is that when a query comes, in my knowledge, to some extent every search engine uses monopolistic pricing. If the demand is too low, then as a monopolist does, supply, i.e. the number of ad-slots is cut or the reserve is raised so that some ads are not assigned space even though the page has abundant of it (the reason is not relevance, because raising the bid does allocate the space to the ad, so I am talking about relevant ads which do not show up due to monopolistically higher prices).
This does not happen when the interest is targetted, which happens with most other forms of advertising.
So I challenged the fellow scientists to come up with some mechanisms, to cut the monopoly of the search engines. If it is a monopoly of a single company, usually society acts to contain it. But when the entire business model is monopolistic then the problem is challenging, and require the scentific exploration to find a benefecial outcome for the society.
There are bunch of other evidences, I had given. For an example,if the cost of making/maintaining search engine falls, while giving essentially the same service, multiple search engines would maintain their current revenue. As a thought experiment, if you split Google, into Google A and Google B, providing exactly the same search pages, and for conceptual purpose ignore the fixed overhead costs, then within the current business model, both A and B would be able to maintain the same total revenue as the single Google, even though none has any competetive advantage.
In any other business, even those with the Network Effects, this won’t be the case. Take your favorite software (the reason I chose software, because this is another business which scales as good as a search business), if two separate companies provide exactly the same bits, none will be able to make any profit, due to bertrand competition.
Why is this important? Because in a monopolistic situation, the auctioneer has information advantage. If you see a used car, only the seller of the used car knows what is wrong with it. Do not expect the seller to give you this information. Similarly a search engine could realize that by not providing accurate information to the bidders, it can charge higher monopolistic rents. The situation is so dynamic, that a bidder can’t learn what’s going on, as usually assumed in repetitive games. The game which will be played tomorrow on a keyword, would at least be little bit different than the game which is played today. The number of keywords is also so large that it is virtually impossible for bidders to keep correctly estimating the information about each one of them. The only thing a bidder knows reliably is that a click is delivered, and the bill is charged below the bid.
This is also one of the most important reason that pay per click is superior to both pay per impression and pay per action, as this is the point when the need of information sharing between a search engine and the bidders is least. Impression to Click information is best observed by the search engine, and click to action is the best observed by a bidder. So choosing pay per click minimizes the dependency, and the need of sharing the information, between a search engine and the bidders.
Yes, as you can see the second principle holds in selling a new car. The salesperson is willing to tell every detail of the new car and as well as competing cars, when a buyer insists. So if a buyer is willing to spend enough time, the buyer could actually make an informed choice. Not true in case of an used car, and in case of bidding for a click.
It seems that there are two separate issues: the “atomic monopoly” and the information asymmetry.
I am not sure that the monopoly explains anything here: to begin with auctions are in a setting of complete seller monopoly and I don’t see anything special here that explains the gain from reducing bidder information.
I agree with your analysis that information asymmetry is key here, and agree with your analysis of the popularity of cpm. But still, what is so special here that the information asymmetry causes the seller to gain from reducing information where the usual analysis shows that on the average sellers gain from releasing more information?
Because in a non-monopolistic situation, the buyers value information, and in order to compete the seller of a non-monopolistic situation would offer the information.
Suppose every searcher who searched Google had also searched Yahoo, i.e., the intent is given to two companies in a marketable form. It is not hard to see that in such a situation Google would have started giving more information instead of Yahoo cutting on information.
You can form a 2 by 2 game, about information sharing to see this. It won’t be hard for you to see that the monopolistic situation, is transferring all the surpluses to the search engines. Minimizing the flow of information to the bidders, and making the entire auction opaque.
So atomic monopoly and the lack of information flow to bidders have some cause and effect relationship with each other.
I still don’t see why withholding information helps. Let us suppose the extreme monopolistic situation: there is only a single search engine in the world and furthermore that no other search engine can ever be started and furthermore that there is no other place to place any type of ad. As far as I can tell, the main point of Milgrom&Weber is that still full information revelation is in the best interest of the seller! ( http://iew3.technion.ac.il/~dov/milgrom_weber.pdf — see description starting with last paragraph on page 1095)
I thought this part is clear to you at least by being in Google, that the bidder lacking crucial information benefits the search engine. If bidders know everything then they may not always target the highest affordable slot, or at least higher than their optimum choice. If they have all the information then they will target their optimum choice, which will hurt the search engine revenue.
Information is money. One viewpoint of market, auction etc is to trade and share the information. When you are investing in a stock market, that’s what you are doing. People lacking the information or lacking the correct information loses. People having the right information, even if it is coincidentally right wins, but in the process they share this information.
Information assymetry is 23 degress tilt of the Earth’s axis, causing all kinds of dynamics.
This is a part which we have to agree to disagree. The information sharing happens, because either a bidder will get that information from competitor which the bidder will value more, or because a part of the information sharing is in the benefit of the seller.
Here is an example, you have to resolve. You have a used car. You are selling it by competitive bidding. You have all the great service record of it. But you also know a big deep problem with the car. Tell me whether you can maximize the price of the car by informing all your bidders about the deep problem with the case. So this is an example where you do not want to share information in a competitive bidding.